Notes: The table presents the treatment effects on child development outcomes. The sample includes children surveyed in the end-line survey (2016). All estimates show results from OLS regressions based on equation (2). All regressions include the following controls: baseline values of the outcomes variables, child gender, child age, the total number of children in the household, the caregiver’s age, the caregiver’s education level (defined as a binary variable equal to 1 if the caregiver has at least primary education and 0 otherwise), the caregiver’s marital status (defined as a binary variable equal to 1 if the caregiver is married or cohabitating and 0 otherwise), and the asset index. This is equal to the first principal component of the following variables: floor materials of the house, roof materials of the house, main source of drinking water, and whether the house of the respondent is owned or rented, as described in Online Appendix Section A.4. All regressions include sampling weights. The dependent variables in columns (1–5) include standardized z-scores of the five ASQ dimensions calculated by subtracting the control group mean and dividing by the control group standard deviation in each survey wave. The dependent variable in column (6) is the child development index calculated by taking the average of the five ASQ z-scores. A full description of the construction of the outcomes is in Section 3 and in Online Appendix A.3. Light treatment is a dummy variable equal to 1 if the respondent (the caregiver) participated to the LT group. Full treatment is a dummy variable equal to 1 if the respondent (the caregiver) participated to the FT group.

p < 10%, **p < 5%, ***p < 1%. Robust standard errors in parentheses are clustered at the sector level. WILD cluster bootstrap with 9,999 replications and residuals drawn from Webb’s 6-point distribution are reported below the estimates (Cameron, Gelbach, and Miller 2008; Roodman et al. 2019). Two tailed p-values from a 5,000 replications Romano–Wolf step-down procedure (Romano and Wolf 2005; Clarke, Romano, and Wolf 2020) are shown below the estimates. A t-test of LT = FT is presented with the statistical significance of the test expressed in p-value. Observations and the R squared are presented at the bottom of the table.