Does the Micro-Macro Paradox hold? An Investigation of the Effects
of Microfinance on Economic Growth.
R. Nijland – s2596172 University of Groningen | FEB Supervised by Prof. C.K.D. Adjasi
July 2019
Abstract: This paper provides an answer to the question whether microfinance has growth effects and strengthens the financial system. Therefore it first identifies transmission channels through which microfinance can affect economic growth and selects successful microfinance institutions in 24 developing countries. Next, it analyzes growth effects of microfinance by applying Propensity Score Matching and by performing system GMM regression models including an interaction term between microfinance and the financial system. Based on the results it follows microfinance stimulates economic growth via financial inclusion, productivity and welfare effects. Moreover, when the effects of microfinance and the financial system are combined, finance-led growth occurs.
Microfinance affects poverty levels in developing countries by providing credit, insurance and savings. Globally we all see developing countries in which microfinance is growing, catching up with developed countries. Whether this effect can be assigned to microfinance is still heavily debated, as are the microeconomic transmission channels through which this would occur. Microfinance increases income and welfare on individual level, and this could sum up to economic growth on country-level. Sometimes microeconomic changes do not lead to macroeconomic changes, which is known as the micro-macro paradox. Investing in microfinance is one of the tools available for policy makers aiming on decreasing worldwide inequality, but therefore it matters whether this paradox is also at play in the field of microfinance. This is still not known, and this paper seeks to provide an answer to this relevant matter.
The relevance of macro- and microeconomic effects of microfinance as decreased poverty and inequality and increased financial inclusion is underlined by the United Nations Sustainable Development Goals. Microfinance mainly affects the first goal of Ending poverty and hunger and the tenth goal of Reducing inequality. On these goals microfinance has been argued to have a positive impact (Setboonsarng and Parpiev, 2008). Microfinance should be an effective tool for decreasing poverty and increasing welfare of the world’s poor. However, in research there is no clear-cut evidence for this claim. Many scholars focused on financial inclusion (Beck et al., 2015), welfare effects (Akotey and Adjasi, 2016), female empowerment (Karlan et al., 2009) or performed case studies on the effectiveness of MFIs. These papers however, have not yet found consensus on the effects of microfinance on poverty levels. In example, the paper by Augsburg et al. (2015) finds no impact of microfinance on overall income, Field (2013) does find an impact on income that depends on the length of a grace period and Ferdousi (2015) finds an overall positive effect on incomes. As a result of these contradicting studies, the discussion of the impact of microfinance has not yet been brought to a conclusion. Given the lack of consensus on the overall effects of microfinance and the importance of poverty reduction and reducing worldwide inequality, it is valuable to investigate the role of microfinance on economic growth.
Much of the literature written on macroeconomic effects of microfinance concerns single country analyses or theoretical frameworks. Raihan et al. (2017) find positive growth effects of microfinance in Bangladesh in their Computable General Equilibrium Model; Buera et al. (2012) generate a Partial and General Equilibrium model of which calibrated results based on data from India yield positive effects of microfinance on growth. The results of these papers lead to the hypothesis that microfinance has positive growth effects. Moreover, based on this literature, the following other hypotheses are expected to hold. First, it is expected this positive effect of microfinance on economic growth occurs through microeconomic channels for transmission. Second, positive interaction effects between the financial system and microfinance are predicted to occur.
This paper completes prior studies by empirically investigating growth effects of microfinance in a cross- country analysis. Therefore it first combines several studies on microeconomic effects of microfinance, to meet this study’s first objective of determining the channels by which microfinance affects economic growth. To quantify the growth effects of microfinance and possible interaction effects, this study selects 35 MFI’s based on selection criteria and information available on MixMarket. These criteria ensure selecting only MFI’s with large outreach, mature age and focus on the very poor and are formed by age, outreach, number of active borrowers, services to the very poor and reliance on donated capital. Then matching procedures compare economic growth between treatment and control groups, after controlling for factors influencing the establishment of a MFI in a certain area. These factors concern economic, poverty-indicating and demographic factors. Next, by performing system GMM regressions on the data, economic growth of countries in which these MFI’s are located, is compared to growth in other developing countries, while controlling for structural factors in the economy. This way the paper tries to estimate the second objective of quantifying the impact of microfinance on economic growth. The third goal of this paper, regarding the presence of interaction effects, is investigated by performing the regression analyses both with and without an interaction term between microfinance and the financial system.
Examining whether the combined effect of microfinance and the financial system exceeds the sum of its components, therefore forms a major contribution of this paper.
The remainder of this paper is organized as follows. Section 1 investigates the main characteristics and methods of MFI’s. Section 2 tries to explain the relation between microfinance and economic growth via identifying three channels of transmission. In Section 3 the methodology of the empirical part is introduced and Section 4 contains the selection of MFI’s, other data and descriptive statistics. Lastly, Section 5 uses matching analyses and system GMM regression to investigate the effects of microfinance on economic growth, after which Section 6 concludes.
1. Microfinance, context and developments
A starting date for microfinance is hard to find. Already long before the introduction of the first formal MFI’s as Grameen Bank (1976), credit unions, village banks and other types of local initiatives were communities or organizations aimed at financing economic activities. The participants of these initiatives gathered once every time period to contribute a certain amount of money to a pot. The total amount of money was allocated to their projects as investment. The first formalization of this process occurred in the 20th century in the form of ROSCA’s, Rotating Savings and Credit Associations. These associations developed further into formal MFI’s. The main difference in the way MFI’s nowadays are organized is formed by the fact they are no longer local processes. In recent times microfinance institutions are often established by developed countries or wealthy donors as means for decreasing poverty. This increases the amount of money available for microfinance, and hence the number of poor lifted out of poverty. However, it is sometimes difficult for these donors to ensure focus on the very poor and to evaluate which lending methods are most effective. This might even lead to a shift from helping the very poor towards profits: mission drift. The sections below discuss these issues further.
1.1. Donor-driven versus commercially driven and mission drift
Microfinance is believed to be a powerful tool for decreasing poverty, since it is focused on helping the poor by providing financial services. Whereas in the past money for microfinance services often originated from the poor themselves, today it also comes from donors and investors. Two main schools of thought can be identified for the origin of the money necessary for these financial services: the Institutionalist and Welfarist approaches.
consider microfinance as opportunity for low risk portfolio diversification. The investments of these investors increase the amount of capital available for the poor. Therefore they are necessary for reaching the goals of microfinance.
The second school of thought is formed by the ‘Welfarist approach’. This group of scholars is against the transformation of MFI’s into for-profit commercial banks funded by MIV’s. They state this transformation causes the aim of MFI’s to shift from poverty reduction to profit maximization. Commercialization implies microfinance becomes based on economic principles, which are the same principles that pushed the very poor out of the banking market before.
