Embed Size (px)
Citation preview
This article was downloaded by: [York University Libraries]On: 11 November 2014, At: 01:23Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: MortimerHouse, 37-41 Mortimer Street, London W1T 3JH, UK
Middle East Development JournalPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/rmdj20
Multidimensional Fuzzy Poverty and Pro-Poor GrowthMeasures in Nonmonetary Dimensions in EgyptBetween 1995 and 2005Valérie Berengera
a CEMAFI, U niversity of Nice-Sophia Antipolis av. Doyen Trotabas 06050 Nice cedex 01Published online: 12 Feb 2014.
To cite this article: Valérie Berenger (2010) Multidimensional Fuzzy Poverty and Pro-Poor Growth Measures inNonmonetary Dimensions in Egypt Between 1995 and 2005, Middle East Development Journal, 2:1, 15-38
To link to this article: http://dx.doi.org/10.1142/S1793812010000204
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose ofthe Content. Any opinions and views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be reliedupon and should be independently verified with primary sources of information. Taylor and Francis shallnot be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and otherliabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to orarising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
Middle East Development Journal, Vol. 2, No. 1 (2010) 15–38c© Economic Research ForumDOI: 10.1142/S1793812010000204
MULTIDIMENSIONAL FUZZY POVERTY ANDPRO-POOR GROWTH MEASURES IN NONMONETARYDIMENSIONS IN EGYPT BETWEEN 1995 AND 2005∗
VALERIE BERENGER
CEMAFI, University of Nice-Sophia Antipolisav. Doyen Trotabas 06050 Nice cedex 01
Received 31 January 2010Revised 15 March 2010
The main goal of this paper is to adopt a multidimensional approach to poverty that goesbeyond focusing on a unidimensional measure of income or expenditure when attemptingto ascertain the main aspects of the living conditions of households. In order to obtainmultidimensional poverty measures the paper uses an approach based on fuzzy sets.This methodology is applied to data from the Demographic and Health Surveys for theyears 1995 and 2005 in order to obtain an aggregated index of the Standard of Livingfor each household. The evolution of the Standard of Living of households between 1995and 2005 as well as the uneven progress registered in the index of education as assessedby Egyptian HDR (2008) lead us to investigate the impact of non- income growth onpoverty. Thus, following the study of Grosse et al. (2008), extended growth incidencecurves (Ravallion and Chen, 2003) are applied to the index of the Standard of Livingand to education in order to assess if progress has been biased in favour of the poor.
Keywords: Fuzzy sets approach; multidimensionality of poverty; pro-poor growth; growthincidence curve.
1. Introduction
Over the last decade, Egypt has moved from a state-led economic system to amarket-oriented one with the adoption of its Economic Reform and StructuralAdjustment program (ERSAP) in 1991. The main objective of this program wasto achieve stable growth through a private sector led and market-oriented strategy,including fiscal reforms, trade liberalization, financial and labour market liberali-sation and privatization.
In order to limit the social costs of such reform, Egypt has adopted a gradualapproach and kept some forms of social justice through the Social Fund for Devel-opment. Nevertheless, poverty remains at high levels in comparison with othercountries in the MENA region (World Bank 2002). According to existing reports,
∗This paper has been finalized within the framework of the project “PROPOORSUDS” fundedby the ANR “French National Agency of Scientific Research”.
15
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
16 V. Berenger
the incidence of poverty increased between 1980–81 and 1990–91, decreased slightlyfrom 1990–91 to 1995–96 and from 1995–96 to 1999–2000, before starting to increaseagain in 2000 from 16.7% to 19.6% by 2005 (World Bank 2007) when the countrywas confronted with economic slowdown and high inflation.
In contrast, the Egyptian Human Development reports (HDR) find that non-monetary indicators of poverty do not follow the same trends. Egypt’s overallHuman Development Index (HDI) at the national level has consistently improvedsince 1993. A closer analysis of the HDI components of health, education and GDPper capita, however, reveals that progress in health and education has been uneven.
Despite the fact that the multidimensional nature of poverty is now well-recognized in the academic community as well as in international developmentinstitutions, studies of poverty in Egypt are still dominated by approaches usinga unidimensional, monetary measure. There is little work in this country or otherMENA countries taking a multidimensional approach to poverty. With this in mind,the main goal of this paper is to adopt a multidimensional approach to poverty thatgoes beyond a single measure of income or expenditure when attempting to discoverthe main aspects of the living conditions of households. This approach provides moreinsight about the various facets of poverty.
In order to obtain multidimensional poverty measures the methodology used isbased on the fuzzy set approach suggested by Cerioli and Zani (1990) followed byCheli and Lemmi (1995) who were the first to propose a multidimensional measureof poverty based on the theory of fuzzy sets. Fuzzy sets theory has the advantageof dealing with the vague and complex nature of poverty. Instead of dividing thepopulation between poor and non poor, the fuzzy approach takes into account acontinuum of situations between these two extremes. The methodology is applied todata from the Demographic and Health Surveys for the years 1995 and 2005 in orderto obtain an aggregated index of the Standard of Living for each household basedon indicators covering the various areas relevant to household living conditions.Due to the additive decomposability of the index, decomposition by sub-group ofpopulation (according to their socio-economic characteristics) and by dimensioncan be performed in order to determine where deprivation is most severe. Logitregression analysis is then used to explore the determinants of this multidimensionalmeasure of poverty, highlighting the key role played by education.
The evolution of the standard of living of households between 1995 and 2005 aswell as the uneven progress registered in the index of education as assessed by theEgyptian Human Development Report (2008) lead us to investigate the impact ofnon-income growth on poverty. For this purpose, the methodology used follows thestudy of Grosse et al. (2008) who were the first to introduce the multidimensionalityof poverty into pro-poor growth measurement. Extended growth incidence curves(Ravallion and Chen 2003) are applied to the index of the standard of living andto education in order to assess if progress has been biased towards the poor.
The paper is organized as follows. Section 2 presents the methodology usedin order to obtain multidimensional poverty measures and the results obtained
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
Multidimensional Fuzzy Poverty and Pro-poor Growth Measures 17
from the application of the methodology based on the fuzzy set approach to DHSdata for 1995 and 2005. In order to improve understanding of the results obtained,Sec. 3 analyzes the distribution of the Standard of Living of the household and thelevel of education between 1995 and 2005 by extending growth incidence curves tonon-monetary dimensions. Section 4 gives concluding comments and policy recom-mendations.
