18
MODELING INFORMALITY FORMALLY: HOUSEHOLDS AND FIRMS SEBASTIAN GALIANI and FEDERICO WEINSCHELBAUM Informality is widespread in most developing countries. In Latin America, 50% of salaried employees work informally. Three stylized facts characterize informality: (1) small firms tend to operate informally while large firms tend to operate formally; (2) unskilled workers tend to be informal while skilled ones have formal jobs; (3) ceteris paribus, secondary workers (a worker other than the household head) are less likely to operate formally than primary workers. We develop a model that accounts for all these facts. In our model, both heterogeneous firms and workers have preferences over the sector they operate and choose optimally whether to function formally or informally. There are two labor markets, one formal and the other informal, and both firms and workers act unconstrained in them. By contrast, a prominent feature of the preexisting literature is that workers’ decisions play no role in determining the equilibrium of the economy. In our model, policies that reduce the supply of workers in the informal labor market at given wages will increase the level of formality in the economy. This has noteworthy implications for the design of social programs in developing countries. We also show that an increase in the participation of secondary workers would tend to raise the level of informality in the economy. (JEL J24, J33) I. INTRODUCTION Recent estimates indicate that around 10% of gross domestic product (GDP) in the United States is produced by individuals or firms that evade taxes (Schneider and Enste 2000). In Latin America, approximately 50% of all salaried employees work informally. An important ques- tion is, therefore: What are the determinants of informality? Unfortunately, three decades of research have not yet yielded a consensus opin- ion on this point. *We thank David Levine, two anonymous referees, sem- inar participants at London School of Economics, Uni- versity of Edinburgh, Universidad Di Tella, Universidad de San Andr´ es, Washington University in St. Louis, Uni- versidad Nacional de San Luis, Pontificia Universidad Catolica de Chile, Fundaci ´ on de Investigaciones Econ ´ omicas Latinoamericanas, LACEA Annual Meetings 2006, EALE Annual Meeting 2007, SOLE Annual Meeting 2008, and ISNIE 2009 for useful comments and Leopoldo Tornarolli from CEDLAS for providing us with the data used in this paper. We are also grateful for the excellent research assis- tance of Paulo Somaini. Galiani: Professor, Department of Economics, Washington University in St Louis, Campus Box 1208, St Louis, MO 63130-4899. Phone (314) 935-5670, Fax (314) 935- 4156, E-mail [email protected] Weinschelbaum: Professor, Department of Economics, Uni- versidad de San Andres, Vito Dumas 284, (B1644BID) Victoria, Provincia de Buenos Aires, Argentina. Phone +54-11-4725-7041, Fax +54-11-4725-7010, E-mail [email protected] The traditional explanation, initiated with the work of Lewis (1954), assumes the existence of segmented labor markets in which some workers do not have access to jobs in the regulated, for- mal sector. These workers are forced to accept lower-paying informal-sector jobs that usually involve inferior working conditions. While this view has become prevalent in the development literature, direct empirical tests of the premise that informal workers would expect higher wages in the formal sector yield mixed results. For instance, Magnac (1991), Maloney (2004), and Pratap and Quintin (2004) do not find com- pelling evidence of segmentation between the formal and the informal sectors using data from Colombia, Mexico, and Argentina, respectively. It is, nonetheless, certainly plausible that the presence of unions might segment the labor mar- ket (Layard et al. 1991). However, even if we believe that unionization has been an important factor in determining labor-market segmentation in the past, the pronounced decline of union- ization throughout Latin America over the last ABBREVIATIONS GDP: Gross Domestic Product VAT: Value Added Taxes 821 Economic Inquiry (ISSN 0095-2583) Vol. 50, No. 3, July 2012, 821–838 doi:10.1111/j.1465-7295.2011.00413.x Online Early publication September 21, 2011 © 2011 Western Economic Association International

GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

Embed Size (px)

Citation preview

Page 1: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

MODELING INFORMALITY FORMALLY: HOUSEHOLDS AND FIRMS

SEBASTIAN GALIANI and FEDERICO WEINSCHELBAUM∗

Informality is widespread in most developing countries. In Latin America, 50%of salaried employees work informally. Three stylized facts characterize informality:(1) small firms tend to operate informally while large firms tend to operate formally;(2) unskilled workers tend to be informal while skilled ones have formal jobs;(3) ceteris paribus, secondary workers (a worker other than the household head) areless likely to operate formally than primary workers. We develop a model that accountsfor all these facts. In our model, both heterogeneous firms and workers have preferencesover the sector they operate and choose optimally whether to function formally orinformally. There are two labor markets, one formal and the other informal, and bothfirms and workers act unconstrained in them. By contrast, a prominent feature ofthe preexisting literature is that workers’ decisions play no role in determining theequilibrium of the economy. In our model, policies that reduce the supply of workersin the informal labor market at given wages will increase the level of formality inthe economy. This has noteworthy implications for the design of social programs indeveloping countries. We also show that an increase in the participation of secondaryworkers would tend to raise the level of informality in the economy. (JEL J24, J33)

I. INTRODUCTION

Recent estimates indicate that around 10%of gross domestic product (GDP) in the UnitedStates is produced by individuals or firms thatevade taxes (Schneider and Enste 2000). In LatinAmerica, approximately 50% of all salariedemployees work informally. An important ques-tion is, therefore: What are the determinantsof informality? Unfortunately, three decades ofresearch have not yet yielded a consensus opin-ion on this point.

*We thank David Levine, two anonymous referees, sem-inar participants at London School of Economics, Uni-versity of Edinburgh, Universidad Di Tella, Universidadde San Andres, Washington University in St. Louis, Uni-versidad Nacional de San Luis, Pontificia UniversidadCatolica de Chile, Fundacion de Investigaciones EconomicasLatinoamericanas, LACEA Annual Meetings 2006, EALEAnnual Meeting 2007, SOLE Annual Meeting 2008, andISNIE 2009 for useful comments and Leopoldo Tornarollifrom CEDLAS for providing us with the data used in thispaper. We are also grateful for the excellent research assis-tance of Paulo Somaini.Galiani: Professor, Department of Economics, Washington

University in St Louis, Campus Box 1208, St Louis,MO 63130-4899. Phone (314) 935-5670, Fax (314) 935-4156, E-mail [email protected]

Weinschelbaum: Professor, Department of Economics, Uni-versidad de San Andres, Vito Dumas 284, (B1644BID)Victoria, Provincia de Buenos Aires, Argentina.Phone +54-11-4725-7041, Fax +54-11-4725-7010,E-mail [email protected]

The traditional explanation, initiated with thework of Lewis (1954), assumes the existence ofsegmented labor markets in which some workersdo not have access to jobs in the regulated, for-mal sector. These workers are forced to acceptlower-paying informal-sector jobs that usuallyinvolve inferior working conditions. While thisview has become prevalent in the developmentliterature, direct empirical tests of the premisethat informal workers would expect higherwages in the formal sector yield mixed results.For instance, Magnac (1991), Maloney (2004),and Pratap and Quintin (2004) do not find com-pelling evidence of segmentation between theformal and the informal sectors using data fromColombia, Mexico, and Argentina, respectively.It is, nonetheless, certainly plausible that thepresence of unions might segment the labor mar-ket (Layard et al. 1991). However, even if webelieve that unionization has been an importantfactor in determining labor-market segmentationin the past, the pronounced decline of union-ization throughout Latin America over the last

ABBREVIATIONS

GDP: Gross Domestic ProductVAT: Value Added Taxes

821

Economic Inquiry(ISSN 0095-2583)Vol. 50, No. 3, July 2012, 821–838

doi:10.1111/j.1465-7295.2011.00413.xOnline Early publication September 21, 2011© 2011 Western Economic Association International

EDGAR
Realce
EDGAR
Realce
Page 2: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

822 ECONOMIC INQUIRY

two decades seems to be a less satisfactoryexplanation of the phenomenon at the presenttime, especially since informality increased inmost countries in Latin America during the sameperiod (Gasparini and Tornarolli 2006).

Another approach emphasizes the role oflabor-market regulation and taxation as the maincause of informality. A pioneering study byDe Soto (1989) describes how informal work-ers develop their own laws and institutions tomake up for the shortcomings of the officiallegal system. It is certainly true that the sizeof the shadow economy is affected by the bur-den of taxes and social security contributions, aswell as by the many other regulations governingthe official economy. The problem with adopt-ing this explanation of informality unreservedly,however, is that it overlooks the fact that work-ers have nonpecuniary preferences for the sectorthey work for, and these preferences play a rolein determining the equilibrium level of formal-ity of an economy. In particular, workers in theinformal sector are not eligible for socially man-dated benefits.

