RESULTS-BASED FINANCING RBF EDUCATION EVIDENCE MOZAMBIQUE Can Information and Incentives Increase School Attendance? MARCH 2018 REACH funded an evaluation that compared the effectiveness of three different interventions designed to increase school attendance among grade six and grade seven girls. The Results in Education for All Children (REACH) Trust Fund supports and disseminates research on the impact of results-based financing on learning outcomes. The EVIDENCE series highlights REACH grants around the world to provide empirical evidence and operational lessons helpful in the design and implementation of successful performance-based programs. Although more children than ever RBF mechanism that has been used Results-based financing has been used to incentivize are starting school in Africa, in many in many developing countries to parents and students to countries dropout rates remain high incentivize individuals to take actions improve school attendance and few students complete their that they may not otherwise take, and achievement. schooling, especially girls. Results- such as attending school or using based financing (RBF) has been preventive health services. CCTs used in many developing countries work by giving individuals a cash Attendance to attempt to incentivize various transfer, conditional on verification M T W Th F stakeholders such as students, that they have completed the parents, and teachers to achieve prescribed behavior. CCTs have been better results. RBF mechanisms shown to be effective in increasing work by linking financial incentives school attendance in many countries, More research is needed to measurable results, for example but their cost and complexity to identify the role that school attendance, dropout rates, makes them difficult to manage for information sharing plays. or student test scores. Conditional countries with limited administrative cash transfers (CCTs) are one such and budgetary capacity. This note was adapted from de Walque, Damien and Christine Valente (2018). Incentivizing School Attendance in the Presence of Parent-Child Information Frictions, Policy Research Working Paper 8476, World Bank, Washington D.C. 2 RFB EDUCATION | EVIDENCE The Results in Education for All were conditional on the girls financial incentives to parents. Given Children (REACH) Trust Fund at the maintaining a high level of school the low cost of providing information World Bank funded an evaluation that attendance. The second intervention and the ease with which such an compared the effectiveness of three provided parents with a cash transfer initiative could be scaled up, this different interventions designed to conditional on their daughters’ may be a promising policy option increase school attendance among attending school. The third simply for countries with limited capacity. girls enrolled in grades six and seven provided students’ parents with In addition, providing financial in Mozambique. The first intervention weekly report cards detailing their incentives directly to students was provided girls with vouchers that daughters’ attendance. These at least as effective as providing three interventions were designed incentives of equal monetary value to to explore (a) whether providing their parents. Furthermore, providing information to improve parental incentives to students was nearly monitoring could be as effective in twice as effective as providing increasing school attendance as information alone. financial incentives and (b) whether financial incentives were more Both the information intervention and effective when given directly to the intervention that provided students students or to their parents. with incentives improved math test scores significantly, although This evaluation found that all three providing the parents with incentives interventions significantly increased did not. This suggests that increased girls’ school attendance. Providing attendance can increase cognitive information alone had a substantial skills under certain conditions, but that effect, which was not significantly CCTs targeted to parents may have different from the impact of giving counterproductive effects. CONTEXT years old. For upper primary school where the intended completion age is 13 years old, the completion rate Despite large increases in enrollment in rural areas by age 19 is only 8 rates in lower primary school grades, percent for girls and 14 percent for most children in Mozambique are still not completing primary education. In boys.1 The slower progress for girls is 2016 only 45 percent of girls and 51 problematic both from an equity point percent of boys completed primary of view and because of the positive Mozambique school, and only 18 percent of both externalities from women’s schooling girls and boys enrolled in secondary on households and society. school, even though net enrollment In 2016, only for grades one to seven exceeded This study was conducted in Manica, a 18% 85 percent for both girls and boys. province located in the Center Region Rural students are also less likely of Mozambique, which is home to 7.5 to complete primary school. In rural percent of the country’s population. of children areas where most children live, only It is close to the national average enrolled in 60 percent of