2. The Assessment of Multidimensional Poverty: A Method Basedon Fuzzy Sets
While the Human Development reports and the World Bank reports treat boththe human development and income concepts of poverty, in both academic andinstitutional spheres, poverty analysis in Egypt is still dominated by the monetaryapproach. The first studies on poverty date from the 1980s when the Egyptianeconomy faced pronounced transformations. With the initiation in the early 1990sof structural economic reforms (known as ERSAP), studies on poverty show signif-icant disagreement between authors in terms of the trend and incidence of poverty(Yamada 2008). The main divergences are due to the methodology used to estimatethe poverty line. According to World Bank (2002), these differences arise from sev-eral technical aspectsa making estimates of poverty provided by different studiesdifficult to compare. In their works, El-Laithy and Osman (1997), the World Bankand the Ministry of Planning (MOP) have resolved these problems. The measure-ment of poverty lines has been improved by defining household specific povertylines instead of a per capita poverty line as explained in the “Poverty Reduction inEgypt-Diagnosis and Strategy” report (World Bank and MOP 2002). Nevertheless,the studies following this approach produce poverty profiles based on a unidimen-sional monetary indicator. Among the most significant and original studies are thoseof Datt et al. (1998) and of Datt and Jolliffe (1999). These analyses follow the con-ventional view of poverty with a focus on insufficient resources for securing basicgoods and services. Since the seminal works of Sen (1985) and Townsend (1979)however, arguments have been advanced to go beyond these money-metric mea-sures and consider other approaches that adopt a broader definition of well-beingincluding a wide array of components ranging from nutrition and calorie intake tofreedom to achieve certain social arrangements (Baliamoune 2006) that are imbed-ded in the promotion of the Millennium Development Goals (MDG) by the UnitedNations.
Whereas poverty profiles and the UNDP reports since 1994 acknowledge thecomplex and multiple facets of poverty, there does not seem to have been muchwork in Egypt taking a multidimensional approach to poverty. To our knowledge,
aThe methods used do not take into account economies of scale within households, consumptionpatterns in different regions, hypothetical cost of the diet for the poor, the differing “basic needsrequirements of different household members (young versus old and male versus female).
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
18 V. Berenger
the only exceptions are the works of Bibi (2004) on Egypt and Tunisia, Bibi et al.(2008) on South Africa and Egypt, Osman et al. (2006) and Berenger et al. (2009)on Egypt, Morocco and Tunisia using a methodology proposed by Deutsch andSilber (2008).
Selecting a multidimensional approach to poverty implies addressing issues thatneed not be faced when taking a unidimensional approach. Thus, in addition tothe problem of determining a poverty line and choosing a measure of poverty, amultidimensional approach forces the researcher to select the relevant dimensionsof poverty, the indicators supposed to represent these dimensions and finally theweights to be given to these various dimensions and indicators. Two alternativestrategies can be identified in the literature in order to implement a truly multidi-mensional poverty analysis. The first one consists of considering several indicatorsof well-being by studying each of them independently. The second one aggregatesthe set of well-being indicators into a single index and then proceeds to povertymeasurements by fixing a poverty line.
Difficulties raised by the choice of the poverty line suggest that it is not easyto achieve a wide consensus in setting such a limit. For this reason, the divisionis established with imprecision and ambiguity, suggesting the importance of con-sidering instead the continuum of situations between poor and non poor. Thus,well-being of individuals can be viewed as a matter of degree, a vague predicatewhich makes the traditional division of the population between poor and non pooran over simplification. This consideration is all the more justified within the multidi-mensional framework of poverty measurement. With such methods, various aspectsof poverty can be captured and summarized in a single index.
Derivation of a single index of poverty using information included in severalindicators, however, raises another problem. For example, in some cases a personcan be in such a state of deprivation in one sphere that he or she is deemed to be poorwhile in other spheres he or she certainly should not be classified as poor. As a result,it is not clear how poverty should be measured to take into account informationprovided by several indicators. The arbitrariness inherent in the identification ofthe poor according to a poverty line and the need to combine various, sometimesvague aspects comprising a full concept of poverty have led to the search for newmethodological tools in order to deal with these two aspects.
2.1. A methodological approach using fuzzy set theory
In the last decade, Cerioli and Zani (1990) followed by Cheli and Lemmi (1995)were the first to propose multidimensional measures of poverty based on the theoryof fuzzy sets introduced by Zadeh (1965) and developed by Dubois and Prade(1980). Fuzzy sets theory has the advantage of dealing with the vague and complexnature of poverty.b This approach applies to situations where one is unable to
bFor more details, see Chiappero-Martinetti (2000).
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
Multidimensional Fuzzy Poverty and Pro-poor Growth Measures 19
determine which elements belong to a given set and which ones do not. Zadehhimself (1965) characterized a fuzzy set (class) as “a class with a continuum ofgrades of membership”.
Let there be a set X and let x be any element of X . A fuzzy set or subset A ofX is characterized by a membership function µA(x) that will link any point of X
with a real number in the interval [0, 1]. µA(x) is called the degree of membershipof the element x to the set A. If A were a set in the sense in which this term isusually understood, the membership function which would be associated to this setwould take only the values 0 and 1. But if A is a fuzzy subset, we will say thatµA(x) = 0 if the element x does not belong to A and that µA(x) = 1 if x completelybelongs to A. But if 0 < µA(x) < 1, x belongs only partially to A.
This simple idea may be easily applied to the concept of multidimensionalpoverty based on micro-level data and to well-being measures at the macro-level(Berenger and Verdier-Chouchane 2007). Then, the methodology involves twosteps.
The first step concerns the definition of the membership function to a given setassociated with each household and indicator. While the membership function cantake several formulations (Lelli 2001; Baliamoune 2006 and Chiappero-Martinetti2006), we consider the “Totally Fuzzy Analysis (TFA)” as defined originally byCerioli and Zani (1990) in contrast to the “Totally Fuzzy and Relative” (TFR)defined by Cheli and Lemmi (1995).c For each household, we select k indicatorswhich reflect the main relevant aspects of living conditions. They can be viewed ascomponents of the Standard of Living of households. The value of the membershipfunction will provide for the household’s deprivation degree relative to a givenindicator, increasing linearly between zero and one.
In a second step, the different degrees of deprivation obtained for each house-hold and each indicator need to be summarized in order to obtain a compositeindex of Standard of Living (SL) for each household. In this perspective, povertycan be defined as an accumulation of “deprivations” or “shortfalls” according tothe different dimensions considered. Inversely, the index value can be interpreted asan accumulation of “effective achievements”. According to Cerioli and Zani (1990),composite indices are defined by taking the weighted arithmetic mean of the mem-bership functions according to the k indicators:d
The weights are based on frequency of symptoms of poverty and are definedas an inverse function of the mean deprivation level relative a given indicator.e
cFor more details about the technical aspects of the TFA, the reader can refer to the paper ofBerenger and Verdier-Chouchane (2007).dAs Chiappero-Martinetti (1996) emphazises’ the function must have a value ranging between themaximum and the minimum and must allow interaction between the various indicators.eSeveral approaches have been suggested in the literature like adopting an agnostic attitude andconsidering that all dimensions are equally important or deriving the weighting from the datausing multivariate statistical methods like PCA. Between these two approaches, an often usedalternative since the study of Cerioli and Zani is the frequency-based weighting suggested byDesai and Shah (1988) that allows to define a ranking between dimensions.