Three stylized facts characterize informal-ity: (1) small firms tend to operate informallywhile large firms tend to operate formally;(2) unskilled workers tend to be informalwhile skilled ones have formal jobs; (3) ceterisparibus, secondary workers (a worker other thanthe household head) are less likely to operateformally than primary workers. In this paper,we present a model of an economy in whichthe sizes and wages of the formal and infor-mal sectors are endogenously determined thataccounts for these three facts. Our model entailsa continuum of types on both sides of the labormarket (i.e., firms and workers). Firms choosewhether to operate formally or informally andworkers choose in what sector they work. Inequilibrium, managerial talent determines sizedualism at the firm level, and human capitalis the factor that determines whether workersare employed in the formal or informal sec-tor. In our model, both heterogeneous firmsand workers have preferences over the sectorsthey operate and choose optimally whether tofunction formally or informally. By contrast, aprominent feature of the preexisting literatureis the idea that workers’ decisions play no rolein determining the equilibrium of the economy.Thus, our model is the first one in which poli-cies that affect workers’ preferences (as well, ofcourse, as firm preferences) over the formal orthe informal sector such as pension plans, public

health issues, and unemployment and any socialbenefits will affect the sizes and wages of theformal and the informal sectors. This has pro-found consequences for the design of policies,especially social programs.

Our main findings are as follows: First, sec-ondary workers, ceteris paribus, are less likelyto operate formally than primary workers are.This is because, whatever the net benefits ofoperating formally may be for a worker hav-ing a given human capital endowment, they willbe lower if another member of the household isalready in the formal sector. This is observed inthe data. Thus, an increase in the participation ofsecondary workers would tend to raise the levelof informality in the economy.1 We believe thatthis effect could partially explain the increasesin informality in Latin America seen over thelast two decades.2

Second, we show that all the regulations—fixed costs—that increase the cost of operatingformally, both for firms and for workers, influ-ence the equilibrium as they would be expectedto do.3 One important instance is that of socialprograms. Any program that affects the sup-ply to the (formal) informal sector will affectthe equilibrium in our model. For example,social programs that target the poor by con-ditioning eligibility on the labor-market statusof each individual will increase (reduce) thesupply of workers to the informal (formal) sec-tor, increasing the size of the informal sectorin the economy. There is ample empirical evi-dence supporting this effect. The expansion ofthe social safety network since the 1990s inLatin America could then also partially explainthe increases in informality seen over the lasttwo decades.

1. Female labor force participation has increased sub-stantially in Latin America in the last five decades. Itincreased from approximately 20% in 1960 to 35% in2000 and to almost 40% in 2008 (World DevelopmentIndicators).

2. For example, in a previous version of this paper wesimulated the rates of formality for Argentina, which isthe Latin American country where informality increased themost during the last two decades. We assumed that only thedemographic composition of the labor force changed overtime (and that the formality rates of all these groups areunaffected by these shifts in the labor force composition)and find that approximately 20% of the rise in informalitybetween the early 1970s and the 1990s could be directlyaccounted by the reduction of the share of the labor forcerepresented by primary workers during that period (Galianiand Weinschelbaum 2007).

3. The same is true for payroll taxes—or any taxationthat affects labor demand in the formal sector of theeconomy—and enforcement.

Page 3: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

GALIANI & WEINSCHELBAUM: INFORMALITY, HOUSEHOLD DECISIONS, LABOR MARKETS 823

Third, as in the previous literature that fol-lowed the lead of Rauch (1991), size dualism ispresent in the equilibrium of the economy wemodel: small firms operate informally whereaslarge firms operate formally. Finally, workerswith low human capital work in the informalsector whereas those with high human capitalwork in the formal sector. Thus, distributions ofboth human capital and managerial talent affectthe equilibrium of the economy. A poor econ-omy with low levels of human capital and lim-ited managerial capabilities is likely to exhibithigh levels of informality. This is also observedin the data.

The rest of the paper is organized as follows.Section II reviews the previous literature oninformality. Section III presents evidence on thenature of informality in Latin America, whichserves as a basis for the analysis offered inthis paper. Section IV portrays the setup of ourmodel, while Section V reports on a set of com-parative static exercises. Section VI studies theeffect of incorporating secondary workers intothe labor market. Finally, Section VII presentsour conclusions.

II. LITERATURE REVIEW

Dualistic models of labor markets have per-vaded the literature since the seminal work ofLewis (1954). Firms and workers are assumedto comply with regulations in the formal sectorbut not in the informal sector. Wages in the for-mal sector are set exogenously and above thecompetitive equilibrium level. The labor mar-ket is then segmented, with only some workersobtaining jobs in the formal sector. The clas-sic model presented in this literature is the onedeveloped by Harris and Todaro (1970).

An excellent formalization of this type ofmodel is provided by Rauch (1991), who devel-ops a model in which the prevailing real wage inthe formal sector is exogenous. Firms can refrainfrom paying the exogenously determined realwage provided they operate on a scale smallerthan a given detection threshold. The productmarket is assumed to be competitive. Workersare homogenous but firms differ in terms ofmanagerial talent, which in turn determines firmsize as discussed by Lucas (1978). Thus, in equi-librium, there is a break in the size distributionof firms. Informal wages are lower than formalwages. However, unemployment does not ariseas in the study by Harris and Todaro (1970)because, in this model, workers’ utilities are not

equalized in equilibrium. Some workers obtainjobs in the formal sector while the others areunderemployed (in the sense that they could beearning more in a competitive equilibrium) inthe informal sector.4 Note, however, that theemphasis in this paper is on size dualism —that is, the existence of a difference in sizebetween the smallest formal-sector firm and thelargest informal-sector firm. By contrast, theearlier theoretical literature focused on labor-market dualism, which is defined as the exis-tence of a difference between the formal-sectorwage and the informal-sector wage for econom-ically identical employees.

Fortin et al. (1997) extends Rauch’s modelby introducing corporate profit and payroll taxesand by assuming that the marginal cost of taxand regulation evasion increases with the size ofthe firm. They derive three forms of dualism thatare consistent with labor-market segmentation:scale, wage, and evasion. In equilibrium, smallfirms in the informal sector benefit by payinglower wages and taxes than firms in the formalsector do, but incur the risk of being penalizedfor noncompliance with such labor regulations.

Thus, in this type of model, homogenousworkers are paid less in the informal sector onlybecause of the presence of labor-market regu-lations that affect the profit function of firmsoperating in competitive product markets.

A significant departure from dualism is foundin the work of Amaral and Quintin (2006), whoexplicitly model the labor market as a competi-tive market. They assume that both workers andentrepreneurs are heterogeneous. As in the pre-vious models, firms differ in managerial talent.Firms in the formal sector pay taxes on prof-its but have access to the credit market, whilethose in the informal sector do not pay taxesand have no access to credit. Therefore, in equi-librium, large firms are formal and small firms

4. Rauch’s model is consistent with many of thehypotheses advanced to account for the formal-sector wagefloor. If the legal minimum wage set by the government isthe binding factor, then, clearly, given the limited adminis-trative resources of third-world governments, the number ofworkers protected by the law can be maximized by focusingenforcement efforts on large firms. If it is union bargainingpower that sets the wage floor, big firms will once again bethe focus since, as is well known, unions concentrate theirorganizing efforts on large companies. Even if one believesthat the efficiency wage hypothesis is the relevant factor, andif the efficiency wage entails minimizing the cost of moni-toring workers (as described by Shapiro and Stiglitz 1984),then it is reasonable to think that small firms will be able tomonitor almost costlessly and therefore do not need to paythe efficiency wage.

Page 4: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

824 ECONOMIC INQUIRY

are informal. Workers only derive utility fromthe wages they receive at work; they have nointrinsic preferences for the sector they workfor (i.e., they do not have nonpecuniary prefer-ences). Thus, there is a differentiated demand forformal and informal employment but labor sup-ply is the same in both sectors. In other words,there is only one labor market in this econ-omy. Therefore, and contrary to the previousliterature, similar workers earn the same amountin the formal and informal sectors. Within thisframework, Quintin (2008) models the impactof limited contract enforcement on the size ofthe informal economy.

Other interesting and related papers areStraub (2005) and de Paula and Scheinkman(2009, 2010). Straub (2005) develops a modelwhere formal firms pay an entry fee to becomeformal. Formal firms have access to a publicgood, while informal firms do not. The pub-lic good is identified with access to a bettercredit market (lower interest rates). In this caseas well, in equilibrium, large firms are formaland small firms are informal. The labor mar-ket is not modeled. For their part, de Paula andScheinkman (2009) assume that firms pay taxeson sales when they operate in the formal sec-tor but, on the other hand, pay lower interestrates on capital. Hence, in equilibrium, there isalso a break in the size distribution of firms. Asin the previous models, firm size is determinedby managerial talent. Workers are homogenousbut, in this model, they are paid the same in bothsectors. Informality is defined as tax avoidance;they test the model with Brazilian data. de Paulaand Scheinkman (2010) analyze the role of valueadded taxes (VAT) in transmitting informality.Through this channel, they find that the level ofinformality of a firm is correlated to the infor-mality of the firms from which it buys inputsor sells its output. They confirm this result alsousing Brazilian data.

