Academia.eduAcademia.edu

Figure 7 – uploaded by Robert Lensink

See full PDFdownloadDownload figure
Notes: Robust standard errors are in parentheses. ***Coefficients significant at 1 per cent, **5 per cent, *10 per cent

Table 7 Notes: Robust standard errors are in parentheses. ***Coefficients significant at 1 per cent, **5 per cent, *10 per cent

Related Figures (7)

Figure 1. Impact of a cash transfer.
Figure 1. Impact of a cash transfer.
Table 1. Sample for the CT experiment dropped out of the analysis. We also considered the households that originally did not want the intervention (C3 group); 22 per cent of this group requested the cash transfer in the midline survey. However, we kept those households in group C3. Overall, the numbers suggest that over time, the number of households that sought to leave the T group was less than the number that wanted into that group, which could be a demonstration effect. on -" % — pees ee ween ay: ‘? . « > «© ee ed
Table 1. Sample for the CT experiment dropped out of the analysis. We also considered the households that originally did not want the intervention (C3 group); 22 per cent of this group requested the cash transfer in the midline survey. However, we kept those households in group C3. Overall, the numbers suggest that over time, the number of households that sought to leave the T group was less than the number that wanted into that group, which could be a demonstration effect. on -" % — pees ee ween ay: ‘? . « > «© ee ed
Notes: Robust standard errors are in parentheses. The regression results refer to a logistic regression, where Y = | if the household remained in the sample (281 out of 300), and Y = 0 if it left (19 out of 300). Table 2. Characteristics of dropouts and respondents ~* We tested whether the attrition was random, by estimating a logit model that explained attrition using a vector of baseline variables (Table 2). When regressing dropouts on a vector of baseline
Notes: Robust standard errors are in parentheses. The regression results refer to a logistic regression, where Y = | if the household remained in the sample (281 out of 300), and Y = 0 if it left (19 out of 300). Table 2. Characteristics of dropouts and respondents ~* We tested whether the attrition was random, by estimating a logit model that explained attrition using a vector of baseline variables (Table 2). When regressing dropouts on a vector of baseline
Table 3. Balancing test between groups Notes: Two-sample Kolmogorov-Smimov test for the equality of the distribution functions at 5 per cent significance level. The majority of these households included self-employed or regular wage earners; a minority reliec on casual labour (defined as non-permanent labour, with no fixed employer or contract from any singl employer) for their income. Casual labourers, such as daily workers or ‘handymen’, often work muctl longer hours than do regular salary or wage earners. In addition, the average household size in th sample was around 5.5 people. Table 4 provides some baseline descriptive statistics for the mait variables in our analyses. Pe SEERA IRS ERs Aa ee teehee STRESSES mee ASC Te eS eet aaes According to Table 3, the ¢ groups of self selected households are similar, such that randomisatio: ypparently resulted in balanced groups. Only in case of PCnonfood the test suggests baselin MW 10 dM de HM erences between C1 and C2. Since, this only holds for one of the variables this does not invalidat sur randomisation.* We also compared the characteristics of the self-selected households and house ds that did not want to take part. Even for this comparison, for nearly all the variables, we found n erences between households that self-selected for the CT and households that initially did not wan 1 CT (group C3). These results may not come as a surprise; all the households in our experiment ar ow the poverty line and live close to one another. Thus there is no reason to expect substantia erences in their observational characteristics across groups. (i ee i e/a ae
Table 3. Balancing test between groups Notes: Two-sample Kolmogorov-Smimov test for the equality of the distribution functions at 5 per cent significance level. The majority of these households included self-employed or regular wage earners; a minority reliec on casual labour (defined as non-permanent labour, with no fixed employer or contract from any singl employer) for their income. Casual labourers, such as daily workers or ‘handymen’, often work muctl longer hours than do regular salary or wage earners. In addition, the average household size in th sample was around 5.5 people. Table 4 provides some baseline descriptive statistics for the mait variables in our analyses. Pe SEERA IRS ERs Aa ee teehee STRESSES mee ASC Te eS eet aaes According to Table 3, the ¢ groups of self selected households are similar, such that randomisatio: ypparently resulted in balanced groups. Only in case of PCnonfood the test suggests baselin MW 10 dM de HM erences between C1 and C2. Since, this only holds for one of the variables this does not invalidat sur randomisation.* We also compared the characteristics of the self-selected households and house ds that did not want to take part. Even for this comparison, for nearly all the variables, we found n erences between households that self-selected for the CT and households that initially did not wan 1 CT (group C3). These results may not come as a surprise; all the households in our experiment ar ow the poverty line and live close to one another. Thus there is no reason to expect substantia erences in their observational characteristics across groups. (i ee i e/a ae
Table 4. Descriptive statistics measured at the baseline
Table 4. Descriptive statistics measured at the baseline
Notes: Robust standard errors are in parentheses. ***Coefficients significant at 1 per cent, **5 per cent, *10 per cent In the second co umns in cereal between the C1 and 7 calories consumed from cere which suggests tha’ the shi Tables 5—7, we find differences in the per capita calories consumed as [ groups. The coefficient for C1 in the regression explaining “per capita als’ (PCCereal) is positive, and significant in the fixed effects regression, ft to a cash transfer leads to a decline in cereal consumption. We also conducted regressions for the three main categories of cereals (rice, wheat and other); their results, presented in tables A4—A6° in the Online Appendix, suggest that the decline in cereal consumption is due mainly to a dec ine in w tion of other cereals also is The Tablec 5—7 alen cnovect that nncoanditiqgnal CT induce oreater ner canita exnenditires on non- heat consumption. However, across all estimation methods, the consump- positively affected by the shift to the cash transfer.
Notes: Robust standard errors are in parentheses. ***Coefficients significant at 1 per cent, **5 per cent, *10 per cent In the second co umns in cereal between the C1 and 7 calories consumed from cere which suggests tha’ the shi Tables 5—7, we find differences in the per capita calories consumed as [ groups. The coefficient for C1 in the regression explaining “per capita als’ (PCCereal) is positive, and significant in the fixed effects regression, ft to a cash transfer leads to a decline in cereal consumption. We also conducted regressions for the three main categories of cereals (rice, wheat and other); their results, presented in tables A4—A6° in the Online Appendix, suggest that the decline in cereal consumption is due mainly to a dec ine in w tion of other cereals also is The Tablec 5—7 alen cnovect that nncoanditiqgnal CT induce oreater ner canita exnenditires on non- heat consumption. However, across all estimation methods, the consump- positively affected by the shift to the cash transfer.
Notes: Robust standard errors are in parentheses. ***Coefficients significant at 1 per cent, **5 per cent, *10 per cent. Table 6. Fixed effects regression
Notes: Robust standard errors are in parentheses. ***Coefficients significant at 1 per cent, **5 per cent, *10 per cent. Table 6. Fixed effects regression
Related topics:
EconomicsDevelopment Studies
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved