Microfinance institutions: a cross-country empirical investigation of outreach and sustainability.
Microfinance institutions (MFIs) confront special challenges of meeting double bottom lines: increasing outreach with financial services (the first bottom line) and attaining financial sustainability (the second bottom line) (Robinson, 2001; Armendariz de Aghion and Morduch, 2005). So, performance assessment of MFIs based on a double bottom line framework provides a better understanding on how far they are able to meet such dual challenges.
Financial sustainability of MFIs is broadly defined to be their ability to cover costs and to continue operations without resorting to philanthropic aid or subsidies existing in different forms (Armendariz de Aghion and Morduch, 2005). Such sustainability can be attained basically through ensuring loan repayments on time, earning enough interest revenue and controlling costs to guarantee efficient use of resources (de Crombrugghe, Tenikue and Sureda 2008). Standard indicators of these three components of operational performance are portfolio-at-risk (PAR), operational self-sufficiency (OSS) and cost per borrower (Armendariz de Aghion and Morduch, 2005).
Understanding the outreach and sustainability scenarios of MFIs, especially on a global landscape, is still demanding, as there has been a demonstrated inadequacy of rigorous empirical investigations. The evaluation literature on microfinance is rich. However, still there are scopes for improvement, especially in terms of sample size, as numerous empirical investigations on MFIs' performance are limited to a relatively smaller number of samples. Among others, for instance, two empirical studies that have examined the determinants of sustainability and outreach of a sample of MFIs are de Crombrugghe, Tenikue and Sureda (2008) and Cull, Demirguc-Kunt and Morduch (2007). The first one is based on a sample of 45 MFIs in India, while the second one utilized a global dataset of 114 MFIs. With better availability of cross-country data, however, the sample size can now be greatly increased. Besides, necessity for more evidence on the precise mechanisms describing numerous performance indicators of various microfinance delivery models is well established (Hermes and Lensink, 2007).
This paper aims at supplementing the literature with an analysis on MFI performance on a global perspective. The sample is much bigger, captures diversity in itself and concerns the present-day situation. Basically, operational aspects of MFI performance have been the main focus in this article. So, the marginal contribution of this exercise is related with examining how the sampled MFIs are able to accomplish sustainability without abandoning their social mission of increasing outreach. Three objectives have motivated this work--to add to the empirical underpinnings of the microfinance literature, to improve management yardsticks and to provide with proper recommendations.
The paper first looks at factors that may have an impact on self-sufficiency of MFIs. Then it proceeds to explore any probable trade-off that may take place if MFIs try to serve the poor--with small loans and low interests--and attain self-sufficiency at the same time. Finally, in order to examine whether there is any contradiction between high repayment rates and profitability or between high repayment rates and cost control, the determinants of loan repayment, profitability and costs have been studied simultaneously.
The study employs a unique cross-country cross-section database of 426 MFIs in 81 countries for the years 2005-2007. As a check on robustness to outliers and other unfulfilled assumptions of regression analysis, we used iteratively weighted least squares--one type of robust regression method--in all estimations. The results we obtained in this exercise are plausible and largely on par with findings of other studies. Concerning the revenue front, for instance, our estimates suggest that raising the interest rate is associated with an increased level of risk of loan delinquency after a certain range. Before that turning point, interest rates can be increased safely and this will not harm the existing sustainability situation. Results in this exercise also indicate alternative ways through which MFIs can cover high operating costs on small and somewhat unsecured loans and ensure sustainability through better financial performance while keeping their focus on the poor.
The paper proceeds as follows. Section 2 deals with various delivery models used by the sampled MFIs. Section 3 describes the data and empirical specifications. Then, Section 4 discusses the estimation results. Finally, concluding remarks are given in Section 5.
The Global Microfinance Delivery Methods
Diverse methods to provide microfinance services globally have divided MFIs broadly into four categories: individual-based lenders, individual and solidarity-group-based lenders, solidarity-group lenders, and village banks.
Cull, Demirguc-Kunt and Morduch (2007) note that individual-based lenders use standard bilateral lending contracts between a lender and a single borrower. Liability for loan repayment rests with the individual borrower solely, with a few minor exceptions. Here, requirements, such as collateral, are less stringent than in standard banking contracts. Solidarity-group lenders employ contracts based on joint liability implemented with 'solidarity groups.' Individual member is given the loan, but if any group member fails to repay, the group as a whole shoulders responsibility. For example, the Grameen model, initiated by the Grameen Bank in Bangladesh, is the pioneering solidarity group lending method. Ac cording to de Crombrugghe, Tenikue and Sureda (2008), it corresponds to a lending method where the institution lends to standardized and fellow-feeling groups of five individuals. They organize weekly meetings and saving is mandatory for members. Credit is not given to all members concurrently, but all hope to have their turn and stand for each other's obligations. The groups are created under supervision of the MFI, according to a well-defined structure to facilitate access to microfinance services (Schreiner, 2002). Again, some MFIs are run using both individual- and solidarity-based methods. Finally, village bank type lending was mainly pioneered by the Foundation for International Community Assistance (FINCA), another microfinance giant. Each branch of village bank lenders has a single and large group and this type of lenders exercise some extent of autonomy (Cull, Demirguc-Kunt and Morduch, 2007).
Table 1 describes the distribution of sampled MFIs across aforesaid four diverse delivery models in six developing regions in the world--EAP, EECA, LAC, MENA, SA and SSA. (1) Loan delivery methods differ across different developing regions. For instance, in the dataset no MFI in the EECA region operates on solidarity-group-based methodology and there is no village bank in the MENA region. Individual-based lending and village banking predominate in the LAC region. In SA, MFIs operating on group-based lending mechanism outnumber other delivery methods.
According to the summary statistics in Table 1 and delivery models in Table 2, the regional distribution of sampled MFIs is reasonably balanced. Out of a total of 426 MFIs, 13% of MFIs come from the EAP and another 13% from institutions from the SA region, while institutions from the EECA and the SSA regions comprise 15% and 17% of the sampled MFIs respectively. The MENA region boasts a comparatively small and new microfinance industry and contributes only 6% of the sample. About 35% of sampled MFIs, the highest, come from the LAC region.