A major feature of the previous literature isthat workers’ decisions are seen as playing norole in determining the equilibrium. On the onehand, it is only the firms that choose optimallywhether to operate formally or informally whilehomogenous workers are constrained to obtainthe best jobs available. On the other hand, bothfirms and workers act unconstrained, but the lat-ter do not derive nonpecuniary utility from thesector they work for. Firms have differentiateddemands for formal and informal employmentbut workers are indeed indifferent about whichsector they work for. Thus, there is only one

labor market in the economy. By contrast, in ourmodel, workers do have nonpecuniary prefer-ences for the sector they work for. Firms choosewhether to operate formally or informally, andworkers choose in which sector to work. There-fore, there are two labor markets in the economyin which both firms and workers act uncon-strained. At equilibrium, managerial talent deter-mines size dualism at the firm level, and humancapital is the factor that divides workers intothose who work in the formal sector and thosewho work in the informal sector. In our model,any characteristic of the economy that changesworkers’ (firms) preferences over the formal orthe informal sector affects the size and wages ofthe formal and the informal sectors.

III. INFORMALITY IN LATIN AMERICA

Formality is usually defined on the basis ofthe right to exercise socially mandated benefitssuch as health insurance and pensions. Wefollow the trend exhibited in the literature anddefine formality on the basis of whether or notan employee pays social security taxes (see,among others, Gasparini and Tornarolli 2006).

In the literature, workers are usually cate-gorized into three groups: formal employees,informal employees, and independent work-ers. Independent workers are distinguished fromemployees for two reasons. First, independentworkers typically differ in their motivations andskills. They tend to be individuals who dis-play entrepreneurial skills and who attach a highvalue to the nonpecuniary benefits of indepen-dent work (see, among others, Blanchflower andOswald 1998). Second, the surveys most com-monly used as a source of data on the laborforce in Latin America collect information onemployees’ social security contributions but noton contributions by business owners or the self-employed. Consequently, it is not possible toestablish whether or not self-employed personsare operating formally or informally.5 We havetherefore abstracted from the decision to workas an independent worker and focus our analy-sis on firms’ and salaried workers’ decisions asto whether or not to operate within the formalsector of the economy.

5. Even if most self-employees operate informally,which it is unlikely to be the case—as illustrated, forexample, by the case of independent professionals—theshare of self-employees in the labor force in LAC is smallerthan the share of salaried workers that operate informally(Gasparini and Tornarolli 2006).

Page 5: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

GALIANI & WEINSCHELBAUM: INFORMALITY, HOUSEHOLD DECISIONS, LABOR MARKETS 825

TABLE 1Formality Rates among Salaried Workers (%)

Country Mean Males Females Household Heads Spouses Other Household Members

Argentina 57.5 60.2 53.9 65.2 59.2 43.7Bolivia 27.7 26.9 29.3 33.5 33.4 15.0Brazil 67.7 68.5 66.7 70.9 68.3 61.7Chile 87.3 90.1 83.0 92.2 84.3 82.4Ecuador 30.3 26.7 37.9 34.2 44.0 20.8El Salvador 54.6 47.2 69.1 52.6 67.7 51.9Guatemala 41.7 41.3 42.5 44.3 43.8 38.1Mexico 42.0 40.6 44.7 43.1 47.5 38.5Paraguay 27.6 26.8 28.8 33.7 36.9 17.8Peru 31.0 32.7 28.3 39.8 35.3 23.3Uruguay 75.4 78.2 72.3 81.5 76.1 65.5Venezuela 60.4 57.6 64.7 67.5 66.9 50.5

Average 50.3 49.7 51.8 54.9 55.3 42.4

Notes: Statistics from the latest survey available for each country. All surveys are for the years 2002, 2003, or 2004. Aworker is defined as formal if he/she pays social security taxes. The same definition is applied to all countries. For Argentina,we excluded the individuals that work in the social program Jefes y Jefas de Familia. Sampling weights are used to computeall the statistics by country in this table. Average statistics are unweighted averages of country rates.

Source: Prepared by the authors based on data from the Socio-Economic Database for Latin American and the Caribbean(CEDLAS).

In Table 1 we present evidence on formalityrates among salaried workers in Latin America.The data are taken from household surveysfor 12 countries (those listed in Table 1) inwhich the existence or absence of social securitycontributions is registered for each employee inthe sample. All the surveys used here are for theyears 2002–2004. On average, approximatelyhalf of all salaried workers hold informal jobs.Formality rates also appear to display a largevariance among countries—ranging from 87%in Chile to 27.6% in Paraguay—and tend toincrease with development. The richest countriesin this sample—that is, Argentina, Brazil, Chile,and Uruguay—all have formality rates above50%, while all the poorest countries exceptEl Salvador (Bolivia, Ecuador, Guatemala, andParaguay) exhibit formality rates among salariedworkers well below that level.

On average, no appreciable differences areobserved in the degree of formalization neitherbetween males and females nor between headsof household and spouses. At first glance, thismight seem counterintuitive, but it is actuallythe natural result of sample selection. Femaleswho participate in the labor market are, on aver-age, more educated than males.6 In Table 2 werestrict the sample to households where both

6. However, this is not true at the margin. Thus, therecent increase of female labor supply observed throughLatin America is capable of causing an increase in themean levels of informality. This is specially so if secondary

members (i.e., the head of the household andthe spouse) are salaried employees and presentformality rates for household heads and spouseswhere each of these groups is divided intothree educational categories: low-, medium-, andhigh-education level. Our findings indicate thatthe likelihood of a salaried worker being for-mally employed increases with the individual’slevel of education. Conditioning on the level ofeducation, heads of household are more likelyto be formally employed than their spouses are,but among highly educated individuals, there areno differences at all.7

We further investigate the results in Table 2.When the head of the household and the spouseparticipate in the labor market, the decisioncould be for both to work in the formal sec-tor (FF), for one to work in the formal and theother in the informal sector (FI), or for bothto be employed in the informal sector (II). Weestimate a multinomial logit model for thesethree potential outcomes focusing on the effectof human capital on the likelihood of observ-ing any of them. In the econometric model weinclude a full set of educational dummies for

workers, ceteris paribus, have lower incentives to operateformally than primary workers.

7. Similar results hold even when we do not imposethis sample restriction in the analysis (Galiani andWeinschelbaum 2007).

Page 6: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

826 ECONOMIC INQUIRY

TABLE 2Formality Rates among Households in Which Both Members Are Salaried Workers, by Household

Status and Educational Level (%)

Household Heads Spouses

LowEducation

MediumEducation

HighEducation

LowEducation

MediumEducation

HighEducation

Argentina 61.2 77.0 84.0 28.8 64.5 82.3Bolivia 8.6 28.3 78.6 5.7 28.2 75.9Brazil 71.2 88.0 92.9 53.8 82.6 92.2Chile 89.9 96.5 97.0 72.9 87.4 93.3Ecuador 21.8 62.3 79.6 16.0 51.8 78.1El Salvador 50.8 81.6 92.8 54.7 85.2 90.0Guatemala 44.5 73.5 79.2 31.3 80.6 85.0Mexico 40.2 62.8 74.1 35.8 63.2 71.7Paraguay 30.9 69.1 66.9 16.6 50.5 64.4Peru 7.9 47.9 67.2 3.7 25.1 63.6Uruguay 80.8 92.5 95.3 63.6 88.0 96.4Venezuela 67.4 80.2 91.6 56.9 85.4 89.8

Average 47.9 71.6 83.3 36.6 66.0 81.9

Notes: See Table 1. The low-education category includes all individuals that have not completed high school; the medium-education classification includes individuals who have exactly completed high school while the high-education categoryincludes all individuals with more than high school (i.e., incomplete or complete tertiary/college education).

Source: See Table 1.

each individual.8 We also include as regres-sors the age and the age squared of both thehead of the household and the spouse, country-fixed effects, and a dummy variable indicatingwhether the household resides in an urban area.We summarize the results in Figure 1. As canbe seen, the probability that the head of thehousehold and his/her spouse work both for-mally increases monotonically in the educationof the couple while the probability that both ofthem are informal decreases monotonically intheir education. Interestingly enough, the like-lihood that one works formally while the otheroperates informally first increases with the edu-cation of the couple and then decreases.

Finally, in Table 3 we present formality ratesby firm size. Overall, small firms tend to oper-ate informally while large firms are substan-tially more likely to operate formally. Thus, insummary, informality is highly concentrated insmall firms, and the likelihood that a worker isemployed in the informal sector of the econ-omy decreases as his/her level of education

8. For each individual, the omitted education categoryis no schooling while the schooling dummies included inthe model are: incomplete primary, complete primary, high-school dropout, high school, incomplete higher education,and complete higher education (i.e., tertiary or collegedegree). The model is estimated using the sample in Table 2.Regression results are available upon request.

rises. This relation between education and for-mality is more pronounced among spouses thanamong households’ heads. In other words, anunskilled or semiskilled spouse is more likely tobe employed informally than an unskilled headof household is. In the next two sections wedevelop a model that accounts for all these facts.