Bisen, Dalton, and Wilson (2012) find nowadays many MFI’s are for-profit organizations funded by MIV’s. On the one hand, this results in more money available for the poor as the Institutionalists argue. On the other hand, poverty reduction is not the main motive of these investors and therefore the focus of microfinance could shift from helping the very poor, towards profit maximization. This mission-drift could exclude the poor from financial markets, as the Welfarists fear.
In essence, both the Welfarist and Institutionalist groups aim on keeping microfinance focused on the very poor. In other words, none of them wants mission drift to occur. How can mission drift be recognized? Mission drift can be defined as ex post changes in the stated preferences to fit unplanned performance outcomes (Copestake, 2007). Augsburg and Fouillet (2013) identify signs of this mission drift as rising interest rates and increasing loan sizes available for a smaller group of clients. Instead of measuring success by the depth of outreach, it becomes measured by the magnitude of its outreach, in which helping the very poor comes in second place. When MFI’s are focussing on profits, the programs offered to the poor lack sustainability, measured as the ability to meet present needs without compromising future needs (Jose and Chacko, 2017). High interest rates necessary for profits of MFI’s result in over-indebtedness of the poor. It now follows that in order to help the very poor, mission drift must be prevented. Actions to decrease MFI’s susceptibility to mission drift include improved goal setting, strategic planning, routine monitoring and structural evaluations of these actions to make them more cost efficient (Copestake, 2007).
donor funded MFI’s are selected. These donors are expected to ensure the focus on helping the poor.
1.2. Lending methods in microfinance
Microfinance institutions traditionally aim on offering credit to the very poor. While starting as credit providers, with time MFI’s developed into complete financial intermediaries by offering savings and insurance services. Necessary for MFI’s to both be able to service the very poor and avoid bankruptcy is offering credit against low interest rates, while achieving high repayment rates. When MFI’s are effective in achieving these high repayment rates against low costs, they can become self-sufficient. This way their future and sustainability does not depend on MIV’s or donors, but they might still be necessary for governance. Moreover, when they operate more efficiently, they are able to help more of the poor out of the poverty trap. Hence high repayment, efficiency and transparency could increase the effects of microfinance on economic growth. For achieving high repayment rates, MFI’s use methods different from those in traditional banking.
The first method they use is group lending with joint liability. This entails MFI’s lending to groups of borrowers, who have to repay for others in case one defaults. Good borrowers will group with other good borrowers using local knowledge. This lowers adverse selection. Moreover, they monitor whether their group members are spending credit for productive purposes. This lowers moral hazard and monitoring expenses for MFI’s. Since low risk types group with other low risk types and vice versa, high risk types more often have to repay for others’ default. This results in higher indirect costs of interest for the high risk types in a separating equilibrium (Gan, Hernandez , and Liu, 2018). Hence group lending uses local knowledge and monitoring to increase repayment rates via decreased adverse selection and moral hazard.
The second method proven by Field, Pande, Papp , and Rigol, (2013) to increase repayment rates is the introduction of grace periods. Most microfinance loans have early repayment. However, even though this is supposed to increase repayment, it decreases the amount invested in longer term investments. Therefore profitability of borrowers’ investments decreases and this on its term decreases the ability to repay. Hence the introduction of grace periods increases long term investments and thus repayment rates. Other mechanisms designed to increase repayment rates are dynamic incentive schemes, in which current actions have consequences for future funding.
provide the possibility to insure against economic and lifecycle risk. Karlan et al. (2014) find farmers are not only constrained by credit, but risk is even a larger hindrance to investment. Therefore insurance is more beneficial than credit, since it protects farmers to low-probability high-loss events. Hence by using saving and insurance products, the poor have more funds available for investments.
It follows the product portfolio within microfinance is expanding. Donors aiming on making their MFI financially independent, should focus on the new lending methods explained above to achieve higher repayment rates at lower costs. When following these new methods, MFI’s might be able to help more of the poor, due to the lower costs and be better prepared for the future. Therefore microfinance is expected to have an increasingly positive impact on economic growth.
2. The impact channels of microfinance - Theoretical overview
There are three main channels by which microfinance is expected to affect economic growth: finance-led economic growth, financial inclusion and welfare theory. All three will be further analyzed in this section.
2.1. Channel 1: finance-led economic growth
The finance-led growth hypothesis states where financial development occurs, economic development follows. This supply-led growth theory started with Schumpeter (1911) arguing developed financial systems are necessary components for economic growth. Confirming this line of reasoning from financial development to economic growth is the research of King and Levine (1993). Their empirical study finds financial development indicators predict subsequent values of economic growth and concludes Schumpeter was right. Since microfinance increases banking facilities available for the poor and strengthens the financial system in developing countries, it is expected to cause economic growth.
of an economy by improving the link between savings and investment. Therefore microfinance is followed by economic growth (Roubini and Sala-i-Martin, 1991).
It now follows that by increasing capital accumulation via removing the need to self-insure, by educating the poor based on higher level education and by increasing the savings-investments link, microfinance leads to economic growth following finance-led growth theory.
2.2. Channel 2: financial inclusion
Microfinance enables the poor to enter the financial system by providing access to credit and other financial services. This is captured by the term “Financial inclusion”. These poor households and enterprises were not able to access financial services offered by commercial banks before, either because of lack of collateral, transaction costs or high perceived risks. Beck, Senbet, and Simbanegavi (2015) show this financial inclusion leads to economic growth.
First, financial inclusion leads to increased sizes of financial markets, which fosters innovation. Innovation or the commercial application of ideas and inventions is, as a changing force from within the system, a main source of growth. First of all, by increasing the size of financial markets in developing countries, financial inclusion attracts foreign banks. These foreign banks are generally more sophisticated and spur innovation. Second, large and stable systems can be exploited by scale economies, for which innovation is needed. It now follows by enhanced innovation and economies of scale, microfinance causes economic growth via financial inclusion.
Moreover, financial inclusion transfers capital and savings of the poor from the informal to the formal financial system. This leads to higher capital accumulation. The formal financial system efficiently reallocates the resources of the poor. More efficient allocation of resources caused by financial inclusion fosters economic growth.
In addition, financial inclusion offers the unproductive poor the possibility to use credit for productivity enhancing activities. These activities lead to more economic growth than other possible purposes for the money, since there exist diminishing returns to scale. Investing an unit of money in a poor person to allow him to start working, yields higher returns than investing this unit in someone who already works. Therefore unlocking the potential of the poor by financial inclusion yields higher per unit returns than investing in wealthier groups of people. Hence financial inclusion spurs economic growth by unlocking the productive potential of the poor.
competitiveness of MSE’s. Hence financial constraints hinder growth, while MSE growth could lead to expansion in small and medium enterprises (SME’s). Financial inclusion decreases financial constraints for MSE’s and henceforth enhances growth (Ishengoma and Kappel, 2011). Moreover, it is proven MSE’s make an increasingly important contribution to employment and national income at time of relative decline in the contribution of larger businesses (Chan and Lin, 2013). It follows MSE growth is hindered by lack of access to credit, while MSE’s are important to national employment and income. Therefore financial inclusion leads to economic growth by supporting productivity of MSE’s.
