Master Thesis: Does Micro-Finance Matter for the Poor? Evidence from Indonesia. N.Nuruzzaman ANR

Transcription

1 Master Thesis: Does Micro-Finance Matter for the Poor? Evidence from Indonesia N.Nuruzzaman ANR Master Thesis submitted to obtain the degree of: Master of Science (MSc) in Economics Supervisor: Prof. A.C. Meijdam Second Reader: Dr. H. G. van Gemert Tilburg University Tilburg School of Economics and Management 2010

2 Abstract This study scrutinizes the link from micro-small-medium credit to poverty and income inequality. In a broader sense, this research can be viewed as an attempt to provide empirical evidence on how financial development affects growth, poverty, and inequality. This research employed data micro-finance and poverty data from 319 regencies in Indonesia at Using distance to Jakarta as the Instrumental Variable for micro-small-medium credit, the author finds that micro-small-medium credit negatively affect the poverty rate and positively affect income inequality. After decomposing micro-small-medium credit and including dummy variable for region, the author finds that micro-credit is negatively correlated to poverty index and income inequality. Small-credit has no effect on both poverty and income disparity. Meanwhile, medium-credit is negatively correlated to poverty index but positively correlated to income inequality. At the end of this research, the author proposes several policy implications from the empirical evidence. Keywords: Micro-Small-Medium Credit, Micro-Finance, Poverty, Income Inequality, Instrumental Variable.

3 Table of Content Introduction Background 1 Research Statement 4 Differentiation of This Research 5 Structure of the Paper 5 Literature Review Financial Development and Poverty 6 Financial Development and Inequality 7 Empirics on Financial Development, Poverty, and Inequality 9 Methodology and Data Methodology 14 Data 20 Result Preliminary Evidence 22 Evidence Using Instrumental Variable 23 Extension of the Research 33 Conclusion and Policy Implication Conclusion 40 Policy Implication 42 Suggestion for Future Research 42

4 Figures, Tables and Boxes Figures Figure 1.1 Number of Population below Poverty Line & Headcount Index Figure 1.2 Micro-Credits and Poverty in Indonesia Tables Table 2.1 Summary on Empirics of Finance-Poverty-Inequality 12 Table 4.1 Preliminary Evidence on Finance-Poverty-Inequality 22 Table 4.2 Regression of Distance to Jakarta on MSM-Credit per Capita 23 Table 4.3 Instrumental Variables on Each Endogenous Explanatory Variable 24 Table 4.4 Correlation Coefficient among Explanatory Variables 25 Table 4.5 Estimation Result Model (i) after Instrumenting Explanatory Variables 26 Table 4.6 Estimation Result Model (ii) after Instrumenting Explanatory Variables 28 Table 4.7 Estimation Result Model (iii) after Instrumenting Explanatory Variables 30 Table 4.8 Result after Decomposition 34 Table 4.9 Estimation Results on Regional Differences 36 Boxes Table 4.10 Decomposition and Interaction Terms 37 Box 2.1 Dynamics of Capital Accumulation and Income Inequality 8 Box 3.1 Absolute VS Relative Poverty Line 18 Box 3.2 The Lorenz Curve and Gini Coefficient 19

5 Introduction The Mother of Revolution and Crime is Poverty Aristotle ( BC) a. Background Poverty might be the most serious problem in the world, from the early age of humankind until current generation. One thing that makes poverty become the main concern is the non isolated impact of poverty. As Aristotle asserted, poverty could be the root of other social problems. Thus, alleviating poverty might solve other social problems such as crime and prostitution. In Indonesia, as in other developing countries, poverty is the main political issue. Every politician will exploit this issue in his/her campaign. This phenomenon is to be expected. There are more than 32 million people being under poverty line in 2009 (BPS, 2010). This is counted as 14.51% of the population in Indonesia. Attracting this group will certainly give an advantage in the national politics. However, based on the data released by Badan Pusat Statistik (Statistic Bureau of Indonesia), poverty in Indonesia has been declining significantly for the last decade. In 1999, the headcount index was 23.55%, which has dropping off as much as 9.04% for a decade. The following figure will exhibit number of population below poverty line and headcount index of poverty in Indonesia. Figure 1.1 Number of Population below Poverty Line and Headcount Index ,43 % 19,14 % 18,41 % 18,20 % 17,42 % 16,66 % 15,97 % 17,75 % 16,58 % 15,42 %14,15 % Headcount Index (%) Number of Population below Poverty Line (in Million) Source: Badan Pusat Statistik,

