Do Poverty Reduction Programmes Foster Education Expenditure? New Evidence from Rwanda

Abstract

The Rwandan Government has implemented various education policies that contribute to higher enrolment in education, but has become aware that these policies might be less effective for children from poor families. This study investigates the contribution of poverty reduction programmes on education expenditure of households. Using a multi-level regression analysis, combining district data on labour markets with detailed expenditure data on 7,230 households, it teases out the effects of several social protection programmes. The results show that access to health insurance and to waged work are positively related, while direct financial or in kind support are negatively related to paying into the children’s schooling. Non-agricultural employment opportunities in particular seem to stimulate education investments. Reducing the vulnerability of households might provide more equal access to these opportunities.

Keywords

Poverty reduction education expenditure resources dilution employment opportunities Rwanda

Introduction

The Government of Rwanda has implemented various policies to stimulate enrolment of children in education with a specific emphasis on improving the gender balance. Among others, fee-free education has been extended from six years to nine years for all and an extra three years for those successful in the first nine. In many respects these policies have met success. Attendance in primary education is close to 100% for both boys and girls, and the completion rates by girls are above those of boys (Table 1). The decrease of gender inequality is supported by the data on the 2011 Primary Leaving Exams: girls made up 54 per cent of the 167,166 registered candidates (Kwizera et al., 2011). Yet the overall completion rates at age 12 are still below 50% and the youth literacy rate (80% for 15–19 years old) is not exceptional for Sub-Saharan Africa and is closer to South than to East Asia (Nabalamba and Sennoga, 2014). This raises a wider concern that not everyone benefits from education as a development path, which has also been forcibly expressed for India by Corbridge et al. (2013) and Drèze and Sen (2013). Nkurunziza et al. (2012) found for Rwanda, that despite improvements, in particular children from very poor and from large households had lower chances of educational enrolment up to 2006. More generally it turned out that Rwanda’s development policy was less suited to target the very poor (Ansoms, 2008). The Rwandan Government decided to introduce a Social Protection Policy that provides various forms of support for the poorest. This raises the question to what extent these policies have also provided better results when it comes to education.

Table 1.

Primary school enrolment and its completion status in 2011.

Gender	Age group	Attend primary school (%)	Age group	Completion of primary school (%)
Male	8–10 years	2,774 (98.8)	13–15 years	440 (17.5)
Male	11–12 years	1,694 (98.6)	16–17 years	549 (35.5)
Female	8–10 years	2,764 (99.4)	13–15 years	561 (22.4)
Female	11–12 years	1,791 (98.6)	16–17 years	661 (41.6)

Source: Prepared by the authors on basis of the information provided by the 2011 Integrated Household Living Conditions Survey (IHLCS 2011).

This paper adds to this debate on the role of poverty reduction strategies in the educational domain in several ways. First, many studies have used educational attainment as a proxy for educational demand (e.g. Pushkar, 2003). It focuses on education expenditure instead of educational attainment because it directly measures the willingness of parents to pay for their children’s education (Vu, 2012) in a context of scarce resources. Accordingly, it investigates to what extent various Social Protection interventions have an impact on the amount of family expenditure on education, controlling for other variables like the number of under school age children in the family, that define this expenditure. Second, it acknowledges that education expenditure is more than just out-of-pocket costs as these should also be seen as an investment that enables access to better paid job opportunities. This is a reason why expenditure is higher in urban than in rural areas. Yet, rather than including urban rural differences it has chosen to specify this opportunity structure at the district level and to include this in a multilevel model, controlling for the fact that poor people tend to live in poor areas. For the analyses a dataset of 7,230 households taken from the 2011 Integrated Household Living Conditions Survey (IHLCS)¹ conducted by the National Institute of Statistics of Rwanda (NISR) and district data from other secondary sources are used. The aim of the research is to show the effect of poverty reduction measures on education expenditure within the wider context of economic development.

The next section provides an overview of the background of the study while the third section outlines the hypotheses on the predictors of household education expenditure. The fourth section explains the methodology and presents the sample population. The fifth section gives the results of the multilevel regression analysis, followed by a discussion and conclusion in the last sections.

Background: Rwanda Vision 2020

Despite a positive economic development trajectory of more than a decade, Rwanda still remains a poor country (Siegel et al., 2011). Even if the situation has improved to some extent, many Rwandans continue to live at levels below the poverty line (NISR, 2012a: 17). Besides, Rwanda has a high population density (467 people per square kilometre) and a young population. Due to a high fertility level, which remained as high as 6.1 children per woman until 2005, two out of three Rwandans are under the age of 25 (Ruberangeyo et al., 2011). The fertility level dropped recently – thanks to among other factors improved access to family planning facilities (Muhoza, 2014) – but, with 4.3 children now, it is still twice the level common in East-Asian countries. The Rwandan population will continue to grow as a result of its current age composition. It will take some time before the country can profit from the so-called demographic dividend as countries in Asia and Latin America already have done (Drummond et al., 2014; Gribble and Bremner, 2012).

The aim to reduce population growth was only one of the outcomes of many internal debates held at the end of the 1990s to address Rwanda’s multi-faceted social, political, and economic challenges (Hayman, 2007: 372). The main policy objectives were presented in the strategic policy paper locally known as Rwanda Vision 2020. The aim of the Rwanda Vision 2020 was to transform Rwanda’s agricultural-based economy into a knowledge-based one within two decades. To achieve those objectives the Government has applied various development strategies such as reducing poverty, educating the youth and offering the population better access to (reproductive) health facilities. It was acknowledged that, in order to succeed, these strategies must pay serious attention to reducing gender disparities (King and Mason, 2001; Zuckerman, 2001). Table 1 illustrates that currently, in quantitative terms, a gender balance has been achieved in school enrolment and in completion rates. Several policies have contributed: the abolishing of school fees; school feeding programmes in food insecure districts; community pressure against non-attendance through the establishment of a Parents Teachers Association (PTA) committee at each school; and support to orphans and other vulnerable children (Nkurunziza et al., 2012). Rwanda now invests 22.1 per cent of its budget in education, a bit above the Fast Track Initiative (FTI) benchmark of 20 per cent (UNESCO, 2012a).

Rwandan children officially start their formal education at the age of 7 years and undergo a primary cycle of 6 years before entering the lower secondary level of 3 years followed by a higher secondary level of 3 years. Since 2003, significant priority has been given to a reform of the education system. As a result, enrolment in primary education has improved impressively since 2005–2006. In 2010–2011 the proportion of youth aged 8 to 12 years in primary school was over 98 per cent (Table 1). In order to create more places for students at secondary level and to increase efficiency, a number of strategies were adopted. The establishment of schools locally known as Nine Years Basic Education schools (9YBE) was one of them. Tuition is fee-free and privately organized schools (through churches in particular) receive Government funding.