children who start lower on a number of indicators, including population density, poverty rate, and secondary school primary school finish by the age of 19, even though the intended age is 10 the primary school dropout rate.2/3 MOZAMBIQUE 3 WHY WAS THE INTERVENTION CHOSEN? CCTs are one of the most important would value a weekly report showing and effective policy tools for increasing their daughter’s attendance, and school enrollment and attendance and 98 percent of those said that the reducing dropout rates. While as of report would improve their ability to 2009 CCTs had been implemented in monitor their child’s attendance. at least 29 developing countries and in nearly all cases have been effective in In addition, there has been little increasing school attendance,4 there research on the household are several unanswered questions dynamics that drive decisions about about the optimal design of these whether children attend school financial incentives. and on how financial incentives may affect these decisions. By One highly debated question is about shedding light on the way in which the role that conditionality plays in schooling decisions are made by the effectiveness of these transfers. both parents and children within the While the requirement (or condition) household, this study can help to to attend school in order to receive inform the optimal design of CCT the cash transfer has the effect of programs, particularly by answering increasing parents’ or students’ the question of whether transfers are returns to education, it also has more effective when given to parents the more subtle benefit of helping or students. Children and parents parents to monitor their children’s may have different views on when it attendance. Although parents cannot is optimal for the children to invest directly observe their children’s in their human capital by attending attendance, they can infer it from the school. A study in Malawi found fact that they receive a conditional that increasing the amount of cash transfer. Parents have been found to transferred to the household, either value this type of monitoring, as in to the parents or to the child, did not Brazil where a CCT program reduced increase the effectiveness of a CCT the information gap between parents incentivizing attendance.6 Another and children and enabled parents study in India that incentivized This study can help to better enforce attendance.5 This students’ performance on a literacy raises the question of whether the test rather than their attendance to inform the optimal information component of CCTs found no evidence that the identity could be an effective policy tool in of the incentive recipient matters design of CCT programs, its own right, even without financial incentives. Initial qualitative surveys on average, but shows that parent incentives are more effective relative particularly by answering in Manica found that both girls to child incentives in cases where the question of whether and their parents take part in the parents have greater abilities to decision about whether to attend teach and motivate their children.7 transfers are more school and that parents have However, no study has investigated incomplete information on their the relative effectiveness of giving effective when given to child’s school attendance. Eighty transfers exclusively to children or percent of parents said that they to parents to incentivize attendance. parents or students. 4 RFB EDUCATION | EVIDENCE HOW DID THE school and a cross for each day missed. The report cards were given spot checks. Crucially, financial incentives were paid only on the INTERVENTION to the girls at the end of each week to show their parents and bring them basis of the report cards and not the spot checks to minimize the WORK? back to school the following week. In cases where girls or parents earned likelihood that the spot check data would be manipulated, a common One hundred and seventy-three financial incentives for attendance, problem when RBF incentives are “complete” primary schools (schools these report cards were used to linked directly to the main outcome offering all seven grades of primary verify attendance. of interest. Other outcomes that the education) were randomized into four intervention compared among the Schools were randomly assigned to four groups were school enrollment groups: three treatment groups and one of the four groups. In schools rates, students’ test scores, child one control group. The schools in in the first group in addition to the autonomy and empowerment each of the three intervention groups report cards, girls who achieved a (whether girls are able to keep their were provided with attendance report 90 percent attendance rate received earnings and whether they are cards for each girl in senior primary money-equivalent vouchers that involved in decisions concerning school (grades six and seven). These could be used to buy items such as their healthcare, school attendance, simple report cards used a coding school uniforms, shoes, and bags. or work outside the house), the that could be easily understood even The vouchers were worth roughly quality of the parents’ attendance by illiterate parents: the teacher drew US$8 per trimester or about US$25 monitoring, and girls’ marital