In conclusion, financial inclusion by microfinance causes economic growth, via fostering innovation, efficient allocation of resources, unlocking the productive potential of the poor and increasing productivity of micro and small enterprises.
2.3. Channel 3: welfare effects
Microfinance is beneficial for economic growth when it enhances welfare for the poor structurally. Many studies investigate the effects of microfinance on the poor’s living standards and conclude microfinance enhances welfare of the poor via the following routes. First of all, microfinance increases living standards of the poor which leads to economic growth. Their productive potential and income rise, which leads to improved health and nutrition. When a society has a higher standard of living and welfare, economic growth follows (Rajan, 2009).
Moreover, microfinance does not only increase the poor’s standards of living in terms of health as described above, but also in terms of education. Adjei, Arun, and Hossain (2009) find microcredit to increase amounts of human capital of the poor, by providing opportunities for higher levels of education. The higher levels of health and education, increase employability and thus income of the poor. It follows the welfare effects of microfinance increase economy-wide employment and income, and economic growth occurs.
Which effect is more likely to dominate? The positive productivity effect caused by higher allocative efficiency will certainly be present. Whether the negative productivity effect occurs, depends on whether new entrepreneurs actually enter the market. These entrepreneurs will enter when the profits of becoming an entrepreneur exceed the opportunity costs of forgoing wage. Microfinance results in an increase in factor prices, hence in wages. Therefore at the same time opportunity costs of becoming entrepreneur increase (you forgo higher wages), and the costs of being an entrepreneur increase (you pay higher wages). Both of these effects limit profitability and thus entry of new entrepreneurs. It now follows microfinance positively affects Total Factor Productivity, since the negative effect caused by entry of new entrepreneurs is dominated by the positive effect of its allocative efficiency (Buera et al., 2012).
In contrast to the line of reasoning described above, some scholars argue microfinance is welfare decreasing. There are two main arguments provided by opponents of microfinance. The first argument is built upon the risk of over-indebtedness. When the poor use their credit to consume instead of for acquisition of productive assets, they have to rely on new loans to repay the first. The second argument is based on economies of scale. MFI’s provide credit to small enterprises, which have no economies of scale and are less efficient than large enterprises. It follows resources are not allocated to their most productive means and economic growth is hindered. However, empirical research by Raihan (2017) shows monitoring ensures credit is used for investment purposes and the main source of this credit is the sector of microfinance itself. Therefore these negative effects of microfinance on welfare are unconvincing.
As a final point, the poor suffer not only from lack of capital, but also from lifecycle and economic risk. Microinsurance decreases these risks, while microcredit provides capital. Providing only credit without offering insurance has little or no impact on their welfare (Akotey and Adjasi, 2016). In this case credit might be used to resolve risky events instead of it being invested in the poor’s income- generating capacity. By combining both microcredit and microinsurance, the poor are insured against risks and enabled to invest in their own productivity. Hence microfinance increases welfare of the poor, which leads to economic growth.
2.4. Overview of empirical literature on growth effects of microfinance
From the three channels above it follows microfinance leads to enhanced economic growth. Microfinance increases the level of financial development, leads to more financial inclusion and has positive welfare effects. Since transmission via financial inclusion and welfare effects both leads to increased productivity, donors looking for effects of microfinance should monitor productivity. The combination of these three channels leads to the hypothesis microfinance positively affects economic growth. This hypothesis is supported in literature. Among the first studies on quantifying the macroeconomic effects of microfinance is the paper of Buera et al. (2012). This article shows microfinance has both a negative and a positive influence on the economy, in which the positive effect prevails.
Microfinance increases factor prices and thus wages. When wages increase, entrepreneurs face higher costs of the labour they hire, which decreases their own income and thus savings. These wages are earned by the poor. In other words, there is a transfer of money from the entrepreneurs to the poor. Whereas the entrepreneurs saved this money, the poor rather consume it. Hence this transfer of money from high-saving entrepreneurs to the low-saving poor hinders capital accumulation and therefore negatively affects economic growth.
However, microfinance also increases Total Factor Productivity, as explained above in the effects of microfinance on welfare. This positively affects economic growth.
Since this empirical study of Buera et al. (2012) shows that the negative effects on capital accumulation are outweighed by the positive effects of increased TFP, it supports our hypothesis.
Consistent with the findings of Buera (2012) is the study of Donou-Adonsou and Sylwester (2017). They find positive effects of microfinance on economic growth, in contrary to no growth effects of traditional bank lending. These growth effects can mainly be attributed to increases in Total Factor Productivity caused by microcredit, since they find no investment effects. However, they emphasize no miracles should be expected and the positive growth effects of microfinance, even though present, are small.
Lastly, Raihan et al. (2017) find microfinance accounts for approximately 12% of total GDP in Bangladesh. They considered only the capital-augmenting part of microfinance and compared current GDP to a simulation in which microfinance did not exist. This impact of microfinance is partly direct, by increasing production. Second, the impact of microfinance is indirect: it decreases the effective price of capital. Lower prices of factor inputs increase production and hence GDP. It now follows this study of Raihan positively relates microfinance to economic growth via lower factor prices.
Moreover, I expect the finance-led growth theory to hold and hence positive effects of financial systems on economic growth. Lastly, when both microfinance and the financial system are positively influencing economic growth, I predict an interaction term between the two increases economic growth as well.
2.5. Drivers of economic growth
Above it is stated microfinance leads to economic growth. In order to test this hypothesis, other drivers of economic growth must be identified and controlled for in the regressions. These other factors are the stock of physical capital, human capital, institutional quality, technology and trade. Their influence on economic growth will be discussed below.
A high stock of money, formed by savings, that is available for investments leads to physical capital accumulation. Education determines the quality of human capital. Moreover, education leads to innovation and higher levels of technology. Together, increases in capital accumulation and human capital increase productivity of capital and labout, which leads to economic growth.
Using capital for investments often includes incurring sunk costs, and uncertainty increases the probability of not recurring those costs. Since high quality institutions decrease uncertainty, institutional quality is also a driver of economic growth (Fedderke and Simkins, 2012).
Lastly, since next to education, exports also increase the level of technology and innovation via technology transfers, exports are a driver of growth. This growth effect of exports is enlarged by lower unit-costs due to economies of scale and scope (Jawaid, 2014).
It now follows the factors above influence economic growth. Therefore the stock of physical capital, human capital, institutional quality, technology and trade are drivers of economic growth, which must be included in the regression models.
