World Bank Document · Contestable Markets, Trade, and Development William J. Baumol and Kyu Sik Lee How Much Does Trade with the South Affect Workers in the North? Adrian Wood Organized

THE WORLD BANK

IO& SEARCOBSERVER

i 7517Contestable Markets, Trade, and Development

William J. Baumol and Kyu Sik Lee

How Much Does Trade with the South Affect Workers in the North?Adrian Wood

Organized Labor, Politics, and Labor Market Flexibilityin Developing Countries

Joan M. Nelson

Technological Change and Commercialization in Agriculture:The Effect on the Poor

Hans P. Binswanger and Joachim von Braun

The Earmarking of Government Revenue:A Review of Some World Bank Experience

William McCleary

Social Security and the Poor:Choices for Developing Countries

Ehtisham Ahmad

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

THIE WORRLD BANK

RESEARCHOBSERVER

EDITOR Ravi KanburCOEDITORS Hans Binswanger, Vittorio CorboCONSULTING EDITOR Philippa ShepherdMEMBERS OF THE EDITORIAL BOARD Shantayanan Devarajan (Harvard University,United States), Mark Gersovitz (University of Michigan, United States), Shuja Nawaz(International Monetary Fund), Vittorio Corbo, Dennis N. de Tray, Enzo Grilli, RaviKanbur, Peter Muncie, George Psacharopoulos, William Tyler

The World Bank Research Observer is published twice a year, in January and July, bythe International Bank for Reconstruction and Development (the World Bank). It is man-aged in the Office of the Senior Vice President, Policy, Research, and External Affairs,and the editorial board is drawn from across the Bank and the international communityof economists. It seeks to inform nonspecialist readers about economic research currentlybeing undertaken within the Bank and areas of economics relevant for developmentpolicy.

The findings, interpretations, and conclusions expressed in this journal are entirely thoseof the authors and should not be attributed in any manner to the World Bank, to itsaffiliated organizations, or to its Board of Executive Directors or the countries they rep-resent. The World Bank does not guarantee the accuracy of the data included in thispublication and accepts no responsibility whatsoever for any consequences of their use.When maps are used, the designations employed are solely for the convenience of thereader and do not imply the expression of any opinion whatsoever on the part of theWorld Bank or its affiliates concerning the legal status of any country, territory, city,area, or of its authorities, or concerning the delimitation of its boundaries or nationalaffiliation.

The World Bank Research Observer welcomes editorial comments and responses. Pleasedirect correspondence to the editor at the address in the copyright notice below.

A subscription form for The World Bank Research Observer appears at the back ofthe journal. Readers with mailing addresses in countries that are not memnbers of theOrganisation for Economic Co-operation and Development are eligible to receive com-plimentary subscriptions on request. Subscription-related correspondence should beaddressed to the Publications Sales Unit at the address in the copyright notice below orto the World Bank European Office, 66, avenue d'Ina, 75116 Paris, France.

© 1991 The International Bankfor Reconstruction and Development/THE WORLD BANK

1818 H Street, N.W.Washington, D.C. 20433, U.S.A.

All rights reservedManufactured in the United States of America

ISSN 0257-3032

Material in this journal is copyrighted. Requests for permission to reproduce articlesshould be addressed to the Director, Publications Department, at the address in the copy-right notice above. The World Bank encourages dissemination of its work and willnormally give permission promptly and, when the intended reproduction is for noncom-mercial purposes, without asking a fee. Permission is not required to make photocopiesfor classroom and similar immediate use.

The World Bank Research Observer is indexed by the Journal of Economic Literature,the Standard Periodical Directory, the Public Affairs Information Service, Inc., and,online, by the Economic Literature Index and DIALOG. It is available in microformthrough University Microfilms, Inc., 300 North Zeeb Road, Ann Arbor, Michigan 48106,U.S.A.

TFlHE WORLD BANK

RESEARCHOBSERVER

VOLUME 6 NUMBER 1 JANUARY 1991

Contestable Markets, Trade, and DevelopmentWilliam J. Baumol and Kyu Sik Lee 1

How Much Does Trade with the South Affect Workersin the North?

Adrian Wood 19

Organized Labor, Politics, and Labor Market Flexibilityin Developing Countries

Joan M. Nelson 37

Technological Change and Commercialization in Agriculture:The Effect on the Poor

Hans P. Binswanger and Joachim von Braun 57

The Earmarking of Government Revenue:A Review of Some World Bank Experience

William McCleary 81

Social Security and the Poor:Choices for Developing Countries

Ehtisham Ahmad 105

CONTESTABLE MARKETS, TRADE,AND DEVELOPMENT

William J. BaumolKyu Sik Lee

The design of policies to improve economic efficiency using the private sectoras principal agent requires a clear understanding of the role of market struc-ture. Contestability analysis, not ten years old, provides a tool for the purpose.The concept can guide the government that wants to have it both ways: to pro-tect the public and smaller firms from the threat of monopolistic behavior bylarge firms, at the same time allowing the larger firms enough freedom to meetthe requirements of efficiency and to exercise entrepreneurship.

Perfect contestability provides a standard for ideal market behavior. A per-fectly contestable market is one in which entry and exit are perfectly costless;in such a market, the mere (perpetual) threat of entry can enforce good conductby incumbents. So long as sunk costs are zero, a potential entrant can undercutany excessive prices (or unnecessary costs) of incumbent firms yet earn anattractive rate of return. Thus perfect contestability precludes excessive profitsand prices as well as waste and inefficiency, and it prevents predatory pricing.

What contestability analysis means for policy in developing countries is thatan economy that wishes to take advantage of available economies of scale canuse the norms of behavior provided by the theory as a guide for regulation ofits larger firms, instead of resorting to nationalization or to stifling restrictionsas the means to protect its infant industries and its consumers. Under this stan-dard, the bounds on the firms' behavior set by the regulations replicate thosethat would be enforced by market pressures in an ideal state of perfect con-testability. The article gives a Nigerian example in which regulatory changesapplying the theory promise to improve the performance of the electric utilitymarket. Such methods can do more to promote the public interest than priva-tization, which often results in replacement of a state monopoly by a privatemonopoly.

The World Bank Research Observer, vol. 6, no. 1 (January 1991), pp. 1-17

© 1991 The International Bank for Reconstruction and Development/THE WORLD BANK

C ontestability analysis, a recent arrival in the literature of economic the-ory, has rapidly aroused attention among those involved in determiningpolicy related to economic regulation and antitrust activity. Yet these

are not the only promising areas in which the new theory might be applied. Italso appears to offer suggestions and ideas for policy related to internationaltrade and developing countries. This article undertakes to suggest some ofthese applications.

Contestability can be considered a generalization of the concept of perfectcompetition, interpreted as a (more or less) ideal state of economic affairs. Un-der contestability, competitive pressures can be supplied by potential entry aswell as by the presence of actual current rivals. As a result, unlike perfect com-petition, perfect contestability is compatible with the presence of economies ofscale and the consequent small number of relatively large firms that are disci-plined into good behavior by the threat of entry. More industries can thus moreeasily approximate the circumstances of contestability than those of competi-tion. In other words, contestability may provide a somewhat more realistic de-scription of the behavior and structure of some industries. And becauseeconomies of scale and oligopoly cannot be wished away or abolished by fiat,contestability constitutes a more attainable goal for government policy towardrestraint of monopoly power, and it suggests some rational rules for control ofmonopolistic industries.

We will show here that freedom of trade is critical in the provision of thebeneficent pressures of contestability in the industries of the developed econo-mies. In addition, we will show that in developing countries that are not pre-pared to expose their infant industries to the unrestrained rivalry of themultinational enterprises or established firms from industrial economies, con-testability analysis can offer some guidelines for policies that effectively denymarket power to the larger domestic enterprises and to those established underforeign sponsorship, without hampering their ability to compete, their efficien-cy, or growth in their productivity.

The Theoretical Framework

Before considering possible applications of contestability analysis, we needto define the concept and outline its virtues and limitations. What is contest-ability, and what can such a state of affairs achieve?

Perfect Contestability as a Standard for the Ideal Market

A perfectly contestable market can be defined as one in which entry and exitare perfectly costless. Freedom of exit in this sense is a critical part of the storybecause a little consideration confirms that it is merely the obverse of costless-ness of entry: if a firm can leave an industry whenever it chooses and can re-

2 The World Bank Research Observer, vol. 6, no. 1 (January 1991)

move all the funds it has invested without loss, its decision to enter willobviously have cost it nothing. In the reverse situation, in which exit does entailsome loss in value of assets (that is, in which some substantial portion of in-vestment is sunk), the decision to enter clearly incurs, at the very least, somecost in the form of entry risk-that is, risk greater than that incurred by a firmalready operating in the arena-and this will inevitably act as a deterrent toentry.

In practice, of course, entry and exit always require some time and this re-quirement can itself become a source of cost. In particular, delay subjects theentrant to the costly risk of retaliation by incumbents who can use the lag pe-riod to reduce prices, mount expensive advertising campaigns, or take othermeasures to make an invader regret its decision to enter. But even so, a marketcan remain highly contestable if it is feasible for an entrant to achieve contrac-tual relations with prospective customers. Such contracts can render the entrantimmune from retaliation. Where this is true, the incumbent can protect itselffrom intruders only by behaving well-for instance, by offering customers rea-sonable prices-all along. Such behavior will indeed foreclose entry opportuni-ties, but only by providing customers with all the benefits that an entrant couldbe expected to bring. In this the mere (perpetual) threat of entry can enforcegood conduct by oligopolists and even by monopolists.

None of this should be taken to imply that perfect, or near perfect, contest-ability is easy to achieve, or that it is pervasive throughout the economy. Insome markets heavy sunk costs are the norm, and contractual arrangementsbetween buyers and sellers are rare and difficult to achieve. Such markets willbe far from contestable. HIow widespread contestability is in practice is an em-pirical issue about which evidence is very limited. Moreover, the contestabilityof particular industries must be judged case by case; of course, no industry willsatisfy fully the theoretical requirements of perfect contestability, and no oneis sure how many can be taken to approximate that norm.

The Virtues of Perfect Contestability

It is convenient to classify under four headings the benefits that perfect con-testability offers to society: (1) absence of excessive (monopolistic) prices orprofits, (2) absence of inefficiency or waste, (3) absence of cross-subsidy orpredatory pricing, and (4) (with the possible exception of the single-firm case)Pareto optimality in pricing-that is, pricing consistent with efficiency in theallocation of economic resources to serve the desires of consumers. Another vir-tue of the contestable markets model, primarily of interest to economic theo-rists rather than to those concerned with application, is that contestabilityfacilitates the construction and analysis of economic models in many of theways (and for the same reason) that perfect competition does. The reason isthat contestable markets impose prices rather than demand curves on the firm.This fact permits the analytic simplification of dealing in parametric prices,

William J. Baumol and Kyu Sik Lee 3

without injection of the complications that arise in taking demand functionsinto account in an analysis of the behavior of the firm.

The most obvious benefit of perfect contestability is its complete preclusionof excessive profits and excessive prices. Even a pure monopolist in a perfectlycontestable market can earn no higher rate of return on its capital in the longrun than is attainable by a small firm under perfect competition. The reasonfor this startling result is that any profit in excess of the competitive rate ofreturn inevitably constitutes a profit opportunity for a potential entrant so longas sunk costs are zero. Excessive profits permit the entrant to undercut theincumbent's excessive prices to a degree that still leaves an attractive return tothe new enterprise. Even if the opportunity proves extremely transitory, costlessexit means that the entrant will have benefited from its incursion and will beprepared to repeat it should the incumbent ever again repeat its folly.

The same reasons underlie the second virtue of perfect contestability, its to-tal preclusion of waste and inefficiency. For if any firm incurs costs greater thanthe lowest permitted by current knowledge and the current state of technology,then a more efficient entrant can benefit in exactly the same way that excessiveprofits permit. The point is that excessive prices are in themselves always anentry opportunity, whether those prices are attributable to excessive profits, tounnecessary costs, or to anything else.

The absence of waste under perfect contestability encompasses more than isapparent at first glance. Obviously, it precludes allocative inefficiency-the sortof waste on which economic theory has tended to focus. That is, it precludesthe use of input x by firm A and input y by firm B when x will yield a morevaluable output if used by B and y will produce more if employed by A. Inaddition, the no-waste attribute of perfect contestability rules out X-inefficien-cy, the portmanteau class of inefficiencies called to our attention by HarveyLeibenstein, which encompasses all forms of inefficiency aside from the alloc-ative inefficiencies most frequently discussed in standard economic theory. Itincludes waste attributable to managerial ignorance, wrongdoing, or lack ofinitiative, among other sources. Perfect contestability is also incompatible withinefficient delays in the introduction of technological improvements becausesuch delays, too, entail unnecessary costs. That is, contestability requires justthe efficient rate of introduction of new productive techniques as they becomeavailable. (Contestability does not, however, eliminate the free-rider problemthat accompanies the acquisition of knowledge or the production of inventions,so that it does not guarantee a socially optimal level of expenditure on researchor an optimal flow of innovations.)

The absence of inefficiency guaranteed by perfect contestability extends toindustry structure. For example, if the cost structure of an industry is such asto make it most efficient, say, for fifteen firms to produce the industry's currentcombination of outputs, then in perfectly contestable equilibrium the industrywill not be made up of fourteen firms or sixteen or some other number differ-ent from the cost-minimizing fifteen enterprises. It is this efficiency property


that has permitted contestability theory to analyze what determines the struc-ture of an industry-that is, what makes one industry an oligopoly, another amonopoly, and a third highly competitive. Since contestability is compatiblewith any of these market forms, its theory can treat the choice among them asan endogenous selection. Moreover, under perfect contestability the choiceamong them will not be rnade on the basis of strategic considerations, but itmust instead always follow the cost-minimization requirements just described.Of course, since in reality contestability may often be absent, strategic and oth-er considerations can be important determinants of the structure of industry inpractice.

The third desirable attribute of perfect contestability, its prevention of pred-atory and cross-subsidy pricing, is a crucial consideration for control of mo-nopoly and market power, and prevention of their extension. Predatory pricingoccurs when an incumbent firm adopts prices that entail a deliberate sacrificeof profit, with the object of either driving a competitor from the market or pre-venting the entry of a new rival. This accomplished, the predator can raise pric-es again to yield excessive profits large enough to make up for the initial profitsacrifice. Since the prospect of excessive future profits is crucial to predation,and perfect contestability always prevents excessive profits by means of the zerocost of entry, with excessive profits impossible, predation must become unat-tractive.

Cross-subsidy, the second bete noire of those concerned with monopoly,raises more subtle considerations. A cross-subsidy is defined to be present whenone set of customers of a firm obtains its products at a price that entails a lossto the firm and that loss is made up for by excessive prices to other customers.The latter are then said to be providing a cross-subsidy to the former. Usually,this is an accusation brought by a firrn complaining that it is subjected to unfaircompetition by being forced to meet the indefensibly low prices permitted bycross-subsidy, while the allegedly subsidizing firm makes up for its subsidies inmarkets for its products where it happens to hold a monopoly.

This is not the place to go into the extensive debate over the circumstancesthat can lead to cross-subsidy or the possibility of rational motives for its adop-tion. Rather, having noted that it is a standard concern of those who deal withthe control of monopolisitic practices, we want to show how it is prevented byperfect contestability. For the purpose we observe first that economists todayusually propose the inicremental cost of a product as the criterion by which tojudge whether the price of the product entails the receipt of a cross-subsidy.That is, if the product is priced so as to yield a total revenue at least equal tothat product's incremental cost-the increase in the firm's total cost attribut-able to the supply of that product-it is said that the consumers of the productare receiving no cross-subsidy because they are paying the entire cost incurredby the firm in supplying the product to them. But under perfect contestability,any product price yielding revenue below incremental cost is incompatible withequilibrium, so cross-subsidy cannot occur under that regime. The reason is


most easily described by illustration. Consider a firm that supplies three prod-ucts, x, y, and z, and assume that the revenue from x is less than its incrementalcost. Then an entrant can open for business and offer to sell only y and z atprices slightly below the incumbent's. By dropping x from its product line, theentrant forgoes revenue from x but avoids x's still greater incremental costs,the result being a net addition to the entrant's profits over and above those for-merly enjoyed by the incumbent. Thus cross-subsidy permits the entrant toprofit by stealing the incumbent's profitable markets away, sharing with cus-tomers of those products part of the supplier's earnings previously going intothe cross-subsidy.

We will not take the space required to explain the rather more complex logicbehind the fourth desirable attribute of a contestable market: its enforcementof optimal prices in markets containing two or more suppliers. From the view-point of such issues as stimulation of growth in a developing country, this isundoubtedly an esoteric matter, not of immediate pertinence. For a full discus-sion, see Baumol, Panzar, and Willig (1988, chap. 11).

Enough has been said to indicate why perfect contestability can be consid-ered an economic ideal whose approximation by market behavior is to be en-couraged when feasible, and whose behavior it may be desirable to replicateby means of policy when markets cannot do the job.

Unsettled Matters in Contestability Analysis

Nothing in the preceding discussion should mislead the reader into believingthat everything is settled and definitive in the field of contestability analysis.On the contrary, reservations expressed by several commentators have suggest-ed that the subject was rather controversial, although the authors and most ofthe commentators seem since to have agreed on positions between the possibleextremes.

Perhaps most important, as has several times been suggested here, we do notknow yet how many markets or industries constitute reasonably close approx-imations to contestability and how many are far from meeting its requirements.A number of studies of individual industries have thrown some limited light onthis matter. Several such studies have indicated that oceanic freight transporta-tion is highly contestable, or at least prospectively so if some restrictive prac-tices are eliminated. Another piece of research has suggested the conclusionthat aluminum production is, in the long run, contestable to a rather surprisingdegree. (For a review of the literature, see Baumol, Panzar, and Willig 1988,chap. 17, sec. 17E.) In contrast, it is now generally agreed that air passengertransportation is considerably less contestable than had been thought at thetime of its deregulation in the United States, although this industry still liesfar from the opposite end of the spectrum (see, for example, Morrison andWhinston 1987). A few more studies for particular industries have been carriedout, but much remains to be determined.

6 TheWorldBankResearchObserver,vol. 6,no. 1 (January 1991)

Note, however, that the analytical power of the theory does not depend onthe ubiquity of contestability; indeed, its policy lessons apply primarily to in-dustries that are not contestable, whose regulation can be aided by contestabil-ity theory, which provides norms of behavior to which the regulated firms canreasonably be held.

A second important gap in knowledge about the subject in general relates tothe behavioral implications of imperfections in the degree of contestability. Ofcourse, real-world instances of perfect contestability, like those of perfect com-petition, probably do not exist. Schwartz and Reynolds (1983) have argued thatin this arena even a little imperfection is apt to be a dangerous thing, producinglarge departures from the ideal behavior to be expected in a regime of perfectcontestability. Opposing arguments have been offered-a series of studies byKessides (see, for instance, Kessides forthcoming) have provided evidence sug-gesting that trifling imperfection brings correspondingly trifling deviation fromideal behavior-but all the participants in the discussion seem to agree that theissue can be settled only empirically.

Similarly, some reservations about the theory, such as those expressed byWeitzman (1983), have been pretty well dealt with in the ensuing discussion,yielding an outcome intermnediate between the initial views of the discussants.Since the matters raised are fairly technical and are unlikely to be of great in-terest to readers of this journal, they will not be pursued here.

Applications

Let us turn now to application of some of these ideas, specifically to inter-national trade and to the efficiency of the public sector in developing countries.

Contestability and International Trade

Some industries are comnpelled by their technology to consist of a few com-paratively large firms within the boundaries of any one country. The reason isthat in those industries technology gives rise to economies of scale and scope.Here the term "economies of scope" (which is yet another concept that derivesfrom the literature on contestability) refers to a state of affairs in which a givenbundle of several (or many) goods or services can be produced more cheaplyby a single multiproduct firm than by a group of more specialized enterprises,each producing only a subset of the industry's products. Where economies ofscale and scope are available, one firm, or at most a small number of firms,will supply consumers' demands for the industry's products most cheaply, sincea multiplicity of small firms would make those economies impossible.

Under these circumstances, the number of firms that is viable in the industryis limited by the extent of the market. If there is something like a minimumefficient scale of operation for a firm in the industry, and if the quantity of the


industry's product demanded at a viable price is only a small multiple of min-imum efficient scale, then only a few enterprises can be expected to survive. If,in addition, entry into the industry requires substantial sunk costs, then neitherthe presence of a large number of firms nor the pressures of contestability maybe able to protect the interests of consumers against prices that exceed com-petitive levels and other related abuses.

It was Adam Smith who wrote of economies of scale being limited by theextent of the market. The same is obviously true of the number of competitorsactually in an industry-the U.S. market faces more competitors if it leaves it-self open to Japanese products than if it does not. But the reason is just a bitless obvious. It follows from the relation of the extent of the market to thetechnologically determined efficient scale of operation of the firm. An oversim-plified example will make the point. Suppose that in industry X a firm achieveslowest cost with an annual output of 10 million units of its product, and thatcosts rise sharply at any other output level. If a price that covers minimum unitcosts (including a normal return on capital) will elicit a demand for 30 millionunits of the product in the U.S. market, it follows that an isolated and sealed-off United States can support at most three firms in the industry. The extentof the market will condemn the industry to the status of a tight oligopoly. Butif an open world market is prepared to purchase 100 million units of the prod-uct at the same price, the industry will be able to support ten competitorsrather than three, and that can be enough to undermine the possibility of col-lusion that a three-firm industry might have invited.

As a crude rule of thumb, then, the number of firms that an industry cansupport is given by the ratio of quantity of product demanded by the marketat the lowest viable product price to the output of a firm with most efficientsize. Because free trade increases the extent of the market, it multiplies thatdemand without changing the technologically most efficient size of firm. As aresult, it increases the ratio of demand to efficient firm output and raises theviable number of participants in the market.

But freedom of trade can also affect the power of potential entry, and thatis where contestability analysis comes in. Even an industry whose activities re-quire an entrant to incur substantial sunk costs may have at its disposal a goodmany firms for which entry and exit are easy. These are firms which alreadyhave laid out the sunk expenditures needed for entry but which for some reasonare not now present in the market. It will be remembered here that sunk costsare pertinent to an optimal decision only before those sunk costs have beenincurred. Once a sunk outlay has been made it becomes irrelevant for a rationaldecision since that decision, by definition, cannot undo such a cost outlay ofthe past. Thus, if in Spain's industry X, production by an entrant requires asunk (and fixed) outlay of $10 million, a new Spanish firm can be deterred fromentering the industry. But it is no impediment to a French firm, already in op-eration in France, which decides to enter the Spanish market. The French com-pany's sunk cost has already been incurred and is not increased by its decision


to enter the market in another country. In this way, international trade can bea potent source of the sorts of pressures that impart contestability to a market.

Long-run trends have worked to enhance the power of such pressures. Withsecularly falling transport costs, with the advent of electronic means of trans-mission that have transformred many services from largely domestic undertak-ings into routine objects of international trade, and with related developments,each country's markets have become increasingly vulnerable to the threat of en-try, potential and actual, from other nations. One set of estimates-the valu-able figures provided by Maddison (1982)-suggests that in the 110 years after1870 the share of exports (and imports) relative to gross national product (GNP)

rose perhaps threefold or fourfold. That is, despite the startling rises in grossdomestic product (GDP) (arid in GDP per capita) over this period, foreign tradehas exploded even more sharply, and by a considerable margin. All of thisopens markets abroad to a country's domestic producers. But in exchange itmakes them increasingly vulnerable to competitive pressures from abroad.

This growth in the extent of the market has undoubtedly transformed someindustries that might otherwise have been subjects of concern to the antitrustauthorities into highly competitive entities. The fact that the U.S. auto industryfor years contained no more than four domestic firms, all of which could haveincurred substantial sunk costs, might have been grounds for attention by an-titrust agencies, but the competition of cars from France, Germany, Japan,Yugoslavia, and a variety of other places, and the threat that still others wouldjump in if domestic autoniakers were to attempt to extract monopoly profitsby considerably overpricing their products, have preserved the competitivenessof the U.S. auto market.

A protectionist policy, in addition to the better-known costs that it imposeson the public, is thus likely to undermine the contestability of some industriesand weaken that of others. Given the virtues of contestability described earlier,that too can impose a heavy burden on the public.

In an article on export-promoting trade strategy, Bhagwati (1988) supportsthe argument that economies of scale will be more fully exploited as interna-tional trade expands, although he acknowledges that systematic empirical evi-dence is not yet available on this effect. A study by Harris (1986), however, isindicative: this showed that "a 3.6 percent rise in GNP could follow from theunilateral elimination of Canadian tariffs, if the economies of scale are fullyexploited" (p. 39).

In a study on the extent of barriers to industrial countries' imports, Nogues,Olechowski, and Winters (1986, pp. 181-99) show that even though tariffbarriers to trade have declined significantly under the General Agreement onTariffs and Trade-from an average of about 40 percent in the mid-1930s to4 to 8 percent after the Tokyo Round of 1979-nontariff barriers have becomemore common. Nontariff barriers include import restrictions, voluntary exportrestraints, decreed import prices, antidumping and countervailing duties,and others. The study's estimates, based on data from the United Nations


Conference on Trade and Development, indicate that for the sixteen industrialeconomies examined, 27 percent of imports were subjected to nontariff barri-ers. The coverage for textiles was as high as 44.8 percent in 1983. The authorsestimated the value of imports subjected to nontariff barriers to be some$231 billion1 (based on 1981 trade flows)-almost half the total imports of thestate-trading East European countries. They also found that nontariff barriersare more prevalent on imports from developing countries than from industrialnations, and they concluded that "the growth of NTBs and their effect on in-ternational trade should be taken very seriously" (p. 197).

After a careful review and analysis of recent debates on export promotionpolicies, Bhagwati concluded: "Export promotion policies emerge with successfrom the detailed scrutiny offered in this article. Equally important is the factthat their successful adoption will require collaborative and intense efforts toensure that the protectionist threat, recently escalating, is not allowed to breakout into actual protection on a massive scale. The multilateral trade negotia-tions offer the only reasonable prospect for maintaining a momentum in favorof a freer world trading system. Failure to pursue them successfully, in a spiritof accommodation and mutual understanding of constraints and needs, willonly undermine what seems like the best mechanism for containing the protec-tionist threat" (1988, p. 48).

Contestable Markets and Public Sector Efficiencyin Developing Countries

The developing countries have been torn between the desire to take advan-tage of any available economies of scale and scope that promotion of largefirms can offer, and the conviction that in business small is beautiful, whereaslarge enterprise is inherently dangerous to society. One expedient widely adopt-ed in developing countries has been recourse to nationalization and politicalcontrol of the activities of the publicly owned enterprises; another has beendetailed governmental regulation to a degree not fully appreciated by casualobservers from industrial countries. For example, few economists in the wealth-ier countries can envisage a regulation prohibiting firms producing automobileparts (and hundreds of other industries) from having a total company invest-ment exceeding some $50,000. Nor can they conceive of a rule that requireseach large firm in the economy to obtain from the government every year alicense that specifies the maximum amount the company is permitted to pro-duce. Nor are they generally aware of cases in which the insurance industry isprohibited from using computers. Yet such rules have been adopted by somedeveloping countries whose per capita incomes are among the lowest in theworld.

It is hard to resist the conclusion that such constricting rules, however goodtheir intentions, significantly impede productive entrepreneurship and system-

10 The World Bank Research Observer, vol. 6,1no. 1 (January 1991)

atically redirect the energies of the entrepreneurs toward the undergroundeconomy. There is good reason to believe that tight regulation and politicalcontrol of nationalized firms in the developing countries have seriously ham-pered economic efficiency and substantially depressed outputs and standards ofliving.

Data for the period since World War 11 (see, for instance, Summers andHeston 1984) suggest that, measured in both productivity levels and per capitaincomes, the industrial coiuntries have been converging on a common standardof living and a common state of efficiency in production. Broadly speaking, theother industrial economies have been approaching (but not so far exceeding)the growth path of the United States, and at the same time, the range of theirperformance has, with some exceptions, been narrowing sharply.

Matters are very different for the developing countries. They have been con-verging neither on the standard of industrial economies nor even on a standardamong themselves. For example, according to Summers and Heston (1984),who report on a sample of seventy-two countries spanning all income groupsover the period 1950-80, real per capita incomes grew at an annual rate of3.1 percent in the countries they classified as industrial, but at only 1.5 percenta year in their low-income country group-less than half that of the industrialgroup. The poorer group ended up considerably further behind than they hadbeen at the beginning of the postwar period.

The record for intragroup convergence is similar. For the industrial groupthe Gini coefficient of real GDP per capita dropped precipitately from 0.302 in1950 to 0.129 in 1980, meaning that over the period the degree of inequalityin the standards of living among those countries fell sharply. But for the poorercountries the Gini coefficient actually rose slightly over the same interval, from0.103 to 0.112. Other independent calculations confirm these results. Clearly,at least some of the less affluent countries were not outstandingly successful inappropriating and instituting productive techniques either from the more afflu-ent economies or from one another.

There is no definitive and undisputed list of explanations for this lacklusterperformance. But it seems widely agreed that a likely culprit is the frequencywith which the poorer countries have restricted the range of activities left opento entrepreneurs. And, at least in part, this must be ascribed to fear of marketpower and of the income inequalities to which it leads. This implies that thenational income pie has inadvertently been kept small by measures intended toprevent entrepreneurial success from slicing the pie unevenly. The statistics sug-gest that the price of pursuing this goal has been high, with whatever domesticequality it has been able to achieve being purchased at the cost of exacerbatedinequality among nations.

What has contestability analysis to do with any of this? The answer is thatit may offer a way out of the dilemma that allows entrepreneurs and firmsthe freedom to innovate, grow, and take advantage of economies of scale andscope and yet protects the public from exploitation by them. For contestability


analysis provides some principles for controlling enterprises suspected of pos-session of market power that offer the public all the protection from monopo-listic exploitation that they would have received if the enterprises had beeneffectively constrained by the forces of market competition, yet do not impedeeconomic efficiency, thwart initiative, or handicap the exercise of entrepreneur-ship.

To understand what contestability theory can contribute to the analysis ofpolicy in this area, we must first inquire a bit further into the roots of the prob-lem. The role of economies of scale and scope is crucial-without the avail-ability of such economies no one would seriously advocate promoting orperhaps even tolerating bigness. It is only because the production techniques insome industries make production by large enterprises less costly to society thanproduction by small firms that developing countries have been willing to livewith larger enterprises. In any case, in industries with diseconomies of scale theproblem is not even likely to arise, because their larger firms with their neces-sarily higher costs will be apt to succumb to the competition of more efficientsmaller enterprises. Thus the problems that justify government control of in-dustry are peculiar to the presence of economies of scale.

The presence of economies of scale, however, poses several special problemsfor pricing policy. First, the marginal-cost-pricing rule that guides pricing underperfect competition-that is, the setting of the price of each product at a levelequal to marginal cost, as market forces require in a regime of perfect compe-tition-becomes a recipe for bankruptcy where economies of scale are present.(Marginal cost is defined as the increment in the supplier's total cost that re-sults from a one-unit increment in that firm's output. In other words, the mar-ginal cost of product X is the first partial derivative of the firm's total cost withrespect to the quantity of X it supplies.) The easiest way to see this is to recallthat in the single-product firm, economies of scale mean that the average costcurve must decline as output increases and that when average cost is fallingmarginal cost must be lower than average cost. In that case, a price equal tomarginal cost cannot possibly cover the firm's average cost and so total revenuemust be less than total cost. With economies of scale, then, marginal-costpricing is incompatible with viable operation. This, and the fact that survivalof multiple firms is neither possible nor desirable where economies of scaleoccur, rules out the concept of perfect competition as a guide for policy in thearena we are considering.

But whereas perfect competition is incompatible with economies of scale,perfect contestability is not. Thus the theoretical concept of perfect contestabil-ity, with the demonstrated benefits it offers to the public, can serve as a properguide to a government that wants to have it both ways-to permit relativelylarge firms to operate where they are appropriate, and to operate with a con-siderable degree of freedom, while extending effective government protectionof the interests of the public.

12 The World Bank Research Observer, vol. 6, no. I (January 1991)

Just what government action does contestability analysis call for in these cir-cumstances? We have seen that contestability prevents prices from going toohigh because beyond some level (that is, beyond the cost of operation of anentrant) they will bring an influx of new competitors. Prices in a perfectly con-testable market can approach this level, but they can never exceed it. Thus theforces of perfect contestability impose a ceiling on prices at their entry-inducinglevels.

Similarly, perfect contestability imposes a floor on prices-one that protectsthe legitimate interests of competitors, just as the ceiling protects the interestsof consumers. We have seen that this floor is set at the level of incremental cost,that is, while price in a perfectly contestable market can be nearly as low asincremental cost, it can go no lower.

Between these price bounds, the contestable firm is free to set its price at anylevel or rather to adapt the price of each of its products to the current state ofdemand for that item. And prices constrained in this way can yield no excessiveprofits; for under contestability excessive profits will always induce entry, sothat excessive profits mean that at least some of the firm's prices must exceedthe proper ceilings-the boundary between profitability and unprofitability ofentry.

What all this means for policy in a developing country is that such an econ-omy using contestability as its guide can protect both consumers and smallerrivals of large firms without recourse to drastic and economically costly mea-sures such as nationalization or highly restrictive constraints on enterprise. Ifcompetitive behavior is what the society wishes to achieve, it can attain thiswith the aid of some fairly straightforward rules, notably the adoption of upperand lower bounds for the prices of firms considered to possess market power,bounds calculated with care to replicate the bounds that would be enforced bymarket pressures under the ideal state of perfect contestability.

An Illustration from the Developing World

We conclude with a discussion of a specific situation to illustrate the appli-cability of contestable market analysis to developing countries. In Nigeria, asin many other developing countries in Africa, Asia, and Latin America, statemonopoly enterprises with substantial capital investment already in place failto deliver infrastructural services such as electric power, water, and telecommu-nications sufficient to meet demand. A World Bank research project that stud-ied the economic effects of infrastructural deficiencies in Nigeria (Lee andAnas 1989) confirmed that the costs of poor and unreliable public infrastruc-tural services fall heavily on manufacturing firms, which have to go to consid-erable expense to overcome these deficiencies. According to the establishmentsurvey conducted for the study, all but 14 of 179 firms have standby generatorsfor electric power. The firms without their own generators are mostly smallenterprises; they do not have the equipment not because the burden of poor


electric supply is less per unit of output for them, but rather because of econ-omies of scale in the generation of electricity. At the same time, the survey in-dicates that small firms are the ones that cannot afford expensive capitalinvestments for boreholes for self-provision of water, for vehicles for shipmentof products, for motorcycles for couriers, and for radio equipment to be usedin lieu of telephones. Hence, the burden of inadequate infrastructural servicesmay be much higher for small firms than large ones.

Nigeria, therefore, affords two extreme manifestations of inefficiency in theprovision of infrastructural services: first, a public sector with heavy capital in-vestment that fails to provide the services; second, the costly provision of thoseservices by the firms. These infrastructural failures have at least two sources.The first is relatively well understood and relates to shortcomings in theprocedures and technology used by the public sector, including problems inday-to-day management, operation, and maintenance of the facilities. The sec-ond, more complex and intransigent, relates to general deficiencies of admin-istration, bureaucracy, planning, billing and revenue collection, and personneltraining, and the lack of appropriate incentives for efficient management. Thissecond set of influences that lead to X-inefficiency, as described earlier, has re-mained a stubborn obstacle to progress over the years because investment inadditional facilities is futile if the necessary institutional organization and lo-gistical support systems are lacking.

For example, the Nigerian Electric Power Authority (NEPA), heavily subsi-dized by the government, had kept the price of electricity at 7 kobos (a littleless than 1 U.S. cent) a kilowatt-hour for the ten years preceding 1989. An in-dication of how much firms are willing to pay for reliable services is the factthat, despite this low rate (compared with 7 cents in industrial countries),individual firms have maintained expensive standby generators even though, onaverage, 75 percent of the generating capacity of the firms remained idle, andthe average cost of self-generation was 4.6 naira (60 cents) a kilowatt-hour. InJuly 1989 the NEPA price was raised to 30 kobos a kilowatt-hour, a fourfoldincrease. How quickly NEPA will improve the quality of services with the aidof this higher rate remains to be seen.

At best, the public sector's performance is likely to improve very gradually.Thus the most promising way to correct these basic inefficiencies, for Nigeriaas for many developing countries, is to combine efforts to improve public sectorperformance with incentives for private sector participation in the provision ofservices. In Nigeria, some minor regulatory changes can yield significant bene-fits to individual firms and households and to the economy. For example,Nigerian manufacturers are not allowed to sell any excess power they may pro-duce to the public agency or to other firms. The cost saving attainable by al-lowing such transactions can be large. More important, however, suchregulatory changes can make the electric utility market more nearly contestable.Large manufacturing firms with unused generating capacity can produce moreelectricity and sell their excess power to other firms. Because of the economies


of scale in electricity generation, the firms will be able to lower their price asthey increase their production of electric power. According to the survey results,at full utilization of their generating capacity, the average cost would declinefrom 1.75 naira a kilowatt--hour for firms with 20 to 49 employees to 40 kobosfor those having 1,000 or more employees. The price could decline until all ex-cessive profits are eliminated. This would put strong pressure on NEPA to per-form and compete with the manufacturing firms engaged in joint production(of electricity and their own products). Suppose such forces of contestabilityset a ceiling price at 40 kobos, the average generating cost of the firms in thelargest category. The customers would still buy electric power from these firmseven though NEP's price is lower at 30 kobos as long as NEPA's service remainsunreliable. If NEPA should deliver reliable power service at 30 kobos, which isunlikely, the manufacturers would exit from the electric utility market.

Contestability would attract third-party vendors (that is, private utility com-panies) into the market if they could compete with the existing manufacturingfirms selling electricity. Moreover, it is quite possible that these entrepreneursspecializing in the generation of electric power will produce electricity at anaverage cost lower than that of the large manufacturing firms, perhaps at30 kobos a kilowatt-hour, either because of economies of scale that can beexploited further or by use of better technology. If so, the manufacturing firmswill no longer find it profitable to produce and sell electric power as joint pro-duction. Thus the utility companies in this example would set the ceiling price.Indeed, a private entrepreneur began producing and selling electricity in theMaroco low-income area in Lagos. NEPA allowed this individual to operate therebecause this slum area was excluded from NEPA's network. The private operatorcharged 30 kobos a kilowatt-hour even in this low-income neighborhood andwas thriving and doing good business even though NEPA's price was only 7 ko-bos-a strong indication of users' willingness to pay for reliable services.

