Chapter 3: Delinquency of Microfinance: A Two-part Probit Analysis of Cross-MFI Data
3.5 Empirical Results and Discussion
3.5.2 Subsample analysis between different microfinance institutions
Table 3.4presents split-sample regressions, where the main regressions are repeated for the four MFIs with two different measurements of loan default. The results, which substantially change across different sub-samples, reconfirm that, in general, married borrowers are associated with lower probabilities of being in default, but such a relation disappears in CACIL Honduras. A possible explanation to the results differ across countries is that the ben- efit of second income brought by marriage might be neutralised by the extra financial bur- den associated with the number of dependents. The summary statistics of our data shows that the average household size in CACIL Honduras is 2.84, while the numbers in the other MFIs are lower than 2.18.
What is more, the influence of marriage is especially pronounced for MICROCRED Madagas- car, where married borrowers have not only a higher repayment rate but also a smaller in- tensity of default (Columns 3). These findings are consistent with the propositions of Dunn and Kim (1999) and Vogelgesang (2003) stated previously. Hypothesis 1 is accepted, in terms of ‘marriage’. The most likely reason why married borrowers have better repayment records should be the high-risk tolerance brought by the community property regime.
However, when we look at the other variables of marital status, ‘cohabitation’ is only signifi- cant in MICROCRED and FINCA, while ‘divorced’ and ‘widowed’ are only significant in MI‐ CROCRED. Therefore, we may infer that in some countries a sense of responsibility might be irrelevant to the probabilities of default and delinquent, as there are no differences between single and cohabitation.
Regarding the other demographic characteristics such as gender, age and education level, the results are distinct from one country to another. Gender and age are insignificant in most of the subsamples. In FINCA Peru, female borrowers are associated with a lower proba- bility of default as expected but a higher intensity of default. On the other hand, in CACIL Honduras, significant convex relations between age and loan default have been found in both hurdles. This finding contradicts Hypothesis 2 that the youngest and oldest groups of borrowers have the highest probabilities of repayment. It could be because the oldest bor- rowers have higher financial commitments to their family and business expenses, while the youngest borrowers are less responsible in repaying their loans and lack of experience in business and financing.
In this section, we also found that a higher education level is not always significant and posi- tively related to better repayment performance across different MFIs. It is detected in the subsample of MICROCRED Madagascar only. In terms of CACIL Honduras, the clients who completed primary education have shorter periods of delayed repayment at the 1% level. However, there is no relationship between education level and the probability of loan de- fault. These findings imply that, in many countries, financial literacy and financial awareness, which highly associate with repayment performance, are generated from other sources in- stead of formal education at school.
The results for the variables of loan purposes in Table 3.4are inconsistent to those in Table 3.3. We reconfirm that investing in agricultural business is associated with lower probability of delinquency in the subsamples of INSOTEC Ecuador and FINCA Peru. However, in CACIL Honduras, it is found that engaging in agriculture has a higher intensity of delinquency. In
terms of the probability of delinquency, the difference between the agricultural and service sectors is insignificant.
There are two possible explanations for the inconsistent results. First, borrowers involved in agricultural businesses usually use credit to purchase seeds, fertiliser, pesticides, livestock, machinery, etc. They need at least four months to one year to receive the revenue from har- vesting. Hence, they cannot pay back the loan with a high repayment frequency, such as the weekly repayment plan with two weeks grace period which may sound feasible to the other business sectors. In our data, all microfinance institutions in the sample apply an indiscrimi- native and fixed frequency of repayment (monthly) to all clients. Considering the payback periods of different sectors, the institutional management should re-evaluate and modify their lending system to ease the burden on specific groups of borrowers.
Second, while agriculture is claimed to be the safest sector due to high social control and low volatility, it is in line with the prevailing weather conditions and indeterminate natural disasters. May 28 of 2009, an earthquake with a moment magnitude of 7.3 (very strong) oc- curred at the Caribbean Sea, 320 kilometres northeast of Tegucigalpa, the capital of Hondu- ras. It caused an estimated $37 million worth of damage. Mar 25 of 2010, the National Con- gress of Honduras approved to declare a national emergency caused by a prolonged drought and famine. 7,000 families suffered from severe food shortages. Paradoxically, Honduras also experienced flooding and excessive rains in other regions within the same period. Ac- cording to the statistics proved by Knoema, the cereal production had decreased by 9.9%, and the primary vegetable production had decreased by 7.8% in 2010. With the strong inter- ventions of natural disasters, the regression results related to agriculture in Honduras may be bias and misleading. In fact, Honduras’ vulnerability to natural disaster kept increasing dramatically in the recent decades. Its nominal losses were estimated at $4.7 billion, nearly half the losses for the entire Central American since 1974 (IADB, 2009). Therefore, the risk of investing in agricultural businesses may vary from a country to another.