Panorama Socioeconomico, December 2009, vol.27, iss.39, pp.100-112. Available From:Link to Source
This study evaluates the impact of microcredit with a distinction between NGO- and bank-based programmes, because they address different market shares. It uses probit estimations to assess whether microcredit clients are better off than non-clients in terms of individual income. The authors use data for Chile and Brazil to compare microcredit clients with a control group of people with similar characteristics. Because formal survey data were available only for clients, the control group was created using non-survey methods: propensity score matching (PSM) was used to partially address selection bias.
The Chilean data were collected in February and March 2002. Random samples from the bank Bandesarrollo (N = 51) and the NGO Propesa (N = 30) led to a total of 81 clients for the treatment group. The control group was constructed from the Chilean survey of national socioeconomic characteristics, which was run in November 2000 (CASEN 2000), and consists of 715 employees aged 17–66. PSM was based on individual characteristics (gender, age, marital status, educational attainment), household characteristics (household head, household income), enterprise characteristics (sector, number of employers, working hours, income from firm), and region.
In Brazil, data were collected in February and March 2002. Two banks (Microcred and Socialcred; N = 46), and three NGOs (CAEPE, Bancri, and Banco Povo Santo Andre; N = 152) provided information about 198 microcredit clients. The control group was constructed from the Brazilian National Survey on Households, run in September 1999 (PNAD 1999), which consisted of 34,887 observations. Apart from overall estimates (control group C1), this larger sample allowed for a distinction between employers (C2; N = 9,186) and salaried workers (C3; N = 25,701), which led to better comparisons than was the case in the Chilean data. Matching was based on the same type of (proxy) variables. The Brazilian analyses were corrected for inflation by adjusting the mean income in the PNAD sample with the general price index.
In both Chile and Brazil, there were large differences in income between microcredit clients from banks and NGOs. These different market shares did affect outcomes. The Chilean data presented a positive, but non-significant increase in average income for the total sample, indicating a 25-percent higher income for microcredit recipients (Ch$ 68,568). Those in the bank-based programme earned a non-significant 38-percent higher income than those without microcredit (Ch$119,432). The NGO-based programme, however, negatively affected the income of microcredit clients by about 50 percent, which is a significant decrease of Ch$219,484.
In Brazil, microcredit was highly effective, and significantly increased the average income of clients by R$1,188. As expected, differential impact existed when using as control group of either entrepreneurs (R$1,351), or salaried workers (R$1,434). Splitting the data into bank-based and NGO-based programmes did not change impact estimates drastically. Impact was highest for bank clients: R$1,913 compared to salaried workers, and R$1,494 compared to employers. For NGO clients, the impact was lower but still high, as well as significant. In contrast to bank clients, NGO clients were better off, compared to employers (R$1,164) and to workers (R$971).
The Brazilian labour market is known for people having more than one job at a time. As a robustness check, the authors re-estimated the average income of the Brazilian control group, using all income sources. The effect of the microcredit programme was somewhat lower but still high and significant. For bank clients, the impact of microcredit was R$1,297 compared to employers and R$1,831 compared to workers. For NGO clients, the effect was R$1,028 compared to employers and R$872 compared to workers.