This article was downloaded by: [University of Sussex Library] On: 05 April 2013, At: 04:48 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Journal of Development Effectiveness Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/rjde20 The impact of Kenya's Cash Transfer for Orphans and Vulnerable Children on human capital The Kenya CT-OVC Evaluation Team To cite this article: The Kenya CT-OVC Evaluation Team (2012): The impact of Kenya's Cash Transfer for Orphans and Vulnerable Children on human capital, Journal of Development Effectiveness, 4:1, 38-49 To link to this article: http://dx.doi.org/10.1080/19439342.2011.653578 PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and- conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.
Transcript
This article was downloaded by: [University of Sussex Library]On: 05 April 2013, At: 04:48Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Journal of Development EffectivenessPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/rjde20
The impact of Kenya's Cash Transferfor Orphans and Vulnerable Children onhuman capitalThe Kenya CT-OVC Evaluation Team
To cite this article: The Kenya CT-OVC Evaluation Team (2012): The impact of Kenya's Cash Transferfor Orphans and Vulnerable Children on human capital, Journal of Development Effectiveness, 4:1,38-49
To link to this article: http://dx.doi.org/10.1080/19439342.2011.653578
PLEASE SCROLL DOWN FOR ARTICLE
Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions
This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representationthat the contents will be complete or accurate or up to date. The accuracy of anyinstructions, formulae, and drug doses should be independently verified with primarysources. The publisher shall not be liable for any loss, actions, claims, proceedings,demand, or costs or damages whatsoever or howsoever caused arising directly orindirectly in connection with or arising out of the use of this material.
Journal of Development EffectivenessVol. 4, No. 1, March 2012, 38–49
The impact of Kenya’s Cash Transfer for Orphans and VulnerableChildren on human capital
The Kenya CT-OVC Evaluation Team∗1,2
Kenya’s Cash Transfer for Orphans and Vulnerable Children (CT-OVC) is a nationalchild-protection programme that provides a flat monthly transfer of Ksh 1500 toultra-poor families with orphans and vulnerable children aged 17 years and younger.A cluster randomised social experiment was conducted in 2007–2009 to evaluate theimpact of this programme. We use these data to provide an in-depth analysis of theeffects of the programme on children’s human capital development. Because basicschooling is free in Kenya and enrolment rates are relatively high, the scope of anunconditional programme such as the CT-OVC may be small. We use data from theevaluation baseline as well as national survey data to make ex-ante predictions aboutwhere the programme is most likely to have a big impact. We compare these predictionswith actual programme impacts as a way of assessing whether the programme has hadthe expected impact on children’s human capital development given the institutionalenvironment. We find that the programme has had an impact on the margins we wouldexpect, and the size of the impact on secondary school enrolment of this unconditionalprogramme is comparable with those from conditional programmes in other parts of theworld. The ex-ante analysis is crucial to understanding where to look to appropriatelyassess the impact of the programme.
Keywords: Social Cash Transfers; Kenya; children’s schooling; Africa
Introduction
The Kenya Cash Transfer for Orphans and Vulnerable Children (CT-OVC) is theGovernment of Kenya’s flagship social protection programme, currently reaching130,000 households and over 260,000 orphans and vulnerable children (OVC) across thecountry. The objective of the programme is to provide regular cash transfer payments tofamilies living with OVC to encourage fostering and retention of children and to pro-mote their human capital development. Eligible households, those who are ultra-poor andcontain OVC, receive a flat monthly transfer of Ksh 1500 (approximately US$20; thiswas recently raised in the 2011/12 Kenya budget). OVC are defined as household resi-dents between zero and 17 years old with at least one deceased parent, or a parent whois chronically ill, or whose main caregiver is chronically ill. Beneficiary households areinformed that the care and protection of the resident OVC is their responsibility for receiv-ing the cash payment. Currently there are no punitive sanctions for non-compliance withthis responsibility, although several districts will test punitive conditions in an upcomingexpansion of the programme scheduled for 2012.
*Email: [email protected]: The views expressed in the Work are those of the Author(s) and do not necessarily reflectthe views of the Food and Agriculture Organization of the United Nations.