6 From the above figure, it is obvious that both number of poor people and headcount index are showing declining trend over the last decade in Indonesia. Any line of reasoning could be possible to explain this, and perhaps none of it was wrong. In general, academia in economics field offers two approaches to answer that question, which are financial and non financial access. With respect to financial access, both theoretical and empirical researches have not reached any consensus to confirm whether increasing financial access would eradicate poverty. The theoretical approach generates an ambiguous prediction. On one side, finance can increase productivity by reducing information asymmetries and transaction cost (Levine, 2005). On the other side, financial development can slow down the saving rates, thus deliver poor economic growth (Gurley and Shaw, 1955). Meanwhile, the empirical researches using cross country data do not give convincing evidence due to measurement error problem, reverse causation, and omitted variable bias (Beck, 2008). However, increasing attention has been paid to micro, small, and medium credit (MSM credit). The case of Grameen Bank in Bangladesh or Bank Rakyat Indonesia in Indonesia can be used as the grounds on exploring the relationship between MSM credit and poverty. The provision of MSMcredit could occur from both formal and informal institutions. Grameen Bank in Bangladesh is the example of MSM credit that being provided by informal institution. Meanwhile, MSM credit of Bank Rakyat Indonesia is offered by formal institution. Grameen Bank is always associated with the early history of micro credit. Founded by Muhamad Yunus, Grameen Bank is a project to provide loans to the poor in Bangladesh. The bank focuses on women and would lend only to the poorest of the poor in the rural areas (Bakhtiari, 2006). The collateral is not needed to apply for a loan. In addition, the bank would also support the poor borrowers in their business. This strategy was claimed as a proven approach on alleviating poverty. The other success story of micro finance takes place in Indonesia. Bank Rakyat Indonesia (BRI) is a state owned bank that is firstly began as agriculture bank in rural areas. Currently, BRI is known as a bank with largest network that can reach the bottom level of the society. It has more than 3500 local units and serves more than 3 million borrowers (Bakhtiari, 2006). Bakhtiari (2006) stated that BRI is the evidence of profitable microfinance intermediary. Although they are different in the sense that Grameen Bank is informal institution while BRI is formal institution, but both banks receive growing attention for the contention that they have been substantiated on eradicating poverty in rural areas. The BRI phenomenon in Indonesia should stimulate the researchers to elaborate the relationship between micro credit and poverty based on data gathered, particularly aggregate data. If we take a look at rough data from Central Bank of Indonesia (BI), it is expected that development of MSMcredit has a correlation with poverty level. The following figure will exhibit the trend of MSM credit in Indonesia, and will compare it with the poverty data. Clearly we could observe that from , the growing of MSM credit is followed by the decreasing trend of poverty, although after that 2

7 poverty and micro credit are following the same pattern. It is also obvious to notice that private credit and poverty followed the opposite direction. Figure 1.2 Micro Credits and Poverty in Indonesia Number of Population below Poverty Line (In Million) Private Credit per GDP (%) Headcount Poverty Index (%) Micro Credit per GDP (%) Source: Badan Pusat Statistik and Bank Indonesia, On the ground of the presumption mentioned above, this research was designed. It was inspired by a very simple question; can we generalize the effectiveness of MSM credit on alleviating poverty in the aggregate level? Moreover, this inquiry basically has its roots on the finance poverty nexus. Questioning the effectiveness of MSM credit on alleviating poverty is parallel to questioning the mechanism on how financial development will benefit the poorest member of the society. The line of reasoning is straightforward. The level of MSM credit implies the outreach of financial system. Ergo, this will indicate the degree of financial development. In addition, this research tries to assess the ability of MSM credit on altering the income distribution. The link between MSM credit and distribution of income is an important subject matter. As Bourguignon (2004) argued that growth might have very limited impact to poverty if growth failed to correct income distribution. Thus, the significance of financial development should be viewed not only on its ability to promote growth, but also how it affects poverty and inequality. 3

8 This research has both theoretical and policy implications. On the theoretical side, this research is expected to provide evidence on the debate of how financial development affects poverty and inequality. On the practical point of view, this research is also expected to evaluate government policy in encouraging the growth of micro finance in Indonesia. However, the implication might be very limited to Indonesia economy and could not be generalized to other countries. b. Research Statement This research prompts two main questions. First question is how financial development, in the form of micro credit, affects poverty. Second inquiry then questions the effect of financial development to income distribution. The link between financial development and poverty does not have to be in line with the link between financial development and inequality. Investigating both channels will generate better understanding on how financial development works in the economy. Thus, there are three main variables involve in this research. They are MSM credit, poverty, and inequality. This research is mainly an empirical research utilizing 4 years data from 319 regencies in Indonesia. The proposed methodology to test several hypotheses is cross section instrumental variable. This methodology is expected to solve measurement error, reverse causality and unobserved variable bias. Thus, instrument is needed in this research. Methodology and Data section will provide the instrument employed and the grounds to exercise that instrument. The result of this research is quite unpredictable. This is due to the ambiguous theoretical predictions. On one side, some economists argued that financial development in general and micro finance in particular could reduce poverty and correct income distribution. On the other side, financial development is expected to hinder the poverty alleviation and worsen income distribution. Specifically, this research uses MSM credit as the main variable of interest. The main distinction between MSM credit and non MSM credit is the maximum amount of credit being provided. Micro credit is credit that less than Rp. 50 million, small credit is credit between Rp. 50 million and Rp. 500 million, and medium credit is credit from Rp. 500 million to Rp. 5 billion. The small amount of this credit reflects the collateral of the debtor. The lower the former implies the lower the latter. This indicates that MSM debtor is a micro to medium company which has under average size of initial wealth endowment. The other characteristic is that the credit is also provided together with business coaching from the bank. The data of MSM credit is obtained via website of Central Bank of Indonesia. However, there is a limitation of using this dataset. This dataset records only micro, small, and medium credit that provided by formal financial institution. The provision of micro credit from informal institution is excluded from the dataset. If informal financial sector plays an important role in the economy, then this might affect the result. Therefore, the expected outcome cannot be generalized to financial system as a whole, but only to formal financial system. 4