Table 3 presents the summary statistics by lending types and demonstrates that the patterns of average revenue and costs vary systematically across various loan delivery methods. In the sample, borrowers from the village banks face the highest (46%) average interest rates while individual-based lenders charge about 29%, the lowest. However, village banks bear the highest rate of operating expenses per dollar lent and, also, labor costs to assets in village banks are the highest among all lending mechanisms. Again, the average loan balance (adjusted by GNI per capita) of these lenders is on the lower side and next to the solidarity-group lenders. Thus, when extending smaller sized loans to poorer clients in remote and inaccessible locations, where the village banks are generally stationed, costs get much higher for village banks. In the sample, revenues outweigh costs of village banks and, as a result, a positive, though very low, margin on the return on assets (0.01) is ensured. The average interest rates charged by either individual-based lenders, or lenders who follow both individual-based and solidarity-group-based methods, are the lowest in this study. Lenders who completely depend on solidarity-group-based methods charge interest rates that are a little higher (34%). In all the cases, however, revenue outweighs expenses and all types of lenders have managed to get a positive rate of return on assets on average.
Loans offered by the solidarity-based lenders are the lowest ($140). Again, village banks deliver smaller sized loans (on average $220) and loans delivered by, for example, individual-based lenders are extremely large ($2130). Microfinance literature suggests that small-sized loans are typically used by poorer borrowers and, therefore, smaller loans indicate better depth of outreach. In that sense, the solidarity-based lenders do far better in reaching the poor, earning an average of 34% of interest on average. However, in terms of another measure of depth of outreach, percentage of women borrowers, village banks perform better, which is evident from the fact that 85% of the borrowers of village banks are women. Solidarity-based lenders reach the women even better (90%). However, this is not quite favorable for the individual-based lenders and only 53% of their borrowers are women.
Data and Empirical Issues
The Empirical Model
The empirical model assumes that several factors may affect MFI performance significantly. Therefore, the objective of the benchmark regressions is to investigate the determinants of several performance indicators of MFIs--profitability or sustainability, repayment status or portfolio quality and cost indicators--primarily to explore why some MFIs perform better than others. The empirical model is:
Y = [alpha] + [beta]'[X.sub.i] + [gamma]'[Z.sub.i] + [mu]'[CV.sub.i] + [u.sub.i] (1)
The dependent variable Y is either a measure of portfolio quality or financial self-sufficiency or cost; X is a set of two baseline variables that are always included in the regression; Z is the variable of interest and CV is a set of two control variables drawn from a pool of possible variables theoretically or empirically linked to financial self-sufficiency. Both Z and CV are drawn from the same pool of variables. [beta]', [gamma]' and [mu]' are the vectors of parameters and u is the usual error term. The index i represents a single observation.
Summary statistics in Table 1 show that although the means are within expected range in most of the cases, there are still wide gaps between the minimum and maximum values for some of the variables. Besides, we suspect that some of the fundamental assumptions of regression analysis--such as, independence of observations, non-influence of outliers and normality, and homoskedasticity of the residual distribution--are not adequately fulfilled in our analysis after taking care of almost all available measures including transformation of variables and so on. Although these are very common in any empirical exercise, there are ways by which one can strengthen the model against unruly data. Robust regression analysis provides an alternative to ordinary least squares when these fundamental assumptions are unfulfilled--fully or partially. Therefore, as a check on robustness to outliers and other unfulfilled assumptions the iteratively weighted least squares--one type of robust regression method--has been used in all estimations. (2)
Managerial, financial and portfolio data for 426 MFIs in 81 countries have been analysed in this paper. Over 93% of the observations pertain to the year 2007, and only a handful of them belong to 2006 (5.87%) and 2005 (0.94%) where 2007 data were unavailable. The database was primarily constructed from the Microfinance Information eXchange Inc. (MIX), the largest global, web-based microfinance information platform promoting transparent data dissemination in the industry. Data were also gathered from the Micro Banking Bulletin (MBB), in which data are standardized and aggregated in ratio format broken down by selected peer groups. As submission to the MBB is voluntary and carries with it an impression of prestige, it is assumed that MFIs submitting data tend to be among the higher performers in their respective peer groups. In order to construct the dataset, the primary objective was to obtain a cross-section of data of as many MFIs as possible from all delivery models and legal statuses. Depending on the level of disclosure, the MIX Market uses 'diamonds' to rank MFI data where a rank of the highest of five-diamonds means the best quality. In this analysis, about 95% of the sampled MFIs are ranked with four and five diamonds, which means that mostly data from only the audited financial statements have been used. Besides, the selection criteria also required MFIs to have available data on all variables, at least 2,000 active borrowers and a satisfactory size of assets as on 2007 (or 2005 and 2006 as applicable).
Portfolio quality. Two at-risk ratios, PAR30 and PAR90, capture the accounting convention that loans exceeding 30 days or 90 days overdue respectively cause an unacceptably high risk of non-repayment. In portfolio quality regressions, PAR30 and PAR90 have been used as the outcome variables basically to explore their determinants. However, in sustainability regressions, one outstanding loan portfolio ratio (PAR30) enters as an explanatory variable and is hypothesized to be inversely associated with financial self-sufficiency.
Sustainability. Micro Banking Bulletin defines OSS as the ratio of unadjusted operating income to unadjusted operating expenses that includes financial expense, loan loss reserve expense and operating expense. OSS measures how well the MFI can cover its costs through operating revenues. We used OSS as a proxy for MFI sustainability, and not FSS, mainly because of two reasons. First, as de Crombrugghe, Tenikue and Sureda (2008) note, simply by applying the inflation rate to own funds, FSS includes a cost for own funds of the MFI. But it is difficult to estimate the equity for own funds of an MFI and the opportunity cost of such funds. Second, Hartarska and Nadolnyak (2007) note that since donors monitor MFIs' OSS and can exercise long-term control due to increased competition for donations, OSS is likely to be a more reliable approximation of financial sustainability of an MFI than FSS. Additionally, as another proxy for MFI sustainability, we used ROA in our models. Being the ratio of net operating income after taxes over average total assets, ROA measures how well the MFI uses its total assets to generate returns.