IV. SETUP OF THE MODEL

We model an economy with three types ofeconomic agents: a continuum of firms, a contin-uum of workers, and the government. Firms areheterogeneous in their managerial talent. Theymaximize profits by choosing whether to oper-ate formally or informally and by hiring labor.Workers are heterogeneous in their endowmentof human capital. They maximize utility bychoosing whether to work in the formal or infor-mal sector. The government, whose behavioris not modeled, collects payroll taxes, providesformal workers and their families with non-pecuniary benefits, regulates formal activities,and conducts activities designed to detect andpenalize informal operations in the economy.

A. Firms

Firms are assumed to produce a homogenousgood that is traded in a perfectly competitive

Page 7: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

GALIANI & WEINSCHELBAUM: INFORMALITY, HOUSEHOLD DECISIONS, LABOR MARKETS 827

FIGURE 1Predicted Probabilities from Multinomial Logit Model, by Education Level

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

No schooling Primary Secondary Higher Education

Both Informal Formal–Informal Both Formal

Notes: This figure is constructed using the estimated coefficients of the multinomial logit model. In order to represent theresults graphically, we settled the age of both household members in 46 and the Mexico fixed-effect and the urban dummy inone. Furthermore, for the purpose of the figure, we also assume that both members of the household have the same educationallevel.

TABLE 3Formality Rates among Salaried Workers, by

Firm Size (%)

Country MeanSmallFirms

Rest ofFirms

LargeFirms

Argentina 57.5 26.5 65.3 87.7Bolivia 27.7 6.3 35.5 57.4Brazil 67.7 35.4 83.1 —Chile 87.3 70.1 88.3 94.6Ecuador 30.3 5.9 23.8 44.5El Salvador 54.6 7.0 35.3 83.6Guatemala 41.7 6.6 47.6 76.2Mexico 42.0 12.1 57.4 —Paraguay 27.6 8.2 39.1 62.4Peru 31.0 4.4 21.8 55.2Uruguay 75.4 39.0 74.9 96.2Venezuela 60.4 21.7 75.8 —

Average 50.3 20.3 54.0

Notes: See Table 1. Firm size is defined as follows(depending on the distinctions made in the available data):Small firm: fewer than four or fewer than five workers; restof firms: more than four or more than five workers. Largefirms: more than 40 or more than 50 workers.

Source: See Table 1.

market using managerial talent (m) and homoge-nous units of labor (l).9 Using an approach

9. For the sake of simplicity, we normalize the price ofthe good to 1.

similar to that used in the previous literature,we assume that there is a continuum of firmsindexed by their innate managerial talent anddistributed with density function g(m) with sup-port in R+.

Each firm in the economy chooses whetherto operate formally or informally and decideshow many units of labor it will hire to produceits output. If a firm operates in the formalsector of the economy, its profit function isgiven by:

�f(m) = f (m, l) − wfl(1 + t) − τ

where f (m, l) is the production function, wf isthe prevailing wage per unit of labor paid toworkers in the formal sector (i.e., firms are wagetakers), t is the payroll tax levied on firms, andτ is a fixed cost incurred by firms that operateformally.

There are a large number of studies in the lit-erature documenting various formal-sector entrycosts. The best-known example is the discus-sion in De Soto’s The Other Path of the cost ofstarting a business in Lima, Peru. De Soto findsthe complexity and stifling nature of Peruvianlaws and regulations to be a powerful disincen-tive for formality and hence a stimulus for theexpansion of the informal sector (see also Stoneet al. 1996).

Page 8: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

828 ECONOMIC INQUIRY

If, on the other hand, a firm operates infor-mally, then its expected profit function isgiven by:

�i(m) = (f (m, l) − wil)(1 − q)

where wi is the prevalent wage per unit of laborpaid to workers in the informal sector and q isthe probability that a firm is caught operatinginformally, in which case it will be closed andwill lose its current profits.

In the following discussion, we assume thatthe production function is Cobb-Douglas withconstant returns to scale, thereby presentingdiminishing returns to labor:

f (m, l) = mβl1−β

Firms choose l in order to maximize profits.Under the assumptions made, the derived labordemand for a firm with managerial talent m is:

lf = m((1 − β)/wf(1 + t))1/β

if it operates formally and:

li = m((1 − β)/wi)1/β

if it operates informally. Note that both demandfunctions are linear in managerial talent. Thus,the size of a firm (assuming that it doesnot switch sectors), measured by l, is linearlyincreasing in m.

Very small firms (i.e., those with managerialtalent m close to 0) operate informally becauseof the fixed cost τ that they would have to incurin order to operate formally. As in the studyby Rauch (1991), we now show that in ourmodel there is size dualism: small firms operateinformally while large firms operate formally.When there are firms operating formally andthe profit functions satisfy the following con-dition, there is a unique cutoff point in firmsize:

d�f(m)/dm > (d�i(m)/dm ∀m.(1)

Condition (1) is equivalent to:

(1/wf(1 + t))(1−β)/β > (1 − q)(1/wi)(1−β)/β.

(2)

Hence, in equilibrium, it is necessary thatinequality (2) be satisfied in order to have firmsoperating formally. Therefore, we establishedthat, in equilibrium, whenever there are both for-mal and informal firms, there is a unique cutoffpoint (m) that satisfies:

�f(m) = �i(m).

Thus, firms with managerial talent m ≤ mchoose to operate informally while firms withmanagerial talent m > m choose to operateformally.

The demand for formal labor is then:

lfd(wf, wi)

=∫ ∞

m

m((1 − β)/wf(1 + t))1/βg(m)dm.

while the demand for informal labor is:

lid(wf, wi) =

∫ m

0m((1 − β)/wi)

1/βg(m)dm

Therefore, the total demand for labor isgiven by:

Ld(wf, wi)

=∫ m

0m((1 − β)/wi)

1/βg(m)dm

+∫ ∞

m

m((1 − β)/wf(1 + t))1/βg(m)dm.

In looking at how a change in wages willaffect total labor demand, it is useful to breakthe impact down into direct and indirect effects.The direct effect is the derivative of total labordemand with respect to wages, holding the cut-off point constant. The indirect effect is thederivative of total labor demand with respectto wages that operates through changes in thecutoff point, holding constant the integrandfunctions.

B. Workers

Workers are endowed with l units of homoge-nous labor to sell in the market. There is a con-tinuum of workers indexed by their endowmentof labor and distributed with density functionh(l) with support in R+. Thus, wages are set forunits of substitutable labor. Each worker has tochoose whether to sell his/her labor to a firmthat operates formally or to a firm that operatesinformally.10 Workers cannot split their units oflabor between different firms. In the first case,the worker receives wf per unit of labor plusmandated social benefits. In the second case, theworker receives only wi per unit of labor.

10. Olsen (1993), cited in Gruber (2000), providessupportive evidence of the presence of worker’s self-selection in the United States. He finds that workers withgreater than expected health needs (wives whose husbandsdo not have health insurance, and who have less healthychildren) self-select into firms that provide health benefits.

Page 9: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

GALIANI & WEINSCHELBAUM: INFORMALITY, HOUSEHOLD DECISIONS, LABOR MARKETS 829

The utility of a worker who chooses to workin the formal sector is11,12:

Uf(wfl, HI ) = wfl + HI − γ

where HI are the nonpecuniary benefits that, forsimplicity’s sake, we equate with a government-provided health insurance package which, antic-ipating the analysis in the next section, weassume covers all members of the worker’sfamily.13,14

Workers also incur a fixed cost γ for work-ing in the formal sector. This includes the directcosts, which are related to (see, among others,De Soto 1989)15 but also, and perhaps moreimportantly, indirect costs such as the unem-ployment insurance and other subsidies, likecash transfers, that a worker would have to dis-regard when he/she accepts a formal job.

When an informal firm is detected, its work-ers do not receive their payment. The expectedutility of a worker who chooses to work in theinformal sector is:

Ui(wil, HI ) = wil(1 − q).

11. We do not tax workers’ formal income becausewe already taxed firms on their payrolls. As usual, theequilibrium of the economy is unaltered if we instead taxworkers’ labor earnings or if both firms and workers areboth taxed. It only matters the total tax on labor earnings.

12. We assume that the utility function is linear inorder to abstract from risk-pooling behavior when, as inSection VI, we allow for situations in which more thanone member per family works. Naturally, this is a theo-retical abstraction and we are by no means attempting tosuggest that risk aversion is not a consideration when house-holds choose the portfolio of sectors where their memberswork.

13. Although, for the sake of simplicity, we equatemandated social benefits with a health insurance package(that provides better health coverage than public hospitals),there are other mandated benefits associated with the formaljobs that affect workers’ payoffs similarly. Examples arepensions and family allowances. Formal workers normallyreceive pensions from retirement age on, and, after theydie, it is common for their surviving spouses, if they arenot receiving a pension themselves, to start to receive apension until their own death. Thus, a formal job providesthe entire family with an old-age pension. Family allowancesredistribute money among formal workers as a function offamily size. Only one worker per family is typically entitledto this type of cash transfer even if both spouses work inthe formal sector.