9 c. Differentiation of This Research Hitherto, there are abundance research papers that investigate finance poverty nexus, either theoretical or empirical. Beck, Demirguc Kunt, and Levine (2007) proved that financial development has been benefited the poor by increasing the poor s income and this, in turn, decreases the rate of poverty. Using fixed effect vector decomposition (FEVD), Akhter and Daly (2009) indicate that financial development is conducive for poverty alleviation, although the instability of financial sector is detrimental to the poor. This research differs from cross country empirical researches mentioned above as this research utilizes single country as an object of research. This brings some advantages. First, this research will not be suffered from the (financial) policy distortion that might affect the result. This is due to the single banking policy in Indonesia, which is governed by Central Bank of Indonesia. In addition, there is no discrepancy in legal system across regencies in Indonesia. Second, there are more than 400 regencies in Indonesia. This number is quite large and therefore will enable researcher to conduct empirical research by exploiting a very extensive dataset. In addition, this research also differs from Geda, Shimeles, and Zerfu (2006) and Hiatt and Woodworth (2006) who conducted micro study to investigate the link between finance and poverty. This research does not exploit micro dataset. Thus, in contrast with Geda, Shimeles, and Zerfu (2006) and Hiatt and Woodworth (2006), through which channel microfinance able to alleviate poverty is not main concern in this research. However, as mentioned above, the outcome of this research can only be generalized in Indonesia. It might not be wise to generalize the result of this research in a broader scope. More detailed on finance poverty inequality researches, both theoretical and empirical, are discussed in next section. d. Structure of the Paper This thesis will consist of five sections. First section is introduction. Second section is literature review, which offers detail explanation on the theory of finance poverty and empirical evidence on that issue. Third section will discuss empirical method to test the hypothesis as well as econometric models and data sources. The fourth section will reveal the result from empirical estimation. In addition, the extension of this research will also be presented in this section. Fifth section will draw conclusion of this research and propose several policy recommendation regarding the issue of finance povertyinequality. 5

10 Literature Review The banker, therefore, is not so much primarily middleman... He authorizes people in the name of society. Joseph A. Schumpeter ( ) This research prompts a main question on whether or not financial development, in the form of micro financing, has been able in alleviating poverty and correcting income inequality. Instead of providing theoretical explanation on finance poverty nexus, this research is intended to come up with the evidence. However, several theoretical predictions are being presented to give big picture on how we might expect the link between finance and poverty. This section consists of three sub sections. First, it will discuss financial development and poverty. Second, theoretical explanation on how financial development affects income distribution will be examined. The last sub section will summarize the empirical evidence on financial development and poverty. a. Financial Development and Poverty There are two conflicting predictions regarding how financial development affects the poor. On one side, some theorists assert that financial development will have detrimental impact on the poor. On the other side, some economists claim that financial development will benefit the poor. This section will elaborate these two competing arguments. The contention that claims financial development has unfavorable effect on the poor, can be traced back to Banerjee and Newman (1993), Galor and Zeira (1993), and Aghion and Bolton (1997). By developing a non linear model of development process, Banerjee and Newman (1993) demonstrate that capital market imperfections will cause the poor to work for a wage rather than self employment. This implies that the poor has a limited ability to exploit investment opportunity due to informational problem in the capital market. Following the same line, Galor and Zeira (1993) assert that capital market imperfections deteriorate the capacity of the poor to invest in education. Thus, capital market may foster economic growth, but it will depend on initial wealth endowment. This denotes the appearance of multiple equilibria on how capital affects economic growth. The contention of these two concepts lies on the idea of asymmetric information in the capital market. The allocation of credit in the capital market suffered from both ex ante adverse selection and ex post moral hazard. Only prospective debtor has full information on the project. Meanwhile, the creditor (the bank) has limited information as well as limited ability to reveal the proper value of business venture. The screening mechanism, which banks build up to separate bad entrepreneurs from good 6