3. Methodology
This paper applies matching procedures and system GMM regression analyses to investigate the following hypotheses regarding the effects of microfinance on economic growth: first, based on the three channels identified in Section 2, it is expected to find a positive aggregate effect of microfinance on economic growth. Second, finance-led economic growth theory is assumed to hold and hence positive effects of financial systems on economic growth are predicted. Third, when both microfinance and a developed financial system are present, an interaction term between the two stimulates economic growth as well.
results and can only be interpreted as indication. This is due to the fact the outcome can be contributed to differences in other factors than the presence of microfinance.
Next, this study uses Propensity Score Matching (PSM). This method uses a logistic regression to generate propensity scores. PSM can be used after implementation of programs and provides the probability of treatment given observed factors. Since the MFI’s already were started many years ago and their establishment was not randomized, PSM is the most appropriate tool for evaluation. In this model treatment consists of hosting a selected MFI and the counterfactual is not having such a MFI. First the model calculates the propensity score, or the probability of treatment given the observed:
e(X) = P(Z = 1|X), 0 < 𝑒(X) < 1 (1)
In Eq. (1), e(X) equals the propensity score, Z is treatment and X represents the covariates. Individuals with equal propensity scores are balanced: they have the same distribution of X. This is expressed by Eq. (2):
Pr {X|Z = 1, e(x)} = Pr {X|Z = 0, e(x)} (2)
The effect of microfinance on economic growth can now be estimated as the mean difference in economic growth between the treatment and control group. This is denoted by Eq. (3).
Treatment effect = E {Y|Z = 1, e(x)} − E {Y|Z = 0, e(x)} (3)
To check whether the results of PSM are robust to changing the method of matching, this study applies Nearest Neighbour Matching(NNM). NNM is a flexible matching approach since it drops functional form assumptions. In this study for every of the 24 countries in the treatment group, only the nearest country in the control group is selected. This approach therefore includes 48 countries. NNM needs a lot of data and could be subject to large sample bias.
For determining the appropriateness of these matching procedures on our data, covariate balance tests are executed. These tests investigate whether treatment and control groups have similar means and variances on the covariates. Moreover, density and box plots graphically show the match between treatment and control. However, even if these tests show the results are reliable, it has to be kept in mind data from developing countries might be flawed. Moreover, on some variables no observations are available for this group of countries. Those variables are left out of the analysis as described in the next Section (4). This limits the reliability of the results. To check whether the results are caused by outliers, PSM is repeated on the dataset winsorized on the 5% level.
(2017). Using this regression method the results of the matching procedures on growth effects of microfinance are checked, and the third hypothesis regarding the interaction effects is investigated. The advantage of using this type of estimator lies in its power to avoid any endogeneity problems and biases originating from weak instruments. The model in this paper could be subject to endogeneity, since there might be reverse causality between economic growth and the financial system. The financial system is a variable of interest for checking the third hypothesis regarding interaction effects. Moreover, it could be the case some factors from outside the model are influencing economic growth. This way the model suffers from omitted variable bias. Lastly, the data from developing countries could be subject to measurement errors. Since these three elements of endogeneity might be present, system GMM is appropriate to use. This method is also efficient in estimating models with large N, small T panels. As this panel has only six periods (each containing three years) and 93 countries, it is a large N small T panel. Other advantages of the system GMM estimation method are robustness to heteroskedasticity and autocorrelation. Moreover, the estimator is valid in models where variables of the right hand side of the equation are correlated with past and current values of the error term.
The system GMM estimation method uses two equations to estimate the results of a model in which fixed effects are removed and the lags of endogenous variables are used as instruments. Consider the original model defined as follows:
yi,t = β𝑖 + θt + β1yit−1+ ∑ β𝑘Xit 𝑛
𝑘=3
+ ε𝑖 (4)
In this model y is the logarithm of GDP per capita and X the set of controls. This set of controls includes human capital measured by primary school attainment, trade openness measured by exports to GDP ratio, capital stock measured by gross capital formation to GDP ratio and institutional quality and it is an empty set in the first estimates of the model. Since the right-hand side of Eq. (4) might be endogenous, the system GMM model transforms the equations. The system GMM regression method differences Eq. (4), resulting in the transformed Eq. (5). The variable of interest, microfinance, is constant over time. Therefore this variable will be added after Eq. (4) is differenced, since it would be removed as time-invariant fixed effect otherwise.
yi,t− yi,t−1 = θt− θt−1 + β1(yit−1− yit−1) + ∑ β𝑘(Xit 𝑛
𝑘=3
− Xi,t−1) + ε𝑖𝑡− ε𝑖𝑡−1 (5)
estimated consisting of the left-hand side of (5) and right-hand side of (4) now including the MFI dummy, as shown in Eq. (6). This equation is needed because the independent variables could follow a random walk. To circumvent this problem, Eq. (6) is estimated and denoted as follows.
yi,t− yi,t−1 = β𝑖 + θt + β1yit−1+ β2MFI + ∑ β𝑘Xit 𝑛
𝑘=3
+ ε𝑖 (6)
In this study two system GMM models are tested. The first model is the Basic Model. In this model both a dummy for MFI and a proxy for the financial system are included. However, there might be biases in this model, since MFI’s form a subset of the entire financial system. Therefore this model also contains estimates of regressions in which the proxy for the financial system is excluded (included), while the MFI dummy is included(excluded). These regressions suffer from omitted variable bias.
The second model is the Enlarged model. This model consists of the Basic Model, enlarged by an interaction term between the financial system and the presence of microfinance. The goal of the Enlarged Model is to investigate whether they are mutually enforcing and whether the results on growth effects of microfinance in the Basic Model are partly caused by interaction effects.
To check the reliability of the results of both the Basic and the Enlarged model, two checks for robustness are executed. In the first check both models are estimated using an alternative measure for the strength of the financial system. The second check estimates the models after winsorizing the data on 1% and 5% levels. This last check ensures the results are not caused by outliers in the dataset.
When applying system GMM, several diagnostic tests are necessary for determining the suitability of the model. In order to identify whether the instruments are valid, the Sargan test of over-identifying instruments is conducted. This test assumes under its null hypothesis any over-identifying requirements are valid. Moreover, the paper uses an AR(2) test to test for second-order serial correlation in the error terms. This test assumes under its null hypothesis errors are not serially correlated beyond lag one.
4. Data
(MixMarket). The data goes from 2000 till 2017 and is clustered in periods of three year to avoid stationarity problems.
4.1 Treatment and control Group
Using the information available on MixMarket the treatment group is selected by applying filters on the full set of MFI’s. The filters stated in Appendix B, are aimed on selecting MFI’s i) focussing on the very poor, that have ii) large outreach, and are iii) mature.
Focus on the very poor is conceptualized by the presence of donated capital and type of client base. These factors decrease the probability of mission drift.
Outreach and the number of borrowers express magnitude; only sufficiently large MFI’s are able to influence economic growth.
Mature age is necessary to account for experience; MFI settlement, borrowers’ investments and growth effects take time.
Data on the aspects above has to be reliable. Therefore only MFI’s with at least two out of the five diamonds for reporting quality are selected.