Privately funded schools are a choice of the parents who have the capacity and the willingness to support the related high costs. In 2013 only 3 per cent of the primary school pupils attended a privately funded school, for secondary school students this percentage was 17 per cent (MINEDUC, 2014).

Lessons learned from the evaluation of the first Vision 2020 strategy (2002–2005) indicate that despite the fact that this strategy did indeed provide policies for enhancing development and growth, the chosen implementation was less suited to target the (very) poor (Ansoms, 2008). After 2005, economic development and poverty reduction policies were re-focused on equitable and sustainable growth, with rural development as an important priority (Ansoms, 2008; Evans et al., 2006). In the second strategy, various policy measures have been re-labelled and re-packed under the term ‘Social Protection’ and the adoption of a special Social Protection Policy at the end of 2005 paved the way for reducing the inequality by targeting poor people in particular (GoR, 2007). This approach is expected to be a prerequisite for making poverty reduction successful (Sebates-Wheeler and Roelen, 2011).

The Social Protection instruments, which essentially comprise different programmes such as the Vision 2020 Umurenge Project (VUP), are geared to support the poor. One Cow Per Family (OCPF) as well as the Community Based Health Insurance (CBHI), were applied to enable poor and vulnerable people to control the forces that condition their lives.

The Vision 2020 Umurenge Project has developed various types of support for the poor. The beneficiaries for each type are selected on the basis of a locally determined Ubudehe² classification³. The direct financial support (VUPDS) are monthly payments to the poorest households that have no or insufficient land (less than 0.25 hectares) and no able-bodied members who can work. The support through public works (VUPPW) gives poor households with able-bodied members access to paid activities (every two weeks) in employment guarantee schemes. Another type of support increases the access of poor households to financial services (VUPFS) and is helping them to become familiar with the banking system and to deal with financial institutions (Kindness, 2011; Siegel et al., 2011),

The OCPF programme – locally known as GIRINKA – launched in 2006, provides poor households with a dairy cow (through a gift or loan) and has various goals. First, it seeks to reduce malnutrition through increasing milk consumption of the rural poor. Second, it provides manure for soil fertility improvement on the land of the beneficiaries, or an opportunity to produce biogas for cooking. Third, it promotes social cohesion through a system where the first born calf is passed on to another household in need. Finally, it should create opportunities for earning an additional income and contribute to a feeling of self-confidence (Bucagu, 2013).

The introduction of the CBHI system enables low-income households to manage their financial risks and reduce their vulnerability in the face of financial shocks due to health problems (Ahuja and Jutting, 2004). Based on household subscription, the CBHI – locally known as Mutuelles de santé – was reinitiated as a pilot project in 1999; its up-take accelerated sharply in 2004–2005 with the adoption of a national policy on Mutuelles and a roll-out of the schemes with the financial and technical support of the development partners (Twahirwa, 2008). Initially the Mutuelles covered the costs of all services and drugs provided by the local health centres and of services in district hospitals, but already in 2006, a co-payment of Rwandan francs (Rwf)200 (US$0.33) per visit at the health centres was introduced. In July 2011, the flat annual premium of Rwf1000 (US$1.66) per household member, was replaced by a stratified system⁴ for the annual premiums according to the Ubudehe poverty classification. This was in order to ensure the financial sustainability of the Mutuelles scheme⁵ while still taking equal access for all members of society to health facilities into account (Binagwaho et al., 2012).

Crucial for achieving Vision 2020 will be to properly link education policies with economic development and labour market policies. The Government of Rwanda (GoR) (2000) argued that the emergence of a viable private sector that develops into the principal growth engine of the economy is absolutely crucial to its development. It was estimated that it will be necessary to create 1.4 million jobs outside the agricultural sector by 2020. A growing demand for educated labour can stimulate parents to invest in their child’s education. ‘To attract foreign investors, the Government of Rwanda has strengthened the country’s institutional position by implementing new business reform legislation containing arbitration laws and regulations to fight bankruptcy’ (KPMG, 2012: 7). Besides, it has reduced administrative barriers by the establishment of the Rwanda Development Board (RDB). The RDB functions as a ‘one-stop’ investment services centre where potential investors can get assistance to obtain required approvals, certificates, work permits, tax incentives and land registrations. ‘As a result of these reforms, Rwanda recorded a major improvement in Doing Business, jumping from the 150^th to the 58^th position between 2007 and 2010’ (Economisti Associati et al., 2011: 7).

Hypotheses on household education expenditure

Education can be viewed conceptually as both consumption and investment. If education expenditure is viewed as consumption, household schooling decisions are determined by interaction of social, cultural and economic factors working through power relations within the household (Al-Samarrai and Peasgood, 1998). According to its socio-economic and demographic position, the household is faced with two constrains: scarcity of (financial) resources; and costs of other competing basic needs such as expenditure on food, water, health, clothing and housing. It is expected that poor and extreme poor households will have more difficulties in dealing with these constraints and consequently spend less on education per child than non-poor households. Important confounding factors with a possible (opposite) impact in this matter are family size and family support.

In large families, children have to compete for the limited parental resources and probably not every child can receive formal education. This interaction between sibling size and parental resources – known as resource dilution (Downey, 1995) – results in less money spent for education per child as the sibling size increases. In many African societies, transfers between families and family members are frequent, which could help poor families to increase consumption or to pay for their child’s education. Researchers have argued that in Sub-Saharan Africa, the extended-family system and the practice of fosterage in particular, redistribute resources across families in a way that buffers educational inequality (Akresh, 2005; Lloyd and Blanc, 1996). Therefore, it is expected that transfers between families could contribute to either higher average expenditure on education (in case of receipts) or lower expenses (in case of donations to others).

Besides financial support from family members or other private persons, several other sources of funding exist in particular for special groups of children in Rwanda. Different Ministries (Ministry of Local Government (MINALOC), Ministry of Education (MINEDUC) and Ministry of Gender and Family Promotion (MIGEPROF)), local and international non-governmental organizations, churches, in collaboration or individual, have programmes focusing on vulnerable children like orphans, street-children, HIV-infected children, and children from HIV–AIDS-infected parents. This support can include payments of educational costs (Paxton and Mutesi, 2012; Ruberangeyo et al., 2011; World Bank, 2011). For children in high education and from poor families, the Government provides school bursaries in the form of loans that must be partially reimbursed after graduation.

If households consider the decision to send a child to school also as an investment, they take into account the expected future returns. If rates of return to education are high – meaning that education pays off with improved opportunities (better paid jobs and more successful private initiatives) – households may choose to invest in education in order to increase the earning capacity and other benefits in the future (Tilak, 2002). The current labour market situation in terms of available jobs that require educated workers is expected to have a positive influence on education expenditure. This will be in particular the case as parents or other household members themselves are already educated and profit from their training. It is expected that households with caretakers that have attained upper secondary education will spend more on educating their children, as has been found in other studies (Handa et al., 2004; Kreft and De Leeuw, 1998; OECD, 2011).