status. a circle for each day a girl attended annually, which is equivalent to These outcomes were measured by eight times the daily wage of an fielding baseline and endline surveys agricultural worker in Manica. In to each household, which also asked the schools in the second group, in for self-reported information about Randomized addition to the report cards, parents expected returns to education, whose daughters achieved a 90 students’ cognitive tests, and Control Trial percent attendance rate received household expenditures on girls’ (173 schools participated) the same amount of cash (around personal items. US$25 a year), which could be used to purchase the same items. The intervention was announced in Treatment Group 1 In the schools in the third group, February 2016 at the beginning of Attendance report card + vouchers the girls and their parents received the school year. The implementing given to girls with 90 percent the report cards with no financial NGO, a well-known development attendance organization called Magariro, visited incentive. Finally, in the control group schools, girls received no attendance each school and informed school Treatment Group 2 report cards or incentives. Girls staff, students, and parents that Attendance report card + cash in all four groups were monitored there would be unannounced visits to given to parents of girls with 90 for attendance using independent collect attendance data one to three percent attendance unannounced “spot checks” times per trimester. In addition, for conducted twice per trimester by the schools in the three treatment Treatment Group 3 Intercampus, Lda, a survey firm. groups, Magariro explained the Attendance report card only relevant intervention, distributed The aim of the intervention was to the attendance report cards, and Control Group compare the proportion of enrolled answered questions. The transfers girls who were in attendance at were distributed at the end of each No interventions used each school in each group, as trimester for the girls who met the measured by the independent 90 percent condition. MOZAMBIQUE 5 WHAT WERE THE RESULTS? Compared to the control group, that providing information as part of Student test scores were improved all three treatment interventions a CCT can have a substantial effect by the girls’ incentives and the significantly increased school on school attendance, independent of information intervention but not attendance. In the control group any financial incentive. by parents’ incentives. While schools, 60.7 percent of enrolled giving CCTs to parents raises girls were in attendance during Incentivizing students directly was students’ attendance, these gains the independent spot checks. In at least as effective as incentivizing in attendance do not translate comparison, attendance was 4.5 parents. The estimated effect of into gains in their test scores. This percentage points higher in schools incentivizing girls was 38 percent finding is consistent with many where only information was provided, larger than the effect of incentivizing other research studies on the same 6.0 percentage points higher in parents, although with the available question. In contrast, both the schools where parents received cash sample size this difference was not information intervention and the incentives, and 8.3 percentage points statistically significant. Furthermore, girls’ incentives increased student higher in schools where girls received incentivizing girls directly was nearly test scores on the Annual Status vouchers. The impact of all three twice as effective as simply providing information, and this difference of Education Report (ASER) math interventions was significant at the was significant. Since children have test by 8.3 percent and 9.3 percent 5 percent level. perfect knowledge of whether they respectively. The magnitude of these Providing information alone had attend school on any given day learning gains was large, roughly a significant effect on school but parents do not, an incentive equivalent to half of the difference in attendance, even without any given to students may have been scores between students completing financial incentive. The estimated more effective because it was a grade six and those completing effect on attendance of providing direct inducement to attend school grade five. These results suggest report cards to parents was roughly whereas the giving an incentive of that increasing attendance is 75 percent as large as the effect of the same size to their parents could beneficial for the development of giving parents financial incentives, only induce them to try to compel students’ cognitive skills but that and this difference was not their children to attend school, with CCTs targeted to parents may have statistically significant. This suggests less than perfect success. counterproductive effects. 6 RFB EDUCATION | EVIDENCE The interventions had no significant is not surprising given the low statistically insignificant. In contrast, effect on school enrollment rates, variation in parents’ responses to in all treatment groups, the number girls’ marital status, girls’ autonomy, these survey questions. Ninety- of absences reported by the parents or teacher absenteeism. Self- seven percent of parents claimed was positively correlated with the reported school enrollment from the that they knew whether their child probability of being absent on the household survey started from a was at school each day, and only day of an independent attendance very high baseline rate of 95 percent. 