Am
TD
a
ARRA
JIIOC
KMGNESB
1
omtpetieemrboa
tp
1h
The Journal of Socio-Economics 41 (2012) 500– 512
Contents lists available at SciVerse ScienceDirect
The Journal of Socio-Economics
j o ur nal homep ag e: www.elsev ier .com/ locate /soceco
ssessment of multidimensional poverty and effectiveness oficrofinance-driven government and NGO projects in the rural Bangladesh
amgid Ahmed Chowdhury ∗, Pundarik Mukhopadhaya1
epartment of Economics, Macquarie University, North Ryde, NSW 2109, Australia
r t i c l e i n f o
rticle history:eceived 20 January 2012eceived in revised form 14 April 2012ccepted 29 April 2012
EL classification:380390290190
a b s t r a c t
This paper has developed a multidimensional model usable in assessing economic, social, political andcultural dimensions of poverty by utilizing primary data collected from 78 villages in Bangladesh usinga participatory approach. Employing the developed model, a comparative analysis has been performedbetween microfinance-driven government (GO) and NGO (non-government organization) projects toexplore their relative effectiveness in enhancing wellbeing of the poor in rural Bangladesh. It is observedthat GO agencies are more effective in enhancing ‘economic wellbeing’ of the poor, whereas NGOs arecontributing more in the ‘social’ aspects of wellbeing. Findings also revealed that, as whole, GO agenciesperform 42% better than NGOs in improving living standards of the rural poor which contradicts withthe existing literature of poverty reduction projects in developing countries.
© 2012 Elsevier Inc. All rights reserved.
eywords:ultidimensional povertyovernmenton government organizationconomic wellbeingocial wellbeing angladesh. Introduction and rationale of the study
The main aim of this paper is to assess the relative effectivenessf microfinance-driven poverty reduction projects run by govern-ent and NGOs in enhancing wellbeing (or reducing poverty) of
he rural poor. However, as there is no composite model incor-orating multi-aspects of poverty in Bangladesh available in thexisting literature, we first developed a multidimensional modelhat incorporates all such indicators as opined by the local poor. Its believed in the paper that poverty is lack of wellbeing and thus byxploring and fulfilling the indicators of poverty, we can necessarilynhance wellbeing of the poor. This is why the model is a povertyodel whose positive outcomes are multi-aspects of wellbeing (or
educed poverty) and the degree of achievement of those indicators
y the respective credit recipients would assess the effectivenessf government and NGOs. This paper describes the model buildingnd comparative analysis methods stated above.∗ Corresponding author. Tel.: +61 433180544.E-mail addresses: [email protected],
[email protected] (T.A. Chowdhury),[email protected] (P. Mukhopadhaya).1 Tel.: +61 298506476.
053-5357/$ – see front matter © 2012 Elsevier Inc. All rights reserved.ttp://dx.doi.org/10.1016/j.socec.2012.04.016
Microfinance has been used as an effective tool for povertyalleviation around the world for decades. This approach not onlycreated poor’s access toward capital, but also allowed them toimprove their business which in turn increased personal incomeand increased personal spending on children’s education, fam-ily illness and improved housing and nutrition (Morduch, 2000;Coleman, 2005). However, several studies also found unconvincedresults about the economic benefit of microfinance (Coleman,2005; Hoque, 2004). Study by Hossain (1988) on Bangladesh estab-lished that microfinance has positive impacts on social indicatorssuch as opportunity for empowerment and decision making rightswhich increase confidence and self-esteem. Pitt and Khandker(1998) indicate that microfinance programs promote the poorhousehold’s investment in human capital through choice of pub-lic versus private schooling and the contraceptive behavior of thefamilies. But again, several other studies (Kabeer and Neoponen,2005) found that the impact of microfinance on social indicatorsis inconclusive. Even though it can be said that the microfinanceorganizations such as government institutions and NGOs startedwith economic solvency to the poor, but this operation has sev-
eral other multiplier effects in the lives of poor which made theseinstitutions more appealing in the development context. However,being positive contributors in the lives of poor, credit providersare not equally effective or rated alike by the donors or governing
Journa
bimpamptdOaecTp
til
T.A. Chowdhury, P. Mukhopadhaya / The
odies. For example, one noticeable change in the credit operations the massive shift of the donor’s funding channel from govern-
ent to NGOs. One prime reason for such shift is reported to beurely ideological, legal and institutional method of fund deliverys it was argued that expansion of NGO involvement in develop-ent is the anti-state intervention nature of structural adjustment
olicy which prefers minimum government and recommendedhat development programs should not be totally controlled byonor and recipient governmental agencies (Schneider, 1988).ther reason behind such shift is said to be the comparativedvantages of the NGOs which made them more effective than gov-rnment in serving poor in the developing countries. However, thatlaim is inconclusive,2 debatable and not universally acceptable.hus such shift in preferred channel for credit delivery raises onerimary question at the foundation of this study:
(i) What are the appropriate measures of effectiveness in themicrofinance-driven projects and are they good enough forsuch assessment?
There have always been debates about the effectiveness ofgovernment and NGOs in the social works and they were com-pared on several grounds. Meyer (1992) and Vivian (1994) intheir studies on Zimbabwe have shown that NGOs are more costeffective than government in reaching to the poor. However,Hashemi (1992) on his study on several Asian countries andCarroll (1992) in his study on Latin American countries foundthat NGOs were more expensive in reaching to the poor andmaintaining popular participation. Riddell and Robinson (1992)on their work on Africa found similar result. Asian DevelopmentBank (2003) reported that NGOs could reach to only 20% of themarginal poor in Bangladesh with such a long presence in thecountry. Korten (1990) found that NGOs are better than gov-ernment in project planning and management in social works.Meyer (1992) also reported that NGOs were more effectivein delivering education and health services compared to gov-ernment in Latin America due to their better implementationcapability. However, Edwards and Hulme (1996) argued thatthis effectiveness was by default and not by design as gov-ernment lacked resources for project implementation on thosecountries. Hasan (1993) in his work on Pakistan found thatNGOs were more effective than public sector in the sanitationprograms. However, Riddell and Robinson (1992) in their studyon 16 Asian and African projects (related to sanitation, immu-nization, etc.) found the opposite results with government’ssuperiority. Howes and Sattar (1992) in their work on BRACprimary schooling concluded that there is no trade off betweenNGO’s cost effective service provision and quality of work. How-ever, LaFond (1995) in his study has shown that long termNGO services suffer from quality to all. Several studies reportedthat NGOs performed much better than public sector in inno-vating and delivering services like oral dehydration (Howesand Sattar, 1992) and agricultural technology development(Farrington et al., 1993a). On the other hand, Farrington et al.(1993b) found that in Latin America NGOs became less innova-tive when they worked as contractors to donors or government.Mahmud and Ahmed (2003) in their studies on delivering nutri-tion services in Bangladesh found that NGOs were cheaper in
delivering the required services compared to government. Deng(2008) reported that NGOs are more likely to deliver more dif-ferentiated and customized services than bulky government2 A study by the World Bank (cited in Narayan et al., 2000) that used Participa-ory Poverty Assessment (PPA) tools concludes that, (i) the state has been largelyneffective in reaching the poor, and (ii) the role of NGOs in the lives of the poor isimited, leaving them dependent primarily on their own informal networks.
l of Socio-Economics 41 (2012) 500– 512 501
units while working for rehabilitation. Edwards and Hulme(1996) also noticed that NGOs are more effective in creatingand delivering social funds intended to mitigate the social con-sequences of economic and structural adjustment packages.However, Fernando (2006) in his study found that the effec-tive interest rate charged by NGOs is more than 28%, which ismuch higher than the interest rate charged by the governmentcredit institutions or even the commercial banks of Bangladesh.More importantly, Kelly (2008) in his study found that NGOs inBangladesh are working as parallel government and account-able to none. In addition, Epstein and Crane (2005) in their workmentioned about several other findings on issues like number ofborrowers, borrower retention rate, financial sustainability ofthe project and repayment rates. However, in most cases therewas no complete domination of effectiveness by either GO orNGOs.
Among all existing criteria, access to and repayment rateof credit and the total amount disbursed toward women areconsidered to be most appealing to donors and policy makers(Kevane and Wydick, 2001; Mayoux, 1999) which we believeare extremely narrow in nature. For example, the repaymentrate cannot be used as the sole indicator of effectiveness in loanutilization, since the source of income may not necessarily befrom the revenue generated by productive investments. Thereis evidence to suggest that poor people, especially women, bor-row money from one microfinance institution (MFI) to pay thedebt burden of another3 (Goldin Institute, 2007; Burra, 2005).In addition, it was reported (Goetz and Gupta, 1996) that eventhough the credit is sanctioned in the name of women, it is inmany extents used by their husbands or male members of thefamily which not only violates the precondition of empower-ment but also questions the validity of effectiveness assessmentbased on the number of women reached.