The kind of relation between children and caretakers could also have an impact on the expected balance between costs and profits of investments in education. This balance between current costs and future returns is probably negative in the case of orphans and foster children who are less likely to be enrolled in school than children who live with their biological parents (Bhalotra, 2003; Siaens et al., 2003).

Considering education as an investment could mean that for some families, the opportunity costs of schooling can be high even under a fee-free schooling regime, due to the loss of family earnings from child labour. However, the most important reason why poor children do not enrol in schools is that their parents cannot afford to pay the direct and indirect expenses that school attendance incur (Caillods et al., 2009). The absence of adequate resources hampers learning, and discourages enrolment and continuation to higher grades (Van der Berg, 2008).

The main research hypotheses for this study are based on the assumption that the Rwandan Social Protection policy as part of the poverty reduction strategies, contribute to decreasing extreme poverty and vulnerability. It is expected that self-resilience and livelihood security increases thanks to the existence of programmes such as VUP, OCPF and CBHI that provide for (indirect) food and income support (Sweetman, 2011). A higher social security is expected to contribute to higher investments (expenditure) in education, in other words in the quality of education for children in Rwanda (Lee, 2004).

The expected contribution of participation in a health insurance programme needs some further explanation. Unexpected health expenses in case of illness of one of its members reduce the available household budget for food, education, farming expenses and other expenditure (Wang et al., 2006). It is expected that households protected from catastrophic health spending by having a CBHI, are more confident to invest in the education of their children compared to households without this insurance. In addition the chance to seek for medical care doubled when Rwandese had a Mutuelles (Lu et al., 2012), which could indicate that having a CBHI contributed to lower indirect costs and reducing the poverty trap resulting from being ill (McIntyre et al., 2006)

The rates of return to education can be analysed from both the demand and supply side of the labour market. For employers, the improvement in education will build a more productive and efficient workforce. For workers, the availability in a district of employment for the better educated is a sign of higher future returns on household education expenditure.⁶ As a consequence higher education expenditure with increasing formal sector employment at district level is expected. The effects of employment opportunities in the public and formal sectors on household’ education expenditure is compared to the effect of wage labour in the agricultural sector. Besides, this distinction points at rural–urban differences, because wage labour on farms is predominant in rural districts, while the other two types of employment are mainly established in urban districts.

Data and methodology

The IHLCS 2011 conducted by the NISR provides socio-demographic and economic data on the members of 14,308 households. The households that had at least one child full-time at school during the 12 months preceding the survey as a research population (7,230 households) were selected.

The dependent variable in this study’s analysis is the Average Household Education Expenditure per Child (AHEEC)⁷ which equals the total household education expenditure divided by the number of children at school. The household education expenditure combines school tuition and registration fees, parent contributions, costs for school uniforms, books and other supplies, school transportation, eventual boarding costs and other schooling expenses. The use of the log transformation of the AHEEC instead of its value in Rwf⁸ is preferred first because of its convenience for transforming a highly skewed variable into one that is more approximately normal distributed and second, because it allows us for the use of the percentage changes in the AHEEC.

To test the hypotheses elaborated on in the previous section, a multilevel model is required. The hypotheses examine the effect of level 2 variables (like the employment opportunities at district level) on the outcome variable (household education expenditure per child) while controlling for other level 1 variables (household characteristics). The logical consequence of this approach is that the variable ‘education’ measures different things, depending on the level of analysis. If the intra-class correlation is not significantly different from zero, no district differences exist for the variable of interest (Kreft and De Leeuw, 1998).

The chosen predictors of the AHEEC are thus classified at household or at district level. The data for the district level (in total 30 districts) are taken from the Rwanda Education Statistics (GoR, 2012a) and the IHLCS3 Thematic Report on Economic Activity (NISR, 2012b).

Table 2 presents the descriptive statistics of the variables used in the model and the accompanying average education expenditure. The results give a picture of the characteristics of the sample population. A further clarification is added below for only those variables with a specific classification.

Table 2.

Descriptive statistics.

Categorical variables	Observations n=7,230 (%)	Mean of AHEEC (US$)	Standard deviation of mean
Occupation of household head (HH)
Farm activities	6,528 (90.3)	8.55	0.63
Non-farm activities	702 (9.7)	34.90	0.84
Family poverty level
Non-poor family	3,540 (48.9)	18.02	0.73
Poor family	1,682 (23.3)	7.00	0.51
Extreme poor family	2,008 (27.8)	4.43	0.49
Number of under school age siblings
None	3,052 (42.2)	17.41	0.75
1 Child	2,023 (28.0)	7.64	0.60
2 Children	1,656 (22.9)	5.60	0.52
3 Children	451 (6.2)	5.17	0.47
4 children	47 (0.7)	5.51	0.52
5 Children	1 (0.0)	0.69
Type of children supported by family
Son/daughter of HH	6,266 (86.7)	7.57	0.60
Son/daughter of HH and other children	964 (13.3)	52.83	0.73
External intervention in child education
Family without external intervention	3,650 (50.5)	7.65	0.59
Family with external intervention	3,580 (49.5)	12.70	0.74
Household CBHI Status
Family without CBHI	1,971 (27.3)	7.06	0.63
Family with CBHI	5,259 (72.7)	11.09	0.69
VUP Public Works (VUPPW)
Family out of VUP Public works	7,136 (98.7)	9.88	0.68
Family into VUP Public works	94 (1.3)	5.23	0.51
VUP direct support (VUPDS)
Family out of VUPDS	7,179 (99.3)	9.75	0.68
Family into VUPDS	51 (0.7)	7.92	0.63
VUP Financial Services (VUPFS)
Family out of VUPFS	7,121 (98.5)	9.75	0.68
Family into VUPFS	109 (1.5)	11.19	0.54
One Cow for Poor Family (OCPF)
Family out of OCPF	6,022 (83.3)	9.86	0.69
Family received a cow	466 (6.4)	10.04	0.65
Family received other animal¹³	742 (10.3)	9.18	0.65
Education level of HH
HH non-educated or up to 5 primary	4,906 (67.9)	8.11	0.63
HH with primary completed	1,867 (25.8)	11.58	0.68
HH with more than primary school level	457 (6.3)	38.36	0.80
Presence of parents
Both parents presents	4,980 (68.9)	8.81	0.64
Father only	377 (5.2)	11.85	0.74
Mother only	1,858 (25.7)	12.50	0.75
Double orphans	15 (0.2)	16.63	0.95
Type of school and level of education
Child at public primary school	3,694 (51.1)	8.18	0.56
Child at private primary school	3,127 (43.3)	7.41	0.63
Child at public secondary school	1,023 (14.2)	49.76	0.57
Child at private secondary school	953 (13.2)	66.97	0.60
Child at public university school	85 (1.2)	270.93	0.52
Child at private university school	58 (0.8)	449.64	0.49
Continuous variables	Minimum	Maximum	Mean (SD)
Log (transfers received)	0	7.40	39.81 (0.68)
School quality at district level
Class-size in primary	65	114	85 (8.31)
Employment status at district level
Percentage of wage farm employment	2.7	22.7	10.18 (4.19)
Percentage of public sector employment	3.2	16.8	9.07 (3.01)
Percentage of formal private employment	0.4	18.7	2.64 (3.68)

The family poverty level is calculated on the basis of the household consumption expenditure including purchases, but also consumption from other sources like own production and payments received in kind.⁹ This measurement of poverty differs somewhat from the one used for the implementation of the social protection policy that is based on the resource constraints of Ubudehe classifications. The mean of the annual AHEEC estimated at US$4.43 for an extreme poor family can appear to be small, but in reality this amount is very high for a family that is struggling to survive with less than US$1.25 per day.

This study distinguishes between households that pay all education costs by themselves and the ones that receive also external funding from relatives outside of the household or from non-related persons and institutions. To avoid endogeneity problems in this study’s analysis, it was checked whether the received transfers¹⁰ were meant for child’s schooling.

For the employment status, the percentages of wage workers in the agricultural sector and percentages of employees in the public and formal private sector were used.

The multilevel linear model examines separate linear regression models in each district, followed by an examination of the relation between the parameters of the models and district characteristics. Thus, this multilevel regression decomposes the total variances into within-district and between-district components. Following Peugh (2010), the question of whether a multilevel model is needed in the case of this study becomes: ‘to what extent is the AHEEC variation present at district level?’ Answering this question involves the calculation of the intra-class correlation (ICC) and of the design effect statistics. Using the AHEEC from the IHLCS dataset, it can reasonably be expected that the AHEEC will vary across households within a district due to household differences in financial resources and motivation.

Results

Table 3 presents the estimates of the multilevel linear regression analysis. Model 1 shows only the intercept and model 2 the results of the predictors of the household expenditure on education on two levels. On district level the class-size in primary schools and employment opportunities in the public sector were dropped. Those variables proved not to be significant. All household characteristics are binary except the transfers received and the number of under school age children which are continuous; the average household education expenditure per child (AHEEC) increases or decreases according to the sign and magnitude of each parameter.

Table 3.

Estimates for multilevel linear regression of average household education expenditure as a function of household and district characteristics.

Fixed effect	Model 1		Model 2
	Estimate	SE	Estimate	SE
Intercept	3.80***	0.06	3.47***	0.06
Household characteristics
Household head (HH) in farm activities (Reference Category (Ref. cat.)HH in non-farm activities			0.14***	0.02
Number of under school year child			–0.04***	0.01
Non-poor family (Ref. cat.)Poor family			–0.15***	0.01
Extreme poor family			–0.26***	0.01
External interventions and social protection programmes
Log (transfers received)			0.05***	0.01
Family without external intervention in education cost (Ref. cat.)Family with external intervention in education cost			–0.03***	0.01
Family without CBHI (Ref. cat.)Family with CBHI			0.04***	0.01
Family out of VUPPW (Ref. cat.)Family into VUPPW programme			0.09**	0.04
Family out of VUPDS (Ref. cat.)Family into VUP direct support			–0.11*	0.06
Family out of VUPFS (Ref. cat.)Family into VUP financial services			0.05	0.04
HH out of OCPF and without other animal (Ref. cat.)HH got one cow from Government or non-governmental organization			–0.05***	0.02
HH got other animal than cow			–0.01	0.02
HH head education level and presence of parents status
HH non-educated or up to 5 primary (Ref. cat.)HH with primary completed			0.05***	0.01
HH with more than primary education level			0.16***	0.02
Both parents present (Ref. cat.)Double orphan			0.20*	0.11
Mother only			0.01	0.01
Father only			0.03	0.02
Control variables
HH supporting their son/daughter only (Ref. cat.)HH supporting both sons/daughter and other childrenChild at public primary school (Ref. cat.)			0.52***	0.02
Child at private primary schools			–0.03***	0.01
Child at public secondary schools			0.54***	0.02
Child at private secondary schools			0.70***	0.02
Child at public university schools			0.64***	0.05
Child at private university schools			0.75***	0.06
Employment status at District level (continuous variables in %)
Percentage of wage farm employments			–0.01***	0.004
Percentage of private formal employments			0.02***	0.005
Random effect	Variance component	SE	Variance component	SE
Residual within district variance σ²_e	0.397	0.007	0.165	0.003
Residual between district variance σ²_μ0	0.088	0.023	0.053	0.019
Variance of wage farm employment σ²_μ2			0.000	0.000
σ_μ0j			–0.004	0.002
Deviance	13,963.8		7,587.6
Number of parameters	2		18

Model 1= model with intercept only, Model 2= model with household and district characteristics, *** significance at 1%; **significance level at 5%; *significance at 10%; SE: Standard error.

A hypothesis test of random variance is useful to assess the necessity of hierarchically structured data. The results from Model 1 (Table 3) show that the null hypothesis (H₀: σ²_e =0) is rejected, suggesting that some significant covariance exists among households in districts. With the existence of covariance, applying traditional regression models violates the assumption of independence of observations and increases type I errors (Kreft and De Leeuw, 1998).

The intercept only or unconditional model, estimates the intercept as 3.80 (US$10.44), which is the average of education expenditure per child across all households and districts. The between-district variance symbolized by σ²_u0 , is estimated as 0.088, p<0.001. The within-district variance symbolized by σ²_e , is estimated as 0.397, p<0.001. Following Hox (2010), the intra-class correlation (ICC), calculated by ρ= σ²_μ0/(σ²_e + σ²_μ0), is 0.18. Thus, 18 per cent of the variance of the AHEEC occurred across districts, which is high. Since the intercept-only model contains no explanatory variables, the residual variances represent unexplained error variance. The ICC measures the degree of similarity within the same district (Muthén and Satorra, 1995). Given the results in this study’s model, the ICC indicates the need for a multilevel approach for the chosen dependent variable.

The covariance between the regression coefficient for household characteristics and the intercept is very small (σ_μ0j=-0.004) and obviously not significant. The deviance in Table 3 is a measure of model fit. When predictor variables are added, the deviance is expected to go down (Hox, 2010) which is the case in model 2. The deviance drops from 13,963.8 in the unconditional model to 7,587.6 in the model with household and district predictors of household education expenditure. Compared to model 1, model 2 shows a better fit: having (D₁-D₂) = 6,376.2 with (m₂-m₁) = 16 degrees of freedom and a p-value below 0.001.¹¹

Testing the hypotheses on household education expenditure it is seen in the model, that families increase their education expenditure by 0.5 per cent (US$0.55)¹² when they receive 1 per cent more transfers or increase them by 4.2 per cent (US$0.49) when they have a health insurance. The families are forced to reduce the expenditure in education by 4.3 per cent (US$0.46) for each additional under school age child, by 15.3 per cent (US$1.43) when they are poor, by 26.1 per cent (US$2.18) when they are extreme poor, and remarkably by 5.2 per cent (US$0.54) when they profit from the one cow per family programme. The household education expenditure increases by 9.1 per cent (US$1.11) for families involved in employment schemes (VUP public works) but decreases by 10.8 per cent (US$1.06) when the family receives direct financial support.