Another example of a minor regulatory change that would benefit Nigerianmanufacturers is permission for the management board of an industrial areato operate and manage "utility pools," which would include a wide range ofservices such as electric power generation, water supply, garbage collection,and telecommunications. Such an arrangement would make it possible to ex-ploit whatever economies of scope are available, along with economies of scale.

On the basis of what we can observe in Nigeria and from similar situationsin many other developing countries, a sensible way of ending the inertia thatpermits continued X-inefficiency in the public sector is to encourage privatesector participation in various infrastructural functions and activities, rangingfrom production of services to distribution, operation and maintenance, andbilling and revenue collection. The extent of private sector participation ina market for the supply of a particular type of activity will depend on the con-testability of such an activity. For example, production and distribution of elec-tricity will incur larger sunk costs than operations and maintenance activities.The contestability framework will make it easier to understand variations in

William ]. Baumol and Kyu Sik Lee 15

the emerging market structures of particular types of activity and will offer pol-icy choices. This seems a more promising strategy than the now popular priva-tization schemes, which often end up by replacing a state monopoly with aprivate monopoly with no regard for the underlying market structure in theeconomy.

The challenge is to provide an environment with appropriate incentive sys-tems for private entrepreneurs to engage in the supply of the pertinent infra-structural services, thus creating the market mechanisms needed for effectiveconduct of activities such as production, distribution, and maintenance. Theopportunity to create and expand such markets is opened by the failure of thepublic sector to perform in these countries and by the willingness of users topay for reliable services when they become available. To the extent that themarkets for particular infrastructural services are contestable because the sunkcosts involved in capital investment are low, it should be possible to liberalizerestrictions against the entry and exit of private entrepreneurs. In this new en-vironment the role of the government can be redefined as that of carrying outappropriate monitoring and supervision of market operations, with little or noprovision of such services by the public sector.

Conclusion

The contestable market framework, not yet ten years old, has already madeits mark on the design of policy. In the United States, it has explicitly been usedas the basis for the design of new approaches to policy and new relationsbetween government and industry. Moreover, it regularly enters the delibera-tions of regulatory agencies, the courts, and legislatures. This article has de-scribed the key elements of the theoretical framework and attempted to extendsome of the lessons learned to the areas of international trade and economicdevelopment. Our illustrative discussion of public-sector inefficiency in Nigeriahas shown applicability of this concept to developing countries. The issue isparticularly pertinent at this juncture, when many developing countries arepursuing development of the private sector and making use of privatization,and when the international trade regime is increasingly affected by the protec-tionist threat. One of the main messages of this article is that the design ofpolicies to improve economic efficiency through private sector participationrequires a clear understanding of the emerging market structure. The contest-ability analysis provides a tool for the construction of such policies.

Note

William J. Baumol is a professor of economics at New York and Princeton universities.Kyu Sik Lee is a principal economist in the Infrastructure and Urban Development Departmentof the World Bank.

1. One billion equals 1,000 million.


ReferencesThe word 'processed" describes works that are reproduced from typescript by mimeograph,

xerography, or similar means; such works may not be cataloged or commonly available throughlibraries, or they may be subject to restricted circulation.

Baumol, W. J., J. C. Panzar, and R. D. Willig. 1988. Contestable Markets and the Theory ofIndustry Structure. Rev. ed. San Diego, Calif.: Harcourt Brace Jovanovich.

Bhagwati, Jagdish N. 1988. "Export-Promoting Trade Strategy: Issues and Evidence." WorldBank Research Observer 3, no. 1: 27-57.

Harris, R. 1986. "Market Structure and Trade Liberalization: A General Equilibrium Assess-ment." In T. N. Srinivasan and J. Whalley, eds., General Equilibrium Trade Policy Modelling.Cambridge, Mass.: MIT Press.

Kessides, I. N. Forthcoming. "Market Concentration and Sunk Costs." University of Maryland.Processed.

Lee, Kyu Sik, and Alex Anas. 1989. "Manufacturers' Responses to Infrastructure Deficiencies inNigeria." Policy, Planning, and Research Working Paper 325. World Bank, Policy, Research,and External Affairs Office, Washington, D.C. Processed.

Maddison, Angus. 1982. Phases of Capitalist Development. New York: Oxford University Press.Morrison, S. A., and Clifford Whinston. 1987. "Empirical Implications and Tests of Contestabil-

ity Hypothesis." Journal of Law and Economics 30: 53-66.

Nogues, J. J., A. Olechowski, and L. A. Winters. 1986. "The Extent of Nontariff Barriers toIndustrial Countries' Imports." World Bank Economic Review 1, no. 1: 181-99.

Schwartz, Marius, and R. J. Reynolds. 1983. "Contestable Markets: An Uprising in the Theoryof Industry Structure: Comments." American Economic Review 73 (June): 488-90.

Summers, Robert, and Alan Heston. 1984. "Improved International Comparisons of Real Prod-uct and Its Composition, 1950-1980." Review of Income and Wealth 30 (June): 207-62.

Weitzman, M. L. 1983. "Contestable Markets: An Uprising in the Theory of Industry Structure:Comments." American Economic Review 73 (June): 486-87.

Williarn J. Baumol and Kyu Sik Lee 17

HOW MUCH DOES TRADEWITH THE SOUTHAFFECT WORKERS IN THE NORTH?

Adrian Wood

Does trade with developing countries have a small and benign effect on work-ers in industrial countries, as most economists have maintained, or a large andadverse effect, as the general public and advocates of protection believe? A re-view of the evidence suggests that neither of these positions is tenable. Themethods that economists have conventionally used to measure the effect ofNorth-South trade are biased downward. The true size of this effect remainsuncertain, but some recent studies suggest that it is much larger than previouslyestimated. Trade with the South has probably significantly altered the sectoralcomposition of employment in the North, shifting workers out of manufactur-ing and into nontraded services. More important, it has probably significantlyworsened the relative economic position of unskilled workers in industrialcountries, and may also have aggravated the problem of reconciling low infla-tion with low unemployment. Even so, the adverse side effects of trade withthe South are much smaller than is popularly supposed. And the popularremedy-protection-is clearly wrong. What is needed instead is more actionby governments to offset the reduction in the relative demand for unskilledlabor through training and education, job creation, and income redistribution.

T he effect of trade with developing countries (the South) on the laborforce of industrial countries (the North) is a matter of practical impor-tance and political controversy. Popular concern about job losses

has caused tariff and nontariff barriers to be erected against the manufac-tured exports of developing countries, making it harder for these countries toachieve prosperity through export-oriented industrialization. In contrast, most

The World Bank Research Observer, vol. 6, no. 1 (January 1991), pp. 19-363 1991 The International Bank for Reconstruction and Development/THE WORLD BANK 19

economists who have studied this issue have concluded that the effect of tradeon employment and wages is not a serious problem. Their findings have en-couraged governments and international organizations to fight against protec-tion for domestic industries, but the evidence has failed to convince largesections of the public. A few economists have also taken a pessimistic view(Bienefeld 1982, Seers 1983, Godfrey 1985, and Gray 1985).

Up to a point there is no disagreement. The public is aware of the gainsfrom trade that economists emphasize: inexpensive imported consumer goodsand well-paid jobs in export industries. But economists recognize too that tradehas uneven effects-positive as well as negative--on particular firms, indus-tries, and localities. Realizing the potential gains from trade thus requires thereallocation of labor and capital, which in some cases is not easy and, it iswidely agreed, deserves some government assistance. The controversial ques-tion is whether these adjustments are small and transitory, relative to othersources of structural change and relative to the gains from trade, or large andenduring.

Economic theory gives an ambiguous and incomplete answer to this ques-tion. It usually assumes that trade cannot reduce aggregate employmentbecause labor is mobile and wages are flexible. But theory suggests that trademay alter the distribution of income between skilled and unskilled workers orbetween labor and capital. And it leaves open the possibility that wage rigidi-ties may aggravate the transitory unemployment associated with changes intrade flows. Moreover, theory is silent on the magnitude and duration of theseside-effects of trade. All this makes empirical research essential. Such researchis not easy or uncontroversial, however, especially because of the difficulty ofassessing how the economy would have evolved in the absence of trade. Thisarticle reviews statistical research on the effect that trade with the South hason workers in the North and discusses the implications of the results for policy.

Import Penetration Ratios

One sort of empirical evidence that is sometimes cited to show that trade inmanufactured goods with developing countries has had little effect on the workforce in industrial countries is the low level of import penetration ratios (WorldBank 1987). These ratios measure the share of imports from developing coun-tries in industrial countries' consumption of manufactures. In a few sectors-notably toys, clothing, jewelry, sporting equipment, and footwear and leatherproducts-this share is substantial (Brodin and Blades 1986; Hughes andWaelbroeck 1981; Berthet-Bondet, Blades, and Pin 1988). But for manufactur-ing in total, the developing-country import share is low. Although the level de-pends on how penetration is measured, typical estimates for 1988 are about5 percent in the United States and 2.5 percent in Japan, with the EuropeanCommunity in between (UNCTAD 1989, chart 4).


Such figures are valuable as an antidote to exaggerated claims and fearsabout manufactured imports from developing countries. But they do not shedmuch light on the effect of such trade on the work force in industrial countriesfor three reasons. First, import penetration ratios relate to output rather thanemployment, a vital distinction because these imports are concentrated inlabor-intensive sectors,-products, and activities. Second, competition from im-ports can displace workers even where the import penetration ratio is low, sincedomestic firms may respond to this competition by introducing new productionmethods that use less labor. Third, the import penetration ratio is a one-sidedmeasure. It neglects the gains in employment generated by increased exports todeveloping countries-which are largely financed by the money that developedcountries pay for their imports from developing countries.

Studies of the Factor Content of Trade

Calculating the factor content of trade is simple in principle. It involves find-ing out how much labor is required to produce the goods that are exported todeveloping countries, and how much labor would have been required to pro-duce domestically those goods that are imported from developing countries.The net effect of trade is then estimated as the difference in labor contentbetween exports and imports, which can be calculated both in aggregate andfor particular industries and categories of workers.

Conventional Methodology and Results

Calculating the actual labor content of exports is reasonably straightfor-ward, although it requires large amounts of data. Calculating the hypotheticaldomestic labor content of imports is inherently more difficult. In practice,economists have usually based the calculation on actual use of labor in do-mestic industries that produce similar goods. For example, the hypotheticallabor content of apparel imports is estimated from the actual amount of laborused (directly and indirectly) in the domestic apparel industry. The underlyingassumption is that all the imports concerned "compete" with domestic pro-duction.

This assumption is clearly not appropriate for imports of many primaryproducts (for example, cocoa or tin). So with few exceptions (UNIDO 1986 andSapir and Schumacher 1985), economists have limited their calculations of thefactor content of trade t:o manufactured products, which are in any event thesource of most public concern. Some economists have also defined manufac-tures narrowly, excluding those (such as refined oil and canned food) that con-sist largely of a primary commodity. There have been other variations inmethodology as well. For example, some economists have omitted indirectlabor requirements for lack of suitable input-output tables. Moreover, factor

Adrian Wood 21

content calculations have been made for many different industrial countriesand time periods.

Although such variations lead to differences in the results, the examples intable 1 show that there is a high degree of similarity among them. The tablecompares the labor content of a representative bundle of exports to developingcountries with that of an equivalent bundle of imports from those countries.The imports contain more labor than the exports, which is to be expected sincethe competitive advantage of developing countries lies in labor-intensive prod-ucts, but the differences are small-only about 20 percent on average.

Because the labor content of manufactured imports exceeds that of manu-factured exports, a balanced expansion of trade with developing countriestends to reduce employment in manufacturing in industrial countries. In fact,however, this sort of trade is not balanced. Industrial countries as a group havealways had a large surplus in this category of trade. Thus some studies haveargued that such trade generally increases employment in manufacturing inindustrial countries (table 2). Exceptions are admitted. For example, theUnited States has a deficit in trade in manufactures with developing countries(especially those in East Asia), but even in this case the reduction in employ-ment is tiny. Similar calculations have been made on the basis of past or ex-pected changes in (rather than levels of) trade: some of these have suggestedpositive effects, others negative effects, but in almost all cases the net effectslook insignificant relative to total employment in manufacturing.

The estimated effects are more significant in particular industries. All thestudies have identified the same sets of winning and losing sectors. The losersinclude food processing, wood products, textiles and clothing, and leathergoods and footwear. These losses have been largely offset, however, by in-creased employment in the machinery and chemicals industries. Of course,

Table 1. Ratio of Labor Content of Exports to Labor Content of Competing Imports

Year Coverage Ratio

1983 Total OECD 0.801985 Germany, Fed. Rep. 0.781985 France 0.801985 Italy 0.991985 Netherlands 0.801985 Belgium 0.851985 United Kingdom 0.78

OECD, Organisation for Economic Co-operation and Development.Note: Data refer to industrial countries' trade in manufactures with developing countries. All calculations

are based on the narrow definition of manufactures, excluding processed primary products. For otherestimates, see OECD 1979, annex II; UNIDO 1978; Balassa 1979, 1986; Driver and others 1984; Kol 1986;Sapir and Schumacher 1985; and Schumacher 1983, 1984.

Source: Balassa 1989, Schumacher 1989.


Table 2. Effect on Employment of Trade in Manufactures with DevelopingCountries, 1983(thousands of employees)

NetRegion Exports Imports effect

United States 662 -795 -133EC 1,508 -644 864Japan 947 -134 813Other OECD 371 -204 167Total OECD 3,488 -1,776 1,712

EC, European Community.Source: Balassa 1989, table 6.5. For similar estimates, see Balassa 1986 and UNIDO 1986, chap. 3.

changes in the sectoral composition of employment mean that workers mustbear the cost of changing jobs. But enormous numbers of workers change jobseach year for other reasons, and the increase that results from trade with de-veloping countries has been estimated to be only a small percentage of totallabor turnover.

Of more concern is the uneven effect that trade has on skilled and unskilledworkers, and on male and female workers (but see Wood 1991). Studies of thefactor content of trade show that industries that export to developing countriesemploy higher proportions of skilled workers than those that compete withimports from developing countries (see table 3). As a result, skilled workersgain from this sort of trade, whereas semiskilled and unskilled workers lose.

Table 3. Skill Content of Trade in Manufactures with Developing Countries(percentage of all employees in trade-related sectors)

Import-Area and year Skill category Exports competing

OECD (1975) Professional and technical 11.2 6.0Other white collar 19.5 16.9Skilled manual 23.2 13.5Other manual 46.0 63.6

Germany, Fed. Rep. (1985) Unskilled and semiskilled 36.7 52.7

France (1985) Unskilled and semiskilled 31.4 42.8

Italy (1985) Unskilled and semiskilled 46.7 51.1

Netherlands (1985) Unskilled and semiskilled 42.6 53.0

Belgium (1985) Unskilled and semiskilled 42.9 50.2

United Kingdom (1985) Unskilled and semiskilled 41.3 51.1

Source: Balassa 1979, Schumacher 1989. For similar results, see UNIDO 1978, tables 26 and 29; de Grauweand others 1979; Sapir and Schumacher 1985; and Schumacher 1983, 1984.

Adrian Wood 23

However, just as these studies have concluded that the net effect of trade ontotal manufacturing employment is tiny, they have also concluded that itseffects on the skill composition of the labor force are insignificant in relationto the overall numbers of workers concerned.

Qualifications and Alternative Results

There are four substantial problems with the results from conventional stud-ies of the factor content of trade, three of which are bound to understate thetrue effects of trade, although one could err the other way. The first problem,which is the only one that might skew the results in the direction of overstate-ment, is that goods imported from developing countries are often much cheaperthan if they had been produced in industrial countries (Lydall 1975). Thus con-sumers buy more of the imports than they would have of the domestic versions,which makes the volume of imports greater than the associated decline indomestic output. As a result, using the level of imports as an indicator of thereduction in domestic output tends to exaggerate the loss of employment. Butsince the data refer to value rather than volume, and imports are cheaper, agiven value of imports tends to displace a greater value of domestic productionof the same goods. Which of these two biases predominates depends on thedegree to which lower prices actually increase consumers' demand.

The second problem concerns those calculations of the factor content oftrade that point to the industrial countries' large surplus in trade in manufac-tures with developing countries as evidence that such trade actually increasesemployment in industrial countries. The unsatisfactory logic of this argumentis apparent from its curious implication that trade in manufactures reduces em-ployment in developing countries, which conflicts head-on with much otherevidence (Krueger 1983). A more reasonable method, given that developingcountries pay for imports of manufactures largely out of their earnings fromexports of primary commodities, is to assume zero manufactured exportsfrom developing countries and an equal reduction in the value of manufacturedexports from industrial countries. This parallel reduction (assumed in manystudies of the factor content of trade) makes the effect on employment in in-dustrial countries appear less favorable. Indeed, even this assumption may betoo favorable, since the South's manufactured exports contain some primaryproducts that would otherwise have been exported unprocessed, and henceeach extra dollar of manufactured exports adds less than a dollar to the South'scapacity to import manufactures from the North.1

The third problem is that conventional studies of the factor content of tradeassume that all imports of manufactures from developing countries competewith identical goods made in industrial countries, and thus that their labor con-tent can be estimated from data on the corresponding domestic sectors. In fact,however, firms in industrial countries no longer manufacture many of the labor-intensive items imported from developing countries. This is true both of fin-


ished goods, especially when distinguished by type and quality, and of compo-nents and assembly operations. In technical terms, then, many imports ofmanufactured goods from developing countries are now "noncompeting."2 Thelikelihood of and theoretical reasons for this outcome were pointed out byKrueger (1977), who also directed some relevant empirical work (Krueger andothers 1981). But the implications of their insight for employment in industrialcountries have been neglectled. What it means is that conventional factor con-tent of trade calculations inevitably underestimate the amount of labor dis-placed by imports and hence understate the adverse net effect of trade with theSouth on the demand for labor in industrial countries. The only question con-cerns the degree of understatement.

To answer this question it is necessary first to assess the proportion of man-ufactured imports from the South that are noncompeting and then to find abetter way of estimating their hypothetical labor content. Both aspects of thistask pose serious difficulties. The first in principle simply involves determiningwhich imported goods are not produced domestically. To do this satisfactorilyin practice, however, would require matched sets of data on trade and produc-tion more detailed than any currently available.3

The next step is to estimate how much labor (and what skills) is actuallyused to produce the goods in the country of origin. These estimates must thenbe adjusted to allow for the likelihood that higher wages-especially forunskilled workers-in industrial countries would induce firms to use less labor-intensive methods of production. The main uncertainty in making this adjust-ment concerns how much production methods respond to the level of wages.Some allowance should perhaps also be made for the inferiority of productiontechnology in developing countries.

Wood (1990a) uses this method of estimating the labor content of noncom-peting imports and tackles some of the other problems of conventional esti-mates of the factor content of trade. Because of the uncertainties involved inaspects of this calculation, the results are expressed in ranges in the originalstudy, but the central results are shown in table 4. Not surprisingly, these esti-mates suggest a larger adverse effect on workers. The demand for labor in themanufacturing sector in industrial countries in 1990 is estimated to be about9 million person-years lower as a result of trade in manufactures with develop-ing countries. Moreover, this reduction appears to be concentrated almost en-tirely on unskilled labor, a result that is consistent with the earlier studies citedin table 3. The net effect on skilled workers (those with education or formaltraining beyond basic general schooling) is positive but small. Still, the net re-ductions in the central case are equivalent to only about 12 percent of totalmanufacturing employment, and only about 6 percent of economywide un-skilled employment. Furthermore, the contraction has not been sudden. Table 4refers to the cumulative effects of trade over two or three decades, although italso refers only to manufacturing.

Adrian Wood 25

Table 4. Effect of Trade with Developing Countries on Demand for Labor inManufacturing in Industrial Countries(cumulative effect to 1990)

Item Exports Imports Net effect

Absolute effect (millions of person-years)Skilled labor 2.0 -1.8 0.1Unskilled labor 2.0 -11.1 -9.2Total labor 3.9 -12.9 -9.0

Proportionate effect (percentage of 1985 manufacturing employment)Total labor n.a. n.a. -12.3

Proportionate effect (percentage of 1985 economywide employment)Skilled labor n.a. n.a. 0.1Unskilled labor n.a. n.a. -5.7Total labor n.a. n.a. -2.8

n.a. Not applicable.Source: Wood 1990a.

The results in table 4, in common with conventional estimates of the factorcontent of trade, are subject to a further basic problem (the last of the fourmentioned earlier), which means that they still underestimate the true effectof trade on the demand for labor. This problem is that studies of the factorcontent of trade neglect the effect of trade on technical progress. For displace-ment of domestic production by imports is not the only way in which compe-tition from low-wage developing countries reduces the demand for unskilledlabor in industrial countries. The same result arises when defensive technolog-ical innovation succeeds in keeping firms in industrial countries competitive byenabling them to produce with fewer workers. In order to detect this indirecteffect of trade, it is necessary to use an alternative method.

Time Series Studies

The main alternative to the factor content method involves examiningchanges over time. How strong a relation (if any) has there been over the pasttwo or three decades between changes in employment and wages in the Northand changes in the extent of trade with the South? This method requires iso-lating the effects of trade from those of other forces that have been acting onemployment and wages in the North.

Conventional Methodology and Results

The usual technique has been to work with an accounting relation that de-composes the sources of change in employment in particular sectors, and in the


economy as a whole, into several components. The first step is to separate the(negative) effects of labor productivity growth from the (usually positive)effects of output growth. TFhe second step is to split up the sources of outputgrowth, distinguishing domestic demand from export demand (which tends toraise domestic output) and import flows (which tend to lower it). The result isa table that shows how much of the actual change in employment over a par-ticular period is attributable to each of these components.

Early applications of this technique are surveyed in UNIDO (1978) and OECD

(1979). See also Krueger (1980) and Kol and Mennes (1983). The most compre-hensive recent calculations of this type, summarized in table 5, are by UNIDO

(1986). They cover the six largest industrial countries from 1975 to 1980 andare based on input-output tables, with alterations in intermediate input coeffi-cients as an additional source of change in employment. Exports and importsare not shown separately, but the net effects of trade with different tradingpartners are distinguished., (As in all the other studies of this type, the changein employment is not disaggregated by degree of skill.)

The table suggests that foreign trade is a minute source of change in totalindustrial employment. Rising net exports to all developing countries on bal-ance increased employment, but the number of jobs created was only about2 percent of the number attributed to increases in domestic demand. Similarly,although increases in net imports from nine selected developing countries onbalance reduced employnment in these six industrial countries, this reductionwas only about 2 percent of the reduction in employment attributed to risinglabor productivity. The effects of foreign trade tend to be larger in particularindustrial sectors than in these aggregate calculations (in which employment

Table 5. Sources of Change in Industrial Employment in Six Industrial Countries,1975-80(thousands of employees)

Source Change

Domestic demand 9,654

Foreign trade -156All industrial countries -503All developing economies 171

Nine developing economiesa -206Input-output coefficients 165Labor productivity -10,259

Total -830

Note: The six industrial countries are France, Federal Republic of Germany, Italy, Japan, United Kingdom,and United States. Industry consists of manufacturing, mining, and utilities.

a. Argentina, Brazil, Hong Kong, Republic of Korea, Malaysia, Mexico, Philippines, Singapore, andThailand.

Source: UNIDO 1986, table 3. 1.

Adrian Wood 27

gains from exports in some industries cancel out losses due to imports in otherindustries). But even at the sectoral level, the UNIDO calculations suggest thatchanges in domestic demand and productivity generally have a much largereffect on employment than changes in foreign trade.

These results are consistent with earlier studies. For example, Cable (citedin OECD 1979, table 4) analyzed employment changes in the United Kingdomduring 1970-75 in those manufacturing industries in which imports from de-veloping countries accounted for 2 percent or more of domestic consumption.He found that total employment in these industries fell about 8 percent overthis period, and that trade with developing countries contributed to the decline.But growth in productivity resulted in more than thirty times more job lossesthan did changes in trade with developing countries (which in more than halfthe industries were actually a source of job gains). Even in the apparel industry,which was especially hard-hit by imports from developing countries, growth inproductivity caused four times as many job losses as trade (UNIDO 1978,table 12). A more recent study of England's apparel industry in the 1970s, how-ever, attributes about half its decline in employment to imports from develop-ing countries (OECD 1985).

Qualifications and Alternative Results

These conclusions assume that changes in labor productivity are unrelatedto foreign trade. Martin and Evans (1981), Beenstock (1984), and Baldwin(1984) have pointed out that this assumption may be misleading, since compe-tition from low-cost imports is likely to prompt an increase in labor produc-tivity. The elimination of inefficient firms, intensified pressures to cut costs andinvest in technology, and the substitution of capital and skills for labor haveall been cited as mechanisms through which foreign competition can raise pro-ductivity.

The two most important such mechanisms are probably defensive labor-saving innovations and changes in product mix. The latter are induced, asexplained above, when competition from the South forces firms in the Northto abandon labor-intensive activities, either by eliminating particular productsaltogether or by splitting up the production process so that only the most cap-ital- or skill-intensive stages are performed domestically. Trade with developingcountries must thus have caused larger reductions in the demand for unskilledlabor in industrial countries than the accounting calculations in table 5 suggest.The question is how much larger.

This question requires disentangling the contribution of North-South tradefrom other causes of growth in labor productivity, including trade among in-dustrial countries. As Krueger (1980) and others have emphasized, capitalaccumulation and autonomous technical progress gradually pull up real wagesin industrial countries, making labor-intensive production uneconomic. Tradewith developing countries has similar effects, which are therefore hard to iso-


late, especially because this sort of trade is probably partly a response, ratherthan an addition, to these internal pressures.

Wood (1989) attempts to isolate the effects of trade on growth in productiv-ity by using data on nontraded sectors (such as construction and services) tocontrol for other internal influences. The evidence shows that labor productiv-ity in manufacturing in industrial countries grew faster after the early 1960s,when manufactured exports from developing countries began to take off. Thisacceleration is consistent with the relocation of labor-intensive production todeveloping countries and with the adoption of defensive labor-saving innova-tions in industrial countries. It could in principle also be due to faster autono-mous technological progress or growth in real wages, but these explanationsseem implausible because growth of productivity in nontraded sectors did notaccelerate. Indeed, the growth of productivity slowed in all sectors after the late1960s, but there was less of a deceleration in manufacturing than in nontradedsectors.

The influence of trade on the demand for labor in manufacturing is thusestimated by comparing the actual use of labor per unit of output with whatlabor use would have been if productivity in manufacturing, relative to produc-tivity in nontraded sectors, had continued to grow at the slower rate of the1950s.4 This method suggests that from the early 1960s to the mid-1980s thecumulative trade-induced reduction in the demand for labor in manufacturingin all industrial countries was between 10 million and 30 million person-years.5

Looked at another way, a result in the middle of this range would imply that,in the absence of expanded trade with developing countries, the manufacturingsector's share of total employment in industrial countries would have remainedconstant from 1965 to 1985 rather than declining (from 29 percent to 22 per-cent, although the absolute number of workers in manufacturing did not fall).6

These results are subject to several possible sources of error. The 1950s arelOt necessarily a good basis for comparison, and there are serious practicalproblems in measuring labor productivity in nontraded sectors. Faster growthof labor productivity in the manufacturing sector may have been due at leastin part to other causes, such as new technology. Moreover, even if the acceler-ation were due to trade, it might have been mainly caused by more competitionamong industrial countries. A similar lack of distinction between North-Southand North-North trade exists in other recent time series studies.

For example, Grossman (1982) estimates the determinants of employmentand wage changes, including competition from imports as well as other domes-tic influences, for nine selected manufacturing industries in the United Statesfrom 1967 to 1979. He measures competitive pressure from imports by the be-havior of import prices in these industries relative to the general level of pricesin the economy. (The usual approach is to look at changes in import quantities,but Grossman's price measure has the potential advantage of detecting com-petitive pressure even in industries that succeed in holding imports down.)Although all of these industries were popularly believed to have been severely

Adrian Wood 29

affected by competition from imports, Grossman's results suggest less effect.Only in one industry (radios and televisions) had price pressure from importsgreatly reduced employment; no import-induced loss in employment wasdetected in two other industries in which developing countries are importantcompetitors (footwear, and dolls, toys, and games), although pressure from im-ports appeared to have reduced wages slightly.

Grossman's findings on footwear conflict with case studies (see Hamilton1989) that show competition from developing countries has greatly reducedemployment in this industry. This reduction in employment is apparent inGrossman's data-a drop of about 4( percent in ten years-but his model at-tributes this contraction to other causes. It is possible that the case studies arewrong about causation and that footwear manufacturing is particularly sensi-tive to economywide increases in wages. But it is also possible that Grossman'smodel is misleading. For example, the lags seem. too short: the employment-reducing effects of lower import prices are compressed into eighteen monthsrather than spread over several years (as installed capacity is gradually elimi-nated or replaced by less labor-intensive equipment).7

Revenga (1989), using a different technique, concludes (as do Galbraith andCalmon 1990) that import competition has had a significant effect on wages inparticular manufacturing sectors. Revenga cites other studies with similar con-clusions regarding employment. All these studies involve the United States,where, several authors have also argued, the widening wage differential be-tween more and less skilled workers over the past decade or two has beencaused by increased exposure to trade (see Murphy and Welch 1988, Bluestone1990). Although North-North trade may have played a part (see OECD 1989),this outcome is remarkably consistent with the studies of North-South tradedescribed earlier, which show that trade with developing countries reduces therelative demand for unskilled labor. Moreover, although there appear to be nocorresponding studies in other industrial countries, there is evidence of increas-ing inequality of wages in the 1980s in Europe as well, most notably in theUnited Kingdom (OECD 1987).

Conclusions and Implications for Policy

Are the effects of trade with developing countries on workers in industrialcountries as small and benign as most economists have maintained? Or are theyas large and adverse as the advocates of protection and the general public areinclined to believe? Empirical evidence suggests that neither of these positionsis tenable. Trade with developing countries has probably had a significant effecton the sectoral composition of employment in industrial countries, and it hasprobably also significantly worsened the relative economic position of unskilledworkers. But its impact cannot conceivably have been as large or as adverse asis sometimes alleged.


Economists have tried to measure the effect of trade with the South on work-ers in the North by looking at the factor content of trade and at the sourcesof change in employment. Both methods, as conventionally applied, suggestthat the net effect is insignificant-at least in relation to total employment andin relation to other sources of structural change. Both methods, however, aresubject to clear downward biases that undermine the credibility of their results.These results are also hard to reconcile with the substantial gains consumersin industrial countries and workers in developing countries have apparently re-alized from trade, and with the strength of the demand for protection in indus-trial countries.8

It is not possible to arrive at a confident or precise conclusion about howlarge the effect has been or how it has affected different groups of workers.Too little work has been done, and the results are diverse and open to variousdoubts. What can be done at present, though, is to establish a fairly confidentupper limit on the magnitude of these effects and to show that even this worstcase is not nearly as bad as has sometimes been suggested.

Trade with the South has had two effects on the composition of the demandfor labor in the North. It has shifted workers out of such traded sectors as man-ufacturing and into nontraded sectors, such as construction and services; andit has reduced the demancd for unskilled labor relative to the demand for skilledlabor. These alterations in the composition of the demand for labor may alsohave contributed to aggregate unemployment because of immobility and otherrigidities in the labor markets of industrial countries. But popular concern thattrade with the South simply reduces the number of jobs in the North is mis-placed. Indeed, expansion of trade can sometimes raise the aggregate levelof employment by increasing efficiency, income, and the demand for labor (seeOECD 1989).

As for the magnitude of the effect on sectoral employment, the largest num-ber mentioned in any study is a cumulative net reduction of 30 million person-years in the demand for labor in manufacturing. Although this number seemslarge, it is actually rather small in relation to total employment in industrialcountries (more than 300 million), especially since this decline has been spreadout over a quarter of a century. On an annual average basis, it would have al-tered the sector of employment of less than 0.5 percent of the labor force,which should be compared with the roughly 20 percent of workers who changejobs each year (OECD 1989). Of course, this net reduction conceals larger grossflows into and out of particular manufacturing industries. Moreover, the shiftsin employment have been more noticeable in recent years because the absolutesize of the annual increases in imports from the South has become larger. Inaddition, employment in other traded sectors-agriculture, mining, and someservices-may have been reduced by trade with the South. But even if the30 million person-years estimate were doubled, this trade-induced increment tostructural change and the reallocation of labor would appear proportionallymodest and unlikely to pose a serious or enduring problem.

Adrian Wood 31

The magnitude of the effect of trade with the South on the skill compositionof the demand for labor is probably similar to that of the sectoral effect. Inmanufacturing, the reduction in demand for less skilled workers has actuallybeen greater than the net reduction in total demand for labor (because the de-mand for skilled labor has increased), but apparently not by much, so the figureof 30 million person-years probably again roughly indicates its upper limit.This is equivalent to approximately 20 percent of the current unskilled laborforce of industrial countries (imperfectly measured as those with less than acomplete secondary education). There is almost no evidence on the other trad-ed sectors, but including them might substantially increase this figure and per-haps even double it (Wood 1990a). This shift, however, appears much smallerwhen one recalls that it has been spread over two or three decades.

Thus the change each year in the skill structure of the demand for labor inthe North caused by trade with the South cannot have been very large, butneither has it been trivially small. Moreover, changes in skill structure pose po-tentially more serious and durable problems than changes in sectoral structure.Expanded employment opportunities for professional, technical, and skilledmanual workers are of little help to unskilled workers whose jobs have disap-peared. The market's solution to this mismatch is a wider wage differential be-tween skilled and unskilled workers, which should increase the relative demandfor unskilled labor and reduce its relative supply by strengthening the incentivesto acquire skills. And there is evidence, particularly in the United States butalso in Europe, that a considerable widening of skill differentials has occurredin the past decade or so, although it is not clear how much has been caused byincreased trade with the South.

This widening of wage differentials should in principle prevent any enduringproblem of unemployment among unskilled workers. But it also increases ine-quality, and for this reason may be resisted-by unions, minimum wage laws,or minimum income provisions in social security systems. This sort of relativewage rigidity would perpetuate the initial shortage of skilled workers and sur-plus of unskilled workers, making it harder for governments to reconcile theconflicting policy objectives of low inflation and low unemployment.9 There isstrong evidence that these two macroeconomic policy objectives have becomeharder to reconcile in industrial countries over the past two decades (Johnsonand Layard 1986). It is not possible to say how much of this problem-if any-has been caused by increased trade with developing countries in conjunctionwith relative wage rigidity. But it seems rather likely that this is part of theexplanation, especially in Europe, where there are more institutional obstaclesto wage flexibility and where there is persistent large-scale unemployment ofunskilled workers.

So the effect of trade with developing countries on workers in industrialcountries may well be more of a problem than most economists acknowledge.But it has been much less of a problem than is popularly supposed. Moreover,the popular remedy, protection, is clearly wrong, even if considered only from


the standpoint of industrial countries, without regard to the damage it inflictson developing countries. Study after study has shown that the cost of protec-tion to society at large far outweighs the gain to those in protected sectors(de Melo and Tarr 1988).

But there are losers from trade, and governments can reduce and resist po-litical pressure for protection by helping the losers in other ways. The usualsort of help has involved assistance in moving from one sector to another,which is highly desirable (see OECD 1989). But the changes that trade causes inthe sectoral composition of the demand for labor are less problematic than theassociated changes in its skill composition. To tackle the problem of reduceddemand for unskilled workers, governments of industrial countries should in-crease expenditures on education, training, and the creation of unskilled jobs.The disposable incomes of low-paid workers should also be raised through thetax and welfare systems. Unfortunately, however, over the past decade mostindustrial countries have shifted their policies in the opposite direction. Thishas harmed both the less prosperous citizens of their own countries and(through greater pressure for protection) the much poorer people in developingcountries.

Notes

Adrian Wood is a professorial fellow of the Institute of Development Studies at the Universityof Sussex. In the course of the research on which this article is based, many people kindly pro-vided advice, data, and publications. Valuable comments on a draft research report were givenby David Evans, Martin Godfrey, Anne Krueger, Jaime de Melo, Sherman Robinson, DieterSchumacher, and Alan Winters, and on this article by Terence Moll and two referees. Researchassistance from Trevor King and financial support from the U.K. Overseas Development Admin-istration and the U.K. Economic and Social Research Council are also gratefully acknowledged.

1. The primary product content of the South's manufactured exports (narrowly defined) was28 percent in 1985 (Wood 1990b). But increases in manufactured exports reduce net primary ex-ports by less than this figure suggests because manufactures imported from the North containsome primary products purchased from the South.

2. The sharp distinction between competing and noncompeting imports has been replaced inrecent computable general equilibrium models by the assumption of imperfect substitutabilitybetween imports and domestically produced goods. Such models have been used to analyze theeffect of trade on employment, but not specifically the effect of North-South trade (OECD 1989).

3. The most disaggregated such data set is compiled by the OECD, which is at the four-digitlevel of the International Standard Industrial Classification (Brodin and Blades 1986; Berthet-Bondet, Blades, and Pin 1988).

4. Looking at interperiod changes in differences in productivity among sectors avoids theobvious objections to considering only the change in productivity in manufacturing over time,or only the difference between growth of productivity in manufacturing and in nontraded sec-tors. In principle, this calculation also requires estimating the effect of trade on the real outputof manufactured goods in industrial countries, but in practice this effect appears to have beennegligible.

5. The range of results reflects the use of alternative sources of data. The upper limit of30 million person-years is more tentative than the central figure and lower limit. One source ac-tually suggests a much higher upper limit, but this result was rejected on grounds of inconsis-tency with other evidence.

Adrian Wood 33

6. These figures relate to all industrial countries together; the experience of individual coun-tries has varied. That the share of employment in manufacturing fell in industrial countries dur-ing this period in a way that could not have been predicted by earlier data (either time series orcross-section) is shown in Syrquin (1988, pp. 239-42) and Syrquin and Chenery (1989, especiallytables 4, 14, 15, and 17).