As can be seen, most of the stated (and available) evalua-tion criteria are designed from the organizational perspectiveand do not measure the aggregate impacts of the credit drivenprojects in the wellbeing of the poor – an unaddressed issue inthe literature. Thus this paper argues that the existing methods(as described above) of effectiveness evaluation are narrow innature as they always ignored comparing the degree of contri-bution by the development partners in enhancing wellbeing ofthe beneficiaries – which is supposed to be the main aim of thecredit-driven projects. While the answer of the first question isdetailed earlier, that leads to the second question of this paper:
(ii) Who is more effective between GO and NGOs in credit-drivenpoverty reduction programs in creating wellbeing of the poorbeneficiaries?
We believe that the results of this type of comparative studywould help the policy makers of donors to make more informeddecision about their preferred channel for fund delivery. Butto do so, the first step would be to develop a multidimen-sional poverty model which can capture the wellbeing issues as(Ravallion, 1998) argued that “a credible measure of poverty (orpoverty model) can be a powerful instrument for focusing theattention of policy makers on the living conditions of the poor”.However, such attempt has never been made in Bangladesh(case of this study) as well as in many poverty-ridden coun-tries of the world. This is why a comparative effectivenessmeasurement between GO and NGOs on enhancing wellbeing
has never been conducted. Thus this study will address tworesearch questions.3 According to Goldin Institute (2007), it is not uncommon for families to carry asmany as five loans, most used to cover old debts, rather than purchase new assets.
5 Journa
1
pni2ibmaia‘oiti
fNtafv(BpBpii
anadcofittImafau
M
tc2
“
i
t
and wellbeing should be judged in terms of the quality and quan-tity of available opportunities, which he termed as ‘genuine choice’(Sen, 1993) and that modification was considered by many asoption of diversity.14 Sen (1993, 1984) always emphasized that the
02 T.A. Chowdhury, P. Mukhopadhaya / The
(a) Which factors contribute to the wellbeing of the rural poorin Bangladesh?
(b) To what extent microcredit providers (government andNGOs) could contribute on those wellbeing issues?
.1. Limitations of poverty assessment in Bangladesh
Even though studies on human wellbeing have revealed thatoverty is multidimensional,4 poverty in Bangladesh is still viewedarrowly in terms of direct caloric intake5 (DCI) and food energy
ntake6 (FEI) in official assessment (Government of Bangladesh,010). In addition to these, a cost of basic needs (CBN) method was
ntroduced in the mid 1990s. The CBN method sets the poverty liney computing the cost of a food basket that enables a household toeet predetermined nutritional requirements, and adds to this an
llowance for basic non-food consumption. Other than measuringncome poverty,7 there are two non-income indicators of povertyre used in Bangladesh, namely an ‘infant mortality rate’ and the
school enrolment ratio’. The infant mortality rate reflects the statef the primary healthcare system of the country, and the pace of itsmprovement over time, while the school enrolment ratio indicateso what extent the country is able to deliver universal education tots people.
Like the government agencies, NGOs have also been workingor poverty reduction in Bangladesh since 1971. Even though theGOs claim (Mahmud, 2008) that they work for social mobiliza-
ion, women’s empowerment and income generation, their mainctivities as with government agencies are limited to basic needulfillment8 through delivering micro-credit and other social ser-ices to the poor. NGOs do not use any other official poverty modelother than CBN) that can address other dimensions of poverty inangladesh. Head count ratio (HCR) based on DCI or CBN methodsrovides change in monetary poverty rate for the whole or regionalangladesh, but these are unable to capture the changes in social,olitical and cultural dimensions of poverty for a specific year. This
s one important limitation of the poverty assessment method usedn Bangladesh.
Furthermore, whilst the HCR based on DCI/CBN method offersn overall measure of the poverty situation in Bangladesh, it doesot split the individual contribution of micro-finance driven GOsnd NGOs or other development partners in improving living stan-ard of the poor. NGOs claim (Ravallion et al., 1999) that theyontribute more in eradicating poverty in Bangladesh becausef their higher disbursement of micro-credit, larger number ofeld workers and greater coverage of geographic areas comparedo government agencies. However, no statistics are available onheir (GO and NGOs) relative performance9 in poverty reduction.n addition, due to the absence of any multidimensional poverty
odel it is not possible to compare the effectiveness of GOsnd NGOs in the microfinance driven poverty reduction projects
rom a wider perspective (including economic, political, socialnd cultural sides) and thus the questions raised above remainednanswered.4 Refer to Table A1 in Appendix A for details. Also see Booysen (2002) andcGillivray and Noorbakhsh (2007).5 In DCI poor households are defined as those with per capita energy intake less
han the standard per capita requirement of energy (1805 kilo calorie and 2112 kiloalorie for extreme and moderate poverty line) (Bangladesh Bureau of Statistics,007).6 FEI method sets the poverty line as the income or consumption level at which
basic needs” are met (Ahmed, 2004).7 In income poverty analysis, statistics on land ownership, consumption and sav-
ngs pattern are available.8 That includes income generation, healthcare and education support.9 How many poor beneficiaries of government and NGOs separately could break
he poverty.
l of Socio-Economics 41 (2012) 500– 512
This paper argues that development of a multidimensionalpoverty model (first objective) through the opinion of the poor canhelp the policymakers to pin-point the specific livelihood needs ofthe people10 thus better targeting poverty and enhancing wellbe-ing.
The approach: While developing a model that represents thewellbeing preferences of a group of people, there may be con-flict in preferences among people. Our method incorporates amultidimensional modeling approach through confirmatory fac-tor analysis where overall goal is to identify those dimensions thataffect perception or behavior, which may not have been readilyevident in the data and cannot be explored through traditionaldescriptive methods.
2. Literature on definition and dimensions of the conceptsused in the paper
2.1. Definition and dimensions of poverty and wellbeing
The traditional definition of poverty with respect to small earn-ing is now viewed as a narrow definition of poverty (Sen, 1982,1993) as that bypasses other social, psychological, cultural, polit-ical and participatory indicators.11 By considering the evolvingconcepts of poverty (as shown in Table A1 of Appendix A), wedefine12 poverty simply yet broadly (multi-dimensionally) as “theinability to participate in society, economically, socially, culturally andpolitically”.
For further understanding of the relation between povertyand wellbeing and their underlying indicators, definition offeredby World Bank (2000) can be helpful which states, “poverty ispronounced deprivation in well-being”, where well-being can bemeasured by an individual’s possession of income, health, nutri-tion, education, assets, housing, and certain rights in a society suchas freedom of speech. Furthermore, Sen (1993) developed13 con-cepts of capability, freedom and functioning as an alternative to thetraditional economic framework of conceptualizing poverty, well-being, vulnerability and human development. Sen (1982) arguesthat utility based evaluation of individual wellbeing might notreveal important dimensions of life and leads to misleading inter-personal comparisons. At the primary stage the major indicatorsused by Sen’s analysis are education (literacy), escape from morbid-ity, longer life expectancy, working properly, health status, politicalactivity, enjoying positive states, etc. (Sen, 1984; Clark, 2005).However, Sen did never subscribe a fixed set of indicators andrevised and broadened the concepts by arguing that the poverty
10 Available studies only explored few indicators and did not offer any validatedmodel for the effectiveness comparison. For further details, see Nabi et al. (1999)and Mahbub and Roy (1999).
11 However, change is visible in the poverty analysis and at least two major changescan be observed from the literature (Shaffer, 2008): (a) the concept of poverty hasbeen broadened from physiological to social model with an attention to vulner-ability, deprivation, inequality and human rights and (b) the causal variables ofpoverty have been enlarged by including social, political, cultural and environmentalconcepts.
12 Similar approach was used by Silver (2007).13 Beginning with the Tanner Lecture ‘Equality of What’, first delivered in Stanford
University in 1979.14 Noticeable works on capability approach and suggested list of capabilities that
can be tested for rural Bangladesh are available in Nussbaum (2005) [some sug-gested items are life, bodily health, sense and imagination, emotions, affiliation,political and material control, etc.] (Alkire and Black, 1997) [suggested items arelife, knowledge, friendship, self determination, etc.] and Clark (2005) [jobs, housing,
Journal of Socio-Economics 41 (2012) 500– 512 503
ibt
cp(
2d
wfnmoiiiopoetissmttsp
ttirsUwtAi
3
tie
trsc
fg
(
dofi
Table 1Dimensions and items for the multidimensional poverty model.