Applying the constraints above, 35 MFI’s in 24 countries met all requirements. These MFI’s are incorporated in the dataset by use of a dummy variable. This dummy equals one for countries in which the selected MFI’s are located, and zero otherwise. Since the selected MFI’s are subject to the requirements above, they are significantly large (sometimes as large as a traditional bank) and efficient MFI’s, active for many years, of high reporting quality and without mission drift. Therefore it is likely these MFI’s are able to affect economic growth. It has to be kept in mind, that even though these high quality MFI’s could influence economic growth, other MFI’s of less quality might fail to do so. The names and spread of the countries in the treatment group are depicted in Figure 3 in Appendix B.
The control group has to be similar to the treatment group. The treatment group consists of developing countries, mainly in Asia and Africa. These countries are all in the low- and low – middle income category of the World Bank. To select countries with similar characteristics, all other countries in these categories of the World Bank are selected as control group. However, some of those countries were removed from the control group due to a lack of data available. Together the treatment and control groups form a dataset consisting of 93 countries in total, of which 24 are in the treatment group and 69 are in the control group. The spread of the countries in this dataset is shown in Figure 4 in Appendix B.
4.2 Selection and control variables
established in a specific country. MFI establishment is likely to be affected by the strength of the financial system, availability of facilities as electricity and education, life expectancy and quality of institutions. The data used on these selection variables originates from data of the World Bank. As control variables for the regressions macroeconomic factors influencing economic growth have to be included in the analysis. From the literature section it follows drivers of economic growth are financial systems, political stability, trade, capital stock and human capital. These drivers are assessed by the following variables in Table 1:
In the analysis the three indicators of political stability are combined into one binary variable named institutional quality. This dummy equals zero for countries of low institutional quality and one for countries of high institutional quality. In this new measure the three indicators are equally weighted. As measure for the financial system domestic credit provided by the financial system will be used, since the other measure might be biased towards large firms. This alternative measure is used to check the robustness of the results. More precise descriptions and abbreviations of the selection and control variables used in the analysis are explained in Appendix A. Summary statistics on the treatment group in Table 2 show MFI’s lend mostly to female borrowers, have over 34,000 active borrowers and a large amount of donated capital. The selected MFI’s take low risks, since portfolio at risk is below the recommended target value of 5% (Caribbean Microfinance Alliance, 2011).
Table 2
Summary statistics on the selected MFI’s
Variable Mean Minimum Maximum Standard Deviation
Donated Capital (USD) 5,195,216 407 42,243,061 8,727,520
Number of active borrowers 326,136 34,054 4,557,601 764,594 Portfolio at Risk > 90 days 2.70% 0.0% 35% 6.2% Percentages of female borrowers 78% 29.94% 100% 22.7%
Table 3 presents some summary statistics on the control and selection variables used in the analyses. Some of the variables seem to be biased by outliers. Therefore Propensity Score Table 1
Control variables for drivers of economic growth Economic
growth
Trade Human Capital Financial Systems Political
Matching and the system GMM will be repeated with winsorized variables as a check for robustness.
Table 3
Summary Statistics
Variable Obs. Mean St. Dev. Min. Max.
Time 1,674 2000 2017 MFI 1,674 0 1 LGDP 1,612 7.051 0.968 4.713 9.677 Inst. 1,674 0 1 SEP 1,339 93.496 12.680 21.723 100 Exports 1,517 30.119 16.733 0.099 102.837 DCPSB 1,520 25.574 22.051 0.0046 156.809 DCPFS 1,507 36.998 34.866 -46.809 215.238 LifeExp 1,581 63.297 8.383 38.702 79.522 Electricity 1,581 55.101 33.692 0 100 LDev 1,638 19.851 1.173 16.336 23.160 AgriEmp 1,620 49.425 22.381 4.600 92.255 Lpop 1,668 16.205 1.768 11.343 21.050 UrbPop 1,650 42.040 18.414 8.246 87.490 LGCF 1,440 10.151 1.168 5.029 12.920 5. Results
It is expected the presence of microfinance positively affects economic growth. To quantify and prove the effects of microfinance on economic growth, Section 5.1 presents the results of the matching procedures. Next, Section 5.2 continues the story by first showing the results estimated via the system GMM Basic Model. Section 5.3 analyzes the effects of microfinance on growth even further by investigating in the Enlarged Model whether they are dependent on the strength of the financial system. These procedures combined test the three hypotheses regarding the effects of microfinance on economic growth.
5.1 Matching results
Figure 1 represents the distribution of the propensity scores for both treatment and control groups and the Common Support Region
Figure 1: The distribution of propensity scores for the Treated and the Untreated Source: author’s computation based on data of the World Bank
For a check of the reliability of the results of the PSM, the left-hand side of Figure 2 depicts the balance plot of the propensity scores of the raw and matched data. From this figure it follows the data could include significant outliers, which might lead to biased results. Therefore the data is winsorized at 5% level and the PSM is repeated. The balance plot in the right-hand side of Figure 2 shows the winsorized data is more in balance and less subject to outliers than the non-winsorized data. Therefore the PSM will be executed based on this winsorized data.
Figure 2: Balance plots of the data for Propensity Score Matching (Right panel is winsorized) Source: author’s visualization based on matching results
0 .2 .4 .6 .8 1
Propensity Score
Untreated Treated
The results of the probit estimation on the winsorized data are shown in Table 4. Table 4
Results of the winsorized probit regression for estimation of the propensity score (treatment in MFI)
MFI Coefficient Robust std. Error P-Value
Financial system 0.016 0.0016 0.000*** Life Expectancy -0.015 0.0070 0.029** Education 0.020 0.0042 0.000*** Access to Electricity 0.0055 0.0018 0.003*** Agricultural Employment 0.0025 0.0027 0.345 Urban Population 0.0049 0.0033 0.132 Institutional quality -0.119 0.119 0.318
Note: significant at respectively the 1%(***), 5%(**) or 10%(*) level. Source: Author’s computation based on data of the World Bank
From the results it follows the financial system, life expectancy, education and access to electricity are likely to affect the establishment of a MFI. Based on the propensity scores of the probit regression PSM is executed. These results are depicted in Table 5 (column 3), together with the results of a T-test (column 1) and NNM (column 2).
Table 5
T-test, Propensity Score Matching & Nearest Neighbour matching Outcome variable economic growth; treatment variable MFI
(1) (2) (3) Treatment effect 0.494*** (0.053) 0.145*** (0.048) 0.124** (0.062) Number of observations 1,612 965 1,183
Note: significant at respectively the 1%(***), 5%(**) or 10%(*) level; SE in parentheses. Source: Author’s computation based on data of the World Bank
From Table 5 follow the estimates of the impact of microfinance on economic growth. The results of these matching procedures provide evidence of microfinance increasing economic prosperity. Under each of the methods in Table 5, the results are statistically significant and provide evidence of a positive relation. The difference between growth in countries with and without microfinance is respectively 49%, 15% and 12%, depending on the method of estimation. The results of NNM, column (2) and PSM, column (3), are quite similar. The results of the T-test, column (1), overstate the effect of microfinance, since it does not take the impact of differences between countries on other aspects into account.
for the standardized difference of institutional quality and its variance ratio. Moreover, the variance ratio of urban population is not close to one either under PSM. For NNM the standardized difference of the financial system is significantly larger than zero; the variance ratio of urban population is significantly below one.