6 percent answered that there had spot check, and the size of this Therefore, it is not surprising that ever been a day when they thought correlation implied large increases this intervention had little effect on their daughter was at school but in parental information–from enrollment decisions, particularly actually she was not. It is likely that 46 percent (in the information given the small size of the transfers this reflects some unwillingness by treatment group) to 72 percent (in and the fact that the intervention parents to acknowledge their lack of the parents’ CCT group) of what was announced close to the start control over their children, especially would be predicted under perfect given that 80 percent of parents information. of the school year. The information methodology and the parents’ CCT said that attendance monitoring None of these interventions had both had a large but statistically would be useful. A more objective an effect on measures of girls’ insignificant effect on the likelihood of measure of the quality of parents’ autonomy and empowerment. girl students getting married. These information about their daughters’ Lastly, although teachers were results suggest that better parental school attendance is the correlation given small amounts of cellphone monitoring may help to prevent early between the number of absences credit as compensation for the extra marriage while girls are in school, but reported by the parents during a work of completing the student they are too imprecisely estimated given month and whether the girl attendance report cards, these three to make it possible to draw definitive was absent during an independent methodologies had no effect on conclusions. attendance spot check at the school teacher absenteeism. This indicates during the same month. In control that their impact on attendance These interventions had no schools, the correlation between and test scores was driven by effect on the self-reported quality these two variables was only 20 the behavior of the parents or the of parents’ monitoring their percent of what would be predicted children and not by differences in daughters’ attendance, which under perfect information and was school environments. The results confirm those from other studies that have found that even small incentives can have substantial effects, which helps to ensure the financial sustainability of such transfer programs. MOZAMBIQUE 7 WHAT WERE qualitative surveys no girl said she was forced to give or sell her voucher a stated attendance target per student in the information only THE LESSONS to someone else. The findings should, therefore, be interpreted in light of the treatment group. This was done to minimize any “public recognition” LEARNED? fact that the vouchers remained with the girls themselves and were valued incentive. Similarly, the possibility of a “salience” effect (in which One critical consideration in any social by them, suggesting that the choice parents and children become more transfer program is determining the of the recipient of the incentive was aware of the importance of school size of the financial incentive. In this meaningful. attendance) was minimized by study, the amount given to either introducing the report cards without Nevertheless, even though the two any accompanying message about girls or their parents was relatively incentives had the same monetary the importance of school attendance. small, up to a maximum of roughly value, they may not have been US$25 annually. This amount was Furthermore, independent equally valued by their recipients. chosen in order to provide sufficient attendance spot checks were Because the vouchers could incentives within the range of other carried out in all schools, including only be used to purchase certain CCTs throughout the world while still those in the control group (although items whereas the cash transfers ensuring sustainability within the the information gathered during could be used to buy anything, Mozambican government’s budget the spot checks was not given to the more fungible cash may have constraint. The results confirm those parents), so if there was a salience been perceived as more valuable. from other studies that have found effect, it is likely to have affected Furthermore, the vouchers had a that even small incentives can have students in control schools as well. negative effect on the purchases of substantial effects, which helps to Nevertheless, it is possible that the other goods, as parents presumably ensure the financial sustainability of total effect of the report cards may substituted their own spending such transfer programs.8 have been greater if they had in fact away from these children. Therefore, been accompanied by recognition, the net effect of the vouchers on In addition, to ensure comparability messages about the importance of girls’ disposable income was only between the two types of financial school attendance, and/or by more 80 percent of the effect of the cash incentive, it was critical