When other factors are controlled for, on average, a household with a head in non-farm activities invests 14.3 per cent (US$1.88) more in education of the children compared to the ones whose household heads are involved in farm activities. The more educated the household head, the higher the investments in children’s education. The proportion of investment in education of the two distinguished groups is respectively 5.2 per cent (US$0.61) and 16.3 per cent (US$2.20) higher compared to investments of a household head without a completed primary education.

As expected, the expenditure increases substantially when a child in the household proceeds to higher education levels. Also the differences in costs between public and private schools show up in the expenditure made. Finally, the results indicate that (the few) double orphans in the sample still profited from special support given to this group.

Model 2 includes also the employment sector composition at district level. The increase of wage farm workers in a district by one unit decreases the household education investment per child by 1.1 per cent (US$0.12), while an increase of formal sector employees in a district by one unit, increased the household education investment by 2.3 per cent (US$0.26).

Discussion

After the reconstruction period in the late 1990s, Rwanda has been on a positive development trajectory ever since. With assistance of its donors, the Government of Rwanda implemented various development policies amongst others to improve the health and education level of the population and to reduce poverty. According to Malmberg (2008), efficient policies that promote health, increases in education and improvements in infrastructure can be pivotal for a shift towards more favourable demographic trends.

The achievements are positive: almost all children regardless of gender enrol in primary education; and the last two Demographic and Health Surveys showed decreasing mortality levels and a fast increasing number of women using modern contraceptives. Regarding the transition from primary to lower secondary level, Rwanda is one of the few Sub-Saharan countries that managed to boost the number of lower secondary students by 25 per cent within a year by the introduction of a Nine Year Basic Education (9YBE) cycle and the elimination of fees for lower secondary school in 2009 (UNESCO, 2012b).

To achieve the Millennium Development Goal number 2, universal completion of primary education, further policy steps had to be taken after the introduction of free education in 2003. This paper has analysed the impact of special programmes for enhancing the ‘social protection’ of very poor families on the investments of these households in the quality of their children, notably by sending them to school.

Applying a multilevel linear regression method on a dataset of households of the 2011 household survey showed the impact of some direct poverty reduction measures on family education expenditure. The impact of poverty on household education investments is illustrated by the large gap between the education expenditure of the poor and of the extreme poor compared to that of non-poor families. Poor people may consider education as a positive investment but the costs are still too high even within a fee-free education context. This study’s results are quite similar to the conclusion of others (Glewwe and Jacoby, 2004; Megumi, 2010; Vu, 2012 ; Ulusoy and Yolcu, 2013) that there is a positive and significant relationship between changes in wealth and an increase in educational expenditure.

This study’s results also show that access to basic health services and access to wage work in employment schemes are related to a modest increase in household education expenditure. In contrast, getting direct financial support or an animal for free is negatively related to the education expenses made by the households involved. Given the methodology adopted care should be taken in interpreting these outcomes as cause and effect, and it is too early to tell whether the various policies are indeed effective or detrimental. Yet most of the coefficients seem plausible in the light of other evidence.

The health insurance coverage was added in the model because empirical evidence shows that CBHI not only has a strong impact on access to health care, but is also associated with a higher degree of financial risk protection (Saksena et al., 2011). The stratified system for annual premiums according to the Ubudehe poverty classification has provided equal access to health facilities, and hence, decreased the utilization gap between poor and wealthier families. Lower incidence of catastrophic income shocks due to illness and related health care expenditures are associated with a significant increase in household budget available for food, asset accumulation and education expenditures. This might account for the positive relation between CBHI participation and the education expenditures of households.

The same line of reasoning can be applied to the positive sign of the employment schemes. According to Siegel et al. (2011), public works can reduce vulnerability, build resilience and increase incomes and food security in a virtuous cycle that links social protection with disaster risk reduction/food security (anti-erosive ditches, radical terraces and valley dams) and climate change.

Access to waged work provided in public employment schemes gives a regular income every two weeks. Since all VUP beneficiaries have their own bank accounts, they can benefit from a range of financial and insurance services, including savings (which are voluntary but highly encouraged) and access to subsidized credit. Savings are used for basic consumption needs, health insurance, school fees and investments in small livestock which are also a form of ‘self-insurance’ against negative impacts of different hazards (Kindness, 2011). Again, more secure household budgets will allow for more investment in schooling.

Public works programmes (VUPPW) have a double dividend because they create off-farm employment. In a country like Rwanda where almost 90% of the active working population is employed in agriculture and the median land size per household is less than one hectare, non-agricultural employment opportunities like VUPPW seem to provide an alternative to on-farm labour. If education expenditure is viewed as consumption, the non-agricultural employment opportunities form an important part of household income diversification and risk reduction strategies (see also Ellis and Freeman, 2004). For rural landless and near-landless households income flows are fluctuating and unpredictable, in particular as almost all farming is rain-fed. Most of those households will depend heavily on non-agricultural income sources (Madaki and Adefila, 2014).

It is plausible that reducing risks of income shocks by a health insurance (Binagwaho et al., 2012) or access to regular paid work and credit make parents more confident to spend money on the education of their children.

That still leaves the question as to why direct financial support has a negative rather than a positive effect on education expenditure. This is in line with the findings of Devereux (2011) that poverty reduction strategies like the VUP programmes have short-run effects and a limited multiplier effect for the beneficiaries. Households that receive direct financial support have no land or less than 0.25 hectares and their members are unable to work because of age, disability or illness. For this type of household, expenditure for basic needs such as food, clothes and housing have a higher priority than expenditure on education; Devereux (2011) states that transfers might be too small, have limited duration, or are given erratically. Under those conditions, households can meet their immediate needs but are not able to improve their livelihood activities in a sustainable way, neither will they develop confidence that their income is stable or will increase in the future. Consequently, conditions that are expected to foster education expenditures are not met.

The negative effect of the one cow policy on household education expenditure is more difficult to analyse. Households that qualify for this programme must have access to 0.75 hectares of land, of which a third is pasture, and also must have or construct a cowshed. The negative score on education expenditure of these households probably points to extra costs that getting a cow may imply, such as purchase of fodder and veterinary care. In case the produced milk is consumed by household members and not sold, the objectives of the GIRINKA programme can be met only partly. It will take time before the household can make extra money out of cattle ownership. Allocating land to fodder production can also be an impossible option as it could compromise the own food production. Fodder availability differs strongly among participating farmers due to differences in available land size and its productivity (Bucagu, 2013), and collecting sufficient fodder for free in densely populated areas is not always an option. Evaluations of the one cow per family policy showed that it can contribute to poverty alleviation (Rutareka, 2011), but as this study’s analysis indicates it does not lead to more household investments in education.

