A Welfare Economic Analysis of the Impact of Microfinance in Bangladesh AKM Ghulam Hussain Department of Economics University of Dhaka Nigar Nargis Department of Economics University of Dhaka April 30, 2008 Correspondence: Nigar Nargis, Assistant Professor, Department of Economics, University of Dhaka, Nilkhet, Dhaka-1000, Bangladesh, Phone: (880) 2-8011151, Fax: (880) 2-8615583, E-mail: nnargis2100@yahoo.com. Abstract Using data of about 2700 households in Bangladesh from a longitudinal survey for the period of 1998-2004, this study undertakes a welfare economic analysis of the impact of microcredit program in Bangladesh. First, the welfare dominance approach shows that aggregate welfare of the rural households improved over the period of 1998-2004. Household income increased across all income percentiles for all of regular, occasional and non-participant groups. Regular participants experienced the lowest welfare gain and non-participants gained most. By the abbreviated social welfare function approach, on the other hand, the welfare outcome appears ambiguous. The reallocation of labor resource from farming to non-agricultural self-employment opportunities resulted in income and productivity growth and poverty fell across all groups by MFI participation status. Average work hour, however, did not increase leaving the status quo of pervasive underemployment in the rural economy of Bangladesh. Moreover, both between and within group inequality increased. When we break down the sample of households by the status of eligibility to participate in a microcredit program, we find that eligible participants performed no better than the eligible non-participants and non-eligible participants fared far worse than non-eligible non-participants. It appears that the contribution of microcredit program participation lies in helping the households at the lower side of the economic strata to keep up with the rest of the society possessing comparable initial endowments. On the higher side, it appears to even detract from their growth potential. Thus this study finds evidence against the popular belief that microcredit is instrumental to uplifting of the rural poor to a higher economic status. There is no denying the fact that poverty would have been alleviated at a faster pace if it was accompanied by income and productivity growth with greater employment opportunities and reduced inequality. 1 I. Introduction The micro credit program in Bangladesh began its journey in the late 1970s and assumed maturity by the mid-1990s. The following decade has seen replications of the program in different parts of the country so numerously that any village without such interventions can hardly be found. The history of microfinance revolution in the last quarter of a century is marked with successes and failures which have been recorded in a wide array of literature1. These studies so far focused on the absolute impact of microfinance in terms of consumption increase (Khadakar, 2001), poverty reduction (Khandakar, Hossain, 1988) and improvement in social indicators such as women empowerment (Matin, 1996). However, none of these studies investigated income growth, employment generation, poverty alleviation and income equalization across the microcredit participant and non-participant groups over time irrespective of their eligibility status for the participation in microcredit program. As such, the overall change in the welfare status of rural households in Bangladesh in the wake of microcredit revolution is yet to be analyzed. In this paper, we first take an overview of the aggregate welfare economic change over time aiming at a more complete understanding of the role microcredit can potentially play in welfare improvement in rural Bangladesh. Against this background, we then undertake the welfare analysis across microcredit participant and non-participant groups by their eligibility status in an attempt to isolate the marginal contribution of microcredit in welfare improvement. 2 This paper addresses the following questions. First, how have the participants and the non-participants of the microcredit program shared the overall economic growth performance? Second, to what extent has the micro credit program stimulated selfemployment activities and income generation, thereby leading to poverty reduction, of its beneficiaries relative to the non-participants? Third, has this growth been accompanied by increased (or reduced) inequality, both between and within participant and nonparticipant groups? The paper is organized in eight sections. Section II lays out the analytical framework for the study. Section III describes the income growth experience of rural households, first overall and then by participation status in the microcredit program. Section IV decomposes household income to trace out the source of income growth. Section V investigates the accompanying growth and redistribution of employment among wage employment, farming, non-farm agricultural activities and non-agricultural selfemployment activities. Section VI reports the resultant state of poverty and inequality associated with the observed income and employment growth. Section VII presents the welfare analysis by the eligibility status of rural households and identifies the marginal contribution that microcredit plays in this respect. The paper summarizes its findings