## Implications for Poverty Reduction in Rural Households in Ghana: Consumption Expenditure

The bottom part of Table 3b shows that the coefficients of rho (p), sigma (o) and lambda or selection hazard (A) -0.298, 0.496 and -0.148 respectively are all significant. But pos = A, however p determines if selection bias exists or not, once p is significantly different from zero, then it can be concluded that selection bias exists in the sample and has been corrected for. Also the negative sign of the coefficient of A shows that, OLS estimation of the impact of access to MF is bias downwards. That is, OLS under estimates the impact of access to MF as the estimator is unable to deal with the issue of sample selection bias. A comparison of the estimates of the coefficient of acesmf using OLS (0.212) and treatment effect (0.397) shows an underestimation of the impact of acesmf by 0.185 or 18.5% using the OLS estimator which does not correct for selection bias. Table 4 shows the t-test of the significance of the difference of weekly consumption expenditure of beneficiaries (lwkexpend) and non-beneficiaries (nlwkexpend) of MFI loans estimated at the means contingent on all the variables that are significant in explaining weekly consumption expenditure as discussed from Table 3b above (Kasena, Bwest, oldsav, amtprof, depend, age, age1, numacty, Primary school, JSS/Middle and acesmf). Thus lwkexpend is mean of log consumption expenditure for beneficiaries contingent on the above variables, nlwkexpend, mean of log weekly consumption expenditure for non- beneficiaries contingent on the above variables while diff is the difference between lwkexpend and nlwkexpend. The mean weekly consumption expenditure for beneficiaries of MFI loans is 3.055 while that of non-beneficiary is 2.952, contingent on all the statistically significant variables. Satisfaction with ATMs

Taking the anti-log these figures mean that beneficiaries of MFI loans spend about GH021.22 per week on basic needs while non-beneficiaries spend GH019.14 per week on basic needs contingent on these statistically significant variables. However, the difference between the mean weekly consumption expenditure for the two groups is GH01.11. The t-test of the null (H0) that the difference in weekly consumption expenditure for the two groups is equal to zero is rejected given the t-test value of 4.0701. This indicates that beneficiaries spend GH01.11 more per week on basic needs than non-beneficiaries.

Using the non-beneficiaries consumption expenditure (nlwkexpend) as a counterfactual outcome for the beneficiaries consumption expenditure (lwkexpend) therefore, it can be said that beneficiaries would have been spending GH019.14 per week on basic needs if they had not benefited from the MFI loans; but they now spend GH021.22 per week on basic needs, that is GH01.11 per week more. This suggests that MF has increased beneficiaries’ consumption expenditure by GH01.11 per week on basic needs. By implication access to MF contributed to poverty reduction.

As a robust check weekly consumption expenditure on food was used as a proxy for poverty. Food is seen as the most basic need of life. See appendix A for the results of the treatment effect with the log of consumption expenditure on food (lexpfood) as the dependent variable. The results indicate that kasena, bwest, oldsav, amtprof, age1, primary school education, numacty and acesmf are all significant in explaining consumption expenditure on food.

**Table 4: Two-sample t-test with unequal variances for weekly Consumption expenditure**

Variable | Mean | Std. Dev. | |

lwkexpend | 3.055 | 0.331 | |

nlwkexpend | 2.952 | 0.409 | |

diff. | 0.103 | 0.169 | |

H0: diff = 0; | T= 4.070 | ||

Satterthwaite’s | |||

degrees of freedom | 835.531 | ||

Obs. | 437 |