Proposed factors Outcome and process based indicators
Economic wellbeing Items related to food intake by familymembers, income, savings, access toelectricity and sanitary latrine and safewater, home and land ownership, landholding size, other household assets,average sick days of the familymembers, morbidity status, capacity towork in daily life, shortage time foodintake, degree of vulnerability withrespect to land and asset ownerships,etc. (outcome-based)
Social wellbeing (includes social,cultural and political aspects)
Influential indicators are access toinformation about natural disaster,loan, education, health and job,information about politics and localand central government, healthcare,education, schooling, freedom of doingsocial, cultural, religious and politicalworks, participation in society andpolitics and voting behavior, decisionat household and work place,experience of robbery and theft,mental stress and feeling of insecurity,etc. (outcome-based)
Effectiveness Indicators like, loan repaymentcapability of the beneficiaries of thatprovider, amount of loan provided,length of borrowing from a particularprovider with repeat borrowing, etc.(institutional or process based)
T.A. Chowdhury, P. Mukhopadhaya / The
ndicators of capability in poverty and wellbeing assessment shoulde area specific which will reflect the social values and culture ofhe local community.
By reviewing other available literature,15 lists of poverty indi-ators that are relevant to the lives of rural poor in Bangladesh arerepared and a summary of those indicators is provided in Table 1see 2nd and 3rd row).
.2. Conceptualizing wellbeing based effectiveness in creditriven projects
In general, effectiveness is determined by an input–output ratiohich is expressed through profit and rate of return, in the case of
or-profit organizations. In the social sector, particularly in microfi-ance driven poverty reduction projects, the effectiveness has beeneasured through the number of beneficiaries reached, amount
f credit delivered, financial sustainability of the project and mostmportantly, by repayment rates of micro-credit16 which are allnstitutional indicators and do not assess the improvement in liv-ng conditions of the poor. We argue that, in analysing effectivenessf credit providers (GOs and NGOs) in the poverty alleviationrograms, it is important to examine the extent to which the devel-pment partners could support the poor for the achievement ofconomic (people always seek to increase the return to the activi-ies they undertake by using the microfinance as increased incomes the security of economic wellbeing), material (such as food andhelter security) and non-material wellbeing (such as, self esteem,ense of control and inclusion, physical security of the householdembers, health status, political enfranchisement, cultural works)
hrough reducing vulnerability (savings to cope with that, shockime support) and mobilize them in social activities beyond mea-uring the quantity of profit made by disbursing micro-credit to theoor.
Even though the effectiveness of the agency will be measuredhrough their contribution in social and economic wellbeing ofhe poor (outcome-based study), another part of the model willncorporate process-based and institutional indicators such as loanepayment rate, frequency of defaulting, repeat borrowers rate,ize of the loan, length of borrowing, etc. (refer to row 4, Table 1).nique feature of our model is thus the incorporation of bothellbeing indicators (poverty model) and other existing institu-
ional indicators of effectiveness assessment in a single model. list of both outcome and process-based indicators is presented
n Table 1.
. Questionnaire and data collection
To build the model, primary data was collected through a ques-ionnaire survey approach. The questionnaire was of mixed mode,ncluding questions having multiple options, organized with a Lik-rt type scale and few were dichotomous in nature.17
A formal questionnaire consists of 8 sections (including first sec-ion that contains general and demographic information of the
espondents) and 62 questions was developed. Second and thirdection of the questionnaire includes questions regarding healthonditions and educational status of the respondents respectively.riendship, self determination, education, income, family and friends, religion, food,ood cloths, recreation, safety, etc.].15 See also, Sen (1982, 1993, 1984), Narayan et al. (2000), and Ruggeri Laderchi2001).16 See for instance, NGOAB and PKSF websites and Goldin Institute (2007).17 There was no open-ended qualitative question in the questionnaire. However,uring face to face surveys, additional relevant qualitative comments and opinionsf the beneficiaries were noted which were also utilized to validate the quantitativendings of the paper.
Fourth section incorporates information about access to utilities anddifferent types of information whereas section 5 details individ-ual asset and income related questions. Section 6 highlights theissues relevant to empowerment and decision making and section7 includes items relevant to security in every day. Section 8 reflectsthe opinion about support services and includes the informationabout the indicators listed in row 4 of Table 1.
The multi-stage sample selection procedure and model buildingprocess is discussed in Fig. 1 that shows the name of the dis-tricts covered. After selecting the districts, we then chose upazillafrom each district based on two major criteria namely less liter-acy rate and severity of natural calamity as only those informationwere available. Villages were selected randomly from the chosenupazillas with a single criterion that they are more distant fromthe district headquarters such that the grass-root level and mostdisadvantaged people can be surveyed.
Data were collected randomly from local Bazars (small markets),individual homes and while walking in the muddy streets of thevillages. To ensure symmetry, respondents were chosen randomlyfrom the microfinance-driven poverty alleviation projects and byconsidering that each respondent has only one credit scheme (valueof the loan is between USD100–150) and are not taking any otherpaid services from the development partners. In addition, house-holds with less than half an acre of land were eligible for the study(similar approach was used by Pitt and Khandker, 2001). A total offive hundred and sixty-two usable questionnaires were found fromseventy-eight villages and 23 upazillas of eight districts namely Lal-monirhat, Kurigram, Nilphamari, Gaibandha, Potuakhali, Barguna,Jhalokathi and Jessore. Among these usable questionnaires, thereare 292 (52%) male respondents and 270 (48%) female respondents.While 37% of the 562 people were GO beneficiaries, the rest was
NGO beneficiaries.
504 T.A. Chowdhury, P. Mukhopadhaya / The Journal of Socio-Economics 41 (2012) 500– 512
Note : Upazi llas are th e lo west lev el of administrati ve gov ernm ent in Banglade sh, HCR stands for Head
Count Rati o. *E xplor atory factor an alysis (E FA) is a stati stica l metho d that helps deter mine und erlyi ng
dimens ion for a set of measured ite ms. **CFA allo ws th e resea rche r to te st the hyp oth esis th at a
relati onship between the observed it ems and th eir under lying dim ension(s) exi sts . CFA tests whether a
specified set of dim ensi ons is influ encing resp onses in a predi cted way .
Poverty prone and
natural disaste r affec ted
divisions ch osen. 3
division s are Baris al,
Rajshahi and Khul na
Base on HCR> 0.60, 12 di strict s were selecte d from 3
divisions. They are: 3 district s from Barisal division
(Barguna , Jhalokat hi & Potuak hali), 5 from Rajsha hi
(Gaiba ndha , Kurigra m, Lalmonirha t, Nil pha mari &
Rangpur) an d 4 fr om Khulna (B agerhat, Jes sore,
Khulna and Satkhira)
9 dis tric ts fina lly se lected shar ing
common ec onom ic, social and
natural featur es. These are: Barg una,
Jhal okathi, Potuakhal i, Gaiban dha,
Kurigram, Lalmonirha t, Nilphamar i,
Jessore, & Satk hira
At lea st 2 Upazilla s from each district
based on li terac y and severi ty of natura l
disaster are selecte d. A total of 23
Upazil las sur veyed from 8 dis tri cts
Random se lection of at least 3
villages from each Upaz illa. A total
of 78 villages were chosen
562 qu estionn aire s were filled in.
Quest ionna ire co ntai ns 56 wel lbe ing plus
6 serv ice rela ted variables
Explorat ory Factor Analys is* is
perfor med and fi nall y 11 fact ors and
43 ite ms were selecte d
Confirmatory Factor Analysis **
performed and finally 11 facto rs
and 36 item s reta ined
Mea surement mode l preparation and
vali dation tested fo r the model. Finally,
9 fac tor and 29 ite ms reta ined.
Discriminant and nomological validity
of the model te ste d
Fin alised Stru ctural m odel with 9
factor s an d 29 items
Invaria nce of the model te sted
between GOs an d NGOs
beneficiar ies
Fig. 1. Sample selection, data collection and model building Procedures. Note: Upazillas are the lowest level of administrative government in Bangladesh, HCR stands for headc rminet lying
r
4
4
wwimiipafv
4
r
(hab
(m
ount ratio. *Exploratory factor analysis (EFA) is a statistical method that helps deteest the hypothesis that a relationship between the observed items and their underesponses in a predicted way.
. Model development methodology
.1. Model purification through exploratory factor analysis18
At the primary stage of developing the model, we have 56ellbeing and 6 support service related items with two broaderellbeing dimensions (economic and social wellbeing) and one
nstitutional effectiveness factor. As the Kaiser–Meyer–Olkin (KMO)easure of sample adequacy (MSA) value is high as 0.763, the data
s good enough for the study. The result of EFA shows that these 62tems belong to 16 individual factors. We then continued the EFAurification until we get all the items with loading greater than 0.50s well as no cross loaded items in the study.19 Finally, a total of 11actors represented by 43 items were selected with 68.34% of totalariance explained.
.2. Individual confirmatory factor analysis (CFA)
These 43 items and 11 factors were then put to CFA with aestricted rule of deleting items with a loading less than 0.40 and
18 EFA detects: (a) items that influence social and economic dimensions directly;b) items that have an indirect relationship to social and economic wellbeing, butave direct relationships to other intermediary dimensions that directly affect socialnd economic dimensions; and (c) items that are less relevant to the study of well-eing in Bangladesh according to the opinions of the beneficiaries.19 Similar rules were followed in Marketing literature by Shimp and Sharma1987); in Psychology literature by MacCallum and Austin (2000); in Research
ethodology by Hair et al. (2009).
underlying dimension for a set of measured items. **CFA allows the researcher todimension(s) exists. CFA tests whether a specified set of dimensions is influencing
accepting the individual CFA models with good fit statistics.20 Inaddition, judgemental views were taken into consideration becauseit is important to determine the significance of droppable variablesbased on existing literature and qualitative observations from fieldstudy. Following the stated criteria, 7 other items21 were droppedfrom further study through whole CFA.