14. There is a large literature in the United States thatshows that health insurance is an important determinant ofjob mobility (Gruber and Madrian 2004). Madrian (1994),for example, estimates that health insurance reduces mobilityacross jobs for as much as 25%. See also Gruber (2000).

15. An important direct cost associated to work in theformal sector could be transportation costs. These costscould be particularly high in large urban areas whereunskilled workers tend to reside further away from whereformal firms tend to agglomerate.

When there are workers in both the formaland informal sectors and the utility functionssatisfy the following condition, then there is aunique cutoff point that determines that workerswith low levels of human capital will be in theinformal sector and that those with high levelsof human capital will be employed in the formalsector:

dUf(l)/dl > dUi(l)/dl ∀ l.(3)

Condition (3) holds if and only if16,17:

wf/wi > (1 − q).(4)

The cutoff point is the level of l that makesworkers indifferent as to whether they work inthe formal or informal sector. We denote thislevel of l by l, which is determined by:

Uf(wfl, HI ) = Ui(wil, HI )

or

wfl + HI − γ = wil(1 − q)

and solving for l we obtain:

l = ((γ − HI )/(wf − wi(1 − q))).

Note that workers with l > l prefer to workin the formal sector,18 thus workers with l ≤l operate informally while workers with l < loperate formally.

The supply of formal labor is:

lfs(wf, wi) =

∫ ∞

l

lh(l)dl

while the supply of informal labor is:

lis(wf, wi) =

∫ l

0lh(l)dl.

Accordingly, the total supply of labor is:

Ls(wf, wi) =∫ l

0lh(l)dl +

∫ ∞

l

lh(l)dl = Ls

16. Note that the formal-sector wage premium is notsigned. This result also holds if we instead assume that whenan informal firm is detected, workers receive their paymentwith a positive probability but without certainty.

17. Juarez (2008) presents evidence, exploiting exoge-nous variability in access to free health care and prescriptiondrugs in Mexico City, that informal female workers are paidapproximately 20% more than similar formal female work-ers in order to compensate for the lack of fringe benefits ininformal jobs.

18. Note that if condition (4) holds and HI ≥ γ, allworkers would prefer to work in the formal sector. Addi-tionally, as is plausible, if the worker’s utility is such that itassigns positive utility to mandate social benefits only aftera minimum level of income is achieved (wl), then condition(4) guarantees a cutoff point regardless of the sign of HI -γ.

Page 10: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

830 ECONOMIC INQUIRY

Because labor is inelastically supplied, wagesdetermine the amount of labor that is suppliedto the formal or informal sector, but not the totalquantity supplied.

C. Equilibrium

We restrict our analysis to equilibria wherethere are firms and workers operating in boththe formal and informal sectors. An equilibriumis a (wf; wi; lf

s; lis; lf

d; lid) vector such that there is

market clearing in both the formal and informallabor markets:

lfs(wf, wi) = lf

d(wf, wi)(5)

lis(wf, wi) = li

d(wf, wi)(6)

and where wages satisfy inequalities (2) and (4).A very interesting characteristic of this equi-

librium is that it does not sign the formal-sectorwage premium: wf can be higher, equal, orlower than wi (see inequality (4)). This resultcontrasts with the previous literature, where apositive formal-sector wage premium is identi-fied with the presence of labor-market dualismin the economy, while the absence of a wagepremium is taken to be indicative of the preva-lence of a competitive environment in the labormarket.

To represent an equilibrium for this economygraphically, we first derive the locus of wageswhere total labor demand is equal to total laborsupply. That is, the wage pairs (wf, wi) aresuch that Ld(wf, wi) = Ls. From this condition,we derive an implicit function H(wf, wi) = 0that we will denote wH

f (wi), which yields adownward-sloping curve, as we assume that thederivative of total labor demand with respect toboth wages is negative (Figure 2). Any equilib-rium in this economy must lie somewhere on

FIGURE 2Locus of Possible Equilibria

B

A

wi

wf

this curve.19 While total labor demand and sup-ply are equal along this curve, their compositionis not. Each point maps to a different (lf

s; lis;

lfd; li

d) vector. For example, for a high informalwage (and hence a low-formal wage) such aspoint A, firms would demand relatively moreformal workers than informal ones. However, atthat relative wage, a large proportion of workerswould prefer to be informal rather than formal.Thus, a point such as A would not be an equi-librium of that economy. The opposite situationwould arise at a point such as B.

In the same figure, we can also draw the func-tions wf(wi) showing the wage pairs that wouldalign labor demand and supply within each sec-tor of the economy. Differentiating Equation(5) with respect to both wages, we obtain:

−∂lfs(wf,wi)

∂wi− ∂lfd(wf,wi)

∂wi

∂lfs(wf,wi)∂wf

− ∂lfd(wf,wi)

∂wf

= dwf

dwi(7)

which can easily be seen to be positive. Dif-ferentiating (6) with respect to both wagesgives us:

−∂lis(wf,wi)

∂wi− ∂lid(wf,wi)

∂wi

∂lis(wf,wi)∂wf

− ∂lid(wf,wi)

∂wf

= dwf

dwi(8)

which is also positive.Given that total labor supply is inelastic, we

know that:

∂lfs(wf, wi)/∂wi = −∂li

s(wf, wi)/∂wi

and

∂lfs(wf, wi)/∂wf = −∂li

s(wf, wi)/∂wf.

And as we assumed that total labor demand isnegatively sloped with respect to both formaland informal wages, it holds that:

∂lfd(wf, wi)/∂wi < −∂li

d(wf, wi)/∂wi

and

−∂lfd(wf, wi)/∂wf > ∂li

d(wf, wi)/∂wf.

19. This always holds if the direct effects predomi-nate over the indirect effects. If there is a case wherethe indirect effect is positive and predominates over thedirect effect, wH

f (wi) is not necessarily a function (i.e.,for a given informal wage there could be different for-mal wages that satisfy wH

f (wi)). If this were the case,there could be multiple equilibria in the economy thatmight generate interesting economic effects; a discussionof such effects extends beyond the scope of this paper,however.

Page 11: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

GALIANI & WEINSCHELBAUM: INFORMALITY, HOUSEHOLD DECISIONS, LABOR MARKETS 831

FIGURE 3Equilibrium Wages

lFD+ lI

D = lFS+ lI

S

wi

wf

lID = lI

S

lFD = lF

S

wi*

wf*

Therefore, it follows that (8) > (7). Figure 3presents these three curves together andgraphically illustrates an equilibrium in thiseconomy.20

V. COMPARATIVE STATICS

We now explore how the equilibrium wedescribed in the previous section changes inresponse to changes in the parameters.21 We willthen look at how the equilibrium of the economychanges when we allow for the incorporation ofsecondary workers.

A. Changes in Human Capital

We will first look at how different changesin the distribution of worker’s human capitalaffect the equilibrium of the economy. In thisexercise, we hold total human capital constant.The distribution of human capital can changein three qualitatively different ways. First, thenew distribution can maintain the total amountof l both above and below the original cutoffpoint that divides workers between the informaland formal sectors as a function of their humancapital. Second, the new distribution can shift ltoward workers above the original cutoff point.Third, the new distribution can shift l towardworkers below the original cutoff point.

20. All the equilibrium points we analyze in this paperare in the area defined by inequalities (1) and (4). We havenot graphed these two relationships in order to keep thefigure simple.

21. The reader can read a more complete analysis ofthese comparative static results in the previous version ofthis paper (Galiani and Weinschelbaum 2007).

In the first case, the equilibrium of the econ-omy remains unaltered. At prevailing wage lev-els, workers’ choices would remain unchangedand the aggregate formal and informal labor sup-ply and demand would remain balanced.

If, on the other hand, l is redistributed towardworkers in the formal sector, then the curveof equilibrium points in that sector shown inFigure 3 shifts downward, whereas the curvefor the informal sector shifts to the right. Asthe total supply of labor represented by work-ers above the original cutoff point increases,at prevailing wage levels, there is excess sup-ply in the formal sector. This means that, fora given informal wage, equilibrium in the for-mal sector is only achieved if formal wages godown. The same reasoning applies to the infor-mal sector. Finally, because neither the totalamount of human capital nor the distribution ofmanagerial talent change, the function wH

f (wi)representing the wage levels that align the totalsupply of labor with the total demand for laborremains the same. We illustrate these changesin Figure 4. Thus, in the new equilibrium, theinformal wage increases while the formal wagedecreases. Managerial decisions also change. Atthe new wage levels, the managerial talent cutoffpoint is also lower than in the previous equi-librium. Therefore, more firms choose to oper-ate formally. Moreover, the firms that originallywere operating formally increase their total labordemand while those that remain in the informalsector reduce their total labor demand. Thus, achange in the distribution of human capital suchas this increases the size of the formal sector.The opposite happens if l is redistributed towardworkers in the informal sector.