The Government of Kenya, with technical and financial assistance from UNICEF,designed and began implementing the CT-OVC as a pilot in 2004. After a three-yeardemonstration period, the programme was formally approved by Cabinet, integrated intothe national budget and began expanding rapidly in 2007. Prior to programme expansionin 2007, UNICEF designed a social experiment to track the impact of the programme on arange of household welfare indicators including child health and schooling. The evaluationwas contracted to a private consulting firm, Oxford Policy Management, and entailed arandomised longitudinal design, with a baseline household survey (and related communitysurvey) conducted in mid 2007 and a 24-month follow-up in 2009 (Ward et al. 2010). Theethical rationale for the design was that the programme could not expand to all eligiblelocations at the same time, so locations whose entry would occur later in the expansioncycle could be used as control sites to measure impact. Thus within each of seven districtsacross the country (Kisumu, Migori, Homa Bay, Suba, Nairobi, Garissa and Kwale), fourlocations were identified as eligible, and two were randomised out of the initial expansionphase and served as control locations. Targeting of households was carried out in all fourlocations (per district) according to standard programme operation guidelines.
This article reports results on the impact of the programme on children’s humancapital. It is only the second published study that provides experimental impacts of anAfrican unconditional cash transfer programme on children’s schooling. The experimentalevidence to date on the relationship between schooling and cash transfer programmes is pri-marily based on evidence from conditional cash transfer programmes from Latin America(Schultz 2004, Handa and Davis 2006, World Bank 2009). But cash transfer programmesin Africa tend to be unconditional rather than conditional. That fact, along with the distinctsub-Saharan context (higher poverty, lower supply of services, higher prevalence of HIV),means that the evidence from Latin America and other countries on the impact of condi-tional cash transfers is unlikely to be valid for sub-Saharan Africa countries. Meanwhilecash transfers are expanding rapidly in the region, with national programmes or smallgovernment-led demonstrations now in place in Ghana, Burkina Faso, Liberia, Nigeria,Zambia, Malawi, Mozambique, Zimbabwe, Tanzania, Ethiopia, Lesotho and Uganda (plusthe universal programmes already in existence in Namibia, South Africa and Botswana).Given the significant public funds now being devoted to such programmes in Africa, it isimportant to begin to understand their behavioural impact to assess whether they representan effective development policy instrument in the region.
In terms of impact evaluation, as mentioned earlier only one published study existsthat provides experimental evidence on the impact of a national cash transfer in Africa onchildren’s human capital. This study, based on the Mchinji (Malawi) Social Cash TransferScheme, reports a 4 percentage point increase in school enrolment among interventionhouseholds relative to experimental controls (Miller et al. 2010) although the sample size issmall (400 households in each study arm) and the difference significant at 10 per cent only.On the other hand, a quasi-experimental evaluation of South Africa’s Child Support Grant,based on difference-in-differences (DD) propensity score matching, reports a statisticallysignificant impact of 7 percentage points on primary school enrolment (Samson et al.2010).
The present study adds to the literature on the impact of cash transfers on schooling inAfrica in several important ways. First, it is the only the second experimental impact assess-ment of a cash transfer in this context, and with a school-age sample of nearly 5000 childrenhas much greater power to detect significant effects than the Mchinji evaluation reportedin Miller et al. (2010). Second, we explore a range of schooling outcomes for differentage groups to understand the complete behavioural impact of the programme on children’s
Dow
nloa
ded
by [
Uni
vers
ity o
f Su
ssex
Lib
rary
] at
04:
48 0
5 A
pril
2013
40 The Kenya CT-OVC Evaluation Team
human capital. Finally, and unlike other impact evaluations we are familiar with, we usethe baseline data to provide ex-ante predictions about expected impacts, and then use theseas benchmarks to assess the success of the programme on schooling outcomes. This isespecially important in the Kenyan context because primary schooling is free so moneymay not be the main factor inhibiting demand among younger children. We argue that thisapproach should be used by all evaluations, when possible, as a way of judging the successof a programme.
