Growth, Redistribution and Poverty Changes in Cameroon

Growth, Redistribution and Poverty Changes in Cameroon:A Shapley Decomposition Analysis

Francis Menjo Baye*Department of Economics and Management, University of

Yaounde II, PO Box 1365, Yaounde, Cameroon

This paper studies the decomposition of poverty changes in Cameroon.Specifically, it reviews theoretical frameworks for growth–redistributiondecomposition analyses, presents the data and poverty measures and esti-mates the growth–redistribution components of changes in measuredpoverty by the Shapley value-based approach using Cameroon’s householdsurveys. By all the Pa class of measures, poverty increased significantlybetween 1984 and 1996. The growth components overaccounted for theincrease, although shifts in national, rural and semi-urban distributionsmarginally mitigated the worse effects on the population. A decline inmean incomes as well as adverse distributional shifts contributed to a sig-nificant increase in urban poverty during the same period. These findingscorroborate the general information in the literature that growth effectstend to dominate the effects of changes in the distribution of income.These results illustrate the potential contribution of distributionallyneutral growth in household incomes to poverty alleviation in Cameroon.Although redistribution also has an important role to play, it should beaccepted that there must be severe limits to what can be achieved bygrowth neutral redistribution. Growth in household incomes appearsmore likely to be essential for long-term poverty reduction and will bemore effective if poverty alleviation programmes are targeted disproportio-nately in favour of rural and semi-urban areas.

1. Introduction

Many developing countries are now addressing concerns for thepoor in addition to pursuing growth objectives as envisioned in

# The author 2006. Published by Oxford University Press on behalf ofthe Centre for the Study of African Economies. All rights reserved. Forpermissions, please email: [email protected]

*Corresponding author. E-mail: [email protected]

JOURNAL OF AFRICAN ECONOMIES, VOLUME 15, NUMBER 4, PP. 543–570doi:10.1093/jae/ejk010

their poverty reduction strategy papers. This developmentapproach anchors on the broadening of the initial objectives ofstructural adjustment to take on board social considerations, as gov-ernments and donors now share the same opinion that adjustmentefforts cannot be sustainable if the needs of the poor were ignored.1

After a period of sustained growth which Cameroon experiencedup to the middle of the 1980s, accomplishing an annual averagegrowth of 7% over a 10-year period (Government of Cameroon,2003, p. 1), the situation deteriorated from 1986 onwards and thecountry suffered a severe economic and social crisis. The crisiswas blamed on poor macro-economic performance occasioned, atleast in part, by a slump in world market prices of its main exportcommodities from 1986, which exposed the structural deficienciesof the country, and by the overvaluation of the cooperation finan-ciere en Afrique (CFA) franc against the dollar—a currency inwhich most of Cameroon’s exports are quoted. The short-termeffects of the ensuing policies designed to achieve macro-economicstability are thought to have equally deteriorated the welfare ofCameroonians, especially those at the lower half of the distributionof living standards. The harshness of the crisis led to the abandon-ment of the planned development system pursued since the early1960s, and the adoption of the IMF/World Bank medium-term struc-tural adjustment programmes (SAP) from 1988.2 The crisis also led toconsiderable shortfalls in public finances, making it difficult for thegovernment to pursue its development strategy with vigour. Evensome of the infrastructural achievements deteriorated for lack ofmaintenance. Many of the rural infrastructures, notably develop-ment projects put in place by the state collapsed, thereby aggravatingthe poverty of the people who benefited from those services.

Economic indicators continued to deteriorate and the steadydecline in incomes led to a 40% fall in per capita consumptionbetween 1985/86 and 1992/93. The external debt stock rose fromless than one-third to more than three-quarters of GDP between1984/85 and 1992/93. Investment declined from 27% to less than11% of gross domestic product (GDP) over the same period(Government of Cameroon, 2003, p. 2).

1 See Cornia et al. (1995) and Woodward (1992) for a discussion of this view, as wellas the conventional view frequently associated with the World Bank and the IMF.

2 For a succinct presentation of the development planning policies executedthrough 5-year Development Plans in Cameroon, see Baye and Fambon (2001).

544 F.M. Baye

Cameroonian authorities tried to cope with the economic crisis byreducing public expenditures, including cuts in producer prices of tra-ditional exports in 1991, and a 60% cut in civil service wages in 1993.These measures did not, however, stimulate growth. The authoritiesalso tried to cope by borrowing more to finance budget deficits inthe system. Unfortunately, debt servicing grew rapidly and startedcrowding out investments (Mbanga and Sikod, 2002). The provisionof public services declined markedly due to lack of investment andpoor performance of state-owned enterprises. Government reducedbasic health and education funding and this led to a major declinein health delivery systems and school enrolment (Khan andNoumba, 2001). Restructuring of state and semi-state enterprises, afreeze in increments and recruitment in the public service and staffredundancies caused a surge in unemployment, which affectedmostly women and young people. The marketing of traditionalexport commodities was liberalised in 1992–4, thus exposingfarmers to the volatility of world market prices. These moves werejudged by the political entrepreneurs, subjected to the conditionalitiesof the donor community, to be too slow to effect the much-neededadjustment and hence recourse was made to a measure that includedboth expenditure-reducing and expenditure-switching components.

The adverse international environment reflected in the slump inworld market prices of commodity exports in the late 1980s and theearly 1990s, and its implications for government revenue, pro-duction, consumption and relative prices, together contributed tocompel the franc zone countries to yield to the 50% devaluationof the CFA franc in January 1994. Being a centre-piece of adjust-ment, the devaluation was intended to reduce expenditure onimports as well as re-allocate resources away from non-tradable totradable commodities with a view to propping up the global com-petitiveness of the economy. Subsequent to the 1994 devaluationof the CFA franc, Cameroon achieved macro-economic stability.3

Yet, rural incomes were slow to improve because much of theacreage under coffee and cocoa had been abandoned, in addition

3 Despite this achievement, implementation of the post-devaluation reforms wasat first non-committal, and the government failed to capitalise on the windowof opportunities opened by the currency devaluation. It was only in 1996 thatthe government gave a firm mandate to a new team to work with the inter-national community, notably the IMF and the World Bank, to tackle the structuralproblems affecting the economy.

Growth, Redistribution and Poverty Changes in Cameroon 545

to the typically low short-run elasticities of supply of these com-modities. Moreover, salary cuts, the devaluation of the CFA francwith its short-term inflationary effects and the retrenchment ofpublic sector workers eroded the real purchasing power of mostCameroonians.

To better inform public debate in the aftermath or during policychanges that affect living standards, it is necessary to measure theevolution of poverty and income distribution, notably the decompo-sition of observed changes, with a view to assessing the importanceof factors explaining them. In the dynamic decomposition proposedby Datt and Ravallion (1992), for instance, the factors contributingto changes are variations in growth and redistribution. But,growth and redistribution in this standard decomposition do notform a partition because a residual element is usually includedto ensure the identity of the decomposition. This residual con-stitutes a ‘black box’, which is definitely of interest to both analystsand policy makers if opened and its contents attributed accordinglyto meaningful components. This residual measures the interac-tion effect between the two components of the povertydecomposition.