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
20 V. Berenger
In this manner, a more important weight is given to the indicators that are morewidespread. As an illustration, if access to improved water represents a fundamentalbasic need, it follows that a lack of access to this service must be considered asymptom of poverty.
According to the given formulations above, the higher the Standard of Living(SL) for a given household, the closer to zero is the value index. Inversely, the closerthe value of the composite index is to one, the higher is the degree of deprivationrelative to SL.
Then, it is possible to derive a general fuzzy poverty index by averaging theindividual membership functions. According to Cerioli and Zani, this index rep-resents the proportion of individuals (households) “belonging” in a fuzzy sense tothe poor subset (p. 282). In addition, due to the additively decomposable nature offuzzy indices, they can be broken down by dimensions and sub-group of populationto provide important basic information about the level and the structure of povertyand particularly about dimensions that contribute most to the state of deprivation.Unlike the traditional approaches that advocate transfer policies, this approach iscapable of providing a comprehensive analysis of the main causes of poverty forthe design of structural socio-economic policies that could alleviate poverty in thelong run.
2.2. Evolution of fuzzy multidimensional poverty in Egypt
between 1995 and 2005
2.2.1. Data description
Although household budget surveys have been collected regularly in Egypt, theyare retained within government agencies and are not made public. However, theuse of Demographic and Health Surveys (DHS) initiated by the U.S. Agency forInternational Development (USAID) offers an alternative instrument. DHS havebeen collected for several countries and for several time periods. As the data arestandardized for all countries, it makes it possible to draw comparisons across timeperiods within a country and across countries. However, as the intention of thesesurveys is to present information on health, education and nutritional status, theydo not include information on income and expenditure data which is a seriousdrawback. In the absence of monetary indicators, we have used individual andhousehold schedules from DHS 1995 and 2005. The indicators have been selectedfollowing the distinction made by Cheli and Lemmi (1995) between “cause” and“effect” indicators. The former provide a measure of the risk of poverty, while thelatter capture the degree of real unsatisfied basic living conditions.
We have considered 18 indicators of “effect” that make it possible to apprehendpoverty as deprivation of standard of living (Annex 1). The same indicators orproxies have been selected for each year in order to ease comparisons across time.They are classified in four dimensions in order to circumvent measurement errorsimbedded when using single indicators. Thus, the dimension Durable includes the
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
Multidimensional Fuzzy Poverty and Pro-poor Growth Measures 21
ownership of a radio, washing machine, television, phone, heater, cooking fuel (ormodern stove), sewing machine, etc. The dimension Housing conditions refers to thequality of house occupied by the household as well as the living comfort. It includesthe number of rooms per capita in order to determine overcrowding, the existenceof a kitchen as a separate room and floor construction materials. The dimensionService refers to indicators related to the access of basic services captured by theaccess to electricity, to improved water source and the kind of toilets facilities asproxy for the presence of a sewerage system. Finally, other Properties include theownership of a dwelling and a car as proxies of the ownership of assets.
2.2.2. Results of fuzzy poverty measures
The methodology presented above is applied to the two data sets to obtain fuzzypoverty measures as well as its decomposition by dimensions. Table 1 presentspoverty measures in terms of standard of living including all four dimensions andby each dimension separately.
The results in Table 1 indicate that 28.46% and 18.19% of households are struc-turally poor in 1995 and 2005 respectively and that poverty has decreased overtime in each dimension for the overall sample. Estimations of indices by dimen-sion indicate that Properties emerges as the most serious dimension of poverty forthe two years considered, followed by Housing and then by Durable. Whereas thevalue of the dimension Properties remains stable over time, the Housing dimensionrecords significant improvements between 1995 and 2005 (more than 16%), followedby Durable and Service dimensions (12% and 8% respectively).
Moreover, decompositions by dimension and by area of residence make it possi-ble to provide interesting information about the symptoms and causes of poverty.The results displayed in Table 2 reveal that poverty is higher in rural than in urbanareas in general as well as in each dimension except Properties. While the percent-age of poor households in terms of Properties is higher in urban areas, rural areasare characterized by bad housing conditions since 57.19% and 32.90% of house-holds are poor in 1995 and 2005 respectively. Whereas urban poverty records mod-est changes, poverty declines significantly in rural areas in all dimensions except
Table 1. Statistical indicators of poverty in standard of livingand by dimension.
% Poor Standard-error Critical value
Years 1995 2005 1995 2005 1995 2005
Total 28.46 18.19 0.152 0.111 0.365 0.262Durable 33.07 21.10 0.233 0.209 0.445 0.325Properties 41.68 40.38 0.352 0.376 0.195 0.165Housing 40.84 24.81 0.280 0.219 0.395 0.345Service 14.25 6.00 0.173 0.083 0.255 0.235
Source: Author’s calculations from DHS 1995 and 2005.
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
22 V. Berenger
Table 2. Fuzzy poverty measures by dimension and area of residence.
Urban (%) Rural (%)
Years 1995 2005 1995 2005
Total 13.97 14.94 41.22 21.19Durable 15.94 13.84 48.47 27.88Properties 52.94 52.46 22.04 28.53Housing 18.80 16.78 57.19 32.90Service 2.17 2.77 24.71 9.11
Source: Author’s calculations from DHS 1995 and 2005.Note: Urban and rural areas cover respectively 7397 and 8170 house-holds in 1995 and 10555 and 11417 households in 2005.
Properties where poverty increases between 1995 and 2005. The most significantimprovements concern the Housing and Durable dimensions. Although the povertygaps between urban and rural areas fall, rural poverty remains higher than urbanpoverty.
Finally, decompositions by region of residence indicate that poverty is not onlya rural but also a regional phenomenon (Annex 1, Table 1.A). Thus, in the presentinstance, rural Upper Egypt experiences deprivations in all dimensions exceptProperties. On the other hand, urban Governorates and urban Lower Egypt havethe lowest percentage of fuzzy poor households except in the case of Properties.Whatever the region, poverty has decreased between 1995 and 2005 while the WorldBank report (2006) indicates an increase in (unidimensional) monetary poverty inUpper Egypt.
2.2.3. Determinants of multidimensional fuzzy poverty
Due to its definition, the TFA method does not make use of a poverty line andcannot theoretically be used to separate the sample into “poor” and “non-poor”.However, it is possible to adopt a dichotomous approach to derive a critical valuefrom the cumulative distribution of the deprivation index in terms of standardof living and its components (see the last column of Table 1). This critical valueserves as a threshold to estimate the number of households experiencing a genuinedeprivation in a particular dimension.f
The dichotomisation of the approach is also useful in order to identify the factorsthat contribute to increase the risk of poverty using logistic regression models.Insofar as our poverty measures have been based on “effect” indicators, we now
fThe critical value (or breaking value) µj crit associated to indicator j can be defined as:
F (µj crit) = 1 − µj
With F the cumulative distribution function and µj the average value of indicator or dimensionj which indicates, in a dichotomous perspective, the proportion of poor households according toj (see Berenger et al., 2007). This critical value can serve in order to derive the traditional FGTpoverty measures (see Berenger, 2008).