and concludes in Section VIII 3 II. Analytical framework The data used for the analysis comes from a panel of about 2700 households that was collected during 1997-2004 in four waves to evaluate the socio-economic impact of microcredit in a quasi-experimental set-up involving comparison between program and control villages (dispersed throughout the country) and between participants and nonparticipants in the program villages (Zohir et al. 2001; Rahman et al. 2005).1 Household participation status was defined in terms of current membership of either member of a household with microfinance institutions (MFIs). The households were grouped into eligible participants, eligible non-participants, non-eligible participants and non-eligible non-participants. As the criterion for eligibility varies across MFIs, the ownership of 50 decimals or less of cultivable land was used to define eligibility. The sample households were broadly classified into three groups on the basis of individual affiliation with MFIs: (i) regular participants: participants in any microcredit program in all four rounds of the longitudinal survey; (ii) occasional participants: who may drop out for a period but may rejoin; (iii) non-participants: those who never participated in any microcredit program over the observation period. For the present analysis, we maintain this classification throughout the paper. Within the participant categories, the convenience of occasional participation as opposed to the persistent dependence on microcredit is expected to result in better performance of 1 However, no control villages could be discerned when the same villages were revisited in 2004, as is expected given the rapid rate of expansion of the microfinance institutions (MFIs) in rural Bangladesh. 4 occasional participants than the regular participants. We have considered the changes over two surveys conducted in 1997-98 and 2004 to allow a considerable length of time to elapse between the beginning and the end of the observation period such that the long term welfare impacts are accounted for. The welfare impact of changes in income distribution can be captured using the welfare dominance approach due to Hadar and Russell (1969) and Saposnik (1981). As the theorem says, “… one distribution X first-order-dominates another distribution Y for the class of anonymous, increasing social welfare functions if and only if the income of the person in each rank in X is at least as great as the income of the person with the corresponding rank in Y and strictly greater someplace” (Fields 2005). By the axiom of anonymity of individuals in the income distribution, this theorem does not require that the same person be followed over time. What the welfare dominance method suggests is that we make welfare comparison between the beginning and the end of the observation period undergoing some program intervention on the basis of the comparison of income percentiles for example. If income increases at every percentile, we can conclude that welfare has improved. Since this approach involves overall welfare comparison, we have used total household income for determining the implied rank dominance. For a deeper understanding of the rural labor market outcomes due to microcredit program participation, we couch the analysis in the comparative static framework 5 following Fields (2005). Fields studied theoretically the welfare consequences of labor market policy changes in modern-versus-traditional sector or urban-versus-rural sector environment of the Harris-Todaro model. On the other hand, the present analysis is limited to the rural sector characterized by traditional land based or dependent employment vis-à-vis growing self-employment opportunities, presumably created by the extensive network of the MFIs in the country. In a paradoxical labor market where per capita income is one of the lowest in the world but reported official unemployment rate is no more than 5%, it is clearly the extent of underemployment rather than the unemployment rate that should be its health indicator. Thus the criterion of welfare judgment regarding employment performance is that the increase in the number of employed should be accompanied by growth of work intensity which would be reflected in the growth in annual average working hour per person. Given wage rate, average annual hours worked can, therefore, be used as a proxy of effective employment. This two-dimensional employment growth should cause household labor earnings to grow and poverty to fall. If the process of all these simultaneous changes takes place with lower inequality of income, we are able to conclude unambiguously that social welfare has increased. The class of social welfare functions that summarize these welfare judgments is named abbreviated social welfare function, which can be formally expressed as follows in the present context: W = f (household labor earnings; annual average work hour; poverty; inequality). 