The hypothesis of resources dilution is confirmed by a decrease in household education expenditure per child for each addition of an under school age child in a household. The effect of this dilution can materialize as a drop-out of an older sibling, but also by a reduction of expenses on school items or by a change of type of school attended. In a situation with limited resources, the presence of young siblings could push their older brothers out of school to assist in the family’s economic activities, and push older sisters out of school to perform domestic chores at home (Greenspan, 1992). In Rwanda, both mechanisms seem to occur. By combining gender and the parental co-residence status, Nkurunziza et al. (2014) found that girls without a mother and boys without a father had higher chances of dropping out. As this study analysed the average expenditure on education per child in the household (total expenditure divided by number of school-aged children) it was unable to show whether households spent more on their sons than on their daughters. In the African context, the dilution of parental resources is less problematic because the extended family system and the practice of fosterage redistribute resources among family members, and buffer educational inequality (Akresh, 2005). Yet, there should be awareness that many (very) poor people also have (very) poor siblings, and this study’s analyses show that many children are left without extended family support.

This study’s analysis confirmed that the education level of caretakers positively affects the education expenditure. The willingness to allocate resources for the education of their children is also stronger among parents that work outside the agricultural sector. Together these findings support the hypothesis that the household will invest more in education when it expects and is familiar with possible future returns. This is in line with the conclusion of Vu (2012) in Vietnam that households where the household heads have a higher level of education or have professional jobs, enhances the probabilities of educational expenditure.

Not surprisingly, the availability of formal employments in the district pushes the household to invest more in the education of children. If education expenditure is viewed as investment, households take into account the expected future returns. In line with the results of Wiggins and Hazell (2011), who show that unskilled labour like construction, pottering, agriculture and many personal services, provides low returns while skilled labour such as medicine, teaching, accounting and administration gives high-returns activities, this study concludes that the presence of non-agricultural job opportunities fosters educational expenditure.

Conclusion

The Rwanda Vision 2020 targeted a creation of 3.2 million off-farm jobs for the year 2020. While state and donor funds must contribute to this aim, the backbone of the process should be the investments of a growing middle class of Rwandan entrepreneurs. In addition to its success to attract foreign direct investment, the Rwandan Government launched in 2012 a HANGA UMURIMO¹³ Program (HUP) with a purpose to nurture an entrepreneurial culture among Rwandans and foster the emergence and growth of a locally-based business class (GoR, 2012b). More attention is also given to Technical and Vocational Education and Training (TVET) policy geared to provide the economy with qualified and competitive workers needed to achieve the targets formulated in the Rwandan Vision 2020 (GoR, 2008).

McCord (2013) argues that the attempt to reconcile many policy objectives is likely to result in sub-optimal performance, which in all likelihood will not necessarily be effective in terms of the outcomes to be achieved in the long run . This raises the question whether Rwanda should continue to apply social protection policies, or aim exclusively for the growth of non-agricultural employment.

In the light of this study’s findings, transforming the economy and extending employment in non-agriculture sectors will certainly help parents to diversify their strategies to meet their numerous needs, and would also convince them of the future returns of education investments. Yet the price of increased inequality could be high. This study has shown that poverty alleviation instruments which help to reduce the vulnerability of extreme poor households (in establishing subsistence security and protection against financial shocks) should continue as well in order to help provide more equal access to a proper education. Combined with a further reduction of the fertility rate the social protection policies could help poor households to invest more in the education of the children in the short run, hoping that they too will benefit from a further diversification of the national economy in the longer run.

Footnotes

Acknowledgements

Thanks are extended to the anonymous reviewers for their stimulating comments, and to the National Institute of Statistics of Rwanda (NISR) who provided the data.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: We thank the Hewlett Foundation and the Netherlands Organization for Scientific Research for their financial support (grant number: W07 40 202 00).

Notes

Author biographies

Joseph Nkurunziza is Lecturer in the Department of Applied Statistics, University of Rwanda. His research interests are in the field of education economics, poverty alleviation, sustainable development and development planning.

Annelet Broekhuis is senior researcher of International Development Studies, Utrecht University, The Netherlands. Her research focuses on population issues and livelihood studies in Sub-Sahara Africa.

Pieter Hooimeijer is full professor of Human Geography and Demography at Utrecht University, The Netherlands. Contextual effects on educational commitment and achievement is one of his research interests.

References

Ahuja

Jutting

(2004) Are the poor too poor to demand health insurance? Journal of Microfinance /ESR Review 6(1): 1–20.

Akresh

(2005) Adjusting Household Structure: School Enrolment Impacts of Child Fostering in Burkina Faso. IZA Discussion Paper No. 1379. Available at: https://www.econstor.eu/dspace/bitstream/10419/20675/1/dp1379.pdf (accessed 13 October 2013).

Al-Samarrai

Peasgood

(1998) Educational attainments and household characteristics in Tanzania. Economics of Education Review 17(4): 395–417.

Ansoms

(2008) Striving for growth, bypassing the poor? A critical review of Rwanda’s rural sector policies. The Journal of Modern African Studies 46(1): 1–32.

Bhalotra

(2003) Child Labour in Africa. OECD Social, Employment and Migration Working Paper No. 4, OECD, 2003. Available at: http://www.aktiv-gegen-kinderarbeit.de/files/2011/01/Studie_Kinderarbeit_Afrika2003.pdf (accessed 11 June 2013).

Binagwaho

. (2012) Mutual health insurance and its contribution to improving child health in Rwanda. Passauer Diskussionspapiere: Volkswirtschaftliche Reihe, No. V-66–12. Available at: http://www.wiwi.uni-passau.de/fileadmin/dokumente/lehrstuehle/lambsdorff/Papers/Passauer_Diskussionspapiere_Beitrag_V-66–12.pdf (accessed 17 October 2013).

Bucagu

(2013) Tailoring Agroforestry Technologies to the Diversity of Rwandan Smallholder Agriculture. PhD Thesis, Wageningen University, The Netherlands.

Caillods

. (2009) Overcoming the Obstacles to EFA. Paris: UNESCO/IIEP. Available at: http://www.unesco.org/iiep/eng/research/basic/PDF/ADEACaillods_en.pdf (accessed 12 May 2013).

Cho

(2003) Using multilevel analysis in patient and organizational outcomes research. Nursing Research 52(1): 61–65.

10.

Corbridge

Harriss

Jeffrey

(2013) India Today: Economy, Politics and Society. Cambridge, UK: Polity Press.

11.

Devereux

(2011) The Vision 2020 Umurenge Programme: A Pathway to Sustainable Livelihoods for Rural Rwandans? Paper presented at the Annual IPAR Research conference ‘Improving the Lives of Ordinary Rwandans: Evidence for Policy’. Kigali Institute of Education, 8 December 2011. Available at: http://www.ipar-rwanda.org/index.php?option=com_docman&task=doc_download&gid=53&Itemid=151 (Accessed 22 November 2014).

