Implications for Poverty Reduction in Rural Households in Ghana: Results of OLS estimation and Probit estimation
Table 2 shows the results of OLS estimate of equation. The dependent variable is lwkexpend2 which is the natural log of wkexpend. Thus lwkexpend2 measures the percentage of weekly expenditure on basic needs due to the independent variable(s). From the Table kasena, bwest, depend, oldsav, amtprof, age, Primary, JSS/Middle, Post secondary, numacty and acesmf all are statistically different from zero and each of them has a positive relationship with the percentage of weekly expenditure (proxy for poverty).
The coefficient of acesmf (0.2122) implies that beneficiaries of MFIs loans on the average spend 21.2% weekly on basic needs more than non-beneficiaries of MFIs loan. However, due to the issues of selection bias and endogeneity OLS estimate of the coefficient of acesmf (0.2122) is bias and for that matter does not provide an accurate measure of the impact of access to MF. In the light of this, the treatment effect model is estimated which controls for sample selection bias. Bank-customer
Tables 3a and 3b show the results of the treatment effect model, which have been simultaneously estimated using the access to MF (acesmf) and the outcome (weekly consumption expenditure (wkexpend2)). Table 3a shows the results of access to MF equation with acesmf as the dependent variable. While Table 3b shows the results of the weekly expenditure equation: with the natural log of weekly expenditure on basic needs, a proxy for poverty (lwkexpend2) as the dependent variable. The results of access to MF using a probit estimation method showed that Bwest, oldsav, assets and frnsours are all statistically different from zero at 1% significance level while kasena and amtprof are also statistically different from zero at 5% significance level. However, lonsours is also different from zero at 10% significance level.
The coefficients of kasena and Bwest are -0.490 and -0.979 respectively. These suggest that respondents from the Kasena Nankani and Bawku West Districts have less probability of accessing or taking loans from MFIs than their counterparts from the Talensi-Nabdan, Bongo and Builsa Districts. Both Kasena Nankani and Bawku West Districts have very vibrant market centres and good number of MFIs. It stands to reason therefore that people from these two Districts could have over the years become more financially self reliant and would therefore be less likely to borrow from a MFI than their counterparts in the reference Districts. The result also shows that respondents from Kasena Nankani District have higher probability of accessing MFIs loans than those from the Bawku West District. The coefficient of oldsav is -0.00213, this means that an increase in old or initial savings reduces the probability of accessing or taking an MFI loans. This could be so because those who have substantial amount of saving may not have to borrow again from an MFI. Though initial saving is usually a criterion for grant of loans, the results suggests that individual who are able to raise the initial amount required as a start-up capital for their agro-processing business have less probability of borrowing from an MFI.
Table 2: Results of OLS estimation of impact of access to MF (dependent variable: weekly consumption expenditure)
Variables | Coefficient | Robust Std. error | P-Value |
Kasena Nankana District (1/0) | 0.341*** | 0.072 | 0.000 |
Bawku west District (1/0) | 0.548*** | 0.063 | 0.000 |
Age in years | 0.052*** | 0.016 | 0.001 |
Age-squared | -0.001*** | 0.000 | 0.001 |
Number of dependants in household | 0.033** | 0.014 | 0.017 |
Number of people in household | -0.006 | 0.008 | 0.477 |
Has received MFI loan(1/0) | 0.212*** | 0.050 | 0.000 |
Number of borrowing sources | -0.009 | 0.051 | 0.855 |
Amount of profit | 0.002*** | 0.001 | 0.003 |
No. of i ncome generating activities | 0.045** | 0.019 | 0.019 |
Initial savings | 0.000* | 0.000 | 0.092 |
Value of physical assets | 0.000 | 0.000 | 0.537 |
Primary school(1/0) | 0.187*** | 0.057 | 0.001 |
JSS/Middle school(1/0) | 0.182** | 0.083 | 0.030 |
SSS/Secondary school(1/0) | -0.082 | 0.138 | 0.554 |
Post Secondary school(1/0) | 0.336* | 0.179 | 0.061 |
Constant | 1.335*** | 0.349 | 0.000 |
Number of obs. | 437 | ||
R-squared | 0.345 | ||
Root MSE | 0.498 |
Table 3a: Results of Probit estimation of determinants of access to MF(dependent variable: access to MF)
Variables | Coefficient | Std. error | P-Value |
Kasena Nankana District (1/0) | -0.490** | 0.218 | 0.025 |
Bawku west District (1/0) | _{–}0 979*** | 0.235 | 0.000 |
Age in years | 0.017 | 0.042 | 0.688 |
Age-squared | -0.000 | 0.001 | 0.944 |
Dependants in household | 0.046 | 0.039 | 0.227 |
Number of people in household | 0.024 | 0.025 | 0.345 |
Number of borrowing sources | -0.331* | 0.198 | 0.095 |
Number of friends with loans | 0.298*** | 0.037 | 0.000 |
Amount of profit | 0.004** | 0.002 | 0.024 |
No. of income generating activities | -0.069 | 0.072 | 0.338 |
Initial savings | -0.002*** | 0.001 | 0.001 |
Value of physical assets | 0.001 | 0.000 | 0.537 |
Primary school(1/0) | 0.118 | 0.185 | 0.521 |
JSS/Middle school(1/0) | 0.149 | 0.214 | 0.485 |
SSS/Secondary school(1/0) | -0.292 | 0.309 | 0.345 |
Post Secondary school(1/0) | 8.274 | 7.395 | 0.999 |
constant | -1.206 | 0.906 | 0.183 |
Number of obs. | 437 |
Table 3b: Results of the treatment effect model (dependent variable: weekly consumption expenditure)
Variable | Coefficient | Std. error | P-Value |
Kasena Nankana District (1/0) | 0.342*** | 0.070 | 0.000 |
Bawku west District (1/0) | 0.548*** | 0.066 | 0.000 |
Age in years | 0.048*** | 0.014 | 0.001 |
Age-squared | -0.001*** | 0.000 | 0.000 |
Dependants in household | 0.029** | 0.013 | 0.022 |
Number of people in household | -0.009 | 0.009 | 0.318 |
Number of borrowing sources | 0.004 | 0.067 | 0.957 |
Amount of profit | 0.001** | 0.006 | 0.025 |
No. of income generating activities | 0.048** | 0.021 | 0.025 |
Initial savings | 0.000* | 0.000 | 0.064 |
Value of physical assets | 0.000 | 0.000 | 0.593 |
Has received MFI loan(1/0) | 0.397*** | 0.093 | 0.000 |
Primary school(1/0) | 0.170*** | 0.064 | 0.008 |
JSS/Middle school(1/0) | 0.173** | 0.074 | 0.020 |
SSS/Secondary school(1/0) | -0.122 | 0.110 | 0.267 |
Post Secondary school(1/0) | 0.202 | 0.308 | 0.511 |
Constant | 1.395 | 0.295 | 0.183 |
Observations. | 437 | ||
^{rho(}P^{)} | -0.298** | 0.120 | |
sigma( o ) | 0.496*** | 0.018 | |
Lambda( A) | -0.148** | 0.062 | |
Wald chi2 | 223.23*** | ||
Log likelihood | -518.74472 | ||
Wald test of indep. eqns. (rho = 0): chi2 | 4.39** |