6 Each of the argument in the welfare function is a vector of three elements representing the corresponding outcomes for three groups of the population: regular, occasional and non-participants of the microcredit program. So f i j would denote the conditional partial derivative that represents the change in W in response to a given change in the i th argument for the j th group (j = regular, occasional, non-participant). The spill-over effects of microcredit program enjoyed by the non-participants are included in their corresponding partial derivative. The welfare evaluation criteria can be written in the order of the arguments in the welfare function as: f1 j > 0 ; f 2j > 0 ; f 3j < 0 ; f 4j < 0 . Without any ad hoc functional specification of the abbreviated social welfare function, its partial derivatives can be identified with the change of outcome over time conditional on the participation status of households or individuals. The sign of the change in W can be ascertained provided all the changes are aligned in the same direction. It should be noted that while increasing income increases welfare unambiguously, increasing work hour may not necessarily do so. This is because there is a disutility from work arising from lower time for leisure. Nevertheless, we maintain the criterion that the partial derivative for work hour is positive in view of the fact that in an economy with severe underemployment and income shortfall from the minimum requirements, increase in work hour leads to increasing welfare. 7 The ranking of these outcomes across the groups would indicate whether one group is doing better than the other, in general. Their relative performance can then be attributed to the participation status once we break down the analysis by the eligibility status of the participant and non-participant households. Thus a normative evaluation of the impact of microcredit program on the rural labor market conditions in Bangladesh can be performed. III. Growth in household income2 The annual income of households is calculated by adding the sales value of annual household production of crops, livestock and fishery, earnings from wage employment, non-agricultural self-employment, and other sources such as service, help from relatives, relief, food for education, remittance, house rent, etc. The average annual household income grew at an annual compound rate of 3.88% to rise from Tk.48,195 in 1998 to Tk.60,546 in 2004 (Table 1). Comparing the percentiles of annual household income between 1998 and 2004, we find that growth occurred across all the income strata. This change in household income distribution is shown in Figure 1a by the upward shift in the curve of annual household income percentiles. In particular, the median income increased from Tk.36,008 to Tk.41,550 marking an annual compound rate of growth of 2.41%. This rank dominance of the income distribution in 2004 over that in 1998 implies that the income of the household in each rank in 2004 is greater than the income of the household with the corresponding rank in 1998 for every percentile. In other words, the household that ranks poorest in each distribution has a higher real income in 2004 than in 1998, and 2 All the income figures are converted to 2004 constant prices. 8 likewise for the second poorest household, the third, and so on. It in turn implies that poverty diminished in 2004 compared to 1998. The remarkable growth over 1998-2004 was, however, not uniform across the income distribution of the sample households. It concentrated at the upper half of the households, peaking for the top 1% of the income distribution and exponentially decreasing up to the bottom income stratum (Figure 1b). It suggests that income inequality increased in the rural economy during the period under observation. Two questions are pertinent as far as of the role of microcredit is concerned. First, how this unequal growth performance was shared by the regular, occasional, and nonparticipants of the microcredit program. In other words, what the contribution of between group inequality was to the observed unequal household income growth. Second, to what extent this unequal growth is a result of within group inequality in income. Comparing the annual average household income levels by participation status, we find that non-participants ranked highest followed by the occasional and then the regular participants at both the beginning and the end of the observation period (Table 1). The same order is observed in the gain in average income levels over the period under observation. The gain experienced by non-participants is twice as high as that of regular participants. As a consequence, the average income gap widened between these groups over time. 9 While all the three groups of households across all the income percentiles experienced growth in household income, the overall growth performance seems to have been primarily driven by growth at the top quartile of households as shown in Figure 2a. Among the three groups, regular participants experienced the least increase. In order to have a closer look at the growth differences across groups, we have blown up the figure from 1st to 75th percentiles for both years in Figure 2b. It shows that the growth level curves of occasional and non-participants crisscrossed each other at several points below the 75th percentile suggesting similar income gain of this two groups at the lower three quartiles. At the top quartile, however, non-participants dominated. IV. Decomposition of household income To figure out what stunts the growth of household income for regular participants compared to occasional- and non-participants, we resort to the decomposition of total household income by two broad sources—endogenous labor income and exogenous income. By endogenous labor, we mean engagement of household members in wage employment, household agricultural activities including farming and non-farm agricultural activities, and non-agricultural self-employment activities within their locality. Income from this source is likely to depend on, among others, household participation status in any microcredit program. Exogenous income, on the other hand, comes largely from foreign remittance, besides services outside the locality, rental income from assets, and transfers. This is the unearned income accruing to the individuals living in the household and is unlikely to depend on their MFI participation status. 