Table 3.

Correlates of attrition from target baseline sample to estimation sample.

	Coefficients	SE	p-value
T-N	−0.016	0.021	0.444
T-A	0.015	0.025	0.542
T-AN	0.028	0.017	0.089
T-ANG	−0.019	0.020	0.348
Woman's age at baseline (years)	0.000	0.000	0.778
Man's age at baseline (years)	−0.001	0.001	0.304
Woman's education at baseline (years)	0.006	0.002	0.025
Man's education at baseline (years)	−0.007	0.002	0.000
Number of adults in HH at baseline	0.006	0.006	0.333
Number of children under 15 in HH at baseline	−0.001	0.006	0.923
Size of operated land at baseline (acres)	−0.018	0.005	0.001
Wealth index	0.008	0.003	0.013
Constant	0.194	0.042	0.000
Upazila fixed effects	YES
Number of observations	3,264
Test of joint significance
F-statistic: Treatment arms	1.913
p-value	0.105

Note: Dependent variable is coded as 0 if HH is included in sample (Control/T-N/T-A/T-AN/T-ANG only) and 1 if excluded from sample; Standard errors adjusted for clustering at block level. All treatments targeted to men and women.

Definition of treatment arms: T-N = Nutrition Behavior Change Communication (BCC) training delivered to women and men by agricultural extension agents (AEAs) from the Ministry of Agriculture. T-A = Agricultural Production training delivered to women and men by AEAs. T-AN = Agricultural Production + Nutrition BCC training delivered to women and men by AEAs. T-ANG = Agricultural Production + Nutrition BCC training delivered to women and men by AEAs + gender sensitization activities for women and men conducted by Helen Keller International (HKI).