11 entrepreneurs, is through collateral provision. Consequently, only entrepreneurs with high wealth endowment can provide high value collateral. Credit, hence, is being allocated based on initial wealth endowment, not on investment opportunity. Ergo, wealth deficient entrepreneurs are being excluded from capital markets. In addition, Aghion and Bolton (1997) confirm the claim that poor entrepreneurs are being discouraged to borrow funds from capital markets. Aghion and Bolton (1997) find that as effort supply diminished when entrepreneur borrows more, the unit repayment must be commensurately escalated to guarantee that lender will obtain the equal expected repayment. This in turn, will depress wealth deficient entrepreneurs on gaining advantage over capital market because the poorer the entrepreneurs, the higher his/her repayment. On the other side, some theorists argue that better functioning financial system will have positive effect on the poor. By ameliorating information and transaction cost, financial development expands the allocation of capital, benefited larger portions of population, including the poor (Levine, 1997). Durnev, Morck, and Yeung (2001) argue that capital markets create incentives for investor to gather information on projects return, hence allocate capital investment efficiently. Moreover, specifically on micro finance, theorists argue that financial development in the form of micro credit is able to foster the capability of the wealth deficient entrepreneurs. There are at least five mechanisms through which micro credit benefited the poor. First, micro finance helps the poor increasing the income through intensifying production capability and protecting against income risk (Morduch and Haley, 2002). Second, Simanowitz and Walter (2002) find that the clients of micro finance institution experience consumption smoothing and increasing in ability to sustain over time through risk diversification. In addition, through micro finance the poor could have better access on nutrition (Swope, 2005). Access to micro finance in rural areas could help the poor on maintaining their consumption level in the hard time. Thus, stable consumption pattern will increase the probability of the poor to attain better nourishment. This in turn, will increase productivity of the poor. Fourth, access to finance will help the poor to gain access on education (Morduch and Haley, 2002). From dynamic perspectives, this will facilitate the poor to leave vicious circle. The last mechanism works through empowering the woman in rural areas. Women in rural areas are being prioritized to get the loan from micro finance institution such as Grameen Bank. Increasing the financial access for women could broaden the source of production capability in the society (Simanowitz and Walter, 2002). b. Financial Development and Inequality In line with financial development and poverty, the theoretical link between finance and inequality is vague. In one side, we might expect that financial development able to correct income distribution to be more equal. In the other side, financial development could worsen income inequality due to capital 7

12 market imperfections. This section provides conceptual review of the link between financial development and income inequality. Galor and Zeira (1993) argue that credit constraint, which is inherited in financial development process, will impede the poor to gain advantages from human capital investment. The access to financial assets will be influenced by the initial wealth of the agents. This implies that only wealthy agent in the economy able to invest in human capital through education. The poor, who inherits low level of initial wealth endowment, could not attain credit due to high cost of borrowing. This in turn, limits the poor to invest in education. Thus, as long as credit constraint is binding, the financial development will be in favor of wealthy agent. In this scenario, the financial development will worsen income distribution. The work of Galor and Zeira (1993) also initiate the possibility of multiple equilibria in the relationship between capital accumulation and economic growth. Box 2.1 will illustrate the general idea of Galor and Zeira (1993). Box 2.1 Dynamics of Capital Accumulation and Income Inequality Where: X t+1 A X t = wealth of current generation X t+1 = wealth of subsequent generations B Xn and Xs = the wealth which are being inherited in the long run. A = long run equilibrium of wealthy agents Xn f g h Xs X t B = Long run equilibrium for wealth deficient agents Agents who inherit more than f but less than g will invest in human capital but not all their subsequent generations will continue in the skilled labor sector in the future. After some generations their descendants turn back into unskilled workers and their inheritance will converge to Xn in the long run. Agents who inherit more than g in period t could invest in human capital. Thus, their descendants will be able to gain skill and become skilled workers. This lasts form generation to generation. This implies that this group can maintain their inheritance in level Xs in the long run. In other words, this group is able to maintain high level of wealth in the long run. Ergo, there are two groups in the economy in the long run. The first group is the wealthdeficient group that has limited ability to invest in education and thus stay poor in the long run. The second group is dynasty that is able to invest in education, be skilled workers, and continue being rich in the long run. The wealth deficient dynasty remains poor because this group has no access to credit due to capital market imperfections. Source: Galor and Zeira (1993) 8

13 In addition, Galor and Moav (2004) assert that financial development will be the source of income inequality if the primary source of economic growth is physical capital accumulation. In this contention, Galor and Moav (2004) argue that although it might have positive impact on economic growth, but financial development will enlarge the income disparity because financial development will enhance development by routing resources to individual whose marginal propensity to save is higher. However, in the later stage of development, human capital accumulation offsets the inequality effect of financial development and thus narrows down the income gap between the poor and the rich. Ergo, based on this proposition, financial development is expected to worsen income inequality if physical capital acts as engine of growth in the earlier stage of development. Other predictions come from Claessens and Perotti (2007). Reviewing evidence on the relationship between finance and inequality, Claessens and Perotti (2007) exhibit that there is unequal access to financial asset, which leads to economic inequality. This economic inequality, in turn, will cause infeasibility of financial reform. Thus, vicious circle of development might be expected due to initial wealth inequality that generates unequal access to financial assets. Moreover, Claessens and Perotti (2007) find a threshold for the development of financial institution that can guarantee the accomplishment of financial reform. On the other side, if financial development is able to facilitating allocation of capital, monitoring investment, ameliorating risks and mobilizing savings (Levine, 2004), then it would be able to select wealth deficient entrepreneurs with excellent investment opportunity. If financial development benefits the poor, then it should be able to correct income disparity. However, this claim assumes perfect information in capital market. Nevertheless, Greenwood and Jovanovic (1990) propose inverted U shape curve on the relationship between financial development and inequality. This proposition follows the idea of Kusnetz (1955) hypothesis regarding the process of economic development. Greenwod and Jovanovic (1990) argue that at the early phase, financial development widen the income gap as financial resources being allocated to wealthy agents who needs liquidity service. When financial structure becomes more settled, financial development has higher capability to distribute the resources equally, hence cut down the income disparity. Greenwood and Jovanovic (1990) argument is also helpful to explain the idea behind Kusnetz curve. Grenwood and Jovanovic (1990) contention generates the idea that the phase of financial development is crucial to clarify the link between inequality and economic growth. c. Empirics on Financial Development, Poverty, and Inequality In contrast to previous section, which reviews several theoretical works, this section will provide empirical evidences on finance, poverty, and inequality. The explanation will be divided into two subsections. First sub section will briefly explain cross country empirical studies. Second sub section will elaborate country specific studies. 9