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
Multidimensional Fuzzy Poverty and Pro-poor Growth Measures 23
consider “cause” indicators that refer to the socio-economic characteristics of thehouseholds: sex and age of the head of household, number of children under 5 yearsold, highest level of education and marital status of the head of household and thearea of residence (urban/ rural).g
The results obtained from the logit regressions for the two years indicate thatthe age and the education level of the head of household as well as the number ofchildren and the area of residence have statistically significant coefficients (Annex2, Tables 2.A and 2.B) Surprisingly, gender of the head of household does not sig-nificantly influence the risk of poverty regardless of the specification of the model.At best, its impact is slightly significant in 1995 when considering the inclusionof interaction terms between gender and marital status. Thus, in 1995, house-holds whose head is a single woman have, ceteris paribus, a higher probabilityof being poor than a single man. In the same way, this probability is also higherwhen the head of the household is a widow. On the other hand, the risk of beingpoor decreases, ceteris paribus, for married and divorced women. These results dif-fer to a certain extent from those assessed in World Bank (2002). According tothis report, female-headed households actually have lower poverty incidence andpoverty gaps than male-headed households, especially in rural areas. This rathersurprising result may be explained by the low percentage of female-headed house-holds and by the fact that the vast majority of them are widows and old havinga greater command over resources through social transfers than the population ingeneral.
The findings support the conclusion of the other studies on poverty that takea unidimensional monetary approach. But unlike those studies, they provide moreprecise information about the spheres in which deprivations are the most severe.In all cases, despite the adoption of important structural reforms, poverty hasdecreased over the period. These findings can be related to the fact that the imple-mentation of economic reform and structural adjustment program (ERSAP) wasrelatively slow and gradual. In addition, the government of Egypt kept some formsof social justice such as the social insurance system and special programmes throughthe Social Fund for Development.
However, despite impressive progress in rural areas, poverty remains a ruralphenomenon especially in Upper Egypt. A large proportion of rural population isstill deprived in terms of basic social needs as revealed by our results. Moreover,education emerges as one of the main determinants of poverty. The governmentshould, therefore, pursue and strengthen its efforts to bridge the development gapbetween urban and rural areas as well as between regions.h In addition, policies to
gAlthough the employment status of the head of household might have an impact on the proba-bility of being poor, the quality of data does not make it possible to include it in our analysis.hIn 1994, the Ministry of Rural Development initiated a rural development program called Shoroukassisted by USAID. The program encourages local development through local planning and civilsociety engagement. For more details, see World Bank, Policy note, Report No. 36432-EG, June2006.
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
24 V. Berenger
alleviate poverty have to benefit and empower the poor through universal access toeducation as well as improvements in the quality and efficiency of the educationalsystem.
3. The Assessment of the Pro-Poorness of Growth in NonMonetary Dimensions
The decline in poverty between 1995 and 2005 raises the question of the linksconnecting growth, inequality and poverty. The link between patterns of growth andpoverty is an important issue that has attracted the attention of academia as wellas development practitioners since the most important goal of development duringthe last decade has been poverty reduction. Beside the recognition of the multiplefacets of poverty, an emerging consensus is that growth alone is necessary but notsufficient for poverty reduction. The notion of pro-poor growth is concerned with theinterrelationship between three concepts: poverty, growth and inequality. Broadlyspeaking, the goal is a development strategy that enables the poor to benefit fromand contribute to the growth process. As a result, it differs strikingly from thegrowth strategies advocated by the trickle down theory popularized in the 1970s.However, there remains a lively debate over its definition and its measurement asit depends on the social evaluator’s definition of what constitutes pro-poor growth(Kakwani and Pernia 2000; Ravallion and Chen 2003 and Kraay 2006).
The traditional tools to assess the pro-poor nature of growth focus on a sin-gle, monetary measure of poverty. For instance, in the case of Egypt, the mostcomprehensive studies on this topic have been performed by Kheir-El-Din et al.(2006) and by El-Laithy et al. (2009). In the first study, the authors study theEgyptian growth experience over the period 1990/91–2004/05 along three levels ofaggregation. The results obtained at the household level using expenditure datafrom HIECS (Household Income and Expenditures Surveys) indicate that over thewhole period, the expenditure distribution has markedly improved with a declinein poverty incidence. But this pattern has not been uniform over sub-periods. Newand updated findings are obtained in the study of 2009 using HIECPS panel dataconducted by CAPMAS which make it possible to track household consumptionand living standards over 2005–2008. This study shows that due to rapid economicgrowth, Egypt has achieved poverty reduction. However, economic growth has notresulted in universal improvements of living conditions.
All these studies share in common the fact that poverty is conceived in termsof a lack of income or expenditure and can therefore be measured by a single,monetary indicator. However, if poverty is to be considered and measured takinga multidimensional approach, it is then necessary to look at the pro-poor natureof growth beyond the sole focus on the monetary aspect of poverty. With respectto that issue, Grosse et al. (2008) have proposed to make use of the tools devel-oped for pro-poor growth to investigate the distribution of change in non-monetarydimensions of well-being.
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
Multidimensional Fuzzy Poverty and Pro-poor Growth Measures 25
3.1. Extension of growth incidence curve to non-monetary
dimensions
In order to assess pro-poorness of growth in non-monetary dimensions, Grosse et al.(2008) suggest an extension of the growth incidence curve (GIC) in order to takeinto account the changes in the distribution of non-income indicators of well-beingconditional or not on the distribution of a monetary indicator. The GIC have themerit of providing a graphical tool and an intuitive approach to pro-poor growth.Thus, by definition, GIC indicate the rate of growth by percentile across the dis-tribution for a given period.
Following Grosse et al. (2008), GIC can be expressed in relative as well as inabsolute terms. In particular, absolute GIC are useful for non-income dimensionssince they are usually measured on a discrete rather than a continuous basis becausecomparisons of growth rate can be misleading. Indeed, non-income dimensions,such as, education or health, can have upper limits so it may be the case thatsome households have reached the upper limit so that further improvement is notpossible. At the other end of the distribution, relative measures based on growthrates tend to overstate the improvement for households with low starting levels.Consequently, the improvement for households with a low level of achievement in aparticular non-economic dimension might not have the same interpretation as thesame variation for households less deprived in that dimension.
Following Grosse et al., the absolute non-income growth incidence curve(NIGIC) is defined as follows:
Absolute NIGIC = ct(p) = yt(p) − yt−1(p)
with y the value of the non-income indicator. Absolute NIGIC shows the absolutechanges for each percentile. It follows that if NIGIC is negatively sloped, it indicatesstrong absolute pro-poor growth. In the same way as the PPGR (pro-poor growthrate), pro-poor change (PPCH) is the area under the NIGIC up to the headcountratio H :
PPCH = cpt =
1Ht−1
∫ Ht
0
ct(p)dp
By comparing PPCH with the change in mean (CHIM), the distribution of changeswith respect to the non-income attribute is said to be pro-poor in the strong absolutesense if PPCH exceeds CHIM.