Table 6
Covariate balance statistics on matching procedures
Outcome variable economic growth; treatment variable MFI
Propensity Score Matching Nearest Neighbour Matching
(1) (2) (1) (2) Financial System 0.064 0.959 0.184 1.065 Life expectancy -0.003 1.079 0.070 0.912 Human Capital -0.067 0.758 0.084 1.036 Electricity 0.027 1.061 0.007 0.986 Urban Population -0.022 0.973 -0.067 0.842 Agricultural Employment -0.022 0.854 -0.001 0.940 Institutional quality 0.161 1.300 0.030 1.052
with positive growth effects of microfinance identified by Donou-Adonsou and Sylwester, (2017). However, they do not investigate the effects in a cross-country analysis as this study. It is of interest for practitioners of MFI’s and policy makers to know via which channel transmission occurs. The results of the PSM do provide evidence of channel 1. Based on Table 4 we can argue developed financial systems increase the probability of high quality MFI’s, which on its turn leads to economic growth. Hence this line of reasoning supports channel 1. Of course the various channels via which microfinance could influence economic growth are not mutually exclusive. It needs to be noted the distinction between channel 2 and 3 is a fine line. Financial inclusion increases productivity, but welfare effects affect productivity as well. So far, the results cannot with confidence indicate which one of these channels is most pronounced. Whether growth occurs via channel 1 – finance led economic growth, channel 2 – financial inclusion or channel 3 – welfare effects, cannot be stated with certainty. Practitioners of MFI’s searching for effects of their work in microfinance therefore can trust it leads to economic growth, and focus on the first channel, since that channel is supported by the PSM. However, based on literature the effects microfinance has on productivity via financial inclusion and welfare effects are still very likely to occur. Therefore donors looking for effects of their investments in microfinance should monitor changes in productivity of MFI’s clients.
Till now evidence is presented in favour of both the hypothesis microfinance stimulates economic growth as of the first channel of transmission of the finance-led growth theory. To verify this positive relation between microfinance and economic growth even more, the next section elaborates on the system GMM regression model. In addition to the matching model, this system GMM model will check the hypothesis of whether interaction effects between microfinance and the financial system occur.
5.2 System GMM results of the Basic Model
The Basic Model represents an analysis of the influence of microfinance and the financial system separately to check the results of the matching procedures. In the next section, the Enlarged Model will also estimate these effects, but add a measure for the impact of microfinance and the financial system combined. Table 7 presents the results of the estimation of the Basic Model using system GMM regression methods.
Table 7
Basic Model: effect of microfinance and the Financial system on economic growth System GMM
Dependent Variable: Log of GDP per capita
Financial system -0.011*** (0.001) 0.001 (0.001) -0.0005 (0.001) MFI -9.981*** (1.004) 0.272* (0.146) 0.427*** (0.158) LPop -0.050 (0.045) -0.025 (0.040) -0.064 (0.043) LDev 0.283*** (0.062) 0.203*** (0.053) 0.228*** (0.054) LDevt – 1 -0.257*** (0.061) -0.208*** (0.054) -0.197*** (0.055) Human Capital 0.000 (0.003) 0.000 (0.003) -0.001 (0.003) Trade 0.006*** (0.002) 0.005*** (0.001) 0.006*** (0.002) Institutional Quality 0.071 (0.085) 0.089 (0.075) 0.097 (0.076) Capital Stock 0.488*** (0.038) 0.508*** (0.036) 0.498*** (0.036) Constant 4.746*** (0.318) 1.152*** (0.129) -0.850 (0.755) -0.450 (0.654) -0.642 (0.666)
Sargan test p-value 0.000 0.000 0.6406 0.1882 0.4046
AR(1) test p-value 0.000 0.0004 0.0317 0.0357 0.0335
AR(2) test p-value 0.002 0.0003 0.4325 0.3466 0.3057
Number of groups 92 89 84 82 82
Number of observations 448 423 365 350 350
Note: significant at respectively the 1%(***), 5%(**) or 10%(*) level; SE in parentheses; all control variables
are potentially endogenous and treated as such in the system GMM.
Source: Author’s computation based on data of the World Bank
growth, while the financial system doesn’t. The coefficient of microfinance in this full Basic Model indicates microfinance increases GDP per capita by 42.7% compared to countries without microfinance. This seems quite large and therefore several tests for robustness will be executed in the next section. However, only MFI’s of high quality and significant size are selected. The MFI’s are serving many borrowers for many years with high amounts of donated capital, which could also explain this large number. When the controls are included in the regressions, all tests for valid over-identifying restrictions and serial correlation except for the AR(1) test fail to reject the null hypothesis. This can be derived from the fact the p-values are insignificant at 1% and 5% levels.
The covariates capital stock, development aid and trade have a significant and positive effect on economic growth and the lag of development aid is significant and negative, while the coefficients for the level of institutional quality, population size and human capital are insignificant. The lag of GDP per capita has a significant and positive effect on the current value of GDP per capita.
Concluding on the main research question, the Basic Model shows microfinance significantly and positively influences economic growth. Whether this effect remains after inclusion of an interaction term, will be investigated next.
5.3 System GMM Results of the Enlarged Model
The Basic Model confirmed the story of the positive effects of microfinance on growth, started by the matching procedures. It considered the effect of microfinance and the financial system separately and found microfinance to increase economic growth. However, the positive effect of microfinance on growth of 42.7% seems rather large. Therefore the Enlarged Model continues and investigates whether an interaction term further clarifies the impact of microfinance on economic growth. When in this model the coefficient of the interaction term is positive (negative) and significant, microfinance magnifies (diminishes) the effect of the financial system on economic growth. Table 8 depicts the results of the Enlarged Model.
Table 8
Enlarged Model: effect of the interaction variable on economic growth System GMM
Dependent Variable: Log of GDP per capita
LPop -0.045 (0.034) -0.041 (0.035) LDev 0.183*** (0.044) 0.172*** (0.045) LDevt – 1 -0.182*** (0.047) -0.186*** (0.049) Human Capital 0.000 (0.002) -0.0004 (0.002) Trade 0.003** (0.002) 0.003* (0.002) Institutional Quality 0.107* (0.063) 0.103* (0.062) Capital Stock 0.512*** (0.029) 0.515*** (0.030) Constant -0.286 (0.589) -0.058 (0.608)
Sargan test p-value 0.5526 0.4170
AR(1) test p-value 0.0255 0.0240
AR(2) test p-value 0.9063 0.8540
Number of groups 81 81
Number of observations 338 338
Note: significant at respectively the 1%(***), 5%(**) or 10%(*) level; SE in
parentheses; all control variables are potentially endogenous and treated as such in the system GMM.