Some effort has been made in the past in constructing povertylines and profiles for Cameroon (Lynch, 1991; Njinkeu et al., 1997;Government of Cameroon, 1996; World Bank, 1995; Fambon et al.,2000, 2001). Knowledge on how poverty changes by socio-economiccharacteristics reflected through the level of education, employmentstatus and region of the country is unfortunately still at its infancy(Institute of Statistics, 2002; Baye, 2004; Fambon et al., 2004). To thebest of our knowledge, the exact decomposition of changes inpoverty into growth and redistribution components is still poorlyunderstood by both analysts and decision-makers in Cameroon.This paper attempts to fill some of these knowledge gaps by per-forming exact decompositions of changes in aggregate measuredpoverty. The procedure hinges on the Shapley value, which is awell-known solution concept in the theory of cooperative games.Our contributions are, therefore, entirely empirical.

The main objective of this paper is to investigate the character-istics of poverty in the period 1984–6, how poverty has changedand some of the factors explaining the changes. The specific objec-tives are (1) to review theoretical frameworks for growth–redistri-bution decomposition of poverty changes, (2) to investigate the

546 F.M. Baye

growth and redistribution effects of changes in aggregate poverty and(3) to derive policy implications on the basis of the analysis. The restof the study is in four main sections. Section 2 reviews the theoreticalframeworks for decomposition analyses. Section 3 presents the dataand other information necessary for implementing the methodology.Section 4 reports the results based on the Shapley value, and Section 5outlines some concluding remarks.

2. Theoretical Frameworks for Growth–RedistributionDecomposition

The adoption of the enhanced growth and poverty reduction strat-egies proposed by the donor community in sub-Saharan Africa inthe 1990s is indicative of the belief that economic growth helps toalleviate poverty. In the context of absolute poverty measurement,growth is expected to augment the incomes of, at least, some ofthe poor, hence leading to a fall in measured poverty using any con-ventional poverty index. This tendency can, however, be reinforcedor even reversed if economic growth is accompanied by redistribu-tion, which is pro-poor or pro-rich, respectively. Even for a givenrate of economic growth, a neutral or pro-poor redistributionpattern is preferable in terms of its impact on poverty. In whatfollows, theoretical frameworks for decomposition that separatethe effects of growth and redistribution on changes in measuredpoverty between two dates are reviewed.

2.1. The Datt–Ravallion Approach (1992)

The dynamic decomposition of measured poverty between twodates, t and tþ n, which allows one to rigorously quantify the rela-tive importance of growth versus redistribution, is proposed byDatt and Ravallion (1992). This decomposition procedure hasbeen widely applied or reviewed (Datt and Gunewardena, 1997;Canagarajan et al., 1997; McKay, 1997) and modified for the effectof a change in the poverty line by Ali (1997) and Shorrocksand Kolenikov (2001). As this is not an exact decomposition, thereis always a residual component that captures the interactionbetween growth and redistribution.

Essentially, the Datt–Ravallion methodology decomposes agiven change in aggregate poverty between two dates, t and tþ n,


into a growth component denoted by G(t, tþ n, r ), a redistributioncomponent D(t, tþ n, r ) and a residual R(t, tþ n, r ), where r is areference period (which may be t or tþ n ). The growth component,G(.), gives the impact on poverty of the change in the mean incomewhile holding the Lorenz curve constant at the reference level Lr.The redistribution component, D(.), gives the change in povertydue to a change in the Lorenz curve while keeping the meanincome at the reference level mr. The residual, R(.), measures theeffect of interaction between growth and redistribution terms onpoverty.

Assuming a fixed absolute poverty line, a poverty measure that isadditive and time-consistent may be defined as Pt ¼ P(Z/mt, Lt),where m is mean expenditure (or income) and L is a vector fullydefining the Lorenz curve. The change in poverty can be decom-posed as in equation (1).

Ptþn � Pt ¼ Gðt; tþ n; rÞ þDðt; tþ n; rÞ þ Rðt; tþ n; rÞ; ð1Þ

where

Gðt; tþ n; rÞ ¼ PðZ=mtþn; LrÞ � PðZ=mt;LrÞ;

Dðt; tþ n; rÞ ¼ PðZ=mr; LtþnÞ � PðZ=mr;LtÞ;

Rðt; tþ n; rÞ ¼ Gðt; tþ n; tþ nÞ � Gðt; tþ n; tÞ

¼ Dðt; tþ n; tþ nÞ �Dðt; tþ n; tÞ:4

The residual term is interpreted as the difference between the growth(redistribution) components evaluated at the terminal and initialLorenz curves (mean incomes), respectively. The residual can be rela-tively high whenever the marginal effects on the poverty index ofchanges in the mean (Lorenz curve) also strongly depend on theprecise Lorenz curve (mean),5 and as the reference period isswapped, the sign of the residual changes while conserving its

4 In each case, the first two arguments in the parentheses refer to the initial andterminal dates of the decomposition period, and the last argument makes explicitthe reference date r with respect to which the observed change in poverty isdecomposed.

5 That is, when the poverty measure is not additively separable in m and L.

548 F.M. Baye

magnitude. The Datt and Ravallion (1992) decomposition approach isthe base of the other proposed approaches presented subsequently.

2.2. The Axiomatic Approach due to Kakwani (1997)

The decomposition proposed by Datt and Ravallion (1992) includesa residual term and uses the ‘benchmark period’ concept, whichleads to an asymmetrical consideration of initial and final periods.To overcome these two limitations, Kakwani (1997) develops anaxiomatic approach which eliminates the residual term and givesa symmetrical evaluation of the initial and final periods. Given apoverty change between two periods, t and tþ n, this changecould entirely be explained by growth (Gt,tþn) and redistribution(Dt,tþn) components, expressed implicitly as: DPt,tþn ¼ f (Gt,tþn,Dt,tþn). To determine the nature of the decomposition, Kakwani(1997) proposes three axioms:

Axiom 1 If Dt,tþn ¼ 0, ) DPt,tþn ¼ Gt,tþn, and if Gt,tþn¼0 ) DPt,tþn ¼

Dt,tþn

Expressed in words, if the growth effect (redistribution effect) is zero,

then the change in poverty is entirely accounted for by a change in

inequality (change in mean income). In this context, if growth and

redistribution effects are simultaneously equal to zero, then poverty does

not vary between the two periods.

Axiom 2 If Gt,tþn � 0 and Dt,tþn � 0, ) DPt,tþn � 0, and if Gt,tþn � 0 and

Dt,tþn � 0, ) DPt,tþn � 0

Expressed in words, if growth and redistribution effects take the same

sign, then the poverty change also takes the same sign.