14 Cross country Studies Early cross country studies on finance, poverty, and inequality was conducted by Beck, Demirguc Kunt, and Levine (2007). Using legal origin and geographical latitude as instrumental variables to deal with reverse causality bias and measurement error, they construct econometrics model to test whether or not financial development narrow down income gap and eradicate poverty. Beck, Demirguc Kunt and Levine (2007) find that countries with higher private credit per GDP enjoy faster decline on poverty rates and income inequality. This implies that better functioning financial system will enable country to raise income of the poor disproportionately, compare to the average income, which results in higher level of poverty alleviation and lower level of income inequality. Second cross country study on financial development and poverty is a research conducted by Akhter and Daly (2009). Using panel data of 54 countries from 1993 to 2004, Akhter and Daly (2009) undertake the research to clarify the link between financial development and poverty reduction. In their research, Akhter and Daly (2009) employ fixed effect vector decomposition (FEVD). They find that financial development contributed to poverty reduction around the world. However, they also find that financial development brings instability that can be detrimental to the poor. Both of these two studies measure the effect of financial development on poverty and income disparity in aggregate level. Hence, both of these studies did not measure the access to finance, micro aspect of finance and poverty. From both studies we can infer that financial development is able to alleviate poverty, but through which channel it alleviates poverty has not been revealed yet. However, both studies can be generalized due to the fact that these two studies utilize data from multi countries as the object of research. On the other side, Cull, Demirguc Kunt, and Morduch (2008) conducted cross country study, but using micro level data. Exploiting dataset from 346 world leading micro finance institutions, Cull, Demirguc Kunt and Morduch (2008) find that profit maximizing investor that invest their capital in micro finance institutions are less interested on allocating loan to poorest member of society as well as women. Hence, profit maximizing micro finance institutions charge higher interest on loan to cover high transactional cost as transaction unit becomes smaller. These evidences indicate that micro finance does not fully address informational problem in capital market, and thus contradict the results of previous studies. Country specific Studies Tam (1988) provides case study on rural finance in China during early period of liberalization. Tam (1988) finds that there has been significant progress on china s rural financial institution after the period of liberalization. Although distortions still appear, mainly through government intervention, financial liberalization brings advantage to rural entrepreneurs by providing easy access to credit. This in turn facilitate rural entrepreneur to lift out the poverty line. 10

15 In Indonesia, Panjaitan Drioadisuryo and Cloud (1999) could have evaluated the government program for the poor called Small Farmer Development Program (SFDP) and micro credit from Bank Rakyat Indonesia (BRI), a state owned commercial bank in Indonesia. They observe the evidence that SFDP along with micro credit from BRI has been successful on reaching poor women in the society and provide them with credit and training. Both credit and business training enable poor woman to boost their income through developing new business such as trading goods, producing handicraft, and weaving traditional cloth. In addition, Panjaitan Drioadisuryo and Cloud (1999) also find that microcredit has multiplier effect to the poor as micro credit enable poor family to attain better nutrition and repay family loans. Ergo, these findings confirm the positive impact of micro finance to poverty reduction. Another micro econometrics study conducted in Vitenam s rural credit market. Using several methodologies, they examine both formal and informal credit markets. They found that productionspecializes entrepreneur tend to demand credit from formal financial institution, whereas the clients of informal financial institution are more diverse. They also observe that the reputation, dependency ratio of household, and the amount of credit application determine the credit rationing by rural bank. Nevertheless, Duong and Izumida (2002) draw conclusion that elasticity of output to credit was very high. This result implies that financial access will have positive impact on productivity and poverty alleviation. Burgess and Pande (2005) conducted econometric study in India, which attempts to investigate the effect of banking deregulation on poverty reduction. After controlling for health and development spending, fraction of legislator, state and year dummy, Burgess and Pande (2005) observe that opening bank in rural unbanked area is associated with poverty alleviation in India. In addition, Burgess and Pande (2005) found that reduction in rural poverty is related to increasing saving mobilization and credit provision in rural areas. The fifth country specific study to be discussed is research conducted by Geda, Shimeles, and Zerfu (2006). Exploiting household panel data in Ethiopia, they asses the finance poverty nexus. After controlling for endogeneity variables, they show that financial access is a crucial factor in consumption smoothing of the poor. Thus, finance does matter to alleviate poverty in Ethiopia. Moreover, Geda, Shimeles, and Zerfu (2006) observe the evidence of poverty trap due to liquidity constraints. They also propose policy maker to extend the access to finance especially in the rural areas. Hiatt and Woodworth (2006) perform the analysis of micro finance, provided by three Central American NGOs, on poverty alleviation. They observe the increasing income of new micro credit participants since they started participating. Moreover, Hiatt and Woodworth (2006) also observe that entrepreneurs who stayed in the program earned higher income than entrepreneurs who left the program. This evidence confirms the ability of micro credit to assist the poor through raising their production capacity. 11

