The appendix provides additional tables and figures to support the analysis of various datasets. **Table A3a** presents simulation results for 30 observations in Sri Lanka, with each method based on 10,000 simulations. The simulated treatment effect is an increase of 1,000 Sri Lankan Rupees in profits for the treatment group, which is about 25% of the average baseline profits. **Table A3** shows the power of detecting a treatment effect using different methods for ENE data. The simulated treatment effect is a 920 Peso increase in income for the treatment group, which is about 20% of the average baseline income. Another simulation indicates that one-third of randomly selected children in the treatment group who would have dropped out do not. **Table A3v** details the power of detecting a treatment effect for IFLS expenditure data. The simulated treatment effect is an increase of 0.4 in ln household expenditure per capita, corresponding to about half a standard deviation or moving a household from the 25th to the 50th percentile. **Table A3w** outlines the power of detecting a treatment effect for LEAPS math test score data. The simulated treatment effect is an increase of one quarter of a standard deviation in the test score. **Figure A4a** to **Figure A4t** present various p-values and balance checks for different datasets at follow-up versus baseline, including Sri Lanka data, ENE income data, IFLS expenditure data, and LEAPS math test score and height z-score data. These figures also include results for different sample sizes (30, 100, 300 observations) and additional sample size results for other datasets.The appendix provides additional tables and figures to support the analysis of various datasets. **Table A3a** presents simulation results for 30 observations in Sri Lanka, with each method based on 10,000 simulations. The simulated treatment effect is an increase of 1,000 Sri Lankan Rupees in profits for the treatment group, which is about 25% of the average baseline profits. **Table A3** shows the power of detecting a treatment effect using different methods for ENE data. The simulated treatment effect is a 920 Peso increase in income for the treatment group, which is about 20% of the average baseline income. Another simulation indicates that one-third of randomly selected children in the treatment group who would have dropped out do not. **Table A3v** details the power of detecting a treatment effect for IFLS expenditure data. The simulated treatment effect is an increase of 0.4 in ln household expenditure per capita, corresponding to about half a standard deviation or moving a household from the 25th to the 50th percentile. **Table A3w** outlines the power of detecting a treatment effect for LEAPS math test score data. The simulated treatment effect is an increase of one quarter of a standard deviation in the test score. **Figure A4a** to **Figure A4t** present various p-values and balance checks for different datasets at follow-up versus baseline, including Sri Lanka data, ENE income data, IFLS expenditure data, and LEAPS math test score and height z-score data. These figures also include results for different sample sizes (30, 100, 300 observations) and additional sample size results for other datasets.