Impact of microcredit program on poverty reduction among rural households in Saki East Local Government area of Oyo State , Nigeria

It is widely acknowledged that access to financial services plays a critical role in helping poor people strengthen their economic potentials, increase their asset base and diminish their vulnerabilities to external shocks. However, poor people have very limited access to financial services provided by the formal financial institutions due to the procedures outlined for accessing such loans. In recent time, Microcredit has received global attention owing to its claim in alleviating poverty. This study investigates the impact of microcredit program on poverty reduction among rural households in Saki East Local Government area of Oyo State using logit model and propensity score matching techniques to prove the impact. Findings revealed a negative, non-significant impact of microcredit on poverty reduction among rural households. Therefore, it is recommended that microcredit program should be implemented along with other poverty alleviation strategies.


Introduction
Microfinance is a term used to describe financial services for those without access to formal banking.It incorporates the loans often at interest rate of 25% or more, to individuals, groups, and small business i.e.Microcredit.More recently; it has also been extended to include the provision of savings accountsmicrofinance-as well as insurance and money transfer services.The main feature of microcredit is lending decisions are based on the ability of the client's microenterprise, to effectively apply the loan, to increase revenue and in turn repay the loan (www.microrate.com).
Microcredit is a successful way to increase the economic growth and, it has also empowered women by increasing their poverty gap and reduces their extent of vulnerabilities (Chavan and Rumakumar, 2002).The microcredit program proved to be influential in many developing countries (Hossain, 1988;Otero, 2000).There are some shortcomings as well as reported earlier by different scientists (Adams & Von Pischke, 1992;Buckley, 1997).
In view of the above findings, concerning the positive impact of microfinance program on the poor, there are still controversies trailing the methodological approach being used to measure the impact estimate.In this study, the method of Propensity Score Matching was used to match the treated individuals (microfinance users) with the control group i.e. non-users of microfinance (Counterfactual group).Thus, a relevant issue for empirical investigation is to examine the impact of microcredit program driven by Non-Governmental Organizations (MFIs) on poverty reduction among rural households.Specifically, the study will attempt to; i.A cross sectional data of users and nonusers of microcredit program were obtained.Respondents were selected by random sampling.Ago-amodu, Sepeteri and Ogbooro were purposively selected based on their clients' relative length of loan cycles.Users and non-users of microcredit program in three selected communities were randomly selected proportionately from a total of eight (8) wards in the three communities.A total of one hundred and twenty-nine respondents comprising both the users and non-users of microcredit (NGO) were selected.Here, a household is considered to have access to microcredit program, if the household head or the spouse is participating in the microcredit program.Three micro-finance institutions-LAPO, SEAP and Grooming Centre-are distributed within the three local government areas that constitute the old Ifedapo local government.A SEAP microfinance institution was purposively selected because it is the only microfinance institution with the widest coverage within the local government area.

Propensity score matching
In this approach, access to microfinance is considered as a "treatment" so that an average treatment effect on the treated is estimated using the propensity score matching as explained previously (Rawlings and Schardy, 2002;Esquivel and Huerta-Pineda, 2007;Ravallion, 2001).Thus in this study, the critical assumption that was made in using this methodology, is that the decision to be treated (to access microfinance), although not random, ultimately depends upon observable variables.Therefore, for estimating the impact of microfinance services on expenditure and other indicators, two groups are identified, those with access to microfinance (denoted as Ti =1 for household i and those without Those with access to microfinance (treated) are matched to those without (control group) on the basis of the propensity score: (probability of accessing microfinance (intervention) given observed characteristics).
given xi, as a vector of pre-intervention control variables.If the are independent over all i, and the outcomes due to intervention access are independent of access to intervention given xi, then outcomes are also independent of access to intervention given just as they would be if interventions were given randomly.After computing the propensity score, the ATT effect (τ) is estimated as follow; Y1 is the potential outcome if the individual is treated; Y0 is the potential outcome if the individual is not treated.The known matching algorithms in the literature are nearest neighbor, kernel matching, radius matching, local linear regression matching etc.However, in this study, nearest neighbor and kernel matching were used, since the propensity score is a continuous variable.

