Jeffery C. Tanner and Dr. Gary M. Woller, Romney Institute of Public Management
Microfinance is a growing instrument of poverty alleviation in the field of poverty alleviation, wherein a presumeably poor client borrows money to us as seed capital in starting or growing a small business. In continuation with the ongoing debate surrounding microcredit impact, this paper submits data and analysis showing returns to time spent in a village bank.
For this study a unique data extraction technique using multi-stage clustered, random, and stratified sampling was employed. The data were gathered via survey during the summer of 2000 from borrowers in Honduras and El Salvador associated with FINCA (Foundation for International Community Assistance) International and its local affiliates.
The analysis takes on a variety of functional forms in attempt to best model the data. This paper describes the different findings for Ordinary Least Squares (OLS), Seemingly Unrelated Regression Equations (SURE), and Least Absolute Deviation (LAD) regressions. I also attempted to model the data in an E-Gamma and a Box-Tao distribution of the error terms.
The surprising but key result coming out of this analysis is the large, negative coefficient on the es2yr and h2yr variables from the reduced models in the OLS, SURE, and LAD regressions. This highly statistically significant result implies that as a borrower stays a member of a microfinance bank longer and receives more loans (given in four month cycles), the net profits from her loan-subsidized business are reduced. It’s a puzzling discovery. From economic theory, we would naturally expect that if a person has more money to use in a business, the business will necessarily grow and become more profitable. If the person is actually losing money by receiving a loan in a village bank, they would drop out. The fact that they are still in the bank after two years implies that this figure is misleading.
The verity of this result of business being less profitable in the absolute (to say nothing of being less profitable on the margin) is further underscored by the positive coefficient on the constant in all regressions. It is of particularly high magnitude and significant beyond the 99.9% level for the reduced equation SURE model for El Salvador. As the “2yr” parameter is a dummy variable, the constant exhibits the impact of being in the bank for only one year. According to these regressions, those being in a village bank for two years can expect lower monthly net profits than those in the bank for one year. There are several explanations that could shed light on this phenomenon, most of them stemming from the fungibility of money in LDC households.
This premise of debilitating debt is supported by the negative coefficient on the “current/last” loan variable, which is highly significant and of large magnitude in almost all of the models. The negative coefficient suggests that as the size of the current loan increases, business productivity decreases. I speculate that this is due to the large repayment schedule on the debt.
The size of the total loan that a person has received has a positive effect on business profitability. I believe this is because a business is still receiving benefit from the influx of capital from previous loans, but it now does not have to pay interest for that influx. It is in essence receiving a free long term benefit at a short term cost. Thus the coefficient on the last loan variable is negative while the coefficient on the total loan variable is positive.
The education parameter is positive generally significant at the 90% level and for the Honduras models, and seems to have a joint significance with the Honduras “last loan” parameter. The implications of this positive coefficient seem straight forward: as a woman’s education increases, she is better able to make her business profitable.
The “retail” parameter employed so frequently in the El Salvador models is always negative. While it is not statistically significant in the LAD model, it is in all the other models. It seems clear from the negative sign of this parameter, that retail businesses are not as profitable as nonretail businesses. The trend of retail enterprises being less profitable becomes even more obvious when one considers the coefficient of the constant on the reduced El Salvador models. “Retail” is a dummy variable; thus the positive, often statistically significant coefficient on the constant suggests that these individuals run more profitable businesses than those who run retail businesses or who have been borrowing for two years, or both.
Evaluating the nestedness of the error distributions would be extremely interesting in trying to pin down microfinance effectiveness, though it is clearly of secondary importance to gathering reliable data and developing a solid model. To understand effectiveness, however, we must evaluate how well borrowers are doing some time after finishing the microfinance program, as is suggested by the negative “last loan” coefficient and the positive “total loan” coefficient in this paper.
If real net profits of the businesses are decreasing, then the high dropout rates we see in MFIs the world over are largely explained—loans simply ceased to be profitable for the business over time. The financial accounting costs (to say nothing of economic and opportunity costs) outstripped the benefit of the loan. This inverse effect is quite exciting and certainly merits further probing into exactly why this phenomenon occurs.