FIGURE 4Comparative Statics

lFD+ lI

D = lFS+ lI

S

wi

wf

lID = lI

S

lFD = lF

S

wf**

wf*

wi* wi

**

Page 12: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

832 ECONOMIC INQUIRY

We now study what happens when we redis-tribute managerial talent among firms but wehold total managerial talent constant. Mutatismutandis, the argument is similar to the one pro-vided above for the changes in worker’s humancapital. Given that labor demand is linear inmanagerial talent (when firms do not switchfrom one sector to the other), a redistributionof managerial talent among firms belonging tothe same sector does not change either total orsectorial labor demand. A redistribution of man-agerial talent from one sector to the other insteadchanges both total labor demand and its compo-sition between sectors. When m is redistributedtoward firms in the formal sector, in the newequilibrium, there are more workers and laborin the formal sector. The opposite happens whenm is redistributed toward firms in the informalsector.

B. Changes in Taxes and Enforcement

As expected, an increase of the payroll taxrate t induces firms to reduce their demand forformal labor. In the new equilibrium the shareof informal labor would increase. Naturally, theopposite happens if payroll taxes are reduced.

An increase in the probability that a firmis detected operating informally (q) lowers thecutoff point of formality as a function of man-agerial talent, which implies that more firmswould choose to operate formally, while fewerfirms would choose to operate informally. Theworkers’ cutoff point for formality as a func-tion of human capital also decreases, inducingmore workers also to shift into the formal sec-tor. Thus, in the new equilibrium, the size ofthe formal sector increases. Again, the oppositeoccurs when q is reduced.

C. Changes in the Cost of Operating Formally

Finally, we study what happens to informal-ity when the cost of operating formally changes.If the cost to firms of operating formally (τ)declines, at prevailing wage levels, the util-ity of operating in the formal sector increases,whereas the utility of operating in the informalsector remains constant. Thus, the formality cut-off point for firms therefore declines, more firmsoperate formally, increasing the relative wageof the formal sector and attracting labor to it.Thus, in the new equilibrium, the size of theformal sector increases. The opposite happensif the cost to firms of operating formally (τ)increases.

If the workers’ fixed cost of operating for-mally (γ) is lower (or the nonpecuniary ben-efits of working in the formal sector HI arehigher), at prevailing wage levels, the utility ofworking in the formal sector increases whilethe utility of working in the informal sectorremains constant. Hence, there is a reductionin the formality cutoff point for workers. Thus,in the new equilibrium, the size of the formalsector increases. The opposite happens if theworkers’ fixed cost of operating formally (γ)increases.

This last result is quite relevant to our mainargument in this paper. Note that it also includesall the policies that affect workers’ preferencesfor one or the other sector. For example, a reduc-tion in the quality of health insurance pack-ages or a deterioration of the pension systemwould induce workers to increase their pref-erences over the informal sector. Thus, these“negative” changes in policies will increase thesize of the informal sector in the economy,and certainly, they likely played a role in therecent increase of informality in Latin Amer-ica, especially since the debt crisis of the early1980s.

As we discussed it earlier, an increase in theworkers’ costs to operate formally might be theresult of indirect costs such as the unemploy-ment insurance and other subsidies, like cashtransfers, that a worker would have to disre-gard when he/she accepts a formal job. In LatinAmerica, there is ample evidence that socialprograms aiming to alleviate poverty induceworkers to move to informality. For example,Gasparini et al. (2007) present evidence show-ing that a social program in Argentina that pro-vides cash transfers to unemployed householdheads induces its beneficiaries to increase sub-stantially the likelihood of working in the infor-mal sector. Using matching techniques, theyestimate that beneficiaries of this program are5 percentage points more likely to operate infor-mally than a similar control group.22 Camachoet al. (2009), using panel data, investigate theeffect of the expansion of a noncontributoryhealth insurance program for the poor in Colom-bia on the likelihood that a worker operatesinformally. They find that this program increasesthe probability of working in the informal sectorby 4 percentage points.

22. Using instead a difference-in-differences estimator,Jauregui (2009) estimates an even larger effect for thebeneficiaries of this program on their likelihood of workinginformally.

Page 13: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

GALIANI & WEINSCHELBAUM: INFORMALITY, HOUSEHOLD DECISIONS, LABOR MARKETS 833

VI. INCORPORATING SECONDARY WORKERSINTO THE LABOR MARKET

We will now turn our attention to what hap-pens to the equilibrium when it is subjected toa major shock, which we produce by allowinghouseholds to have a second worker in the labormarket. The question of differential incentivesbetween first and secondary workers has beenpreviously explored in labor economics. Gruberand Madrian (2004) provide a review of the dif-ferentiated effects of health insurance on laborsupply and job mobility in the United States.They show that access to health insurance isan important factor in the labor supply deci-sion of secondary workers. However, all thiswork focuses only on the household decisions,ignoring the equilibrium effects of those deci-sions. In our model, instead, heterogeneous firmsand heterogeneous workers interact in determin-ing the equilibrium of the economy. Therefore,we study the effects of the presence of sec-ondary workers in the sizes and wages of theformal and the informal sector instead of onlyconsider the household decisions taking equi-librium variables as given. Thus, our paper isthe first study in which workers’ decision playsa role in shaping the equilibrium size of theformal and informal sectors and hence, wherethe differential incentives between primary andsecondary workers affect the equilibrium of thelabor market. Moreover, the difference of incen-tives between first and secondary workers hasnot been studied before in the informality liter-ature. Let us consider first the household deci-sion involved in this scenario. Throughout thissection, we assume that both workers in thesame household have the same level of humancapital.23 When two members of a householdparticipate in the labor market, the decisioncould be for both to work in the formal sector,for one to work in the formal and the other inthe informal sector, or for both to be employedin the informal sector. We will compute theexpected utility for each of these possibilitiesand then analyze the corresponding householdchoices.

The utility of a household that has twoworkers in the formal sector is:

Uff(wfl, HI ) = 2wfl + HI − 2γ.

23. Although in the data, the actual simple correlationof years of education for couples for 23 countries in LatinAmerica is approximately 0.7, this assumption is only madefor expositional purpose. All our results are robust to therelaxation of this assumption.

A salient feature of the problem is that, what-ever the net benefit of working in the formalsector may be for a worker possessing a levelof human capital l, it will be lower for a secondworker because she/he already enjoys some por-tion of the nonpecuniary benefits generated bythe first worker. For this reason, ceteris paribus,second workers are more likely to operate infor-mally than primary workers. The expected utilityof a household that has one worker in the formalsector and the other in the informal sector is:

Ufi(wfl, HI ) = qlw f + (1 − q)(wf + wi)l

+ HI − γ.

Finally, the expected utility of a household thathas two workers in the informal markets is24:

Uii(wil, HI ) = (1 − q)22lwi + 2(1 − q)qw i l.

Analyzing the derivatives of the utility func-tions for each case with respect to human capital,we find that:

dUff/dl > dUfi/dl > dUii/dl

whenever wf /wi > (1 − q)

which is the same condition we used beforein order to establish that, in equilibrium, therewas a cutoff point dividing workers between theformal and informal sector.

In this case, there are two worker cutoffpoints. Both workers in households endowedwith low levels of human capital choose to workin the informal sector up to the first cutoff point,while both workers in households endowed withhigh levels of human capital and above the sec-ond cutoff point choose to work in the formalsector. Workers in households with human cap-ital levels between the two cutoff points chooseto diversify sectors, with one worker operatingformally and the other informally.25

The first cutoff point is determined by:

Ufi

(wfl, HI

)= Uii

(wf l, HI

)

and, solving for our functional forms, we obtain:

l = ((γ − HI )/(wf − (1 − q)wi)).

24. To simplify the explanation, we assume that theprobabilities of detection are independent for workers fromthe same household. However, since we assume risk neutral-ity, removing this assumption would not change the results.

25. Note that when members of a household differ intheir human capital, if they decide that only one operatesformally, it will always be the one with higher humancapital.

Page 14: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

834 ECONOMIC INQUIRY

This cutoff point (l) that makes a householdindifferent between having both workers inthe informal sector and one worker in eachsector is equal to (l), the cutoff point thatmakes a household with only one worker in thelabor market indifferent between working in theformal and the informal sector. This is becausethe decision to operate formally or informally isnot influenced by the presence of an informalworker in the household.

The second cutoff point is obtained by equal-izing the utility function when both workersoperate formally with the utility function whenone worker operates formally and the other oper-ates informally:

Uff

(wfl, HI

)= Ufi

(wfl, HI

)

and, solving for our functional forms, we obtain:

l = (γ/(wf − (1 − q)wi)).