10 As shown in Table 1, the growth of endogenous labor income of regular participants fell short of that for the other two groups of households. As a result, despite beginning with the higher average of the three, regular participant households ended up with the lowest average endogenous labor income by 2004. Thus it appears that the growth in endogenous labor income was not equalizing across participation status of households. In terms of average exogenous income, regular participants began from an initially disadvantageous position and maintained it through 2004 as shown in Table 1. Nonparticipant households took the lead in both endogenous labor income and exogenous income growth. In terms of per capita income (total household income divided by household size), regular and occasional participants did not perform any better in 2004 than in 1998 (Table 2). Non-participants, on the other hand, added Tk.1,400 to per capita income on average. However, the comparison of income growth performance in terms of per capita income in the traditional rural setup in Bangladesh is misleading in the present context. It understates the welfare gain from income growth for young households with proportionately larger number of young children compared to the senior households with larger proportion of working age adults. To address this concern and keeping in view the presumption that the role of microcredit is mostly reflected in the growth performance of endogenous labor income, we have looked at the endogenous labor income per working member of the household. This variable is essentially a measure of individual labor productivity. By this measure, as shown in Table 2, regular participant households made the least progress among the three groups. Occasional participants added to their labor 11 productivity by almost twofold and non-participants added by more than threefold of the productivity gain experienced by regular participants.3 V. Growth and redistribution of employment In an attempt to understand why microcredit program participants did not experience as much growth in labor productivity as the non-participants, we investigated the pattern of allocation and reallocation of their labor resource by type of employment. The annual average hours worked by individuals in the sample households did not register any increase between 1998 and 2004—as shown in Figure 3, it changed from 1,289 to 1,285 hours. This low level of annual work hour reflects the severity of underemployment in the rural sector of Bangladesh. It is expected that microcredit would widen the scope of self-employment on top of the existing employment opportunities. So work hour would increase in the aggregate as well as per working member of a household, especially in the participating households. Although average work hour did not rise, a remarkable change occurred in the mix of hours by type of employment—a change that reveals the dynamics of the redistribution of labor resource occurring between these two years. While the average annual working hour decreased in wage employment and farming, it remained almost the same in nonfarm agricultural activities and increased significantly in non-agricultural selfemployment activities, such as, trading and micro-enterprises. 3 The greater increase in productivity than in per capita income over 1998-2004, as shown in Table 2, is partly explained by the slower growth in the number of working members per household than household size. This is due to the fact that the children born during the six years under observation do not readily turn into working members whose income generating capacity can be potentially influenced by participation in a microcredit program. Nor do children demand equal amount of household resources as adults. 12 Similar pattern is observed in the proportion of total annual household work hours to different types of employment as shown in Table 3. The fall in the proportion of hours in farming from 15% to 10% in six years with concomitant rise in non-agricultural selfemployment from 38% to 46% indicates that the incentives to withdraw from subsistence farming and to move towards more rewarding self-employment activities outside the agricultural sector are at work. The continual reallocation of labor resource resulted in the proportion of hours employed in self-employment activities, including both non-farm agricultural and non-agricultural types, to increase from 0.54 to 0.61 contributing to the overall growth in labor productivity of rural households. The pattern of inflow of labor to self-employment activities, both in terms of average annual hours worked per person and the allocation of hours, is found to be similar across the