Targeting in the CT-OVC
The programme employs a three-stage targeting process to select eligible households.In stage one, districts are chosen for inclusion into the programme based on overallpoverty levels and the prevalence of HIV/AIDS (directly related to OVC). In stage two,a community-based targeting process is implemented, led by members of the communitycalled the Location OVC Committees (LOCs). LOCs are responsible for compiling a listof eligible households within their designated location based on the eligibility criteria indi-cated above. Once this list is compiled, all members of the LOC meet to decide whichhouseholds qualify or not by discussing the eligibility and needs criteria. This prelimi-nary eligibility list is then sent to Nairobi. Enumerators then return to those householdsidentified by the LOCs as eligible and collect more detailed information on house-hold demographic composition, caregiver characteristics, and a series of proxy variablesintended to assess the household’s relative poverty status. A proxy means test is under-taken to ensure that households meet the poverty criteria, but since far more ultra-poorand eligible households are identified by the LOCs than can be served by the programme,an additional ranking system is employed to identify families with greater vulnerabilityso they can access the programme first. The ranking system first prioritises child-headedhouseholds (younger than 18 years of age) and, among them, households with more OVC,followed by the eldest caregivers and, within them, households with more OVC. Based onthis ranking and the total resources available for the location, a final list of programmerecipients is generated and validated by a community assembly, where programme offi-cers explain the rules of the targeting system and then announce each name out loud in theestablished order according to the priority criteria. The community can then raise concerns,doubts, and questions regarding the ranking of households for programme eligibility. Thesecases are reviewed and resolved before the final eligibility list is produced and householdsinvited for registration. The targeting performance of the programme is reported in Handaet al. (2012) in this special issue.
The evaluation sample
The evaluation sample was drawn from the programme eligibility lists compiled by thecommunity and ranked by the Ministry of Gender, Children and Social Development inthe seven selected districts. Households in either arm (Intervention, Control) were sur-veyed prior to their knowledge that they were selected into the programme. A total of1540 and 754 households were interviewed in intervention and control locations respec-tively at baseline. Table 1 provides some descriptive statistics for the two groups to assesstheir comparability. On two characteristics only, caregiver age and adults with completeStandard 8 schooling, there is a significant difference in means between the two groups butoverall the samples are well balanced, particularly in terms of monetary poverty and living
Dow
nloa
ded
by [
Uni
vers
ity o
f Su
ssex
Lib
rary
] at
04:
48 0
5 A
pril
2013
Journal of Development Effectiveness 41
Table 1. Means of selected characteristics of intervention and control households in evaluationsample.
Intervention (n = 1540) Control (n = 754)
DemographicsHousehold size 5.68 5.66Number of children age 0–17 3.39 3.53Number of orphans age 0–17 2.57 2.50Caregiver is male 0.14 0.13Caregiver chronically ill 0.142 0.155Caregiver age in years∗ 48 41No adults in household with Standard
8 schooling∗0.50 0.38
Poverty and living conditionsMonthly per-capita expenditure (Ksh) 6506 6571Walls made of mud/dung/grass 0.80 0.84Dirt floor 0.69 0.72Has electricity 0.06 0.07No linked toilet 0.57 0.54Unsafe drinking water 0.48 0.50Number of poultry owned 3.50 4.12Number of goats owned 0.65 0.72
∗Statistically significant difference in means.Source: Authors’ own calculations from baseline evaluation data.
Table 2. Mean characteristics of children aged six to 17 at baseline.
Intervention (n = 4082) Control (n = 2108)
Age (years) 11.35 11.04Female 0.44 0.48Currently enrolled 0.83 0.82Ever enrolled 0.88 0.86Current grade for those in school 4.63 4.46
conditions. Table 2 shows child-level descriptive statistics at baseline for residents age fiveto 17. Again these are comparable across the two arms and none of the mean differencesare statistically significant. Eighty-three per cent of children are currently enrolled, and themean grade for those currently in school is around 4.5.