The NIGIC can be understood in two different ways. The first way, calledunconditional NIGIC (UNIGIC), investigates the changes in the distribution of thenon-income indicators regardless of the ranking of the household with respect toincome or wealth. The UNIGIC allows tracking progress towards the achievementsof MDGs for each percentile of the distribution. The second way is conditionalNIGIC (CNIGIC) and considers the absolute changes of the non-income indicatorbased on the ranking of the household in the income distribution. The CNIGIC
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
26 V. Berenger
provides information on how the improvements in non-income well-being indica-tors are distributed across the income distribution. It is of particular interest if weregard progress in non-income dimension for lower income groups as being moreimportant from the point of view of social welfare than the increases for higherincome groups.
3.2. The pro-poorness of changes in standard of living and
education achievements of Egyptian households
3.2.1. Pro-poor nature of changes in standard of living
This methodology can be applied to the above distribution of deprivation scores ofthe standard of living defined for each household using the fuzzy set approach. Thescore obtained can be conceived as an indicator of the standard of living for eachhousehold. It can be viewed as an indicator of the achievements made possible byincome. Thus, Fig. 1 displays the distribution of changes in the standard of livingbetween 1995 and 2005 for all households and by urban and rural areas.
The shape of the curves indicates that progress in the standard of living has ben-efited the poor in a strong absolute sense. Based on the critical value for 1995, thevalues of PPCH are 0.1596, 01674 and 0.1444 respectively for all households, ruralones and urban ones, and are greater than the absolute mean changes CHIM (0.103,0.1253, 0.081 respectively). In particular, we observe that household standard ofliving in rural areas (PPCH of 0.1674) is catching up with the standard of living
Fig. 1. Absolute NIGIC for household standard of living: 1995–2005.Source: Author’s calculations from DHS 1995 and 2005.
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
Multidimensional Fuzzy Poverty and Pro-poor Growth Measures 27
in urban areas (PPCH of 0.1444). This finding supports the results obtained frompoverty measures presented in Sec. 2.
3.2.2. Extension to the education dimension
The investigation of the determinants of multidimensional poverty has highlightedthe impact of area of residence and level of education on the probability of beingpoor. Education is one of the components of the HDI for assessing the relativeperformance of countries in terms of human development. Moreover, several stud-ies have shown that education would be the best indicator for the design of anindex to capture non-economic aspects of human well-being (McGillivray 2005).Finally, UNDP’s country report on education in Egypt in 1999 emphasizes thecritical role of education as a powerful instrument for achieving and sustainingeconomic growth, reducing poverty and enhancing equity. On this matter, therehave been significant improvements in Egypt’s overall HDI score since 1996. How-ever, a closer analysis of the HDI components reveals that progress in HDI hasnot always been consistent (Egypt’s Human Development Report 2008). While thelife expectancy index has steadily increased with the largest improvement after2001, the improvements in the education index have been more uneven. These var-ious reasons point to the importance of investigating the distribution of changeswith respect to education. As education indicator, we consider the average years ofschooling per household for all household members and then for all adults and allchildren separately.i Table 3 shows achievements in education for 1995 and 2005.Thus, average levels of education increase and the Gini indices fall.j Progress is moresignificant in rural than in urban areas in spite of the fact that gaps with urban areaspersist.
Figure 2 displays the absolute NIGIC for education for all households and byarea of residence. We do not find strong absolute pro-poor progress as the slopes ofthe absolute curves are not negative and are even positive for the poorest deciles. In
Table 3. Average households’ education: 1995–2005.
Households Total Urban Rural
Years 1995 2005 1995 2005 1995 2005
Mean 4.91 5.72 6.49 6.99 3.45 4.54GINI 0.42 0.36 0.35 0.31 0.45 0.37% no education 7.10 3.49 3.65 3.18 10.33 6.94
Source: Author’s calculations from DHS 1995 and 2005.
iIn Egypt, primary education starts at age six and consists of five (six since 2004) years of schooling.A further three-year period, known as the preparatory stage is considered as basic education andis compulsory. The second stage not compulsory includes another three years.jAs pointed out by Grosse et al. (2008), variations of Gini indices are related to the fact thathouseholds with high level of schooling are quite close to the upper limit and the households withlow education level are catching up.
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
28 V. Berenger
Fig. 2. Absolute UNIGIC for average education per household: total, urban and rural.Source: Author’s calculations from DHS 1995 and 2005.
Table 4. Average years of education for the young per householdand by area of residence.
Young Total Urban Rural
Years 1995 2005 1995 2005 1995 2005
Mean 4.89 5.30 5.82 5.83 4.25 4.89GINI 0.41 0.38 0.37 0.35 0.44 0.39% no education 9.13 7.88 4.74 6.31 11.64 8.82
Source: Author’s calculations from DHS 1995 and 2005.
all cases, increases in the education levels for households between the 20th and the80th percentile are relatively identical. Except for the 10th percentile, the averagelevel of education has increased more in rural than in urban areas. This findingmight be explained by the initially high gaps between the education levels in thetwo areas. In the same way, whereas the education level increases exponentiallyfor the highest percentiles in rural area, it decreases in urban areas which may beexplained by measurement errors as well as by the fact that no further progress ispossible when the upper limit is reached.
The above analysis did not take into account the composition of the householdaccording to the age of its members. It may, however, be more revealing to considerthe distinction between the average education of the young (6–19) and the adults(older than 20 years). Education achievements between 1995 and 2005 for the youngand for adults are reported in Tables 4 and 5 respectively.
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
Multidimensional Fuzzy Poverty and Pro-poor Growth Measures 29
Table 5. Average years of education for adults per householdand by area of residence.
Adults Total Urban Rural
Years 1995 2005 1995 2005 1995 2005
Mean 5.77 7.08 8 8.84 3.70 5.46GINI 0.50 0.41 0.39 0.32 0.57 0.46% no education 19.95 13.52 9.45 8.19 29.04 17.75
Source: Author’s calculations from DHS 1995 and 2005.
Fig. 3. Absolute UNIGIC in education for the young: total, urban and rural.Source: Author’s calculations from DHS 1995 and 2005.
Whereas the average level of education of the young is flat in urban areas, itincreases by 0.64 years in rural areas. Gini indices as well as the percentage of youngwithout education support the significant improvements evidenced in rural areas.In the same way, Table 5 shows that adults’ average level of education is higherthan that of the young except for adults living in rural areas in 1995. Inequalityis higher among adults than among the young with a higher decrease for adults inrural areas over time.