7. Thus the effect of import competition may be captured mainly in the time trend term, whichis the only industry-specific independent variable other than the import price index. In a subse-quent paper applying the same methodology to the U.S. steel industry, Grossman (1986) testslags of up to twenty-four months, but his model suggests full adjustment within eighteen months.The results in this paper imply that import competition substantially reduced employment,accounting for about a quarter of the actual decline in employment in the steel industry during1976-83 and about a half during 1979-83. This intensified competition, however, was mainly theresult of the appreciation of the dollar, and over the full period an unexplained time trend ac-counted for more of the decline in employment.

8. Baldwin (1984) summarizes the results of many studies of protection in industrial countries.Trade barriers tend to be highest in labor-intensive industries with high proportions of unskilledworkers, low average wages, and large imports from developing countries.

9. More precisely, such wage rigidity would tend te increase the "natural" or "non-accelerating-inflation" rate of unemployment. For a simple account of how a decline in the de-mand for unskilled labor relative to that for skilled labor could indirectly increase aggregateunemployment, see Wood (1988).

References

The word "processed" describes works that are reproduced from typescript by mimeograph,xerography, or similar means; such works may not be cataloged or commonly available throughlibraries, or they may be subject to restricted circulation.

Balassa, Bela. 1979. "The Changing International Division of Labor in Manufactured Goods."Banca Nazionale del Lavoro Quarterly Review 130: 243-85.

. 1986. "The Employment Effects of Trade in Manufactured Goods between Developedand Developing Countries." Journal of Policy Modelling 8, no. 3: 371-90.

--. 1989. New Directions in the World Economy. London: Macmillan.

Baldwin, R. E. 1984. "Trade Policies in Developed Countries." In R. W Jones and P. B. Kenen,eds., Handbook of International Economics, vol. 1. Amsterdam: North-Holland.

Beenstock, Michael. 1984. The World Economy in Transition. 2d ed. London: George Allen andUnwin.

Berthet-Bondet, Claude, Derek Blades, and Annie Pin. 1988. "The OECD Compatible Trade andProduction Data Base 1970-1985." OECD Department of Economics and Statistics WorkingPaper 60. Paris. Processed.

Bienefeld, Manfred. 1982. "The International Context for National Development Strategies." InManfred Bienefeld and Martin Godfrey, eds., The Struggle for Development. Chichester, U.K.:Wiley.

Bluestone, Barry. 1990. "The Impact of Schooling and Industrial Restructuring on Recent Trendsin Wage Inequality in the United States." American Economic Review, Papers and Proceedings(May): 303-7.

Brodin, Anders, and Derek Blades. 1986. "The OECD Compatible Trade and Production DataBase 1970-1983." OECD Department of Economics and Statistics Working Paper 31. Paris.Processed.

de Grauwe, P., W. Kennes, T. Peeters, and R. van Straelen. 1979. "Trade Expansion with LessDeveloped Countries and Employment: A Case Study of Belgium." WeltwirtschaftlichesArchiv 115, no. 1: 99-115.


de Melo, Jaime, and David Tarr. 1988. "Welfare Costs of U.S. Quotas on Textiles, Steel andAutos." Policy, Planning, and Research Working Paper 83. World Bank, Policy, Research,and External Affairs Office, Washington, D.C. Processed.

Driver, Ciaran, Barry Naisbitt, and Andrew Kilpatrick. 1984. The United Kingdom EmploymentEffects of Trade Expansion with EEC and NiCs. National Economic Development Office Work-ing Paper 17. London: HMSO.

Galbraith, J. K., and P. D. P. Calmon. 1990. "Relative Wages and International Competitivenessin U.S. Industry." Working Paper 56. Lyndon B. Johnson School of Public Affairs, Universityof Texas, Austin. Processed.

Godfrey, Martin. 1985. Global Unemployment. Brighton, U.K.: Wheatsheaf.

Gray, J. P. 1985. Free Trade or P'rotection? A Pragmatic Analysis. London: Macmillan.

Grossman, Gene M. 1982. "The Employment and Wage Effects of Import Competition in theUnited States." NBER Working Paper 1041. National Bureau of Economic Research,Cambridge, Mass. Processed. Reprinted in Journal of International Economic Integration 2(1987): 1-23.

. 1986. "Imports as a Cause of Injury: The Case of the U.S. Steel Industry." Journal ofInternational Economics 20 (May): 201-23.

Hamilton, Carl B. 1989. "The Political Economy of Transient 'New' Protectionism."Weltwirtschaftliches Archiv 125, no. 3: 522-46.

Hughes, Helen, and Jean Waelbroeck. 1981. "Can Developing-Country Exports Keep Growingin the 1980s?" The World Economy (June): 127-48.

Johnson, G. E., and P. R. G. Layard. 1986. "The Natural Rate of Unemployment: Explanationand Policy." In Orley Ashenfelter and Richard Layard, eds., Handbook of Labor Economics,vol. 2. Amsterdam: North-Holland.

Kol, Jacob. 1986. "Key Sectors, Comparative Advantage and International Shifts in Employ-ment." Paper presented at Eighth International Conference on Input-Output Techniques,Sapporo, Japan, July 28-August 2. Processed.

Kol, Jacob, and L. B. Mennes. 1983. "Trade and Industrial Employment: An Accounting forGrowth Approach with an Application to the Netherlands." In Istvan Dobozi and PeterMandi, eds., Emerging Development Patterns: European Contributions. Budapest: EuropeanAssociation of Development Institutes.

Krueger, Anne 0. 1977. Growth, Distortions and Patterns of Trade among Many Countries.Princeton Studies in International Finance 40. New Brunswick, N.J.: Princeton UniversityPress.

. 1980. "LDC Manufacturing Production and Implications for OECD Comparative Advan-tage." In I. Leveson and J. W. Wheeler, eds., Western Economies in Transition. Boulder, Colo.:Westview.

- . 1983. Trade and Employment in Developing Countries. Vol. 3, Synthesis and Conclu-sions. Chicago: University of Chicago Press.

Krueger, Anne O., H. B. Lary, T. Monson, and N. Akrasanee, eds. 1981. Trade and Employmentin Developing Countries. Vol. 1, Individual Studies. Chicago: University of Chicago Press.

Lydall, H. E 1975. Trade and Employment. Geneva: International Labour Office.

Martin, J. P., and J. M. Evans. 1981. "Notes on Measuring the Employment DisplacementEffects of Trade by the Accounting Procedure." Oxford Economic Papers 33: 154-64.

Murphy, Kevin, and Finis Welch. 1988. "The Structure of Wages." University of Chicago, Uni-versity of California at Los Angeles, and Unicon Research Corporation. Processed.

OECD (Organisation for Economic Co-operation and Development). 1979. The Impact of theNewly Industrializing Countries on Production and Trade in Manufactures. Paris.

1985. Costs and Benefits of Protection. Paris.

1987. "Occupational Differentials in Earnings and Labor Demand." OECD EmploymentOutlook (September): chap. 3.

Adrian Wood 35

. 1989. "Trade and Employment." Paris. Processed.

Revenga, Ana L. 1989. "Wage Determination in an Open Economy: International Trade and U.S.Manufacturing Wages." Harvard University, Cambridge, Mass. Processed.

Sapir, Andre, and Dieter Schumacher. 1985. "The Employment Impact of Shifts in the Compo-sition of Commodity and Services Trade." In OECD, Employment Growth and StructuralChange. Paris.

Schumacher, Dieter. 1983. "Intra-Industry Trade between the Federal Republic of Germany andDeveloping Countries: Extent and Some Characteristics." In P. K. M. Thakaran, ed., Intra-Industry Trade. Amsterdarr: North-Holland.

. 1984. "North-South Trade and Shifts in Employment." International Labor Review 123,no. 3 (May-June): 333-48.

. 1989. "Employment Impact in the European Community Countries of East-West TradeFlows." International Employment Policies Working Paper 24. International Labour Office,World Employment Programme, Geneva. Processed.

Seers, Dudley. 1983. The Political Economy of Nationalism. Oxford: Oxford University Press.

Syrquin, Moshe. 1988. "Patterns of Structural Change." In Hollis Chenery and T. N. Srinivasan,eds., Handbook of Development Economics, vol. 1. Amsterdam: North-Holland.

Syrquin, Moshe, and Hollis Chenery. 1989. Patterns of Development: 1950 to 1983. World BankDiscussion Paper 41. Washington, D.C.: World Bank.

UNCTAD (United Nations Conference on Trade and Development). 1989. Trade and DevelopmentReport. New York.

UNIDO (United Nations Industrial Development Organization). 1978. "The Impact of Trade withDeveloping Countries on Employment in Developed Countries." UNIDO Working Paper onStructural Change 3. Vienna. Processed.

. 1986. Industry and Development Global Report. Vienna.

Wood, Adrian. 1988. "How Much Unemployment Is Structural?" Oxford Bulletin of Economicsand Statistics 50, no. 1 (February): 71-81.

. 1989. "North-South Trade and the Demand for Labor". Final Report of ESCOR ResearchScheme R4258. Overseas Development Administration, London. Processed. Also forthcomingfrom Oxford University Press.

. 1990a. "The Factor Content of North-South Trade in Manufactures Reconsidered."Institute of Development Studies, University of Sussex. Processed.

. 1990b. "What Do Developing-Country Manufactured Exports Consist Of?" Institute ofDevelopment Studies, University of Sussex. Processed.

-. 1991. "North South Trade and Female Labor in Manufacturing: An Asymmetry."Journal of Development Studies 27, no. 2 (January).

World Bank. 1987. World Development Report 1987. New York: Oxford University Press.


ORGANIZED LABOR, POLITICS,AND LABOR MARKET FLEXIBILITYIN DEVELOPING COUNTRIES

Joan M. Nelson

Organized labor is usually viewed as an obstacle to labor market adjustment.But unions' responses to adjustment programs in fact range from militantopposition to acquiescence or even explicit cooperation. Three sets of variablesshape these responses: the strength and characteristics of the union movementitself; economic cycles; and political institutions and their ties to unions.

Strength of the labor movement: In industrial democracies, an aggressivestance on wages tends to be associated with moderately strong unions.Small or weak unions are less militant, as might be expected; more sur-prisingly, large and powerful unions also tend to be more moderate, pri-marily because of their greater participation in consultation anddecisionmaking at the national level. The experience in developing coun-tries is somewhat different: only a few of these have strong labor move-ments, and among these, militancy is common. The large share of wagelabor in the public sector complicates comparisons with industrial nations,since governments as employers clearly behave differently from private em-ployers, particularly in hard times.

* Economic cycles: Depression almost invariably reduces militancy in devel-oping as well as industrial nations.

* Political institutions: The nature of the political regime-democratic or au-thoritarian-is only roughly associated with how governments handlelabor relations, and correspondingly with the options available to unions.More important in shaping unions' behavior are the nature of the politicalparty system and how unions are connected with parties. The conditionsneeded to gain workers' cooperation are analogous to those which encour-age business to invest: political stability, a voice in policy that affects their

The World Bank Research Observer, vol. 6, no. 1 (January 1991), pp. 37-56C 1991 The International Bank for Reconstruction and Development/THE WORLD BANK 37

interests, and, arising from these, the confidence that current sacrifices willultimately yield a fair share of future benefits.In developing as in industrial countries, the probable response of labor

unions is crucial to government decisions about wages, prices, and employ-ment. Organized labor is usually seen as a potential obstacle to labor mar-

ket adjustment, a source of rigidities impeding shifts in relative wages andreallocation of labor among sectors, and a center of resistance to stabilizationmeasures that entail compression of demand. Yet in some countries, labor con-federations have cooperated with the government and private employers to fa-cilitate adjustment. What makes organized labor confront or cooperate, andwhat empowers it to slow or stall attempts to increase labor market flexibility?What political circumstances and approaches have persuaded organized laborto acquiesce in such attempts in the past?

Both theory and evidence on these questions are fragmentary, and partly be-cause of poor data, much of the literature on organized labor in developingnations is too descriptive and narrowly focused to provide an adequate basisfor comparative analysis. Most of the analytic and comparative researchappropriate for the purpose is based on experience in industrial democracies.This article therefore reviews key findings from research on industrial countriesand considers the extent to which they are relevant to developing nations. Thefocus is on two categories of variables that affect labor militancy: the size,strength, and structure of the union movement itself, and the economic cycle.The discussion then turns to how political institutions and stability shapeunion behavior.

Union Structure and Union Militancy

Classic theory assumes that free play of market forces will ensure flexibilityin labor markets and that decentralized wage bargaining should restrain wageincreases. Conversely, one expects strong unions and centralized wage bargain-ing to be associated with comparatively rapid growth of real wages in the or-ganized sectors and with relatively high levels of unemployment. Indeed, oneinfluential line of analysis argues that over time, powerful organized interestgroups (including but not confined to labor unions) may lead to such rigidities,such inability to adapt to changing circumstances, as to cause the decline ofgreat powers (see Olson 1982).

Evidence from Industrial Democracies

Recent research on labor movements in advanced industrial democraciesposes puzzles for these assumptions. This research finds that during the 1960sand 1970s countries with large, highly centralized union movements tended toexperience low wage push, low unemployment, and low inflation.1 Moreover,


strike activity (as a measure of union militancy) was negatively correlatedwith strong labor movements and highly centralized wage bargaining.Cameron's (1984) findings are an example. He developed five indices of thepower of organized labor in eighteen industrial nations, using data from 1965through the early 1980s. The indices measured union membership relative tototal labor force, concentration or fragmentation of union confederations, theirpower in collective bargaining, the scope of such bargaining, and the extent ofarrangements for worker participation in decisions at the firm level (such asworks councils and worker representation on company boards). Cameronfound that each of these indices correlated negatively with strike activity, levelsof unemployment, increases in both nominal and real earnings, and consumerprices between 1965 and 1982. Some of the correlations-with the organiza-tional unity of labor and workers' participation in decisions in particular-were strong (Cameron 1984, table 7.7).

More recent studies suggest that the relationship between centralized wagebargaining and militant wage demands may be humpbacked (shaped like aninverted U) rather than monotonic (decreasing steadily). Wage push is leastevident where bargaining is highly decentralized (conducted largely at the levelof individual firms), and where it is highly centralized; wage pressures seemstrongest where bargaining is moderately centralized (at the sector or industrylevel). Calmfors and Driffill (1988, pp. 17-24) find a humpbacked relation be-tween centralization-defined as "the extent of inter-union and inter-employercooperation in wage bargaining with the other side"-and several indices ofemployment and inflation. Freeman (1988) uses the degree of wage dispersionacross industries as a strong indicator of market structure and finds that coun-tries with very high or very low dispersion, which are usually the countries withhighly centralized or highly decentralized bargaining in their labor markets,have increased employment more rapidly and wages less rapidly than countrieswith intermediate degrees of centralization.

What mechanisms are at work here? Two different explanations are offeredfor these empirical findings. Calmfors and Driffill see shifts in incentives(rational expectations about gains and losses) as the principal impetus for in-transigence in negotiations, depending on the degree to which wage bargainingis centralized. When bargaining is conducted at the firm level, unions have littlemarket power. An increase in wages is likely to bring a large drop in employ-ment, because the isolated firm cannot raise output prices and remain compet-itive. The market power ofunions increases if wage bargaining is conducted atthe industry or sector level. But as the centralization of wage bargaining in-creases, greater power must be balanced against greater risk that wage gainswill affect price levels, thus eroding the real value of nominal wage hikes. In-creased centralization on the union side of the negotiating table usually goesalong with correspondingly increased centralization on the employer side, andemployers also face shifting incentives for resisting or accommodating uniondemands at different levels of aggregation. The outcome is a nonmonotonic or

Joan M. Nelson 39

humpbacked relation between increased centralization and labor militancy:unions have the strongest incentives to bargain aggressively at intermediate lev-els of centralization.

An alternative explanation rests on the theory of democratic corporatist ar-rangements for determining wages and broader social policies developed in the1970s and 1980s, mainly with reference to Western and Northern European na-tions. This use of the term corporatist may call for explanation. The same termoften refers to certain kinds of political systems, including some-especiallyduring the 1970s in Latin America-in which relations between unions andgovernments are entirely different from those described in the following.Nevertheless, this is the label usually used to refer to the pattern of labor mar-ket arrangements discussed here, and I have followed this usage rather thancompounding confusion by inventing a new label.2

Even when the concept of corporatism is used narrowly with reference tolabor market arrangements, definitions and interpretations of its implicationsvary. But four characteristics are common to most or all of the competing ver-sions: (1) in a corporatist system, each major interest in society is representedby one or a very few central, or peak, associations-in the case of labor, con-federations; (2) these associations include as members most of the relevantgroups and individuals; in the case of labor, a high proportion of the laborforce is unionized, and most or all unions belong to the peak confederation(s);(3) the peak association(s) have considerable authority or control over theirconstituent organizations, and these in turn have considerable authority or in-fluence over their individual members; and (4) peak association officials partic-ipate actively in government decisions affecting their constituents' interests.The arrangements for participation vary from country to country but entaildiscussion and informal or formal bargaining among associations representingvarious interests and the state.

For labor, such arrangements mean that unions have sufficient economic andpolitical muscle to encourage their incorporation into national decisionmakingcircles. (They may have gained this strength historically, by gradual growth ofunions and federations, or the government may have directly aided the growthof unions as part of an elite strategy to preempt more radical influences.) Wagerestraint and acquiescence in adjustment are a response to generous socialsecurity and welfare arrangements that buffer workers against the vicissitudesof adjustment and exchange the immediate advantage of a pay increase for em-ployment security and a variety of longer-term benefits.

At the same time, in a strong corporatist system top labor leaders have theauthority within the unions to ensure that the rank and file will accept agree-ments made by the national elite. And discussion and bargaining with otherdecisionmakers lead labor leaders to recognize national interests that constrainpursuit of labor's narrow or short-term interests. (More radical interpretationswould argue that not only rational persuasion, but co-option-the desire tokeep the material benefits and status of associating with the national elite-is

40 The World Bank Research Observer, vol.6, no. 1 (January 1991)

a principal incentive to cooperate in such a case.) Recognition of broader na-tional interests encourages acquiescence in policies of wage restraint, to pro-mote national economic stability and growth and thereby sustainedemployment and gradually rising real wages. Negotiating wage agreements atthe national level also means that individual unions have less scope and needto worry about losses relative to other unions. 3

In several of the smaller Northern and Western European nations, labor'scooperation with broader economic policies reflects the clear recognition thatnations highly dependent on trade must maintain international competitiveness(Katzenstein 1985). To take one illustration, in Belgium in February to Marchof 1982, labor agreed to skip an automatic indexed increase in wages followingdevaluation and to replace later indexed rounds with lower lump-sum wageadjustments to reach an agreed degree of deindexation. Further real wage in-creases were banned through 1984 and thereafter regulated by a norm of com-petitiveness taking into account the weighted average of Belgium's seven mostimportant trade partners (see Theo Peters in Williamson 1985, p. 99).

The following section will argue that virtually no developing countryapproximates the labor organization and wage negotiation arrangements of thecorporatist model. But the principles that make the model work will prove use-ful in understanding problems of labor relations and labor market reform indeveloping nations, and they may suggest broad guidelines for improved laborrelations. In essence, corporatist institutions affect the range of informationand options available to all parties in the negotiations and thereby alter incen-tive structures. More specifically, they (1) broaden the array of options undernegotiation to include a wide range of social policies along with wage and non-wage compensation; (2) lengthen time horizons, largely by increasing confi-dence both in government economic management and in labor's continuedaccess to decisions affecting its interests; and (3) increase the sophistication andcomplexity of economic implications taken into consideration by union, busi-ness, and government representatives.

Corporatist labor market arrangements cannot be replicated in most devel-oping nations, in part because the structure of labor markets and the largereconomy bear no resemblance to the usually small industrial democracies inwhich these arrangements have evolved, and in part because the arrangementshave grown out of historical features quite unlike those of most developing na-tions. Moreover, most developing nations cannot afford the employment ben-efits and other social benefits that are key elements in the corporatist bargain.Indeed, in some developing countries the attempt to provide organizedworkers-a relatively small fraction of the labor force with generous benefitsmodeled on those of industrial nations has contributed to both economic dif-ficulties and social tensions.

For all these reasons the corporatist model is not a practical blueprint foremulation, though there may be a few developing countries that could evolvealong those lines. Rather, it is a heuristic model that elucidates some of the

Joan M. Nelson 41

mechanisms and relations that have helped some nations channel the self-interest of a sizable and well-organized interest group in ways compatible withbroader national interests. Where similar arrangements are unlikely to emerge,functional equivalents are needed-institutions, procedures, and policies thatbroaden the array of options, lengthen time horizons, and increase the sophis-tication of all key actors in labor markets.

Applicability of the Evidence to Developing Countries

Virtually no developing country replicates the features of the prototypicalstrong corporatist system with respect to labor. In many of the countries ofNorthern and Western Europe, unions claim as members two-thirds or moreof the work force. Reliable and comparable data on union membership aremissing for most developing countries, but those for which data are availablesuggest a span ranging downward from about a third of the labor force. In themid-1980s Argentine and Venezuelan unions claimed about a third of the laborforce, Jamaican unions a quarter, Mexican unions roughly a fifth.4 OutsideLatin America, save for Mauritius and Sri Lanka (both with roughly a third ofthe labor force unionized), union membership is mnuch lower. Some of the mostindustrialized (and rapidly industrializing) developing countries, above all theEast Asian "tigers," have in varying degrees repressed union movements.

Even the comparatively large union movements in some Latin Americancountries do not usually have the additional characteristics of strong central-ized movements. In Brazil and Mexico, for example, rivalries among confeder-ations are strong. The power of Mexico's largest confederation, theConfederation of Mexican Workers, is reduced, quite aside from the effects ofaging leadership, by its structure: average union size is small (about 130 mem-bers), and less than half of the confederation's members are in powerful indus-trial unions such as railroads, petroleum, or mining. Nearly half are inenterprise-level unions, which are in turn parts of functionally diverse, region-ally based federations. The Confederation of Mexican Workers must oftencompete with rival confederations active in the same geographic areas, and thegovernment can manipulate these rivalries to undermine the confederation, asthe government of Miguel de la Madrid did in 1S983. Therefore it is extremelydifficult to coordinate labor action within an industry. Mexican unions rarelyhave strong representation, such as works councils, at the enterprise level(Middlebrook 1989, pp. 209-17). Similar weaknesses plague labor movementsin many other countries. In short, few if any developing countries have labormovements able to play their side of the corporatist strategy in labor marketpolicies.

If data were adequate to develop indices of union strength and centralizationof wage bargaining for most developing nations, virtually all would fall in thelow and moderate segments of a global scale arraying labor movements fromweakest and least centralized to strongest and most centralized. (These dimen-


sions do not necessarily covary neatly.) Both classic liberal expectations andthe forecasts of recent proponents of a humpbacked relationship based on theexperience of industrial nations predict little militancy and wage push at thelow end of this scale and greater militancy and wage push toward the center.5

Evidence from developing countries accords with these expectations.Walton and Ragin (1988, table 10.3) found a strong correlation between the

proportion of the labor force in unions (as of 1975) and the severity of protestsagainst austerity between 1976 and 1987 in twenty-six countries in LatinAmerica, Africa, and Asia. In many developing nations, labor movements arenot only a small part of the labor force but are also weak and fragmented.Strikes (especially wildcat strikes) may be quite frequent but are likely toinvolve few workers and have little effect on broader wage levels. But unionsin some developing countries are strong enough to exercise considerable wagepush in particular sectors or industries. Unions in strategic sectors may exercisetremendous economic and political leverage, even if the labor movement for thenation as a whole is not strong. Copper miners in Chile and Zambia, tin minersin Bolivia, oil workers in Mexico, Nigeria, Venezuela, and other countries, and,to a lesser degree, bauxite and alumina workers in Jamaica have been in a po-sition to throttle a large fraction of national export earnings and governmentrevenue. They have often used their leverage to maintain wages markedly outof line with the rest of the economy, with a variety of distorting effects on theindustry and the economy as a whole. In Mexico, despite the pattern of incor-poration of Mexican labor confederations, especially the Confederation ofMexican Workers, their exclusion from economic decisions under the de laMadrid government, and their inability to block massive drops in real wages,certain Mexican unions were able to block government reforms. For instance,the powerful petroleum union used threats of sabotage and massive resistanceto force the government to back down on its plans to eliminate the union's con-trol over subcontracting as part of a rationalization scheme. In a few develop-ing countries with sizable industrial sectors, including several Latin Americannations, Turkey, and Tunisia, not merely unions in strategic industries but la-bor movements more generally have periodically been militant, wielding con-siderable economic and political clout. In the framework of theories based onindustrial democracies, these countries fall somewhere in the central rangeof the scale linking centralization to militancy and wage push: their militantbehavior conforms with the hypothesis of a humpbacked relation.

A further characteristic of unions in developing countries may encourage atendency toward militancy. Both the corporatist theories and Calmfors andDriffill's theory of union bargaining strategies implicitly assume unions haveconsiderable analytic capacity to assess the implications for labor of currenteconomic trends and alternative wage settlements. Few industry-level unions ornational confederations in developing countries have economic staffs compara-ble to those in Western European countries. Moreover, available data are oftenunreliable, and analysis and forecasting are particularly difficult where inflation

Joan M. Nelson 43

and other aspects of economic performance are highly volatile. The result maybe to increase uncertainty and risk associated with wage restraint and reducethe attractiveness of trading off wage increases for social welfare benefits.6

One other important aspect of the structure of labor movements in develop-ing countries is that the proportion of the labor force in public service jobs ismuch higher than in most industrial democracies (see table 1). In severalAfrican nations public sector workers made up 60 to 80 percent of nonagri-cultural workers; in India the figure was 72 percent (Heller and Tait 1984,table 22).

The high proportion of wage labor in the public sector means that the sec-tor's wage and employment policies are important both in their own right andalso because of spillover effects into the private sector. Heller and Tait (1984,p. 35) estimated that "central government decisions on wages and salaries indeveloping countries are likely to affect 15 to 40 percent of employed workersin the urban labor market and therefore have a pervasive 'leverage' effect ondomestic unit wage costs."

How do large public sectors affect wage push and employment flexibility?The topic is little studied. Two questions are central: Are public sector unionsintrinsically weaker (or stronger) than their private counterparts? And doesgovernment, in its role as employer, operate with different goals and constraintsthan do private employers? The answer to the first question is ambiguous. Theanswer to the second is clearly affirmative.

Public sector workers in industrial societies are often assumed to be in aweaker position to exert wage pressure. Historically, they have often been re-

Table 1. Public Sector Workers as a Percentage of Nonagricultural Employment

Type of public employment

Nonfinancial TotalCategory of Number of Central State public publiccountries countries' government and local enterprises employment

OECD 16 8.7 11.6 4.1 24.2

Developingcountriesb 35 23.4 4.0 13.9 43.9

Africa 16 30.8 2.1 18.7 54.4Asia 5 13.9 8.0 15.7 36.0Latin America 10 20.7 4.2 5.5 27.4

OECD, Organisation for Economic Co-operation and Development.Note: Data are from 1979 or 1980.a. The number of countries shown for each category is that for which data were available on central

government employment. Not all countries had data for each category of workers, and the number ofobservations therefore varies somewhat within each country category for different items.

b. This category incorporates data for a few countries in the Middle East.Source: Adapted from Heller and Tait 1984, table 1.

44 The World Bank Researcb Observer, vol. 6, no. 1 (January 1991)

stricted from forming unions. Even where employees' associations are legal andlarge, many categories of public sector workers are legally barred from strikingand risk dismissal and arrest if they do strike. Their position is weaker still ifopportunities for alternative employment in the private sector are extremelylimited, as is true in many of the poorest countries.

To balance these handicaps, do public sector workers and their associationshave leverage that private workers do not? Freeman (1986, p. 42), discussingstate and local government unions in the United States, argues, "A fundamentaldifference between public and private sector collective bargaining is that publicsector unions, more so than private sector unions, can influence ... employerbehavior through the political process."

But it is not clear why this should be so. Public sector workers may indeedform voting blocs or lobby or demonstrate for their demands. But so mayorganized private sector workers. Strikes, work stoppages, or slowdowns in vi-tal public services hold the public hostage, generating strong political pressureon the government to hasten a settlement. But the outcome may be repressionrather than concessions to the strikers, depending on the political circumstanc-es and particularly on the degree of public sympathy with the strikers.

A more convincing argument is based on patronage ties between politiciansand public sector workers. Patronage implies a reciprocal commitment: the ex-change of jobs for political support. Many national and local politicians havebuilt up their support bases primarily through this mechanism. Sharp declinesin real wages and, even more clearly, large-scale reductions in the work forcemean the disintegration of support coalitions. In Africa, moreover, patronageoften has tribal overtones and is endorsed, indeed mandated, by the strong tra-ditional expectation that a "big man" protects and advances his ethnic broth-ers. A severe wage squeeze or extensive firings betray that trust.

Comparisons of public and private sector wage levels in developing countriesoffer ambiguous evidence as to whether public sector workers (and their unionsor associations) are handicapped or advantaged. Any such comparisons mustbe approached with caution. Workers' compensation packages are complex, in-cluding base pay, varied bonuses, and nonwage compensation such as subsi-dized housing, insurance, and other benefits. And it is hard to control acrosssectors for differences in education and skill composition. There may alsobe important contrasts between the wages and security of civil servants andthose of workers in stalte enterprises or local government. Ehrenberg andSchwarz (1986), in their review of U.S. research on the issue, note considerableevidence that the relative wage gap between union and non-union workers issmaller in the public than in the private sector, but they go on to suggest severalconsiderations that cast doubt on the finding. Comparing public to private em-ployees (without considering the union versus non-union dimension), they notethat U.S. federal government workers are paid somewhat better than compara-ble private sector workers; this is particularly clear for women and for non-white males. The finding holds, but less strongly, for state and local employees,

Joan M. Nelson 45

except perhaps for male local government workers. Taking into account non-wage compensation (which most studies ignore) would probably strengthenthese findings.

For what they are worth, the "stylized facts" turn up geographical differenc-es. In some Latin American countries in the mid-1980s, public sector workersseem to have been somewhat less well paid than their private sector counter-parts (Lopez and Riveros 1989, Riveros and Sanchez 1990). In Sub-SaharanAfrica in the postindependence decades, public sector workers tended to besubstantially better paid, but during the 1980s real wages in the public sectorfell dramatically, perhaps erasing or even reversing their earlier advantage(Lindauer, Meesook, and Suebsaeng 1988). The regional contrasts are probablyattributable more to differences in the relative weight of public and private sec-tor unions than to variation in the militancy of public sector unions. Privatesector unions are large and important in a number of Latin American countries,presumably driving up private sector wage levels at least in highly unionizedsectors; in Africa employment and unions in the private sector are much lessimportant relative to their public counterparts.7

More broadly, it is surely the behavior of governments as employers morethan the behavior of unions that distinguishes wage determination in the publicsector from processes in the private sector, above all during hard times.Governments do not have to cope with profit and loss sheets but instead areconstrained by fiscal and monetary considerations. Recession and fiscal crisispit pressure from public sector workers to maintain wages and perquisitesagainst political pressures for other uses of public funds. Trends in personnelcosts as a percentage of total public expenditures could be viewed as a roughmeasure of the political priority accorded to public sector workers. In most ofLatin America and Africa in the 1980s, that percentage rose substantially. Un-der intense fiscal pressure, maintaining public employment has been given veryhigh priority; indeed, especially early in the crisis, governments often tried tohire countercyclically (Lopez and Riveros 1989, p. 20; Riveros and Sanchez1990, p. 35). As the need for long-term reform became inescapable, a few gov-ernments laid off large numbers of workers, but rnost such cases entailed gen-erous compensation. Ghana, for example, paid the equivalent of roughly twoyears' total compensation to workers stripped from the notoriously overstaffedCocoa Board. Guyana cut almost a quarter of its civil service and public en-terprise workers but awarded severance pay so large that the government hadto borrow heavily to cover the costs (Nunberg 1988, pp. 9-11). Buffering lower-paid workers has also been stressed: not only have average real wages in thepublic sector dropped sharply, but, especially in Africa, differences betweenthe highest- and lowest-paid workers have been cut dramatically. Union pres-sure obviously played a role in these patterns. But more fundamentally, the pat-terns reflected the different incentives and constraints motivating governmentsas employers, in contrast to those governing the actions of private employers.


The Economic Context and Labor Militancy

The most plausible and best-supported proposition about the effects of eco-nomic cycles on organized labor is that bargaining power and militancy declinein hard times and increase in prosperity. Strikes vary with economic cycles:when labor markets are slack and protracted unemployment looms, unionsoften consent to wage regulation; conversely, good times make labor more ag-gressive (see Lange 1984 and Kennan 1986). Labor militancy is likely to havehigher payoffs in tight labor markets, and widespread wage drift (induced byworkers or employers) and other forms of noncontracted gains are more prob-able. For instance, in Norway's postwar history of wage regulation, break-downs were most likely during periods of rapid economic growth. Conversely,Epstein's analysis (1988) of union responses to official austerity programs be-tween 1976 and 1984 in seven Latin American countries with large and impor-tant labor movements found fairly strong statistical links between risingunemployment and fewer and smaller strikes.

The rapidity with which workers' priorities shift, during economic crises,from maintaining real wages to protecting their jobs tends to reflect the historyof economic growth in the country in question. "The longer the time since thelast structural crisis, the more workers are likely to expect that strong economicconditions . .. will continLe, and that downturns will be only temporary"(Lange 1984, p. 117). In Costa Rica, for example, the economic tailspin of1979-82 broke a virtually uninterrupted span of growth from the early 1960s.After Alberto Monge's government's remarkably effective stabilization effortof 1982-83 had clearly taken hold, popular pressure to restore real incomes totheir precrisis peak mounted rapidly.

Lange's discussion of the rational bases of workers' consent to wage regula-tion adds one modification that is important for our concerns. Although pessi-mism about short-term economic prospects usually encourages cooperationwith wage regulation, the effect is less predictable when labor confronts "struc-turally bleak prospects."

On the one hand, it might be thought that workers would be stronglyinclined to regulate wages, for they would think that failing to do so wouldcondemn the national economy to sustained recession or even depression,and their own wages to real decay. On the other hand, the bleak economicprospects might lead workers to think that there is little to be gained by re-straint-things will not get better anyway without major disruptions to theireconomic lives-and that it is therefore best to try to get as much as possiblewhile it can still be had (Lange 1984, p. 115).

Lange argues further that, in such situations, a key variable shaping workers'behavior is whether they believe that "the fruits of their restraint will redoundto their future economic advantage." This perception, in turn, rests on thecredibility of the governrnent's broader recovery program and on implicit or

Joan M. Nelson 47

explicit assurances that their interests will be protected during industrialrestructuring. In short, when the long-term economic outlook is grim, politicalrelationships and expectations become much more important in determininglabor's responses.

Political Structure, Labor Militancy, and Labor Influence

Power is relational. The strength of governments, their attitudes and supportbases, and their strategies toward labor markets are as important in shapingoutcomes as the strength, autonomy, and orientation of labor movements them-selves. And unions adjust their actions to the political context. Several dimen-sions of the political context demand consideration: type of regime; the partysystem and the place of organized labor in that system; and more transitorypolitical circumstances, particularly electoral cycles and changes in regime.

These same factors may also affect governments' commitment to adjustmentin general and to labor market flexibility in particular. For this brief survey thediscussion is simplified by assuming government commitment to measures(1) to contain wage increases to a significant degree, as an important elementin a stabilization package, and (2) to encourage increased labor market flexi-bility in the medium term. In other words, government commitment to flexiblelabor markets is taken as a given in this article, and the emphasis is on howvarying political contexts shape government tactics and labor responses.

Regimes and Tactics

Most of the research reviewed earlier in this article is based on evidence fromindustrial democracies with governments widely accepted by their citizens aslegitimate, with competition among two or more parties that constitute poten-tial governments, and with broad legal and actual protection for civil andpolitical rights. The type of regime, in effect, is held constant. In developingcountries, types of regime vary widely.

One would expect the type of regime to be most obviously reflected in thetactics governments choose, once convinced of the necessity of measures orga-nized labor is likely to resist. To cope with that resistance, three basic types oftactics are available: persuasion, partial compensation, and containment. Socialpacts combine persuasion and compensation. Compensation need not take theform of immediate wage increases; social pacts often trade short-run wagerestraint for social benefits (in effect, deferred improvements in the security andquality of life). Another form of compensation is reform of labor laws to pro-vide labor with assurances of greater freedom to organize or with better accessto decisionmakers (in effect, improving prospects of later economic gains).Union pressure can be contained not only by repression (or the threat of repres-sion) but also by co-opting labor leaders or by encouraging internal divisionsand the emergence of rival unions or federations.


Broadly, democratic governments rely primarily on persuasion and compen-sation; authoritarian regimes are less reluctant to use containment in generaland repression in particular. So unions in authoritarian systems-if they arepermitted to operate legally at all-are likely to be more cautious and less mil-itant.

But type of regime is too crude a variable to be a reliable predictor of gov-ernments' choice of tactics in dealing with labor. Both authoritarian and dem-ocratic governments display a wide range of legal and institutionalarrangements governing union activity. Some democracies have laws hedgingunion activity quite tightly, and many democratic governments have been will-ing to ignore or put down labor protests. In Sri Lanka in 1980 an attemptedgeneral strike was broken by dismissing large numbers of public sector strikers.In Bolivia in 1985 the newly elected government of Victor Paz Estenssoro actedpromptly and decisively to squelch the miners' union that had repeatedly de-stroyed earlier stabilization efforts. In Jamaica in the same year Edward Seagafaced down an unprecedented general strike in which both major union con-federations (each linked with one of the two major parties) protested his aus-terity program.

Conversely, some long-established one-party or dominant-party governmentsgrant unions considerable influence, although this may be punctuated withepisodes of tension or repression. Examples include the long history of KennethKaunda's relations with the Zambian copper miners; the dominant Institution-al Revolutionary Party's long-term management of labor relations and policiesin Mexico; Neodestour's relationship with the major Tunisian labor federation;and Juan Velasco's military government's cultivation of support from Peruviantrade unions in the late 1960s and early 1970s.