It is easy to verify that l = l < l, as is illustratedin Figure 5. This means that the presence of aformal worker in the household reduces the like-lihood that other members of the household willchoose to operate formally. This result can onlybe generated in a model where workers chooseoptimally the sector they work for.

These cutoff points demarcate three regionsin the support of the distribution of human cap-ital. In Region 1, both households with oneworker and households with two individualsworking would choose to operate informally.In Region 2, households with only one workerwould choose to operate formally while house-holds with two workers would choose to diver-sify between sectors. Finally, in Region 3, bothone-worker and two-worker households wouldchoose to operate formally.

We will now consider how the equilibriumof the economy changes when households havea secondary worker. In particular, we will ana-lyze how the equilibrium changes when someof the workers in the distribution are secondaryworkers.

When some workers with human capital inRegion 1 or 3 are secondary workers, there is

FIGURE 5Household Cutoff Point Distribution

l l= l

1 32

l

no change in the equilibrium. This is becausein these regions both primary and secondaryworkers choose the same sector (informal inRegion 1 and formal in Region 3). However,when workers with human capital in Region 2are secondary workers, they choose to work inthe informal sector, whereas they would chooseto work formally if they were the only workerin the household. Thus, when some of theworkers in Region 2 are secondary workers, atthe equilibrium wages of the economy with onlyone worker per household there is an excesssupply of labor in the informal labor market andan excess demand in the formal. Thus, informalwages decrease. Therefore, some formal firmsmove to the informal sector and some informalworkers move to the formal sector. Thus, in thenew equilibrium, the informal sector expands.The effects of the existence of these secondaryworkers are the same as those produced by aredistribution of human capital toward informal-sector workers (Section V(A)).26

A. Empirical Evidence

We now investigate empirically the hypoth-esis that spouses, ceteris paribus, are morelikely to work informally if the head of house-hold works formally. Relying on our theoreticalframework, in this section, we derive a con-sistent likelihood estimator for this interestingparameter.

Consider first the estimation of a probitmodel of the decision of secondary workers as aresponse to the head of household’s sector deci-sion (i.e., a dummy variable that equals 1 if thehead of the household has a formal salaried job).Inconsistent estimates are obtained since includ-ing the primary worker’s endogenous responsein the secondary worker’s latent model ren-ders the estimation inconsistent. This is clearlyshown by the following recursive bivariate pro-bit model (Greene 2003):

y1 = 1[β1x1 + ε1 > 0](9)

y2 = 1[β2x2 + αy1 + ε2 > 0](10)

where Equations (9) and (10) are, respectively,the primary and secondary workers’ partici-pation equations. Following the literature, weassume that (ε1, ε2) is distributed as bivari-ate normal with mean zero, unit variance, andcov(ε1, ε2) = ρ, and is also independent of

26. The comparative static results, presented inSection V, are robust to the existence of secondary workers.

Page 15: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

GALIANI & WEINSCHELBAUM: INFORMALITY, HOUSEHOLD DECISIONS, LABOR MARKETS 835

(x1, x2). The parameter vector β1 can be con-sistently estimated by means of a simple probitmodel for Equation (9). However, if ρ �= 0, thenε2 and y1 are correlated, and a probit modelfor Equation (10) is inconsistent for both β2and α. Thus, when the unobservables that deter-mine both choices are correlated, a probit modelfor Equation (10) will produce inconsistent esti-mates of the parameters of the model, includ-ing the parameter of interest, α. Consider, forexample, the case where ρ > 0 (which is ourcase as estimated in Table 4 below), maybebecause the place of residence plays a rolein determining the choice that workers make,implying that a probit model for Equation (10)might even produce positive estimates of α, evenwhen the true parameter values were negative.In spite of this, the joint distribution of (y1, y2)given (x1, x2) is easily obtained and estimatedby the method of maximum likelihood.27

In Table 4, we present the results of esti-mating this recursive bivariate probit model.We employ the dataset used in Table 2 inSection III. Thus, the sample includes onlythose households in which both the head andspouse are salaried employees. This sampleselection provided us with 32,011 householdobservations.

We present three alternative specifications. InModel 1, for the spouse equation, the dummyvariable that equals 1 if the head of the house-hold has a formal salaried job, DF, is inter-acted with the three levels of education definedin Table 2. The model also includes as controlvariables the logarithm of wages, which mea-sures the logarithm of labor earnings of work-ers in their occupation, and that, in our setup,helps controlling for the individual’s unobserv-able endowment of labor l, education dummies,age and age squared, a dummy variable indi-cating whether residence is urban or rural, andcountry-fixed effects. The spouse equation alsoincludes the education dummies of the head ofthe household (HH) to control for sample selec-tion. The coefficients of interest—that is, thoseassociated to DF in the spouse equation—areall negative but only those interacted with low-and medium-education levels are statisticallysignificant at conventional levels of significance.

27. Note that, as is always the case in estimating a probitmodel, a necessary condition to identify α is that y1 is not aperfect predictor of y2. Moreover, to identify the parametersof the model, it is also necessary for the set of variables inx1 to contain at least one variable excluded in x2 (Green2003).

Thus, the evidence suggests that spouses, ceterisparibus, are more likely to work informallyif the head of household works formally. Themarginal effects range between 6% and 13%.These estimates are commensurate with theaverage differences presented in Table 2. Thecoefficient of correlation for the error terms inEquations (9) and (10) is quite high (0.521). Thisindicates that the bias of estimating the probitmodel for Equation (10) could be quite severeand could even reverse the sign of the trueparameter α. Indeed, this is what happens whenwe estimate the probit model for Equation (10)alone. All other coefficients have the expectedsign.

The results are robust across specifications.In Model 2, we add dummies for the totalincome per capita quartile of the household inthe spouse equation (we are unable to calcu-late this variable for 44 observations, whichleave us with 31,967 observations). Finally, inModel 3 we add household size variables ascontrol variables in the spouse equation. All thecoefficients of interest remain unaltered acrossspecifications.

VII. CONCLUSIONS AND POLICY LESSONS

In this paper, we present a model of an econ-omy in which the sizes and wages of the formaland informal sectors are endogenously deter-mined. This model entails a continuum of typeson both sides of the labor market (i.e., firmsand workers). Firms choose whether to oper-ate formally or informally and workers choosein what sector they work. In the equilibrium,managerial talent determines size dualism atthe firm level, and human capital is the factorthat determines whether workers are employedin the formal or informal sector. Thus, in ourmodel, both heterogeneous firms and workershave preferences over the sectors they operateand choose optimally whether to function for-mally or informally. We emphasize the role ofhousehold choice in shaping the outcomes ofthe economy we study. Using our model, weshow that an increase in the participation ofsecondary workers, ceteris paribus, would tendto raise the level of informality in the econ-omy. Furthermore, we believe that the setupof our model is rich enough to be used infurther work to study a broad set of generalequilibrium effects by introducing dynamics andinstitutional arrangements not addressed in thispaper.

Page 16: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

836 ECONOMIC INQUIRY

TA

BL

E4

Rec

ursi

veB

ivar

iate

Prob

itM

odel

s

12

3

Coe

ffici

ent

Stan

dard

Err

ors

Mar

gina

lE

ffec

tC

oeffi

cien

tSt

anda

rdE

rror

sM

argi

nal

Eff

ect

Coe

ffici

ent

Stan

dard

Err

ors

Mar

gina

lE

ffec

t

Spou

seE

quat

ion

DF×

low

educ

atio

n−0

.158

∗∗0.

064

−0.0

58−0

.166

∗∗∗

0.03

9−0

.062

−0.1

60∗∗

∗0.

027

−0.0

60D

med

ium

educ

atio

n−0

.344

∗∗∗

0.06

6−0

.131

−0.3

40∗∗

∗0.

053

−0.1

30−0

.327

∗∗∗

0.05

1−0

.125

DF×

high

educ

atio

n−0

.141

0.12

2−0

.052

−0.1

410.

092

−0.0

53−0

.124

0.09

0−0

.046

Log

ofw

ages

0.90

2∗∗∗

0.10

10.

787∗

∗∗0.

106

0.81

2∗∗∗

0.08

7M

ediu

med

ucat

ion

0.49

7∗∗∗

0.07

20.

466∗

∗∗0.

068

0.45

7∗∗∗

0.06

7H

igh

educ

atio

n0.

318∗

0.18

60.

343∗

0.18

00.

332∗

0.17

8A

ge0.

027∗

∗0.

013

0.03

6∗∗

0.01

40.

045∗

∗∗0.

011

Age

squa

red

−0.0

003∗

∗0.

0001

−0.0

005∗

∗∗0.

0002

−0.0

006∗

∗∗0.

0001

Edu

catio

nH

Hdu

mm

ies

Yes

Yes

Yes

Inco

me

quar

tile

dum

mie

sN

oY

esY

esN

umbe

rof

depe

nden

tch

ildre

n<

18&

Num

ber

ofch

ildre

n<

18N

oN

oY

es

Hea

dof

Hou

seho

ldE

quat

ion

Log

ofw

ages

HH

0.61

6∗∗∗

0.05

60.