regular, occasional and non-participants (Table 3 and Figures 4a, 4b, and 4c). Within group comparison of allocation of work hours reveals two important facts. First, regular participants worked longer hours than the other two groups, particularly in nonagricultural self-employment. Second, in both years, regular participants allocated the largest share of their work hours to self employment including non-farm agricultural and non-agricultural activities. By 2004, they were spending two-thirds of their time in self employment in contrast to 60% by occasional participants and 56% by non-participants. This pattern of household labor allocation, along with the fact that their labor productivity 13 remained lower of the three groups, suggests that their marginal return to labor in self employment activities is low and also experienced slower growth. VI. Poverty and inequality Poverty by all of the four measures shown in Table 4 decreased from 1998 to 2004.4 The incidence of poverty and the rate of movement out of poverty are, however, not uniform across the MFI participation status. By the head count index, the poverty gap ratio, and the Sen index, regular participants were the poorest group in both years. By the FosterGreer-Thorbecke index of order 2, although they ranked the least poor of the three groups in 1998, they turned into the poorest by 2004. Moreover, regular participants made the slowest progress in the movement out of poverty and in narrowing the shortfall from the poverty line income. The performance of occasional participants was close to or even better than the non-participants by this criterion. The slower rates of growth in income and productivity of regular participants observed in the earlier sections explain why regular participants did not make as much progress as the other two groups in poverty reduction. We would have observed a greater rate of reduction in poverty if it were accompanied by reduction in inequality. The shifting out of the Lorenz curve drawn on the basis of the distribution of per capita income in Figure 5, however, indicates that inequality increased between 1998 and 2004. The overall Gini coefficient increased from 0.38 in 1998 to 0.40 in 2004 indicating rise in inequality (Table 5). The within group Gini coefficients show 4 These measures are based on total income per capita. 14 that inequality increased among both regular and non-participants, more so among nonparticipants than among regular participants. It did not change much among occasional participants. However, it is not clear at this point what role microfinance plays in the observed trend in inequality. It can be understood using the decomposition of Gini coefficient into between- and within-group coefficients, as shown in the lower panel of Table 5. The between-group inequality measure represents what the degree of inequality would have been if the total income of each group were equally distributed among the group members.5 Thus this part of inequality is attributable to the segregation of population into different groups, by microfinance intervention in this case. As shown in Table 5, about one-third of overall inequality in income can be explained by inequality within the groups of regular, occasional and non-participants in both years. The contribution of between-group inequality, on the other hand, increased from 13.5% in 1998 to 20.9% in 2004. Thus it appears that microcredit intervention did not contribute to bridging the inequality between participants and non-participants. As far as labor market outcome is concerned, we can specifically investigate the between group inequality of labor productivity. The emphasis on labor income per working individual allows us to abstract from the effects of changes in household composition or any redistributive policy on household income (Burtless 1999). This way the impact of 5 The measure of between-group inequality is obtained by assigning the group mean per capita income to all individuals within the group and calculating the inequality of that smoothed income distribution. 15 microcredit intervention on inequality through self-employment generation can be isolated. To this end, we calculate group-wise average of productivities, assign the average productivity to each individual within the corresponding group, and plot Lorenz curve on the basis of the group mean so that within group inequality is smoothed out. The resulting Lorenz curve is shown in Figure 6, which displays two kinks dividing the population into the three comparison groups under consideration. Regular, occasional and non-participant groups are represented in order of their average productivity level. The slope of each line segment represents the average productivity of the corresponding group. The movement of the curve over 1998-2004 away from the line of equality implies increasing inequality of labor productivity between groups, which in turn contributes to the increase in overall inequality of household income distribution reflected in Figure 5. VII. Welfare economic analysis by eligibility status With a view to identifying the net welfare impact attributable to microcredit program participation, we compare the income and employment outcomes across participants and non-participant households that are comparable by eligibility status. For eligible households, total household income and