Figures 3 and 4 present the unconditional NIGIC for education for the youngand adults respectively. For the young (Fig. 3), the absolute NIGIC shows no clearpro-poor growth trend. Whatever the area of residence, the poorest percentiles interms of education have not improved their educational attainment and we evenfind deterioration for the poorest urban young. In line with the preceding results,the improvements are more significant in rural areas whereas the curve is virtually
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
30 V. Berenger
Fig. 4. Absolute UNIGIC in education for adults: total, urban and rural.Source: Author’s calculations from DHS 1995 and 2005.
flat and close to 0 for the young in urban areas. Moreover, except for the extremepercentiles of the distribution, improvements in rural areas have been distributedequally across the education groups and the poorest have not benefited dispropor-tionately from the improvements.
Figure 4 for adults mirrors the picture that was found for the young in thatthe strongest increases are in rural areas and for some medium groups but not forthe poorest. Unlike Fig. 3, the progress in urban areas for adults starts to decreasefrom the 60th percentile as indicated by the downward slope of the curve reflectingthe fact that those with a relatively high education level have already approachedthe upper limit.
Although we have not used monetary measures in this paper and have not basedour measures of poverty on a unidimensional indicator, we can use our fuzzy depri-vation indicator obtained for each household in order to derive NIGIC conditionalon our multidimensional indicator of the standard of living. The conditional NIGIC(CNIGIC) allows us to check whether the progress in education has reached themost deprived population groups. As shown in Fig. 5, the highest improvementsoccur in the first part of the distribution until approximately the 50th percentile ofthe most deprived households and slightly increase with the standard of living. Onthe other hand, in urban areas, households around the 60th percentile are facinga deterioration in their educational attainment which may be due to the nature ofthe indicator or to reform of the education system.
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
Multidimensional Fuzzy Poverty and Pro-poor Growth Measures 31
Fig. 5. Absolute CNIGIC in education: total, urban and rural.Source: Author’s calculations from DHS 1995 and 2005.
More interesting are the results obtained for the young and adults. As shown inFig. 6, the conditional NIGIC indicates improvements in education for the low tailof the distribution until the 50th percentile. However, the curve is relatively flat forthe young in urban areas and deterioration of the education occurs for the highestpercentiles. In contrast, in rural areas, improvements in education are higher andseem to be equally distributed rather than biased in favour of the lowest percentilesof the distribution. This finding is also reflected in comparing PPCH with CHIM.Except for the entire sample, whatever the area of residence, young people living inthe poorest households in terms of standard of living have not been able to improvetheir education level by more than the average. Indeed, the PPCH for the wholesample and by area of residence based on the critical value of the fuzzy index for1995 (PPCHall = 1.04, PPCHrural = 0.33 and PPCHurban < 0) are lower than thevalues of the absolute change in the mean (CHIMall = 0.36, CHIMrural = 0.57 andCHIMurban = 0.35) for rural and urban areas.
Finally, the conditional NIGIC for adults in Fig. 7 is virtually flat with lowerfluctuations than for the young meaning that the distribution of changes for adultsis not pro-poor in a strong absolute sense. The distribution of changes by area of res-idence shows different patterns. The increase in education is higher in urban than inrural areas for the first 20th percentiles of the most deprived. While adults betweenthe 60th and the 70th percentiles show deterioration in urban areas, the same per-centiles reveal the strongest improvements in rural areas. These findings need to be
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
32 V. Berenger
Fig. 6. Absolute CNIGIC in education for the young: total, urban and rural.Source: Author’s calculations from DHS 1995 and 2005.
Fig. 7. Absolute conditional NIGIC in education for the adults: total, urban and rural.Source: Author’s calculations from DHS 1995 and 2005.
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
Multidimensional Fuzzy Poverty and Pro-poor Growth Measures 33
treated with caution since the households belonging to the same percentile of thedistribution do not have the same standard of living in rural and urban areas.
Thus, as indicated in Sec. 2, deprivation in rural areas is higher than in urbanareas. Although the patterns of the conditional NIGIC make it difficult to claimstrong pro-poorness of changes, the higher changes in education level in rural thanin urban areas revealed by the NIGIC suggest that the most deprived seem to havebenefited from improvements. The progress in educational attainment in rural areasboth for the young and for adults has, however, been equally distributed acrossthe distribution of education with the highest increase for some of the mediumgroups and not the poorest groups. The shape of the conditional NIGIC is morevolatile making it more difficult to conclude whether the improvements in educationare focused more on the initially poor in terms of standard of living. Unlike theunconditional NIGIC, conditional NIGIC indicate improvements occurring at theleft tail of the distribution until the 60th percentile for the entire samples. Thissuggests that the expansion of education has been to the benefit of the poor to amodest extent.
4. Conclusion
Unlike the existing studies on poverty that focus on income or expenditure as thesole dimension of poverty, we have taken a multidimensional perspective in orderto investigate poverty in multiple, non-monetary dimensions. Using a methodologybased on the fuzzy set approach, fuzzy poverty measures have been estimated interms of a composite measure of the standard of living and its four componentdimensions. Our findings indicate that poverty in terms of the standard of livingdecreased over the period 1995–2005 with the most significant improvements occur-ring in the housing component followed by access to durable goods and to basicinfrastructure. Decompositions by area of residence show that poverty declined sig-nificantly in rural areas despite the fact that gaps with urban areas persist. Thesefindings do not support those based on income as the sole dimension of povertyreported by Kheir-El-Din (2006) who conclude that poverty in 2004/05 is not sig-nificantly different from that in 1995/96. Analysis of the determinants of povertyin terms of the standard of living also shows that the level of education has asignificant impact on the risk of being multidimensionally poor.
The decline in poverty between 1995 and 2005 shown in this paper, led us toexamine the extent to which changes in the standard of living and education havebeen biased in favour of the poor. Thus, following Grosse et al. (2008), we inves-tigated the distribution of the change in non-monetary dimensions of well-being.Applied first to the fuzzy index of the standard of living, we find that the poor bene-fitted in a strong absolute sense. In particular, household standard of living in ruralareas is catching up with that in urban areas. Second, the distribution of changesin educational attainment has been analyzed across education groups. The resultsreveal that in spite of significant progress in educational attainment in rural areas
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
34 V. Berenger
both for the young and for adults, the improvements have been equally distributedacross the distribution of education with the highest increase being for some mediumgroups and not for the poorest groups in terms of education. And third, unlike theresults for the unconditional NIGIC just discussed, the conditional NIGIC indicatesimprovements occurring at the left tail of the distribution until the 60th percentilefor the entire sample. This suggests that the expansion of education has been tothe benefit of the poor to a modest extent. However, these results do not give anyindication of the quality of the education system, a critical problem in Egypt.k
These findings need to be linked to the social and economic context of Egypt.Indeed, despite the fact that the Egyptian government introduced significant eco-nomic reforms, it has not developed a national strategy to reduce poverty untilrecently. It has designed and implemented ad hoc interventions aimed at povertyreduction, resulting in a complex social protection system, considered to be highlyinefficient, that includes subsidies on multiple private and public goods and othersocial transfers in cash. Given the concern for the widening gap between urbanand rural areas, the government has established several programs over the period.The Shorouk program implemented in 1994 was expected to address all aspects ofdevelopment. However, three-quarters of the fund have gone to infrastructure andsanitation. The Social Development Fund originating in the early 1990s that wasestablished to alleviate negative impacts of ERSAP on the youth and poor segmentsof the population is involved in small and micro credit, public works projects andhuman resource development. In addition to these funds, the Emergency Fund wascreated in 1998 in order to meet urgent needs in rural areas.