Naming the factors: In order to establish inter-factor relationshipin the measurement model, we name the factors according to theircorresponding items as listed in Table 2.
4.3. The measurement model: construction and purification
The measurement model shows the extent to which all factorsand measured items as a whole are operational and compatible as amodel and assesses the construct validity.22 After running the firstmeasurement model with 11 factors and their corresponding 36items, the result was found to be not-admissible due to negativecovariance of the ‘Empowerment factor’ with other factors of the
model.23 In addition, loading values of the items of the ‘Empow-erment’ factor are too low when grouped in the model with otherfactors and items. One probable reason of non-significance of the20 For instance, small and significant Chi-square values, p values greater than 0.05,GFI and CFI greater than 0.90, RMSEA less than 0.05 and Hoelter value more than 200(Byrne, 2001).
21 All these items have loading less than 0.30.22 Construct validity refers to the extent to which operationalizations of dimen-
sions do actually measure what the theory says they do.23 This problem factor (Empowerment) has been identified by a trial and error
process by checking the GOF values of the model by deleting one factor and itsitems at a time.
T.A. Chowdhury, P. Mukhopadhaya / The Journal of Socio-Economics 41 (2012) 500– 512 505
Table 2Factors and their outcome items after CFA.
Items (outcomes) retained through CFA Proposed name of the factor
Access to natural disaster, loan, education, health and job related information Access to general informationSharing political and government information Access to governance informationFreedom of performing cultural, religious and political works FreedomHome ownership, land holding size and status Asset buildingAverage sick days of male and female, morbidity and capacity to work normally Human capability buildingFood intake per day by male, female and kids Core need fulfillmentDecision at household, experience of theft and robbery and shortage time food consumption VulnerabilityMonthly income and savings, access to electricity and sanitary latrine Economic wellbeingDecision at job, mental stress and feeling of insecurity Social wellbeingLoan repayment status, amount of loan taken and length of borrowing by the beneficiaries EffectivenessVoting by male and female beneficiaries, choice of preferred candidates Empowerment
ith G
‘Befi(
tTet
Fig. 2. Measurement model for effectiveness analysis. Chi-square is 1564.578 w
Empowerment’ factor is less relevant of its items in the model.24
ased on statistical results and evidence of less relevance, ‘Empow-rment’ factor and its corresponding 3 items were dropped from
urther study. The purified measurement model which is shownn Fig. 2 with 10 factors and 33 items was found to be operationalrefer to the GOF indices with Fig. 2). However, it was observed24 For example, our survey result explored that about 96.3% of the respondents castheir vote regularly and 98% of the regular voters vote for their preferred candidate.his shows that from the voting point of view, beneficiaries are quite empow-red thus these items and ‘Empowerment’ factor were identified as less relevanto poverty model of Bangladesh.
FI, CFI, PClose and RMSEA values are 0.851, 0.881, 0.000 and 0.066 respectively.
that some further adjustment of the model is needed by consider-ing large modification and error index values (especially the onesgreater than 15).
A large correlation was found between loan repayment rate(e27) and the ‘Human capability building’ factor. This is because,with a better capability of the individual, earnings and conse-
quently timely loan repayment capacities improve. In addition tothis, few larger correlation and modification index values wereobserved too.2525 Explanations on all correlations are not included here. We will be happy toprovide the detailed results to any interested reader.
5 Journa
woastm
4
o
(
(
(
iulatvetnu‘orj(asomtdwtr
4
t(vlm70b
t
r
v
(
(
NGO beneficiaries. Simultaneous multiple group method (dividingthe whole data into two groups of beneficiaries of GOs and NGOs)
06 T.A. Chowdhury, P. Mukhopadhaya / The
By considering required modifications, our measurement modelas purified further. Purified measurement model has a chi-square
f 1005.461 with a PClose value 0.090. In addition, GFI, CFI, RMSEAnd Hoelter values of 0.901, 0.927, 0.049 and 243 & 254 respectivelyignal that the purified measurement model has a satisfactory fito keep all the items and proceed to the next level of structural
odeling.
.3.1. Validity of the measurement modelFor the construct validity – which judges the overall validity –
f the model, we checked that:
a) the items in a single dimension are all correlated to each otherand their corresponding dimension (called convergent validity),
b) two dimensions are in fact different from each other and thateach one of them contains certain phenomena that are notfound in the other (called discriminant validity) and
c) the dimensions and their items behave as they should within asystem of related coordinates with another sample character-istic (such as age of the respondents of GO and NGOs) (callednomological validity).
Results for the convergent validity analysis26 show that alltems except ‘decision making at household level’ has loading val-es greater than 0.40 (acceptable level) which satisfies the factor
oading criteria. Except ‘Vulnerability’ all other factors have aver-ge variance extracted values more than 40% (range is from 43%o 85%) that necessarily guarantee the evidence of convergentalidity. Finally, construct validity values range from 0.49 to 0.94xcept ‘Vulnerability’ factor which suggest adequate reliability ofhe measurement model. However, we did not check discrimi-ant and nomological validities for the measurement model due tonsatisfactory ‘construct reliability’ of the ‘Vulnerability’ factor. The
Vulnerability’ factor is relatively less significant, perhaps becausef its less-relevance to other factors and items. For example, (a) ouresult shows that more than 91% of ‘household decisions’ are madeointly thus this item has less importance in the whole model andb) because of low income level of the respondents, nothing valu-ble is available to be ‘theft from their home’ and thus this itemeems less important for the model too. Two out of three itemsf ‘vulnerability’ factor were found to be less important, whichade this factor less significant for the whole model. Therefore at
his stage, ‘Vulnerability’ factor and its corresponding items wereropped due to low extracted value of average variance (36.98%)ith less construct reliability (0.27). Deletion of ‘Vulnerability’ fac-
or further proves that vulnerability is not a dimension of povertyather it is a symptom of poverty.27
.3.2. Validity testing for the finalized measurement modelIn the revised measurement model, item named ‘Access to elec-
ricity’ was decided to be dropped due to a critical ratio of 1.47which is less than acceptable value of 1.96) and low factor loadingalue. Probably this particular utility service is still considered as auxury good by the poor beneficiaries in Bangladesh. With the above
odifications, our finalized measurement model has chi-square of33.018 with GFI, CFI, PClose, RMSEA and Hoelter values of 0.918,.953, 0.837, 0.047 and 283 & 297 respectively which indicate a
etter fit.Notable results of the purification are: (a) all items have satisfac-ory factor loadings with average variance extracted values greater
26 It includes factor loading testing, average variance extracted test and constructeliability measures. Detailed results can be provided upon request.27 Similar findings can be seen in DFID’s Sustainable Livelihoods Model, whereulnerability is not considered itself as a dimension of poverty.
l of Socio-Economics 41 (2012) 500– 512
than 45%, (b) construct validity values ranging from 0.736 to 0.981(which is another indication of construct validity) and (c) constructreliability of the factor titled ‘Economic wellbeing’ has increasedto 0.831 from 0.567 due to the elimination of the item ‘Access toelectricity’. For further validation of the model, discriminant andnomological validities were performed.28
4.4. Constructing the structural poverty model for effectivenessassessment
The preliminary structural model is constructed in a way that,‘Effectiveness’ of the development partners is measured by theircontribution in improving ‘Economic’ and ‘Social’ wellbeing of thebeneficiaries. In the one hand, ‘Economic wellbeing’ is influenced by‘Core need fulfilment’, ‘Human capability building’ and ‘Asset build-ing capability’ of the beneficiaries and on the other hand ‘Socialwellbeing’ is the result of ‘Access to general information’, ‘Accessto governance information’ and ‘Freedom’. Similar explanations canbe given to demonstrate the relation among measured items andtheir corresponding factors.29
By running the preliminary structural model, we have exploredseveral new correlations between a few factors and measured itemswhich require specific interpretations.
a) Relations between length of borrowing (item-28) and loan repay-ment rate (item-27) with ‘Human capability building’ factor. Therate of repayment depends heavily on an individual’s physi-cal and intellectual capabilities. For example, less sick days andbetter capacity to work will ensure more work days, thus moreearning and consequently better repayment rate.
b) We found a relation between the factor ‘Freedom’ and ‘incomeper month’ (item-21) which means our respondents believe thatfreedom of doing things depends on the level of income. Thatmeans better earned people are freer than an insolvent person,or better earning people are less socially excluded.