endogenous labor income do not vary significantly across participants and non-participants in either year, although differences exist in exogenous income levels (Table 6a). Of particular importance is the growth in endogenous labor income that appears to be not much different by participation status— Tk.6,568 for regular participants, Tk.7,187 for occasional participants and Tk.6,007 for 16 non-participants.6 In contrast, among non-eligible households, non-participants experience significantly larger growth in endogenous labor income—three times as much as for regular participants and 1.4 times as much as for occasional participants (Table 6b). Similarly, when we compare labor productivity, we observe no significant difference between regular, occasional and non-participants among the eligible households-Tk.1,490 for regular participants, Tk.1,839 for occasional participants and Tk.1,457 for non-participants (Table 7). Non-eligible non-participants, in contrast, performed far better than the non-eligible regular and occasional participants (Table 7). Taking ownership of cultivable land as the proxy for initial resource endowment and assuming that eligible households are alike in that respect, we can infer that the contribution of microcredit to income and productivity growth is negligible. For noneligible households, the outcome is even worse. It is the dispersion in the performance of non-eligible participants and non-participants that is contributing to the increase in the between group inequality observed in the previous section. The productivity growth across groups with increased between group inequality suggests that the efficiency gain in the allocation of labor resource was not redistributive. This finding contrasts the observation by Armendáriz and Morduch (2005, p.34) that provision of access to financial services to the poor can both enhance their opportunities relative to the non-poor and increase aggregate productive efficiency. The optimism, spelled out by 6 The hypothesis that these differences in average endogenous labor income over time are equal across the comparison groups cannot be rejected at 1% level of significance. 17 Aghion, Caroli and Peñalosa (1999) a few years ago that microcredit intervention might have started off the virtuous circle in which redistributive policy can be used to reduce inequality, which in turn would accelerate growth and thereby automatically induce further reductions in inequality, seem to have not come true. VIII. Summary and Conclusion Using data of about 2700 households in Bangladesh from a longitudinal survey for the period of 1998-2004, this study undertakes a welfare economic analysis of the impact of microcredit program in Bangladesh. First, the welfare dominance approach shows that aggregate welfare of the rural households improved over the period of 1998-2004. Household income increased across all income percentiles for all of regular, occasional and non-participant groups. Regular participants experienced the lowest welfare gain and non-participants gained most. By the abbreviated social welfare function approach, on the other hand, the welfare outcome appears ambiguous. The reallocation of labor resource from farming to non-agricultural self-employment opportunities resulted in income and productivity growth and poverty fell across all groups by MFI participation status. Average work hour, however, did not increase leaving the status quo of pervasive underemployment in the rural economy of Bangladesh. Moreover, both between and within group inequality increased. When we break down the sample of households by the status of eligibility to participate in a microcredit program, we find that eligible participants performed no better than eligible non-participants and non-eligible participants fared far worse than non-eligible 18 non-participants. It appears that the contribution of microcredit program participation lies in helping the households at the lower side of the economic strata to keep up with the rest of the society possessing comparable initial endowments. On the higher side, it appears to even detract from their growth potential. Thus this study finds evidence against the popular belief that microcredit is instrumental to the uplifting of the rural poor to a higher economic status. There is no denying the fact that poverty would have been alleviated at a faster pace if it was accompanied by income and productivity growth with greater employment opportunities and reduced inequality. The findings from this study lead us to enquire why participation in microcredit program is so widespread and persistent despite the insignificant contribution it makes in enhancing the economic wellbeing of the participants in comparison to the nonparticipants with respect to income, productivity and employment growth, and income equalization. It clearly asks for further research to find out whether the contribution of microcredit is limited to the improvement of economic well-being in the short run through consumption smoothing, sans investment in gainful economic activities. 19 References Aghion, Beatriz Armendáriz de, and Morduch, J. 2005. The Economics of Microfinance, MIT Press. Aghion, Philippe, Eve Caroli, and Cecilia Garcia-Peñalosa. 