Effect of Rental Markets on Poverty Reduction (Dynamic Panel Regression)
| Variable (dependent variable: Dummy for nonpoor, 1 if income is greater than the national poverty line,a 0 otherwise) | (1) | (2) | (3) | (4) |
|---|---|---|---|---|
| Dummy for nonpoor in prior survey year (1 = above the poverty line) | 0.107*** (2.78) | 0.102** (2.55) | 0.115*** (3.02) | 0.108*** (2.75) |
| Dummy for nonpoor in prior survey year × Rented-in land (α) | -0.046 (0.69) | -0.034 (0.50) | ||
| Rented-in land (β) | 0.044 | 0.047 | 0.061 | 0.060 |
| (1.01) | (1.08) | (1.21) | (1.21) | |
| Log of total agricultural assets | 0.033** | 0.027 | 0.032** | 0.027 |
| (2.25) | (1.49) | (2.24) | (1.52) | |
| Log of adult equivalents | -0.203*** | -0.230*** | -0.202*** | -0.230*** |
| (7.12) | (3.17) | (7.07) | (3.18) | |
| Land endowment per capita | 0.058*** | 0.098*** | 0.057*** | 0.097*** |
| (4.29) | (3.80) | (4.21) | (3.75) | |
| Log of livestock value | 0.017*** | 0.013*** | 0.017*** | 0.013*** |
| (6.31) | (3.55) | (6.27) | (3.54) | |
| Female-headed household ( = 1) | -0.113* | -0.085 | -0.110* | -0.085 |
| (1.84) | (0.38) | (1.79) | (0.37) | |
| Death of household head since prior survey ( = 1) | 0.055 | 0.036 | 0.052 | 0.035 |
| (0.72) | (0.21) | (0.69) | (0.20) | |
| Log of household head’s age | 1.110 | 1.280 | 1.081 | 1.270 |
| (0.77) | (0.73) | (0.76) | (0.72) | |
| Log of household head’s age squared | -0.140 | -0.161 | -0.137 | -0.160 |
| (0.76) | (0.71) | (0.74) | (0.70) | |
| Household head completed primary school ( = 1) | 0.048 | 0.045 | 0.047 | 0.044 |
| (1.27) | (1.08) | (1.25) | (1.07) | |
| Log of rainfall | 0.156 | 0.179 | 0.158 | 0.181 |
| (0.23) | (0.25) | (0.23) | (0.26) | |
| Log of rainfall in the previous period | 0.547 | 0.598 | 0.526 | 0.583 |
| (0.75) | (0.81) | (0.72) | (0.78) | |
| Test for α+β=0(p-value) | — | — | 0.82 | 0.74 |
| Hanson J overidentification test (p-value) | 0.44 | 0.31 | 0.45 | 0.27 |
| Observations | 2,396 | 2,396 | 2,396 | 2,396 |
Note: Absolute value of t - statistics in parentheses. All standard errors adjusted for clustering effects at the village level. Village dummies were added to control for local fixed factors that affect the change in income, time dummies were included to control for time trend of income growth, and interactions of village and time dummies are included to control for any policy change over time that may affect the income. All the models are estimated using GMM-IV estimation method. Columns 1 and 3: First differencing equation, ΔYijt- 1, ΔRijt and their interaction term Δ(Yijt- 1 × Rijt) are treated as endogenous variable with ΔYijt- 1 being instrumented by Yijt- 2, ΔRijt by Rijt -1, and Rijt -2 and Δ(Yijt-1×Rijt) by (Yijt-2× Rijt-1). Columns 2 and 4: First differencing equation, ΔYijt-1 and ΔRijt and their interaction term Δ(Yijt-1×Rt) are again treated as endogenous and instrumented the same way as Columns 1 and 3. Δ(Log of value of agricultural assets), Δ(Log of adult equivalents)t, Δ(Land endowment per capita)t, Δ(Log of livestock value)t, and Δ(Female head)t are all treated as predetermined, and they are instrumented by the level of themselves lagged by one period and lagged by two periods (i.e., xt-1 and xt-2). The inverse probability weighting approach was used to account for potential attrition bias in all the regressions.
↵a The national poverty line for rural Kenya was KSh 1,239/month/person in 1997, KSh 1,347/month/person in 2000, KSh 1,490/month/ person in 2004, and KSh 1,598/month/person in 2007.
↵* Significant at 10%;
↵** Significant at 5%;
↵*** Significant at 1%.