Cost indicators. The CPB ratio measures the value of total monetary and in-kind inputs required to produce a given level of output, as measured by borrowers. A higher ratio indicates that more money and in-kind contributions are required to produce a given number of loan clients. CPB, used as an explanatory variable in sustainability regressions, is hypothesized to be inversely associated with MFI sustainability. OED ratio is defined as the ratio of operating cost over gross loan portfolio. It is important to differentiate OED from CPB because CPB is in terms of borrowers and OED is expressed in dollar terms. However, in sustainability regressions, OED is used as an explanatory variable and is hypothesized to be inversely associated with financial self-sufficiency. Similar definitions and hypothesised link also apply for the TCD ratio.
Explanatory Variables. Several other variables meet the criteria for inclusion in Z and CV. These independent variables can be grouped into some broad categories as described below.
Yield. Following Christen (2001), the variable yield is always included in X in equation (1) as a determinant of MFI performance. We utilised the nominal (unadjusted for inflation) portfolio yield as a proxy for the interest rate charged on loans which is widely used in the literature. An increase in the (nominal) portfolio yield is hypothesized to be associated with a deteriorating portfolio quality and an increased level of financial self-sufficiency. This variable is a better proxy because the MFI can make a decision on what nominal interest rate to charge, but not on the real interest rate. Real interest rate, although adjusted for inflation, becomes clear only ex post.
Breadth and depth of outreach. NAB is an indicator of both breadth of outreach and scale. Conventionally, scale is inversely related with costs and positively with profitability, because the fixed costs of production are amortized across a larger number and value of outputs. The number of borrowers is thus hypothesized to be positively associated with financial self-sufficiency. Average loan size (LBB)--defined as the average gross loan portfolio divided by NAB--is a proxy for depth of outreach. Smaller loans generally indicate better depth of outreach (Schreiner, 2002), and LBB is hypothesized to be positively associated with financial self-sufficiency. LBBGNI, another proxy for depth of outreach, normalizes the variable for different price and income levels found in different countries, thereby making cross-country comparisons more valid. This variable is also hypothesized to be positively associated with financial self-sufficiency.
Productivity. BPSM--the staff productivity ratio--measures the total number of staff required to produce a given level of output, as measured by borrowers. A higher ratio indicates that fewer staff is required to produce a given number of borrowers. Clearly, it is hypothesized to be positively associated with financial self-sufficiency. BPLO, or the loan officer productivity ratio, measures the number of loan officers required to produce a given level of output--the number of borrowers. A higher ratio means that fewer loan officers are required to produce a given number of borrowers. BPLO is also hypothesized to be positively associated with financial self-sufficiency.
Details on variables and their summary statistics are presented in Table 1.
Results and Discussions
The Role of Interest Rate Revisited
Interest rates charged by MFIs have important implications for the agency theory. Generally, owing to the agency problems of adverse selection and moral hazard, interest rates charged to the borrowers affect the financial performance of MFIs mainly in two ways. First, interest rates affect loan delinquency rate and, second, they have impacts on overall financial sustainability level. Demand forces may also affect this relationship between interest rates and sustainability: excessively high interest rates may reduce credit demand and thus profits (Cull, Demirguc-Kunt and Morduch, 2007). To support this view, Karlan and Zinman (2008) and Dehejia, Montgomery and Morduch (2009) refute the common claim that the poor borrowers are interest rate inelastic, which suggests that the demand forces are truly in action.
Table 4 demonstrates that interest rates affect MFIs' loan repayment significantly and negatively, suggesting, all else equal, deterioration of portfolio quality at relatively low interest rates. This result is similar to one aspect of those obtained by Ahlin and Townsend (2007), but quite opposite to those of Cull, Demirguc-Kunt and Morduch (2007) in that their results only apply for the individual-based lenders.
These crude signs of regression coefficients, however, cannot explain all underlying intricacies and the undercurrents warrant some elaboration. Figure 1 and Figure 2 portray the relationship between loan delinquency and yield (estimated from models 1 and 3 of Table 4) and plot yield against the predicted values of PAR30 and PAR90. Clearly, the turning point is around an interest rate of 30 percent in our sample. U-shaped relationships confirm that PAR30 (or PAR90) declines first as interest rates increase and then increases as interest rates go further beyond the turning point. (3) Drawing mainly on de Crombrugghe, Tenikue and Sureda (2008), a U-shaped association can plausibly be explained by either selection effects or repayment effects. They match with the standard adverse selection effects of high interest rates as described in the literature.
Arguments for the selection effects and repayment effects proceed as follows. De Crombrugghe, Tenikue and Sureda (2008) note that imposing low interest rates first destabilizes MFIs' screening efforts and then weakens the application of self-selection norms of the borrowers. Consequently, owing to a lack of critical foresight of repayment cost and ability, MFIs may lend to less-creditworthy, or non-creditworthy, borrowers. (4) As well, repayment effects are justified since MFIs that fancy targeting the very poor are generally flexible on selection criteria, which adversely affect clients' repayment potentials. Besides, poor borrowers consider loans as donation or gift at low interest rates, and this adversely affects their willingness to repay the loan. At the optimized interest rates (the rate where again the curve is bending upward in Figure 1) clients should want to repay most. On the contrary, at high interest rates lenders' probability of facing default increases again. However, justifiably, some extent of obscurity in terms of causality remains in this connection that needs further investigation.
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
As expected, portfolio yield is positively and significantly related with both measures of financial self-sufficiency--OSS and ROA--across all models in Table 5 and Table 6 which suggests that higher interest rates improve financial performance. The coefficients for the yield-squared variable are negative and generally significant, which clearly confirms a nonlinear association that MFI sustainability first increases, and then decreases, with interest rates.
Figure 3 and Figure 4 demonstrate that the projected connection between portfolio yield and sustainability (OSS or ROA) is clearly hump shaped (quite justified with a U-shaped relationship between yield and the loan default rates), suggesting that sustainability of MFIs first increases and then decreases with rising interest rates. Clearly, MFIs' sustainability reaches its maximum at around a yield rate of 35% and 60% in OSS and ROA regressions respectively, which seem very high to call for ethical and other, including mission drift, concerns.