Source: Author’s computation based on data of the World Bank
p-values, and therefore we cannot reject the null hypotheses of zero serial correlation and valid over-identifying instruments.
It now follows that based on the Enlarged model microfinance could positively affect economic growth, only when it interacts with a well developed financial system. This is contrasting to the Basic Model, where microfinance always stimulated economic growth, since this model lacked the interaction term. To investigate whether these results are robust and not caused by outliers, the next section will use an alternative measure for the financial system and winsorize the data.
5.5.1. Check for robustness: an alternative measure for the financial system
The results of Section 5.3-4 use domestic credit provided by the financial system as a measure for the financial system. An alternative measure is domestic credit to the private sector by banks, although this measure might be biased towards large firms. Descriptive statistics on this measure are already included in Table 3 in Section 4.2.
Changing this variable yields results similar to those observed before, depicted in Appendix C. For the Basic Model, including both microfinance and the financial system still results in only microfinance affecting economic growth significantly, whereas the financial system doesn’t. Hence the results of the Basic Model are robust to the change in measurement of the strength of the financial system.
However, in the renewed Enlarged model, the coefficient and p-value for microfinance change, as shown again in Appendix C. Where microfinance used to be significantly affecting economic growth in the Enlarged Model, it now becomes insignificant. The effect of the interaction term on economic growth remains unchanged in this check for robustness. Hence the effect of microfinance on its own has disappeared in this regression, and the only way in which microfinance is affecting growth is via its interaction with the financial system. The positive effect of microfinance on growth identified in the Basic Model, already became dependent on the financial system in the Enlarged Model in the former section, and this dependence increases after changing the measure for the financial system. The results regarding the third hypothesis of microfinance strengthening the financial system and these two factors combined stimulating economic growth therefore still holds. It follows the results of the Enlarged Model are not completely robust to the alternative measure for the financial system, and microfinance only affects growth in countries with developed financial systems.
5.5.2. Winsorize the variables
seem to show the variables are subject to outliers. This idea of outliers is confirmed by means of histograms in Appendix D. Therefore it is valuable to winsorize the variables at 1% and 5% level. Appendix E contains the results of the regressions using winsorized data. At the 1% level the results of the Basic model remain robust. The only change in the results is the increase in significance level of the financial system. However, in the Enlarged model, the significance levels for both the financial sector and microfinance change. Whereas in the non-winsorized model microfinance was significant in combination with the interaction term, now the financial system is significant and microfinance insignificant. Moreover, the coefficient of the interaction term has dropped. It follows the results of the Enlarged model are not robust to winsorizing the data on 1% level. When the 5% level is considered, similar changes occur. Again in the winsorized Basic model all signs and significance levels are identical to the non-winsorized estimates, while in the Enlarged Model microfinance becomes insignificant. On the 5% level, the financial system in the Enlarged model remains insignificant as in the non-winsorized model. Hence the results of the Enlarged model are robust to winsorizing the data on 5% level, except for the significance of microfinance.
5.4 Final results of the system GMM models
In conclusion, this winsorized Enlarged Model confirms the positive growth effects of microfinance, identified by matching analyses, the Basic Model and literature. In this Model, microfinance and developed financial systems of average strength together increase economic growth with 14.4%.
5.5 Arguments behind the results
The line of reasoning started with matching procedures indicating microfinance increases growth by 12-15%. Next, the Basic Model obtained this positive growth effect of microfinance as well, but indicated it as 42.7%. To investigate whether these results were caused by interaction with the financial system or outliers, the Enlarged winsorized Model is estimated. This results again in microfinance positively affecting growth, by 14.4% in an average financial system. This result is similar to both the results of NNM and PSM. Hence so far, all analyses indicated positive effects of microfinance on growth, although different in size. Some reasoning can be applied to explain the results above. The fact microfinance positively affects economic growth is in line with the literature in Section 2.1-3. In this section three channels were identified, namely finance-led growth, financial inclusion and welfare effects. In the theoretical framework of Section 2, Buera et al. (2012) found that by increasing Total Factor Productivity and financial inclusion, microfinance has positive growth effects. It follows the results of matching, the Basic and Enlarged Model confirm hypothesis 1 of a positive effect of microfinance on economic growth.
The second hypothesis regarding finance-led growth theory and channel 1 was supported by matching, although this effect was identified indirectly: the financial system increased likelihood of microfinance, which led to growth. The Basic Model measures this relation directly and shows the effects of the financial system on economic growth are insignificant. Therefore the results of the Basic Model reject finance-led growth and thus the second hypothesis of this paper. This is contradicting to the theoretical findings in Section 2.1 and empirical findings of matching, which led to the expectation of finance-led economic growth to occur. In theory it was stated microfinance and developed financial systems led to economic growth by i) increasing capital accumulation via removing the need to self- insure, ii) educating the poor and iii) strengthening the savings- investment link. Whether this second hypothesis is either confirmed or rejected in the Enlarged Model follows in the next paragraph.
financial system stimulate economic growth and the third hypothesis cannot be rejected. Since in this Enlarged Model microfinance and the financial system combined lead to economic growth, this model does favour the second hypothesis of finance-led growth. Hence the results of the Enlarged Model confirm finance-led growth, just as matching procedures did and in contrast to the results of the Basic Model.
Hence the validity of the effects of channel 1 on growth is ambiguous. However, all models indicate positive effects of microfinance on growth and both channel 2 and 3 are not falsified by the analyses and lead to increased productivity. Therefore donors looking for effects of their activities in microfinance should measure productivity effects on the borrowers.
These findings above are in line with other studies. The effects of microfinance on Total Factor Productivity are extensively investigated (Buera et al. (2012); Donou-Adonsou and Sylwester (2017). Moreover, the effects of productivity on economic growth have also been subject to numerous studies and proven to be growth enhancing(OECD (2015); Rogers (2003). This also follows from Krugman’s (1994) bold statement “productivity isn't everything, but in the long run it is almost everything”. Taken together, these studies, used in the theoretical framework of Section 2, also indicated microfinance via financial inclusion and welfare effects increases TFP and therefore stimulates economic growth. The positive results of the Enlarged Model and PSM regarding finance-led economic growth are in line with other findings in literature. For example the study of Madsen and Ang (2016) in OECD countries shows developed financial systems foster economic growth via production of ideas (R&D), savings and fixed investment. The contribution of the current study lies in connecting these theoretical results of Madsen and Ang (2016) on financial development in general to the field of microfinance in specific.