Axiom 3 If Gt,tþn ¼ 2Gtþn,t and Dt,tþn ¼ 2Dtþn,t, then symmetry is implied.

Expressed in words, if one moves from the terminal period to the initial

period, the effects are of the same magnitude but with opposite signs. This

axiom ensures symmetry between the initial and terminal periods.

As the first two axioms ensure that the generating function f(.) islinear and additive, and the third axiom ensures symmetry,Kakwani (1997) computes the contribution of growth as the meanof two effects: (1) the growth effect when the initial redistribution(Lorenz curve) is maintained and the growth effect when the finalredistribution (Lorenz curve) is maintained. Average redistributioneffects are calculated following the same logic, motatis motandis.


Reacting to the absence of a common framework for decompositionprocedures, Shorrocks (1999) proposes a Shapley value-based coop-erative game theory framework. When applied to growth–redistri-bution decomposition of poverty changes, Shapley value-basedapproach yields exactly the same theoretical and empirical resultsas due to Kakwani (1997). The only differences are that, althoughthe axiomatic approach is limited to two components, the Shapleyvalue-based approach is more general and motivated by well-grounded theoretical underpinnings.

2.3. The Shapley Value Decomposition Framework

2.3.1. Defining the Shapley Value

An important issue in distributive analysis would be how to assignweights to the factors that contribute to an observed level orchange in a measure of living standards. This issue is similar to pro-blems that arise in cooperative game theory, and recent literature indistributive analysis is proposing an attribution according to theShapley value (Shorrocks, 1999; Kabore, 2003; Rongve, 1995;Chantreuil and Trannoy, 1999). The Shapley value (Owen, 1977,Moulin 1988), therefore, provides a recommendation for the divisionof the joint profits or costs of the grand coalition. Let K ¼ f1, 2, . . . , mg

be a finite set of players. Non-empty subsets ofK are called coalitions.To accomplish the division process, the players may form coalitionsand the strength of each coalition is expressed as a characteristicfunction v.

For any coalition or subset S# K, v(S ) measures the share of thesurplus or loss that the coalition, S, is capable of appropriatingwithout resorting to agreements with players belonging to othercoalitions. For each player k, k� S, Shapley (1953) proposes a valueon the basis of the player’s marginal contribution—defined as theweighted mean of the marginal contributions v(S< fkg) 2 v(S ) ofplayer k in all coalitions S# K2 fkg. That is, player k is attributedthe extra amount that he brings to the existing coalition of players.

When the coalition, S, is composed of s elements, we can onlyfind the value they will obtain, v(S ), when the first randomlyordered s elements are exactly the elements of S. The weight ofthe coalition, S, will be measured by the probability that the first selements are all elements of S. This probability is found by dividing

550 F.M. Baye

the number of ordered arrangements of which the first s elementsare all in S by the total number of possible ordered arrangements.The numerator can be obtained by imagining that the first splayers are orderly arranged in a sequence and the remainingm2 s21 players are also orderly arranged in another sequence.

The number of possible ordered arrangements is the number ofpermutations of m players taken m at a time, which is m! By thesame reasoning, as the first s players yield s! number of permu-tations, the remaining m2 s21 players would yield (m2 s2 1)!number of permutations. The number of ordered arrangements inwhich the first s players are all elements of S is thus given bys!(m2 s2 1)!

The Shapley value of player k, denoted by fSk ðK; vÞ, is thus the

weighted mean of his marginal contributions v(S< fkg) 2 v(S )over the set of coalitions S# K2 fkg given by:

fSk ðK; vÞ ¼

Xm�1

s¼0

X

S#K� kf g

Sj j¼s

s!ðm� s� 1Þ!

m!½vðS< kf gÞ � vðSÞ�; ð2Þ

where by convention, 0!¼1 and v(Ø) ¼ 0.6 To apply the Shapleyvalue in distributive analysis, instead of considering m players asin cooperative game theory, we now consider m factors that contrib-ute to the explanation of an observed phenomenon. The Shapleyvalue is symmetric, exact and additive.7

2.3.2. Application of the Shapley Value to Growth–RedistributionDecomposition

Given a fixed poverty line Z, the poverty level at time t may beexpressed as a function P(mt, Lt) of mean income mt and theLorenz curve Lt. Following Shorrocks (1999), the growth factor inthe change in poverty between period t and tþ n is denoted by G ¼

ðmtþn=mtÞ � 1 and the redistribution factor, which is due to the shift

6 The Shapley value can also be interpreted as the expected marginal contributionmade by the player (or factor) to the value of a coalition, where the distribution ofcoalitions is such that any ordering of the players (or factors) is equally likely.

7 For a proof of these Shapley’s axioms in the context of distributive analysis, seeShorrocks (1999. pp. 5–6)


in the Lorenz curve is given by D ¼ Ltþn–Lt.8 The exercise becomes

one of identifying the contributions of growth, G, and redistribu-tion, D, in the decomposition of changes in any poverty measurethat is additively decomposable. As is habitual in most applications,we adopt the Pa class of poverty measures (Foster et al., 1984). Theaggregate change in the Pa class of measures is given as

DPa ¼ Paðmtþn; LtþnÞ– Paðmt; LtÞ¼ Paðmtð1 þ GÞ; Lt þDÞ– Paðmt; LtÞ ¼ vaðG;DÞ: ð3Þ

This is an expression of the change in poverty, DPa, which weneed to decompose into the growth (G ) and redistribution (D ) com-ponents. Because there are just two factors here, the possible elim-ination sequences (permutations) are m! ¼ 2! ¼ 2, given by fG, Dg

and fD, Gg. The marginal contribution of growth, G, and the mar-ginal contribution of redistribution, D, to any poverty change(DP ) and their associated weights are given in Table 1.

The Shapley values of the growth and redistribution componentscan be interpreted from Table 1 as follows:

fSGð2; vÞ ¼ 0:5½vðG;DÞ � vðDÞ þ vðGÞ�; ð4Þ

fSDð2; vÞ ¼ 0:5½vðG;DÞ � vðGÞ þ vðDÞ�; ð5Þ

When growth is absent, G takes the value 0 and the change inpoverty due only to redistribution becomes

vðDÞ ¼ Pðmt;LtþnÞ– Pðmt;LtÞ: ð6Þ

The indication here is that the change in poverty is due only to ashift in the Lorenz curve from Lt to Ltþn, holding mean income con-stant. By the same token, assuming away the redistribution factorby setting D ¼ 0 gives

vðGÞ ¼ Pðmtþn; LtÞ– Pðmt;LtÞ: ð7Þ

This shows that the change in poverty is due to a change in meanincome from mt to mtþn, with the Lorenz curve fixed at Lt. In otherwords, this is a situation of a distributionally neutral growth.

8 As noted by Shorrocks (1999), this is a slight abuse of notation, as the growth andredistribution factors ought to be distinguished from the variables representinggrowth and redistribution. However, as the growth and redistribution factorsare eliminated by setting G and D equal to zero, no serious confusion arises.