[FIGURE 3 OMITTED]
[FIGURE 4 OMITTED]
In order to choose first-rate borrowers and supervise them properly, MFIs indeed require high-quality and dedicated personnel with proper social motivation. This can improve MFIs' efficiency and productivity and the BPLO ratio--i.e., number of borrowers served per personnel (or per loan official)--is very important in this regard. Justifiably, a low ratio ensures better selection and supervision of borrowers. In contrast, higher BPLO ratios negatively affect MFIs' overall performance. The summary statistics in Table 1 show that BPLO is only about 278 in our sample, which is quite on the low side. For example, de Crombrugghe, Tenikue and Sureda (2008) refer that the acceptable range should be 250350 in the Indian context.
The negative significant BPLO coefficients in Table 4 are as expected. These suggest that if the number of borrowers could be increased in terms of staff members engaged in loan services, loan repayment rates can be improved. Also, an increase in BPLO lowers costs, as this productivity indicator's impact on other cost measures (OED, TCD and CPB) results in significantly negative coefficients throughout Table 7. Together, these results suggest that without damaging MFIs' repayment performance there are scopes for either increasing the number of borrowers or decreasing the number of staff members, or both. These findings are quite similar to those of de Crombrugghe, Tenikue and Sureda (2008).
The BPLO ratio also clearly affects sampled MFIs' sustainability scenarios. The positive significant BPLO coefficients in sustainability regressions in Table 5 and Table 6 (models 1 and 2) indicate that increasing the number of active borrowers for a given number of loan officers would contribute to the profitability of MFIs in general. In a few of our model specifications in sustainability regressions (model 3 in both Table 5 and Table 6) we included the number of active borrowers and the number of staff members separately instead of using their ratio. This confirmed the results. We found a positive significant effect of the number of borrowers and a negative significant effect of the number of personnel on profitability (OSS and ROA). The significant contributions of BPLO to portfolio quality, profitability and costs indicate that MFIs in our sample could gain from either staff reduction or from an increase in the number of borrowers. This is consistent with the fact that most MFIs in our sample use group mechanisms (solidarity or some form of solidarity-individual and solidarity mixed) and have room for raising the BPLO ratio within acceptable range as mentioned above (250-350). An alternative to downsizing staff members is to increase the number of active borrowers keeping the volume of staff members fixed. However, this calls for further investigation, including their implications for economies of scale, as the higher number of borrowers per MFI could not only contribute to the scale economies on the staff involved but also on other fixed costs.
Size of Loans
The insignificant and non-monotonic coefficients for LBBGNI in loan delinquency regressions in Table 4 suggest that repayment pattern is by and large unaffected by average loan size. (5) However, the relationships between LBBGNI and OSS, or LBBGNI and ROA, are generally positive significant. Plotting predicted values of OSS against LBBGNI variable produces a hump-shaped relationship between them in Figure 5. Clearly, this hump shape has the implication that MFIs may increase the average loan balances but only up to the point where it starts bending downwards. After that point, profitability is negatively affected by loan size.
[FIGURE 5 OMITTED]
In average cost regressions in Table 7, the elasticity of OED to the loan size variable and the same for the TCD variable are negative significant and on the high side (0.36 and 0.37 respectively), suggesting a cost-decreasing trend. However, in CPB regressions, elasticities of CPB to the loan size variable are positive significant, suggesting increasing costs due to higher loans. As larger loans basically go to individual borrowers, it is apprehended that MFIs expend more on client selection and loan supervising purposes. In our sample, we have seen earlier that the size of loans does not affect profitability. However, these trends in cost increase (per borrower costs) as well as cost decrease (costs per dollar lent) in average cost regressions show that the cost reduction per dollar (elasticity=0.37) takes over the cost increase per borrower (elasticity=0.26). This is contrary to conventional wisdom: increased loan size should minimize per borrower costs as they basically are extended to richer clients.
[FIGURE 6 OMITTED]
[FIGURE 7 OMITTED]
We see in Column 4 of Table 7 that, as expected, the elasticity of CPB to the number of active borrowers is negative, significant and very high (0.63). The negative sign and high significance of the coefficient clearly substantiates that a rise in number of active borrowers (NAB) would help in mitigating the average operating cost. The projected declining average cost against (log) number of active borrowers is plotted in Figure 8. This negative effect of NAB on cost clearly matches with positive significant effect of log of NAB on sustainability measures--OSS and ROA--as presented in Table 5 and Table 6. This has the implication that increasing the number of borrowers would improve the overall financial sustainability position of an MFI. However, regarding another measure through which economies of scale can be assessed--gross loan portfolio--we found no significant impact on the overall financial sustainability of MFIs.
The Delivery Models
Solidarity Groups. Banerjee, Besley and Guinnane (1994) suggest that information flows are more useful in bigger groups to reduce costs, such as monitoring costs. This theoretical claim is substantiated in our empirical results. The coefficient for the solidarity lender dummy is negative in CPB regressions. This suggests that in comparison with the control group (MFIs that operate on individual methods or both individual and solidarity methods), MFIs spend relatively less per borrower on average when they lend to a solidarity group. However, this is insignificant.
Using a shortened model, we found that the coefficient for the solidarity dummy variable is negative and insignificant in the OED and TCD regressions. Again, in profitability or sustainability regressions (OSS and ROA), the coefficients are positive as expected, but again insignificant. The negative significant coefficient for the solidarity group dummy variable both in PAR30 and PAR90 regressions can be explained quite comfortably--for solidarity groups the loan repayment performance is far stronger and better. However, it is unclear whether MFIs get better sustainability through this, as the coefficient for solidarity dummy lost significance in profitability regressions.
Village Banks. The positive significant coefficients for the village bank dummy in both OED and TCD specifications in Table 7 suggest that compared to individual loan methodology, the control category, MFIs spend relatively more per dollar when they apply the village bank style loan method. This supports the conventional claim that, especially for village banking operations, microfinance involves higher costs, which compel them to charge higher interests. In loan delinquency regressions, the coefficient for the village bank dummy is always negative and generally significant as expected, which, like the solidarity groups, indicates that village banks' repayment performance is very strong. This is true for both measures of loan delinquency--PAR30 and PAR90. Again, when we control for other variables, including costs, gender of the borrower and MFI age, the coefficient still retains significance with a negative sign. This confirms that village banks' repayment performance is also strong.