Electoral cycles clearly do influence not only governments' willingness tolaunch or maintain unpopular wage and price measures but also organizedlabor's acquiescence in such measures. This pattern is not exclusive to compet-itive democracies. Thus Ferdinand Marcos postponed stabilization measuresuntil after the important Philippine legislative elections of May 1984. The dela Madrid government, after five years of ignoring Mexican labor unions' an-guished protests, entered into an Economic Solidarity Pact in December 1987,as it became increasingly clear that the elections scheduled for mid-1988 wouldpose an unprecedented challenge to the Institutional Revolutionary Party's con-trol (Middlebrook 1989, pp. 207-8).

In sum, the type of regime clearly affects government handling of labor re-lations and, therefore, the options available to unions. But the more fine-grained features of the relation between organized labor and the government,including the channels and degree of access to decisionmaking circles, countmore in how governments manage labor issues. These arrangements are notsystematically related to the broad distinction between electorally competitiveand noncompetitive regimes.

Joan M. Nelson 49

Party Systems and Labor's Political Role

Party systems and union ties with parties are particularly influential in shap-ing unions' political roles and their relations with governments. Labor move-ments occupy positions in party systems ranging from dominance in a rulingcoalition to virtual exclusion. The examples that follow demonstrate the rangeof possible positions.

* Peronist unions in Argentina dominate their party, and they have exercisedtremendous power during periods of Peronist control of the government.

* A pattern more commonly found is incorporation of unions or segmentsof labor as subordinate parts of a ruling coalition, as in Mexico andSingapore.

* In several Latin American and Caribbean countries, unions have links withboth of two centrist parties, which together dominate political processes:Colombia, Costa Rica, Jamaica, and Venezuela are examples. In these cas-es labor has access to (but does not dominate) the government regardlessof the party in power.

* In Peru until quite recently, and in Argentina, labor has been linked to onestrong party within a system divided by deep social and political cleavages.In Peru, the result was to exclude the American Popular RevolutionaryAlliance (APRA) and its labor allies for many decades. In Argentina,Peronists periodically gained power. In both cases the political system waspolarized and destabilized.

a In Uruguay and (less clearly) the Dominican Republic, unions have beenlinked mainly with weak radical parties: the result is semiexclusion.

• In countries such as the Republic of Korea until quite recently, and Chileunder Augusto Pinochet, organized labor has been effectively excludedfrom tightly limited political arenas.

At any given time, different party systems and links between parties andorganized labor strongly influence the options available to unions to defend orpromote their interests. Over time, patterns of incorporation into the party sys-tem also affect the strength and orientation of the labor movement as a whole.

Broadly, it appears that either incorporation into an established dominantparty or links with parties that alternate in power encourage negotiation andcompromise and dampen confrontation. But incentives for the party in powerto make concessions to labor differ sharply in the two categories. The domi-nant party that incorporates major unions may assume that labor has no plau-sible alternative to cooperation. In more competitive two-party systems accessto both major parties assures labor leaders that their interests will get a hear-ing-the kind of assurance Lange (1984) suggests is likely to moderate the mil-itancy of desperation in very hard times. In contrast to both systems, thepolarized pattern is likely to stimulate militancy, whether or not the party withwhich labor is affiliated stands a good chance of taking office.


Instability and Transitional Regimes

Just as research on the militancy of labor movements in high-income indus-trial democracies holds the type of regime constant, it also implicitly focuseson experience in stable political systems. Many developing countries are muchless stable and predictable: governments frequently change through coups rath-er than scheduled elections, and not merely the people in power but the natureof the political system is prone to change. Unpredictable political settings havefar-reaching implications for the politics of adjustment, including the politicsof labor market adjustment.

In the short run, both elected and "irregular" new governments may havespecial opportunities to adopt needed economic reforms. The new governmentis likely to have a honeymoon period during which it can blame the need forunpopular economic measures on the legacy from its incompetent or dishonestpredecessor. The honeymoon effect may be particularly strong both after a longperiod of political and economic decay and if the new government enters officewith strong popular support. (New governments, of course, vary tremendouslywith respect to both their sense of security and the clarity of their economicgoals and program: both factors affect their ability to take advantage of thehoneymoon period.) The effects of governmental turnover on the system asa whole over time are a different matter: frequent turnover is likely to breedcynicism and to drastically shorten the period of time that all political players,including labor unions, are willing to take into account as they consider theirstrategies and tactics. The result in the realm of labor policy is diminished po-tential for trading deferred benefits for current wage moderation.

The current global wave of democratizing reforms and changes in regimeshighlights the special question of interactions between democratization andeconomic liberalization, including labor market adjustments. A principal char-acteristic of democratization is rapid expansion of popular political participa-tion. Popular demands for improved living conditions may be temporarilymuted by widespread recognition that the country's economic problems requiremajor reforms and will take time to carry out. That recognition seems to beparticularly strong in sorne of the Eastern European countries now retreatingfrom state socialism. But popular demands are likely to be more immediate andinsistent in many of the nations of Latin America and Asia, and in some inAfrica now turning or returning to democracy.

In short, new democracies often face considerable wage pressure. If the newgovernment is viewed as a definitive break with the bad old days, workers arelikely to look for speedy improvement in their situation. But if the new regimeis viewed as fragile, labor may well seek to capitalize on the political openingwhile it lasts. From the government's perspective, too, the desire to consolidateor broaden popular support encourages wage increases and other populist mea-sures.

Ups and downs in real wages obviously reflect not only government wagepolicies but other domestic policies and external trends, but it is striking how

Joan M. Nelson 51

many of the new democracies of the past two decades illustrate the tendencyto early wage concessions. In Portugal after 1973 real manufacturing wages rose13.7 percent in 1974 and 10.3 percent in 1975; nominal wages were hiked muchmore (Schmitt 1981, appendix table 1; Stallings 1981, pp. 108-9). In Spain realwages had been rising rapidly from the 1960s but accelerated after 1974 (Lopez-Claros 1988, p. 4 and chart 6). In Argentina real industrial wages had dropped11 percent in 1980 and 10 percent in 1981; the departing military governmentthen raised real wages by at least 25 percent. (Diaz-Alejandro, in Williamson1985, gives the increase as 29 percent. Later data in ECLAC 1989, table 6, p. 17,record the increase as 25.4 percent.) Raul Alfonsin, campaigning for the presi-dency and probably basing his judgment on the 1980-81 data, promised to raisereal wages, and in 1984 real average manufacturing wages rose a further26.4 percent (ECLAC 1989, table 6). In Uruguay average real wages jumped14.1 percent in 1985 with the return of civilian rule, after having fallen morethan 30 percent between 1982 and 1984, although the new civilian governmentwas firmly centrist and the unions were not traditionally politically powerful.In Bolivia Hernan Siles Zuazo's election in 1982 ended a long period of militaryrule, but Siles's attempts to gain control over the rapidly disintegrating econo-my were repeatedly blocked by militant unions. At. least four times the govern-ment introduced austerity measures only to back off and raise nominal wagesafter general strikes.

In Turkey Turgut Ozal became head of the economic team when militaryrule was imposed in 1980 and was later elected prime minister when civilianrule was restored in late 1983. Wage negotiations had become increasingly con-frontational in the late 1970s as the economic and political crisis deepened, andthe new military government cracked down on the labor unions. Real wages,already falling steeply in 1978 and 1979, dropped a further 25 percent in 1980.As the stabilization and adjustment program took effect, the rate of declineslowed to 7.4 percent in 1981 and 4 percent in 1982. In 1983, as elections torestore civilian rule approached, wage policy was eased and prices on key sta-ples held in line; real wages increased 5.6 percent. The collective bargainingprocess was liberalized early in 1985 (this information is based on an internalWorld Bank report and on Kopits 1987, p. 10).

Labor pressure on new democratic governments is not surprising. Muchmore interesting would be a semisystematic explanation for the experience ofsuch governments in trying to bring wage policies back into line and to estab-lish rules of the game to reduce conflict with organized labor while permittingstabilization and adjustment. The 1980s saw a wide array of strategies, includ-ing successful pacts (Spain), a variety of tactics to weaken or split labor(as Alfonsin initially tried to do), and straight union-breaking (under PazEstenssoro in Bolivia). The results of such strategies in winning labor's coop-eration with adjustment have ranged from considerable success to total failure.Explaining the results would go well beyond the space available here. But at


the core of durable arrangements to encourage moderate labor demands mustbe a modicum of confidence.

Long-Haul Adjustment, Labor Cooperation, and Equity

Skillful government tactics can make a great difference to the effective orineffective management olf labor demands in the short run. But adjustment isa long-haul proposition, likely to extend over a decade in many countries, per-haps considerably longer in some. Resumed growth in those countries whoseeconomies have stagnated or worsened in the past decade will facilitate labor'scooperation, but certainly in much of Latin America and Africa it is realisticto expect a long process of painful reorientation.

The crucial element in gaining workers' cooperation is the belief that theirsacrifice will contribute to general gains and that those gains will be distributedfairly. That belief requires (1) confidence in the government's economic man-agement and at least contingent hope that the government has a plausible gameplan-in countries where external economic trends are very adverse, the plau-sibility of the government's game plan may hinge in part on the probability ofsubstantial external support; (2) confidence that labor's interests will be fairlyrepresented and that labor representatives will have access to and influence indecisionmaking circles; (3) as a prerequisite, reasonable confidence in politicalstability, that is, assurance that changes in administration are not likely tobring dramatic change in labor's position or in the rules of the game.

These requirements are strikingly parallel to the conditions needed to getbusiness to start investing again. Labor militancy-workers' main line of de-fense against declining incomes-can be seen as the counterpart to the businesscaution and capital flight that delayed supply response to partial reformsin many developing countries in the 1980s. Conversely, union acquiescence inwage restraint (to the extent that it is voluntary) is essentially an investment:current income and consumption are forgone in favor of expected greaterreturns later.

In some countries, cooperation might be encouraged by mechanisms to givelabor fuller access to decisionmaking circles, while ways are sought to broadenlabor leaders' grasp of larger national problems. The corporatist arrangementsof Northern and Western Europe are suggestive but cannot be replicated-certainly not rapidly, or imposed primarily from above. (Schmitter 1981,pp. 318 ff., emphasizes that the ruling elite has very limited ability to remoldsystems of party and interest intermediation, which are "the product of verylengthy and complex historical forces ... [and] are also subject to strong emer-gent organizational properties that guide their development and insulate themfrom ameliorative meddlings from above.") But what is widely perceived as along-term crisis simultaneously increases the importance of political influenceson labor relations and opens the possibility of dialogue.

Joan M. Nelson 53

Labor cooperation would also be encouraged if governments-and those ad-vising them-placed more emphasis on increased equity to balance insistencethat labor make sacrifices in the public interest. Recent analyses of the self-destructive "populist cycle" evident over the past forty years in several LatinAmerican nations trace the roots of the cycle to marked and growing incomeinequalities and the resulting political pressures for redistributive policies toraise the incomes of poorer groups. Sachs (1989) puts great emphasis on thispoint; it is less prominent but clear in Dornbusch and Edwards (1989, especiallyp. 5). More broadly, as the corporatist model suggests, both institutionalarrangements and policy outcomes must build up confidence that the benefitsof economic reform and recovery will be widely shared. The modest measureslikely to be feasible to increase equity in the 1990s-more progressive taxation(or simply the enforcement of taxes on the books) and reorientation of socialservices to better serve the working class and the poor-will not eliminate con-flict over which groups bear the costs of adjustmernt. But such modest measuresmight reduce anger and distrust and improve the chances for constructive dia-logue and bargaining.

Notes

Joan M. Nelson is a senior associate with the Overseas Development Council in Washington,D.C.

1. Wage push is inflationary pressure largely generated by a pattern of increases in real wagesoutstripping increases in productivity.

2. For an excellent brief discussion of democratic corporatism and its contrasts with othermeanings of the term corporatism, see Katzenstein 1985, pp. 30-37.

3. This description of corporatist arrangements for labor policies is, of course, stylized. Insti-tutional arrangements, policies, and economic outcomes vary among the nations usuallydescribed as fitting the model, especially the Scandinavian nations, Austria, Belgium,the Netherlands, and Switzerland. Within specific countries, the approach has been more effec-tive at some times than at others. For example, Sweden is often taken as the epitome of effectivecorporatist arrangements of this kind, yet recent trends in Sweden suggest the model is no longerworking very well.

4. The most extensive sets of data on union members as a proportion of the labor force arefound in Taylor and Jodice (1983), and in U.S. CIA (1989), but the former is now dated (referringto the mid-1970s) and some of the data from the latter are startlingly out of line with othersources. For example, U.S. CIA (1989) gives Mexican union membership as 35 percent of the workforce, compared with Middlebrook's (1989) figure for 1979 of 16.3 percent and George Grayson'sestimate (given at a conference on Mexican trade unions at the Overseas Development Councilin 1989) for the late 1980s of 20 percent. Recent data for many countries are also available inthe periodic U.S. embassy reports for individual countries (U.S. Department of Labor variousyears); this source is the main basis for the estimates in the text for Argentina, Jamaica, andVenezuela.

5. A literal interpretation of the corporatist approach would not predict this outcome. But thecorporatist line of explanation for unions and wage determination has concentrated on what wasgoing on in the industrial democracies with very high degrees of centralization, and it has paidlittle attention to the range of incentives and strategies guiding union behavior where wage set-ting is decentralized. Had the focus been more on the decentralized cases (and, even more likely,had it been on developing country cases), the assumed monotonic links among decentralization,


labor militancy, and wage push might have been modified. It seems self-evident that very smalland scattered unions are not likely to exercise effective wage push, and plausible that they willbe only sporadically and briefly militant.

6. This point was suggested by Kevin J. Middlebrook in correspondence with the author.

7. Heller and Tait's comparative study (1984) of public employment and pay is still the mostcomprehensive analysis available for developing nations, but it gives conflicting informationregarding public and parastatal wage levels compared to private nonagricultural wages. Directcomparisons of wage levels suggested that central government and nonfinancial public enterpriseemployees tended to be better paid than workers in private manufacturing (table 8, p. 18), butgovernment employment constituted a larger share of total nonagricultural employment than didgovernment wages as a share of total wages (p. 10).

ReferencesThe word "processed" describes works that are reproduced from typescript by mimeograph,

xerography, or similar means; such works may not be cataloged or commonly available throughlibraries, or may be subject to restricted circulation.

Calmfors, Lars, and John Driffill. 1988. "Bargaining Structure, Corporatism, and Macroeco-nomic Performance." Economic Policy 6 (April): 13-62.

Cameron, David R. 1984. "Social Democracy, Corporatism, Labour Quiescence and the Repre-sentation of Economic Interest in Advanced Capitalist Society." In John H. Goldthorpe, ed.,Order and Conflict in Contemporary Capitalism. Oxford: Oxford University Press.

Dornbusch, Rudiger, and Sebastian Edwards. 1989. "The Macroeconomics of Populism in LatinAmerica." Policy, Planning, and Research Working Paper 316. World Bank, Policy, Research,and External Affairs Office, Washington, D.C. Processed.

ECLAC (Economic Commission for Latin America and the Caribbean). 1989. PreliminaryOverview of the Latin American Economy, 1988. New York: United Nations.

Ehrenberg, Ronald G., and Joshua L. Schwarz. 1986. "Public Sector Labor Markets." In OrleyAshenfelter and Richard Layard, eds., Handbook of Labor Economics. Amsterdam and NewYork: North-Holland.

Epstein, Edward C. 1988. "Austerity and Trade Unions in Latin America." In William L. Canak,ed., Lost Promises: Debt, Austerity, and Development in Latin America. Boulder, Colo.: West-view.

Freeman, Richard B. 1986. 'Unionism Comes to the Public Sector." Journal of EconomicLiterature 24 (March): 41-85.

1. 988. "Labour Market Institutions and Economic Performance." Economic Policy6 (April): 63-80.

Heller, Peter S., and Alan A. Tait. 1984. Government Employment and Pay: Some InternationalComparisons. International Monetary Fund Occasional Paper 24. Washington, D.C.

Katzenstein, Peter J. 1985. Small States in World Markets: Industrial Policy in Europe. Ithaca,N.Y.: Cornell University Press.

Kennan, John. 1986. "The Economics of Strikes." In Orley Ashenfelter and Richard Layard, eds.,Handbook of Labor Economics. Amsterdam and New York: North-Holland

Kopits, George. 1987. Structural Reform, Stabilization, and Growth in Turkey. InternationalMonetary Fund Occasional Paper 52. Washington, D.C.

Lange, Peter. 1984. "Unions, Workers, and Wage Regulation: The Rational Bases of Consent."In John H. Goldthorpe, ed., Order and Conflict in Contemporary Capitalism. Oxford: OxfordUniversity Press.

Lindauer, David, Oey Astra Meesook, and Parita Suebsaeng. 1988. "Government Wage Policy inAfrica: Some Findings and Policy Issues." World Bank Research Observer 3, no. 1 (January):

--1-25.

Joan M. Nelson 55

Lopez, Ramon E., and Luis A. Riveros. 1989. "Macroeconomic Adjustment and the LaborMarket in Four Latin American Countries." Policy, Planning, and Research Working Paper335. World Bank, Policy, Research, and External Affairs Office, Washington, D.C. Processed.

Lopez-Claros, Augusto. 1988. The Search for Efficiency in the Adjustment Process: Spain in the1980s. International Monetary Fund Occasional Paper 57. Washington, D.C.

Middlebrook, Kevin J. 1989. "The Sounds of Silence: Organized Labour's Response to EconomicCrisis in Mexico." Journal of Latin American Studies 21, no. 2: 195-220.

Nunberg, Barbara. 1988. "Public Sector Pay and Employment Reform." Policy, Planning, andResearch Working Paper 113. World Bank, Policy, Research, and External Affairs Office,Washington, D.C. Processed.

Olson, Mancur. 1982. The Rise and Decline of Nations: Economic Growth, Stagflation, andSocial Rigidities. New Haven, Conn.: Yale University Press.

Riveros, Luis A., and Carlos E. Sanchez. 1990. "Argentina's Labor Markets in an Era of Adjust-ment." Policy, Research, and External Affairs Working Paper 386. World Bank, Policy,Research, and External Affairs Office, Washington, D.C. Processed.

Sachs, Jeffrey D. 1989. ."Social Conflict and Populist Policies in Latin America." NBER WorkingPaper 2897. Cambridge, Mass.: National Bureau of Economic Research. Processed.

Schmitt, Hans 0. 1981. Economic Stabilization and Growth in Portugal. International MonetaryFund Occasional Paper 2. Washington, D.C.

Schmitter, Philippe C. 1981. "Interest Intermediation and Regime Governability in ContemporaryWestern Europe and North America." In Suzanne Berger, ed., Organizing Interests in WesternEurope: Pluralism, Corporatism, and the Transformation of Politics. Cambridge: CambridgeUniversity Press.

Stallings, Barbara. 1981. "Portugal and the IMF: The Political Economy of Stabilization." In JorgeBraga de Macedo and Simon Serfaty, eds., Portugal in the 1970s: Political and EconomicPerspectives. Boulder, Colo.: Westview.

Taylor, Charles Lewis, and David Jodice. 1983. World Handbook of Political and Social Indica-tors. 3d ed. New Haven, Conn.: Yale University Press.

U.S. CIA (Central Intelligence Agency). 1989. World Factbook 1989. Washington, D.C.: GPO.

U.S. Department of Labor, Bureau of International Labor Affairs. Various years. Foreign LaborTrends. Washington, D.C.: GPO.

Walton, John, and Charles Ragin. 1988. "Austerity and Dissent: Social Bases of Popular Strugglein Latin America." In William L. Canak, ed., Lost Promises: Debt, Austerity, and Develop-ment in Latin America. Boulder, Colo.: Westview.

Williamson, John, ed. 1985. Inflation and Indexation: Argentina, Brazil, and Israel. Washington,D.C.: Institute for International Economics.

56 The World Bank Research Observer, vol. 6, no.1 (January 1991)

TECHNOLOGICAL CHANGEAND COMMERCIALIZATIONIN AGRICULTUREThe Effect on the Poor

Hans P. BinswangerJoachim von Braun

Do the economic gains brought by technological innovation and commercial-ization in agriculture work their way through to the poor? The prevailing op-timistic view is that they do. But this view is not universal: some hold thatthese forces for change can interact with, or even induce, institutional and mar-ket failure, with adverse consequences for the poor.

Adherents of the pessimistic view point to real-world instances in which thepoor have failed to reap the benefits, or even have lost, from the technologicalchange or commercialization. Where these effects have occurred we find thatthey are mostly attributable to inelastic demand or adverse institutional fea-tures; often, when technology or commercialization has been blamed for thedecline in income of the poor, other-not necessarily connected-policies havein fact been responsible for the damage.

This article contends that the optimistic view is, by and large, correct: nor-mally, technology and commercialization stimulate agricultural growth, im-prove employment opportunities, and expand food supply-all central to thealleviation of poverty. The evidence does not offer much encouragement to anextension of this view-that through "social engineering" the benefits fromtechnology and commercialization can easily be targeted toward the poor; thelimited opportunities for such targeting should of course be seized.

T here are conflicting views on how technological change and commer-cialization can affect the poor. (By the poor we mean the absolute poor:households and individuals that cannot earn enough to meet their basic

The World Bank Research Observer, vol. 6, no. 1 (January 1991), pp. 57-80© 1991 The International Bank for Reconstruction and Development/THE WORLD BANK 57

needs.) As Mellor points out, "One of the most important theoretical and em-pirical findings in analysis of Western economic growth is the identification oftechnological change as a major form of growth" (1986, p. 76). Those who seetechnological innovation as a principal source of growth maintain that there isa good chance (though admittedly no guarantee) that this general growth willbring corresponding growth in the income of the poor. In the same way com-mercialization, as part of the expansion of domestic or international trade, orboth, is seen to raise income levels in general via specialization based on com-parative advantage and economies of scale. Again, there is no guarantee thatcommercialization will enhance the income of the poor, but the presumptionis that a general increase in incomes will usually improve the incomes of thepoor, and thereby their food consumption.

It is often assumed that the poor in low-income countries are subsistenceoriented, detached from crop technology, and ]ittle integrated into markettransactions. This notion is incorrect. The poor are usually well integrated inthe rural labor market: whether hired workers or small farmers, they partici-pate in the exchange economy and, despite the high share of income allocatedto food, their cropping patterns and crop-livestock mixes show large involve-ment in markets. This fact is important for the spreading of effects of commer-cialization and technology in the economy.

The view that technology and commercialization play a major role in stim-ulating agricultural growth and alleviating poverty is now widely accepted, butthere is also a tenacious tradition of pessimism about technology and commer-cialization, whose adherents claim that both of these movements may bring ad-verse consequences for the poorest. (For a comprehensive review of thisliterature related to technological change in cereals, see Lipton and Longhurst1989; and for a critique of the literature on adverse effects of commercializationof subsistence agriculture, see von Braun and Kennedy 1986.) Proponents ofthis view adduce instances in which the poor were unable to participatesuccessfully in the adoption of new technologies during the green revolution inAsia, or in which poor farm workers were displaced by machinery. They arealso concerned about the possible effect of commercialization on the food con-sumption of the poor who produce for markets; for example, they maintainthat milk marketing schemes in India induced milk-producing households tosell their milk instead of feeding it to the household's children. They fear thathouseholds that produce cash crops will not have access to purchased foods iftheir cash crops fail, or if cash crop prices collapse.

A third point of view extends the optimistic position. It holds that technol-ogy and commercialization can and should be "engineered" in such a way thatthe poor can participate in the growth: technology must be directed towardcrops produced or consumed by the poor and must be easy for the poor toadopt; rural development projects must provide marketing assistance and creditto smallholders and poor farmers so that they can more easily participate incommercialization.

58 TheWorldBankResearch Observer,vol. 6,no. 1 (January 1991)

In this article we look first at the evidence of benefits of technological inno-vation and commercialization in the context of an open economy (that is, econ-omies open to trade and capital flow). The assumption of an open economymeans that technical change in a small country will not depress the price of thecommodity. We then explore scenarios under which the poor might lose abso-lutely or relatively from the changes, particularly through price effects (whichwould arise in a closed or partially closed economy) and other second-roundeffects. Finally, we examine the opportunities for deliberately targeting techno-logical and commercial change toward the alleviation of poverty.

Positive Effects of Technological Change andCommercialization

This section discusses conditions under which technical change and commer-cialization, or the combination of the two, have positive effects for the poor,and provides examples of where these positive effects have materialized.

Technology in the Lead

As land becomes less and less available, growth in agriculture depends moreand more on yield-increasing technological change. Increased output per hect-are contributed 70 percent of the production increase in major food crops ofdeveloping countries in the 1960s, 80 percent in the 1970s, and still more in the1980s (Paulino 1986). When a new technology, such as a green revolutionvariety, is introduced into a region, higher farm profits initially accrue to allproducers who adopt it, including poor farmers. If demand is elastic, a supplyresponse to the higher profits will usually lead to sufficient expansion of pro-duction so that demand for agricultural labor increases. Demand for pur-chased inputs and marketing and transport services will also lead indirectly toexpansion of employment. Consumer spending out of the higher profits willfuel demand for rural home goods and also expand the demand for labor.Rural wages will rise ancd workers may migrate from poorer areas to take ad-vantage of expanded opportunities. The green revolution, for example, in thePunjab of India (Bhalla 1983), in the MUDA irrigation scheme in Malaysia(Bell, Hazell, and Slade 1982), and in the Laguna Province of the Philippines(Herdt and Ranade 1976) expanded farm output, nonfarm output, employ-ment, wages, and immigration. Many of the effects of new technology in agri-culture on employment are indirect. The direct effect-increased use of laborper hectare in crop production-appears to have diminished in the secondphase of the green revolution after large creation of employment in the firstphase. Kikuchi, Huysman, and Res (1983) show this pattern for rice villages inthe Philippines.


Commercialization in the Lead

Commercialization, in the same way as technology, can be the prime impetusfor poverty-alleviating growth. Many regions in the developing world that pro-duce commercial crops for domestic or export markets are better off thanregions that are under subsistence production. The poor in the commercial re-gions are frequently better paid and have more secure jobs. Opposite findingsare, however, reported: interregional comparisons of sugarcane-growing areaswith other rural areas in Kenya, for instance, imply that cash cropping mayhave an adverse effect on nutrition (Hitchings 1982). Then again, Kennedy andCogill (1987), in their comprehensive analysis o:f the issues, find no adverseeffects of sugarcane growing on nutrition.

That the poor are often better-off in commercial regions is not sufficient toestablish a causal link between commercialization and poverty, either in im-proving or worsening the lot of the poor. A superior agroclimatic endowment,rather than commercial production, could account for the greater wealth of thecommercialized regions. The cross-regional comparisons, therefore, are incon-clusive; examples of positive effects observed over time afford more convincingevidence of the causal link between commercialization and improvement inthe income or nutrition of the poor. In Guatemala, the opening up of newexport marketing channels for vegetables boosted production of high-valuedlabor-intensive crops. Favorable agroecological conditions, basic infrastructure(roads), cooperative arrangements, and farmers' know-how in traditional veg-etable production helped create a success story. Small farmers (with an averagefarm size of 0.7 hectare) realized large gains in income from specialization, andemployment in agriculture increased by 45 percent (von Braun, Hotchkiss, andImmink 1989). The small farmers that joined the commercialization processalso adopted yield-increasing technology on their now reduced maize and beanfields. In this environment of risky markets for outputs and factors, and absentinsurance markets, farmers insure against risks to food security by maintainingsome self-sufficiency in food production. Thus they may fail to capture themaximum short-term gains from specialization.

Another example of commercialization, rather than technological change,taking the lead may be seen in the rapid expansion of cassava production inNortheast Thailand as a result of policies that opened up opportunities fortrade with the European Community in the 1970s and 1980s (Konjing 1989).

Commercialization in Tandem with Technological Change

Low-income countries that shift their crop mix toward marketed and (inter-nationally) traded crops also showed accelerated growth of yields per unit ofland in their staple food crops (von Braun and Kennedy 1986). True, differentcrops compete for scarce resources, but cash crops can also stimulate the pro-

60 The World Bank Research Observer, vol. 6,no. 1 (January 1991)

Table 1. Effects of Increased Income from Technological Change andCommercialization on Food Consumption and Nutrition(percentage increase)

Location Effect of 10 percent increase in income'of survey On calorie consumption On weight-for-ageareas Affected crop of households of childrenb

The Gambia Irrigated rice 4.9 1.9Rwanda Potatoes 4.7 2.5Guatemala Vegetables for export 3.5 1.1

a. Total expenditure, including value of home-produced foods approximating permanent income, is usedin the respective models. All changes are computed at a level of annual income of US$100 per capita.

b. Z-score values of weight-for-age.Source: von Braun 1989.

duction of staple food if channels for the supply of inputs and the marketingof outputs are opened up and if rural financial institutions are improved in con-nection with cash crops. Such complementarities between commercializationand technological change in staple food were strong in Sub-Saharan Africa(Lele, van de Walle, and Gbetibouo 1989). At the same time, hindering thecommercialization process impedes technological advances in production offood crops.

The complementarities between yield-increasing technological change in sta-ple foods and commercialization of agriculture can be exploited to help allevi-ate poverty. Specialization in labor-intensive crops, by stimulating the adoptionof new technology in staple foods, can create employment on the one hand andreduce food prices on the other. Labor-intensive cash crops can also provideemployment for the landless. Household food security is maintained-beyondthe income effects of cash crops-through improved use of technology in staplefoods by farm households that have moved into cash crops. This in turn candampen potential rises in local food prices. The case of vegetables for exportin Guatemala, referred to earlier, revealed precisely this pattern of complemen-tarity (von Braun, Hotchkiss, and Immink 1989). Similarly, in a case of sugar-cane expansion in the Philippines, sugar farmers had higher maize yields onthe reduced area for maize crops and maintained significant amounts of maizeproduction for subsistence. They consumed much less maize, however, than didspecialized maize farmers, because sugarcane-growing households preferred tobuy more rice in the market (Bouis and Haddad 1990). In a Kenyan case ofsugarcane production, farm households expanded their maize area into fallowland at constant yield levels to maintain subsistence (Kennedy and Cogill 1987).

Gains in real income from technical change or commercialization translateinto food consumption of the poor and nutritional welfare of children. In threecases studied in the Garnbia, Guatemala, and Rwanda, a 10 percent increasein income from a level of US$100 per capita translated into a 3.5 to 4.9 percent

Hans P. Binswanger and Joachimn von Braun 61

increase in households' food energy consumption and a 1.1 to 2.5 percentincrease in anthropometric measures of nutritional status of children (table 1).A large proportion of the households below the US$100 cutoff point fall belowaccepted calories per adult equivalent.

Scenarios under Which the Poor Might Lose

In the following, we test against empirical research seven scenarios of tech-nological change and commercialization under which the poor might lose ab-solutely or relatively. The scenarios range from inherent consequences oftechnological change and commercialization that policymakers have to beaware of (such as the agricultural treadmill effect) to failure of policy (forexample, coerced production).

Scenario 1: Declining Agricultural Prices

When final demand is not infinitely elastic-as was assumed in the openeconomy scenarios discussed so far-the expansion of production made possi-ble by technical change will lead to price declines. Such declines reduce thegains from technical change of producers and have secondary effects on con-sumers and on other regions or countries.

CONSUMERS VERSUS PRODUCERS: THE AGRICULTURAL TREADMILL. How thegains from technology and commercialization are distributed between agricul-tural producers and consumers has been a matter of considerable debate. Thebasic conclusion of partial equilibrium models is that under perfectly elasticdemand, producers can capture all the gains from technology, but under inelas-tic demand consumers gain and producers may either gain or lose. In otherwords, since price elasticities for many agricultural commodities are low, farm-ers work in a treadmill in which the fruits of improved technology forever eludethem and are instead enjoyed by consumers.

The treadmill hypothesis is persuasive, but the treadmill effect on rural in-come distribution is not as stark as the partial equilibrium models predict.First, producers can cushion the impact of declining prices on farm profits bymoving into other crops-that is, by substituting or diversifying. Second,declining prices of food have a positive impact on the income distribution inthe nonfarm sector: poor people spend much of their budget on food, so thatwhen food prices fall their real income rises proportionally more than that ofthe rich. Third, even in rural areas the poor may benefit from the price declineas consumers, while they lose farm profits in their role as producers. Finally,poor workers and poor farmers are also affected by the impact of technologyon demand for labor and wages. To measure the effect of technical change onincomes, therefore, more complex methods of assessment are needed, methods


that take into account the substitution response, the effect on consumers, andthe effect on employment and wages.

Quiz6n and Binswanger (1986a) used a general equilibrium model that in-cludes these three effects to assess the impact of the green revolution on ruraland urban incomes in Indiia. They computed a reference path of the real in-comes of rural and urban income groups in India from 1960 to 1981 (table 2).

For the period as a whole, the figures show agricultural productiongrowing-rapidly at the beginning of the green revolution (1965-70), stagnat-ing from 1970 to 1973, and resuming its rise from 1973. Agricultural terms oftrade rose before the green revolution, stayed fairly constant until 1973, andthen dropped substantially by the end of the period. Employment in agriculturegrew by about 20 percent over the period; wages declined by about 5 percent.The total real wage bill for the period therefore rose by about 15 percent.

1960 to 1970. Seriously depressed in the first half of the 1960s, farm profitsmoved dramatically upward in the early stages of the green revolution (1965/66 to 1970/71) as a result of technical change and improved agricultural termsof trade. Rural income rose along with profits: the rapid gains in production,

Table 2. Simulated Indexes of Income Distribution and Income Sources in India,for Selected Crop Years, 1960-81(1970/71 = 100)

Endogenous variable 1960/61 1965/66 1970/71 1973/74 1975/76 1980/81

Total actual agriculturaloutput 79.3 81.2 100.0 99.4 107.1 119.6

Actual prices (agricultural .nonagricultural goods) 89.8 97.2 100.0 97.7 91.6 76.3

Real residual farm profits 64.2 67.9 100.0 86.0 85.1 76.4

Agricultural employment 98.2 100.1 100.0 112.3 118.8 118.5

Real agricultural wage bill 91.2 95.3 100.0 101.4 104.9 105.4

Rural incomeAggregate 92.9 92.4 100.0 93.6 92.9 94.9First quartile (poorest) 101.0 99.0 100.0 95.9 97.4 107.0Second quartile 96.9 95.8 100.0 94.6 94.8 99.9Third quartile 93.8 93.5 100.0 93.8 93.3 96.3Fourth quartile (richest) 88.5 88.6 100.0 92.4 90.7 88.8

Urban incomeAggregate 89.4 102.3 100.0 99.4 102.2 136.7First quartile (poorest) 91.9 100.4 100.0 98.1 100.7 136.0Second quartile 90.9 102.8 100.0 99.3 102.6 141.9Third quartile 90.2 102.7 100.0 99.7 102.5 139.3Fourth quartile (richest) 87.6 102.3 100.0 99.8 102.2 133.5

Source: Quiz6n and Binswanger 1986a.


used by the government largely to replace imports, were associated with risingprices and thus translated into advances in income. The distribution of ruralincome shifted drastically from the wage bill to profits toward the end of the1960s.

1970 to 1980. The subsequent decade saw a marked decline of profits: to86.0 percent of their 1970/71 level by 1973/74, and to 76.4 percent by 1980/81.Once self-sufficiency in food production was assured, increased output was nolonger needed to replace imports; the surplus grain production had to be ab-sorbed domestically. The gains in production during the late 1970s, therefore,were not reflected in further advances in rural incomes, because the prices ofagricultural products fell. Instead, the productivity gains were transmitted toconsumers by way of declining prices.

During the 1970s urban groups were the princilpal beneficiaries of the com-bination of rapid growth outside the agricultural sector with the declining ag-ricultural terms of trade. Urban incomes in the aggregate rose by about34 percent during the late 1970s. But the rural poor also gained: within the ru-ral sector the incomes of the poorest quartile rose by about 10 percent duringthe late 1970s (in contrast to those of the richest quartile, which declinedslightly-about 2 percent).

Several effects interacted to achieve this outcome. First, the gains of the ruralpoor as consumers outweighed the decline in their incomes as producers orwage earners. Second, the effect of falling wages was partially offset by risingagricultural employment and by some participation in the growth of farm prof-its. Third, the incomes of the rural poor from the rural nonfarm sector alsorose substantially.

In sum, the treadmill was clearly in operation. Once prices started to fall,many of the gains from the green revolution were transmitted to the urban ar-eas. The mechanisms of adjustment to the price declines were not sufficientlypowerful to prevent this outcome. However, none of the rural groups lost inabsolute terms, and the poorest groups experienced an income increase of6 percent over the period. The large farmers (the richest quartile of the ruralincome distribution) gained least-they found their per capita incomes back atabout the same level as in the early 1960s, having lost the production and in-come gains of the late 1960s and early 1970s to population growth and urbangroups. Despite the shift in income distribution from wages to profits duringthe late 1960s, by the end of the period rural income distribution was very sim-ilar to its starting point.

REGIONAL DISTRIBUTION. Can technical change hurt if it is confined to certainregions? Often, environment or geographic location dictate whether a regioncan adopt technical change-for example, the green revolution has largely beenconfined to irrigated zones with good water control.

Partial equilibrium analysis of the distributional consequences of unequalregional access usually describes two regions supplying an inelastic demand in


a national market. Should region A increase supply as a result of technicalchange, prices will drop. Region B will lose because its production is no greater,input costs remain the same, but the selling price has fallen.

The gains in region A vvill be distributed among landowners and workers.In region B, farmers will lose, and labor demand and wages will fall. The larg-est share of the losses will be borne by the factors in most inelastic supply: theimmobile factors of production. Land prices will decline more than wage ratesif some labor can migrate to the gaining region (and contain the rise in thewage rate there).

