Evaluating CTs’ impacts

Results from Malawi’s Zomba CT evaluation

Results from Malawi’s Zomba CT/CCT

The Zomba CT in Malawi, a small program designed explicitly for research purposes, tested the impact of conditional and unconditional transfers on educational, marriage, and fertility outcomes for female adolescents. Results from the program's impact evaluation are presented below.

Schooling impacts of CCT and UCT arms

Results two years into the program showed that the conditionality in the transfers was driving key educational outcomes. Although enrollment improved in both the UCT and CCT beneficiary groups, the improvement for the UCT was less than half (43%) of the improvement of the CCT group. School attendance also improved in the CCT arm over that in the UCT and control arms (Baird, McIntosh, and Özler 2010, see Figure A). The two-year results using independent tests of learning outcomes revealed that the CCT also improved learning outcomes. The UCT was not found to improve these outcomes.

Figure A: Schooling results from Zomba CT

total_terms_enrolled2009_school_attendance

Notes:
Left graph is from Table III in Baird, McIntosh, and Özler (2010). Dependent variable equals total number of school terms (out of six possible) that the beneficiary was enrolled during the program. Sample size is 852. Right graph is from Table V in Baird, McIntosh, and Özler (2010). Sample size is 319. Sample is composed of all respondents with attendance information for at least one of the three school terms. Overall attendance is determined by dividing number of days attended out of number of possible term days for terms with available data. For both graphs, results are from OLS regressions run with robust standard errors clustered at the local (EA) level. Controls included are age dummies, strata dummies, a household asset index, highest grade attended, and a dummy for never had sex, all at the baseline. Regressions are weighted to be representative from the EA target populations. *** indicates result is different from control group at 99% level; ** indicates result is different from control group at 95%. †† indicates CCT result is different from UCT results at 95% level.

Impacts on pregnancy and marriage

The evaluation found that the CCT was not effective in preventing teen pregnancy and marriage. The UCT arm, however, was able to significantly decrease the probability that girls would become pregnant or get married (see Figure C). This impact was primarily due to the UCT’s impact on adolescent girls who dropped out of school after the program began. No effects of the UCT or CCT were found on girls who remained students (Baird, McIntosh, and Özler 2011).

Figure C: Marriage and pregnancy outcomes from Zomba CT

ever_married_mapever_pregnant_map

Notes:
Results are from Table VII from Baird, McIntosh, and Özler (2010). Sample size of unique observations equals 2,089. Graphs show results from fixed-effects regressions using robust standard errors clustered at the local (EA) level and weighted to represent the target population in relevant EAs. *** denotes a result that is significant at 99%; ** a result significant at 95%, and * a result significant at 90%. † indicates that the CCT arm differs from the UCT arm at the 90% confidence level; †† indicates the CCT and UCT are significantly different at the 95% level.

This outcome highlights a trade-off inherent in CCT and UCT programs. The CCT was able to encourage human capital formation among girls who complied with the program requirements, but it kept transfers from girls who dropped out of school and were more vulnerable to adolescent pregnancy and early marriage. The CCT was also able to keep girls from dropping out of school, reducing marriage through this channel, but the overall income effect of the UCT on girls’ marriage was larger than this given the large group of CCT non-compliers. As the authors conclude, “This study makes clear that while CCT programs may be more effective than UCTs in obtaining the desired behavior change, they can also undermine the social protection dimension of cash transfer programs” (Baird, McIntosh, and Özler 2011, 35).