16 Maldonado and Gonzales Vega (2008) detect the mixed evidence on the relationship between microfinance and schooling gaps. Utilizing data from survey of households that participate in micro finance program in Bolivia, they discover that on one side micro finance reduces schooling gaps due to increasing ability of poor household to send their children to school. However, on the other side, microfinance increases production capacity of poor households and thus, increases their demand on child labor. This evidence indicates that micro finance might narrow down the schooling gaps as long as wealth deficient entrepreneurs, who participate in micro finance program, can find substitute of child labor. Conditional on this prerequisite, micro finance has positive impact on poverty alleviation program. Liverpool and Winter Nelson (2010) offer more detail research from Ethiopia, which decompose the poverty into several level. In addition, the link from which micro finance affects poverty is being examined. Liverpool and Winter Nelson (2009) find that the impact of micro finance differs across group of the poor. For the poorest group, there is no relationship between participation in micro finance and the use of technology. Meanwhile, for the other group, participation in micro finance program has positive effect on the use of technology. However, Liverpool and Winter Nelson (2009) find that participation in micro finance has positive impact on consumption and asset growth for both groups. This research clarifies the assertion that micro finance is able to alleviate poverty through the use of technology, consumption smoothing, and asset growth, although the magnitude of the effect differs across groups. However, Coleman (2006) doubts the effectiveness of micro finance on alleviating poverty. Coleman (2006) investigates the performance of micro finance on poverty alleviation in Thailand. Coleman (2006) finds that wealthier villagers are more likely to participate in micro finance program than less wealthy villagers. Coleman (2006) also observes that the wealthiest members of the village society often become committee on selecting the loan applicants. The latter evidence could explain the reason why microcredit usually goes to wealthier entrepreneurs, instead of being allocated to poorest villagers. This evidence implies that micro credit does not fully solve asymmetric information problem in the capital market. If this is valid, then the ability of micro finance to eradicate poverty is being doubted. The following table will present summary of empirical research in the field of financial development/micro finance poverty inequality. Table 2.1 Summary on Empirics of Finance Poverty Inequality No Author(s) Result 1 Beck, Demirguc Kunt, Levine (2007) Cross Country Studies Better functioning financial system will enable country to raise income of the poor disproportionately, which results in higher level of poverty alleviation and lower level of income 12

17 inequality. 2 Akhter and Daly (2009) Financial development contributes to poverty reduction around the world, although it brings instability that can be detrimental to the poor. 3 Cull, Demirguc Kunt, and Morduch (2008) Profit maximizing micro finance institutions charge higher interest on loan to cover high transactional cost. Thus, microfinance does not fully address informational problem. Country Specific Studies 1 Tam (1988) Financial liberalization in China brings advantage to rural entrepreneurs by providing easy access to credit. 2 Panjaitan Drioadisuryo and Cloud (1999) Confirms the positive impact of micro finance to poverty reduction in Indonesia through assisting new business, obtaining better nutrition, and repaying loans. 3 Duong and Izumida (2002) Elasticity of output to credit was very high. Ergo, financial access will have positive impact on productivity and poverty alleviation in Vietnam. 4 Burgess and Pande (2005) Opening bank in rural unbanked area is associated with poverty alleviation in India. 5 Geda, Shimeles, and Zerfu (2006) Financial access is an important factor in consumption smoothing of the poor in Ethiopia. 6 Hiatt and Woodworth (2006) Confirms the ability of micro credit in Central America to assist the poor through raising their production capacity. 7 Maldonado and Gonzales Vega (2008) Mixed evidence on the relationship between micro finance and schooling gaps in Bolivia. On one side micro finance increases ability of poor household to send their children to school. On the other side, increasing production capacity leads to increasing the demand on child labor. 8 Liverpool and Winter Nelson Clarifies that micro finance in Ethiopia is able to alleviate (2010) poverty through the use of technology, consumption smoothing, and asset growth, although the magnitude of the effect differs across groups of poor. 9 Coleman (2006) Wealthier villagers in Thailand are more likely to participate in micro finance program than less wealthy villagers. Micro credit usually goes to wealthier entrepreneurs, instead of being allocated to poorest villagers. This evidence implies that microcredit does not fully solve asymmetric information problem. 13