12.

Downey

(1995) When bigger is not better: Family size, parental resources and children’s educational performance. American Sociological Review 60(5): 746–761.

13.

Drèze

Sen

(2013) An Uncertain Glory: India and its Contradictions. London, UK: Penguin Books.

14.

Drummond

Thakoor

(2014) Africa Rising: Harnessing the Demographic Dividend. Washington, DC. IMF Working paper WP/14?143. Available at: https://www.imf.org/external/pubs/ft/wp/2014/wp14143.pdf (accessed 16 April 2014)

15.

Economisti Associati, Center for Economic and Social Research and The Africa Group LLC (2011) Investment Climate in Africa Program, Four-country Impact Assessment: Rwanda Country Report. Available at: https://www.wbginvestmentclimate.org/uploads/Comparative_Report_20110327.pdf (accessed 12 November 2013).

16.

Ellis

Freeman

(2004) Rural livelihoods and poverty reduction strategies in four African countries. Journal of Development Studies 40(4):1–30.

17.

Evans

. (2006) Independent Evaluation of Rwanda’s Poverty Reduction Strategies (PRSP1). Final Report. Available at: http://www.odi.org.uk/sites/odi.org.uk/files/odi-assets/publications-opinion-files/2141.pdf (accessed 13 October 2013).

18.

Glewwe

Jacoby

(2004) Economic growth and the demand for education: Is there a wealth effect? Journal of Development Economics 74(1): 33–51.

19.

Government of Rwanda (GoR) (2000) Rwanda Vision 2020. Ministry of Finance and Economic Planning, Kigali, Rwanda. Available at: http://www.gesci.org/assets/files/Rwanda_Vision_2020.pdf (accessed 15 May 2013).

20.

Government of Rwanda (GoR) (2007) Economic Development and Poverty Reduction Strategies, 2008–2012. IMF Country Report No. 08/90. Available at: http://www.imf.org/external/pubs/ft/scr/2008/cr0890.pdf (accessed 15 May 2013).

21.

Government of Rwanda (GoR) (2008) Technical and Vocational Education and Training (TVET) policy in Rwanda. Ministry of Education, Kigali, Rwanda. Available at: http://www.wda.gov.rw/publications/tvet.pdf (accessed 29 August 2013).

22.

Government of Rwanda (GoR) (2012a) Rwanda Education Statistics. Ministry of Education, Kigali, Rwanda. Available at: http://www.mineduc.gov.rw/IMG/pdf/2011_RWANDA_EDUCATION_STATISTICS.pdf (accessed 27 November 2013).

23.

Government of Rwanda (GoR) (2012b) Hanga Umurimo Program: Start Small, Grow Big. Ministry of Trade and Industry, Kigali, Rwanda. Available at: http://www.minicom.gov.rw/IMG/pdf/HUP_II_CONCEPT_DOCUMENT_last_version_Nov_2012.pdf (accessed 22 April 2014).

24.

Greenspan

(1992) Poverty in the Philippines: The Impact of Family Size. Asia-Pacific Population & Policy, No. 21. Honolulu: Population Institute of the East-West Center.

25.

Gribble

Bremner

(2012) The Challenge of Attaining the Demographic Dividend. Population Reference Bureau. Policy Brief September 2012. Available at: http://www.prb.org/pdf12/demographic-dividend.pdf (accessed 16 April 2015)

26.

Handa

Simler

Harrower

(2004) Human Capital, Household Welfare, and Children’s Schooling in Mozambique. Research Report 134. International Food Policy Research Institute. Available at: http://www.ifpri.org/sites/default/files/pubs/pubs/abstract/134/rr134.pdf (accessed 18 June 2013).

27.

Hayman

(2007) Are the MDGs enough? Donor perspectives and recipient visions of education and poverty reduction in Rwanda. International Journal of Educational Development 27(4): 371–382.

28.

Hox

(2010) Multilevel Analysis: Techniques and Applications. 2nd ed. New York, NY: Routledge.

29.

Kindness

(2011) Vision 2020 Umurenge Programme. Annual Report 2009/10. Kigali, Rwanda. Available at: http://www.unicef.org/rwanda/RWA_resources_vision2020umurenge.pdf (accessed 16 May 2013).

30.

King

Mason

(2001) Engendering Development Through Gender Equality in Rights, Resources and Voice. World Bank Policy Research Paper. Oxford University Press. Available at: http://www-wds.worldbank.org/external/default/WDSContentServer/WDSP/IB/2001/02/02/000094946_01012505311522/Rendered/PDF/multi_page.pdf (accessed 20 June 2013).

31.

Kwizera

. (2011) Girls Outnumber Boys in PLE Exams. NewTimes, 26 October 2011. Available at: http://www.newtimes.co.rw/index.php?issue=14790&article=46570 (accessed 15 March 2012).

32.

KPMG (2012) Cutting Through Complexity. Rwanda Country Profile. Available at: http://www.kpmg.com/Africa/en/KPMG-in-Africa/Documents/RWanda.pdf (accessed 12 November 2013).

33.

Kreft

De Leeuw

(1998) Introducing Multilevel Modeling. London, UK: SAGE.

34.

Lee

(2004) Sibling Size and Investment in Children’s Education: An Asian Instrument. IZA Discussion Paper Series No. 1323. Available at: http://hdl.handle.net/10419/20591 (accessed 16 May 2013).

35.

Lloyd

Blanc

(1996) Children’s schooling in Sub-Saharan Africa: The role of fathers, mothers, and others. Population and Development Review 22(2): 265–298.

36.

. (2012) Towards universal health coverage: An evaluation of Rwanda Mutuelles in its first eight years. PLoS ONE 7(6): 1–16.

37.

Madaki

Adefila

(2014) Contributions of rural non-farm economic activities to household income in Lire area, Kaduna State of Nigeria. International Journal of Asian Social Science 4(5): 654–663.

38.

Malmberg

(2008) Demography and Development Potential of Sub-Saharan Africa. The Nordic African Institute. Current African Issues No. 38. Available at: http://www.diva-portal.org/smash/get/diva2:241196/FULLTEXT01.pdf (accessed 20 February 2015).

39.

McCord

(2013) Public Works and Resilient Food Systems. ODI, On behalf of Federal Ministry of Economic Cooperation and Development (BMZ), Deutsche Gesellscaft fur Internationale Zusammenarbeit (GIZ) GmbH.

40.

McIntyre

. (2006) What are the economic consequences for households of illness and of paying for health care in low-and middle-income country contexts? Social Science & Medicine 62(4): 858–865.

41.

Megumi

(2010) Household expenditures on children, 2007–08. Monthly Labor Review 133(9): 1–14.

42.