Women as a Target Group
The share of loans extended to women borrowers is a reliable indicator for measuring 'depth of outreach' of an MFI. In PAR30 (not in PAR90) regressions, the coefficient for women borrowers is negative and significant. This suggests that gender of the participants is important indeed and women borrowers' repayment performance is significantly better. When included in the profitability regressions this variable displayed no significant pattern, although the negative sign was retained. It is thus not sure that targeting women adds much per se to the financial performance of an MFI. However, empowerment of women is a different issue that should have various dimensions. It is apprehended, however, that increased women participation is very limited in what it can do to bring improvement on the sustainability front of MFIs in general. Separate future studies should aim to address this relevant issue of concern.
[FIGURE 8 OMITTED]
Scaling-up and Experience of the Institution
Although the MFI-age variable has been included in loan delinquency, sustainability and average cost regressions only in a few specifications was it significant. In PAR30 and PAR90 loan delinquency regressions, positive significant coefficients for the MFI-age variable suggest that as MFIs grow older the risk of loan delinquency rises. These results are similar to those of Cull, Demirguc-Kunt and Morduch (2007) and support the view that defaults may not be uncommon for old MFIs, although, with the growing of age of an MFI, it should have attained the necessary level of efficiency to reduce its loan delinquency rate. One more plausible explanation for that is as follows: Due to scaling-up effects, old MFIs are more prone to loan defaults due to inability to repay (selection effects) and unwillingness to repay (repayment effects) of the borrowers as a consequence of flawed screening mechanism.
The impacts of MFI experience on overall financial performance of MFIs are positive as expected in sustainability regressions. However, these are insignificant. One reason behind this is that the age variable may not have enough dynamism in itself to evenly influence the profitability scenario of an MFI when various types of MFIs are sampled. As expected, both in OED and TCD specifications in Table 7, MFI age negatively and significantly affects MFIs' average cost scenarios, clearly indicating the fact that operating expenses should decrease overtime. Effects related with set-up costs should affect cost structure more favorably over time. Besides, economies of scale owing to size of loans and portfolios could increase with age and, consequently, OED declines. These results are similar to previous studies, including Cull, Demirguc-Kunt and Morduch (2007).
The regional dummy variables do explain some variation in loan delinquency as well as in financial performance. In loan delinquency regressions in Table 4, we found negative and highly significant association between regional dummies and both measures of loan defaults--PAR30 and PAR90--for institutions from EECA and MENA. A negative relation indicates the fact that MFIs in these two regions are performing well in terms of loan delinquency in comparison with the control category, LAC. In sustainability regressions, institutions from EECA and MENA region outperformed those from other regions in terms of OSS. But in terms of ROA, only institutions from MENA region quite convincingly outperformed institutions from other regions. Only in model 3, institutions from the EECA region performed well operationally. So is the case for institutions from SSA in model 1. In average cost regressions, institutions from MENA and SA regions have negative and significant coefficients indicating better performance of MFIs in these regions than those of other regions, as OED is significantly lower in these regions compared with the LAC region. Institutions from the EAP region have a negative significant relation with OED, but a positive significant relation with CPB.
This paper looked at factors which may have impacts on MFIs' profitability or sustainability, portfolio quality and cost indicators to explore whether there is any contradiction between high repayment rates and profitability or between high repayment rates and cost control. Results are reliable and largely on par with findings of other studies. Concerning the revenue front, our estimates suggest that raising the interest rate is associated with an increased level of risk of loan delinquency after a certain range. Before that turning point, interest rates can be increased safely and this will not harm the existing sustainability situation. Therefore, MFIs must make choices between their long-run mission to attain sustainability and ensuring the highest possible repayment rate. A very high rate of interest is as bad as a very low one. Both of them are prone to activating adverse selection or moral hazard effects. The policy implications for these findings finally depend on the quest of what the respective MFIs actually fancy. If they give priority on high repayment rates, the existing rates of interest seem to do well. However, if their priority is for becoming operationally fully self-reliant covering all the costs, they may try higher interest rates but there is a risk of default after a turning point, as specified above.
The significant contributions of the BPLO ratio in terms of portfolio quality, profitability and reduction of costs indicate that MFIs in our sample could gain either from staff reduction or from increasing the number of borrowers. This is consistent with the fact that most MFIs in the sample have room to raise the BPLO ratio within acceptable standards and also that most of them use a solidarity-based loan delivery method. Results show that unlike small loans of a solidarity-group-based method, large loans require more individual attention and supervision. Again, increasing loan-size works only up to a certain point to reduce cost. Therefore, MFIs have to be very careful while deciding on the optimal loan size in order to have a good impact on their expenditure front. Small loans are indicative of better depth of outreach and our data suggest that small loan size can be retained. Increasing the BPLO ratio seems to be the most promising way to reduce costs, especially in group-based delivery models. This way of scaling-up will leave the repayment performance unhurt despite a likely ease of monitoring in group lending. Further, if economies of scale are obtained through this, it will be then primarily by extending the breadth of outreach, not by abandoning the quality or depth of outreach, that is, not by de-emphasizing the focus on the poor.
Anecdotal evidence suggests that the overall sustainability of MFIs can be ensured either by increasing the average loan amount or through raising the loan price, or both. But this increases the concerns for mission drift. However, in line with de Crombrugghe, Tenikue and Sureda (2008), results in this study confirm the availability of alternative ways through which MFIs' can cover high operating costs on tiny and somewhat unsecured loans, and ensure sustainability through better financial performance while keeping their focus on poor borrowers. Needless to mention, gaining self-reliance in this way will certainly improve the images of MFIs at a time when these institutions are increasingly blamed for caring much about profit to be sustainable and less for the extremely poor.
The author would like to thank two anonymous referees for very helpful comments. Remaining errors are, of course, mine.
Ahlin, C., and R.M. Townsend. 2007. "Using Repayment Data to Test Across Models of Joint Liability Lending." Economic Journal 117(517): F11-F51.
Armendariz de Aghion, B., and J. Morduch. 2005. The Economics of Microfinance. MIT Press: Cambridge MA. Banerjee, A.V., T. Besley, and W. Guinnane. 1994. "The Neighbor's Keeper: The Design of a Credit Cooperative with Theory and a Test." The Quarterly Journal of Economics 109(2): 491-515.
Christen, R. 2001. "Commercialization and Mission Drift: The Transformation of Microfinance in Latin America." CGAP Occasional Paper no. 05.