Benchmarking programme impacts
An unconditional cash transfer programme such as the CT-OVC will exert a pure incomeeffect on household demand for human capital. There may also be a small substitutioneffect because at the time of enrolment caregivers are told that the cash transfer is to be usedfor the care and protection of resident OVC. However, there is no monitoring of behaviournor are there punitive sanctions and the transfer is not linked to individual OVC but rather isa flat transfer regardless of the number of OVC in the household. Consequently the substi-tution effect is likely to be negligible. We would therefore expect to see large programmeeffects in outcomes that are sensitive to income or total expenditure; that is, where theincome or total expenditure elasticity is large or where income plays an important role in
Dow
nloa
ded
by [
Uni
vers
ity o
f Su
ssex
Lib
rary
] at
04:
48 0
5 A
pril
2013
42 The Kenya CT-OVC Evaluation Team
restraining demand. In Kenya, government primary schools are free, but there are still out-of-pocket direct costs such as travel costs, food, uniforms and shoes. These out of pocketcosts are larger at the secondary level because of school fees and because the network ofsecondary schools is smaller so the average travel time, especially in rural areas, will begreater. Finally, the opportunity cost of schooling rises dramatically at the secondary level.For all these reasons we expect the CT-OVC to have a larger impact on secondary schooloutcomes (due to higher direct costs), among households that are farther away from schools(greater travel costs), and among older children (due to higher opportunity costs).
To understand more concretely the scope for the CT-OVC to impact children’s humancapital we provide descriptive information on the relationship among children’s school-ing, age and household expenditures. We use baseline data from the Oxford PolicyManagement evaluation sample as well as poor households from the 2006/07 KenyaIntegrated Household Budget Survey (KIHBS). The latter is a nationally representativemulti-topic survey conducted by the Kenya National Bureau of Statistics; poor householdsare those that fall below the Kenya national poverty line as constructed by Kenya NationalBureau of Statistics.
Figures 1 and 2 show the proportion of children that ever attended school and the pro-portion currently enrolled by total (adult equivalent) household expenditure (adeq in thefigure) and age respectively. These are local linear (lowess) regressions. The relationshipwith total expenditure is very strong until about Ksh 1500 per person per month (roughly66 US cents per person per day) and then flattens out. The relationship between everenrolled and age is strong and positive up to about age 10 and then flattens out, whilethe relationship between age and current enrolment follows this same pattern up to age10 but then shows a noticeable downward slope beyond age 14. These graphs suggestthat programme impacts for current enrolment among the entire sample of children maybe positive but very small. However there may be very large effects for very young chil-dren (under age nine or 10) and older children. The graphs suggest a similar pattern of
.3.4
.5.6
.7.8
.91
0.00 1000.00 2000.00 3000.00 4000.00
Monthly per adult equivalent total household expenditure
Ever Attended School by Expenditure
.3.4
.5.6
.7.8
.91
6 8 10 12 14 16 18
age in years
Ever Attended School by Age
KIHBS CT-OVCKIHBS CT-OVC
Figure 1. Ever attended school, household expenditure and age.
Dow
nloa
ded
by [
Uni
vers
ity o
f Su
ssex
Lib
rary
] at
04:
48 0
5 A
pril
2013
Journal of Development Effectiveness 43
.3.4
.5.6
.7.8
.91
0.00 1000.00 2000.00 3000.00 4000.00
Monthly per adult equivalent total household expenditure
Enrolment by Expenditure
.3.4
.5.6
.7.8
.91
6 8 10 12 14 16 18
age in years
Enrolment by Age
KIHBS CT-OVC KIHBS CT-OVC
Figure 2. School enrolment, household expenditure and age.
programme impacts for ever enrolled, except for older children where programme impactsseem unlikely since over 90 per cent of children age 14 and older have been to school atone point in their life.
Figures 3 and 4 show the relationship between household expenditure and two furtherschooling indicators for those currently in school - days missed in the last two weeks andgrades behind. There is a threshold in the number of days missed in the last two weeks attwo days so we construct a binary variable indicating whether or not a child missed two ormore days. We construct the grades behind variable assuming that the normal on-time startof Standard 1 is at age six. We calculate the grade a child ought to be in if he/she began
0.1
.2.3
.4.5
.6
0.00 1000.00 2000.00 3000.00 4000.00
Monthly per adult equivalent total household expenditure
Missed more than 2 days by adeq
L 2 wks (KIHBS), L month (CT-OVC)
KIHBS CT-OVC
Figure 3. Proportion missing two days of school by household expenditure.