(c) We identified correlation between ‘Social wellbeing’ factor and‘freedom to do political works’ (item-8). This relation justifies thatfreer engagement in political activities is an indicator of socialwellbeing.
The above stated and other necessary modifications (additionalcorrelations are shown in Fig. 3) were incorporated in the modeland the goodness of fit values were compared between pre and postmodifications. Finalized structural poverty model demonstratessatisfactory fit values (shown in Fig. 3).
5. Testing the model for invariance across GOs and NGOsfor comparative study
We have accomplished our first objective of developing and vali-dating the multidimensional poverty model in the last section. Nowin order to fulfill the second objective (comparing effectiveness ofGOs and NGOs) it is important to check whether this model and itsindividual items and factors are equally applicable for both GO and
was performed for this purpose and a summary is presented inTable 3 (shown as configural invariance). Remaining fit statistics
28 Interested readers are requested to contact the authors for further details.29 For example, outcome of economic wellbeing are income per month, savings
per month and use of sanitary latrine; outcome of asset building is home ownershippattern, land holding size and land holding status (whether bought new or sold orowned by inheritance).
T.A. Chowdhury, P. Mukhopadhaya / The Journal of Socio-Economics 41 (2012) 500– 512 507
Fig. 3. Final multidimensional poverty model for effectiveness assessment of the development partner. CFI, GFI, RMR, RMSEA and p value of 0.959, 0.922, 0.067, 0.044 and0.983 respectively. Coefficient-H reliability value is 0.909. In the case of the structural model, Cronbach’s Alpha is often underestimated or under-reported (Arbuckle, 2009)thus we used the Coefficient H value.
Table 3Structural invariance tests for GOs versus NGO beneficiaries.
Model tested Model fit measures Model differences
Chi-square df p CFI RMSEA ��2 �df �p
Configural invariance 1150.092 658 0.00 0.944 0.037Metric invariancea 1241.517 686 0.00 0.937 0.037 92.42 28 0.147Scalar invarianceb 1316.385 687 0.00 0.929 0.040 166.29 29 0.136Factor cov. invariance 1232.868 688 0.00 0.938 0.038 82.71 30 0.013Factor var. invariancec 1173.684 666 0.00 0.943 0.037 23.59 8 0.098Error var. invariance 1428.080 687 0.009 0.916 0.044 277.98 29 0.767
Note: Results for configural invariance are the fit values of two group poverty model.a Metric invariance establishes the equivalence of the basic meaning of the construct because the loadings denote the relationship between indicators and latent factor
(Byrne, 2001).b truct.
wcht
a1
It tests for the equality of the measured variable intercepts (means) on the consc It assesses the equality of variances of the factors across the groups.
ere found to be satisfactory30 as well. In addition to these, to
heck the equivalency of the model between GOs and NGOs, weave conducted a number of invariance tests. The results of theseests are displayed in Table 3.30 RMR, GFI, Hoelter are 0.076, 0.902, 315 and 364 respectively. AIC value (1574)nd ECVI value (2.811) were found to be less than that of saturated model (AIC is740 and ECVI is 3.107).
It allows the relative amount of latent factors to be compared between groups.
Table 3 shows that the change in chi-square is only 92.42 with28 degrees of freedom and the change in p value indicates anon-significant difference. Thus two models exhibit full metricinvariance. ��2 is 166.29 with a change in df of 29 which is not sta-tistically significant (as the change in p value is 0.136) thus scalar
invariance between the model is supported too. On the other hand,a ��2 of only 23.59 with change in 8 degrees of freedom (row 6,Table 3) shows only a little difference indicating that factor vari-ances are almost identical between the groups.
508 T.A. Chowdhury, P. Mukhopadhaya / The Journa
Table 4Comparative means of factors for NGO projects.
Estimate S.E. C.R. p
EFFECTIVENESS −0.425 .122 −3.492 ***ACCESS TO GENERAL INFORMATION 0.489 0.076 6.415 ***ACCESS TO GOVERNANCE INFORMATION 0.403 0.073 5.533 ***HUMAN CAPABILITY BUILDING −0.187 0.062 −3.037 .002ASSET BUILDING 0.384 0.051 7.607 ***FREEDOM −0.006 0.064 −.091 .927
N
amp
6r
mGi
gsstaoama
6
ruaTo
wotafimien
eerpip
cm
CORE NEED FULLFILLMENT −0.072 0.034 −2.132 .033
ote: *** means estimates are significant at 5% level.
Given these findings, all factor loadings, variances and covari-nces with additional error covariances of the structural povertyodel are invariant31 across GO and NGO’s poverty reduction
rojects.
. Comparing effectiveness of GO and NGO povertyeduction projects
In this section, validated and invariance checked structuralodel (shown in Fig. 3) will be used to compare the effectiveness ofO’s and NGO’s credit driven poverty reduction projects in enhanc-
ng the wellbeing of the poor beneficiaries in rural Bangladesh.Model assessment: Fit index results of this two-group model are
iven in Table A2 in Appendix A. Results show that the multidimen-ional poverty assessment model exceptionally fits with these twoets of data as CFI and RMSEA values are 0.932 and 0.039 respec-ively. While comparing our model, it was observed that both AICnd ECVI values of our model are smaller than that of saturatedr independence model. In addition, loading values were above thecceptable range (shown in Table 5). Thus we can conclude that ourodel is appropriate for the effectiveness comparison between GO
nd NGO projects.
.1. Comparison based on individual factors and items
We begin the comparison based on the main factors (exogenous)eported in Table 4. It is important to remember that the mean val-es for GO projects were fixed at zero, as we have taken GO projectss the controlled group and NGO projects as the estimated group.hese mean values reported in Table 5 are for the NGO projectsnly.
In spite of the large investment, wider coverage and largerorkforces of NGOs, statistically significant results suggest that,
ut of six poverty and wellbeing indicators, NGOs perform bet-er in three fields (access to general and governance informationnd asset building) whereas, GOs perform better in three otherelds (human capability building, freedom and core need fulfill-ent). However, in one field (‘Freedom’) the gap is marginal and
nsignificant.32 Thus even though government projects are better inmpowering poor people’s freedom, this difference is too small tootice.
It can be readily observed that NGOs are around 49% moreffective compared to government projects in delivering ‘gen-ral information’ (such as, natural disaster, job related, educationelated, loan related and health related information) to the rural
oor, whereas this rate is around 41% in sharing governance relatednformation (information about politics and government actionlans). Interestingly, it was found that the NGO projects are more
31 Invariance of this model was further tested between male and female benefi-iaries and in all cases results were found to be satisfactory which argues that thisodel is robust in comparing between different groups.
32 C.R. value is only −0.091.
l of Socio-Economics 41 (2012) 500– 512
effective (39% more) in helping their beneficiaries to create assets.This is certainly a positive sign of improvement because owner-ship of assets (particularly land and houses) reduces the level ofvulnerability of the rural poor.
GO projects perform comparatively better (19% more) in ‘humancapability building’, especially in reducing morbidity and physicalsickness. This result is quite justified as more rural poor take health-care services from government hospitals (even though hospitalsare remotely located) due to their limited access to NGO and pri-vate healthcare centers caused by financial constraints. However,the GO–NGO effectiveness difference in this field is comparativelysmall (19%) due to the presence of ‘village doctors’. Many rural poorvisit village doctors instead of GO or NGO healthcare centers to availprompt service at the lowest cost, or to get the services on credit.
It can be observed that GO projects perform 8% better in fulfilling‘core needs of the family’ particularly in food intake and providingeducation. Government’s continuing ‘Food for Work and Education’projects are responsible for this result as NGOs do not operate suchprojects. In addition, government’s aged allowance, poor allowanceand pension policy helped in this respects.
To check the overall performance of the organizations, we esti-mated the mean value of the ‘Effectiveness’ factor for NGOs whereEffectiveness is determined by Economic and Social Wellbeingand their respective indicators (including human capability build-ing, core need fulfillment, freedom, etc.) and credit related issues(including loan repayment rate, amount of loan, length of borrow-ing, etc.) (see Fig. 3). The result (−0.425) concludes (see Table 4, 1strow) that, as a whole, GO projects are more effective (by at least42%) than the operations of NGOs in improving living standards ofthe rural poor in Bangladesh. This contradicts the existing litera-ture, which stresses on the NGO domination over Government inpoverty reduction projects in developing countries.
Comparative statistics on remaining factors and individual mea-sured items are reported in Table 5 which shows that, out of 30remaining fields (excluding effectiveness factor), NGOs are superiorin 17 fields whereas GO agencies lead in 13 other fields. Detaileddiscussion for each item is provided in the next section.
7. Discussion and policy implications
The final column of Table 5 reveals that GO projects are moreeffective in improving ‘Economic wellbeing’ of the rural poorcompared to NGOs (loading value of 0.863 for GO and 0.454 forNGOs) whereas, NGOs are better in ‘Social issues’ (loading of 0.238whereas GO loading value is 0.014). But the alarming issue is thatboth GO and NGO projects have less impact on social issues, ascan readily be seen from absolute magnitudes of loading values.This finding further proves the domination of policies aimed forenhancing economic wellbeing in Bangladesh that by-passes socialaspect of poverty.