that the extensive attendance monitoring. transfers on the parents’ income. On vouchers given to the girls were of It is important for policymakers to the other hand, financial incentives roughly the same value as the cash disentangle these potential drivers may be more effective when given transfers given to the parents. The and for these questions could be directly to children because they parents’ cash transfers were equal to explored in future research to find comprise a larger proportion of the value of the girls’ vouchers, and the the optimal combination of design children’s income. Policymakers prices of the items to be purchased should be careful to consider all features to maximize the impact of were the same whether they were of these factors and unintended information and financial incentives paid for with cash or vouchers. consequences when determining on school attendance. Furthermore, these items were chosen the size and recipient of financial based on information collected in incentives and should find out what Lastly, buy-in and support from both focus group interviews with girls incentivizes recipients. teachers and study surveyors was aged between 11 and 15 and their critical in collecting student attendance parents in Manica. The respondents The designers of the intervention data and validating its quality and consistently mentioned these items took care to rule out any alternative reliability. Therefore, similar initiatives as the ones most likely to incentivize interpretations of the outcomes of in the future should ensure that girls to attend school regularly and issuing attendance report cards. teachers and other stakeholders are those that the girls would be able to First, no public ceremonies were equally committed to providing reliable keep for themselves, without being held to acknowledge girls with attendance data that can inform forced to share them with others. In high attendance nor was there parents and incentivize students. Students have CONCLUSION parents about children’s attendance can increase attendance substantially, agency in decisions One of the most important although not as much as giving financial incentives to students. regarding their innovations in social policy in the last few decades has been Given that providing information is own schooling and the introduction of conditional less costly and complex than making financial transfers, this may be a cash transfers given to parents to incentivizing them incentivize prescribed behaviors promising and easily scalable policy such as school attendance. However, option for governments in developing directly may be this study found that students as countries who lack the administrative and budgetary capacity to implement more effective than young as those in grade six have agency in decisions regarding their a conditional cash transfer program. incentivizing their own schooling and that incentivizing Lastly, while providing cash transfers to parents does not translate into test students directly may be more parents. effective than incentivizing their score gains, both the information only parents. This finding is an important methodology and giving incentives lesson about how best to design such directly to girls increased math test financial incentives. Furthermore, scores, with an effect equivalent to simply providing information to roughly half a year of learning. 1 World Bank Education Statistics Database, 2016. 2 INE (2015). Statistics and Social Indicators (2013–2014). National Institute of Statistics, Maputo. 3 MPD-DNEAP (2016). Poverty and Wellbeing in Mozambique: Fourth National Poverty Assessment, Ministry of Planning and Development, National Directory of Studies and Policy Analysis, Maputo. 4 Fiszbein, Ariel and Norbert Schady, eds. (2009). Conditional Cash Transfers: Reducing Present and Future Poverty. World Bank Publications, Washington, DC, USA. 5 Bursztyn, Leonardo and Lucas C. Coffman (2012). “The Schooling Decision: Family Preferences, Intergenerational Conflict, and Moral Hazard in the Brazilian Favelas.” Journal of Political Economy, 120(3), pp.359–397. 6 Baird, Sarah, Craig McIntosh, and Berk Ozler (2011). “Cash or Condition? Evidence from a Cash Transfer Experiment.” The Quarterly Journal of Economics, 126(4), pp. 1709–1753. 7 Berry, James (2015). “Child Control in Education Decisions. An Evaluation of Targeted Incentives to Learn in India.” Journal of Human Resources 50.4: 1051–1080. 8 Fiszbein, Ariel and Norbert Schady, eds. (2009). Conditional Cash Transfers: Reducing Present and Future Poverty. World Bank Publications, Washington, DC, USA. PHOTO CREDITS: Cover: “Island of Mozambique” by GPE/Arnaldo Langa, license: CC BY 2.0 Page 2: Project photo courtesy of the World Bank. Page 3: “Education in Mozambique: students listening in class” by GPE/Arnaldo Langa, license: CC BY-NC-ND 2.0 Page 5: “Education in Mozambique” by GPE/Arnaldo Langa, license: CC BY-NC-ND 2.0 Page 6: Project photo courtesy of the World Bank. RESULTS IN EDUCATION FOR ALL CHILDREN (REACH) worldbank.org/reach REACH is funded by the Government of Norway through NORAD, the Government of the United States of America through USAID, and the Government of Germany reach@worldbank.org through the Federal Ministry for Economic Cooperation and Development.