Dow
nloa
ded
by [
Uni
vers
ity o
f Su
ssex
Lib
rary
] at
04:
48 0
5 A
pril
2013
44 The Kenya CT-OVC Evaluation Team
11.5
22.5
33.5
0.00 1000.00 2000.00 3000.00 4000.00
Monthly per adult equivalent total household expenditure
Grades behind (assuming start at age 6) by adeq
KIHBS CT-OVC
Figure 4. Grades behind by adult equivalent household expenditure (adeq).
school at age six and did not repeat any grade, and subtract from this his/her actual grade.Children who are on schedule will have a value of zero, while a value of two (for example)indicates the child is two years (or two grades) behind schedule. Higher values indicateworse outcomes for both indicators. Figure 3 depicts a fairly flat relationship betweenhousehold expenditure and days missed with even a slightly positive relationship if any-thing - we would therefore not expect to find an impact of the CT-OVC on this schoolingindicator. On the other hand, the number of grades behind (Figure 4) shows a very stronginverse relationship with expenditure, suggesting a large expenditure elasticity and thus thelikelihood of a large programme impact. Although not presented here, we also constructedgraphs by age to see whether there were any patterns that might suggest differential pro-gramme effects by age. There was no clear relationship between days missed and age butthe number of grades behind did increase significantly at ages above 13, suggesting thepossibility of a larger programme effect among older children.
To summarise, our analysis of the relationship among the four schooling indicators, ageand household expenditure provides a fairly clear picture of how the CT-OVC is expectedto impact schooling decisions. The over-impact on school-age children will be small forever enrolled and current enrolment, but potentially much larger for younger children andolder children. We would not expect an impact on attendance (days missed) but expectthe strongest impact of all on the number of grades behind. We would also expect largerimpacts where income is a more binding constraint, such as among households that livefarther away from schools.
Impact of the CT-OVC on schooling outcomes
We estimate programme impacts using the DD estimator for all children age six to 17.We restrict the sample to households that appear in both waves of the data, leaving us with4082 and 2108 children in intervention and control households respectively. The empiricalspecification includes controls for time (baseline or follow-up), treatment status, whetherany adult resident has completed Standard 8 schooling, household size, number of children,age and sex of child and urban residence. The interaction of time and treatment statusprovides the DD impact estimator; we cluster standard errors at the household level ratherthan the location level because we believe the strongest intra-class correlation derives frommultiple children in the same household rather than households within the same location.3
Dow
nloa
ded
by [
Uni
vers
ity o
f Su
ssex
Lib
rary
] at
04:
48 0
5 A
pril
2013
Journal of Development Effectiveness 45
Table 3. Impact estimates on ever and current enrolment by age.
Percentage change from mean 0.56 6.93 −1.55 3.38Observations 9268 1149 5093 4175R-squared 0.091 0.079 0.151 0.03Baseline mean 0.885 0.606 0.838 0.948
Currently enrolledDD impact 0.02 0.047 −0.019 0.078
(1.21) (0.70) (0.91) (3.38)Percentage change from mean 2.38 7.93 −2.29 9.12Observations 9253 1144 5083 4170R-squared 0.022 0.077 0.143 0.055Baseline mean 0.839 0.593 0.828 0.855
Note: Coefficient is DD impact estimate with controls for time, treatment status, head’s schooling, age, sex, andurban residence. t-statistics, accounting for household clustering, reported below impact estimate. Significantcoefficients in bold.