1999. “Inequality and economic growth: The perspective of the new growth theories.” Journal of Economic Literature 37(4): 1615-1660. Burtless, Gary. 1999. “Effects of growing wage disparities and changing family composition on the US income distribution.” European Economic Review 43(4): 853-65. Fields, G.S. 2005. “A welfare economic analysis of labor market policies in the HarrisTodaro model.” Journal of Development Economics 76: 127-146. Hadar, J., and W. Russell. 1969. “Rule for ordering uncertain prospects.” American Economic Review 59: 25-34. Hossain, Mahabub. 1988. “Credit for alleviation of rural poverty: The Grameen Bank in Bangladesh.” Research Report 65, International Food Policy Research Institute, Washington DC. ________. 1991. Labour force, employment and access to income earning opportunities in Bangladesh. Bangladesh Institute of Development Studies, Dhaka. Khandker, S.R. 2003. “Microfinance and poverty: evidence using panel data from Bangladesh”. World Bank Policy Research Working Paper 2945, Washington, DC. McKenzie, David, and Christopher Woodruff. 2003. “Do entry costs provide an empirical basis for poverty traps? Evidence from Mexican microenterprises.” BREAD Working Paper No.020, February. Mahmud, S. 2003. “How empowering is microcredit?” Development and Change: 577605. Rahman, Atiur, et al. 2005. Follow-up Monitoring and Evaluation System Study: Second Poverty Alleviation Microfinance Project. Palli Karm-Sahayak Foundation, April. Sen, A.K., and J. Foster. 1997. On Economic Inequality. Oxford University Press. Saposnik, R. 1981. “Rank dominance in income distributions.” Public Choice 86: 147151. Zohir, S. et al. 2001. Monitoring and Evaluation of Microfinance Institutions. Bangladesh Institute of Development Studies, October. 20 Zohir, S. 2004. “NGO sector in Bangladesh: An overview.” Economic and Political Weekly 39(36), September: 4-10. 21 Notes: 1. 2. All Mean Std Err. N Mean Std Err. N Mean Std Err. N Mean Std Err. N 1998 43,240 988 774 47,369 1159 1,174 55,642 1871 758 48,195 758 2,706 2004 51,537 1387 774 59,162 1470 1,174 73,322 2808 758 60,546 1062 2,706 12,351 17,680 11,793 Growth 8,297 1998 37,543 871 776 36,699 817 1,164 37,520 1239 745 37,033 541 2,683 2004 43,102 1158 773 44,486 1102 1,155 46,431 1879 725 44,208 745 2,648 7,176 8,910 7,787 Growth 5,560 Endogenous labor income 1998 14,406 1288 316 22,653 1516 548 33,256 2244 410 23,701 1011 1,276 N represents the number of sample households that reported household income from any source both in 1998 and 2004. Mean is calculated by dropping the top and bottom 1% observations. Nonparticipants Occasional participants Participation status Regular participants Household income 2004 20,262 1684 344 29,513 1630 587 47,384 2973 451 32,485 1236 1,378 22 8,784 14,128 6,860 Growth 5,856 Exogenous income Table 1: Growth in annual household income (Taka in 2004 constant prices) by MFI participation status and source, 19982004. Mean Std.Err. Mean Std.Err. Mean Std.Err. Occasional participants Nonparticipants All 8,585 49 9,905 121 8,370 70 9,060 57 11,305 155 8,582 72 474 1,400 212 Total annual income per capita 1998 2004 Growth 7,652 7,688 36 65 77 Note: Mean is calculated by dropping the top and bottom 1% observations. Mean Std.Err. Regular participants Participation status 14,559 117 15,870 305 14,103 168 1998 14,400 179 16,220 140 18,739 422 15,644 190 Productivity 2004 15,247 192 848 23 1,661 2,869 1,541 Growth Table 2: Growth in total annual income per capita and productivity (Taka in 2004 constant prices) by MFI participation status, 1998-2004. 0.16 0.16 0.13 0.13 0.46 Wage employment Farming Non-farm agricultural selfemployment Non-agricultural self-employment 0.36 0.32 0.27 Type of employment Occasional participants Regular participants 0.30 0.20 0.18 0.31 0.38 0.16 0.15 0.30 0.53 0.13 0.08 0.27 NonAll Regular participants households participants 1998 0.45 0.16 0.09 0.30 Occasional participants 0.38 0.18 0.12 0.31 0.46 0.15 0.10 0.29 24 NonAll participants households 2004 Table 3: Allocation of total annual household work hours by type of employment, 1998-2004. Table 4: Measures of poverty by MFI participation status, 1998-2004. Participation status Regular participants Change Occasional Participants Change Non participants Change All Year 1998 2004 1998 2004 1998 2004 1998 2004 Change Head count index 76.79 71.70 -5.10 72.88 64.05 -8.83 66.29 57.41 -8.87 72.19 64.43 -7.76 Poverty gap ratio 31.87 28.54 -3.33 31.89 25.73 -6.17 29.28 23.67 -5.61 31.24 25.96 -5.27 Sen index 40.89 36.76 -4.13 40.79 33.38 -7.41 38.14 30.77 -7.37 40.18 33.64 -6.54 FGT(2) 16.47 14.22 -2.25 17.25 13.10 -4.15 16.52 12.42 -4.09 16.88 13.23 -3.65 Notes: 1. The poverty line income in 1998 is Tk.652 per capita per month and in 2004 it is Tk.733 per capita per month (Rahman et.al, 2005). 2. FGT(2) represents Foster-Greer-Thorbecke index of order 2. Table 5: Decomposition of Gini coefficient into between and within group components Gini coefficient All Regular participants Occasional participants Non-participants 1998 2004 0.3781 0.3235 0.3737 0.4242 0.4011 0.3451 0.3679 0.4618 0.1297 (34.3%) 0.0512 (13.5%) 0.1972 (52.1%) 0.1345 (33.5%) 0.0840 (20.9%) 0.1826 (45.5%) Decomposition of Gini coefficient Within Group Between group Residual Notes: 1. The method of decomposition is followed from Sen and Foster (1997), pp.149-156. 2. The percentage contributions of within and between group inequality to the overall Gini coefficient are shown in parentheses below the decomposed Gini coefficients. 