This partial equilibrium model, however, again ignores consumer gains andcannot quantitatively assess the effect of labor mobility. Moreover it ignoresthe ability of farmers in region B to take up production of crops which are dis-placed in region A by the technologically dynamic crops. Quiz6n andBinswanger (1986b) extended the general equilibrium model for India discussedabove to include regional effects. The model allows for migration to respondto regional wage differential. Four regions supplying the same national marketwere analyzed. Two regions benefited massively from the green revolution,while the other two were less able to adapt to technology, owing either to poorwater control or to lack of rain. From simulations of the increases in rice yieldsarising from the green revolution in the technology-adopting regions, Quiz6nand Binswanger estimated that large farmers in the regions that did not adopttechnology lost as the price of rice fell, and their farm costs and yieldsremained the same. But in the same regions the poorest of the two quartiles ofthe rural income distribution gained at least as much from lower food pricesas they lost in farm profits and ended up with a gain overall, or, at worst, nonet loss.

When the simulation was extended to include wheat as well as rice yields,the figures pointed to urban groups and net purchasers of food in rural areasas the beneficiaries. Farm profits rose in the two technology-adopting regionsand fell in the other two--but, again, the poor in the latter regions gained fromthe lower food prices.

Thus scenarios can be modeled in which the poor can lose from technolog-ical change. But the inbuilt advantages of falling food prices often outweighthe disadvantages of falling farm profits. Treadmill effects can also be amelio-rated by diversification to and commercialization of other crops. And the gov-ernment can limit price declines by eliminating or reducing export restrictionswhere these still exist.

THE INTERNATIONAL TRE.ADMILL. Individual regions and countries can ame-liorate or escape the treadmill effect by changing policies that constrain exportsor by investing in infrastructure to reduce the cost of exporting, but theconstraints of worldwide demand cannot be avoided. Prices of agriculturalcommodities are well known to have experienced a declining secular trend


associated with expanded production capacity, 'whose main source has beentechnical change.

The obvious way for a country to avoid losses imposed by more rapid costreduction or technical change in other countries is to accelerate technicalchange beyond the rate of cost reduction elsewhere. Otherwise, loss of costcompetitiveness leads to increased pressure on the balance of payments andfaster depreciation (or slower appreciation) of the exchange rate. The loss ofincome associated with loss of competitiveness thus not only will fall on theproducers of the commodity but will be widely shared among consumersof tradable commodities.

If worldwide technical change occurs rapidly in commodities consumedwidely by the poor in the developing world, such as rice, wheat, and maize,the ensuing decline in real prices will be to the advantage of the poor in low-income countries. But this would not happen if, as often, the country reacts toa loss of competitive advantage by increasing the protection of domestic staples(Krueger, Schiff, and Valdes 1988). (It may also not happen if the country is amajor producer of the commodity so that an appreciation of the real exchangerate compensates for the decline in international prices.)

The poor derive little, if any, of the consumer benefit from technical changein commodities (such as tropical beverages) for which final demand is inelastic,and which are consumed mainly by the developed world. Producers in devel-oping countries impose losses on each other by expanding production. Interna-tional commodity agreements, such as the international coffee agreement, havegenerally failed to control output.

When competitive advantage in a crop is reduced, intercrop substitution ordiversification is once more an avenue for avoiding some of the losses. The ex-tent to which this option is available depends on agroclimatic conditions andon infrastructure for production, marketing, and trade in the alternative com-modities.

Scenario 2: Commercialization of Nonfoods Driving UpLocal Food Prices

Regionally concentrated adoption of cash crops may raise food prices inboth adopting and nonadopting regions of a country. When production of sta-ple food for local consumption is displaced by nonfood cash crops, net foodexports of the region may decline, or a net exporting region may even becomea net importer. Resulting differences in the cost, insurance, and freight (CIF)prices as opposed to free on board (FOB) prices for the region will be larger themore deficient the transport infrastructure is. If income gains resulting fromthe switch are not locally accrued by the poor, the poor may lose to the extentthat they are net purchasers of food. Most of the poor even in rural Africa arenet purchasers of food. The effects on prices for nonadopting regions depend


on market infrastructure, and possibly on the response of trade policy. Withfree trade, imports of food will substitute for domestic production and foodprices will not rise beyoncd levels determined by the CIF price. But if trade ofan export crop is controlled and foreign exchange is channeled through a sys-tem of government controls, allocation of foreign exchange for food importsmay be constrained, food prices would rise, and the poor would suffer.

In a case study of Benin, von Oppen (1989) shows that expanded cotton pro-duction in the north of the country reduces food crop exports from the northto the south. It then depends on effective demand of households and appropri-ate government response to translate the increased export earnings from cottonpartly into food imports to forestall reduced availability of food at householdand country levels. No case studies exist that evaluate how policy has actuallyresponded in such situations. Although potentially relevant, this scenario,therefore, remains hypothetical.

Scenario 3: Late Adoption of Technology

The speed with which tenants and small farmers-a large proportion of thepoor in many countries--adopt new technologies has been intensely studiedsince the green revolution. An early review of the literature on the adoption ofhigh-yielding seed varieties (Ruttan and Binswanger 1978) suggested that nei-ther farm size nor farm tenure has been a serious constraint on adoption.Although different rates of adoption by farm size and tenure have beenobserved, the available data implied that, within a few years of introduction,the lags in adoption due to size or tenure usually disappeared. Of course thenonadopters will have forgone the potential gains of early adoption and mayalready have suffered as a consequence. These conclusions have not been al-tered by more recent research.

Unlike that of seeds, the cost of fertilizers, herbicides, and other yield-raisinginputs can significantly impede adoption by small farmers. Typically, smallfarmers use fewer of these inputs per unit of labor than do large farms, but notnecessarily fewer per unit of land, depending on the steepness of the negativerelation between farm size and productivity.

The adoption is not just an issue of factor ratios, however, but an issue ofthe overall efficiency of use and the relative speed of growth in production.A survey by Berry and Cline (1979) shows that the use of inputs by small farm-ers is as efficient as by large farmers or more so. Econometric evidence fromthe Indian Punjab (Sidhu 1972) indicates that new wheat technology was notstrongly biased in either a labor-saving or a capital-saving direction. Small andlarge farmers achieved approximately equal gains in efficiency. Data from thePakistan Punjab and the Philippines indicate that although small farmers facemore constraints on obtaining irrigation and credit than large farmers, theseconstraints are not large enough to cause any significant differences in yieldsbetween the two categories of size (Ruttan and Binswanger 1978, p. 388).


The pattern of late adoption of cash crops and new market channels is sim-ilar to that for new production technology and staple foods. Of the smallhold-ers in Guatemala referred to earlier (von Braun, Hotchkiss, and Immink 1989)who started producing vegetables for export, disproportionately more of thelarger farms (larger than 1.5 hectares) were early adopters; the smallest trailedbehind. Similarly, smallholders in Kenya and the Philippines were significantlyslower to adopt sugarcane (Kennedy and Cogill 1987, Bouis and Haddad 1990).

Trailing behind in the adoption process is not necessarily a problem forefficiently allocating resources and ensuring the income security of the poor.For one thing, it avoids the risks of early adoption. But waiting too long canmean being shut out of the opportunity when market channels have bottlenecksrelated to organization or capacity. Capacity of an established sugarmill, forinstance, may be filled up rapidly by the (bigger) early adopters. The exportvegetable cooperative in Guatemala has effectively closed enrollment of newmembers because of concerns about bottlenecks in handling and capacity forcold storage.

The instruments usually advocated for accelerating adoption by small farm-ers are extension and credit. Reform of extension and credit policies to removediscrimination against the poor would go a long way toward alleviating theadverse effects of economies of scale in the use of technology.

Scenario 4: The Trap of Committed Expenses

Committing capital to perennial crops or other long-term investment suchas milch cattle, crop-specific irrigation, or housing reduces the capacity to ad-just to technological breakdown, price risk, or disruption of markets. Whenreturns to capital do not materialize, the fixed capital resources (invested, forinstance, in tea bushes or coffee) cannot be switched to new productive tasks.These risks are highest for the poor, but the empirical evidence shows that thepoor rarely specialize completely. Smallholders in Guatemala, for instance,maintained about half of their crop land for staples, as did smallholder sugar-cane-producing households in Kenya.

The same sort of risks as those of committed expenses arise when a projectfor technological change and commercialization attracts households to migrateto a new area and the project then collapses. In the Gambia the collapse of arice irrigation project led to disinvestment in housing and community servicesand even to increased divorce rates (Webb 1989).

Scenario 5: Implications for Women and Children

New production technology and new marketing opportunities in agriculturecan have profound implications for control of resources and division of laborin rural households. Evaluating the effect of technology and commercializationon poverty from the perspective of households in the aggregate does not cap-


ture the intrahousehold and gender-related effects. Both the burdens and thebenefits of technological change and commercialization need to be assessed atthe household level to judge the effect of gender, and both gender and childwelfare are relevant in this respect (Leslie and Paolisso 1989). The effect onpoor women will be more complex the more dramatic the change, the sharperthe division of labor and labor markets according to gender, and the greaterthe separation of control over farm resources.

A comparative analysis of commercialization in five cases (in the Gambia,Guatemala, Kenya, the Philippines, and Rwanda) showed a much reduced roleof women in the new technologies or commercialized crops, even if they wereimportant contributors to farm production before the change (von Braun,Kennedy, and Bouis 1989). Two examples (from the Gambian and Guatemalananalyses) highlight the conflicts and tradeoffs. These argue, in opposition tothe new conventional wisdom, that there is considerable income pooling withinthe household, so that women gain, though less than proportionally, from theincreased income of men.

In the Gambia, where rice was a traditional women's crop, the study foundthat women's access to new technology for rice irrigation was hampered be-cause it was harder for them to hire the necessary labor. Their work burdenincreased more than that of men. At the same time, technological changeresulted in increased household income, despite women's relative-in some sub-groups, absolute-loss of personal income (von Braun, Puetz, and Webb 1989).The increased household income in turn brought increased caloric consumptionand reduced seasonal fluctuations in weight for women.

In the Guatemalan export vegetable cooperative, households' food consump-tion-including women's--improved despite largely male control of incremen-tal income. The effects on] child welfare illustrate the complexity of some ofthe tradeoffs: on the one hand, the expanded employment resulting from thechanges increased seasonal use of child labor to such an extent that local com-munities became concerned about school participation, and changed schoolschedules and the timing of vacations. On the other hand, effects on incomepermitted communities to invest in schooling and the improvement of childwelfare under a cooperative system (von Braun, Hotchkiss, and Immink 1989).

Scenario 6: Eviction of Tenants and Effects on Land Markets

The profitability of new crops or of crops grown with improved technologymay increase the landowners' incentive to evict tenants and move to owner op-eration of farms. An examnple fostered by administrative ruling is found in anarea studied in the Philippines where contracts for sugarcane growing were notgiven to tenants but only to landowners. As a result, landlessness expanded andthe status of tenants deteriorated in an area around the sugarmill (Bouis andHaddad 1990).


The green revolution made farming in Pakistan more profitable. A failure ofland rents to adjust immediately to higher profits may have been instrumentalin the increased self-cultivation by landowners and the increases in farm sizethat followed. Results of surveys and censuses show that tenant cultivationdeclined sharply during the green revolution. Between the 1960 and the 1980censuses, the number of pure tenant farmers declined from 2.0 million to1.1 million, while the area under tenant farms declined from 19 million acresto 10 million acres.

Two other things accelerated the decline of tenant farming in Pakistan. First,aggressive subsidization of mechanization increased purchases of tractors andother technology.1 The tractors made large farms less dependent on bullockdrivers or tenants, or both. A very large decrease in tenant labor hours is re-ported by McInerney and Donaldson (1975) in a "before and after" study oftractorization. The World Bank financed loans for the purchase of large trac-tors at substantial subsidies to farmers. Land ceilings or tenancy laws inPakistan either did not exist or were ineffective, and the 202 farms surveyedgrew on average from 18.2 to 44 hectares. Of the additional land acquired,32.3 percent came from reduction in land rented out, 28.6 percent from in-creased renting, 26.2 percent from reclamation and improvement, and 13 per-cent from purchases. Each tractor replaced an average of 4.5 tenants.

Changes in land tenancy legislation in 1959 and 1972 were the second impe-tus to the decline of tenant farming. The new laws were increasingly adverseto landowners (Nabi, Hamid, and Zahid 1986) so that renting out land becameless profitable and more risky. The decline in tenancy was thus brought aboutby a combination of technological change and bad policy. It is hard to assesshow much each contributed to the trend in this particular instance; the relationis easier to disentangle in two similar examples from Africa. The first is a caseof extensive eviction of land users in Sudan as a direct result of promotion oflarge-scale mechanized sorghum production (Elhassan 1988). The governmentdecided to allocate land in large sizes to farmers who were prepared to investin mechanized farming. For a rural population dependent on wages, this meantdeclining employment with little opportunity to provide for household or com-munity food security. The effects were worst for the poorest: rates of child mal-nutrition in these areas were found to be significantly higher than in thetraditional rain-fed sector (Sudan 1988). Here the eviction, which took placebefore the new technology was introduced, can convincingly be attributed tobad policy. A similar instance, in which the tenants were evicted after the newtechnology was in place, is the case of CADU (a program for promotion of im-proved agricultural inputs and market integration launched in the late 1960s inEthiopia). Here the outcome is attributable not to bad policies but to theabsence of good supplementary policies. Almost all the tenants in areas underCADU were evicted once the new technology was disseminated to large produc-ers and landowners (Cohen 1975). Policies and programs to provide tenants


with access to technology and to protect their tenancy rights have been lackingin these areas.

Scenario 7: Coerced Production or Forced Procurement

One of the worst outcomes of commercialization is associated with coercedproduction. Governments or powerful monopsonistic procurement partnersmay respond in this way in an attempt to shift losses from an ill-designedscheme for commercialization to farm producers, or when the monopsonisticprocurement agent is trying to capture excessive profits. Since the poor usuallyhold a very weak position in the political arena, they are particularly vulnerableto such perverse policy.

One such case of coerced production evolved out of an unsuccessful tea pro-duction scheme in northwest Rwanda (there are several successful schemes inthe country). Smallholders were talked into producing tea that did not turn outto be profitable for them. Supply was therefore not forthcoming, and theprocessing capacities established for the scheme were underutilized. To increasecapacity utilization, the parastatal tea factory then expanded its tea plantationby expropriating small farms in the vicinity of the factory (von Braun, de Haen,and Blanken forthcoming). Schemes for area allotment with procurement reg-ulations that result in coerced production are widespread. Examples includecotton and rice schemes in Egypt and cereal programs in parts of China.

The opposite side of the coin of enforced production is exclusion from pro-duction opportunities. In colonial times, bans on cash crop production by localsmall farmer populations were ubiquitous. Export crops were specifically re-served for settler farmers in Kenya, Tanzania, and other East African colonies.Even when cultivation of cash crops was not explicitly prohibited, extension,credit, and marketing services were not available to the native smallholder. InZimbabwe, for example, these services were reserved for large-scale (white)farmers. Independence swept away most of these constraints, opening up theopportunities to all farmers. But the constraints on participation of smallhold-ers in commercial crops are reemerging in the form of new regulations passedby indigenous elites. Such a case is the reservation of tobacco production forthe estate sector in Malawi, with small farmers allowed to participate in thelucrative tobacco economy only by contractual arrangements with estate own-ers (Lele, van de Walle, and Gbetibouo 1989).

Targeting Technological Change and Commercialization

An extension of the optimistic view of technology and commercialization-that, through growth, they bring benefits to the poor and that most adverseeffects are caused by bad policy and can be rectified by appropriate policy-isthe idea that these forces can be instruments specifically directed towardalleviating poverty. Before we review the pros and cons of specific possibilities


for targeting, two problems of this form of "social engineering" need to be em-phasized.

First, the research necessary for developing poverty-targeted technologymay have opportunity costs. Introducing the extra element of targeting into re-search and development originally oriented exclusively to growth introducesconstraints that may result in forgone growth. The hypothetical tradeoff thenis whether untargeted, but possibly higher, agricultural growth has more po-tential to alleviate poverty than development and dissemination of poverty-targeted technology.

Second, poverty-targeted technology and market development cannot be as-sessed in isolation. They need to be ranked against alternative instruments forthe alleviation of poverty that may be available, such as targeted health andnutrition interventions or rural employment programs. Comparing these instru-ments is not, of course, easy.

Below we discuss six targeting possibilities. Demonstrating feasibility of anyone of them must not be taken as synonymous with proving effectiveness-especially in view of the important role, stressed earlier in this article, playedby rural growth in the alleviation of poverty.

Targeting by Agroclimatic Potential

Green revolution technology reduced poverty in the favorably endowedregions where it could be adopted (Hossain 1988), and commercialization didthe same in areas with known or hidden potential. It is therefore tempting toaccelerate technological change and commercialization in agroclimatically"poor" regions in the hopes of reaching out to the poor. The developing coun-tries and the donor community have undertaken research and launched projectsto replicate green revolution success in such low-potential areas as the Sahel,the semiarid zones in India, and the humid tropics of Africa. So far the invest-ment has not had very powerful effects-complex environmental constraints inthese zones cannot be easily overcome but there have been some successeswith sorghum, ragi, and perhaps millet breeding in India, and with hybridmaize in East Africa. In the high-potential areas, research for wheat and ricecould build on a long history of (re)search for technological change; further-more, costs of adaptive research are lower in those areas which enjoy a highdegree of water control. But research at the regional level has perhaps been go-ing on for too short a time and at a scale insufficient to tap the hidden possi-bilities for low-potential areas.

Targeting Foods Consumed by the Poor

The poor consume substantial amounts of certain staple foods that are con-sidered inferior. Directing research toward roots, tubers, and coarse grains isone possible means of exploiting the treadmill effect in favor of poor consum-


ers. In comparison with research on wheat and rice, the research history forthese crops is rather short, and the jury is still out for assessing the effectivenessof this approach. Moreover, such targeting, even if successful, may in somecases become less powerful. Recent evidence suggests that low-income house-holds are shifting into more time-saving staple foods. Cases in point are theincreasing consumption of rice by the urban and rural poor in West Africa(Delgado 1989) and of wheat by low-income households in Sri Lanka (Senauer,Sahn, and Alderman 1986). The cost of women's time appears to be an impor-tant factor in these changes.

Many of the poor, however, still consume a lot of what they produce. Whentechnical change occurs in crops produced and consumed by the same house-holds, the treadmill effect is irrelevant, as is the potential conflict between netbuyers and sellers. Increased levels and stability of yields of subsistence cropspromise to mitigate the chronic as well as transitory food insecurity of the sub-sistence household. The pressure from increasing population rapidly raises de-mand for yield-increasing technology in subsistence crops where food and labormarkets are risky. For instance, in Rwanda, the production of sweet potatoesrises rapidly when ratios of people to land increase (von Braun, de Haen, andBlanken forthcoming). And in many parts of the Sahel, the drought sequenceof the 1980s (a transitory problem) led to a rapid shift from sorghum into earlymillet, with lower mean yields but higher drought resistance. Efforts by theConsultative Group on International Agricultural Research and national re-search systems to increase productivity in sweet potatoes and improve droughtresistance in coarse grains remain important avenues for research.

Targeting Nutrients and Dietary Components

Attempts to improve the nutrition of the poor through technical advance orcommercial development have focused on plant breeding and on nutritionaltargeting in rural development.

From the conventional wisdom about the nature of the malnutrition problemin the 1960s and 1970s many efforts sprang to increase the protein content andthe content of certain amino acids, such as lysine, in cereals. The InternationalAgricultural Research Centres (IARCs) and many national programs participat-ed. By the early 1980s most breeding activities whose goals were to increasecertain nutrients had been abandoned. The nutritional traits had low heritabil-ities and competed with the achievement of other attributes, such as yield andresistance to disease and pests (for a review of the experience, see Pinstrup-Andersen, Berg, and Foreman 1984; for other proposals, see Lipton andLonghurst 1989). Nutritionists shifted emphasis from the protein gap to energydeficiency. It has become apparent that the diversity of diet of the poor hasfrequently been underestimated in the past: nutritional characteristics of com-modities need to be seen in the context of the total diet of the poor.


In their plant breeding, IARCs nowadays generally consider minimum stan-dards of nutrient content in addition to other characteristics of consumptionand processing (palatability, preparation requirements, storage characteristics,and so on). Concern for meeting the poor's specific constraints on resources(for instance, time, cooking energy, absorbability by young children) may bemore relevant than the nutrient content of a specific commodity, and these mayoffer some scope for targeting.

An example of nutritional targeting in rural development is dairy develop-ment schemes. The rationale is that these schemes directly contribute to im-proving the nutrition of dairy producers by making more milk available forhome consumption. In most instances, however, the effects on consumptionand nutrition are more indirect. Poor dairy smallholders sell expensive calories(milk) and increase net purchases of cheap calories, thereby improving foodconsumption (Alderman, Mergos, and Slade 1987). Thus the favorable effectson nutrition arise from the link between commercialization and income ratherthan directly from the effect of technology on production. Facilitating thepoor's access to the technology and to marketing outlets makes the nutritionalbenefits possible.

Speeding Up Adoption of Technology

We noted before that late adoption of new teclnology and market opportu-nities is a problem with particular repercussions for the poor. Treadmill effectstend to push prices down, and late adopters forgo the early rents brought bytechnology and commercialization, as well as find procurement and processingopportunities blocked or closed by the time they enter the arena. (Note, how-ever, that encouraging early adoption carries with it greater risks-the risksthat the innovators' rents reward.)

Extension and credit are the instruments usually advocated for speedingadoption by the poor. Another promising instrument is the marketing cooper-ative. Policies to institute early cooperative marketing arrangements can bepowerful and sustainable tools for targeting marketing opportunities to thepoor. The Guatemalan example of the export vegetable cooperative and theIndian dairy cooperative system are both cases in point.

Targeting Poor Producers

Targeting the poor frequently means targeting farm households with highlabor-land ratios or low capital-labor ratios, or both. The poor are also moresubject to credit constraints than large farmers. It follows that poor farmerswould more readily adopt technologies that do not require high capital-inputratios-for example, disease-resistant varieties would be easier to adopt thanpesticides to combat the diseases. Emphasis in plant breeding on resistance and


tolerance to pests, diseases, and moisture stress would therefore be especiallybeneficial to poor farmers without lessening their advantage to large farmers.

The much discussed question of whether green revolution varieties are supe-rior to traditional varieties only with high doses of fertilizers has clear impli-cations for the poor. Poor farmers would be deterred if large doses of fertilizerwere needed. Most results from experiment stations where the environment isheld constant suggest that high-yielding varieties outyield traditional varietiesat both high and low fertilizer doses.

Rental markets, credit markets, and tied contracts allow the poor to rent outlabor and obtain capital, so that even if endowment ratios differ stronglyamong farmers' groups, factor-use ratios are often much closer (Ryan andRathore 1980). If a small tenant farmer can get the landlord to pay for fertilizer,he or she may still adopt the optimum dose. Resources will be efficiently allo-cated, although the landlord is likely to extract the benefit of the fertilizer fromthe tenant by writing the rental contract accordingly. Better targeting technol-ogy to the purchasing power of the poor would therefore still improve equity.However, the efficiency gains that could be achieved by better-targeted technol-ogies would be small.

Even in cases in which the poor specialize in some enterprise, targeting thatenterprise for technology clevelopment may not benefit them, if richer farmersrespond to the enhanced profitability of the enterprise. Jodha's (1985) intensivesurveys of Rajasthan villages illustrate the problem. Raising sheep and goatswas a traditional occupation of the local and seminomadic tribal group inRajasthan; one would therefore have thought that benefits from research onsheep and goats would accrue primarily to these poor groups. Following theland reform in Rajasthan in 1952, the profitability of raising sheep and goatsincreased sharply because households no longer had to pay land revenue to feu-dal landowners and were not charged for the use of common property resourc-es. Prices rose too. Jodha found that, as a result, by 1963-64 many high-castehouseholds in his study villages were engaged in sheep and goat production,whereas in 1955 no higher-caste households had invested in these enterprises.Any researchers who thought that benefits from sheep and goat research in1950 would have benefited only the poorer households in the community wouldhave been disappointed.

Targeting by Gender

In view of the disadvantaged situation of women in many rural environ-ments of the developing wvorld, development organizations and research insti-tutes have been giving increasing attention to the potential for directing thebenefits of technology and commercialization to women.

Experience with a program for irrigated rice development in the Gambiaspecifically targeted toward women underlines the difficulties. In the Gambia,women are the traditional rice growers, and it was thought that technology for


Table 3. Rice-growing Technology and Women Farmers in the Gambia,Wet Season, 1985

ImprovedIrrigation rain-fed Traditional

Variable by pump system technology

Fields under women's control (percent) 10.0 77.0 91.0Yield (tons per hectare) 5.9 2.5 1.3Cost of inputs (U.S. dollars per hectare)a 294 154 20Women's labor (as a percentage of

unpaid family labor) 29.0 60.0 77.0

a. Variable cost of inputs (seed, fertilizer, irrigation, hired labor, transportation, mechanized landpreparation). Dollar converted at parallel exchange rate (US$1 = dalasi 6).

Source: Survey by the International Food Policy Research Institute, Washington, D.C., and Programming,Planning, and Monitoring Unit, the Gambia, 1985/86.

this crop would therefore directly benefit poor women. But men responded tothe increased profits available from growing rice, with the result that a stronginverse relation developed between the level of technology and the control ofwomen over the newly improved crop (table 3). The reason was that the tech-nology changed the very nature of the cropping arrangements and induced alarge shift of male labor from communal agriculture into the newly developedrice perimeters to overcome labor bottlenecks. Constraints on women's accessto credit and to hired labor exacerbated the problem (von Braun, Puetz, andWebb 1989). The outcome suggests that the full range of constraints underwhich women operate needs to be understood when targeting by gender is at-tempted.

In this connection, it is worthwhile to consider ex ante implications of un-targeted new technology for poor women in rural areas. In semiarid Indiaalmost all hand weeding is done by hired women, and earnings from handweeding make up a significant share of women's wage income. Thus the pri-mary effect of any reduction in this task by herbicides would be to reduce thework and income opportunities of the most disfavored labor group, femaleagricultural laborers. Technological development, which would make herbicideuse more cost-effective, would certainly have adverse effects on female laborers(Binswanger and Shetty 1977).

Conclusion

How then do the optimistic and the pessimistic viewpoints fare? Even withwell-functioning markets for factors and products, it is easy to construct sce-narios in which poor producers lose from technical change. All these scenariosdepend on highly inelastic demand. The resulting agricultural treadmill is a re-ality with important regional and international dimensions. Its potentially


serious damage is often diluted by inbuilt compensating effects. In particular,its favorable effects for consumers-especially given that the majority of thepoor are net purchasers of food-should be taken into account when weighingits disadvantages for small nonadopters. Once the consumption effects andother general equilibrium effects are included in the assessment, the treadmilleffects are usually seen to be diffused (although commodities such as coffee,principally consumed by developed countries, do not produce these benefits forpoor consumers).

Commercialization and specialization, in contrast, are usually introduced forcommodities for which demand is elastic-often as a means of bypassing theproblem of inelastic demand faced by traditional commodities. It is thereforehard to construct scenarios in which commercialization by itself-unaided byfailures of institutions, policies, or markets-brings adverse consequences.

The relative seriousness for the poor of the various scenarios differs; theworst outcomes arise when several scenarios or effects coincide. Late adoptionof new technology is a case in point. The risks associated with the new tech-nology discourage the poor farmer from adopting it early; in conjunction withtreadmill effects, late adoption is likely to injure the profits of poor farmers orclose doors for them, or both.

Many of the adverse scenarios arise not because of the inherent nature ofthe technology or the commercialization opportunity but because of bad policy.Constraints on trade, coercion in production, and ill-advised tenancy laws aregovernment actions that may turn a promising opportunity into a disaster forthe poor. The answer to many of these issues is policy reform rather than thereversal or deceleration of technological advance and commercialization(Nerlove 1988).

Possibly adverse effects such as late adoption can also be mitigated by gov-ernment action. Credit policies and extension services are often needlessly bi-ased against the poor. Government policies can facilitate market or capacityexpansion where doors have been closed and help the poor to seize the oppor-tunities and derive the benefits.

The conclusion that policy changes can either avert or mitigate adverseeffects is based on the assumption that the policy and institutional response canbe exogenous and independent of the technology or the expanded commercial-ization. In some cases, however, institutional changes and policy responses arenot exogenous but reflect existing conflict among social groups. The perverseresponses then are a logical outcome of these conflicts and cannot be alteredby a benevolent policy. Some of the cases of tenant eviction fall in this category.Where institutional and policy responses are endogenous in this way, more pes-simistic conclusions are warranted about the benefits of technology and com-mercialization for politically weak poor groups. An important issue for furtherempirical research, therefore, is the extent to which these responses are endog-enous (de Janvry and Sadoulet 1988).


The case studies discussed afford little support for the view that technolog-ical progress and commercialization in and of themselves are harmful to thepoor. Nor, at the other extreme, does the evidence encourage the idea that itis easy to engineer them successfully into instruments of targeted poverty alle-viation. We are thus back to the view that technological change and commer-cialization do expand opportunities, bringing Large general benefits whosecomplex distributional implications are hard to predict. Targeting is difficultbecause of unpredictable distributional implications, and its scope is often lim-ited by technological tradeoffs or agroclimatic constraints.

However, opportunities for cost-effective targeting do exist and should beseized whenever possible. Equally important, if not more so, is the eliminationof policies and intervention that alone, or in interaction with technical changeand commercialization, harm the poor, and that imay needlessly bias the avail-ability of productive support services against them.

Note

Hans Binswanger is chief of the Agricultural Operations Division in the CountryDepartment II (Central America) of the World Bank. Joachim von Braun is director of the FoodConsumption and Nutrition Policy Program of the International Food Policy Research Institutein Washington, D.C.

1. Certain forms of mechanization, especially when heavily subsidized, are bad for the laborincome of the poor. The effects of mechanical technology ate not considered in this article. Fora review, see World Bank (1987).

References

The word "processed" describes works that are reproduced from typescript by mimeograph,xerography, or similar means; such works may not be cataloged or commonly available throughlibraries, or they may be subject to restricted circulation.

Alderman, Harold, George Mergos, and Roger Slade. 1987. Cooperatives and the Commercial-ization of Milk Production in India: A Literature Review. Working Paper on Commercializa-tion of Agriculture and Nutrition 2. Washington, D.C.: International Food Policy ResearchInstitute.

Bell, Clive, Peter Hazell, and Roger Slade. 1982. Project Evaluation in Regional Perspective.Baltimore and London: Johns Hopkins University Press.

Berry, R. A., and W. R. Cline. 1979. Agrarian Structure and Productivity in Developing Coun-tries. Geneva: International Labour Office.

Bhalla, G. S. 1983. The Green Revolution and the Small Peasant: A Study of Income Distributionamong Punjab Cultivators. New Delhi: Concept Publishing.

Binswanger, Hans, and S. V. R. Shetty. 1977. "Economic Aspects of Weed Control in the Semi-Arid Tropical Areas of India." Occasional Paper 13. International Crops Research Institute forthe Semi-Arid Tropics, Economic Department, Hyderabad, India. Processed.

Bouis, Howarth, and Lawrence Haddad. 1990. Effects of Agricultural Commercialization onLand Tenure, Household Resource Allocation, and Nutrition in the Philippines. ResearchReport 79. Washington, D.C.: International Food Policy Research Institute.

Cohen, John M. 1975. "Effects of Green Revolution Strategies on Tenants and Small Landownersin the Chilalo Region of Ethiopia." Journal of Developing Areas 9, no. 3: 335-58.

78 The World Bank Research Observer, vol.6, no. 1 (January 1991)

de Janvry, Alain, and Elisabeth Sadoulet. 1988. "Alternative Approaches to the Political Economyof Agricultural Policies: Convergence of Analytics, Divergence of Implications." Paper forTwentieth International Conference of Agricultural Economists, Buenos Aires, August.Processed.

Delgado, Christopher. 1989. "Why Is Rice and Wheat Consumption Increasing in West Africa?"Paper for Nineteenth European Seminar of the European Association of Agricultural Econo-mists, Montpellier, France, June. Processed.

Elhassan, A. Mohammed. 1988. "The Encroachment of Large-Scale Mechanized Agriculture:Elements of Differentiation among the Peasantry." In Tony Barnett and Abos Abdelkarim,eds., Sudan: State, Capital, and Transformation. New York: Croom Helm.

Herdt, R. W., and C. G. Ranade. 1976. "The Impact of New Rice Technology on the Shares ofFarm Earnings, Laguna and Central Luzon, Philippines." Paper 76-1. International RiceResearch Institute, Los Banos, Philippines. Processed.

Hitchings, Jon. 1982. "Agricultural Determinants of Nutritional Status among Kenyan Childrenwith Model of Anthropometric and Growth Indicators." Ph.D. diss., Stanford University,Stanford, Calif. Processed.

Hossain, Mahabub. 1988. Nature and Impact of the Green Revolution in Bangladesh. ResearchReport 67. Washington, D.C.: International Food Policy Research Institute.

Jodha, N. S. 1985. "Population Growth and the Decline of Common Property Resources inIndia." Population and Development Review 2, no. 2: 247-64.

Kennedy, Eileen, and Bruce Cogill. 1987. Income and Nutritional Effects of the Commercializa-tion of Agriculture in Southwestern Kenya. Research Report 63. Washington, D.C.: Interna-tional Food Policy Research Institute.

Kikuchi, M. A., A. Huysman, arid L. Res. 1983. Agriculture, Labor Absorption and Migration:Comparative Histories of Two Philippine Rice Villages. International Rice Research InstituteResearch Paper 90. Los Banos, Philippines.

Konjing, Chaiwat. 1989. Trends and Prospects for Cassava in Thailand. Working Paper onCassava 6. Washington, D.C.: International Food Policy Research Institute.

Krueger, Anne O., Maurice Schiff, and Alberto Valdes. 1988. "Agricultural Incentives in Devel-oping Countries: Measuring the Effect of Sectoral and Economywide Policies." World BankEconomic Review 2, no. 3: 255-71.

Lele, Uma, Nicolas van de Walle, and Mathurin Gbetibouo. 1989. "Cotton in Africa: An Analysisof Differences in Performance." Paper prepared for Conference on Development and Agricul-tural Growth in Africa: Lessons from MADIA (Managing Agricultural Development in Africa),Annapolis, Md., June. Processed.

Leslie, Joan, and Michael Paolisso. 1989. Women's Work and Child Welfare. Boulder, Colo.:Westview.

Lipton, Michael, and Richard Longhurst. 1989. New Seeds and Poor People. London:Hutchinson. Baltimore: Johns Hopkins University Press.

McInerney, John P., and Graham E Donaldson. 1975. The Consequences of Farm Tractors inPakistan. Staff Working Paper 210. Washington, D.C.: World Bank.

Mellor, John. 1986. "Agriculture on the Road to Industrialization." In John P. Lewis andValeriana Kallab, eds., Development Strategies Reconsidered. U. S. Third World PerspectivesS. New Brunswick, N.J.: Transaction Books.

Nabi, Ijaz, Naved Hamid, and Shahid Zahid. 1986. The Agrarian Economy of Pakistan: Issuesand Policy. Oxford: Oxford University Press.

Nerlove, Mark. 1988. Modernizing Traditional Agriculture. Occasional Paper 16. Panama:International Centre for Economic Growth.

Paulino, Leonardo A. 1986. Food in the Third World: Past Trends and Projections to 2000.Research Report 52. Washington, D.C.: International Food Policy Research Institute.


Pinstrup-Andersen, Per, Alan Berg, and Martin Foreman, eds. 1984. International AgriculturalResearch and Human Nutrition. Washington, D.C.: International Food Policy Research Insti-tute. Rome: United Nations Subcommittee on Nutrition.

Quiz6n, Jaime, and Hans Binswanger. 1986a. "Modeling the Impact of Agricukural Growth andGovernment Policy on Income Distribution in India." World Bank Economic Review 1, no. 1:103-48.

. 1986b. "The Impact of Agricultural Growth and Selected Government Policies on theDistribution of Income in India: Regional Analyses." World Bank, Washington, D.C.Processed.

Ruttan, Vernon W., and Hans P. Binswanger. 1978. "Induced Innovation and the Green Revolu-tion." In Hans P. Binswanger and Vernon W. Ruttan, eds., Induced Innovation: Technology,Institutions, and Development. Baltimore: Johns Hopkins University Press.

Ryan, J. G., and M. S. Rathore. 1980. "Factor Proportions, Factor Market Access and the De-velopment and Transfer of Technology." In Indian Agricultural Research Institute, Proceedingsof the Seminar on Economic Problems in Transfer of Agricultural Technology. New Delhi.

Senauer, Ben, David Sahn, and Harold Alderman. 1986. "The Effect of the Value of Time onFood Consumption Patterns in Developing Countries: Evidence from Sri Lanka." AmericanJournal of Agricultural Economics 68, no. 4 (November): 920-27.

Sidhu, Surjit S. 1972. "Economics of Technical Change in Wheat Production in Punjab, India."Ph.D. diss., University of Minnesota. Processed.

Sudan, Republic of. Ministry of Health and Social Welfare. 1988. Sudan Emergency and Recov-ery Information and Surveillance System (SERISS). Khartoum.

von Braun, Joachim. 1989. "Production, Income, and Employment Effects of Commercialization:Implications for Household Food Security and Lessons for Agricultural Policy." Paper contrib-uted to the IFPRI/INCAP Policy Workshop. Antigua, Guatemala. March 9-11. Processed.

von Braun, Joachim, Hartwig de Haen, and Juergen Blanken. Forthcoming. Commercializationof Agriculture under Population Pressure: Production, Consumption and Nutritional Effectsin Rwanda. Research Report. Washington, D.C.: International Food Policy Research Institute.

von Braun, Joachim, David Hotchkiss, and Maarten Immink. 1989. Nontraditional ExportCrops in Guatemala: Effects on Production, Income, and Nutrition. Research Report 73.Washington, D.C.: International Food Policy Research Institute.

von Braun, Joachim, and Eileen Kennedy. 1986. Commercialization of Subsistence Agriculture:Income and Nutritional Effects in Developing Countries. Working Paper on Commercializa-tion of Agriculture and Nutrition 1. Washington, D.C.: International Food Policy ResearchInstitute.

von Braun, Joachim, Eileen Kennedy, and Howarth Bouis. 1989. "Comparative Analyses of theEffects of Increased Commercialization of Subsistence Agriculture on Production, Consump-tion, and Nutrition." Report to USAID. International Food Policy Research Institute,Washington, D.C. Processed.

von Braun, Joachim, Detlev Puetz, and Patrick Webb. 1989. Irrigation Technology and Commer-cialization of Rice in The Gambia: Effects on Income and Nutrition. Research Report 75.Washington, D.C.: International Food Policy Research Institute.

von Oppen, Mathias. 1989. "Regional Production and Interregional Trade of Agricultural Com-modities in Benin." Paper for Nineteenth European Seminar of the European Association ofAgricultural Economists, Montpellier, France, June. Processed.