NGOs perform better in providing all types of general informa-tion (job, health, education, natural disaster information, etc.) to therural poor, especially information regarding loan sources and nat-ural disasters even though the loading values for GOs and NGOsare quite close (loading differences are 0.057 and 0.106 respec-tively). GO projects lead in providing political information to therural poor, however, the variation (only 0.003) is not that wide withNGO projects. Our result also shows that poor people obtain bet-ter information about the activities of government from the NGOs(loading value is 0.889 for NGO and 0.850 for GO). Noticeable resultsare the higher loading values of both GOs and NGOs in provid-
ing education and health information to the rural poor (all valuesare higher than 0.89) which demonstrate that both the serviceproviders perform exceptionally well in these social dimensions.For instance, Bangladesh Rural Advancement Committee (BRAC)
T.A. Chowdhury, P. Mukhopadhaya / The Journal of Socio-Economics 41 (2012) 500– 512 509
Table 5Standardized regression weights for GO and NGO poverty reduction projects.
Measured items Factors in the model Estimates for Comments
GOs NGO GO/NGO ratioa
EFFECTIVENESS ← ECONOMIC WELLBEING .863 .454 1.90 GOs are twice effectiveEFFECTIVENESS ← SOCIAL WELLBEING .014 .238 0.05 NGOs are 17 times better,
however, values are lowAccess to job info ← ACCESS TO GENERAL INFO .927 .948 0.97 Marginal differenceAccess to health info ← ACCESS TO GENERAL INFO .950 .970 0.97 Almost sameAccess to education info ← ACCESS TO GENERAL INFO .898 .928 0.97 Similar performanceAccess to loan info ← ACCESS TO GENERAL INFO .747 .804 0.93 Similar performanceAccess to natural disaster info. ← ACCESS TO GENERAL INFO .657 .763 0.86 NGOs are more effectiveAccess to political information ← ACCESS TO GOVERNANCE INFO .966 .963 1.00 Equal effectivenessAccess to government info ← ACCESS TO GOVERNANCE INFO .850 .889 0.95 Similar performanceCapacity to work normally (4-5 hours at stretch) ← HUMAN CAPABILITY BUILDING .931 .801 1.17 GOs are more effectiveMorbidityb ← HUMAN CAPABILITY BUILDING −.685 −.727 0.94 GOs are betterAverage sick days maleb ← HUMAN CAPABILITY BUILDING .359 .450 0.79 GOs perform much wellAverage sick days femaleb ← HUMAN CAPABILITY BUILDING .579 .570 1.01 GOs are marginally
effectiveLand holding status ← ASSET BUILDING .800 .855 0.93 Similar effectivenessHome ownership ← ASSET BUILDING .656 .614 1.17 GOs are more effectiveFreedom in doing Political work ← FREEDOM .672 .647 1.03 GOs are more effectiveFreedom in doing Religious work ← FREEDOM .645 .623 1.04 GOs dominanceFreedom in doing Cultural work ← FREEDOM .866 .983 0.88 NGOs are betterIncome/month ← ECONOMIC WELLBEING .452 .510 0.88 NGOs perform betterSave/month ← ECONOMIC WELLBEING .415 .610 0.68 NGO’s dominanceUse of sanitary latrine ← ECONOMIC WELLBEING .309 .353 0.87 NGOs are betterFood intake by kids ← CORE NEED FULLFILLMENT .941 .894 1.05 GOs are betterFood intake by male ← CORE NEED FULLFILLMENT .918 .891 1.03 GOs are effectiveDecision making opportunity at job ← SOCIAL WELLBEING .358 .616 0.58 NGOs are far effectiveMental stressb ← SOCIAL WELLBEING .430 .677 0.63 GOs are much effectiveLength of borrowing credit ← EFFECTIVENESS .595 .460 1.30 GOs are more effectiveLoan repayment status ← EFFECTIVENESS .544 .454 1.20 GOs are far betterAmount of loan taken ← EFFECTIVENESS .553 .793 0.69 NGOs are much betterLand holding size ← ASSET BUILDING .885 .960 0.92 Similar effectiveFeeling of insecurity ← SOCIAL WELLBEING .644 .811 0.79 NGOs are more effective
GOs e
shjr
apNtfvlabct
hhppti(vp
o
a Values bigger than one show GO domination and vice versa.b For average sick days, morbidity and mental stress, ratio lower than one means
chooling,33 free primary and secondary education by government,ospitals and clinics of Gonoshashto Kendra and government’s ‘Sur-
er Hashi34’ (major funding by USAID) and ‘Sobuj Chata’ clinics areesponsible for these findings.
GO projects were superior in building ‘human capabilities’mong rural poor. The results show that the beneficiaries of GOrojects (loading 0.931) have better ‘capacity to work’ compared toGO beneficiaries (loading of 0.801). Similar results were found in
he case of items like ‘morbidity’ and ‘average sick days per month’or male members of the family. This finding is in line with our pre-ious findings of GOs domination in core need fulfillment (refer toast row of Table 4). However, one interesting finding is that, ‘aver-ge sick days per month for female beneficiaries’ are less for NGOeneficiaries (with loading 0.570) compared to that of GO benefi-iaries (loading 0.579). This necessarily proves NGO’s more focusoward women empowerment.
NGOs dominate in ‘asset building’ aspects, particularly in ‘landolding size’ (loading is 0.960 compared to GOs’ 0.885) and ‘landolding status’ (0.855 compared to GOs’ 0.800). This means moreoor beneficiaries supported by NGOs could buy new land com-ared to those supported by GOs. On the other hand, it was foundhat, the ‘home ownership pattern’ is better in the case of GO (load-ng 0.656) beneficiaries compared to the recipient of NGO benefits
loading 0.614). This is an indication that the GO projects target sol-ent beneficiaries with more assets which are used for the collateralurposes.33 As of December 2009, 32,000 primary schools with 32,937 teachers were inperation to cater to the needs of 984,440 children where 65% were girls.34 They have coverage on 61 districts in Bangladesh with 320 static clinics.
ffectiveness and vice versa (values reported in bold).
Results show that GO beneficiaries enjoy more ‘freedom inperforming their political and religious activities’ whereas NGObeneficiaries are better off in cultural works (loading 0.983 com-pared to 0.866 of GOs). This is because NGOs conduct formalgroup meetings more frequently, thus their beneficiaries havemore opportunity for social and cultural engagement. Thereforewe recommend that GO projects should introduce the provisionof frequent meetings with the beneficiaries to explore suggestionsfrom them which can be useful in better implementation of theprojects. Similar recommendations were made by the study ofGoldin Institute (2007) and Bunning (2004) for other countries.
It was found that NGOs are more effective in ‘creating employ-ment’, thus helping to generate more income to the beneficiaries(loading is 0.510 compared to 0.452 for GOs). However, smallerloadings suggest that both GOs and NGOs need to improve thisparticular aspect. Development partners should provide with con-sultation and training to the beneficiaries about better utilization ofthe loan amount such that better output can be expected from theprojects. Not only income generation, NGOs are found to be moreeffective than GO projects in ‘creating savings’ of the beneficiaries(loading is 0.610 compared to 0.415) and this is because NGOs havemandatory saving scheme per week for the beneficiaries.
Interestingly it has been observed that, in all cases of ‘core needfulfilment’ (particularly food intake by the family members), GO
projects perform better than that of NGOs. This is primarily becausethe rate of interest charged by NGOs is much higher than the GOs(Fernando, 2006).35 Thus NGO beneficiaries are less fed despite35 Also see, Interest rates policy for MFIs streamlined in The Financial Express onApril 29, 2009. Check, The Independent on March 5, 2004.
5 Journa
ta
t(gb
b(bnimr
NiiotbprTdwls
tTos
TE
10 T.A. Chowdhury, P. Mukhopadhaya / The
heir higher earnings, because a major portion of their income leaksway on higher interest payments.
Results show that provision of health and hygiene (use of ‘sani-ary latrine’) is quite small by both GOs (loading is 0.309) and NGOsloading of 0.353). We recommend that more emphasis should beiven to the awareness and conscious building programs amongeneficiaries with respect to health and hygiene.
It is observed that in ‘decision making process at job place’, NGOeneficiaries are more empowered compared to GO beneficiariesloading value is 0.616 compared to 0.358). Two explanations cane offered. First, more NGO beneficiaries run their own small busi-esses or invest in farming, thus their decision making opportunity
s more. Second, due to more social engagement through groupeetings, NGO beneficiaries are better informed about their social
ights.It was observed that the ‘mental stress’ is more apparent among
GO beneficiaries (loading 0.677) compared to that of GOs (load-ng 0.430). This may be due to the excessive repayment pressuremposed by the NGO field workers on the beneficiaries. At the timef survey many NGO beneficiaries reported that they had to repayhe installment even if this means they go without food. It has alsoeen observed at the time of survey that, to tackle the repaymentroblem and to pay the installment of one NGO, many beneficia-ies borrowed money from rural money lenders (called Mohajon).hey argued that mohajons are more flexible than NGOs as theyo not ask for a weekly repayment. However, the end result is notelcome as the beneficiaries cannot repay to either NGOs or money
enders and are trapped in chronic poverty with endless mentaltress.