552 F.M. Baye

The full expressions of Shapley value contributions for thegrowth and redistribution effects of a change in the Pa class ofpoverty measures (DPa) are given by equations (8) and (9),making use of equations (3–7), as

fSaGð2; vÞ ¼0:5½Paðmtþn;LtþnÞ– Paðmt;LtÞ

–fPaðmt; LtþnÞ– Paðmt; LtÞg

þ fPaðmtþn; LtÞ– Paðmt;LtÞg�

¼0:5½Paðmtþn;LtþnÞ– Paðmt;LtþnÞ

þ Paðmtþn; LtÞ– Paðmt; LtÞ�; ð8Þ

fSaDð2; vÞ ¼0:5½Paðmtþn;LtþnÞ– Paðmt;LtÞ

–fPaðmtþn;LtÞ– Paðmt; LtÞg

þ fPaðmt;LtþnÞ– Paðmt;LtÞg�

¼0:5½Paðmtþn;LtþnÞ– Paðmtþn;LtÞ

þ Paðmt; LtþnÞ– Paðmt; LtÞ�: ð9Þ

It can easily be seen that overall change in poverty is the sum of thegrowth and redistribution components given by the Shapley contri-butions:

DPa ¼ faGð2; vÞ þ faDð2; vÞ ð10Þ

Table 1: Application of Shapley Value to Growth-Redistribution Decomposition

S ss!ðm � s � 1Þ!

m!v(S < fkg) 2 v(S )(marginal contributions)

Number of elementsin S before G0 Ø 0 0.5 v(Ø < fGg) 2 v(Ø) ¼ v(G ) 2 01 fDg 1 0.5 v(D< fGg) 2 v(D ) ¼ v(G, D )-v(D )

Number of elementsin S before D0 Ø 0 0.5 v(Ø < fDg)-v(Ø) ¼ v(D ) 2 01 fGg 1 0.5 v(G< fDg) 2 v(G ) ¼ v(G, D ) 2 v(G )


3. Data, Poverty Measures and Poverty Lines

3.1. Presentation of Household Data

The analysis of poverty in this text is based on two householdsurveys: the 1984 budgetary and consumption survey (BCS)(Government of Cameroon, 1984), September 1983–September1984, and the 1996 Cameroon Households Consumption Survey(CHCS) (Government of Cameroon, 1996), February–April 1996,carried out by the government’s Statistics Office. These snapshotsrepresent points before and during SAP in which householdsurveys are available.

These surveys vary in duration, 1 year for the first survey and 3months for the second, and in sample size, 5474 households for thefirst and 1731 for the second. They are similar in (1) the partitioningof the various regions, in the sense that the 1984 survey could easilybe regrouped to mimic the structure of the 1996 survey, and (2) thesampling techniques used. The recall period in both surveys is 7days. During the 1984 survey, price data were systematically col-lected from the nearest markets, which were used to value the pro-ducts consumed by households. In the case of the 1996 survey, pricedata were only available for the urban areas, and hence some adjust-ments were made to obtain unit prices for the rural areas.9

The 1984 food and total expenditures were scaled up, employingthe food and total consumer price indices, respectively, to expressthem in terms of 1996 prices to enable us use the poverty lines com-puted from the 1996 survey for the two periods.10 For all practicalpurposes, these surveys are considered suitable for the presentstudy. The welfare indicator used is expenditure per adult equival-ent. Because the composition of households by age was captured bythe surveys, we followed previous studies in Cameroon to attribute

9 More detailed description of the content of 1996 data, their sampling propertiesand other features can be found in Fambon et al. (2000).

10 Ideally, we would have used the same methodology to compute poverty linesusing the 1984 data, but the food items in these data are not disaggregated toenable the calculation of calorie-intake per adult equivalent, which is crucialin the methodology. Fischer indices could have been the first best alternativeapproach to use in scaling up the 1984 food and total expenditures, butFischer indices also require data on disaggregated quantities, which are unavail-able. We therefore consider our choice of the food and aggregate consumer priceindices as a second best approach. The most delicate step here is the deflation ofhousehold incomes or expenditures. Overestimating the deflator implies overes-timating the decrease in mean incomes and overestimating the increase inpoverty. The reverse is true if the deflator is underestimated.

554 F.M. Baye

adult equivalent scales of 1 to all adults (15 years old and above)and 0.5 to children (less than 15 years old).

3.2. Poverty Measures

The poverty indices used belong to the Pa class of povertymeasures, specifically P0, P1 and P2, which represent the headcount,the poverty-gap and the squared poverty-gap indices, respectively(using the terminology of Foster et al., 1984). The headcount indexmeasures the incidence of poverty, the poverty-gap index measuresthe depth of poverty and the squared poverty-gap index measuresthe severity of poverty.11

3.3. Poverty Lines

The procedure we follow in deriving the poverty lines to separatethe poor from the non-poor is a blend of the food energy intake(FEI) and cost-of basic-needs (CBN) methods, which consists oftwo stages (Fambon et al., 2001; Baye, 2005): calculating a foodpoverty line by the FEI method and evaluating the correspondingnon-food expenditures required to scale-up the food poverty lineto determine the overall poverty lines using a hybrid of the CBNapproach.12 In particular, to obtain the food poverty line, use ismade of ‘non-parametric’ regressions of food expenditures oncalorie intake per adult equivalent, which follow the logic of para-metric regressions proposed by Greer and Thorbecke (1986), andfor non-food expenditures, non-parametric regressions of totalexpenditures on food expenditures, which follow the logic of para-metric regressions proposed by Ravallion (1992), Ravallion andBidani (1994) and Ravallion (1998).

The first stage consists of estimating the food poverty line, whichcorresponds to the expected food expenditures needed to meetsome minimum level of calorie intake. This intake level is estimatedby the FAO for Cameroon at 2400 kcal per adult per day. By using anon-parametric regression,13 the food poverty line (ZF) was

11 For the implications and shortcomings of these measures, see Baye (1998).12 For a more comprehensive presentation of this approach, with illustrations, see

Baye (2004).13 The advantages of a non-parametric regression over a parametric one are that:

(1) they do not impose a priori functional forms and (2) the procedure appliesa local weighting process that attributes smaller weights as the absolute gapsbetween individual calorie intakes and the predetermined minimum increase.


estimated at 255.95 CFA francs per day per adult equivalent for 1996(Table 2). The non-parametric regressions were performed usingDAD4.3-A: software for Distributive Analysis developed byresearchers at Universite Laval.14 Table 2 presents comparativefood poverty lines computed parametrically and non-parametrically for the capital city and three rural areas.

The second stage involves estimating the total expenditures ofthose whose food expenditures equal the food poverty line. In prac-tice, however, to obtain the aggregate poverty line, an amountequal to the food poverty line can be projected from the X-axis tointersect the estimated regression line of total expenditures on foodexpenditures. The corresponding point on the Y-axis defines theaggregate poverty line. An aggregate poverty line (ZU) of 533.87CFA francs per day per adult equivalent was adopted for 1996.