Only recently has poverty reduction been incorporated in the framework ofthe National five-year plan for socio-economic development. The poverty reductionaction plan proposes interventions in the form of policy changes as well as programsthat directly or indirectly affect poverty. According to our findings, the structuralnature of poverty needs to foster the transfer of assets to the poor groups of thepopulation by enhancing the access to basic infrastructure in rural areas and socialservices rather than safety net programs. As suggested by the relatively low pat-tern of pro-poor changes in education, alleviation of chronic and intergenerationalpoverty might be addressed by considering transfers conditional upon educationattendance as in the Progresa (Oportunidades) program initiated in Mexico in1997 which targeted only rural areas.l
Although more precise policy recommendations would require a deeper analy-sis, two main recommendations can be drawn from our results. The first one refersto education. In particular, sources of growth in Egypt are volatile and unpre-dictable since they are dependent on revenues from oil and on remittances. From
kSee, World Bank (2007), “Arab republic of Egypt improving quality, equality, and efficiency inthe education Sector: fostering a competent generation of youth,” Report No. 42863-EG, June.lProgresa (Oportunidades) set up in 1997 is aimed at alleviating poverty by considering multi-ple dimensions of poverty. Its main characteristic is that cash transfers are targeted directly tohouseholds on the condition that they send children to school and visit health centers on a regularbasis. For more details, see Skoufias et al. (2001).
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
Multidimensional Fuzzy Poverty and Pro-poor Growth Measures 35
this perspective, in such a labor-abundant economy, human capital development,whether in education or training, that targets the poor in order to enhance theircapabilities to achieve better living conditions and to fully participate in the pro-cess of development is the means to guarantee a labor-intensive growth strategyand sustainable poverty reduction. The second refers to regional policy. Despite thesignificant improvements evidenced in rural areas, the remaining gaps with urbanareas should be taken into account by policy makers. Reforms in rural areas needto be reinforced considering the regional dimension of poverty.
Acknowledgments
The author would like to thank the participants at the 16th ERF Annual Conferencethat took place in Cairo (Egypt) on November 07-09 2009 for their useful commentsand suggestions.
References
Baliamoune-Lutz, M. 2006. “On the measurement of human well-being: Fuzzy-set the-ory and Sen’s capability approach.” In Understanding Human Well-being, eds. M.McGillivray and M. Clarke, 119–138. Washington D.C.: Brooking Institution Press.
Berenger V. and A. Verdier-Chouchane. 2007. “Multidimensional analysis of well-being:Standard of living and quality of life across countries.” World Development, 35(7):1259–1276.
Berenger, V. 2008. “Evolution de la pauvrete multidimensionnelle en Egypte entre 1995 et2005: une approche basee du la theorie des ensembles flous.” Working Paper FEMISE31-06R, May.
Berenger V., J. Deutsch and J. Silber. 2009. “Multidimensional poverty measurement andthe order of acquisition of durable goods and access to services. The case of Egypt,Morocco and Turkey.” Presented at the MEEA Conference, Nice-Monaco in March2009.
Bibi, S. 2004. “Comparing multidimensional poverty between Egypt and Tunisia.” Cahiersde Recherche du CIRPEE, no. 0416.
Bibi, S. and A. El Lagha. 2008. “Comparaisons ordinales robustes de la pauvrete multidi-mensionnelle: Afrique du Sud et Egypte.” Revue d’economie du developpement, 22(1):5–36.
Cerioli, A. and S. Zani 1990. “A fuzzy approach to the measurement of poverty.” In Incomeand Wealth Distribution, Inequality and Poverty, eds. C. Dagum and M. Zenga.Heidelberg: Springer-Verlag.
Chiappero-Martinetti, E. 2000. “A multidimensional assessment of well-being based onSen’s functioning approach.” Societa Italiana di Economia Publica Working Paper,Rivista Internazionale di Scienze Sociali, 108(2): 207–231.
Chiappero-Martinetti, E. 2006. “Capability approach and fuzzy set theory: Description,aggregation and inference issues.” In Fuzzy Set Approach to Multidimensional PovertyMeasurement, eds. A. Lemmi and G. Betti, 93–113. New York: Springer.
Cheli, B. and A. Lemmi. 1995. “A totally fuzzy and relative approach to the multidi-mensional analysis of poverty.” Economic Notes by Monte dei Paschi di Siena, 24(1):115–134.
Datt, G., D. Jolliffe and M. Sharma. 1998. “A profile of poverty in Egypt: 1997.” FCNDDiscussion Paper 49, August.
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
36 V. Berenger
Datt, G. and D. Jolliffe. 1999. “Determinants of poverty in Egypt: 1997.” FCND DiscussionPaper 75, October.
Desai, M. and A. Shah. 1988. “An econometric approach to the measurement of poverty.”Oxford Economic Papers, 40: 505–522.
Deutsch, J. and J. Silber. 2008. “The order of acquisition of durable goods and the multidi-mensional measurement of poverty.” In Quantitative Approaches to MultidimensionalPoverty Measurement, eds. N. Kakwani and J. Silber, 226–243. Palgrave-Macmillan.
Dubois, D. and H. Prade. 1980. Fuzzy Sets and Systems: Theory and Applications. Boston:Academic Press.
El-Laithy, H., H. Abou-Ali, R. Yemtsov, S. Al-Shawarby and D. Marotta. 2009. “Wasgrowth in Egypt between 2005 and 2008 pro-poor?” From static to dynamic povertyprofile. Presented at the Workshop on “Multidimensional Poverty and Pro-PoorGrowth in the Mena Countries,” 11–12 June, Nice.
Grosse, M., K. Harttgen and S. Klasen. 2008. “Measuring pro-poor growth in non-incomedimensions.” World Development, 36(6): 1021–1047.
Kakwani, N. and E. Pernia. 2000. “What is pro-poor growth?” Asian Development Review,18(1): 1–16.
Kheir-El-Din, H. and H. El-Laithy. 2006. “An assessment of growth, distribution andpoverty in Egypt: 1990/91–2004/05.” Working paper 115, World Bank, December.
Kraay, A. 2006. “When is growth pro-poor? Evidence from a panel of countries.”Journalof Development Economics, 80: 198–227.
Lelli, S. 2001. “Factor analysis vs fuzzy sets theory: Assessing the influence of differenttechniques on Sen’s functioning approach.” Discussion Paper Series DPS 01.21, Cen-ter for Economic Studies, Catholic University of Leuven, Belgium.