Table 3 shows DD impact estimates for ever and current enrolment by age and theresults conform to our ex-ante expectation about what we might reasonably see from theprogramme. Among all children there is no programme effect on either outcome. However,among older secondary school children (age >12 years) the programme effects are signif-icant and meaningful – for current enrolment there is a 7.8 percentage point programmeimpact representing a 9 per cent increase over the baseline mean. Table 4 reports the DDimpact estimates for grade progression and grade-for-age. Grade-for-age was explainedearlier and is defined only for children currently in school (higher values are worse out-comes); grade progression is a binary variable indicating whether a child moved aheadfrom the previous to the current year. This is defined only for children who were enrolledin both 2008 and 2009. The results in Table 4 are derived using follow-up data only onchildren who were enrolled at baseline. Three of the six coefficients in this table are sig-nificant at the 10 per cent level and the largest quantitative impacts are again found amongolder children. For example, children in intervention households are 0.096 fewer gradesbehind (about 7% at the mean); these children are also 5 per cent more likely to progress tothe next grade. We experimented with alternative definitions of on-time school entry, usingages five and seven, but this did not affect the results shown in Table 4.
Grade-for-age is affected by successful promotion so it is not surprising that the impactsfor progression and grade-for-age are consistent. However, the CT-OVC could also pullsome children back into school, which could decrease grade-for-age if these children havebeen absent from school for a number of years. The programme can also reduce drop-outs, which will affect grade-for-age. Table 5 explores some of these possible pathwaysby estimating treatment effects for the probability of dropping out and returning to schoolusing the follow-up data only. A child is defined as a drop-out if she/he was enrolled atbaseline but not at follow-up, while a returner is one who was not enrolled at baseline butwas enrolled at follow-up. Results indicate that the CT-OVC has increased the probabilityof returning among older children (age >12 years) by around 2 percentage points.
One-quarter of all children age five to 17 are not OVC according to the programmedefinition. This is because many orphans are cared for by aunts and uncles and so live with
Dow
nloa
ded
by [
Uni
vers
ity o
f Su
ssex
Lib
rary
] at
04:
48 0
5 A
pril
2013
46 The Kenya CT-OVC Evaluation Team
Table 4. Impact estimates on grade progression and grade-for-age.
Note: Cross-section estimates using follow-up data only and children who were enrolled at baseline. Coefficient oftreatment dummy variable with t-statistic (adjusted for clustering at household level) in parenthesis. Regressionsalso control for head’s schooling, urban residence and age and sex of child.
Table 5. Impact estimates on drop-out and returning to school.
Note: Cross-section estimates using follow-up data only. Coefficient of treatment dummy variable with t-statistic(adjusted for clustering at household level) in parenthesis. Regressions also control for head’s schooling, urbanresidence and age and sex of child. Significant coefficients in bold.
cousins or other relatives who are not OVC. We have repeated the analysis reported aboveusing only the 75 per cent of children who are defined as OVC and find identical results forall outcomes except for school enrolment of children older than age 12, where the impactestimate increases slightly to 8.0 percentage points.
Heterogeneous treatment effects by school costs
Theory suggests that the impact of the CT-OVC would be especially strong among house-holds for whom schooling is relatively more expensive, either because of higher time ormoney costs. The community questionnaire reports the distance to the nearest primary andsecondary school. It also reports whether uniforms and shoes are required by the schooland, in the case of primary schools, whether there are ‘extra fees’ associated with attendingthe school. Using this information we construct two indicators that reflect the ‘price’ ofschooling separately for secondary and primary schools. The first is a dummy variableindicating whether the school (either primary or secondary) is more than 2 kilometresaway as this was a clear threshold in the distribution of distances. The second is a costindex, which is the sum of dichotomous variables indicating whether the school will notallow students to attend without shoes, without uniforms, and for primary schools only,whether or not extra fees are charged.
Dow
nloa
ded
by [
Uni
vers
ity o
f Su
ssex
Lib
rary
] at
04:
48 0
5 A
pril
2013
Journal of Development Effectiveness 47
The cost index for primary schools ranges from zero (no extra costs) to three with amean of 1.5. About one-quarter of children must pay extra fees, 50 per cent must wearuniforms and 77 per cent must wear shoes. Only 10 per cent of children live more than2 kilometres from a government primary school. At the secondary level, the cost indexranges from zero to two with a mean of 0.56, and about one-half of all children live morethan 2 kilometres from the nearest government secondary school.