25 45,125 914 46,813 2329 45,456 1398 9,441 10,453 10,351 Household income 2004 Growth 44,499 8,157 1416 Endogenous labor income 1998 2004 Growth 31,889 38,456 6,568 825 1217 483 482 28,772 35,959 7,187 718 1039 619 612 26,944 32,951 6,007 1206 1509 333 326 29,499 36,007 6,508 508 695 1,437 1,419 1998 12,705 1320 169 17,371 1841 236 23,604 3161 144 16,802 1105 547 Exogenous income 2004 Growth 15,878 3,173 1852 192 23,417 6,045 2322 273 28,072 4,467 3275 189 21,903 5,101 1388 650 Participation status Regular participants 26 Household income Endogenous labor income Exogenous income 1998 2004 Growth 1998 2004 Growth 1998 2004 Growth Mean 56,527 64,538 8,011 47,160 51,265 4,104 18,763 26,992 8,229 Std Err. 2542 3164 1836 2354 3150 3284 N 293 292 292 147 154 Occasional Mean 61,981 74,877 12,896 45,618 54,482 8,864 26,894 35,346 8,451 participants Std Err. 2206 2673 1466 2113 2273 2290 N 542 538 538 309 310 NonMean 71,799 97,739 25,940 45,699 57,710 12,012 39,305 62,889 23,584 participants Std Err. 2940 4906 1926 3190 3138 4533 N 410 410 397 269 262 All Mean 63,913 79,954 16,042 45,975 54,746 8,771 29,508 43,588 14,080 Std Err. 1493 2101 992 1485 1618 2073 N 1,244 1,240 1,226 722 726 Notes: 1. N represents number of households. 2. Eligibility is defined by the ownership of less than or equal to 50 decimals of cultivable land. Table 6b: Growth in annual household income by MFI participation status and source for non-eligible households as of 1998. All Nonparticipants Occasional participants Mean Std Err. N Mean Std Err. N Mean Std Err. N Mean Std Err. N 1998 36,342 893 483 35,105 956 628 36,360 1707 347 35,684 628 1,457 Table 6a: Growth in annual household income by MFI participation status and source for eligible households as of 1998. Participation status Regular participants All Nonparticipants Occasional participants Regular participants Participation status Mean Std.Err. N Mean Std.Err. N Mean Std.Err. N Mean Std.Err. N 1998 13,402 210 1138 12,660 188 1394 13,014 390 675 12,886 131 3205 Eligible households 2004 14,892 227 1,242 14,498 213 1507 14,471 395 710 14,615 147 3461 1,729 1,457 1,839 Growth 1,490 Non-eligible households 1998 2004 16,010 16,092 336 357 840 915 15,527 17,204 261 320 1552 1706 18,321 22,087 471 718 983 1,024 16,320 18,069 192 245 3366 3648 Table 7: Growth in annual household income by MFI participation and eligibility status of households as of 1998. 82 27 1,749 3,766 1,676 Growth Annual household income (2004 prices) 0 50000 100000 150000 200000 250000 300000 350000 400000 0 10 20 30 50 60 1998 2004 Percentile of sample households 40 70 Figure 1a: Shifting of household income distribution, 1998-2004. 80 90 100 28 Growth in annual household income (2004 prices) 0 20000 40000 60000 80000 100000 0 10 20 30 50 60 Percentile of sample households 40 70 80 Figure 1b: Growth in household income (Taka in 2004 constant prices), 1998-2004 90 100 29 Growth in annual household income (2004 prices) 0 20000 40000 60000 80000 100000 120000 140000 160000 0 10 20 30 50 60 Regular Occasional 70 Non-participants Percentile of sample households 40 80 90 100 Figure 2a: Growth (Taka in 2004 constant prices) in annual household income by participation status, 1998-2004. 30 0 4000 8000 12000 16000 20000 24000 0 15 Regular 45 Occasional Non-participants Percentile of sample households 30 60 75 31 Figure 2b: Growth (Taka in 2004 constant prices) in annual household income at the bottom 75 percentiles by participation status, 1998-2004. Growth in annual household income (2004 prices) Annual average hours worked 0 500 1000 1500 2000 2500 1,584 Wage employment 1,698 451 Farming 634 303 1998 2004 Type of employment Non-farm agricultural self-employment 297 Non-agricultural selfemployment 1,775 2,059 All sectors 1,289 1,285 Figure 3: Annual average hours worked by individuals by types of employment, 1998-2004. 32 0 500 1,000 1,500 2,000 2,500 Wage employment 1,627 1,613 446 Farming 589 298 1998 2004 Type of employment Non-farm agricultural self-employment 279 (a) Regular participants Non-agricultural selfemployment 1,863 2,097 All sectors 1,412 1,424 Figure 4: Annual average hours worked by individuals by types of employment and MFI participation status, 1998-2004. Annual average hours worked 33 Annual average hours worked 0 500 1,000 1,500 2,000 2,500 1,595 Wage employment 1,740 448 Farming 629 306 1998 2004 Type of employment Non-farm agricultural self-employment 296 (b) Occasional participants Non-agricultural selfemployment 1,722 2,050 All sectors 1,291 1,265 34 Annual average hours worked 0 500 1,000 1,500 2,000 2,500 1,527 Wage employment 1,710 460 Farming 690 304 1998 2004 Type of employment Non-farm agricultural self-employment 316 (c) Non-participants Non-agricultural selfemployment 1,722 2,006 All sectors 1,149 1,154 35 Share of total annual household income 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 Line of equality 0.5 0.6 Share of population 0.4 0.7 0.8 0.9 2004 1998 Figure 5: The Lorenz curves of per capita income in 1998 and 2004. 1 36 Share of endogenous labor income 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 Occasional participants 0.2 0.3 0.5 0.6 Share of population 0.4 Regular participants Regular participants 0.7 0.8 Occasional participants Nonparticipants 0.9 2004 1998 Nonparticipants Figure 6: The Lorenz curves of labor productivity in 1998 and 2004. 1 37