McGillivray, M. 2005. “Measuring non-economic well-being achievement.” Review ofIncome and Wealth, 51(2): 337–364.
Osman, M., E. Zakareya and W. Mahrous. 2006. “Targeting the poor in Egypt: A ROCApproach.” Presented at the 13th ERF Annual Conference. “Oil: Its Impact on theGlobal Economy,” Kuwait, 16–18 December.
Ravallion M. and S. Chen. 2003. “Measuring pro-poor growth.” Economics Letters, 78:93–99.
Salem, S. and J. Gleason. 2005. “An examination of poverty reduction in Egypt: Con-tributing factors, sustainability and lessons.” USAID, Pro-Poor Growth, Tools andCase Studies for Development Specialists, January.
Sen, A. 1985. Commodities and Capabilities. Amsterdam: North Holland.Skoufias, E., B. Davis and S. de la Vega. 2001. “Targeting the poor in Mexico: An eval-
uation of the selection of households for PROGRESA.” International Food PolicyResearch Institute, FNCD Discussion Paper, no. 103, March, 1–47.
Townsend, P. 1979. Poverty in the United Kingdom. Hardsmonsworth: Penguin Books.UNDP. 1999. Egypt Human Development Report on Education 1998/99. New York.UNDP. 2008. Egypt Human Development Report. New York.World Bank. 2002. Arab Republic of Egypt, Poverty Reduction in Egypt. Vol. 1, no 24234-
EGT.World Bank. 2006. Arab Republic of Egypt, Challenges and Priorities for Rural Develop-
ment. Policy Note, Report no. 36432-EG, June.World Bank. 2007. Arab Republic of Egypt Improving Quality, Equality, and Efficiency
in the Education Sector: Fostering a Competent Generation of Youth. Report No.42863-EG, June.
Yamada, T. 2008. “Sustainable development and poverty reduction under Mubarak’s pro-gram.” IDE, Institute of Developing Economies, Discussion Paper no. 145.
Zadeh, L.A. 1965. “Fuzzy sets.” Information and Control, 8: 338–353.
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
Multidimensional Fuzzy Poverty and Pro-poor Growth Measures 37
Annex 1: List of Deprivation Indicators Considered for EachHousehold using Data from DHS 1995 and 2005.
Durable Goods
• Radio (bin.)• Washing machine (cat.)• Television (cat.)• Phone (bin.)• Refrigerator• Heater• Ventilator• Type of cooking fuel (electric oven, 2005)(cat.)• Sewing machine
Properties
• Bicycle (bin.)• Car• Dwelling(cat.)
Housing conditions
• overcrowding: number of room per individual• floor construction materials• Existence of kitchen as a separate room
Access to basic services
• Type of toilet facilities (cat.)• Improved water source• Electricity (bin.)
Table 1.A. Fuzzy poverty measures by regions and dimensions (%).
Total Durable Properties Housing Service
Years 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005
Urban 22.40 14.26 19.87 12.21 63.44 57.17 25.75 13.86 3.80 1.57Governorats
Lower Egypt 21.65 14.44 24.14 13.85 52.03 54.72 24.99 13.80 5.01 1.92Urban
Lower Egypt 28.44 16.82 39.69 22.29 30.23 32.14 41.67 21.38 15.16 6.51Rural
Upper Egypt 25.42 16.10 27.36 16.12 49.33 46.17 35.31 20.92 9.02 3.96Urban
Upper Egypt 37.14 24.05 44.94 31.68 27.38 26.40 63.14 40.88 25.15 10.42Rural
Frontier 26.44 18.60 25.90 19.63 42.33 32.36 25.07 20.75 20.21 12.47
Source: Author’s calculations from DHS 1995 and 2005.
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
June 15, 2010 9:11 WSPC/MEDJ S1793-8120 S1793812010000204
38 V. Berenger
Annex 2: The Determinants of Multidimensional Poverty
Table 2.A. Determinants of multidimensional poverty in 1995 and 2005.
1995 2005
Wald test Wald test
Poverty Coefficients Stat. Pr. Coefficients Stat. Pr.
Intercept 2.3952 138.88 < 0.0001 1.6755 80.1155 < 0.0001Sex (female =1) 0.1095 1.1781 0.2777 −0.0386 0.1511 0.6975Age −0.9059 80.3520 < 0.0001 −1.0102 121.92 < 0.0001Age*Age 0.0767 40.0044 < 0.0001 0.0951 73.5603 < 0.0001Children (less than 0.0615 11.0465 0.009 0.0523 7.2035 0.0073
5 years old)Education −0.9269 1166.1115 < 0.0001 −0.7365 1039.22 < 0.0001Area (rural= 1) 1.0921 619.99 < 0.0001 1.0697 634.723 < 0.0001Married −1.0782 89.7244 < 0.0001 −0.7867 52.2505 < 0.0001Divorced −0.2269 1.0239 0.3116 −0.1028 0.2403 0.6240Widow −0.7492 25.7472 < 0.0001 −0.4631 11.5148 0.0007Number of 15567 21972
observations
N.B. Statistics of global significance tests: LR1995 (9) = 3220.86 with Prob. < 0.0001 andLR2005 (9) = 2797.88 and Prob. < 0.0001.
Table 2.B. Determinants of multidimensional poverty with interaction terms.
1995 2005
Wald test Wald test
Poverty Coefficients Stat. Pr. Coefficients Stat. Pr.
Intercept 2.3724 133.85 < 0.0001 1.6779 77.8564 < 0.0001Sex (female= 1) 0.7364 3.8251 0.0505 0.0263 0.0092 0.9237Age −0.9247 82.5968 < 0.0001 −1.017 122.77 < 0.0001Age*Age 0.0787 41.6891 < 0.0001 0.0958 74.27 < 0.0001Children (less than 0.0608 10.7659 0.001 0.0521 7.128 < 0.0001
5 years old)Education −0.9270 1163.88 < 0.0001 −0.7371 1038.607 < 0.0001Area (rural =1) 1.0934 620.795 < 0.0001 1.07 634.85 < 0.0001Married −1.0128 71.1826 < 0.0001 −0.772 43.245 < 0.0001Divorced 0.2050 0.3228 0.5699 −0.4676 1.5527 0.213Widow −0.7677 17.200 < 0.0001 −0.4674 7.53 0.006Woman/married −0.7726 3.5751 0.0587 −0.2931 0.711 0.399Woman/divorced −1.1406 4.2606 0.039 0.4423 0.808 0.368Woman/widow −0.5303 1.7106 0.1909 −0.0447 0.0216 0.883Number of 15567 21972
observations
N.B. Statistics of global significance tests: LR1995 (9) = 3226.6239 with Prob. < 0.0001 etLR2005 (9) = 2800.74 and Prob. < 0.0001.
Dow
nloa
ded
by [
Yor
k U
nive
rsity
Lib
rari
es]
at 0
1:23
11
Nov
embe
r 20
14