Webb, Patrick. 1989. "When Projects Collapse: The Impact of Agricultural Project Failure in TheGambia from a Household Perspective." International Food Policy Research Institute,Washington, D.C. Processed.

World Bank. 1987. Agricultural Mechanization: Issues and Options. World Bank Policy Study.Washington, D.C.


THE EARMARKINGOF GOVERNMENT REVENUEA Review of Some World BankExperience

William McCleary

Economic theory provides some justification for earmarking. Some economistsargue that by assigning revenue from specific sources to specific purposes, agovernment can facilitate agreement about increasing both revenue and expen-diture in cases in which there would be no consensus about raising either sep-arately. Earmarking may also protect high-priority programs from shiftingmajorities, inefficiency, and corruption.

Most economists, however, have been skeptical about earmarking. In prac-tice, it is difficult to achieve pricing and taxation arrangements that will allo-cate resources appropriately for the service in question and yet require fewadministrative decisions. Often, efficient pricing and taxing lead to unbalancedbudgets for the earmarked fund and hence to interdependence with the generalbudget. The independence of earmarked funds is further compromised becausethey so often depend on the government for additional, non-earmarked sourcesof funding, or for reiterated endorsement of decisions about prices or taxes forthe earmarked sources. Governments also frequently withhold the allottedfunds, fail to make necessary changes in prices or taxes, or simply suspend theearmarking arrangements.

This article looks at the arguments for and against earmarking and goes onto illustrate the discussion with case studies from World Bank experience. Thelessons from the real world appear to bear out the skepticism of the majorityabout earmarking: in general, it has not worked very well. The article, there-fore, concludes by cautioning against the practice except under certain definedand restrictive conditions.

The World Bank Research Observer, vol. 6, no. I (January 1991), pp. 81-1043 1991 The International Bank for Reconstruction and Development/THE WORLD BANK 81

E armarking is the practice of assigning revenue-generally through statuteor constitutional clause-from specific taxes or groups of taxes tospecific government activities or areas of activity. Sometimes there is lat-

itude for earmarked funds to be supplemented by revenue from other sourcesor to be diverted to other uses. The distinguishing feature of earmarking, atleast in its pure form, is the desire to protect a certain category of expenditurefrom the vagaries of the political process by linking it to a particular source ofrevenue. Once this link is established, the system supposedly runs on automaticpilot, and the level of revenue over time drives the level of expenditure.

Principles of Earmarking

At first glance, earmarking appears to apply the benefit principle of taxation(see Bird 1978); that is, the beneficiaries of the government activities pay thetaxes or charges. But benefit taxation can exist independent of earmarking, andvice versa: benefit taxes may be added to the central pool, or revenue may beearmarked for activities that do nothing for the contributing taxpayers.

Four broad types of earmarking, categorized by the specificity of tax sourceand expenditure, are shown in table 1. The most prevalent forms (type A) aregasoline taxes and motor vehicle fees for expenditure on highways and contri-butions of employers and employees to social security and unemployment in-surance funds. Revenue sharing between various levels of government (type D)is also common. A subcategory of type A forms "strong earmarking," in whichthe beneficiaries clearly pay for the goods or services provided. These goods orservices have the characteristics of private goods-few or no external benefitsor costs that affect persons other than the users, no claim by recipients for spe-cial treatment on grounds of income distribution, and no significant inefficiencyresulting from implementing a charge. Transactions involving these goods giveoff signals to the market about the amount desired and the willingness to pay.Strong earmarking could, in fact, be extended to public goods as well, provideda way could be found to assess what people would be willing to pay for alter-native amounts of the good, which would allow a decision to be reached abouthow much to provide and how to divide the tax bill for that quantity amongcitizens according to their preferences. It is also worth noting that public en-terprises-to the extent that they are allowed to retain control over profits-represent a form of strong earmarking; the purity of each case would dependon its independence from the budget for subsidies or other special financingarrangements.

The remaining cases of type A earmarking and all cases of types B, C, and Dare examples of weaker earmarking: the connection between those assessed andthe beneficiaries is tenuous or nonexistent, and objectives related to redistribu-tion or other aspects of social welfare commingle with allocative objectives.The designation of a liquor tax for financing education, for example, gives no

82 The World BankResearch Observer, vol.6, no. 1 (January 1991)

Table 1. Varieties of Earmarking

Type Revenue Expenditure Examples

A Specific tax Specific end use Gasoline taxes andor fee motor vehicle fees for

highway investments.Social security,

unemployment funds.Support of public

enterprises.

B Specific tax Broad end use Lottery proceeds andor fee sin taxes (on tobacco and

alcohol) to financesocial sector programs.

Taxes and royalties frompetroleum to financedevelopment.

C General tax Specific end use Fixed percentage of totalrevenue devoted tospecific programs(such as education).

Revenue sharing for aspecific purpose.

D General tax General end use Revenue sharing.

useful information about the appropriate level of education and represents astraight transfer between two groups of citizens. The same is true of earmark-ing a fixed (and arbitrary) percentage of general revenue for education. Thelatitude for supplementing earmarked funds from general revenue, or divertingthem to other uses, can weaken all forms of earmarking. In such instances, itis not clear what purpose earmarking serves, because decisions about whetherto use funds for the earmarked purpose rather than some other objective ap-pear to dictate the allocation of resources at the margin.

The remainder of this discussion concentrates on efforts to provide fairlyspecific goods or services with revenue from a clearly designated tax or fee(types A and B in table 1). It excludes special cases of earmarking, such as so-cial security, public enterprises, and revenue sharing between levels of govern-ment. The article starts out with a general discussion of the issues: a sketch ofthe cases against and for earmarking as they have appeared in the literature;an investigation of the feasibility-and the pros and cons-of developing rulesfor setting prices and taxes and for regulating how proceeds will be spent onearmarked goods; and an examination of whether earmarking is justified whenthere is little or no connection between the beneficiaries of public expenditureand the payers of taxes or prices. This discussion is then illustrated withsome lessons from World Bank experience with several prominent cases of

William McCleary 83

earmarking: highway funds, betterment taxes to finance local expenditure, andthe extensive earmarking programs in Turkey and Colombia. The final sectiondraws some preliminary conclusions and suggests some criteria that earmarkedfunds should meet.

The Case against Earmarking

Among economists and public administrators, earmarking has but few sup-porters. The litany of principal objections is summed up by Deran (1965):

* It leads to a misallocation of resources, with too much being given to ear-marked activities and not enough to others.

* It hampers effective budgetary control (depending to some degree onwhether provisions are embedded in statutes or in the constitution).

* It infringes on the powers and discretion of the legislative and executivebranches of government.

* It introduces inflexibility into budgets: changes come only after a lag, andearmarking systems continue after their usefulness has been served.

Earmarking, in short, is seen as infringing on discretion; by reducing thescope of the executive and legislative branches' command over the allocationof resources, it builds some rigidity into the system and reduces flexibility. Ear-marked expenditures, exempt from fluctuating general revenue, may attract adisproportionate share of funds at times when resources are short (provided, ofcourse, that sources of earmarked funds are more stable than those of generalrevenue). Earmarking has also been criticized because it removes expenditurefrom close public scrutiny. Earmarked expenditure may not have to satisfy thesame rigorous evaluation as other budgetary expenditure, so that funds may bediverted to low-priority projects or squandered on needless overhead.

The foregoing arguments have some merit. At the very least, they suggestthat expenditure of earmarked money should be formally evaluated and strictlyaccounted and audited and that there should be periodic review of whether theearmarking is still warranted.

The Case for Earmarking

The case against earmarking is not persuasive under all circumstances. Thedefects cited by its critics are the virtues cited by its proponents, who arguethat rigidity and limitations on the possibility of reallocating resources cansometimes be desirable. Moreover, the flexibility of financing programs fromgeneral funds can be overstated. Monies are not readily moved from one cate-gory of expenditure to another: spending on debt service, social insurance, and,often, administration and security take precedence; expenditure programs, oncestarted, take on a life of their own. Some countries have established safeguardsagainst corruption or the misuse of funds so strict that it is difficult to begin


programs or to obtain timely funding for them once approved. In such cases,earmarking is an escape from overly rigid procedures in the general budget(see, for example, Bird 1982).

That earmarking is common in the real world-and that some programs areapparently successful-warrants further investigation into the merits of thepractice. Adherents of earmarking-few as they are-cite a number of advan-tages (see Deran 1965):

* Earmarking applies the benefit principle of taxation (at least in somecases).

* Earmarking gives more assurance of minimum levels of financing for publicservices that governments consider worthy, thus avoiding periodic hagglingwithin the bureaucracy or between the bureaucracy and the legislature overappropriate levels of funding.

* Greater stability and continuity of funding may lead to lower costs becauseof speedy completion of projects.

* By linking taxation with spending, earmarking may overcome resistance totaxes and help to generate new sources of revenue.

To the extent that there is overlap between payers and beneficiaries, ear-marking takes on some of the virtues of benefit taxation: the simultaneous link-ing of the expenditure and revenue sides of the budget; the provision ofadequate financing for public goods and services; and the elimination of excessdemand or shortages that occur when goods or services are provided below(marginal) cost or free. In other words, benefit taxation, like the price mecha-nism, provides appropriate signals for the efficient allocation of resources andfinancing for government services.

In the long search for a fair system of taxation, the potential for linkingspending decisions to decisions on taxes and prices has been a recurrenttheme-from Adam Smith through Wicksell and Lindahl down to Samuelson,Musgrave, and Buchanan (Wicksell 1958, Lindahl 1958, Samuelson 1954 and1955, Musgrave 1959, Buchanan 1968). In a fair tax system, these economistscontend, people would pay according to what they received, and taxes wouldbe set according to the marginal benefits received by taxpayers. Expenditure-both in total and for individual public goods and services-would be expandedso long as, at the margin, benefits received by all individuals exceeded marginalcosts (since the marginal benefits and costs reflect the gains and losses to soci-ety from expanding the provision of any good). Thus total spending (and itscomposition), and total taxation and its distribution across taxpayers, wouldbe determined simultaneously.

As elegant as such a system would appear in theory, its design in practicehas proved problematic. It leaves open two questions: first, how to handle taxesor expenditures that have distributional objectives (that is, how to financethe redistribution), and second, how to induce the public to reveal its prefer-ences in relation to pure public goods-especially if people know that such

William McCleary 85

revelations might increase tax bills. Although the budgetary process-throughthe sensitivity of elected officials to voters' desires-to some extent performsthe function of revealing preferences, it is not at all clear how accurately ittranslates the public's wishes into concrete expenditure and tax programs.Nonetheless, this strand of public finance theory points up, first, the impor-tance of regarding expenditure on the one hand and tax and financing on theother as an entity, and, second, the fact that this link is crucial to decisions onresource allocation. Indeed, the "public-choice economists"-James Buchanan,Charles Goetz, and others-would argue that in some cases the link is essentialto reaching any collective decision to expand government activity.

Criticism of earmarking rests mainly on a notion of government as a singlewill, or government decisionmaking as a perfect reflection of the wishes of thepopulation, under which decisions about spending and taxation are made suchas to improve welfare (even if implicitly). Analogously to utility maximizationby households, if choices about the allocation of public resources are madesuch that the last dollar spent yields the same additional benefits to societyregardless of what use it is put to (that is, such that the net marginal socialbenefits across all uses are equated), any constraints on choice would lead to alower level of welfare. But if one drops the notion of a single will and recog-nizes that political processes are imperfect and that societies consist of manygroups with differing preferences, benefit taxation and earmarking may takeon a more favorable coloring as a means of accommodating differences at apoint in time and over time. (In fact, if voters have identical preferences or ifthe same group of voters had the median preferences with respect to both howfunds are spent and the level of taxes-that is, the size of the budget-then theend results of financing from general funds and earmarking are the same.)If collective decisions are nothing more than the expression of individualchoices through constitutionally agreed rules, financing from general funds mayhave drawbacks. Spending and taxes are considered separately, with the levelof (planned) spending determined annually at budget time on the basis of pro-jected revenue from a tax system that has been implemented piecemeal over anumber of years. In this process, the only links between expenditures and tax-ation are the marginal adjustments in either that must be made to accommo-date macroeconomic stabilization.

These characteristics of financing from general funds, coupled with unstablemajority coalitions, mistrust among competing groups, or general tax weari-ness, may make it difficult to raise additional resources. Voters and legislatorswill be unwilling either to vote for more taxes without assurances about howthe money will be spent or to vote for higher spending without details abouthow the tax burden will be distributed. Both ]Buchanan (1963) and Goetz(1968) have shown that earmarking enforces a tie-in, which ensures that taxeswill be used for certain purposes and that the relationship will be stablethrough time. Thus earmarking can sometimes break through the impasse andallow money to be raised that otherwise might not have been forthcoming.


Another argument for earmarking has been that it facilitates cost recoveryfor goods or services for which charging a price is feasible. Cost recovery, theargument goes, makes more sense if the monies raised are retained for partic-ular public goods or services in the same sector of the economy. Beneficiarieswill be more willing to pay if they know their money will be used for activitiesthat directly benefit them. Officials who provide the,good or service will beencouraged to enforce the collection of fees if they know that their clients willbenefit or that more and better services will be provided. In this way, earmark-ing may improve both collection and perhaps even the way the money is used,since concerned users and officials are better monitors of performance than aremore distant authorities. (For example, a recent study of the Philippines-Jimenez, Paqueo, and de Vera 1988-showed that schools that rely more heavi-ly on local sources of finance have lower costs, holding enrollments and qual-ity-related variables constant.)

The arguments just outlined point to earmarking as a means of improvingthe performance of public sector institutions. They suggest too that under cer-tain circumstances, the practice may be a better instrument of resource alloca-tion than financing from general funds, or it may be a way around cumbersomeand inflexible general budgetary procedures. The arguments have some meritbut are not compelling because they do not give a clear indication of the cir-cumstances under which earmarking is justified or the criteria it would have tomeet to be considered satisfactory.

Efficiency and Automaticity in Pricing

Neither the theoretical nor the practical literature gives much guidanceabout what performance criteria earmarking should meet. The theoreticalarguments are confined to deriving circumstances under which earmarkingmight prove desirable, and they say little about how earmarking might be im-plemented or what might determine whether earmarking arrangements weresuitable or not. The practical literature evaluates the experience of earmarkingin various countries and economic sectors, using broad criteria in assessing thegrowth and mix of earmarked expenditures. On the basis of such qualitativecriteria, earmarking usually receives a mixed to negative rating.

Earmarking arrangements must meet two broad criteria. First, the price andtax arrangements must lead to appropriate allocation of resources for the sec-tor-that is, the financing must promote an economically efficient result.Second, these arrangements must lead to reasonably automatic financing forthe earmarked expenditures-that is, financing that is independent of develop-ments in the general budget. Earmarked funds that have to depend on frequentdecisions about prices and taxes in the executive or legislative branch (or onadditional funds from the general budget) are no longer insulated, and suchearmarking begins to resemble financing from general funds.

William McCleary 87

The traditional theory of earmarking calls for balanced budgets-chargespaid by beneficiaries (taxpayers) will be just sufficient to cover the (current andcapital) cost of providing the good or service. ]For example, in Samuelson's(1954 and 1955) general equilibrium model of the efficient provision of privateand social goods, all goods are provided under constant returns to scale (thatis, equal proportional increases in all inputs lead to an equal proportional risein output), the quantity of each social good consumed is the same for all tax-payers (by definition), and its output is expanded so long as the sum of themarginal benefits over all taxpayers exceeds the (constant) marginal cost of itsprovision.' The equilibrium solution requires that quantities of all public andprivate goods be determined, that taxes be levied on each taxpayer for eachsocial good according to the marginal benefit received, and that the sum of tax-es paid for each social good be equal to the total cost of supplying that good.Although there is no earmarking in the sense of segregated funds, earmarkingin fact takes place because every public expenditure comes with its own sourceof finance, and no public good would be provided without such financing. Withconstant returns, the budget for each public good would be balanced.

Analogous is the case of the public sector providing pure private goods un-der conditions of constant returns to scale-leaving aside for the moment ques-tions of economies of scale, indivisibilities (or those investments which addlarge, nonincremental amounts to output), externalities, and public sectorfinancial constraints. Given the level of demand for the good, the optimum out-put involves the equality of price with short-run and long-run marginal costs(and hence with short-run and long-run average costs). A price in excess ofshort-run marginal cost provides an incentive to increase production using vari-able factors of production; a price in excess of long-run marginal cost wouldlead to increased investment to expand capacity and lower (short-run) costs ofproduction. The optimally adjusted capacity, for a given demand, requires theequalities just listed and results in total revenue equaling total cost. Earmarking(and cost recovery) would in this case involve a balanced budget-but only ifcapacity had been adjusted to the optimum level2 and assuming constant re-turns to scale and no indivisibilities, externalities, or financial constraints. (Theresults would, however, still hold in the face of fluctuating demand. In the caseof so-called peak load pricing, users at nonpeak times pay a price equal toshort-run marginal cost, whereas users at peak times pay a price sufficientto cover the variable and fixed costs of peak use. Optimality thus involves abalanced budget; see Williamson 1966.)

Without the restrictive assumptions, economic optima and earmarkingwould entail unbalanced budgets. In other words, any earmarking rule that re-quired a balanced budget would be incompatible with optimality. The reasonsfor this are well known and need only be outlined here:

* Returns to scale. With increasing returns to scale, the long-run optimumoutput still requires the price to equal the short-run and long-run marginalcosts, both of which fall short of short-run and long-run average costs.


That is, optimum adjustment to any given level of demand will entail con-tinuous losses for an iindefinite period. With decreasing returns to scale,marginal costs exceed average costs, and adjustment to demand will entailexcess profits.

* Indivisibilities. With constant returns to scale but with lumpy investmentsthat add nonmarginal increments to output, the output at which price isequal to both short-run and long-run marginal costs-for any given levelof demand-may be impossible to attain, and the appropriate price willeither exceed or fall short of long-run marginal costs and entail continuousexcess profits or losses.

* Externalities. Optimization in the presence of externalities requires equal-ity between marginal social benefits and marginal social costs. If the goodor service in question is produced under constant returns to scale, externaleconomies (that is, with benefits also affecting nonusers) involve a higheroutput than would be generated by considerations of private demand andsupply, along with a subsidy to consumers to induce them to consume thelarger quantity. External diseconomies (that is, with costs imposed on per-sons other than users) in turn would involve a tax on output, which resultsin a lower equilibrium output than would be generated by considerationsof private demand and supply.

• Public sector financial constraints (pure taxation). To raise revenue to fi-nance general operations in the public sector, the government may imposeexcise and sales taxes (and may set prices of public enterprise products atabove marginal cost)--that is, financial constraints may dictate indirecttaxation, which drives a wedge between the price consumers pay and mar-ginal (private and social) costs. Any money raised by this indirect taxationbelongs to the general budget.

In the instances described above, earmarking, to be economically efficient,would require inequality between revenue and outlay for variable and fixed fac-tors of production-and hence some sort of relationship with the general bud-get (or perhaps with financial markets, if borrowing or lending is allowed).Quite simply, allowing levels of spending in these cases to be driven solely bythe level of earmarked revenue would lead to a misallocation of resources. Forexample, the excess profits generated at the optimum level of output in the caseof increasing costs are not a signal for increased investment since capacity isalready correct; the excess profits should either revert to the general budget or,at a minimum, be held in interest-bearing financial assets until increased realinvestment can be justified. Similarly, the losses dictated by considerations ofoptimality in the case of either decreasing costs or positive externalities wouldhave to be met from general funds. In theory, therefore, in many cases econom-ic efficiency would predicate an unbalanced budget; and in these cases it wouldbe undesirable to run a fund independent of the general budget.

Theoretical considerations aside, experience has shown that arrangementsfor automatic financing are difficult to devise. One obvious reason is that often

William McCleary 89

such funds also depend on additional, non-earmarked sources of revenue. Theamount of such monies will then depend on general budgetary consider-ations-such as the availability of resources and the comparative desirabilityof various expenditures-and the advantage of earmarking for the allocationof resources at the margin is no longer evident. Second, earmarked revenuesare easily eroded, especially during inflation. To the extent that any fund de-pends on a specific tax, an administered price, or an ad valorem tax linked toan administered price, it depends on periodic price adjustments by the govern-ment. By failing to adjust, the government can force the fund to become moredependent on financing from the general budget: or give itself the leeway toraise other taxes. Third, experience has shown that governments will simplyoverride earmarking arrangements if necessary (see the discussion of highwayslater in this article).

The cases of earmarking discussed so far have a large, if not total, overlapbetween the payers of prices and taxes and the beneficiaries of the expenditure.Several real-world cases do not meet this condition. For instance, the use of so-called sin taxes (on alcohol or tobacco) or lottery proceeds to finance the socialsectors and the setting aside of a fixed proportion of income taxes for specificpurposes seem to exhibit no such link. There is no connection between theamount of revenue raised and the appropriate level of expenditure or services;indeed, because the goods or services are provided to users below cost, thereis a tendency toward excess demand, which must be counteracted by nonpricedevices (such as queueing, congestion, and entrance examinations). In addition,since most of these earmarked funds are supplemented by funds from thegeneral budget, the purpose they serve is unclear. Earmarked sources are sup-plying an arbitrary amount of resources for the sector; the amount of addition-al resources coming from the general fund depends on the needs of the sectoras against those of other sectors. The main advantage cited by proponents ofthese taxes is that government can extract more resources from taxpayers bycommitting them to specific purposes. But given fungibility and given that ear-marking has little way of ensuring the appropriate level of output in such cases,such commitments appear to serve political ends rather than be rooted in soundcriteria of efficiency or distribution.

Examples of Experience with Earmarking

The World Bank's experience with earmarked funds for highways and forlocal government programs, and its experience in two countries where ear-marking has been widespread (Turkey and Colombia), illustrate and extend thearguments just propounded.

Highway Funds

Highways provide a particularly fruitful area for the study of earmarking.First, from the point of view of decisions about pricing and cost recovery,


certain characteristics of the sector make it interesting: for instance, financinghighways raises questions of returns to scale and indivisibility, external dis-economies in the form of congestion and pollution, and income-elastic demandfor fuel, which opens up the possibility of luxury taxation. Second, taxes onfuel and vehicles form a significant fraction of government revenue in bothdeveloped and developing countries, in both of which groups there are numer-ous examples of earmarking in the form of highway funds.

The World Bank's initial enthusiasm for earmarking for highway expendi-ture has cooled. Early Bank work called for earmarking in some countries ongrounds that more stable funding would encourage entrepreneurship and helpreduce unit costs of construction by speeding completion; in addition, somecrude statistical work purported to show a correlation between earmarking andthe proportion of investment devoted to highways (Eklund 1967). But by the1980s, Bank experience with road funds-described in an internal World Bankpolicy paper in 1985 as disappointing almost without exception-had led tothe conclusion that with some qualifications, earmarking ought to be avoided.It was conceded that, when mismanagement was rampant in the public sector,earmarking might be better than the alternative if (1) there were a public agen-cy with the demonstrated capacity to carry out the program; (2) funds wouldbe devoted entirely to a maintenance program for which economic prioritieshad been carefully articulated; (3) there were clear controls (before and afterthe project) to prevent the diversion of funds; and (4) the earmarking arrange-ment would be of finite duration. But even this qualified position is not unas-sailable. Despite the safeguards, it is hard to be sure that all maintenanceprograms will take precedence over construction projects or that, given fungi-bility, earmarking really does lead to an improved mix of highway expendi-tures.

Subsequent evaluations (Johansen 1989, McCleary and Tobon 1990) of theBank's experience in five countries-the Central African Republic, Colombia,Ghana, Mali, and Zaire--confirm the skepticism. Only in Ghana has earmark-ing had some success. There, the road fund appears to have ensured moretimely distribution of funds and sharp increases in allocations for maintenanceand rehabilitation. The deterioration of the road network during a long periodof economic mismanagement threatened the government's economic recoveryprogram, which strengthened the government's resolve to ensure adequatefunding. In the other four countries, road funds have- at least for a time-helped to get around cumbersome budgetary procedures, and the more reliablefunding has made it easier to use private contractors through competitive bid-ding. Generally, officials in agencies receiving earmarked funds support ear-marking arrangements; often such arrangements cut through red tape, allowmore timely receipt of funds, and permit better planning. But it does not followthat society also benefits from such arrangements unless priorities for spendingcan be set and adhered to. Otherwise, earmarking creates a new inefficiency to

William McCleary 91

correct an old one. The quest for larger allocations and greater stability offunding appears to have been more elusive.

The lessons of Bank experience have been that road funds are not the easyor obvious solutions to underfunding that they may appear to be. The surveyshave found, among other things, that:

* It is hard to set appropriate prices for road funds; specific taxes or advalorem taxes linked to administered fuel prices quickly erode road fundsin an inflationary environment. A pricing arrangement that will generateadequate resources for highways generally requires periodic decisions byofficials.

* Allocations for road funds appear to depend on the condition of the gen-eral budget. When money is tight, earmarked funds may be temporarilyfrozen (as happened in Ghana) or diverted to other uses (as in Mali), orthe government and public enterprises may stop paying their fuel bills andhence fuel taxes (as they did in Zaire).

* Road funds are more vulnerable to general budgetary problems the morethey depend on periodic changes in fuel prices or tax rates and (obviously)the smaller the share of earmarked taxes in the road funds' resources.

* The adequacy of road funds' resources as a whole is no guarantee of anappropriate mix of maintenance, rehabilitation, and new investment.

The purported insulation of road funds from general budgetary problems isthus largely illusory, since there are many ways governments can tap into ear-marked monies when budget conditions get tight. (The experience in theUnited States bears this out; see Short 1989.) Only when a government is com-mitted to relatively high spending on roads (as in Ghana) will the funds besafe-and financing from general funds could have brought about much thesame outcome.

Earmarking by Local Governments

The prospects for earmarking seem brighter in the context of decentralizedlocal government. Beneficiaries are easier to identify, and the preferences of vot-ers and users are easier to articulate. Because the benefits are localized, thereis less opportunity to pass the cost of programs to nonbeneficiaries. Thus, atthe local level, earmarking occurs in the guise of special assessments (specificexpenditures linked to changes in tax rates or betterment and valorization tax-es); special-purpose governments (such as special districts for water and sew-erage or for schools financed from fees or property taxes); and various servicesprovided for a charge (for example, local transport, garbage removal).

One point is worth stressing. The link between expenditure and revenue inthese arrangements is extremely close. In fact, often the level of governmentactivity or service is proposed first and the financing arrangement decided af-terward. This substantially reduces the danger of a mismatch between appro-


priate levels of revenue and activity and the corresponding danger of over- orunderinvestment, which may occur when spending is linked to a source of rev-enue whose level is set more or less arbitrarily; highway expenditures and fueltaxes are an example of such a linked pair. These advantages may, however, bereaped at the expense of the prime virtues of earmarking: automaticity and in-sulation from general budgetary considerations. The more often officials mustmake decisions about appropriate levels of expenditure and revenue, the morethey and the voters will be forced to choose among competing uses for thefunds, and the less certain financing becomes.

Colombia's municipal valorization tax (similar to what are called specialassessments or betterment taxes in other countries) represents a rather pureform of earmarking at the local government level (see Rhoads and Bird 1967,World Bank 1977, and Bird 1984). This purity results from:

1. the substantial overlap of taxpayers with beneficiaries

2. the use of benefit-cost considerations (analysis would be too strong aword) and the conscious and conscientious attempt to assess taxesaccording to how people benefit from the project

3. the strong link between valorization revenue and expenditure (inprinciple, no supplement from or diversion to the general budget).

In addition, valorization appears to reverse the usual sequence of earmark-ing, in which available revenue drives levels of expenditure: instead, the iden-tification of desirable projects, often within the scope of prepared investmentprograms, stimulates efforts to raise financing from prospective beneficiaries.

Once a project suitable for valorization taxation is selected, a zone of influ-ence-the area over which benefits are expected to be felt-is demarcated.Benefits are estimated as the total resulting rise of site values and allocatedacross the properties within the zone according to formulas that take intoaccount characteristics such as size, shape, topography, frontage, distance fromthe project, and various economic factors. Over time, the scope of valorizationhas been extended to include the construction of streets and parking facilities,local paving, urban development, and the establishment of green spaces. Inprinciple, there is nothing to stop authorities from attempting to recapturesome or all of the (net) benefits from the project, but in practice full recoveryis sought only for the costs of land, construction, and some interest and ad-ministrative expenses. Problems that the poor or people with few liquid assetsmight have in paying the tax are taken into account: the poorest can be ex-empted altogether, and others are allowed to stretch payment over several yearsif it would otherwise absorb a significant fraction of their annual income.

The tax has some obviously desirable features-the connection between ben-eficiaries and taxpayers; the effort to select projects in which the gains clearlyexceed the costs; and the flexibility with which it has been adapted to differingcircumstances. As a tax on the unimproved value of land, it has no adverseeffect on incentives; indeed, the combination of the tax and improvement serves

William McCleary 93

to raise income and increase the potential profitability of investments, whichstimulate incentives. The tax has raised sizable sums and financed significantimprovements in Colombia's cities.

But valorization does have some drawbacks, primarily having to do withcollection. First, quite aside from the fact that authorities attempt to recoveronly the cost of the project, collections fall far short of 100 percent. The rea-sons appear to be cost overruns, generous exemptions (for example, theCatholic church, charitable institutions, and public enterprises), and generousaccommodations for the poor and those with liquidity problems. Difficultieshave been greatest for projects designed to improve conditions in low-incomeareas. Second, the concept of the zone of influence has its limitations: in casesin which many of the beneficiaries fall outside the zone, it has been difficult tocollect because people within the zone have been reluctant to pick up the addedburden. Collection is also problematic when several projects simultaneously af-fect a number of overlapping zones. Third, partly because of these collectionproblems and partly because of administrative deficiencies in planning and im-plementing projects, growth of valorization has proved unstable, characterizedby periods of stagnation followed by sizable increases.

Despite its imperfections, valorization deserves a favorable rating. Its popu-larity and its ability to raise revenue justify its continuation and extension. Itsweaknesses could be reduced by cutting back exemptions, charging interest(this was in fact introduced in 1981), and imposing penalties on arrears.Furthermore, if many people cannot pay and a key objective is to subsidizethem, experience has shown that such objectives are served better by explicitfinancing from the general budget than within the context of the valorizationor betterment taxes.

Earmarking in Turkey and Colombia

Turkey and Colombia are prolific earmarkers. In both countries, earmarkinghas gotten out of hand, bringing significant misallocation of resources. The ear-marking policies of the two countries differ, however, in several respects. InColombia earmarking generates a much larger share of government revenueand is of longer standing than in Turkey, where earmarking is a recent responseto budgetary stringency. Turkey's earmarking is mainly off budget, whereasColombia's is almost entirely on budget.

TURKEY. Some of Turkey's extrabudgetary funds are quite old but most are ofrecent origin, mainly a response to the budgetary problems of the 1980s. As aconsequence, extrabudgetary funds have grown, according to World Bank data,from the equivalent of 8 percent of central government revenue (1.4 percent ofgross national product, GNP) in 1983 to more than 20 percent of governmentrevenue (more than 4 percent of GNP) in 1987-this in a country where central


government revenue as a share of GNP, including extrabudgetary funds, hasbeen falling, from 21 percent in 1980 to 18 percent in 1987.

The purposes and sources of funding for the principal extrabudgetary fundsare varied (see table 2), ranging from quite specific-support for investment inthe defense industry, housing subsidy, fertilizer subsidy-to more general-export and investment subsidies, promotion of exports through marketing andadvertising-to even more general-an extrabudgetary fund to raise moneyfor other extrabudgetary funds. For example, the biggest fund, the PublicParticipation Fund, which builds and sells participation in large infrastructuralprojects, receives money from the petroleum consumption tax and can borrowdomestically. The second biggest, the Mass Housing Fund, provides credit forsmall housing units and land acquisition and receives the bulk of its financingfrom earmarked taxes-from variable import levies, the supplementary valueadded tax, petroleum product taxes, and the $100 exit tax on Turkish citizens.Revenue for the extrabudgetary funds is obtained from tax (68 percent) andfrom nontax (32 percent) sources. Some funds have authority to borrow, butthe amounts involved have so far been quite small. Principal tax sources havebeen import duties (18 percent of total revenue), export taxes (13 percent),petroleum taxes (20 percent), and financial transactions (10 percent), followedby excises on tobacco, alcohol, and beverages. Nontax revenue has includedthe proceeds of lotteries and gambling houses, various fees and penalties, op-erating income from infrastructural facilities, and interest income on financialassets.

In general, extrabudgetary funds are established by decree following approv-al by the Council of Ministers. The decree specifies the purposes, eligible ex-penditures, and the sources of revenue and rates of tax, surcharge, or subsidyinvolved. In some instances the decree is so specific that the Council of Minis-ters determines the charter, policies, operating framework, and budget of theextrabudgetary fund, which has no separate management board and is merelyan account held at the central bank. In other instances, the fund has a separateboard and management body. No central authority supervises the extrabudget-ary funds.

Turkish earmarking has some serious drawbacks. First, it contributes littleto government efforts to raise more resources. The rise of the extrabudgetaryfunds has been sapping central government revenue and expenditure by pre-empting an increasing share of public revenue. For example, import levies onparticular products reduce imports and the amount of import duties flowing tothe central budget from them. Other special indirect taxes can reduce budget-ary revenue by reducing demand or profit.

Second, because there is no central organization, the policies of some ex-trabudgetary funds run counter to those of others or to those of the govern-ment. For example, reliance on export taxes and import duties is contrary tothe government's goal of a more open, export-oriented trade regime. Specialimport levies obstruct the move to reduce quantitative restrictions and tariffs,

William McCleary 95

Table 2. Turkey's Principal Extrabudgetary Funds, 1985-86

ProjectedDate Revenue, revenue, Principal sources

Fund established 1985 1986 of revenue Principal uses

Defense Industry 1985 n.a. 194.6 Indirect: tax on Support to investmentSupport Fund fuel, tobacco; direct in the domestic defense

tax on corporate and industrypersonal income;nontax revenue

Development and 1984 62.2 113.1 Tax on luxury Price support ofSupport Fund (DSF) imports and transfers animal feedstock and

from other funds social programs

EEC Funda 1982 1 3.0b 0 Tax on imports of Transfer to otheriron, steel, and some funds (mainly DSF)chemicals fromthe ECa

Export Encouragement 1984 22.9 3 1 .0 c Tax on imports and Support to privateFund exports and nontax investment and

revenue transfer to other funds(mainly RUSF and DSF)

Export Improvement 1980 0.6 0 .8c Contributions from Promotion of exportsFund private industry through marketing and

advertising

Financing Fund 1980 22.4 30.0c Deferred tax on Support to privatecorporate income investment

Investment Goods 1985 n.a. - Foreign credits and Credit to firmsManufacturing transfers from other that manufactureEncouragement Fundd funds investment goods and

to their importer-clients

Mass Housing Fund 1984 212.4 315.7 Tax on luxury Subsidized housing(MHF) imports; share of credit

supplementary valueadded tax andtransfers from otherfunds

Mutual Assistance 1986 n.a. - Transfers from Income transfer toand Support Fund budget and other poor(Poor People's Fund) funds; surcharge on

corporate andpersonal income tax;nontax revenue

Continued on page 97

and they introduce the potential for arbitrary, uneven protection across variousparts of the economy. At the same time as the government is encouraging pri-vate initiatives and giving incentives to invest, one extrabudgetary fund isdrawing on taxes on bank loans to the private sector to subsidize exports andinvestments. Until recently, one fund was taxing agricultural exports while an-other was subsidizing them at the same rate.

Third, the administration of extrabudgetary funds in Turkey strays far fromthe benefit principle. In virtually no case are the beneficiaries of governmentexpenditure financing the spending themselves (the use of part of the taxes on


Table 2 (continued)

ProjectedDate Revenue, revenue, Principal sources


Petroleum 1984 52.0 108.8 Tax on fuel FinancingConsumption consumption investments of local

government andhighway and sportdirectories

Petroleum Exploration 1973 7 5 .3b 90.0 Tax on domestic Transfer to otherFund petroleum funds (mainly SPSF)

Public Participation 1984 228.7 421.7 User charges on Public investment inFund public infrastructure; infrastructure

issuance of revenue-sharing certificates;share of petroleumconsumption tax

Petroleum Price 1979 6 5 .4 b 88 .3c Tax on imported Transfer to otherStabilization Fund petroleum funds (mainly MHF)

Resource Utilization 1984 112.3 172.2 Tax on bank loans Export and investmentSupport Fund (RUSF) subsidy

Selective Credit Fund 1970 3 5 .0e 5.0 Central government Export and investmentconsolidated budget subsidy (now through

transfer to RUSF)

Support and Price 1980 267.7 280.0 Tax on imports and Fertilizer subsidyStabilization Fund agricultural exports(SPSF)

Tax Administration 1985 n.a. 50.0 Central government Improvement in taxDevelopment Fund consolidated budget administration

Total 1,134.9 1,901.2 n.a. n.a.Total (net of interfund

transfers) 965.0 1,700.0 n.a. n.a.

- Not available.n.a. Not applicable.a. Taxes on imports from the European Communities were abolished in July 1986.b. Data are provisional.c. These are estimates based on an increase of 35 percent over 1985.d. As of July 1986 this fund had not yet started functioning.e. Figure refers to 1984 and is not included in the total.Source: Estimates based on World Bank data.

petroleum to finance highways is a minor exception). Earmarking instead rep-resents the arbitrary assignment of revenue from certain taxes for financingsupposedly desirable government programs. Under these circumstances, reve-nue gives no guidance as to what level of the service is desirable. Finally, andrelatedly, it is not at all clear what criteria for project evaluation the extrabud-getary funds must meet and how rigorous these are compared with those forregular budgetary expenditures. In sum, given the size of the extrabudgetaryfunds and the lack of coordination and control, there is a potential for a sig-nificant misallocation of resources.