18 Methodology and Data The Poor will always be with you. John 12: 8 This section will present the methodology employed and data used in this research. As mentioned in the first section, instrumental variable will be utilized to obtain precise estimate on finance poverty nexus. The first part will describe the purpose of applying instrumental variable, conditionals required to be a proper instrument, econometric models and the hypothesis. The second part will briefly explain the nature of data and the sources of data. a. Methodology This subsection will explain the econometric issue and how to overcome several econometrics problems. Furthermore, this subsection will present the model and elaborate the main hypothesis. Econometric Method To test the relationship between finance and poverty, one can employ a simple regression model, OLS. However, OLS will suffer from biases, which are omitted variable bias, reverse causality bias, and measurement error (Beck, 2008). All these biases will hinder the result of estimation. This implies that researcher cannot fully depend on the result as it will be less precise and accurate. Omitted variable bias arises in the model where relevant variable is being excluded from the model. If one relevant variable being excluded from the model, then the estimation will treat the beta estimate of omitted variable to independent variable as fixed when calculating expected beta estimate (Wooldridge, 2009). This will lead to bias in the beta of variable of interest as expected beta estimate will not be equal to actual beta. Reverse causality bias appears in the model as financial development might depend on the level of poverty. It is possible that the bank s decision to open branch or expand the business to the place where level of poverty is lower or higher. This bias will cause the endogenous variable to be correlated to the error term. In turn, this correlation will impede the true relationship between endogenous variable and dependent variable. The other problem is measurement error. Measurement error problem arises when the researcher cannot collect true data on the variable that affects economic behavior (Wooldridge, 2009). If measurement error appears in the explanatory variable, then the estimator will be biased and inconsistent as the error term is correlated to endogenous explanatory variable (Wooldridge, 2009). 14

19 In the field of finance growth nexus, which also can be applied to finance and poverty, omitted variable bias, reverse causality bias and measurement error can be overcome by several methodologies. Beck (2008) mentions some useful approaches such as cross section instrumental variables, dynamic panel approach, time series econometrics, and difference in difference estimate. Although they still contain several problems, but those methodologies at least able to reduce bias and thus, enable researcher to obtain better estimate. This research is using cross section instrumental variable to obtain accurate estimation. The reason is that dynamic panel approach and time series require a very large dataset (Beck, 2008). Meanwhile, difference in difference estimate can be employ if natural experiment is available. Thus, cross section instrumental variable is the main method that feasible to be employed in this research. The following section will explain the mechanics of cross section instrumental variable and what is the proper instrument for this research. Instrumental Variable Instrumental variable is being useful to solve the above three biases. It can overcome those biases by separating the component of endogenous explanatory variable that is correlated to error term from the component that is not related to error term (Beck, 2008). In addition, instrumental variable recognizes the presence of omitted variable (Wooldridge, 2009). To acquire consistent estimator, additional variable is needed. This new variable, which is called instrumental variable, is expected to convey additional information. However, this new variable should satisfy two assumptions. First, instrumental variable is uncorrelated with error term. Second, this instrumental variable is correlated with endogenous explanatory variable that is being instrumented (Wooldridge, 2009). There is no technique to test the first assumption, which requires instrumental variable to be uncorrelated with error term. Instead of using statistical technique, the first assumption should be verified theoretically (Verbeek, 2008). This implies that instrument variable is uncorrelated to error term by construction. However, The second assumption can be tested by assessing the correlation coefficient between instrumental variable and endogenous explanatory variable. In the case that instrumental variable is weakly correlated to endogenous explanatory variable; substantial bias will be appeared in the IV estimator (Wooldridge, 2009). In this setting, OLS is more useful than the IV method as IV estimator will be less efficient than OLS estimator. In turn, this poor instrumental variable will thwart the consistency of the result (Wooldridge, 2009). Ergo, the instrumental variable should be employed to deal with omitted variable bias, reverse causality bias, and measurement error problem. The instrument should satisfy two criteria. First, it should be 15