4. The Empirical Results

Both national and zonal poverty levels, their changes and decompo-sitions are performed using DAD4.3-A. The headcount index rose by35% points from about 35% in 1984 to 70% in 1996, whereas thedepth and severity of national food poverty rose by 19 and 11%points, respectively (Tables 3 and 4). These food poverty changeswere found to be significantly different from zero at the 1% level(Table 4). By components, distributionally neutral growth accountedfor more than 35% points, and the distributional shifts moderatedthe outcome to register just a 35% points’ increase in the incidenceof national food poverty. A similar tendency is observed withthe changes in the depth and severity of poverty. The indicationhere is that the increase in national food poverty between 1984 and1996 was overwhelmingly accounted for by a large decline inmean incomes and the redistribution effects only marginally miti-gated worse effects of poverty on the population. A similar tendencyis captured in both the rural and semi-urban areas.

The results obtained by this method are less affected by the presence of ‘outliers’in the data and thus do not suffer significantly from specification bias that orig-inates from a ‘wrong’ functional form. For an exposition of the non-parametricestimation procedure, see Yatchew (1998), Silverman (1986), Nadaraya (1964)and Watson (1964).

14 See, for example, Duclos et al. (2003).

556 F.M. Baye

In the rural areas, for instance, the incidence, depth and severityof food poverty increased by 39.2, 23.4 and 14.2% points, respect-ively (Table 4). But for the attenuating effects of the distributionalshifts, poverty would have hit the rural dwellers even harder.This is so because the distributionally neutral decline in meanincomes accounted for the increases in the incidence, depth andseverity of rural poverty of 40.2, 25.3 and 15.8% points, respectively,which surpassed the changes in rural poverty (Table 4). In the semi-urban areas, the increase in the headcount, poverty-gap andsquared poverty-gap indices of 31.8, 16.3 and 9.2% points wasmore than accounted for by the distributionally neutral growtheffects. The redistribution effects had a positive impact on povertyalleviation, reflecting reducing inequality in semi-urban areas, butit was too small to counter the impact on poverty of the sharpdecline in mean incomes.

Table 2: Food Poverty Lines

Region Cost of calorie functionðln FEj ¼ aþ bCj þ vjÞ

Food poverty lines(in CFAF per adultequivalent)

a b R2 Parametric(per day)

Non-parametric(per day)

Yaounde (capital city)(N ¼ 344)

5.125(29.3883)

2.90 � 1024

(4.8043)0.46 337.31 336.64

Rural forest(N ¼ 210)

4.293(31.1793)

4.97 � 1024

(7.6288)0.65 241.24 258.81

Rural H-plateaux(N ¼ 200)

4.382(47.5876)

2.94 � 1024

(11.1143)0.72 162.00 170.50

Rural savannah(N ¼ 203)

4.686(35.1032)

2.45 � 1024

(5.6322)0.39 195.20 203.82

National(N ¼ 1666)

4.866(73.6607)

2.78 � 1024

(12.4692)0.38 252.95 255.95

Source: (Extracted from Baye (2005), Table 1).Notes: The figures in parenthesis represent t-ratios. R2 measures the goodness-of-fitof the regression. N is the number of households.


Table 3: Zonal Food Poverty Indices in 1984 and 1996 in Cameroon (ZF ¼ 255.95)

Zone 1984 1996

Populationshare

P0 P1 P2 Populationshare

P0 P1 P2

Urban 0.1119(0.0062)

0.0284(0.0071)

0.0046(0.0013)

0.0014(0.0005)

0.2945(0.0268)

0.4578(0.0279)

0.1603(0.0134)

0.0760(0.0079)

Semi-urban 0.1824(0.0280)

0.2680(0.0343)

0.0906(0.0157)

0.0419(0.0089)

0.0518(0.0238)

0.5859(0.1613)

0.2534(0.0790)

0.1342(0.0470)

Rural 0.7056(0.0289)

0.4206(0.0246)

0.1355(0.0108)

0.0598(0.0058)

0.6537(0.0429)

0.8129(0.0340)

0.3691(0.0261)

0.2022(0.0187)

National 1.0(0.00)

0.3488(0.0194)

0.1126(0.0084)

0.0500(0.0044)

1.0(0.00)

0.6965(0.0291)

0.3016(0.0198)

0.1615(0.0134)

Source: Calculated from BCS 1984 and CHCS 1996 Survey Data.Note: P0, P1 and P2 represent the headcount, the poverty-gap and the squared poverty-gap indices, respectively.

558F.M

.Baye

A different picture is, however, painted in the urban centreswhere the significant increase in the headcount, poverty-gap andsquared poverty-gap indices was reinforced by both distribution-ally neutral growth and distributional shifts. The main mechanismof the observed adverse redistributional components of povertychanges over the period under consideration seems to have beenthe relative downturn in formal sector employment following theretrenchments in the early 1990s of public and semi-public sector

Table 4: Growth and Redistributional Effects of Changes in Food Poverty in Cameroon(ZF ¼ 255.95)

Zonesand Pa

Shapley approach

Growthcomponent

Redistributioncomponent

Total change

UrbanP0 0.3832 (0.0146) 0.0462 (0.0146) 0.4294a (0.0288)P1 0.1322 (0.0141) 0.0235 (0.0141) 0.1557a (0.0135)P2 0.0609 (0.0068) 0.0138 (0.0068) 0.0747a (0.0079)

Semi-urbanP0 0.3811 (0.0375) 20.0633 (0.0375) 0.3178a (0.1650)P1 0.1899 (0.0671) 20.0272 (0.0671) 0.1628b (0.0806)P2 0.1125 (0.0438) 20.0202 (0.0438) 0.0923b (0.0479)

RuralP0 0.4025 (0.0199) 20.0103 (0.0199) 0.3923a (0.0419)P1 0.2527 (0.0348) 20.0191 (0.0348) 0.2336a (0.0283)P2 0.1583 (0.0216) 20.0160 (0.0216) 0.1424a (0.0195)

NationalP0 0.3535 (0.0131) 20.0058 (0.0131) 0.3477a (0.0350)P1 0.2057 (0.0246) 20.0167 (0.0246) 0.1890a (0.0215)P2 0.1255 (0.0147) 20.0140 (0.0147) 0.1115a (0.0141)

Source: Calculated from BCS 1984 and CHCS 1996 Survey Data.Notes: Figures in parentheses represent standard errors. Stratification and cluster-ing in the surveys were taken into consideration when setting the sample designs.P0, P1 and P2 represent the headcount, the poverty-gap and the squared poverty-gap indices, respectively.aSignificance at 1% level.bSignificance at 5% level.


workers and the consequent overcrowding in the urban informalsector. The indication of these observations is that the structure ofincome distribution in urban centres is such as to exacerbate thenegative impact on the welfare of the poor. The implication of thefindings is that growth would have a significantly positive impacton poverty alleviation and policies that redistribute in favour ofthe rural and semi-urban areas could marginally enhance the posi-tive effects of growth on poverty reduction.