William McCleary 97

COLOMBIA. Most of Colombia's earmarking, unlike Turkey's, comes from old,established funds created at least two decades ago (see table 3; this sectiondraws on McCleary and Tobon 1990). Earmarking has increased since thenmainly as a consequence of revenue sharing and rather buoyant sources of rev-enue for some of the major funds. Colombia's earmarking is a product of its

Table 3. Colombia's Principal Earmarked Funds(millions of Colombian pesos)

RevenueDate Principal sources


National funds

Highway Funds 1967 46,493 59,600 Fuel taxes Transportation invest-(national and other) ment: 75% for national

highways, 10% forneighborhood roads, 10%for national railways, and5% for urban transport

PROEXPO 1967 27,038 43,200 Levy on imports Subsidized credit forexports

National Coffee 1940 17,886 14,005 Taxes on value Economic and socialFund of coffee exports investments in coffee-

and retention growing regions; variouson volumne equity and financial

investments; coffee pricestabilization

National Apprentice- 19S7 18,256 23,195 Payroll taxes Technical trainingship Service

Family Welfare 1968 17,831 23,534 Payroll taxes Programs for protectionInstitute of children and family

Departmental funds

Sectional Health - Taxes on beer, Health programsServices { 78,461 - wine, liquor

Sectional Sports - Taxes on cigarettes Support of sportsCommissions J activities

Municipal funds

Municipal Improve- 1921 6,967 - Tax on increases Investment in roads,ments in unimproved water, sewers, parks, and

value of land so on(valorization tax)

Intergovernmental transfers

Departments and 1971 75,838 109,862 Share of total Education and healthTerritories national revenue programs

minus revenueearmarked forother purposes

Municipalities 1968 55,590 80,432 Share of value Specified municipaladded tax revenue investments

- Not available.Source: McCleary and Tobon 1990.


political history-a long struggle over the degree of centralization and signifi-cant periods of political tension and violence. Earmarking has become a mech-anism for committing government to continue specific activities regardless ofan unsettled political climate and for getting around elaborate budgetary con-trols designed to prevent misuse of funds. Earmarking in the form of the auto-matic sharing of specified sources of central government revenue is an attemptto reduce the extent of centralization and to support increased activity on thepart of departments (provinces) and, more recently, municipalities.

Earmarking has grown rapidly during the past two decades. The proportionof national government revenue earmarked rose from 11 percent in 1970 to28 percent in 1986. The proportion of revenue earmarked at all levels of gov-ernment rose from 21 percent to 32 percent over the same period. Whereas de-partmental was as important as national earmarking in 1970, a surge occurredin earmarking of central government revenue, which now accounts for 75 per-cent of earmarking at all levels of government. The largest sources of earmark-ing at present are revenue sharing for education and health and for specifiedmunicipal investments (35 percent of all earmarked revenue); departmental sintaxes (on alcohol, tobacco, and gambling) for health and sports (21 percent);the fuel tax for spending on highways and other transportation projects(12 percent); payroll taxes to finance training and programs to promote familywelfare (10 percent); the import levy to finance subsidized credit for exports(9 percent); various levies on coffee exports to cover economic and social in-vestments in coffee-growing regions and other investments (5 percent); and themunicipal valorization tax (2 percent). These seven sources account for about94 percent of earmarking in Colombia.

Colombia's earmarking differs from Turkey's in several respects other thanits origins and significance in public resources: first, Colombia's earmarkedfunds are largely within existing budgets so that their revenues and expendi-tures do not so often obscure government activities or the government's effecton the economy; second, the system relies less heavily on import duties andother taxes on international transactions as a source of financing; and third, agreater (though still fairly small) proportion of earmarking would meet the testof being benefit related (for example, the fuel tax, municipal valorization tax,and, more loosely, the coffee tax).

That earmarking has become excessive has long been recognized inColombia, but there has been little success in reducing its scope. The 1981Mission on Intergovernmental Finance (headed by Richard Bird and EduardoWiesner) took a generally skeptical but pragmatic view of earmarking, attrib-uting its origin primarily to Colombia's complex political and budgetary situ-ation (see Bird 1984). The mission recommended evaluating earmarking caseby case, eliminating earmarking that did not fulfill the benefit principle or wasnot a primary source of finance. The Commission on Public Expenditure (1986)took a much harsher view, accepting the Bird-Wiesner view that earmarkingought to be limited to cases that meet the criteria of the benefit principle but

William McCleary 99

finding that principle absent in virtually all cases. This commission thereforeconcluded that the inefficiency of Colombia's budgetary processes was not suf-ficient justification for earmarking if there was no assurance that the appropri-ate amount of resources would be devoted to earmarked activities-if oneinefficient practice replaced another, there was no reason to expect an improve-ment in national welfare. The commission recommended that earmarkingought to be abolished. (The commission's definition did not include revenuesharing, payroll taxes for social security, and profits from public enterprises, allof which could be considered earmarking under a broader definition.) So farthe government has done little to implement these recommendations; in fact, afew small earmarked funds have been added.

Colombia's earmarking is too complex for blanket recommendations.Nevertheless, recent budgetary reform in Colombia makes reductions in thescope of earmarking more feasible politically than in the past, and some piece-meal reforms, if adopted fully, would eliminate about 45 percent of traditionalearmarking (in value terms). Remaining would be revenue sharing, three majorcases of earmarking, and perhaps some minor flnds. The objective would beto eliminate the most obvious cases of inefficiency.

The suggested reforms would apply, first, to those funds whose revenue andexpenditure do not fit the benefit principle of taxation. These funds include im-port duties for subsidized export credit (PROEXPO); payroll taxes for varioussocial welfare objectives (such as the National Apprenticeship Service and theFamily Welfare Institute); departmental taxes on and profits from alcohol,tobacco, and gambling used for health, welfare, and sports; and a number ofminor funds. Second, the remaining major funds could be adjusted to movethem closer to the benefit principle. For instance, spending from the NationalCoffee Fund could be limited to the coffee sector and coffee-growing areas, fueltaxes could be adjusted more often to generate the resources required for high-ways, and the valorization tax could be extended to additional Colombian mu-nicipalities. Third, the government could increase the effectiveness of revenuesharing, an important element in its effort to decentralize activities to the de-partmental and municipal levels, by changing the formulas to remove anoma-lies (such as the undue favoritism given to areas with small populations),increasing incentives to generate resources at the local level, and taking mea-sures to increase the capacity of municipalities to absorb larger resources.

Conclusions and Recommendations

However much the International Monetary Fund and the World Bank havecondemned it in the past, there is no general presumption against earmarking.In theory, as Buchanan, Goetz, and other public-choice economists haveshown, earmarking may make agreement possible on increasing revenue andexpenditure when there would be no consensus about either separately. It may


protect high-priority programs from shifting majorities, administrative ineffi-ciency, and corruption (although experience shows that often apparentlylow-priority programs also get protected). There are examples of successfulearmarking, clustered mainly in the areas of revenue sharing, social security,special assessments (betterment or valorization taxes), and special districts(for example, for water and sewerage and for education). Earmarking in otherareas-requiring specific taxes to finance narrowly defined governmentprograms (such as highwaoys, housing, and training)-has been notably lesssuccessful.

The setting of taxes and prices under earmarking should meet the same testsof efficiency and equity as the setting of prices for public sector goods in gen-eral, and it should provide a reasonably automatic source of financing, inde-pendent of general budgetary considerations. Meeting these two criteriasimultaneously is difficult. Efficient pricing and taxation is likely to dictate un-balanced budgets for the earmarked fund and hence an interdependence withthe general budget. In practice, governments generally override earmarkingarrangements if they need the resources; the protection given by earmarkingtends to evaporate when budgets get tight.

Earmarking works best when it is an extension of the benefit principle oftaxation. But since the objectives of efficiency and equity under benefit taxationcan usually be met through financing from general funds, earmarking needs tobe justified on other grounds-for instance, that it is needed to protect levelsof the service, increase the collection of revenue, or improve the quality of theservice. By the same reasoning, earmarking is not desirable for programs inwhich redistribution or social welfare is emphasized because the connectionsbetween revenue and appropriate levels of activity are tenuous. Nor does ear-marking work well in the case of pure public goods, since taxpayers will notreveal their preferences through voluntary individual payments.

The World Bank's long-standing skepticism about highway funds is amplyjustified by experience. Highway officials like them because they simplify bud-getary procedures and appear to make funding more certain, but experienceshows that the financing they provide is not automatic and remains dependenton government judgments, which change as budgetary considerations dictate.

Earmarking appears to work more successfully in the context of local gov-ernment, where the correspondence of beneficiaries and taxpayers is closer andwhere users can more easily express their preferences by voting. Periodic votesor assessments link expenditure and revenue for specific activities and providea check on the appropriateness of decisions.

The scope of earmarking is restricted, at any level of government, by thelimited number of cases to which the benefit principle applies; the costs of frac-tionalization of decisionmaking; and possible tradeoffs with revenue mobiliza-tion for the general budget. In countries such as Turkey and Colombia, whereearmarking has become extensive, in many cases the practice is not justifiedand ought to be abolished.

William McCleary 101

In sum, earmarking cannot be discarded out of hand, but there are certainlygrounds for skepticism about how well it is likely to work. If past experienceis any guide, any proposed scheme of earmarking carries its own set of poten-tial problems even as it shelters one type of government expenditure from com-petition for funding with another. Questions arise about the adequacy ofresources to meet the needs of a given endeavor, the adequacy of the institutiondesignated to carry out the earmarked activity, the lack of control over priori-ties for spending or administrative outlays, and the possible conflicts with thegovernment's ability to raise resources for the general budget or with other gov-ernment policies. The clear implication is that each earmarking scheme shouldbe required to satisfy a series of tests to ensure that it does represent an im-provement.

* Is there a strong link between beneficiaries and the payers of taxes andprices?

* Is earmarking necessary, in addition to benefit taxation or charges, to en-sure levels of service or to improve the collection of revenue or the qualityof service?

* Will the arrangements for pricing, taxing, and other financing for the ear-marked expenditure elicit resources adequate for present and expected lev-els of demand?

X Will the price or tax arrangements have significant distortionary effects-deadweight losses, inflationary effects-on the allocation of resources?

X Is there an appropriate investment program and a clear set of rules to reg-ulate investment decisions, the mix of spending on capital, maintenanceand rehabilitation, and administrative overhead?

* Are there accounting and auditing safeguards against the misuse or diver-sion of funds?

• Are the expenditure program and its financing consistent with the govern-ment's policies on macroeconomics and the allocation of resources; better,is there a government agency that oversees extrabudgetary funds and en-sures that their activities are consistent with policy?Is there an agency (or prospective agency) with the capacity to plan, eval-uate, and carry out the program?

* Is there a cutoff date for deciding whether the earmarking arrangementsshould be continued?

Satisfying each of these questions would constitute a requirement formidableenough to limit the scope of earmarking quite stringently in practice.

NotesWilliam McCleary is the lead economist for the Pakistan and Turkey Department in the

Europe, Middle East, and North Africa Regional Office of the World Bank.1. This, of course, abstracts from the problem of free riders. If all public goods must be con-

sumed in equal quantities by all, and one person's participation in the benefits of any public


good does not affect the benefits received by others, there are no incentives for taxpayers to re-veal their preferences, especially when large numbers are involved. They benefit whether theypay or not.

2. All short-term equilibria (with short-run marginal cost unequal to long-run marginal cost)involve either excess profits or losses.

ReferencesThe word "processed" describes works that are reproduced from typescript by mimeograph,

xerography, or similar means; such works may not be cataloged or commonly available throughlibraries, or may be subject to restricted circulation.

Bird, Richard M. 1978. "A New Look at Benefits Taxation." In H. C. Recktenwald, ed.,Tendences a Long Terme du Sector Public: Secular Trends of the Public Sector. Paris: EditionsCujas.

. 1982. "Budgeting and Expenditure Control in Colombia." Public Budgeting and Finance2, no. 3 (Autumn): 87-99.

. 1984. International Finance in Colombia: Final Report of the Mission on Intergovern-mental Finance. Cambridge, Mass.: Harvard University, Law School, International Tax Pro-gram.

Buchanan, James M. 1963. "The Economics of Earmarked Taxes." Journal of Political Economy71 (October): 457-69.

. 1968. The Demand and Supply of Public Goods. Chicago: Rand McNally.

Commission on Public Expenditure. 1986. Informe Final. Bogota.

Deran, Elizabeth Y. 1965. "Earmarking and Expenditures: A Survey and a New Test." NationalTax Journal (December): 354-61.

Eklund, Per. 1967. "Earmarking of Taxes for Highways in Developing Countries." EconomicsDepartment Working Paper 1. World Bank, Washington, D.C. Processed.

Goetz, Charles J. 1968. "Earmarked Taxes and the Majority Rule Budget Process." AmericanEconomic Review (March): 128-36.

Jimenez, Emmanuel, Vicente Paqueo, and Maria Lourdes de Vera. 1988. "Does Local FinancingMake Primary Schools More Efficient? The Philippines Case." Policy, Planning, and ResearchWorking Paper 69. World Bank, Policy, Research, and External Affairs Department, Washing-ton, D.C. Processed.

Johansen, Frida, ed. 1989. "Earmarking, Road Funds and Toll Roads: A World Bank Sympo-sium." Infrastructure and Urban Development Department Discussion Paper 45. World Bank,Washington, D.C. Processed.

Lindahl, Erik. 1958. "Just Taxation: A Positive Solution" and "Some Controversial Questions inthe Theory of Taxation." In Richard A. Musgrave and Alan T. Peacock, eds., Classics in theTheory of Public Finance. London: Macmillan.

McCleary, William A., and E. LJribe Tobon. 1990. "The Earmarking of Government Revenuesin Colombia." Policy, Research, and External Affairs Working Paper 425. World Bank,Washington, D.C.

Musgrave, Richard A. 1959. The Theory of Public Finance. New York: McGraw-Hill.

Rhoads, William G., and Richard M. Bird. 1967. "Financing Urbanization by Benefit Taxation."Land Economics 43 (November): 403-12.

Samuelson, Paul A. 1954. "The Pure Theory of Public Expenditures." Review of Economics andStatistics (November): 387-89.

. 1955. "A Diagrammatic Exposition of a Theory of Public Expenditure." Review ofEconomics and Statistics (November): 350-66.

Short, J. W. 1989. "Earmarking for Transportation: A View of the U.S. Experience." In Johansen.

William McCleary 103

Wicksell, Knut. 1958. "A New Principle of Just Taxation." In Richard A. Musgrave and AlanT. Peacock, eds., Classics in the Theory of Public Finance. London: Macmillan.

Williamson, Oliver E. 1966. "Peak Load Pricing and Optimal Capacity under IndivisibilityConstraints." American Economic Review (September): 810-27.

World Bank. 1977. Valorization Charges as a Method For Financing Urban Public Works: TheExample of Bogota, Colombia. Staff Working Paper 254. Washington, D.C.


SOCIAL SECURITY AND THE POORChoices for Developing Countries

Ehtisham Ahmad

There is an urgent need to provide an effective safety net for the poorest insocieties ranging from socialist countries undergoing reform to Sub-SaharanAfrican economies. This article examines social security systems in industrialcountries and explores their relevance to developing countries. The objectiveof social security is defined broadly as public action, including that by commu-nities, to protect the poor and vulnerable from adverse changes in living stan-dards. Relevant instruments include employment and income guarantees, andalso such formal policy instruments as assistance, social insurance, and familyallowances. The article highlights issues that arise in providing social securityin developing countries, particularly its effectiveness in protecting the targetgroups.

T ihe objective of social security is to protect the poor and vulnerable andto ensure that they have an acceptable standard of living. Social securitymay also entail smoothing consumption and reducing risk or spreading

income over the life cycle. Often there is a redistribution of income amonggroups with differing needs.

In industrial countries social security is associated with such programs aspensions, unemployment, and health care, but these associations are not par-ticularly helpful in assessing its role in developing countries. Experience is rel-evant to the extent that it provides a history of the preconditions and effectsof some of the social security instruments that have evolved during this century.And it shows that policymakers have tried very different combinations of pol-icies, which reflect different perceptions of need and vulnerability. The relevantpolicy will depend on identifying vulnerable groups as well as administrative

The World Bank Research Observer, vol. 6, no. 1 (January 1991), pp. 105-127(© 1991 The International Bank for Reconstruction and Development/THE WORLD BANK 105

and political constraints. Public policy is not synonymous with governmentprovision, since there is clearly a role for local authorities as well as communityaction to help target scarce resources.

The first part of the article examines variations in social security programsin industrial nations and attempts to determine their effects. Comprehensivesocial security schemes may slow population growth, for example, but theeffects are long term and quite complex. Similarly, the effect on savings hasbeen the subject of considerable debate (see Atkinson 1987). The inconclusive-ness of this analysis-even with good data for developed countries and exten-sive models-suggests that it will be difficult to assess the effect of suchprograms in developing countries. But social security is not a preserve of therich, and the governments of developing countries, in designing such programs,need to bear in mind the likely incentive effect.

Ahmad and others (1990) argue that social security can be provided even inlow-income countries and that it is as indispensable to an equitable develop-ment strategy as to economic growth. The second part of the article examinestraditional social security mechanisms in developing countries and considersthe role of the state in augmenting and formalizing these indigenous efforts.Although there are extensive formal programs in many Latin American coun-tries, the cost of providing benefits is beginning to impose an intolerable burdenon already stretched budgets (Mesa-Lago 1990). In other countries, however,ranging from Brazil to China and India, aging populations, growing urbaniza-tion, and the rising number of nuclear families haove weakened the traditionalsupport provided by the family and increased the need for formal provision.Before governments expand their programs, however, it is important to re-appraise the objectives of formal social security schemes and evaluate the mostcost-effective forms of public intervention.

The third part of the article assesses the scope for formal social security indeveloping countries and examines whether these programs should be publicor private. A number of instruments are likely to be needed, but the appropri-ate mix will differ from one developing country to another and will probablybe substantially different from the programs in any developed country (seeAtkinson and Hills 1990).

The Evolution of Social Security in Industirial Countries

A study on social security (ILO 1984) describes three stages in its modernevolution. According to this view, the initial response was paternalistic: privatecharity and poor relief were provided for the indigent, but harsh conditions andstigma made this form of provision politically unacceptable. As a reaction, inthe second phase, social insurance schemes were developed based on compul-sory premia that entitled the participants to pensions and sick pay. In time,these programs were expanded to include wider coverage and more contingen-


cies. In the third stage, the concepts of prevention and universality were intro-duced "with the aim of maintaining and enhancing the quality of life"(ILO 1984, p. 17). Until the turn of the century, charity remained the mainstayfor the poor in Europe, and in England charity was largely organized by thechurch. In 1834 the workhouse system was established under the Poor LawAmendment Act, an early example of a self-targeted provision (that is, a mea-sure that automatically ensures that only eligible recipients will receive bene-fits). But the stigma attached to recipients excluded many of the needy, and intime, both the poor laws and charitable support of the disadvantaged came tobe seen as inadequate and demeaning. Beveridge (1942), the architect ofthe modern welfare state, notes that "since the beginning of the presentcentury ... there has been a strong movement against the form and spirit of theold poor law" (p. 211).

The identification of age and infirmity as causes of poverty and deprivationled to the development of social insurance as an important policy instrumentto guarantee minimum living standards and replace income in times of adver-sity. This insurance followed the precedents established by certain trades thathad set up compulsory programs to provide pensions and sick pay for employ-ees in the public sector and in some private sector occupations. In time, thesecompulsory programs were expanded to include more occupations and a widerrange of contingencies. (In some countries the occupational origin of social se-curity is still retained in the form of separate funds for various occupations.)

The evolution of different sets of policies that make up formal social securityin industrial countries was a result of changing perceptions about who was vul-nerable. These policies were also a political response to the need for stability,as was Bismarck's introduction of social insurance in Germany. Different con-figurations of interest groups influenced the direction of social security policiesin industrial countries.

In Britain, for example, an old-age pensions act was introduced in 1908. TheNational Insurance Act followed in 1911. This legislation, which correspondedto Bismarck's reforms in Germany, provided pensions and sick pay to all work-ers. The law implied a fundamental change in philosophy, since it replacedmeans-tested benefits and assistance with social insurance. But not until afterthe Great Depression, in 1934, did Britain pass an unemployment act to providegeneral assistance for all unemployed workers. And the public response to bat-tlefield deaths after the outbreak of World War II prompted the introductionof means-tested supplementary provisions for widows.

The Beveridge proposal was to make national insurance coverage universaland reduce dependence on assistance, given "the strength of popular objectionto any kind of means test" (Beveridge 1942, p. 12). As a precondition for thesuccess of the reform, Beveridge argued for the inclusion of "family allowanc-es," which were neither social insurance nor means-tested benefits and whichincluded basic benefits for children. Although not all of the Beveridge proposalswere accepted, the main idea was adopted, and the government projected an

Ebtisham Abmad 107

82 percent increase in social security as a share of total assistance programs by1975. The share of family allowances would account for about 12 percent ofthe total, largely at the expense of means-tested assistance (which made up 100percent of public welfare at the turn of the century). But although the govern-ment increased the scope of social insurance after 1949, there was a decline inthe share of basic benefits over time and a rise in means-tested assistance forbasic benefits from 1.1 million payments a week in 1949 to 2.7 million in1969-70 (Atkinson 1989). Since 1979 there has been a marked decline in thereal level of basic benefits as well as in national insurance. Means-tested assis-tance, which accounted for about 10 percent of social security expenditures in1969-70, increased to more than 30 percent by 1985, which reflects stringentnew measures to determine eligibility, including more rigorous screening, moredetailed evidence of income, and more reliance on established tests. This expe-rience suggests that the stages predicted by ILO (1984) are not inevitable. In-stead of moving to the third phase, Britain, at least, appears to be regressing.

Formal social security policies differ markedly among industrial countries,depending on the objectives of the government. Because most systems involvecombinations of measures, and because there are often substantial interactionsthat determine whether the basic objective of protecting the vulnerable hasbeen achieved, it can be misleading to evaluate the overall effect by focusingon one program. The essential test of a social security system is how well itpays benefits to the intended target groups. Thus in the case of income poverty,it is necessary to examine to what degree social security provisions contributeto a reduction in poverty. A failure to meet the objective suggests inadequaciesin design. For instance, when benefits are based on a set of contingencies, cov-erage may not include all the eligible poor. Thus in countries of the Organisa-tion for Economic Co-operation and Development (OECD), where socialsecurity was first developed to fight poverty among the aged and disabled, therehas been a growing concern about the incidence of poverty in families withchildren (see Burtless 1986 for a discussion on the United States).

The Luxembourg Income Study suggests that 24 percent of elderlyAmericans were poor in 1979. This figure fell to 12 percent in 1987 as a directresult of the growth in social security retirement benefits in the United States.The proportion of children living in poverty, however, has remained constant,at a fifth of all U.S. children. In Britain, however, where retirement benefitshave been declining, more than a third of the elderly were poor in 1979, butonly 9 percent of the children were poor (Coder and others 1989). (Universalfamily allowances in Britain include benefits for children, whereas those in theUnited States do not.)

The concern with targeting social security effectively is relevant in the OECD

as well as in developing countries, given rising costs, growing budget deficits,and the political sensitivities associated with the "mniddle-class capture" of ben-efits (see LeGrand and Winter 1987). Efficient targeting (see Weisbrod 1969) in-volves not only covering only those in the target g:roup (vertical efficiency) but


also including all individuals in the target group (horizontal efficiency). Unfor-tunately, there is a tension between these two aspects of targeting, and ensuringvertical efficiency by stringent means testing often adversely affects horizontalefficiency. It is not clear how worrying vertical inefficiency may be. As notedin the following section, the interaction of the tax and social security systemsmay permit the government to "claw back" the benefits from middle-class ben-eficiaries. What is important is to examine the overall tax and benefits system.

The Effect of Formal Social Security Programs

Theoretical and econometric work on the effect of social security on behav-ior in industrial countries has focused on the inefficiencies-such as in the de-cision to work-generated by social welfare programs and on possible adverseeffects of government pensions on the accumulation of capital. Other studiesquestion whether there has been a reduction in the private transfers that mightotherwise have taken place and whether social security displaces privatesavings.

In the labor market, questions about participation in the work force, numberof hours worked, and age of retirement have come under scrutiny. For employ-ers, the level of social security contributions required affects decisions on hiringand laying off workers and on whether to hire part- or full-time staff. Theevidence, however, is somewhat inconclusive. Atkinson (1987), in his reviewof the econometric literature, notes that seven studies of pensions in theUnited States used the same Longitudinal Retirement History Survey. Fourstudies found that pensions had a significant effect on the decision to retire,and three found that pensions were insignificant or economically unimportantfactors in the choice. Atkinson suggests that the models probably reflected theactivities of a subset of the population (white, married, working males) andthat the retirement decisions of the black, single, female, or self-employed werenot represented.

Many studies suggest that government social security programs have hadnegative effects, but this outcome may derive in part from inappropriately de-signed programs. If this is so, developing countries considering similar optionsshould carefully assess program design. A case in point is unemployment in-surance, under which unconditional benefits are likely to encourage a longerjob search and delay reentry into the job market. Atkinson (1987) suggests thatunemployment benefits should not be unconditional. If regulations specify theamount a worker must contribute and require the worker to look for a job,and if the duration of benefits is limited, there may be a greater incentive torejoin the labor market. Employment guarantee schemes are a nonexclusivealternative to unemployment insurance in poor countries or those undergoingstructural change, as in Eastern Europe.

Evidence also warns against adopting certain measures that have beenineffective, such as income-tax deductions for children. As the value of the

Ehtisham Ahmad 109

deductions increases with the marginal tax rate, the richest taxpayers benefitmost. This concession is likely to be highly regressive and can be replaced bybetter-targeted transfers directed at children. Th,e connections among policiesshould make it possible to foresee and offset any negative effects of particularoptions.

Lessons for Developing Countries

The notion of three stages in the evolution of social security systems(ILO 1984) is likely to be misleading-or at best irrelevant-if used as a guideto policy. To have had extensive experience in the public provision of socialsecurity and in the administration of a welfare state is no guarantee that thesystem will progress from one stage to another. This may be seen in the recentpolicy changes in Britain. Nor is it necessary to wvait until a country reaches aparticular level of gross domestic product (GDP) per capita before introducingsocial security measures.

Another important lesson from the experience of industrial countries is thatthe interaction among various programs has implications for the overall designof social security systems. Thus financing assistance through the general taxstructure in addition to earmarked wage and payroll taxes might result in un-expectedly high marginal tax rates, with undesirable consequences. Further-more, the interaction among formal social security and other programs mayalso have important financial implications. Governments need to examinewhether it would be cheaper to act to avoid the contingency of unemploymentor sickness, for example, or to pay for unemployment insurance and healthbenefits. This brings into sharp focus the relation between social security andfood relief, intervention in the labor market, and other health and educationprograms.

Finally, it should be noted that the methods of analysis and the modelingtechniques that have been used to examine the costs and benefits of variousprograms will prove to be valuable in the context of developing countries.

Social Security in Developing Countries

Public policy to ensure social security in developing countries should bebased on what individuals and households do to protect themselves, particu-larly in revenue-constrained economies. This means examining savings behav-ior and smoothing fluctuations in consumption, as well as assessing transfersand support mechanisms provided by the extended family or community. Thereis evidence that the poor, whose incomes are seldom above subsistence level,are able to cope with some fluctuations in income. But their ability to accumu-late and liquidate assets, or to rely on family and friends, is limited, particu-larly in the case of major and repeated shocks.


Traditional Social Security

For such contingencies as old age, the family is the principal source of sup-port in societies as diverse as rural China (see Ahmad and Hussain 1990), India,and Sub-Saharan Africa. This dependence typically reinforces the need for sonsand sharply limits the prospects for lowering the birthrate. Problems arise whenthere are no male offspring or when there is a breakdown in filial concern, giv-en the constraints of nuclear families. Iliffe (1987) notes that in many parts ofAfrica the word for poor is synonymous with that for the lack of kin or friends.

Support for the indigent who have no family assistance tends to be based oncommunity-level welfare and food security. In the Gambia, for instance, foodfor the needy and indigent is based on the Islamic precept of zakat (an ear-marked levy on wealth, sometimes provided to the mosque). In addition, a sys-tem of labor transfers provides half the total labor at the community level(von Braun 1990). This local provision is very effective in identifying the indi-gent, and it is relatively costless. Since resources are generally locally provided,however, events that affect the whole community make it less likely that theindigent will be covered in times of stress. Thus some degree of risk poolingacross regions or over time is desirable.

At the community level there are many fairly sophisticated survival strategiesthat are equivalent to social security. Such informal schemes are found in tra-ditional fishing villages, which are among the poorest communities in manyparts of Asia, Latin America, and Sub-Saharan Africa (see Platteau 1990).Fishing is a high-risk occupation. Daily income is subject to abrupt variation(production risks are often independent of price risks), and fishermen also riskthe loss of-or damage to-their lives and assets. Because the work is so phys-ically demanding, fishermen have relatively short working lives.

The high probability of low or zero catches precludes the payment of fixedwages. For the same reason, a fixed rental contract would not be attractive tothe fishermen. Instead, the wage system is based on share contracts. Such con-tracts, which divide the catch (often fifty-fifty), provide insurance to both own-ers and fishermen. Furthermore, a form of sickness coverage is providedequivalent to one share (or half a share) of the current income of the fishingunit (a benefit often observed in Sub-Saharan Africa). Many small-scale fisher-ies also provide for the subsistence needs of old fishermen or of poor familiesthat have lost all of their male income-earners. Preempting the rights of theindigent resembles an informal tax system designed for a social security pur-pose. In Kerala, India, risk pooling also takes the form of interest-free loans toneedy households. This transfer can be reciprocal because the loss of fishingincome is uncorrelated among the various boats. The important feature is thatsurplus earners give free credit in anticipation of future contingencies. Themoral reciprocity commitment implicit in the acceptance of free credit is en-forced by the community.

The most important social insurance mechanism at work in small-scalefisheries-particularly in Sub-Saharan Africa-is the practice of sharing

Ebtisham Abmad 1ll

income by offering employment. In Senegal, owners and crews often accom-modate unemployed fishermen despite the evident reduction in income. Theworkers' (and possibly the owners') remuneration varies inversely with thefirm's level of employment.

The drought in western Sudan shows the vulnerability of traditional socialsecurity practices. Family- and community-based survival strategies that wereeffective in 1980 and for a time thereafter weakened by 1984. In the worst-hitvillages, self-help mechanisms such as communal work parties, gift giving, vil-lage emergency funds, and annual contributions to the poor broke down com-pletely. In better-endowed villages, emergency funds-usually collected bytaxing sales of subsidized government sugar-were used to support survivalstrategies, such as transporting villagers to seek work (Teklu and von Braun1989).

These examples show relatively sophisticated social security mechanismsthat incorporate aspects of redistribution and insurance operating in poor com-munities. But the resources for such schemes are locally based, and the prac-tices cannot withstand sustained shocks. Moreover, the social cohesion thatunderlies these arrangements is increasingly susceptible to the challenges ofcontemporary society. There is a role for the state to provide formal socialsecurity by pooling risks across localities. Given stringent fiscal constraints,however, it is important to effectively target formal social security mechanismsto reinforce rather than supplant traditional support mechanisms.

Formal Social Security

Developing countries have had a variety of experiences with formal socialsecurity designed to provide for such contingencies as sickness, old age, andunemployment. Since governments typically face budget constraints, the choiceof policy is of central importance. But expenditures for the poorest and thoseat risk from starvation need not involve a tradeoff with investments in humancapital or the generation of assets. In many developing countries defense spend-ing often accounts for between a third and a half of current budget expendi-ture, and this expenditure is usually at the expense of programs that benefit thepoor. Table 1 shows that most Asian, North African, and OECD countries spendmore than 3 percent of GDP on defense. Although this percentage is consider-ably lower than that devoted to welfare, assistance, and social security in in-dustrial countries and Latin America, it is more than most other developingcountries spend to protect vulnerable groups. Some developing countries sub-sidize inefficient industries and prestige projects at great cost. Despite some lee-way-which may involve reallocating domestic expenditures as well as aid-budgets are not unlimited, and social welfare programs have to be financedfrom domestic resources.

An early example of a formal social security system that generates its ownresources is zakat, the Islamic provision for the indigent, the financing for


Table 1. Social Security and l)efense Expenditures as a Percentage of GDP, 1975, 1984

1975 1984

Country Social security Defense Social security Defense

Latin AmericaArgentina 4.9 2.0 6.3 3.4Bolivia 3.2a 2.4 1.7 6.6Brazil 7.0 1.3 6.4 0.8Chile 8.2 5.7 13.6 4.0Colombia 1.Oa 1.0 3.0 1.4Costa Rica 5.1 0.5 2.6 0.9Mexico 3.4 2.2 0.5 0.6Uruguay 10.4 2.6 11.3 2.6

AfricaBurkina Faso 0 2.0 1.1 2.7Cote d'lvoire 0.3a 1.2 1.1 1.2Mali - 1.4 2.1 2.1Morocco 1.5 4.6 1.9 6.6Tunisia 4.4 1.8 2.5 3.3

Middle EastIran 1.5 13.2 2.7 7.1Israel 10.8 26.7 11.9 23.6

AsiaChina n.a. 8.0 5 .3b 8.2India 1.9a 3.3 1.1c 3.6Korea, Rep. of 0.8 4.7 0.9 4.9Malaysia 0.7 6.9 1.8c 3.6Singapore 0.3 5.3 0.2 5.3Thailand 0.7 2.8 0.6 3.9

Centrally planned economyHungary 12.2a 2.5 12.4 15.8

OECD countries

France 14.9 3.8 17.5 4.0United Kingdom 8.4 4.9 11.7 5.2United States 7.7 5.9 7.6 6.4

- Not available.n.a. Not applicable.a. ILO 1981-83.

b. Liu 1986.

c. 1983 data from ILO 1981-83.Source: Social security statistics, IMF various years; 1975 defense statistics, Stockholm International Peace

Research Institute 1985; 1984 defense statistics and 1975 data on Costa Rica, Mexico, Burkina Faso, andMali, U.S. Arms Control and Disarmament Agency 1987.

which is obtained by the tax on wealth mentioned earlier. For agricultural landthere is a tithe (ushr), which carries the benefits of insurance, redistribution,and moral suasion. Local administration makes it more likely that the tax will

Ehtisham Ahmad 113

be paid and that only the indigent (for instance, unsupported widows, orphans,and the aged) will receive payments. Since land taxes have not been particularlyimportant in this century, there is some scope for raising additional revenue ifthe funds are earmarked for local social security; and not for general govern-ment expenditure.

EMPLOYMENT GUARANTEES AND UNEMPLOYMENT. Creating employment as aresponse to economic shocks has been a long-standing element of public policythroughout the world. During recessions and other crises, governments in in-dustrial countries have often set up public works programs to assist the unem-ployed. Britain implemented such schemes in the face of economic recessionand food shortages at the turn of the century, and the United States followedsuit in the 1930s. Because insurance and. credit markets often provide little pro-tection against deprivation in developing countries, intervention in the labormarket is often used to prevent destitution. Such schemes address a significantcause of poverty, provide rapid assistance to the most needy through self-selection, and can be designed to improve infrastructure.

Two countries, India and Chile, have successfully implemented such pro-grams. Maharashtra offers a striking example of an employment guaranteescheme. When an extended drought in 1972-73 cut cereal production by 50 per-cent, the government established a relief program based on employment andcash transfers. Employment was restricted to unskilled manual work. Thoseunable to work and without able-bodied relatives were eligible for relief. Thecurrent program builds on the original famine policy, but it provides year-roundguaranteed employment. In essence, the scheme combines income with an in-surance component. Since agricultural production is highly variable, the guar-antee provides coverage for these situations as well as for sudden or unexpectedshocks.

In 1987-88 the government of Maharashtra created 150 million man-days ofemployment at a cost of 2.6 billion rupees (11 to 12 percent of state expendi-tures). Part of the financing was provided by levies on professions and ear-marked taxes on sales and motor vehicles. These taxes, which are generatedprimarily in urban areas, provided 35 percent of the funds required. The em-ployment guarantee scheme has a significant element of redistribution fromsuch industrial cities as Bombay to rural and poorer municipal areas.

An interesting feature of the scheme is that it attempts to generate produc-tive assets. The government maintains a roster of eligible projects, with prioritygiven to infrastructure and other risk-reducing activities, such as soil andwater conservation, reforestation, road works, and flood protection. Localsupervisors oversee the programs. Assistance is perceived as a right, and thisstrong tradition mobilizes political support and a tight administrative structure,which includes supervision and evaluation.

With an open-ended guarantee such as Maharashtra's, targeting is necessary.The government has achieved this targeting through the self-selection mecha-


nisms of relatively low wages and hard manual labor. A low minimum wage,payable in cash and in subsidized grain, is guaranteed. The location of projectsand the periodicity of payments encourage the participation of the poorest seg-ments of the population. About 40 percent of the recipients are women, whoare paid the same wage rates as men (Echeverri-Gent 1988). Higher wageswould erode the self-targeting aspects of the scheme (a factor to be consideredas nationwide implementation in India is under way, and there is evidence ofsome leakage of benefits to nontarget groups).

The Programas de Empleo (EEP), instituted in 1974, was Chile's approach toself-targeted employment. Although Chile already had one of the most exten-sive formal social security systems in the developing world, the program playedan important role during two major crises. During the 1974-77 shock, theindex of real wages per capita (1970 equals 100) declined to 61 in 1974 and54 in 1975 and 1976. In 1976 the EEP employed about 6 percent of the laborforce at a time when unemployment was 16 percent. Unemployment fell in sub-sequent years, but the government maintained the EEP at the same level until1982, when Chile was forced to adjust again. The index of real earnings fellfrom 96 in 1981 to 69 in 1982, and unemployment climbed to 21 percent. TheEEP was doubled in 1983 to cover 13 percent of the labor force (together withunemployment, this amounted to 31 percent of the work force). Because theEEP was self-selecting, the total budgetary cost was only 1.4 percent of grossnational product at the height of the crisis in 1983. A survey suggests that themajority of workers in two of the main schemes were in the lowest incomequintile. The EEP was gradually run down as the labor market tightened andwas withdrawn completely in 1988, when unemployment returned to about6 percent.