Cull R., A. Demirguc-Kunt, and J. Morduch. 2007. "Financial Performance and Outreach: A Global Analysis of Leading Microbanks." Economic Journal 117(517): F107-F133.
De Crombrugghe, A., M. Tenikue, and J. Sureda. 2008. "Performance Analysis for a Sample of Microfinance Institutions in India." Annals of Public and Corporate Economics 79(2): 269-299.
Dehejia, R., H. Montgomery, and J. Morduch. 2009. "Do Interest Rates Matter? Credit Demand in the Dhaka Slums." Financial Access Initiative, working paper.
Hartarska, V., and D. Nadolnyak. 2007. "Do Regulated Microfinance Institutions Achieve Better Sustainability and Outreach? Cross-country Evidence." Applied Economics 39(10-12): 1207-1222.
Hermes, N., and R. Lensink. 2007. "The Empirics of Microfinance: What Do We Know?" The Economic Journal 117(517): F1-F10.
Karlan, D., and J. Zinman. 2008. "Credit Elasticities in Less-Developed Economies: Implications for Microfinance." American Economic Review 98(3): 1040-1068. Robinson, M. 2001. The Microfinance Revolution. The World Bank: Washington DC.
Schreiner, M. 2002. "Aspects of Outreach: A Framework for the Discussion of the Social Benefits of Microfinance." Journal of International Development 14(5): 591-603.
Ashim Kumar Kar, Hanken School of Economics, Helsinki, Finland
(1) Table 1 provides elaborations on these acronyms of the developing regions as classified by the World Bank.
(2) Ordinary least squares with Huber-White standard errors and Median regressions--one type of quantile regression--are also run, but all of them came up with mostly similar results.
(3) We estimated other models too, but all of them give a similar shape.
(4) Such problems seem to have plagued many state development banks in the past (Armendariz de Aghion and Morduch 2005; chapter 1).
(5) However, we observe positive significant links between loan size and two other delinquency measures-write-off ratio (WOR) and loan loss reserve ratio (LLRR)--using a slightly different dataset, which indicates that bigger amounts of loans lead to higher risks in repayment.
For further information on this article, contact:
Ashim Kumar Kar, Department of Economics, Hanken School of Economics, Arkadiankatu 7 (Eco nomicum), FIN-00100 Helsinki, Finland Tel: +358 (0)44 208 2885 E-mail: email@example.com
Table 1. Variable description and summary statistics Variable Description Observations Mean Panel 1: Performance indicators OSS: Operational self-sufficiency 426 1.20 ROA: Return on assets 424 0.03 PAR 30: Portfolio at risk, 30 days 402 0.05 past due PAR 90: Portfolio at risk, 90 days 397 0.03 past due OED: Operating Expenses per dollar 426 0.23 lent CPB: Cost per borrower 425 4.46 TCD: Total cost per dollar lent 425 0.23 Panel 2: Explanatory variables Yield 425 0.33 LBB: Loan balance per borrower 426 1054.48 LBB-GNI: Loan balance per 426 67.07 borrower, adjusted by GNI per capita BPLO: Borrowers per loan officer 422 277.71 MFI-age 425 14.33 Size: Log of total assets 426 16.86 NAB: Number of active borrowers 426 125010.40 Personnel 425 779.16 CPB: Cost per borrower 425 4.46 LGLP: Log of gross loan portfolio 426 16.57 OED: Operating Expenses per dollar 426 0.23 lent For-profit status 426 0.38 Offers savings services dummy 426 0.54 Village Bank lender dummy 426 0.12 Solidarity lender dummy 426 0.09 Labour costs to assets 424 0.09 Loans to assets ratio 426 0.77 Women borrowers 390 0.66 EAP: East Asia and the Pacific 426 0.13 EECA: Eastern Europe and Central 426 0.15 Asia LAC: Latin America and the 426 0.35 Caribbean MENA: Middle East and North Africa 426 0.06 SA: South Asia 426 0.13 SSA: Sub-Saharan Africa 426 0.17 Variable Description Std. Dev. Minimum Maximum Panel 1: Performance indicators OSS: Operational self-sufficiency 0.30 0.24 3.36 ROA: Return on assets 0.06 -0.25 0.31 PAR 30: Portfolio at risk, 30 days 0.06 0.00 0.45 past due PAR 90: Portfolio at risk, 90 days 0.04 0.00 0.30 past due OED: Operating Expenses per dollar 0.18 0.01 1.16 lent CPB: Cost per borrower 1.11 1.10 7.74 TCD: Total cost per dollar lent 0.22 0.01 2.88 Panel 2: Explanatory variables Yield 0.16 0.01 1.07 LBB: Loan balance per borrower 1953.05 31 31596 LBB-GNI: Loan balance per 109.85 2.1 1430.28 borrower, adjusted by GNI per capita BPLO: Borrowers per loan officer 173.56 24 2183 MFI-age 8.61 2 52 Size: Log of total assets 1.59 13.48 22.43 NAB: Number of active borrowers 549482.40 2292 6397635 Personnel 3045.16 7 38545 CPB: Cost per borrower 1.11 1.10 7.74 LGLP: Log of gross loan portfolio 1.61 12.88 21.83 OED: Operating Expenses per dollar 0.17 0.01 1.16 lent For-profit status 0.49 0 1 Offers savings services dummy 0.50 0 1 Village Bank lender dummy 0.32 0 1 Solidarity lender dummy 0.28 0 1 Labour costs to assets 0.06 0.00 0.51 Loans to assets ratio 0.15 0.08 1.02 