The incidence of overall poverty in rural areas is higher than thatof urban areas over the period under review. Rural areas contribu-ted relatively about 86% of national aggregate poverty in 1984 andabout 80% in 1996 (Table 5). This drop in the rural contribution tonational poverty could be explained, at least in part, by thedecline in the population share of rural areas from about 71% in1984 to 65% in 1996 associated with net migration from rural tourban areas. Yet, a comparison of relative contributions and popu-lation shares in both periods indicates that the rural areas are dis-proportionately suffering from aggregate poverty. Table 5 alsoshows the contraction of the semi-urban population from 18 to 5%over the same period. This apparent depopulation of the semi-urban areas may be attributed to the reclassification of villages, aswell as to net migration to the main cities between 1984 and 1996(Baye, 2004).

The incidence of overall national poverty increased from 39 to68% between 1984 and 1996.15 The change in national incidence ofpoverty of 29% points is significantly different from zero at the1% level (Table 6). The depth and severity of aggregate povertyincreased significantly by 14 and 8% points, respectively.

As with the decomposition of the determinants of food povertychanges, the growth components overwhelmingly dominate theredistribution components both at the national and zonal levels(Table 6). With the exception of the headcount index at the nationallevel, the other measures at the national level and all measures atthe rural and semi-rural areas indicated opposing effects betweengrowth and redistribution in explaining the poverty outcomes.

15 According to UNDP (1998), official estimates of poverty rates in Cameroon in1984 and 1996 were 40 and 50.5%, respectively. One of the criticism of these esti-mates is that they were generated using poverty lines whose food componentswere computed with minimum daily consumption needs composed only ofthree food items.

560 F.M. Baye

Table 5: Zonal Overall Poverty Indices in 1984 and 1996 in Cameroon (ZU ¼ 533.87)

Zone 1984 1996

Populationshare

P0 P1 P2 Populationshare

P0 P1 P2

Urban 0.1119(0.0062)

0.0150(0.0048)

0.0024(0.0010)

0.0007(0.0004)

0.2945(0.0268)

0.3675(0.0314)

0.1163(0.0126)

0.0509(0.0066)

Semi-urban 0.1824(0.0280)

0.2961(0.0281)

0.0998

(0.0126)

0.0466(0.0068)

0.0518(0.0238)

0.6032(0.1152)

0.2096(0.0520)

0.0946(0.0270)

Rural 0.7056(0.0289)

0.4764(0.0238)

0.1578(0.0115)

0.0711(0.0064)

0.6537(0.0429)

0.8265(0.0357)

0.3426(0.0246)

0.1748(0.0166)

National 1.0(0.00)

0.3918(0.0190)

0.1298(0.0087)

0.0588(0.0048)

1.0(0.00)

0.6798(0.0313)

0.2691(0.0188)

0.1342(0.0119)

Source: Calculated from BCS 1984 and CHCS 1996 Survey Data.Note: P0, P1 and P2 represent the headcount, the poverty-gap and the squared poverty-gap indices, respectively.

Grow

th,Redistribu

tionandPoverty

Chan

ges

inCam

eroon561

For all the poverty measures, although the growth componentsoveraccounted for the increase in poverty between 1984 and 1996,shifts in rural and semi-urban distributions mitigated worseeffects of the crisis and reforms that caused poverty to rise. By allpoverty measures, shifts in urban distribution, in contrast,reinforced poverty in that sector. Indeed, both a decline in meanincomes and adverse distributional shifts contributed to the

Table 6: Growth and Redistributional Effects of Changes in Overall Poverty in Cameroon(ZU ¼ 533.87)

Zonesand Pa

Shapley approach

Growthcomponent

Redistributioncomponent

Totalchange

UrbanP0 0.2920 (0.0145) 0.0606 (0.0145) 0.3525a (0.0318)P1 0.0862 (0.0110) 0.0277 (0.0110) 0.1139a (0.0126)P2 0.0359 (0.0052) 0.0143 (0.0052) 0.0503a (0.0066)

Semi-urbanP0 0.3316 (0.0325) 20.0245 (0.0325) 0.3071a (0.1186)P1 0.1452 (0.0578) 20.0354 (0.0578) 0.1098b (0.0535)P2 0.0778 (0.0310) 20.0297 (0.0310) 0.0480 (0.0279)

RuralP0 0.3574 (0.0220) 20.0072 (0.0220) 0.3502a (0.0430)P1 0.2178 (0.0339) 20.0330 (0.0339) 0.1848a (0.0272)P2 0.1323 (0.0208) 20.0287 (0.0208) 0.1037a (0.0178)

NationalP0 0.2830 (0.0133) 0.0050 (0.0133) 0.2880a (0.0366)P1 0.1549 (0.0228) 20.0156 (0.0228) 0.1393a (0.0207)P2 0.0911 (0.0138) 20.0157 (0.0138) 0.0754a (0.0129)

Source: Calculated from BCS 1984 and CHCS 1996 Survey Data.Note: Figures in parentheses represent standard errors. Stratification and clusteringin the surveys were taken into consideration when setting the sample designs.P0, P1 and P2 represent the headcount, the poverty-gap and the squared poverty-gap indices, respectively.aSignificance at 1% level.bSignificance at 5% level.

562 F.M. Baye

significant increase in urban aggregate poverty during the periodunder study. The suppression of the public urban transportsystem during this period implies that the welfare of householdswho use this means of transportation deteriorated further.

A central observation emanating from this paper is that the signifi-cant increase in poverty between 1984 and 1996 was due more to thereduction in average incomes than to the rise in inequality. Suchresults illustrate the potential contribution of distributionally neutralgrowth in household incomes to poverty alleviation. As remarkedby McKay (1997), this is not to deny that redistribution also has arole to play, but there must be severe limits to what can be achievedby redistribution in the absence of growth. This is reflected in thegrowth neutral marginal redistributional effects. In this context,growth in household incomes is likely to be essential for long-termpoverty reduction. This will, however, be much less effective if thefruits of growth tend to be highly skewed towards the richest house-holds. Growth in incomes is, therefore, needed and it should be distri-butionally sensitive in favour of the under-privileged.

Our results and those of others using data from other developingcountries (Table 7) indicate that growth contributes overwhelminglyto poverty changes. These results are based on micro-economicgrowth, notably household income/expenditure, which has a directlink with measures of poverty. Economic growth generally refers tomacro-economic growth or GDP growth measured from nationalaccounts data, which can be realised without necessarily leadingto poverty reduction, especially if redistribution is skewed to theright.