These programs illustrate the possibilities for social security in the face ofshort-term economic and agricultural shocks. Both programs perform the func-tion of unemployment insurance common in industrial countries. The self-targeting helps reduce disincentives to work, which feature prominently in thediscussion of unemployment in industrial countries, and also limits budgetaryoutlays. To this extent such programs are also part of the formal programs dis-cussed in the following section.

SOCIAL INSURANCE AND ALTERNATIVES. Concern for the care of the elderlyand disabled has been an important factor in the development of social securityinstitutions in industrial countries. Providing for an aging population is easierif the program starts when the number of retirees relative to the working pop-ulation is still small. Models incorporating the dynamics of population and sec-toral change can evaluate the medium-term costs of particular options. Formalcare for the elderly in rnany developing countries can be compared with theneed for coverage in Britain when old-age pensions were introduced. At the be-ginning of the century .5 percent of the British population was age sixty-fiveor older. This ratio at present is estimated to be 4.1 percent in India and

Ehtisham Ahmad 115

5.9 percent in China (6.5 percent and 9 percent, respectively, are older than six-ty in the two countries). The proportion of the population older than sixty-fivein 2030 is projected to be 8.1 percent in India and 14 percent in China(12.5 percent and 21 percent are over sixty). The numbers in China are stag-gering, rising to more than 220 million older than sixty-five by 2020. The pro-portion of the population older than sixty-five in some Latin Americancountries by 2030 is projected to be even higher: 8.6 percent in Argentina,10 percent in Brazil, and 11 percent in Uruguay.

The Latin American countries, with relatively mature social security pro-grams and aging populations, present an interesting set of issues. At first mostcountries offered full funding of social insurance through trust funds that gen-erated surpluses when the number of retirees was small in relation to the work-ing population. These surpluses eroded over time as the trusts were forced tobuy government paper at below-market interest rates as well as by inappropri-ate investments. The practice of cross-subsidizing overruns on expensive healthcare with surpluses from the retirement accounts, along with the proliferationof benefits conferred when the schemes were in surlplus, depleted the trusts. Theschemes have now become mainly pay-as-you-go, with relatively high contri-bution rates. Table 2 shows that the combined contributions of workers, em-ployers, and the government are more than 30 percent of the wage bill in manycountries. Thus further increases in the payroll tax may not be desirable. Highcontribution rates lead to evasion, which reduces the base for social securityand increases the incentive to hire workers from the informal sector (who thendo not have coverage either).

During inflation, when contributions are not indexed and benefits are, gen-eral revenues could come under pressure to meet the projected deficits of thesocial security system. This is not a particularly easy problem to resolve, givenexisting budgetary deficits and the inability to raise fresh taxes. Nor is it likelyto be equitable to rely primarily on general revenues, which affect the wholepopulation (including the very poor), whereas social security benefits accruelargely to employees in the formal sector, who are likely to be better-off in anycase. The solution to this set of issues entails examining the entire system ofbenefits, funding, and taxes, and the implications for the population as awhole. The prospect of increasing coverage to expand the contribution rate rel-ative to liabilities is limited, since coverage is almost universal in such maturestates as Chile and Costa Rica.

Mesa-Lago (1990) argues that Latin American countries that have the widestsocial insurance coverage also have the most progressive incidence of the ben-efits of social security. A few countries have managed to extend coveragebeyond the formal urban sector, either because they have a modern unionizedrural sector (as in Chile and Costa Rica) or because new methods of financewere designed based on agricultural taxes and a redistribution from the formalurban sector. Thus in Brazil low-income rural residents are eligible for primaryhealth care and minimum pensions; Costa Rica, Uruguay, and some Caribbean


countries provide noncontributory pensions and health care for indigents.In countries with relatively low coverage, social security is more inequitable. InColombia, where only 16 percent of the population is covered, employers andthe state provide most of the finance. The often argued need for a close corre-spondence between contributions and benefits leads to a discriminationbetween the users of social insurance and the recipients of public assistance.This correspondence is a significant barrier to the extension of the social secu-rity system.

In China, income maintenance and social security have been successfullyprovided (see Ahmad and Hussain 1990) at relatively low levels of incomethrough enterprise-based employment guarantees in urban areas in conjunctionwith strict controls on mobility. Enterprises also provide pensions, housing, andhealth care in urban areas for more than 100 million people, along with rationsof essential food and consumer goods. Although these guarantees ensure min-imum living standards for a large proportion of the urban population, the en-terprise-based system has been a severe constraint on the pace of economicreform in China. The continuation of reforms in China (as in other centrallyplanned economies) depends greatly on whether economywide social insurancecan effectively replace enterprise-based provision.

With the introduction of the responsibility system after 1979, enterprises hadto be relieved of the burden of supporting retired workers (the ratio of workersto dependents was often as high as one to one in some older enterprises). Thisissue has become an essential ingredient in making firms responsive to the pricemechanism. Furthermore, the possibility of bankruptcy can be made functionalonly if the enterprises do not carry excessive nonproduction costs and if theincome of unemployed workers can be protected by unemployment insurance.Unemployment insurance is generally limited in duration. In a period of struc-tural change, with a substantial expansion of unemployment, self-targeted em-ployment provision is likely to be the only cost-effective method of providingfor the transitionally unemployed.

In rural areas of China the extended family provides much of the old-agesupport. There is also a system for distressed individuals (widows, orphans,and the elderly without family or kin support) called wu bao (five guarantees).As with Victorian poor relief, the emphasis has been more on preventing thenonpoor from getting relief than on ensuring that all of the eligible are covered.There is a severe social stigma attached to wu bao that restricts take-up rates.Thus rural social security essentially remains based on the ability to work andon the support of sons in old age (Jia 1988, Ahmad and Hussain 1990). Oneof the major objectives of wu bao, which was to restrict population growth byproviding an alternative to family support, has not been achieved.

Although China's public policy had not focused on providing support topoor individuals in rural areas, a system of regional targeting effectively sup-plied grain to chronically poor areas. This policy was effective before 1978 be-cause there was little mobility across regions and considerable equality within

Ebtisham Ahmad 117

00

Table 2. Social Insurance and Social Security Financing, 1983, 1987

Contribution as a percentage of total revenue, 1983Legal contribution as a percentage of salary, 1987 Surplus as a

Government percentageCountry Worker Employer Government Worker Employer and taxesa Investment Other of receiptsb

Latin AmericaArgentina 11.0 12.5-15.0 7.8-10.6 34.5 27.2 36.0 2.0 0.3 3Bolivia 1.5 1.5 1.5 25.5 34.8 24.2 12.4 3.1 23Brazil 8.5-10.0 10.01-19.2 c 15.2 74.7 8.0 0 2.1 1Chile (old) 19.1-20.0 0 c d

<- Chile (new) 13.3-13.6 0 od 29.9 2.0 47.1 19.1 1.9 14E Colombia 2.2 4.3 0.3 25.2 54.1 11.8 8.6 0.3 4

Costa Rica 2.5 4.8-7.3 0.3 28.4 47.0 18.6 5.3 0.8 31t7 Mexico 1.5 4.2 c 21.7 57.5 8.1 5.4 7.3 108 Uruguay 10.0-13.0 12.0-20.0 c 76.6 7.6 14.9 0.4 0.5 12a Africa

g Burkina Faso 4.5 4.5 0 23.9 75.6 0.1 0.4 0 37o Cote d'lvoire 1.6 2.4 0 61.9 17.7 0 20.4 0 43g Mali 1.6 2.4 0 13.7 75.5 1.9 0 8.8 3Q Morocco 1.7 3.4 0 8.4 78.4 0 13.2 0 43< Tunisia 1.3 2.5 0 n.a. n.a. n.a. n.a. n.a. n.a.a> Middle Easto Iran 7.0 20.0 3.0 17.4 49.8 25.8 6.2 0.7 64Z Israel 2.9 4.4 1.1 18.9 37.5 32.7 8.7 2.1 9

Continued on page 119

Table 2 (continued)

S ~~~~~~~~~~~~~~~~~~~~~~~Contribution as a percentage of total revenue, 1983Q- Legal contribution as a percentage of salary, 1987 Surplus as a

Government percentageCountry Worker Employer Government Worker Employer and taxes' Investment Other of receiptsb

AsiaIndia 6.3-8.1 7.6-8.8 1.5 2.0 95.3 0.3 2.2 0.2 47Malaysia 9.5 11.5 0 24.3 45.3 0.6 29.2 0.5 71Singapore 0-25.0 10.0 0 38.9 40.9 0.1 19.7 0.5 49Thailand n.a. n.a. n.a. 0 93.1 0.7 6.2 0 12

Centrally planned economiesChina 3.0 e 15.0-25.5 c n.a. n.a. n.a. n.a. n.a. n.a.Hungary 3.0-15.0 10.0-40.0 40d 14.9 47.1 38.0 0 0 d

OECD countriesFrance 6.5 8.2 0 21.5 50.4 24.0 1.4 2.6 2United Kingdom 3.9-9.0 3.9-10.5 14 17.9 23.9 49.8 2.1 6.3 3United States 5.7 5.7-11.4 c 22.6 34.3 28.6 8.2 6.3 12

n.a. Not applicable.Note: Data are for salaried employees; the self-employed pay a much higher percentage.a. Includes public health care in countries with national health systems.b. Surplus is defined as the excess of legal contributions and receipts over expenditures expressed as a percentage of receipts. In most cases it is difficult to disentangle

those payments made on behalf of public sector employees and those in which a deficit is being financed through additional taxes or general revenues. Only unambiguousdeficit coverage is indicated by (d).

c. Budgetary contributions: taxes and subsidies.d. Deficit coverage.e. Contributions from contract employees only.Source: Derived from U.S. Department of Health and Human Services 1987 and ILO 1981-83.

communes. Thus identifying a poor county was an effective way to targetpoor people. In recent years, however, regional inequalities have increased(see Ahmad and Wang 1989). The use of regional targeting thus means thatsome of the nonpoor are likely to benefit, and the poor in better-off countiestend to be missed.

China faces a significant problem in providing an effective safety net forworkers in informal and service activities who are not covered by the formallabor insurance system, as well as for the increasing number of employees intownship and village enterprises. The wu bao system is incapable of providingfor individuals on a large scale. But extending coverage is not without benefits,since an expansion of the base spreads the risk and ensures better intergener-ational redistribution. With the increase in life expectancy and the rapidgrowth in the number of indigent, considerations of equity will dictate impor-tant changes in China's social security system. This lesson applies as well toother centrally planned economies in the process of reform.

Formal social security coverage in countries such as India is fairly patchy.Guhan (1988) observes that formal sector (mainly public) employees are cov-ered by reasonable sickness, maternity, pension, and survivor benefits, as wellas with subsidized housing. In addition, public employees are not vulnerableto unemployment. Self-employed and informal sector workers, however, whoconstitute nearly 90 percent of earners, get virtually no benefit from socialinsurance or social assistance. Even in the industrial sector, evasion, defects inlegislation, inadequacies in coverage, and difficulties in enforcement delay ordeny benefits to those entitled. State-level pensions, designed to provide benefitsto those below the poverty line, also suffer from shortcomings. Guhan notesthat pensions cover probably no more than a fifthi of the old people under thepoverty line who should qualify for these pensions. There are great differencesin coverage across states. For instance, rural widows are likely to be among themost vulnerable groups in India, but only rarely, as in Kerala, is there a formalpension provision for widows. The level of coverage in most states is low partlybecause eligible beneficiaries lack information, but also because of restrictivequalifying conditions relating to age, destitution, and domicile and because ofad hoc financial ceilings. Furthermore, Guhan concludes, corruption and pro-cedural delays restrict access. The main concern is how to devise an effectivesocial security system that can be targeted to relevant groups without excessiveexclusion. This will require employment guarantees as well as local identifica-tion and care for the indigent or those unable to participate in the labormarket.

A number of instruments are likely to be important for effective social secu-rity in most developing countries. These range from employment guarantees(in lieu of or in addition to unemployment benefits) even in countries withestablished formal systems, such as Chile, to social insurance for basic healthcare and retirement in countries with a large formal sector, such as China andIndia. Local redistributive mechanisms, whether community-level or based on


formal tax and allocation mechanisms, will require an element of central gov-ernment support to insure against covariate risks. The government's ability toidentify the vulnerable and needy as well as political and administrative con-straints will determine the effectiveness of the system.

Prospects for Formal Social Security

Governments need to evaluate the objectives of social security and decidewhether to rely on public or private insurance. In addition, the difficulties ofadministering these programs, particularly in the informal sector, cannot beoveremphasized. Agricultural and self-employed workers are hard to trace. Itis difficult to keep track of people who move back and forth and whose workhistories are not on record. Similarly, distributing benefits to these recipients isa challenge. Thus governments need to look at the consequences of alternativepolicies for incentives, government budgets, and households in different cir-cumstances.

Public or Private?

The public provision of formal social security has been criticized by observ-ers on the right and the left of the political spectrum. The former argue thatthe public sector is incapable of an efficient response and that it would be pref-erable to turn to private insurance and rely on the market. One of the principaleffects of state-financed social security, however, is the universal reduction inuncertainty. For some risks, such as industrial accidents, the probability andcosts are assessable and private insurance may be provided on an actuarialbasis. But problems of moral hazard and adverse selection may prompt a dif-ferent response from the government and the private sector. Moreover, privateinsurance does not apply to such events as calamities, unemployment, and abreakdown of family support for the aged.

Traditionally, the family has played a major role in providing security inmost countries, and it continues to perform the same function today. In manyparts of the world, these intergenerational transfers have led to pressures toproduce male offspring. The resulting effects on population growth are wellknown, as is the tendency for an unequal distribution of entitlements withinthe family. When support is lacking, the community typically provides insur-ance, either through a system of gifts and mutual obligations or through well-defined power relations, such as those between landlords and tenants thatsometimes guarantee employment and entitlements. Such arrangements arelocalized and do not offer the advantages of pooling risk across regions or cli-matic zones. Thus in major crises, such as widespread losses in entitlements,social security measures at both the community and the family level are likelyto be insufficient. Furthermore, migration and the proliferation of nuclear

Ehtisham Ahmad 121

families, particularly in urban areas, are likely to restrict the effectiveness offamily-based care.

That the provision of charity contributes to the public good (in the sensethat private transfers produce externalities) is well known, since the benefits donot accrue to the persons making the transfer. Moreover, "we might all of usbe willing to contribute to the relief of poverty, provided everyone else did. Wemight not be willing to contribute the same amount without such assurance"(Friedman 1962, p. 191). There is thus a considerable scope for everyone beingmade better-off by the state through different forms of intervention as well asthrough direct income support.

As already noted, the main elements that are relevant in choosing betweenpublic and private alternatives in a developing country are administrative costs,adverse selection, and moral hazard. If administrative costs are lower for pri-vate insurance than for public provision, privatization would be less costly forthe insured and would eliminate distortions for those unable to afford privateinsurance. But government schemes may have the advantages of economies ofscale, lower advertising costs, and so on.

Although these arguments assume that perfect competition is sustainable inthe insurance market, a competitive equilibrium may not exist, given adverseselection (where the probability of the event is known to the individual but notto the company). Thus "pooling equilibria" may not be stable, in the sense thatwhen the premium is based on the expected claims for the population, high-riskindividuals will take out insurance and low-risk individuals will not. And asSpence (1978) noted, if separate premia are devised, policies designed forlow-risk individuals would be bought by high-risk customers, thus introducingnegative externalities based on information. This adverse selection is one of thereasons elderly Americans cannot easily buy medical insurance. As "the priceof insurance rises the people who insure themselves will be those who are in-creasingly certain that they will. need the insurance" (Akerlof 1970, p. 492).One solution is to make insurance compulsory to ensure that low-risk purchas-ers do not opt out. The inefficiencies involved may be far less than under thealternative.

A more serious problem is that of moral hazard, when individuals can influ-ence the probability of the insured event in a way that cannot be monitored bythe insurance company. Voluntary unemployment is a common example. Thisagain leads to a breakdown in the market for insurance.

Considerable work on insurance markets examines alternative situations, in-cluding separating equilibria (with different contracts for higher- or lower-riskgroups), and equilibrium with mixed strategies. But "the central message is inthe fact that the market for insurance is unlikely to be perfectly competitive"(Atkinson 1989, p. 116). This brings into question the relevance of the privatealternative even in industrial countries.

Given that administrative costs make up a significant part of formal socialsecurity systems, the choice between public and private provision has to take


into account not only the relative costs but also such issues as market structure.There are a number of hybrid alternatives. One is the Swedish system of no-tionally private unemployment insurance. The system is state regulated, and90 percent of the funding comes from the government, although the adminis-tration is private. "The system is therefore more usefully regarded as apublic scheme whose administration has been hived off to the private sector"(Barr 1987, p. 16).

There is also a trend in many countries to encourage private insurance forthe well-to-do, supplemented by social insurance for the rest of the population.This option is likely to have a major drawback, in that private insurers maycream off all but the highest-cost groups, making social insurance excessivelyexpensive and insufficiently diverse (Spence 1978). A further issue that privateinsurance does not address is the redistribution of income that is inherent inthe social insurance system. This question leads to an investigation of theinterface between social insurance and tax policy.

Alternative Public Programs

Developing countries have had to devise ways to target benefits to reach thepoor, who are not covered by formal social insurance. The alternatives areto rely on means-tested benefits, particularly at the local level, where identifi-cation is less of a problem, and on benefits that are independent of a contrib-utory framework, possibly targeted to particular groups such as children or theelderly.

The literature on optimum taxation, relating to the choice of the tax sched-ule (Atkinson 1987, 1989; Atkinson and Stiglitz 1980), is useful in assessingmeans testing. The provision of a uniform benefit shifts the budget constraintsuch that the marginal tax rate remains zero below the tax threshold. If ameans-tested benefit is provided to individuals below the threshold and is com-pletely withdrawn as incomes rise above the threshold, such individuals facea 100 percent marginal tax rate on benefits. The choice is thus between a risingmarginal tax rate at the bottom of the schedule, with social insurance, and afalling marginal tax rate, with means-tested transfers. Similar issues arise in thechoice between income testing and basic benefit schemes, and these involve bal-ancing the objectives of efficiency and distribution. If the decisions of thelow-income groups relating to work, savings, and so on are unaffected by themarginal tax rate and do not affect the private provision of insurance, or pri-vate transfers, then full withdrawal of benefits would be desirable. If, however,these decisions are distorted in some way, then the rate of withdrawal woulddepend on the efficiency cost of the distortion relative to greater target efficien-cy. The desirability of either falling or rising marginal tax rates depends on thespecification of the problem, and it is not possible to be categorical about eitherex ante.

Ehtisham Ahmad 123

This discussion concerns the form of the benefit. Equally important hurdlesare the administration and costs of determining eligibility, both in time spentgathering information, filing for benefits, and examining claims, and in the stig-ma that is often associated with such claims. Means testing often brings aboutincomplete take-up of benefits (although this can arise without means testing).This result is seen as an advantage by some planners, who advocate ordeals asa means of improving target efficiency, and it is common, for instance, in manyfood-for-work programs. But to ensure that all eligible recipients receive bene-fits requires local information and identification. Regional targeting to providefor all individuals in specific poor regions may well be cost-effective and couldbe an important supplement to social security where regional deprivation isidentifiable. China's experience shows the need for such an option and its lim-itations.

Certain transfer schemes are prone to abuse, and the possibility of fraudu-lent claims has to be addressed. To some extent this is the reason that self-selecting targeting mechanisms, such as inferior-quality rations or work tests,are used. But how administrators are rewarded also affects overall arrange-ments. For instance, it matters whether rewards to administrators are based onreaching target groups or on considerations of cost. If cost is the criterion,officials may save money by not reaching target groups.

If it is the citizen's right to be provided for by the state, basic benefits couldbe targeted to all the elderly or to families with children. Thus a redistributivearrangement to provide a pension equal to the poverty line (140 rupees a retireein India, 250 yuan in China) to all those over sixty-five in India and Chinawould amount to less than 2 percent of GDP. More targeted versions, with ben-efits accruing only to the indigent over sixty-five, would have to be based onlocal information and administration. Under both options, a redistributiveguarantee could be provided to the indigent and the most vulnerable withoutprecluding additional insurance (social or otherwise) for those who desire it.The government would not be obligated to provide this insurance, which wouldfree resources for other productive activities or for benefits to vulnerablegroups.

The strategy of providing basic benefits avoids the costs associated with de-termining a person's employment status and means. It can also supplementsocial insurance by providing funds for low-income workers (Atkinson 1989).If basic benefits are provided at a level that does not induce withdrawal fromthe labor force, a properly targeted social insurance scheme, say for sicknessor unemployment, would have no adverse income effects on labor supply(see the analysis of tagging in Akerlof 1978). The standard issues are, first, ifthe transfers are adequate to live on, there will be incentive problems, and,second, the arithmetic dictates that the tax rate will be high. This points tocombining policies to make the social security system effective. Political choicesand administrative constraints are likely to determine the final outcome.


Conclusion

Effective social security requires a number of policy instruments. But not allof the policies used in industrial countries are relevant for developing countries.Thus administrative constraints and the effects on incentives to work dictatethat unemployment insurance may be inappropriate even in the more advanceddeveloping countries with established formal programs. This does not meanthat nothing can be done. Self-selecting employment schemes can provide abasic safety net and also be designed to retain workers and add to productiveinfrastructure. Formal programs to protect vulnerable groups, such as the sickand the aged, are also likely to have positive long-term consequences, reducinguncertainty and lessening the demand for larger families. To some extent, theseprograms as well as social insurance schemes are likely to be financed by con-tributions and by pooling risks across generations and groups. These contribu-tions would not generally accrue to government revenues. The experience withsocial insurance, however, suggests that it is a function of formal sector em-ployment. Minimum noncontributory pensions targeted solely by age wouldneed to be funded through general revenues. Further targeting at the local levelto identify the indigent may also be possible. Such programs could build oncommunity arrangements where they exist but would depend on local admin-istration and financing.

Note

When this article was written, Ehtisham Ahmad was a senior economist on the staff ofthe World Development Report at the World Bank. He was on leave of absence from the LondonSchool of Economics, where he was director of the Development Economics ResearchProgramme. This paper was greatly influenced by the work of and by discussions with TonyAtkinson. Comments by Jean Dreze, Ravi Kanbur, and Nick Stern are gratefully acknowledged.Nonetheless, responsibility for the views in this paper rests entirely with the author.

References

The word "processed" describes works that are reproduced from typescript by mimeograph,xerography, or similar means; such works may not be cataloged or commonly available throughlibraries, or may be subject to restricted circulation.Ahmad, Ehtisham, Jean Dreze, John Hills, and Amartya Sen, eds. 1990. Social Security and

Welfare in Developing Countries. New York: Oxford University Press.Ahmad, Ehtisham, and Athar Hussain. 1990. "Public Action for Social Security in China." In

Ahmad and others.Ahmad, Ehtisham, and Yan Wang. 1989. "Inequality and Poverty in China: Institutional Change

and Public Policy, 1978-1988." Background paper for World Bank, World DevelopmentReport 1990 (New York: Oxford University Press, 1990). Processed.

Akerlof, G. A. 1970. "The Market for Lemons: Qualitative Uncertainty and the MarketMechanism." Quarterly Journal of Economics 84:488-500.

. 1978. "The Economics of Tagging." American Economic Review 6:8-18.

Ehtisbam Abmad 125

Atkinson, A. B. 1987. "Income Maintenance and Social Insurance." In Alan Auerbach andMartin Feldstein, eds., Handbook of Public Economics. Volume 2. Amsterdam: North-Holland.

. 1989. Poverty and Social Security. Brighton: Wheatsheaf Press.

Atkinson, A. B., and J. Hills. 1990. "Social Security in Developed Countries: Are There Lessonsfor Developing Countries?" In Ahmad and others.

Atkinson, A. B., and J. E. Stiglitz. 1980. Lectures on Public Economics. Maidenhead, U.K.:McGraw-Hill.

Barr, Nicholas. 1987. "The Welfare State As an Efficiency Device." LSE/STICERD Welfare StateProgramme Working Paper 22. London School of Economics. Processed.

- . 1988. "The Mirage of Private Unemployment Insurance." LSE/STICERD Welfare State Pro-gramme Working Paper 34. London School of Economics. Processed.

Beveridge, Lord William. 1942. Social Insurance and Allied Services. Cmnd. 6404. London: HMSO.

Burtless, G. 1986. "Public Spending for the Poor: Trends, Prospects and Economic Limits." InS. H. Danziger and D. H. Weinberg, eds., Fighting Poverty. Cambridge, Mass.: HarvardUniversity Press.

Coder, John, Lee Rainwater, and Timothy Smeeding. 1989. "Inequality among Children and theElderly in Ten Modern Nations: The United States in an International Context." AmericanEconomic Review (May): 320-24.

Echeverri-Gent, John. 1988. "Guaranteed Employment in an Indian State: The MaharashtraExperience." Asian Survey 28 (December).

Friedman, Milton. 1962. Capitalism and Freedom. Chicago: University of Chicago Press.

Guhan, S. 1988. "Social Security in India: Looking One Step Ahead." Bulletin of tbe MadrasDevelopment Seminar Series 18:438-58.

Iliffe, John. 1987. The African Poor. Cambridge: Cambridge University Press.

ILO (International Labour Organisation). 1981-83. The Cost of Social Security: Twelfth Interna-tional Inquiry. Geneva.

- . 1984. Into the Twenty-first Century: The Development of Social Security. Geneva.

IMF (International Monetary Fund). Various years. Government Finance Statistics Yearbook.Washington, D.C.

Jia, Aimei. 1988. "New Experiments with Elderly Care in Rural China." Journal of Cross-Cultural Gerontology 3:139-48.

LeGrand, J., and D. Winter. 1987. "The Middle Classes and the Defence of the British WelfareState." In R. E. Goodin and J. LeGrand, eds., Not Only the Poor. London: Allen and Unwin.

Liu, W. T., ed. 1986. China Social Statistics. New York: Praeger.

Mesa-Lago, Carmelo. 1990. "Formal Social Security in Latin America and the Caribbean." InAhmad and others.

Platteau, J.-P. 1990. "Traditional Systems of Social Security and Hunger Insurance: Some Lessonsfrom the Evidence Pertaining to Third World Village Societies." In Ahmad and others.

Spence, M. 1978. "Job Market Signalling." In P. A. Diamond and M. Rothschild, eds., Uncer-tainty in Economics. New York: Academic Press.

Stockholm International Peace Research Institute. 1985. World Military Expenditures and ArmsTransfers. Stockholm.

Teklu, Teklu, and Joachim von Braun. 1989. "Coping with Disastrous Drought: The 1984-85Famine in North Kordofan, Sudan." International Food Policy Research Institute, Washington,D.C. Processed.

U.S. Arms Control and Disarmament Agency. 1987. World Military Expenditures and ArmsTransfers. Washington, D.C.

U.S. Department of Health and Human Services. Social Security Administration. 1987. SocialSecurity Programs throughout the World. Washington, D.C.

126 The World Bank Research Observer, vol. 6,no. 1 (January 1991)

von Braun, Joachim. 1990. "Social Security in Sub-Saharan Africa." In Ahmad and others.

Weisbrod, B. A. 1969. "Collective Action and the Distribution of Income: A CollectiveApproach." In U.S. Congress, joint Economic Committee, The Analysis and Evaluation ofPublic Expenditures. Washington, D.C.: GPO.

Ehtisham Ahmad 127

Subscription CouponIf you are not already a subscriber to The World Bank Research Observer, you

may begin a subscription by completing and returning this coupon to W;orld BankPublications.

You may also use this coupon to subscribe to The World Bank EconomicReview. This journal, directed at professional economists, publishes original researchthat emphasizes empirical applications and testing. The findings reported are basedon rigorous economic methods and require some background in economic theoryand analysis. Articles are accessible to general economists and not just those involvedin academic research.

Subscriptions to both The Economic Review and The Research Observerare available on a complimentary basis to readers with mailing addresses in non-OECDcountries. For readers with mailing addresses in OECD countries, discounts areavailable for three-year subscriptions and subscriptions to both journals.

Indicate below the type of subscription desired. Please complete both sides ofthis subscription coupon.

For Readers with Mailing Addresses in Non-OECD Countries:Please begin my complimentary subscription to El The World BankEconomic Review U The World Bank Research Observer (check eitheror both boxes). l

For Readers with Main Addresses in OECD Countries:(Check appropriate box).

One Year Thre Years- - ' diuls' -stlaflonIr1 ndivi institutious

The World BankEoomic $25 0 $45 a $ 65 O $125 0ReOiew (three issues per year)

The World BankResearch 20 U! $35 b $ 50 S $95 -Observer(two issues per year)

BothJournals 40 B $70 U $100 U $175 U

Check metbod qfpaymltU Enclosed is my check for US . . , payable to

World Bank Publications.

Charge my U VISA o MasterCard U3 American Express

Credit Card Number

E Signature Expiration Date

UPlease bill me. (InstitutionaL orders only)

Purchase order number

Please complete both sides of this subscription coupon.

Send to: World Bank PublicationsBox 7247-8619Philadelphia, PA 19170-8619, U.S.A.

Please print clearly

Name

Title

Organization name

Organization address

County _ Postal code

Please check only one.O 01 International, regional organization (including field offices), that is, WHO, UNDP

O 02 Central banks, finance and planning ministriesO 03 Other government agencies (including embassies)O 04 Nongovernmental organizations (N(OS) and other associationsO] 05 Research institutionsO 06 University facultyO 07 librariesO 08 BookstoresO 09 Commercial banks and other financial institutions (including savings and

loan associations, security dealers, and so forth)l 10 Business - manufacturing (goods) Where both apply check 'goods"

El 11 Business - consulting (services)El 12 StudentO 13 News media

El RO General economics: theory, history, systemsO RI Economic growth: development, planning, fluctuationsO] R2 Economic statisticsO R3 Monetary and fiscal theory and institutionsO R4 International economicsO R5 Administration, business finance, marketing, accountingO R6 Industrial organization; technological change; industry studiesEl R7 Agriculture, natural resources, rural developmentO R8 Labor, population and human resourcesO R9 Welfare programs; consumer economics; urban and regional economics

The most usefil, comprelhensive, and detailed single set of statistics available ontlhe debt of dteveloping countries.

--Robert H. CassenDirector, Intemational Development Centre, University of Oxford

World Debt Tables 1990-91* Featuring complete and up-to-date data on debt for over100 developing countries

* Includes the latest 1989 data and projections through 1999

* Plus periodic supplements throughout the year as new databecome available

When you need current and reliable data on developing country debt,you will find no better source than World Debt Tables 1990-91. Thisnew edition gives you instant access to data gathered through the WorldBank's exclusive Debtor Reporting System. You also get from topWorld Bank economists an analysis of the world debt situation that willhelp you put the data into perspective.

World Debt Tables 1990-91Volume 1. Analysis and Summary TablesVolume 2. Country Tables

Complete two-volume set and supplements:664 pages / ISBN 0-8213-1725-3 / US$125.00

Available separately:Volume 1. Analysis and Summary Tables:172 pages / ISBN 0-8213-1723-7 / US$16.95

To order World Debt Tables 1990-91, list title, stocknumber, and price on the coupon that follows.

Trends in DevelopingEconomies 1990

The recent economic performance of some 90 developingeconomies -- arranged alphabetically for easy reference -- isassessed in the second edition of the World Bank's newest annualstatistical series.

Including ...* A comprehensive table of economic and social indicators for

1965-89* Descriptive analyses for the majority of countries* Highlights of the problems and prospects for each country's

economy and the main elements of its development strategy.

The series complements the World Development Report, whichlooks at global and regional economic trends and their implicationfor future prospects of the developing countries.

The profiles of each country are drawn from the Bank's data baseand economic analyses of its member countries.

632 pages / ISBN 0-8018-1648-6 / US$19.95 / Order Stock #11648

_

By exploring the effectiveness of rapid smiall-farin growth in redtcing poverty --antd the scope of social action to improve that ectiveness -- thi.s splendjd hookslwuld capture the attention of policymakers an d researchers.

--Michael Lipton, Institute of Dcveloprncnt Studics, University of Sussex

The Great AscentThe Rural Poor in South AsiaInderjit Singh

This new book challenges the widely held view that the rapid growth of smallfarms in South Asia has done the poor more harm than good. By criticallyreviewing recent academic literature and the World Bank's operationalexperience in Bangladesh, India, and Pakistan, Inderjit Singh shows thatagricultural growth has benefited all classes of the poor. He concludes that the"great ascent" from poverty to a more materially rewarding life has begun.

The report examines in detail a variety of programs intended to help the poor. Itevaluates agricultural research, extension, and training activities, as well asprograms in dairying, poultry farming, commercial fishing, and forestry. Mr.Singh also examines the failure of land reforms to reduce poverty.

Published for the World Bank by The Johns Hopkins University Press472 pages / ISBN 0-8018-3954-81 US$39.95 ! Order Stock #43954

Of Related Interest by Inderjit Singh

Land and Labor in South Asia244 pages / ISBN 0-8213-1129-8 / US$ 12.95 / Order Stock #20033

Small Farmers in South AsiaTheir Characteristics, Productivity, and Efficiency82 pages / ISBN 0-8213-1090-9 / US$6.95 / Order Stock #20031

Tenancy in South Asia128 pages / ISBN 0-8213-1091-7 / US$10.95 / Order Stock 20032

To order The Great Ascenzt or other publications byInderjit Singh, list title, stock number, and price on thecoupon that follows.

The World Bankand the Environment

First Annual Report, Fiscal 1990This report reviews the Bank's progress in combating environmentaldegradation during July 1989 - June 1990. The main goal of the Bank'seffort has been the integration of environmental concems into itsdevelopment work.

The Bank emphasizes five problem areas in its operations ...

* Destruction of natural habitats* Land degradation* Degradation and depletion of fresh water resources* Urban, industrial, and agricultural pollution* Degradation of "the global commons" such as atmospheric and

marine pollution

Describes not only the Bank's successes in its environment work, but alsothe problems it has encountered.

112 pages / ISBN 0-8213-1641-9 / US$7.95 J Order Stock #11641

Also Available from the World Bank

Environmental Health Components for Water Supply, Sanitation,and Urban ProjectsJames A. Listorti

Technical Paper 121 / 156 pages / ISBN 0-8213-1537-4 / US$9.95 / Order Stock #11537

The Greenhouse EffectImplications for Economic Development

Erik Arrhenius and Thomas W. Waltz

Discussion Paper 78 / 26 pages / ISBN 0-8213-1520-X / US$4.95 / Order Stock #11520

Conserving the World's Biological Diversity

Jeffrey A. McNeely, Kenton R. Miller, Walter V. Reid, Russell A. Mittermeier, andTimothy B. Werner

204 pages / ISBN 0-8213-1384-3 / US$14.95 / Order Stock #11384

To order any of these publications, list title, stock number, andprice on the coupon that follows.

New from the World Bank'sDiscussion Paper Series

International Migration and Development inSub-Saharan AfricaVolume 1. OverviewVolume 2. Country AnalysisSharon Stanton Russell, Karen Jacobsen, and William Deane StanleyISBN 0-8213-1642-7 (Volume 1) / US$10.95 / Stock #11642ISBN 0-8213-1643-5 (Volume 2) / US$10.95 / Stock #11643Provides a "state of knowledge" analysis of the relation of internationalrmigration to development in the region.

Redefining the Role of Government in Agriculture forthe 1990sOdin Knudsen and John Nash, with contributions by James Bovard, Bruce Gardner,and L. Alan WintersISBN 0-8213-1690-7 / US$7.95 / Stock #11690Questions the massive government intervention in the private production andmarketing of agricultural goods throughout the world.

Debt Management SystemsDebt and International Finance DivisionISBN 0-8213-1696-6 / US$13.95 / Stock #11696Examines the benefits of and obstacles to computer-supported debtmanagement systems.

Problems of Developing Countries in the 1990sVolume 1. General TopicsVolume 2. Country StudiesF. Desmond McCarthy, editorISBN 0-8213-1632-X (Volurne 1) / US$13.95 / Stock #11632ISBN 0-8213-1633-8 (Volume 2) / US$15.95 / Stock #11633Identifies many of the leading economnic issues that developing countries willface in the 1990s.

To order any of these discussion papers, list title, stocknumber, and price on the coupon that follows.

Use this coupon to orderany of the titles listed on the previous pages.

Title Stock Number Price

Subtotal US$Shipping and handling US$ 3.50

Airmail surcharge outside USA (US$6.00 per copy) US$Total US$

Check your method of payment. Orders from individuals must be accompanied bypayment or credit card information. Ask local distributors about forms ofpayment acceptable to them.

O Enclosed is my check payable to World Bank Publications in U.S. dollars.

Charge my OI VISA a MasterCard FI American ExpressThese credit cards are accepted for orders addressed to Philadelphia. (Check withyour local distributor about acceptance of credit cards in your country.)

Credit card accournt number Expiration date

Signature

Institutional customers only:a Send invoice and reference purchase order no. _

Please print dearly

Name

Title

Firm

Address

City State Postal Code_ _

Country Telephone

Customers in the United States: Detach this form and mail to World Bank Publications, P.O.Box 7247-8619, Philadelphia, PA 19170-8619, U.S.A.

Customers outside the United States: Contact World Bank Publications, 1818 H Street, NW,Washington, DC 20433, U.S.A. for a list of World Bank pub]lications distributors. Write tothe local distributor in your area for information on prices and terms of paymenL Couponsreceived in Washington from countries with authorized distributors will be returned to thecustomer. Customers in countries with no authorized distributor, return your coupon toWorld Bank Publications, P.O. Box 7247-8619, Philadelphia, PA 1917048619, U.S.A.

Documents

World Bank Document · Contestable Markets, Trade, and Development William J. Baumol and Kyu Sik Lee How Much Does Trade with the South Affect Workers in the North? Adrian Wood Organized