Women borrowers 0.25 0.03 1 EAP: East Asia and the Pacific 0.34 0 1 EECA: Eastern Europe and Central 0.36 0 1 Asia LAC: Latin America and the 0.48 0 1 Caribbean MENA: Middle East and North Africa 0.24 0 1 SA: South Asia 0.34 0 1 SSA: Sub-Saharan Africa 0.38 0 1 Source: Author's calculations, based on data collected from the Microfinance Information eXchange (MIX), Inc. Table 2. Microfinance delivery models by region Developing region Delivery model EAP EECA LAC MENA SA SSA Total Individual 15 27 73 3 6 8 132 Individual-Solidarity 26 38 53 22 21 48 208 Solidarity 9 0 2 2 21 3 37 Village Bank 6 1 20 0 9 13 49 Total 56 66 148 27 57 72 426 Source: Author's calculations, based on data collected from the Microfinance Information eXchange (MIX), Inc. Table 3. Summary statistics by lending type Individual Ind.-Solidarity Variable Mean SD Mean SD OSS 1.22 0.26 1.20 0.29 ROA 0.03 0.05 0.03 0.06 PAR30 0.05 0.05 0.05 0.06 PAR90 0.03 0.03 0.03 0.04 LBBGNI 4.04 1.06 3.62 1.07 MFI-age 16.23 10.48 13.64 7.62 Size 17.80 1.63 16.61 1.39 For profit dummy 0.58 0.49 0.33 0.47 Yield 0.29 0.15 0.32 0.13 Labour costs to assets 0.07 0.05 0.09 0.05 LBB (in thousands) 2.13 3.09 0.73 0.78 Women borrowers 0.53 0.20 0.65 0.24 Personnel (in thousands) 0.91 3.46 0.61 2.53 BPLO 253.24 165.06 276.05 145.54 OED 0.17 0.13 0.23 0.15 CPB 257.33 286.42 119.27 104.89 Borrowers (in thousands) 98.19 325.01 103.30 473.89 Loans to assets ratio 0.77 0.15 0.78 0.14 Solidarity Village bank Variable Mean SD Mean SD OSS 1.19 0.24 1.12 0.42 ROA 0.04 0.05 0.01 0.10 PAR30 0.04 0.07 0.04 0.05 PAR90 0.03 0.05 0.03 0.03 LBBGNI 2.56 0.71 2.72 0.97 MFI-age 14.03 8.32 12.41 6.24 Size 16.38 1.46 15.80 1.09 For profit dummy 0.22 0.42 0.16 0.37 Yield 0.34 0.17 0.46 0.22 Labour costs to assets 0.10 0.06 0.15 0.10 LBB (in thousands) 0.14 0.07 0.22 0.12 Women borrowers 0.90 0.22 0.85 0.17 Personnel (in thousands) 1.83 5.21 0.35 0.59 BPLO 291.06 102.83 341.63 296.26 OED 0.25 0.19 0.39 0.27 CPB 32.11 26.24 71.92 48.11 Borrowers (in thousands) 427.24 133.09 61.20 97.24 Loans to assets ratio 0.70 0.20 0.77 0.14 Source: Author's calculations, based on data collected from the Microfinance Information eXchange (MIX), Inc. Table 4. Portfolio quality estimation (robust regressions) Variable PAR30 PAR30 Yield -0.10 ** (0.034) -0.12 ** (0.037) Yield-squared 0.13 *** (0.034) 0.15 *** (0.036) LBBGNI -0.003 (0.002) 0.0003 (0.002) Log BPLO -0.01 ** (0.003) -0.005 (0.003) EAP -0.01 * (0.005) -0.02 *** (0.005) EECA -0.026 *** (0.004) -0.023 *** (0.005) MENA -0.021 *** (0.006) -0.016 ** (0.006) SA -0.005 (0.005) -0.001 (0.005) SSA 0.006 (0.005) 0.002 (0.005) Village Bank lender -0.013 ** (0.005) -0.011 * (0.004) Solidarity lender -0.018 ** (0.006) -0.01 (0.005) Offers Savings Services 0.004 (0.003) 0.005 (0.003) Loans to assets ratio -0.03 ** (0.01) -0.015 (0.011) Labour costs to assets -0.04 (0.04) -0.03 (0.05) ratio For-profit dummy -0.006 (0.003) -0.007 * (0.003) Women borrowers -0.016 * (0.007) Log MFI-age 0.006 * (0.003) Log OED 0.008 (0.005) Constant 0.14 *** (0.024) 0.07 * (0.03) Observations 397 364 Adjusted R-squared 0.22 0.27 Variable PAR90 PAR90 Yield -0.04 (0.025) -0.07 ** (0.027) Yield-squared 0.03 (0.027) 0.06 * (0.027) LBBGNI -0.002 (0.001) -0.0009 (0.001) Log BPLO -0.008 *** (0.002) -0.006 ** (0.002) EAP -0.004 (0.003) -0.01 ** (0.004) EECA -0.02 *** (0.003) -0.015 *** (0.003) MENA -0.014 ** (0.005) -0.011 ** (0.004) SA -0.006 (0.004) -0.005 (0.004) SSA 0.002 (0.003) -0.002 (0.003) Village Bank lender -0.009 * (0.004) -0.006 * (0.003) Solidarity lender -0.012 ** (0.004) -0.006 (0.004) Offers Savings Services 0.005 (0.002) 0.004 (0.002) Loans to assets ratio -0.03 *** (0.007) -0.014 (0.008) Labour costs to assets 0.006 (0.03) -0.026 (0.033) ratio For-profit dummy -0.006 ** (0.002) -0.006 ** (0.002) Women borrowers -0.008 (0.005) Log MFI-age 0.005 * (0.002) Log OED 0.008 * (0.003) Constant 0.10 *** (0.018) 0.06 ** (0.02) Observations 392 358 Adjusted R-squared 0.20 0.25 All models are estimated via robust regression (weighted least squares) method, with heteroskedasticity consistent robust standard errors. Standard Errors are given in the parentheses. * Significant at 10% (p<0.05); ** significant at 5% (p<0.01); *** significant at 1% (p<0.001). Table 5. Sustainability regressions (robust regressions) Variable OSS OSS Yield 0.51 * (0.25) 2.47 *** (0.24) Yield-squared -0.29 (0.27) -1.37 *** (0.24) Log LBBGNI 0.07 *** (0.01) 0.005 (0.01) Log BPLO 0.13 *** (0.02) 0.04 * (0.02) EAP 0.05 (0.04) -0.001 (0.04) EECA 0.05 (0.03) 0.07 * (0.03) MENA 0.18 *** (0.05) 0.07 (0.04) SA -0.04 (0.04) -0.14 *** (0.04) SSA -0.13 *** (0.03) -0.03 (0.04) Village Bank lender -0.07 * (0.04) -0.05 (0.03) Solidarity lender 0.07 (0.04) 0.02 (0.033) Loans to assets ratio 0.38 *** (0.076) -0.076 (0.074) Labour costs to assets ratio 0.06 (0.30) For-profit dummy 0.008 (0.02) Women borrowers -0.02 (0.05) Log of MFI age 0.001 (0.017) Log OED -0.45 *** (0.03) Log CPB -0.037 * (0.015) Log PAR30 Log NAB Log number of personnel Constant -0.19 (0.17) 1.89 *** (0.20) Observations 421 383 Adjusted R-squared 0.23 0.60 Variable OSS Yield 0.65 * (0.28) Yield-squared -0.55 (0.29) Log LBBGNI Log BPLO EAP 0.08 * (0.04) EECA 0.09 * (0.03) MENA 0.17 *** (0.05) SA -0.03 (0.05) SSA -0.016 (0.036) Village Bank lender -0.066 (0.038) Solidarity lender 0.06 (0.04) Loans to assets ratio 0.33 *** (0.08) Labour costs to assets ratio For-profit dummy 0.009 (0.025) Women borrowers Log of MFI age Log OED Log CPB Log PAR30 0.06 *** (0.017) Log NAB 0.06 ** (0.022) Log number of personnel -0.10 *** (0.024) Constant -0.32 (0.22) Observations 424 Adjusted R-squared 0.21 All models are estimated via robust regression (weighted least squares) method, with heteroskedasticity consistent robust standard errors. Standard Errors are given in the parentheses. * Significant at 10% (p<0.05); ** significant at 5% (p<0.01); *** significant at 1% (p<0.001). Table 6. Sustainability regressions (robust regressions) Variable ROA ROA Yield 0.20 *** (0.04) 0.47 *** (0.05) Yield-squared -0.10 * (0.04) -0.25 *** (0.046) Log LBBGNI 0.01 *** (0.002) 0.005 (0.003) Log BPLO 0.015 *** (0.004) 0.008 * (0.004) EAP 0.004 (0.006) -0.002 (0.007) EECA 0.007 (0.005) 0.01 (0.006) MENA 0.03 *** (0.007) 0.016 * (0.007) SA -0.003 (0.006) -0.016 * (0.01) SSA -0.012 * (0.006) -0.005 (0.007) Village Bank lender 0.002 (0.006) -0.004 (0.006) Solidarity lender 0.01 (0.007) 0.008 (0.007) Loans to assets ratio 0.09 *** (0.01) 0.04 ** (0.01) Labour costs to assets ratio -0.015 (0.06) For profit dummy -0.006 (0.004) Women borrowers -0.013 (0.009) Log MFI-age 0.003 (0.003) Log OED -0.046 *** (0.006) Log CPB -0.006 (0.003) Log PAR30 Log NAB Log number of personnel Constant -0.20 *** (0.028) 0.015 (0.039) Observations 420 382 Adjusted R-squared 0.29 0.47 Variable ROA Yield 0.18 *** (0.04) Yield-squared -0.10 * (0.05) Log LBBGNI Log BPLO EAP 0.003 (0.006) EECA 0.01 * (0.006) MENA 0.02 * (0.01) SA -0.015 (0.008) SSA -0.004 (0.006) Village Bank lender -0.002 (0.006) Solidarity lender 0.007 (0.007) Loans to assets ratio 0.085 *** (0.01) Labour costs to assets ratio For profit dummy -0.002 (0.004) Women borrowers Log MFI-age Log OED Log CPB Log PAR30 0.001 (0.003) Log NAB 0.01 * (0.03) Log number of personnel -0.006 (0.004) Constant -0.15 *** (0.035) Observations 423 Adjusted R-squared 0.27 All models are estimated via robust regression (weighted least squares) method, with heteroskedasticity consistent robust standard errors. Standard Errors are given in the parentheses. * Significant at 10% (p<0.05); ** significant at 5% (p<0.01); *** significant at 1% (p<0.001). Table 7. Average cost regressions Variable Log OED Log TCD Log LBBGNI -0.361 *** (0.023) -0.374 *** (0.023) Log BPLO -0.517 *** (0.038) -0.510 *** (0.039) Village Bank lender dummy 0.258 *** (0.067) 0.290 *** (0.069) Solidarity lender dummy -0.039 (0.081) -0.058 (0.083) Log MFI-age -0.171 *** (0.037) -0.173 *** (0.037) EAP 0.175 ** (0.066) 0.139 * (0.067) EECA -0.055 (0.066) -0.091 (0.067) MENA -0.267 ** (0.087) -0.272 ** (0.088) SA -0.360 *** (0.070) -0.359 *** (0.071) SSA 0.652 (0.063) 0.620 (0.064) Log NAB Log personnel Constant 7.408 *** (0.258) 2.751 *** (0.264) Observation 421 420 Adjusted R-squared 0.59 0.59 Variable Log CPB Log CPB Log LBBGNI 0.261 *** (0.031) 0.182 *** (0.030) Log BPLO -0.525 *** (0.052) Village Bank lender dummy -0.119 (0.092) -0.115 (0.085) Solidarity lender dummy -0.207 (0.110) -0.178 (0.101) Log MFI-age -0.061 (0.050) -0.083 (0.048) EAP -0.842 *** (0.090) -0.935 *** (0.084) EECA 0.126 (0.089) 0.026 (0.083) MENA -0.675 *** (0.118) -0.610 *** (0.110) SA -1.902 *** (0.095) -1.890 *** (0.094) SSA -0.665 (0.085) -0.593 (0.079) Log NAB -0.633 *** (0.051) Log personnel 0.671 *** (0.052) Constant 7.143 *** (0.352) 7.500 *** (0.342) Observation 420 423 Adjusted R-squared 0.73 0.77 Source: Author's calculations, based on data from the MIX (Microfinance Information eXchange). Robust standard errors are in the parentheses. Significance: * p<0.05; ** p<0.01; *** p < 0.001. Log/Log regressions have been used to measure elasticities.
|Printer friendly Cite/link Email Feedback|
|Author:||Kar, Ashim Kumar|
|Publication:||Journal of Small Business and Entrepreneurship|
|Date:||Jun 22, 2011|
|Previous Article:||Entrepreneurial orientations of women business founders from a push/pull perspective: Canadians versus non-Canadians--a multinational assessment.|
|Next Article:||The role of attributional explanatory style in the perceived outcomes of entrepreneurial venture failure.|