20 indirectly correlated with poverty and inequality. Second, it should be directly related to microfinance variable. In cross country study, legal origin and geographical latitude usually are being employed as instrumental variable to explain financial development (Beck, Demirguc Kunt, and Levine, 2007). Obviously, this research cannot use legal origin as instrumental variable because legal system is indifference across regencies in Indonesia. The Geographical latitude differs across regencies. However, the difference is very small. Thus, both legal origin and geographical latitude cannot be useful as instrumental variable. The other possibility is by using distance to Jakarta as instrumental variable. The argument is that the further away the regency from Jakarta, as financial center in Indonesia, the lower the magnitude of micro credit per capita. The line of reasoning following this argument is straightforward. The further away a regency from Jakarta, the higher the cost for a Bank to open a branch in that regency. Thus the number of bank s branch in the regency will be negatively correlated with the distance from that regency to Jakarta. As the number of bank goes down, the amount of micro credit being allocated will be dropped because financial institutions capability to collect fund is being declined. Theoretically, the distance to Jakarta will have no direct effect to poverty level in the regency as well as its income distribution. Meanwhile it could have direct effect on micro credit per capita. If this contention is hold, then theoretically distance to Jakarta can serve as a proper instrument for microcredit per capita as it satisfies both requirements to be good instrument. Econometric Model and Hypothesis The econometric model is being established based on the idea from the previous section. The econometric models along with the hypothesis to answer main inquiry on this research are presented in this section. The main econometric models to test the link between finance, poverty, and inequality are HCI i = α 1 + β 1 MSM i + γ 1 GRC i + δ 1 DEV i + θ 1 SCH i +ε i PGI i = α 2 + β 2 MSM i + γ 2 GRC i + δ 2 DEV i + θ 2 SCH i +е i GIN i = α 3 + β 3 MSM i + γ 3 GRC i + δ 3 DEV i + θ 3 SCH i +v i (ii) (iii) (i) Where: HCI i = headcount poverty index in regency i PGI i = poverty gap index in regency i GIN i = gini coefficient in regency i MSM i = micro small medium credit per capita in regency i GRC i = growth of GDRP per capita in regency i 16

21 DEV i = Development expenditure per capita in regency i SCH i = Percentage of Population with Primary School or less in regency i ε i, е i, v i = Error term As mentioned earlier in the previous sub section, instrumental variables are needed to deal with reverse causality bias and measurement error problem. In this research, all explanatory variables are endogenous. Thus, four instrumental variables are needed to avoid over/under indentified problem. Economic growth is instrumented by total local government expenditure for the reason that government expenditure of the regency will have multiplier effect to its economic performance. Meanwhile, government expenditure will indirectly affect poverty or income inequality through economic growth. Development expenditure per capita is instrumented by local government income after excluding income from natural resources. How much government of regency will expend on development expenditure is determined by the local government income. We cannot expect that local government income has an effect on poverty/inequality. Government income might have indirect effect on poverty through government spending on poverty program. It is possible that the relationship appears the other way around. We can expect that the poorer the regency, the lower the tax that can be collected. However, this relationship works indirectly through the taxes. Percentage of population with primary education or less is instrumented by Public Expenditure on Education for the reason that public expenditure on education will directly affect the number of educated people. Meanwhile, public expenditure on education will not have a direct effect on poverty and income inequality. This can be true by assuming there is no large difference on education expenditure across time. This implies that today public expenditure on education reflects past public expenditure on education. Thus, the econometric models to instrument endogenous explanatory variables are: MSM i = α 4 + φdis i + ω i GRC i = α 5 + ρgov i + µ i DEV i = α 6 + λgic i + ξ i SCH i = α 6 + πped i + υ i (iv) (v) (vi) (vii) Where: DIS i = distance to Jakarta from capital of regency i GOV i = total government expenditure in regency i 17

22 GIC i = government revenue after excluding income from natural resources in regency i PED i = public expenditure on education in regency i ω i, µi, ξi, υ i = error term The variable of interest is micro small medium credit per capita (MSM). Hence, the magnitude of β 1, β 2, β 3 is the main interest in this research on determining the link from microfinance to poverty and inequality. If β 1 = 0, then micro small medium credit brings no effect on headcount poverty index. If β 1 is positive (negative), consequently micro small medium credit has positive (negative) effect on headcount poverty index. In line with β 1 to headcount poverty index, β 2 and β 3 follow the same logic. Operationalization of Variables In contrast with Beck, Demirguc Kunt, and Levine (2007) this research is employing absolute measurement of poverty. The poverty line being used is the poverty line created by Badan Pusat Statistik Indonesia (Central Bureau of Statistics). The reason is that absolute poverty will be useful on making comparison among several areas (Cutler, 1984). In this case, the poverty line is the minimum amount of money to attain certain amount of needs. The second reason is that the absolute poverty line measures the poorest of the poor. It is important to distinguish between absolute and relative poverty line, as this will influence the interpretation of results. The following box 3.2 will exhibit the difference between absolute and relative poverty measurement. Box 3.1 Absolute VS Relative Poverty Line Absolute Poverty Line: Conditions of failure to meet the essentials of physical existence Measures the poorest of the poor Data based on consumption survey Free from cross cultural bias Relative Poverty Line: Income relative to others income Measures the poor relative to well being of others Data based on Households Income Easy to attain, for instance from data on income tax This research is exploiting poverty data from BPS Indonesia. Thus it is an absolute poverty line as BPS defines the poor based on its consumption. The main advantage of using absolute poverty line is that the data will be standardized because it is based on minimum living standard. In addition, it measures the poorest of the poor in society. However, the consumption data cannot be obtained effortless. The statistic bureau has to conduct household consumption survey, which is more difficult and more expensive than to collect income data. Sources: Cutler (1984) and Roemer and Gugerty (1997) After defining the poverty line, then both headcount poverty and poverty gap index are being determined. Headcount index is calculated by dividing the number of people who live under poverty line 18