The empirical approach is to interact each of the two ‘price’ variables with the DDvariable and look for a statistically significant coefficient, which would indicate a hetero-geneous treatment effect; that is, a difference in the impact of the programme depending onthe ‘price’ of schooling for that household. We expect higher prices to reduce schooling.But if the programme mitigates this negative relationship then the interaction term will bepositive (except for grades behind where a mitigating effect would lead to a negative inter-action coefficient because higher values are worse). Table 6 shows results for children agedunder 13 years and here we begin to see the real benefit of the CT-OVC programme forprimary-aged children. In all cases except grade progression the programme significantlymitigates the negative effect of primary school costs. In other words, the programme hasa particularly strong impact among households that face larger price barriers to primaryschooling. Table 7 repeats this exercise for older children and secondary school ‘prices’.Again there are strong indications that the programme also mitigates some of the negativeimpacts of secondary schooling, particularly for grades behind where the marginal effectis −0.324.
Overall the cost-mitigating effects of the programme are much stronger at the primarylevel but recall that the overall programme effects (on all children) are much larger atthe secondary level. This is consistent with the institutional environment in Kenya wherethe largest cost barriers are at the secondary school level and not at the primary schoollevel. Thus the programme has its strongest impact among secondary school children, but
Table 6. Impact estimates by cost of primary schooling (age <13 years).
DD interacted with: Ever enrolled Currently enrolled Progression Grades behind
Note: Significant coefficients in bold. T-statistics below estimates in parenthesis.
Dow
nloa
ded
by [
Uni
vers
ity o
f Su
ssex
Lib
rary
] at
04:
48 0
5 A
pril
2013
48 The Kenya CT-OVC Evaluation Team
also has important positive impacts at the primary level among those households that facehigher price barriers.
Discussion and conclusion
We report only the second published experimental estimates of the impact of an Africanunconditional cash transfer programme on children’s schooling. We go beyond most pro-gramme evaluations of such programmes by first understanding the margins along whichthe programme might be expected to impact schooling decisions given the institutionalenvironment and the observed relationship between income and schooling indicators. Thisis vital to understand in the Kenyan context where primary schooling is free and enrol-ment rates are very high. Our ex-ante analysis suggests that the programme will have themost impact among older children in secondary level, and on schooling efficiency (grade-for-age). The subsequent impact analysis is consistent with these predictions. While nomeaningful impact is found at the primary level, large ones are found at the secondarylevel for overall school enrolment and grade-for-age. Further probing of the data unveilsthat the programme has managed to encourage school returners.
A recent World Bank (2009) review of the impact of conditional cash transferson schooling reports impact estimates from a set of countries (see Table 5.1; 2009,pp. 128–129). Estimates on school enrolment range from 1.9 to 21.4 percentage points forsamples that include males and females, but the age groups vary. In three instances, esti-mates are reported for slightly older children who are more comparable with the secondaryage group of 13–17 reported in this paper. Those three impact estimates are 12 (Bangladeshages 11–18, females only), 5.6 (Colombia ages 14–17) and 5.2 (Turkey, secondary school-ing) while the comparable estimate we report here is 7.8 percentage points. While this is acrude comparison because, as is well known, impacts will vary by the size of the transfer(as a proportion of recipient household income or consumption), baseline schooling lev-els and institutional context (whether schooling requires large out-of-pocket expenses), itnevertheless illustrates that the schooling impacts of the Kenyan unconditional CT-OVC iswell within the range of impacts observed elsewhere for conditional programmes.
The lack of strong impacts at the primary level is to be expected given the institutionalsetup in Kenya. However, costs can still vary depending on distant to schools and enforce-ment of policies around shoes and uniforms. We look for, and find, strong evidence thatthe programme has mitigated the adverse effect of these costs at the primary level. Thus,among households living over 2 kilometres from a primary school the treatment effect oncurrent enrolment is 19 percentage points higher, and 6 percentage points higher for eachunit increase in the primary school cost index. These more nuanced results indicate thatthe programme is having an important positive impact on schooling among householdsthat are most price constrained – those with secondary school age children and those fac-ing unusually high primary school costs despite the ‘no-fee primary school’ policy of theGovernment of Kenya.