616∗

∗∗0.

057

0.61

6∗∗∗

0.05

7M

ediu

med

ucat

ion

HH

0.28

4∗∗∗

0.03

50.

284∗

∗∗0.

035

0.28

4∗∗∗

0.03

6H

igh

educ

atio

nH

H0.

245∗

∗0.

105

0.24

4∗∗

0.10

50.

245∗

∗0.

106

HH

age

0.05

9∗∗∗

0.01

70.

058∗

∗∗0.

017

0.05

8∗∗∗

0.01

6H

Hag

esq

uare

d−0

.000

8∗∗∗

0.00

02−0

.000

8∗∗∗

0.00

02−0

.000

8∗∗∗

0.00

02C

orre

latio

ner

ror

term

s0.

521∗

∗∗0.

083

0.51

2∗∗∗

0.06

90.

505∗

∗∗0.

062

Urb

an/r

ural

dum

my

Yes

Yes

Yes

Cou

ntry

-fixe

def

fect

sY

esY

esY

es

Not

e:St

anda

rder

rors

clus

tere

dby

coun

try

are

repo

rted

.∗ S

igni

fican

tat

10%

;∗∗

sign

ifica

ntat

5%;

∗∗∗ s

igni

fican

tat

1%.

Page 17: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

GALIANI & WEINSCHELBAUM: INFORMALITY, HOUSEHOLD DECISIONS, LABOR MARKETS 837

A salient feature of the previous literatureon informality is that worker decisions play norole in determining the equilibrium of the econ-omy. It is only the firm that chooses optimallywhether to operate formally or informally, andeither homogenous workers are constrained toobtain the best jobs (as in the dualistic mod-els) or heterogeneous workers are indifferentabout which sector they participate in. This isclearly insufficient, as it ignores the role thathousehold choices play in determining the lev-els of formality in an economy. Naturally, asour model shows, we do not claim that infor-mality is only the result of household’s choices.However, ignoring the role of them could bemisleading and may well induce to disregardimportant (unintended) consequences of policiesand interventions.

At the policy front, the main lesson fromour paper is that governments should not con-sider only labor demand but also labor supplywhen tackling informality. Policies that reducethe supply of workers in the informal labor mar-ket at given wages will increase the level offormality in the economy. This is a very impor-tant lesson to take into account when designingwelfare policies in developing countries. Con-ditioning welfare transfers on the labor-marketstatus of the potential beneficiaries will increasethe level of informality of the economy.

Governments seem to want to increase thesize of the formal sector for different reasons: toincrease tax collection, to improve the produc-tivity of the economy and to provide workerswith the nonpecuniary benefits normally linkedto formal jobs. Specific policy recommendationswould differ depending on where the emphasisis placed. Both our model and the data show thatunskilled workers tend to operate informally, soif the provision of health and other benefits areexclusively attached to formal jobs, these work-ers will not be covered by them. Thus, policiesthat attempt to achieve universal coverage arelikely to be considered to reach the poor. Ofcourse, these policies will reduce the incentivesto operate formally, increasing the size of theinformal sector.

Finally, when formal jobs provide non-pecuniary benefits, there is the question ofwhether to narrow access to them just to theworker or to extend them to his/her family.If we consider the incentives that this deci-sion creates, it might appear that our analysissuggest that these benefits should be specified

by the worker in order to enhance the incen-tives of secondary workers to operate formally.However, at least two considerations are in orderhere. First, not extending the nonpecuniary ben-efits of formal jobs to the family of the workerwould reduce the incentives of primary workersto operate formally, and hence, the overall effectover formality of specifying benefits by workerinstead of households is ambiguous. Therefore,even if we restrict the analysis to the effect ofincentives, the answer to this question is casespecific. Second, while covering the populationwith these benefits is considered important, theanswer to this question is beyond the scope ofour paper, but it is also case specific. Never-theless, if that is the case, we probably need tothink about informality in terms of householdsand not individuals.

REFERENCES

Amaral, P., and E. Quintin. “A Competitive Model of theInformal Sector.” Journal of Monetary Economics, 53,2006, 1541–53.

Blanchflower, D., and A. Oswald. “What Makes anEntrepreneur?” Journal of Labor Economics, 16, 1998,26–60.

Camacho, A., E. Conover, and A. Hoyos. Effects of Colom-bia’s Social Protection System on Worker’s Choicebetween Formal and Informal Employment. Colombia:Universidad de los Andes, mimeo, 2009.

de Paula, A., and J. Scheinkman. The Informal Sector: AnEquilibrium Model and Some Empirical Evidence fromBrazil. Princeton: Mimeo, 2009.

. “Value Added Taxes, Chain Effects and Informal-ity.” American Economic Journal: Macroeconomics,2(4), 2010, 195–221.

De Soto, H. The Other Path: The Invisible Revolution in theThird World. New York: Basic Books, 1989.

Fortin, B., N. Marceau, and L. Savard. “Taxation, WageControls and the Informal Sector.” Journal of PublicEconomics, 66, 1997, 293–312.

Galiani, S., and F. Weinschelbaum. “Modeling InformalityFormally: Households and Firms.” Working Paper No.47, CEDLAS, Argentina. SSRN. 2007.

Gasparini, L., F. Haimovich, and S. Olivieri. “Labor Infor-mality Effects of a Poverty Alleviation Program.”Working Paper No. 53, CEDLAS, Argentina, 2007.

Gasparini, L., and L. Tornarolli. “Labor Informality inLatin America and the Caribbean: Patterns and Trendsfrom Household Surveys Microdata.” Working Paper,CEDLAS, Argentina, 2006.

Greene, W. Econometric Analysis. Englewood Cliffs, NJ:Prentice Hall, 2003.

Gruber, J. “Health Insurance and the Labor Market,” inHandbook of Health Economics, Vol. 1A, editedby A. Culyer and J. Newhouse. Amsterdam: North-Holland, 2000.

Gruber J., and B. Madrian. “Health Insurance, Labor Sup-ply, and Job Mobility: A Critical Review of the Liter-ature,” in Health Policy and the Uninsured, edited byC. McLauglin. Washington DC: Urban Institute Press,2004.

Page 18: GALIANI, S.; WEINSCHELBAUM, F.---2011--- Modeling Informality Formally, Households and Firms. Economic Inquiry, Volume 50, Issue 3, 21 SEP 2011

838 ECONOMIC INQUIRY

Harris, J. R., and M. P. Todaro. “Migration, Unemploy-ment, and Development: A Two-Sector Analysis.”American Economic Review, 60, 1970, 126–42.

Jauregui, I. El Plan Jefes de Hogar Desocupados y laInformalidad Laboral. Argentina: Universidad de SanAndres, mimeo, 2009.

Juarez, Laura. “Are Informal Workers Compensated for theLack of Fringe Benefits? Free Health Care as anInstrument for Formality.” Mimeo, 2008.

Layard, R., S. Nickell, and R. Jackman. Unemployment:Macroeconomic Performance and the Labour Market.Oxford: Oxford University Press, 1991.

Lewis, W. “Economic Development with Unlimited Suppliesof Labour.” Manchester School, 22, 1954, 139–91.

Lucas, R. “On the Size Distribution of Firms.” Bell Journalof Economics, 9, 1978, 508–23.

Madrian, B. “Employment-Based Health Insurance and JobMobility: Is There Evidence of Job-Lock?” QuarterlyJournal of Economics, 109, 1994, 27–54.

Magnac, T. “Segmented or Competitive Labor Markets?”Econometrica, 59, 1991, 165–87.

Maloney, W. “Informality Revisited.” World Development,32, 2004, 1559–78.

Olsen, C. “Health Insurance and Adverse Selection in theLabor Market.” Mimeo, 1993.

Pratap S., and E. Quintin. “Are Labor Markets Segmentedin Developing Countries?” European EconomicReview, 50, 2006, 1817–41.

Quintin, E. “Contract Enforcement and the Size of theInformal Economy.” Economic Theory, 37, 2008,395–416.

Rauch, J. “Modeling the Informal Sector Formally.”Journal of Development Economics, 35, 1991, 33–47.

Schneider, F., and D. Enste. “Shadow Economies: Size,Causes and Consequences.” Journal of EconomicLiterature, 38, 2000, 77–114.

Shapiro, C., and J. Stiglitz. “Equilibrium Unemploymentas a Worker Discipline Device.” American EconomicReview, 74, 1984, 433–44.

Stone, A., B. Levy, and R. Paredes. “Public Institutionsand Private Transactions: A Comparative Analysis ofthe Legal and Regulatory Environment for BusinessTransactions in Brazil and Chile,” in EmpiricalStudies in Institutional Change, edited by L. Alston,T. Eggertsson, and D. North. New York: CambridgeUniversity Press, 1996.

Straub, S. “Informal Sector: The Credit Market Chan-nel.” Journal of Development Economics, 78, 2005,299–321.