Our results also show that the ‘length of borrowing’ is larger in
he case of GO beneficiaries (loading is 0.595 compared to 0.460).hat means there are at least 30% more repeat borrowing in the casef GO beneficiaries compared to NGOs. This is an important mes-age for the development partners that charging higher interestable A1volving approaches of poverty and wellbeing.
Approach Conceptualization and limitations
Physiological approach Families being in primary poverty if their total earninsufficient to obtain the minimum necessities for
(Rowntree, 1901).Basic needs approach Minimum specified quantities of such things that a
necessary to prevent ill health, undernourishment
like (Streeten et al., 1981).Social deprivation approaches Poverty should be measured in terms of judgments
members of a particular society as the way they viereasonable and acceptable standard of living.
Human poverty approach Poverty can involve not only the lack of necessitiesmaterial wellbeing, but the denial of opportunitiesliving a tolerable life (UNDP, 1997).
Social exclusion approach It involves lack or denial of resources, rights, goodsservices, and the inability to participate in the sociewhether in economic, social, cultural, or political ar(Levitas et al., 2007).
Participatory poverty approach It argues that the statistics on income, consumptiohealth care, education do not represent other livingsuch as the lives of poor women in domestic violenthe role of woman in the family decision making pr(Chambers, 1983).
Human rights approach of poverty It is the understanding that every human being hasrights and that is not a charity or even a privilege (2003).
l of Socio-Economics 41 (2012) 500– 512
may cause less number of beneficiaries and more defaulters in thelong run. All these explanations are further supported by anotherfinding that the loan repayment rate is higher in GO projects (load-ing 0.544) compared to NGOs (loading 0.454). Perhaps GOs’ flexibleloan repayment schemes and lower interest burden made benefi-ciaries less loan defaulter. NGOs need to revise their interest ratesin line with GO rate and in addition, NGOs need to consider re-scheduling their loan repayment process. It was also observed thatthe NGOs deliver more loans and larger amounts of loans to thebeneficiaries (loading value is 0.793 compared to 0.553). However,our previous findings suggest that there is no direct correlationbetween loan size and living standard enhancement of the benefi-ciaries. Thus loan size may not matter to all beneficiaries; rather, itsbetter utilization with flexible repayment schedule would be moreeffective.
In summary, it can be claimed that GO projects need to concen-trate more on ‘social wellbeing’ issues whereas NGOs need to becareful about ‘economic issues’ particularly interest burdens andcore needs fulfillment. It should be noted that social sides of poorbeneficiaries are often ignored and both GOs and NGOs requiremore investment (with additional donor support) in social sides ofliving, particularly building awareness about social, cultural, reli-gious and political rights.
8. Conclusion
This paper has developed and validated a multidimensionalmodel of poverty for living standards comparison of poor bene-ficiaries. As invariance analysis was successful for the model, it wasutilized to compare the effectiveness of microfinance-driven GO
and NGOs in Bangladesh. It has been observed that as whole GOagencies are more effective in improving welfare of the poor bene-ficiaries compared to NGOs. However, our survey results show thatGO agencies need to concentrate more on social issues, especiallyIndicators in literature
ings areliving
This approach heavily relies on the money metric utility plus itaccounts a specified proportion of the food expenditure for non-fooditems.
reand the
Inadequate fulfillment of different basic needs (including hunger,education, child and maternal health, etc.) starting from food to lifeexpectancy and mortality.
by thew a
Social deprivation includes indicators relevant to human rights,freedom and participation including decent housing, good workingconditions, accumulated wealth, access to credit, family relationships,access to social networks, caring friends or relatives, etc.
of for
Poverty as the lack of basic human wellbeing such as illiteracy,malnutrition, shorter life span, poor maternal health, and illness frompreventable diseases (UNDP, 1997).Poverty includes lack of access to goods, services, information andinfrastructure like energy, sanitation, education, communication, puredrinking water, etc.
andtyenas
Includes the concepts of human rights, social participation, socialintegration, cultural activities and political aspects including voice topolitical participation, personal security, the rule of law, freedom ofexpression and equality of opportunity (Tilly, 2006).
n, issuesce orocess
Aspects of wellbeing and the quality of life-security, self respect,justice, social life, decision making, political participation, etc.Participation means that it is my right to be involved in makingdecision, planning and reviewing an action that might affect me.
someUNDP,
Getting basic educational facility, training and health-care is afundamental right of every citizen.Human rights based development thus requires: (i) participation andtransparency in the decision making, (ii) non-discrimination in thesocial, political and economic life, (iii) empowerment and (iv)accountability of the actors like state and private sector.
T.A. Chowdhury, P. Mukhopadhaya / The Journa
Table A2Summary of goodness of fit statistics for the GO–NGO structural model.
Fit index Effectivenessmodel
Saturatedmodel
Independencemodel
NPAR 219 928 116CMIN (chi-square) 1312.869 0.000 9669.639df 709 0 812p 0.000 – 0.000NFI 0.864 1.000 0.000IFI 0.933 1.000 0.000TLI 0.922 – 0.000CFI 0.932 1.000 0.000RMSEA 0.039 – 0.140AIC 1750.869 1856.000 9901.639ECVI 3.127 3.314 17.681MECVI 3.230 3.751 17.736
ocmosGbapsca
A
p
A
R
A
A
AABB
B
B
B
C
CC
C
D
E
Results of the Case Study Evaluations. Working Papers No. 68. Overseas Devel-
Hoelter 0.05 331 – 52Hoelter 0.01 343 – 54
n empowerment building of the poor through group meeting pro-esses and employment creation. On the other hand, to reduce theental stress of the beneficiaries, NGOs need to concentrate more
n their loan delivery, rate of interest and repayment schedulesuch that these would not be a burden for the beneficiaries. BothOs and NGOs need to consider human capability building to makeeneficiaries more capable of earning throughout the year as it islways believed that economic solvency is more important to theoor people. In addition, both GOs and NGOs’ less contribution toocial aspects of poverty is disturbing. However, their remarkableontribution in health and education in rural Bangladesh should beppreciated.
cknowledgement
Authors wish to acknowledge the anonymous reviewers of thisaper for their insightful and constructive comments.
ppendix A.
See Tables A1 and A2.
eferences
hmed, M.U., 2004. Socio-demographic correlates of rural poverty in Bangladesh:a case study of Gaibandha Sadar and Tanore Upazillas. Bangladesh e-Journal ofSociology 1 (2).
lkire, S., Black, R., 1997. A practical reasoning theory of development ethics: fur-thering the capabilities approach. Journal of International Development 9 (2),263–279.
rbuckle, J.L., 2009. Amos 18 User Guide. Amos Development Corporation, USA.sian Development Bank (ADB), 2003. Annual Report of ADB.angladesh Bureau of Statistics, 2007. Statistical Year Book.ooysen, F., 2002. Measuring poverty with data from the Demographic and Health
Survey (DHS): an application to South Africa. Journal for Studies in Economicsand Econometrics 26 (3), 53–70.
unning, A., 2004. A microcredit success story: Women make the difference. Devel-opment Bulletin 64, 63–66.
urra, N., 2005. Women and microcredit: some challenges. Note prepared for theNational Commission of Farmers in India.
yrne, B.M., 2001. Structural Equation Modelling with Amos-Basic Concepts, Appli-cations and Programming. Lawrence Erlbaum Associates, Inc., USA.
arroll, T., 1992. Intermediary NGOs: The Supporting Link in Grassroots Develop-ment.
hambers, R., 1983. Rural Development: Putting the Last First. Longman, Harlow.lark, D.A., 2005. The capability approach: its development, critiques and recent
advances. Global Poverty Research Group-WPS-032.oleman, I., 2005. Defending microfinance. Fletcher Forum of World Affairs 29,
181–190.
eng, G., 2008. Public Donations to Help Disaster Relief in the Aftermath of theSichuan Earthquake in China. KOF Swiss Economic Institute in Zurich and theRadboud University in Nijmegen.
dwards, M., Hulme, D., 1996. Too close for comfort? The impact of official aid onnongovernmental organizations. World Development 24, 961–973.
l of Socio-Economics 41 (2012) 500– 512 511
Epstein, M.J., Crane, C.A., 2005. Alleviating global poverty through microfinance: fac-tors and measures of financial, economic, and social performance. Draft versionretrievable.
Farrington, J., Bebbington, A., Wells, K., Lewis, D., 1993a. Reluctant Partners? NGOs,the State and Sustainable Agricultural Development. Routledge, London.