This potential mismatch between national accounts and house-hold data is due to conceptual and methodological differences.The relevant literature has raised this problem, but householdsurveys are generally assumed to be more accurate and indepen-dent than national accounts (Ravallion, 2001; Deaton, 2004). Thetwo may be related, and this relationship may be closer in LDCsthan in developed countries (McKay, 1997), but the link betweenGDP growth and poverty reduction is likely not to be a systematicone (Bocconfuso and Kabore, 2004). Consequently, the nature ofgrowth and the transmission mechanisms of the fruits of growthto the poor need to be better understood via improved conceptual-isation and implementation of policies that encourage pro-poor orshared growth.


Table 7: Growth–Redistribution Decomposition Results Reported for Other Less Developed Countries (LDCs)

Author Method Country Period Povertymeasure

Totalchange

Components R

G D

Bigsten et al. (2003) Datt–Ravallion Ethiopia 1994–1997 P0 25.7 210.6 5.9 21.0P1 23.6 25.3 2.8 21.1P2 22.3 23.2 1.5 20.6

Kakwani (1997) Kakwani Thailand 1988–1994 P0 216.27 220.31 4.04 —P1 26.10 27.96 1.86 —P2 0.57 23.94 4.51 —

Kabore (2003) Datt–Ravallion Burkina Faso 1994–1998 P0 0.9 2.27 21.59 0.27P1 0.16 1.26 21.42 0.00P2 0.17 0.27 20.84 0.05

Datt–Ravallion Senegal 1995–2000 P0 218.8 235.0 3.89 12.3P1 29.57 219.0 9.62 20.19P2 25.90 211.2 8.65 23.34

Kakwani Burkina Faso 1994–1998 P0 0.9 2.40 21.45 —P1 0.16 1.26 21.42 —P2 0.17 0.70 20.87 —

Kakwani Senegal 1995–2000 P0 218.8 228.8 10.0 —P1 29.57 219.1 9.52 —P2 25.90 212.9 6.98 —

Sources: Bigsten et al. (2003), Kakwani (1997) and Kabore (2003).Notes: G, growth; D, redistribution; and R, residual.

564F.M

.Baye

5. Concluding Remarks

This paper attempted to investigate the characteristics of poverty inthe period 1984–6, how poverty evolved over this period and thefactors explaining the various changes. Specifically, it (1) reviewedtheoretical frameworks for growth–redistribution decompositionanalyses and (2) investigated the growth and redistribution effectsof changes in both food and aggregate poverty. The procedureadopted is a unified conceptual framework that is amendable tomost kinds of decompositions in distributive analysis and is freeof hazy concepts like residual or interaction terms. The paper dis-cussed the effects of growth and redistribution on changes inmeasured poverty between 1984 and 1996 using the Shapleyapproach.

The decomposition of food and aggregate poverty changes indi-cated that the growth components overwhelmingly dominated theredistribution components both at the national and zonal levels.Most measures at the national level and all measures at the ruraland semi-urban areas indicated opposing effects between growthand redistribution in explaining poverty outcomes. For all thepoverty measures, although the growth components overaccountedfor the increase in poverty between 1984 and 1996, shifts in ruraland semi-urban distributions mitigated worse effects of the crisisand initial reforms that caused poverty to rise. Shifts in urban dis-tribution, in contrast, reinforced poverty in that sector. Indeed,both a decline in mean incomes and adverse distributional shiftscontributed to the significant increase in urban poverty during theperiod under review. In general, knowledge of how much ofobserved changes in poverty are due to changes in the redistribu-tion as distinguished from growth in average incomes is criticalfor public policy and debate. These issues are particularly usefulin the context of poverty reduction under resource constraints, war-ranting the use of informed targeting schemes.

The implication of these findings is that growth would have a sig-nificantly positive impact on poverty alleviation and policies thatredistribute in favour of the rural and semi-urban areas could mar-ginally enhance the positive effects of growth on poverty allevia-tion. Such outcomes would curb massive migration into the maincities of the country and perhaps reduce the skewness in incomedistribution in the urban centres.


The decomposition analysis of poverty changes in this papercorroborated the general literature findings that growth effectstend to dominate the effects of changes in the distribution ofincome (Datt and Ravallion, 1992; McKay, 1997; Bigsten et al.,2003; Kakwani, 1997; Kabore, 2003). The results illustrated thepotential contribution of distributionally neutral growth in house-hold incomes to poverty alleviation in Cameroon. The temptationis resisted, however, to deny that although redistribution has animportant role to play, there must be severe limits to what can beachieved by redistribution in the absence of growth. This is reflectedin the marginal redistributional effects registered in the paper.Under all practical considerations, growth in household incomesappears more likely to be essential for long-term poverty reductionwhich will, however, be much less effective if the fruits of growthtend to be highly skewed towards the richest deciles. The bottomline is an advocacy for growth-based policies that create opportu-nities for the poor to increase their incomes.

References

Ali, A.A.G. (1997) ‘Dealing with Poverty and Income DistributionIssues in Developing Countries: Cross Regional Experiences’,paper presented at the AERC Bi-Annual Research Workshop,Nairobi, December 1996, Revised January 1997.

Baye, M.F. (1998) ‘Inequality and the Degree of Poverty among thePublic Sector Workers in Cameroon’, Nigerian Journal ofEconomic and Social Studies, 40 (3): 433–52.

Baye, M.F (2004) ‘Non-Parametric Estimates of Poverty Lines:Measuring Rural-Urban Contributions to Poverty Changes inCameroon’, paper presented in the CSAE Conference onGrowth, Poverty Reduction and Human Development inAfrica, Oxford, 21–22 March 2004.

Baye, M.F. (2005) ‘Alternative Methods for Setting Poverty Lines:Measuring Poverty in Cameroon’, Pakistan Economic and SocialReview, XLIII (1): 107–32.

Baye, M.F. and S. Fambon (2001) ‘The Impact of Macro and SectoralPolicies on the Extent of Poverty in Cameroon’, in F.M. Baye,S. Fambon and F. Sikod (eds), Globalisation and Poverty: The Roleof Rural Institutions in Cameroon, Tokyo, Japan: FASID.

566 F.M. Baye

Bigsten, A., B. Kebede, Shimeles and M. Taddesse (2003) ‘Growthand Poverty Reduction in Ethiopia: Evidence from House-hold Panel Surveys’, World Development, 31 (1): 87–106, alsoavailable on http://www.handels.gu.se/epc/archive/00001686/01/gunwpe0065.pdf.

Boccanfuso, D. and T.S. Kabore (2004) ‘Macroeconomic Growth,Sectoral Quality of Growth and Poverty in DevelopingCountries: Measure and Application to Burkina Faso’, paper pre-sented in the Conference on African Development and PovertyReduction: The Macro–Micro Linkage, Somerset West, SouthAfrica, 13–15 October 2004.

Canagarajan, S., J. Ngwafon and S. Thomas (1997) ‘The Evolution ofPoverty and Welfare in Nigeria’, Policy Research Working PaperNo. 1715, The World Bank.

