Table 3.

Impacts of REU Intensive and Moderate Treatments on Adoption Measures at Endline, Mozambique and Uganda

	Mozambique						Uganda
	Adopted OFSP		Share of OFSP in SP Area		Total Area, OFSP (acres)		Adopted OFSP		Share of OFSP in SP Area		Total Area, OFSP (acres)
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
Panel A: Intensive Treatment versus Moderate Treatment
Intensive Treatment	0.659*** (0.043)	0.629*** (0.046)	0.647*** (0.040)	0.622*** (0.042)	0.091*** (0.017)	0.095*** (0.015)	0.617*** (0.040)	0.624*** (0.030)	0.438*** (0.027)	0.428*** (0.023)	0.168*** (0.018)	0.169*** (0.017)
Moderate Treatment	0.691*** (0.033)	0.650*** (0.037)	0.612*** (0.033)	0.587*** (0.033)	0.100*** (0.021)	0.092*** (0.020)	0.579*** (0.071)	0.595*** (0.039)	0.414*** (0.039)	0.410*** (0.040)	0.113*** (0.015)	0.117*** (0.014)
Additional Covariates?	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes
Test H0: Model 1 = Model 2 (p-value)	0.464	0.625	0.438	0.449	0.731	0.883	0.649	0.542	0.615	0.688	0.016	0.018
Panel B: Average treatment effect of both interventions
Treated	0.676*** (0.032)	0.639*** (0.036)	0.628*** (0.029)	0.604*** (0.030)	0.096*** (0.015)	0.094*** (0.015)	0.607*** (0.034)	0.617*** (0.025)	0.432*** (0.023)	0.424*** (0.020)	0.154*** (0.014)	0.156*** (0.014)
Additional Covariates?	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes
Endline Control Mean	0.093		0.073		0.017		0.054		0.027		0.008
Number of obs.	610	610	534	534	610	610	975	975	751	751	975	975

Note: All models are single difference models at endline. Baseline levels of adoption and area planted with OSP were very low, and so were omitted from these models. The share of OFSP in SP area has 59 missing observations in Mozambique and 224 missing observations in Uganda because these households did not grow any sweet potato. Tests of equality of impact of Model 1 and Model 2 are adjusted Wald tests. Average treatment effects reported at the bottom of the table are average impacts over Model 1 and Model 2 using the same specification for that column in a separate regression. Additional covariates included in some specifications, all measured at baseline, are whether the household had access to off-farm work, the number of male and female adults, whether the household head was male, whether or not a nutrition promoter lived in the household, whether the household grew sweet potato in 2006, and per-capita expenditures. All regressions include strata-level fixed effects. Standard errors are clustered at the village level in Mozambique and the farmer group level in Uganda. Asterisks ***indicate significance at the 1% level. For Mozambique, wild bootstrapped p-values appear in appendix table S.1.

Source: REU Baseline and Endline Survey Data, Mozambique and Uganda.