Farrington, J., Lewis, D., Satish, S., Miclat-Teves, A. (Eds.), 1993b. NGOs and the Statein Asia: Rethinking Roles in Sustainable Agricultural Development. Routledge,London.
Fernando, A.N., 2006, May. Understanding and Dealing with High Interest Rateson Microcredit: A Note to Policy Makers in the Asia and Pacific Region. AsianDevelopment Bank Publication.
Goetz, S.M., Gupta, S.R., 1996. Who takes the credit? Gender, power and controlover loan use in rural credit programs in Bangladesh. World Development 24(1), 45–63.
Goldin Institute, 2007. Improving microcredit programs: listening to recipients.Research conducted jointly with Nijera Kori and Unnayan Onneshan.
Government of Bangladesh (GOB), 2010. Economic Review. Ministry of Finance,Peoples Republic of Bangladesh.
Hair, F.J., Black, W.C., Babin, B.J., Anderson, R.E., 2009. Multivariate Data Analysis: AGlobal Perspective, 7th ed. Pearson Publication.
Hasan, A., 1993. Scaling-Up the OPP’s Low-Cost Sanitation Program. Orangi PilotProject, Karachi.
Hashemi, S., 1992. State and NGO Support Networks in Rural Bangladesh: Con-cepts and Coalitions for Control. Center for Development Research, Copenhagen,Denmark.
Hoque, S., 2004. Micro-credit and the reduction of poverty in Bangladesh. Journal ofContemporary Asia 34 (1), 21–32.
Hossain, M., 1988. Credit for Alleviation of Rural Poverty: The Grameen Bank inBangladesh. IFPRI Research Report 65. IFPRI, Washington, DC.
Howes, M., Sattar, M.G., 1992. Bigger and better? Scaling-up strategies pursued byBRAC 1972–1991. In: Edwards, M., Hulme, D. (Eds.), Making a Difference: NGOsand Development in a Changing World. Earthscan, London.
Kabeer, N., Neoponen, H., 2005. Social and economic impacts of PRADAN’s self helpgroup microfinance and livelihoods promotion program: analysis from Jhark-hand, India. Imp-Act Working Paper No. 11.
Kelly, A., 2008, February. Growing Discontent. The Guardian.Kevane, M., Wydick, B., 2001. Micro-enterprise lending to female entrepreneurs:
sacrificing economic growth for poverty alleviation? World Development 29,1225–1236.
Korten, D., 1990. Getting to the 21st Century: Voluntary Action and the GlobalAgenda. Kumarian Press, West Hartford.
LaFond, A., 1995. Sustaining Primary Health Care. Earthscan, London.Levitas, R., Pantazis, C., Fahmy, E., Gordon, D., Lloyd, E., Patsios, D., 2007. The
multi-dimensional analysis of social exclusion, a research report for the socialexclusion task force, viewed 30 June 2007.
MacCallum, R.C., Austin, J.T., 2000. Applications of structural equation modelling inpsychological research. Psychological Research Journals 51, 201–226.
Mahbub, A., Roy, R.D., 1999. Anaemic towards wellbeing. Working Paper No. 20.BRAC-ICDDR. B Joint Research Project, Dhaka.
Mahmud, W., 2008. Social development in Bangladesh: pathways, surprises andchallenges. Indian Journal of Human Development 2 (1).
Mahmud, K.M., Ahmed, S., 2003. Relative efficiency of government and non-government organisations in implementing a nutrition intervention programmea case study from Bangladesh. Public Health Nutrition 6 (1), 19–24.
Mayoux, L., 1999. From access to empowerment: gender issues in micro-finance.CSD NGO Women’s Caucus Position Paper for CSD-8.
McGillivray, M., Noorbakhsh, F., 2007. Composite indexes of well-being: past,present and future. In: McGillivray, Mark (Eds.), Human Well-Being: Conceptand Measurement. Palgrave Macmillan, Basingstoke, England, pp. 113–134.
Meyer, C., 1992. A step back as donors shift institutions building from the public tothe “private” sector. World Development 20 (8), 1115–1126.
Morduch, J., 2000. The microfinance schism. World Development 28 (4), 617–629.Nabi, R., Datta, D., Chakraborty, S., Belgum, M., Chadhury, N.J., 1999. Participatory
Poverty Assessment in Bangladesh. PREM, World Bank, Washington.Narayan, D., Walton, M., Chambers, R., 2000. Voices of the Poor: Crying Out for
Change. Oxford University Press, USA.Nussbaum, M.C., 2005. Well-being, contracts and capabilities. In: Manderson, L. (Ed.),
Rethinking Well-Being. API Network, Perth, pp. 27–44.Pitt, M.M., Khandker, S.R., 1998. The impact of group-based credit programs on poor
households in Bangladesh: does the gender of participants matter? Journal ofPolitical Economy 106 (October), 958–996.
Pitt, M.M., Khandker, S.R., 2001. Credit programs for the poor and seasonality inrural Bangladesh. Research funded by the World Bank project “Impact of Tar-geted Credit Programs on Consumption Smoothing and Nutrition in Bangladesh”under research grant RPO 681-09.
Ravallion, M., 1998. Poverty Lines in Theory and Practice. Living Standards Measure-ment Surveys. Working Paper No. 133. World Bank, Washington, DC.
Ravallion, M., Shah, S., Wodon, Q., 1999. Bangladesh: From Counting the Poor toMaking the Poor Count, A World Bank Country Study. The World Bank, Wash-ington.
Riddell, R., Robinson, M., 1992. The Impact of NGO Poverty-Alleviation Projects:
opment Institute, London.Rowntree, B.S., 1901. Poverty: A Study of Town Life. Macmillan, London.Ruggeri Laderchi, C., 2001. Participatory Methods in the Analysis of Poverty: A Crit-
ical Review. QEH Working Paper 62. Queen Elisabeth House, Oxford.
5 Journa
SSSS
S
S
S
S
T
UUV
W
nomic Inequality, Applied Economics, Journal on IncomeDistribution, Journal of Asian Economics, Asian EconomicsJournal, Economic Record, Netherlands International Law
Review, Journal of Contemporary Asia. Invited Speaker Fulbright Program, SouthernEconomic Association Meeting, WHO Pacific Island Regional Meeting.
12 T.A. Chowdhury, P. Mukhopadhaya / The
chneider, B., 1988. The Barefoot Revolution. IT Publications, London.en, A.K., 1982. Choice, Welfare and Measurement. Basil Blackwell, Oxford.en, A.K., 1984. Resources, Values and Development. Basil Blackwell, Oxford.en, A.K., 1993. Capability and wellbeing. In: Nussbaum, M.C., Sen, A. (Eds.), The
Quality of Life. Oxford University Press, pp. 30–53.haffer, P., 2008. New thinking on poverty: implications for globalization and
poverty reduction strategy. RER (47), 192–231.himp, T.A., Sharma, S., 1987. Consumer ethnocentrism: construction and validation
of the CETSCALE. Journal of Marketing Research XXIV (August), 280–289.ilver, H., 2007. The process of social exclusion: the dynamics of an evolving concept.
CPRC Working Paper 95.treeten, P., Burki, J.S., Haq, M.U., Hicks, N., Stewart, F., 1981. First Things First: Meet-
ing Basic Human Needs in Developing Countries. Oxford University Press, NewYork.
illy, C., 2006. The politics of exclusion. Paper produced for the World Bank ‘Movingout of Poverty’ project.
NDP, 1997. Human Development Report. Oxford University Press.NDP, 2003. Human Development Report – 2003.ivian, J., 1994. NGOs and sustainable development in Zimbabwe: no magic bullets.
Development and Change 25, 181–209.orld Bank, 2000. World Development Report 2000/2001: Attacking Poverty. World
Bank, Washington, DC.
Tamgid Ahmed Chowdhury has received his PhD in Eco-nomics from Macquarie University, Sydney, Australia andspecialized in poverty, inequality and microfinance oper-ations. Before joining the PhD program, he served as anAssistant Professor at Daffodil International University,Bangladesh. He is the author of two text books; one inDevelopment Economics and another in Marketing. Hehas several publications in reputed international journals
such as Oxford Development Studies, South Asian Journal ofPopulation and Health, Journal of ICMA, Dhaka UniversityJournal of Marketing, etc. Dr. Chowdhury has presentedpapers in several international conferences held in USA,Australia, South Korea and Thailand.l of Socio-Economics 41 (2012) 500– 512
Pundarik Mukhopadhaya [PhD, University of New SouthWales]: currently employed as Senior Lecturer, MacquarieUniversity, Sydney, Australia. I have research collabora-tions with Stockholm School of Economics, NottinghamUniversity, National University of Singapore and IndianInstitute of Development Studies. Consultancy clienteleincluded UNESCO, the World Bank and WHO. Publica-tions involved 1 book, 15 chapters in books, 25 academicpapers in international refereed journals on theoreticaland empirical economics including Researches on Eco-
