Variable (dependent variable: Log of gross crop output in KSh) | Village Fixed Effect | Random Effect | Household Fixed Effect |
---|---|---|---|

Log of area cultivated | 0.584^{***} (18.92) | 0.570^{***} (17.33) | 0.499^{***} (12.16) |

Log of adult equivalence | 0.053^{*} (1.97) | 0.064^{**} (2.24) | 0.090^{***} (2.52) |

Log of agricultural assets | 0.074^{***} (7.07) | 0.075^{***} (6.98) | 0.046^{***} (3.58) |

Log of total cost (except for seed) | 0.030^{***} (4.02) | 0.029^{***} (3.60) | 0.014 (1.60) |

Log of total seed expenditure | 0.170^{***} (8.17) | 0.165^{***} (7.70) | 0.119 (5.42)^{***} |

Log of household head’s age | - 0.475 (0.33) | - 0.389 (0.28) | 0.346 (0.21) |

Log of household head’s age squared | 0.066 (0.37) | 0.056 (0.32) | - 0.051 (0.23) |

Household head with primary education | 0.021 (0.66) | 0.020 (0.63) | - 0.002 (0.05) |

Female-headed household ( = 1) | - 0.085^{***} (2.69) | - 0.072^{**} (2.50) | - 0.051 (0.84) |

Death of adult since prior survey ( = 1) | 0.114 (1.61) | 0.078 (1.17) | 0.065 (0.88) |

Log of anual rainfall | 0.419^{***} (3.82) | 0.503^{***} (3.96) | 0.516^{***} (4.14) |

Lagged log of annual rainfall | 0.152^{**} (2.59) | 0.158^{***} (2.66) | 0.159^{***} (2.83) |

Distance to extension service | 0.001 (0.34) | -0.000 (0.12) | 0.000 (0.07) |

Distance to motor road | 0.011 (1.24) | 0.010 (1.20) | 0.005 (0.53) |

Distance to nearest piped water source | - 0.003 (0.95) | - 0.002 (0.99) | - 0.002 (0.69) |

Distance to the nearest electricity supply | 0.002 (0.62) | 0.001 (0.41) | - 0.000 (0.09) |

Population intensity | - 0.075^{*} (1.67) | - 0.115^{*} (1.88) | |

Credit constrained ( = 1) | - 0.079 (1.12) | - 0.078 (1.20) | - 0.063 (0.96) |

Observations | 4,617 | 4,617 | 4,617 |

R-squared | 0.67 | 0.32 |

*Note:*Standard errors adjusted for clustering effect at village level. Village dummies, year dummies, and their interaction terms were included in the OLS and random effects models. Year dummies and the interaction terms of year and village dummies are included in the fixed effects model. Inverse probability weighting is used to account for potential attrition bias (see methods section for description). KSh, Kenyan shillings.↵* Significant at 10%;

↵** Significant at 5%;

↵*** Significant at 1%.