Chantreuil, F. and A. Trannoy (1999) ‘Inequality DecompositionValues: the Trade-off Between Marginality and Consistency’,mimeo, Universite de Cergy-Pointoise.

Cornia, G.A., R. Jolly and F. Stewart (eds) (1995) Adjustment with aHuman Face: Protecting the Vulnerable and Promoting Growth,Oxford: Oxford University Press and UNICEF.

Datt, G. and D. Gunewardena (1997) ‘Some Aspects of Poverty inSri Lanka: 1985–90’, Policy Research Working Paper No. 1738,The World Bank.

Datt, G. and M. Ravallion (1992) ‘Growth and RedistributionComponents of Changes in Poverty Measures: ADecomposition with Application to Brazil and India in the1980s’, Journal of Development Economics, 38: 275–95.

Deaton, A. (2004) ‘Measuring Poverty in a Growing World (orMeasuring Growth in a Poor World)’, Mimeo: PrincetonUniversity.

Duclos, J.Y., A. Araar and C. Fortin (2003). ‘DAD 4.3-A: Software forDistributive Analysis/Analyse Distributive’, MIMAP program,International Development Research Centre, Government ofCanada and CREFA, Universite Laval.

Fambon, S., A.A. Amin, F.M. Baye, I. Noumba, I. Tamba andR. Tawah (2000). ‘Reformes economiques et pauvrete auCameroun durant les annees 1990’, Rapport final projet collabor-atif sur la pauvrete AERC, Nairobi, Kenya, May.


Fambon, S., A.A. Amin, F.M. Baye, I. Noumba, I. Tamba andR. Tawah (2001) ‘Pauvrete et Repartition des Revenus auCameroun durant les Annees 1990’, Cahier de Recherche,No. 01–06 de l’ Universite Laval, format PDF, available onhttp://www.crefa.ecn.ulaval.ca/cahier/liste01.html.

Fambon, S., F.M. Baye, I. Noumba, I. Tamba and A.A Amin (2004).Dynamique de la Pauvrete et de la Repartition des Revenus auCameroon Durant les Annees 80 et 90, Final Report, AERC,Nairobi, Kenya.

Foster, J., J. Greer and E. Thorbecke (1984) ‘A Class of DecomposablePoverty Measures’, Econometrica, 52 (3): 761–76.

Government of Cameroon (1984) Data Base of Budgetary andConsumption Survey, DSCN, Ministry of the Economy andFinance, Yaounde.

Government of Cameroon (1996) Data Base of ECAM I, DSCN,Ministry of the Economy and Finance, Yaounde.

Government of Cameroon (2003) The Poverty Reduction StrategyPaper, Yaounde: Ministry of Economic Affairs, Programmingand Regional Development.

Greer, J. and Thorbecke, E. (1986) ‘A Methodology for MeasuringFood Poverty Applied to Kenya’, Journal of DevelopmentEconomics, 24: 59–74.

Institute of Statistics (2002) Preliminary results of ECAM I andECAM II, Ministry of Economic Affairs, Programming andRegional Development.

Kabore, T.S. (2003) Dynamique de la pauvrete: Revue des approchesde decomposition et application avec des donnees du BurkinaFaso, available on http://132.203.59.36/PEP/Group/pmma-train/files/I-dynamique-pauvrete-kabore.pdf.

Kakwani, N. (1997) ‘On measuring Growth and InequalityComponents of Poverty with Application to Thailand’,Discussion Paper, School of Economics, The University of NewSouth Wales.

Khan, S. and I. Noumba (2001) ‘Rural Health Care and EducationalServices in Cameroon’, in F.M. Baye, S. Fambon and F. Sikod(eds), Globalisation and Poverty: The Role of Rural Institutions inCameroon, Tokyo, Japan: FASID.

568 F.M. Baye

Lynch, S.G. (1991) Income Distribution, Poverty and ConsumerPreferences in Cameroon, Washington, D.C: Cornell Food andNutrition Policy Program.

Mbanga, G. and F. Sikod (2002) ‘The Impact of Debt and Debt-Service Payments on Investments in Cameroon’, Final Reportsubmitted to AERC, Nairobi, Kenya.

McKay, A. (1997) ‘Poverty Reduction Through Economic Growth:Some Issues’, Journal of International Development, 9 (4): 665–673.

Moulin, H. (1988) Axioms of Cooperative Decision Making, Cambridge:Cambridge University Press.

Nadaraya, E.A. (1964) ‘On estimating regression’, Theory ofProbability and its Aplications, 10: 186–90.

Njinkeu, D., G. Kobou and I. Noumba (1997) ‘Structural Adjustmentand Poverty in Cameroon: A labor Market Analysis’, ICEG FinalReport, Nairobi, Kenya.

Owen, G. (1977) ‘Values of Games with Priori Unions’, in R. Heimand O. Moeschlin (eds), Essays in Mathematical Economics andGame Theory in Honour of Oskar Morgenstern, New York:Springer Verlag.

Ravallion, M. (1992) ‘Poverty Comparisons: A guide to conceptsand methods’, LSMS Working Paper No. 88, The World Bank.

Ravallion, M. (1998) ‘Poverty Lines in Theory and Practice’, LSMSWorking Paper No. 133, The World Bank, Washington DC.

Ravallion, M. (2001) Measuring Aggregate Welfare in DevelopingCountries: How well do National Accounts and Surveys Agree?Working Paper, The World Bank.

Ravallion, M. and Bidani, B. (1994) ‘How Robust is a PovertyProfile’?, The World Bank Economic Review, 8 (1): 75–102.

Rongve, I. (1995) ‘A Shapley Decomposition of Inequality Indices byIncome Source’, Discussion Paper No. 59, Department ofEconomics, University of Regina, Regina.

Shapley, L. (1953) ‘A Value for n-Person Games’, in H.W. Kuhnmand A.W. Tucker (eds), Contributions to the Theory of Games,Vol. 2, Princeton University Press.

Shorrocks, A.F. (1999) ‘Decomposition Procedures for DistributionalAnalysis: A Unified Framework Based on Shapley Value’,Mimeo: Department of Economics, University of Essex.


Shorrocks, A.F. and S. Kolenikov (2001) ‘Poverty Trends in Russiaduring the Transition’, mimeo, WIDER and University of NorthCarolina.

Silverman, B.W. (ed.) (1986) Density Estimation for Statistics and DataAnalysis, London: Chapman and Hall.

Watson, G.S. (1964) ‘Smooth Regression Analysis’, Sankhya, Series A,26: 359–72.

Woodward, D. (1992) Debt, Adjustment and Poverty in DevelopingCountries, London: Pinter Publishers in Association with Savethe Children.

World Bank (1995) Cameroon: Diversity, Growth and PovertyReduction, The World Bank Report No. 13167-CM.

Yatchew, A. (1998) ‘Non-parametric Regression Techniques inEconomics’, Journal of Economic Literature, 36: 669–721.

570 F.M. Baye

Documents

Growth, Redistribution and Poverty Changes in Cameroon