What is the external validity of these results for unconditional cash transfer pro-grammes in sub-Saharan Africa? We believe there is very strong external validity for threereasons. First, the results reported here are based on a rigorous and credible research designand so can be taken as causal impacts of the programme. Second, the institutional con-text in Kenya is very similar to other sub-Saharan African countries currently initiatingsuch programmes. That is a context where primary schooling is mostly free and almostuniversal, significant drop-out occurs at the transition from primary to secondary edu-cation, and out-of-pocket costs for secondary education are high. And third, the design
Dow
nloa
ded
by [
Uni
vers
ity o
f Su
ssex
Lib
rary
] at
04:
48 0
5 A
pril
2013
Journal of Development Effectiveness 49
features of the CT-OVC are typical of other programmes in the region and are feasi-ble in terms of human resource requirements and financial costs. These features includeunconditional transfers, community involvement in beneficiary selection, targeting basedon vulnerability and poverty rather than just poverty, and a transfer level that representsabout 20 per cent of mean consumption of the beneficiary population.
Acknowledgements
The authors thank the following individuals for their contributions to the programme and theevaluation: Nils Riemenschneider, Clare O’Brien, Ian MacAuslan and Jack Willis.
Notes1. The Kenya CT-OVC Evaluation Team is comprised of (in alphabetical order) Carlos Alviar
(UNICEF-Ghana), Benjamin Davis (FAO-Rome), Sudhanshu Handa (University of NorthCarolina), Alex Hurrell (Oxford Policy Management), Ahmed Hussein (Ministry of Gender,Children & Social Development), Daniel Musembi (Ministry of Gender, Children & SocialDevelopment), Samuel Ochieng (Ministry of Gender, Children & Social Development), RogerPearson (UNICEF-Ethiopia), Luca Pellerano (Oxford Policy Management), Aly Visram (OxfordPolicy Management), and Patrick Ward (Oxford Policy Management). The corresponding authorfor this article is Sudhanshu Handa.
2. The evaluation of the Kenya CT-OVC was conducted by Oxford Policy Management undercontract to UNICEF – Kenya and data collection was implemented by Research SolutionsAfrica. The evaluation was overseen by a steering committee comprising technical staff from theMinistry of Gender, Children & Social Development of the Government of Kenya, UNICEF –Kenya, UNICEF-ESARO, DFID and the World Bank. The results that appear in this article arethe culmination of over three years of intellectual, technical, financial and operational efforts ofa large and dedicated team and authorship is listed jointly to recognise the important contribu-tion of the key individuals in producing the results reported here. Ahmed Hussein is the contactperson for the CT-OVC.
3. There are 203 communities in 28 locations across seven districts. We choose to cluster at thehousehold level because we believe that from a behavioural perspective the most important cor-relation occurs at the household level. We have repeated our estimates with standard errorsclustered at the community level and significance remained the same except for ‘ever enrolled’among children older than age 12, where significance drops to 0.07.
ReferencesHanda, S. and Davis, B., 2006. The experience of conditional cash transfers in Latin America and
the Caribbean. Development policy review, 24 (5), 13–36.Handa, S., et al., 2012. Targeting effectiveness of social cash transfer programs in three African
countries. Journal of development effectiveness, forthcoming.Miller, C., Tsoka, M., and Reichert, K., 2010. Impacts on children of cash transfers in Malawi.
In: S. Handa, S. Devereux and D. Webb, eds. Social protection for Africa’s children. London:Routledge Press, pp. 96–116.
Samson, M., et al., 2010. Impacts of South Africa’s child support grants. In: S. Handa, S.Devereux and D. Webb, eds. Social protection for Africa’s children. London: Routledge Press,pp. 117–145.
Schultz, T.P., 2004. School subsidies for the poor: evaluating the Mexican Progresa poverty program.Journal of development economics, 74 (1), 199–250.
Ward, P., et al., 2010. Kenya CT-OVC programme operational and impact evaluation 2007–2009.Oxford: Oxford Policy Management.
World Bank, 2009. Conditional cash transfers: reducing present and future poverty. Washington,DC: World Bank.