Concluding, both the results of matching procedures and system GMM regression models show microfinance positively affects economic growth. Whereas matching procedures estimate this effect as 12-15%, the Basic Model overstates this by estimating it as 42.7%. After including the interaction term in the Enlarged Model, it turns out microfinance does positively influence growth, when it interacts with a developed financial system. Since this effect depends on the strength of the financial system, it cannot be quantified in general. However, for an averagely developed financial system it equals 14.4%. This is similar to the effect estimated by matching procedures.
6. Conclusions and recommendations
The research question of this paper therefore lies in identifying whether microfinance stimulates economic growth. This research question is formulated as follows: What are the effects of microfinance on economic growth? Therefore the paper first questions by which transmission channels microfinance could foster growth. Afterwards it selects successful MFI’s and uses methods for data- analysis to quantify the effects of microfinance on economic growth. Even though other scholars already focussed on this subject, this paper distinguishes itself in several fashions. First, it combines theoretic channels for transmission with robust selection of MFI’s and empirical analyses. Moreover, it analyzes growth effects using both matching and system GMM regression models in a cross-country analysis. Lastly, it adds to literature by investigating the interaction effects between microfinance and the financial system.
Before using data to empirically investigate the research question, the following hypotheses were expected to hold. First, microfinance was expected to positively influence economic growth. Second, it was expected these growth effects occurred by finance-led economic growth, financial inclusion and productivity and welfare effects. Lastly, positive interaction effects between the financial system and microfinance were predicted to occur. These three hypotheses were empirically tested by matching procedures and system GMM regressions. These analyses yielded the following results. Matching procedures estimated the effect of microfinance on economic growth as approximately 12-15%. These procedures provided evidence of finance-led economic growth. These results are confirmed by system GMM regressions. These regressions including an interaction term and without outliers show the influence of microfinance on growth depends on the strength of the financial system. In an averagely developed financial system microfinance increases GDP per capita by 14.4% compared to countries without microfinance. The results on the relation between the financial system, microfinance and economic growth are robust to changing the measure for the strength of the financial system. Regressions including this alternative measure, also predict positive effects of microfinance on GDP per capita.
The findings regarding the growth effects of microfinance might seem rather large. However, similar results are found in literature. Moreover, the analysis only includes successful MFI’s with a large outreach, which are already active for many years with high amounts of donated capital. These MFI’s are not likely to be subject to mission drift due to donors and hence are lifting many of the very poor out of the poverty trap.
economic growth. In all our models, several types of matching, system GMM Basic and Enlarged, winsorized and non-winsorized, independent of the measure for the financial system, microfinance positively influences economic growth. In addition, the finance-led economic growth theory on itself does not hold, but microfinance and developed financial systems combined enhance economic growth.
Hence the research question of how microfinance translates into growth can be answered as follows. Microfinance stimulates economic growth by financial inclusion, welfare and productivity effects and finance- led growth theory holds due to interaction effects between microfinance and the financial system. It follows microfinance on individual level, translates into economic growth on country-level. This implies the micro-macro paradox does not hold. The results have consequences for several groups of interest. First of all, donors looking for effects could consider growth effects. However, it is difficult for them to measure economic growth and to estimate the part caused by MFI. Therefore donors searching for effects of their donations should focus on productivity effects caused by channel 2 and 3, since these are likely to be responsible for part of the growth effects of microfinance. Moreover, policy makers aiming on lifting the poor out of the poverty trap while fostering economic growth should consider microfinance as an option, and keep in mind this effect operates via developed financial systems.
References
Abadie, A., & Imbens, G. W., 2011. Bias-corrected matching estimators for average treatment effects. Journal of Business & Economic Statistics 29(1), 1-11.
Adjei, J. K., Arun, T., & Hossain, F., 2009. The role of microfinance in asset-building and poverty reduction: The case of Sinapi Aba Trust of Ghana. Working paper no. 87. University of Manchester, Brooks World Poverty Institute (BWPI), Manchester.
Akotey, J. O., & Adjasi, C. K., 2016. Does microcredit increase household welfare in the absence of microinsurance?. World Development 77, 380-394.
Arellano, M., & Bover, O., 1995. Another look at the instrumental variable estimation of error-components models. Journal of Econometrics 68(1), 29-51.
Augsburg, B., De Haas, R., Harmgart, H., & Meghir, C., 2015. The Impacts of Microcredit: Evidence from Bosnia and Herzegovina. American Economic Journal: Applied Economics 7(1), 183-203.
Augsburg, B., & Fouillet, C.,2013. Profit empowerment: The microfinance institution’s mission drift. In The Credibility of Microcredit (pp. 199-226).
Awaworyi Churchill, S., Korankye Danso, J., & Nyatefe, E., 2018. Microfinance institution performance: Does the macro economy matter?. Economic Papers: A Journal of Applied Economics and Policy 37(4), 429-442.
Bateman, M., 2010. Why doesn’t microfinance work? The destructive rise of local neoliberalism. Zed Books, London.
Bateman, M., & Chang, H. J., 2009. The microfinance illusion. Available at SSRN 2385174.
Beck, T., Senbet, L., & Simbanegavi, W., 2015. Financial inclusion and innovation in Africa: An overview. Journal of African Economies 24(Suppl 1), 11.
Bisen, A., Dalton, B., & Wilson, R., 2012. The social construction of the microfinance
industry: A comparison of donor and recipient perspectives. Cosmopolitan Civil Societies: An Interdisciplinary Journal 4(2), 62-83.
Bos, J. W., & Millone, M. (2015). Practice what you preach: Microfinance business models and operational efficiency. World Development, 70, 28-42.
Buera, F. J., Kaboski, J. P., & Shin, Y., 2012. The macroeconomics of microfinance (No. w17905). National Bureau of Economic Research.
Chan, S., & Lin, J., 2013. Financing of micro and small enterprises in China: An exploratory study. Strategic Change, 22(7-8), 431-446.
Copestake, J. (2007). Mainstreaming microfinance: social performance management or mission drift?. World development, 35(10), 1721-1738.
Donou-Adonsou, F., & Sylwester, K., 2017. Growth effect of banks and microfinance: Evidence from developing countries. The Quarterly Review of Economics and Finance 64, 44-56.
Dupas, P., & Robinson, J., 2013. Why don't the poor save more? Evidence from health savings experiments. American Economic Review 103(4), 1138-1171.
Easterlin, R. A., 2000. The Worldwide Standard of Living Since 1800. Journal of Economic Perspectives 14, 7-26.
Fedderke, J., & Simkins, C., 2012. Economic growth in South Africa. Economic History of Developing Regions 27(1), 176-208.
Ferdousi, F., 2015. Impact of microfinance on sustainable entrepreneurship development. Development Studies Research 2(1), 51-63.
Field, E., Pande, R., Papp, J., & Rigol, N., 2013. Does the classic microfinance model discourage entrepreneurship among the poor? Experimental evidence from India. The American Economic Review 103(6), 2196-2226.