Results and discussion
Table 1 shows the summary statistics of socio-economic variables.Among the microcredit users, a household head has an average age of 42.5 years with an average household size of 8 members.Also, the average years of formal education among the microcredit users are 8.Most of them are found in trading activities with average of 17 years of experience.On the other hand, among the non-users of microcredit program, a household head has an average age of 45 years and household size of 8 members.The predominant activity among the non-users is farming with 19.6 mean years of experience.A non-user' household has an average of 9 years of formal education.The above analysis suggested that non-users of microcredit are more educated and experienced occupationally, than microcredit users and are therefore able to generate more income through improved productivity.This is corroborated by Mpuga (2004).

Propensity Score Matching
The result of the propensity score matching, showing the overall gain (ATT) in expenditure per capita of microcredit receiving households is presented in Table 2 above.The average overall gains in expenditure per capita of microcredit receiving households using the kernel matching is (N205.65).This implies that microcredit users are worse off by the amount in parenthesis and are made poorer by participating in microcredit program.This findings is consistent with Hulme and Mosley (1996).

Determinants of access to microcredit by the household heads
The logit regression model was used to analyze the determinants of access to microcredit.Two distinct models, used to capture the determinants of access to microcredit, were presented in table 3. The second columns of the table, gives the result of logit model with socio-economic variables, credit variables and aggregate social capital index (multiplicative index), while the third column of the table, gives the result of the logit model using the socio-economic variables and disaggregated social capital variables, to identify the significant coefficients of social capital indices.The two models are adequate, considering their chi-square values, which are highly significant at less than one 1% level.In summary, the coefficient of educational status, primary occupation, occupational years of experience, interest charged, time lag of credit delivery, payback period, credit distance, household membership density, decision making and cash contribution index, significantly influenced the households' access to microcredit, at their respective levels.These findings, are consistent with Mpuga (2004).

Poverty profile of the sampled households by socio-economic characteristics
The decomposition of poverty based on several characteristics, was done in order to relate poverty to changes in those characteristics.The characteristics are socioeconomic and participation in microcredit program.The table 4 reveals the poverty status of the sampled households, using the socioeconomic characteristics and participation status of microcredit program.It is evident, from the table, that female headed households are more vulnerable to poverty than their male counterparts as evident in World Bank report (2007).Households with lower level of education, who engage in trading with higher household size and whose age brackets are found above or below the productive age limit, are more vulnerable to poverty than do the otherwise households.On the other hand, microcredit receiving households show higher incidence of poverty, depth and severity than non-microcredit receiving households.

Determinants of poverty among rural households
The response of households' poverty is better captured when expressed in percentage rather than the unit of measurement of the variables.Following Tobit decomposition framework as suggested by Mc Donald and Moffit (1980), the effects of changes in the explanatory variables (Xi) on the probability of being poor, was determined.Table 5 shows the coefficient of marginal effects of probability of being poor among households, in the study area.The second and third columns of table, present the result of the Tobit model with multiplicative social capital index and additive social capital variables respectively.Among the socio-economic variables, the coefficients of age, household size, primary occupation, occupational years of experience and food expenditure are all significantly affect the probability of being poor by the households.On the other hand, the coefficient of internal heterogeneity within the local associations only, significantly affect the probability of being poor; i.e households involving in local level associations, with diverse internal heterogeneity are less likely to be poor, than households with lower level of internal heterogeneity among their members.

Conclusions
It is evident that, the formal financial systems in Nigeria provide services to about 35% of the economically active populations while the rest 65% are often served by the informal sector through non-governmental Organizations-Microfinance institutions (CBN, 2005).The study used both primary and secondary data.Logit regression model was used to generate the propensity scores i.e. the predicted probability of microcredit access.Also, the propensity score matching was also used to find the impact estimate (ATT), Average Treatment effect on the treated.The ATT estimate shows a negative impact of microfinance Thus, implementing microfinance program-as a poverty reduction strategy, in isolation is a necessary, but not sufficient condition for poverty alleviation.It is therefore recommended, that education of the beneficiaries, infrastructure, and capacity development are critical to any poverty alleviation strategy.
The study was conducted in Saki East Local Government Area of Oyo State.Saki East Local Government used for this study, was carved out of the defunct Ifedapo Local Government in I996, under the then military government of General Sanni Abacha.It is one of the Thirtythree (33) local governments in Oyo State.It has an area of 1569km 2 with a population of 108,957, by the 2006 Census.The headquarter of the local government is the ancient town of

Table 1 .
Summary statistics of socio-economic variables.

Table 2 .
Result of the Propensity Score Matching, Showing the Overall gain (ATT) in Monthly Per Capita Expenditure of the Microcredit receiving Households.

Table 4 .
Poverty profile of sampled households by socio-economic characteristics.