MINEDUC (2014) Educational Statistical Yearbook 2013. Kigali, Rwanda: Ministry of Education.

43.

Muhoza

(2014) Excess Fertility and Family Planning in Rwanda. PhD Thesis, Utrecht University. Available at: http://dspace.library.uu.nl/bitstream/handle/1874/302985/Muhoza.pdf?sequence=1 (accessed 18 May 2015).

44.

Muthén

Satorra

(1995) Complex sample data in structural equation modelling. Sociological Methodology 25: 267–316.

45.

Nabalamba

Sennoga

(2014) Analysis of Gender and Youth Employment in Rwanda. African Development Bank, pp 1–35. Available at: http://www.afdb.org/fileadmin/uploads/afdb/Documents/Publications/Rwanda_-_Analysis_of_Gender_and_Youth_Employment.pdf (accessed 16 May 2015).

46.

NISR (2012a) The Evolution of Poverty in Rwanda from 2000 to 2011: Results from the Household Surveysx. Kigali, Rwanda. Available at: http://www.statistics.gov.rw/publications/evolution-poverty-rwanda-2000-2011-results-household-surveys-eicv (accessed 15 May 2013).

47.

NISR (2012b) EICV3 Thematic Report: Economic Activity. Kigali, Rwanda. Available at: http://statistics.gov.rw/publications/eicv-3-thematic-report-economic-activity (accessed 27 November 2013).

48.

Nkurunziza

Broekhuis

Hooimeijer

(2012) Free education in Rwanda: Just one step towards reducing gender and sibling inequalities. Education Research International 2012: 1–11.

49.

Nkurunziza

Broekhuis

Hooimeijer

(2014) Sibling and gender effects on youngsters’ second chance to complete primary education in Rwanda. (Unpublished paper).

50.

OECD (2011) How Does Education Affect Employment Rates? In Education at a Glance 2011: Highlights. Paris: OECD. Available at: https://dx-doi-org.web.bisu.edu.cn/10.1787/eag_highlights-2011-en

51.

Paxton

Mutesi

(2012) School Funding and Equity in Rwanda. Final Report. Rwanda: Institute for Policy Analysis and Research. New York, NY: UNDP. Available at: http://rwandapedia.rw/cmis/browser?id=workspace%3A//SpacesStore/4917e151-e06c-4e5f-a0e3-09823767917c (accessed 11 July 2015).

52.

Peugh

(2010) A practical guide to multilevel modelling. Journal of School Psychology 48(1): 85–112.

53.

Psacharopoulos

Patrinos

(2004) Returns to investment in education: A further update. Education Economics 12(2): 111–145.

54.

Pushkar

(2003) Schooling and educational attainment: Evidence from Bangladesh. Education Economics 11(2): 129–153.

55.

Ruberangeyo

Ayebare

De Lamine de Bex

(2011) Rwanda Social Protection: An Ongoing Process. Excerpt from Sharing Innovative Experiences, Volume 18: Successful Social Protection Floor Experiencies, ILO - SU/SSC (UNDP) - National experts, 2011, pp. 333-360. Volume 18: Successful Social Protection Floor Experiences. Available at: http://tcdc2.undp.org/GSSDAcademy/SIE/Docs/Vol18/SIE_v18_ch15.pdf

56.

Rutareka

(2011) How a Simple Cow Gives Hope to Many Families? The Rwanda Focus, May 2011. Available at: https://karinganire.wordpress.com/2014/04/15/girinka-how-a-simple-cow-gives-hope-to-many-rwandan-families/

57.

Saksena

. (2011) Mutual health insurance in Rwanda: Evidence on access to care and financial risk protection. Health Policy 99(3): 203–209.

58.

Sebates-Wheeler

Roelen

(2011) Transformative social protection programming for children and their careers: A gender perspective. Gender & Development 19(2): 179–194.

59.

Siaens

Subbarao

Wodon

(2003) Are Orphans Especially Vulnerable? Evidence from Rwanda. Washington, World Bank. Available at: http://www.crin.org/docs/Are%20Orphans%20Especially%20Vulnerable%20-%20Evidence%20From%20Rwanda.pdf (accessed 15 May 2013).

60.

Siegel

Gatsinzi

Kettlewel

(2011) Adaptive social protection in Rwanda: ‘Climate-proofing’ the vision 2020 Umurenge programme. IDS Bulletin 42(6): 71–78.

61.

Sweetman

(2011) Introduction. Gender& Development 19(2): 169–177.

62.

Tilak

JBG

(2002) Determinants of Household Expenditures on Education in Rural India. Working Paper Series No. 88. New Delhi, India. Available at: http://www.aserf.org.in/presentations/WP88.pdf (accessed 20 June 2013).

63.

Twahirwa

(2008) Sharing the burden of sickness: Mutual health insurance in Rwanda. Bulletin of the World Health Organization 86(11): 823–824.

64.

Ulusoy

Yolcu

(2013) Household education expenditure of families at primary education level. Journal of Education Sciences Research 3(1): 1–27.

65.

UNESCO (2012a) Rwanda EFA Profile. Available at: http://www.unesco.org/new/fileadmin/MULTIMEDIA/FIELD/Dakar/pdf/EFA%20country%20profile%20%202102-%20Rwanda.pdf (accessed 20 June 2013).

66.

UNESCO (2012b) Youth and Skills: Putting Education to Work. Education for all Global Monitoring Report 2012. Paris, France. Available at: http://unesdoc.unesco.org/images/0021/002180/218003e.pdf (accessed 15 May 2015).

67.

Van der Berg

(2008) Poverty and Education. Education Policy Series 10. The International Institute for Educational Planning, Paris, France and The International Academy for Education, Brussels, Belgium. Available at: http://www.iiep.unesco.org/fileadmin/user_upload/Info_Services_Publications/pdf/2009/EdPol10.pdf (accessed 15 May 2013).

68.

H Q

(2012) Determinants of educational expenditure in Vietnam. International Journal of Applied Economics 9(1): 59–72.

69.

Wang

Zhang

Hsiao

(2006) Ill health and its potential influence on household consumptions in rural China. Health Policy 78(2): 167–177.

70.

Wiggins

Hazell

(2011) Access to Rural Non-farm Employment and Enterprise Development. Background Paper for IFAD Rural Poverty Report. Available at: http://www.ifad.org/rpr2011/background/5.pdf (Accessed 28 November 2014).

71.

World Bank (2011) Rwanda: Education Country Status Report. Toward Quality Enhancement and Achievement of Universal Nine Year Basic Education, an Education System in Transition; a Nation in Transition. Washington, World Bank. Available at: http://www.poledakar.org/pdf/Resen%20Rwanda.pdf ttp://www.poledakar.org/pdf/Resen%20Rwanda.pdf (accessed 21 June 2013).

72.

Zuckerman

(2001) Why Engendering PRSP, Reduces Poverty and the Case of Rwanda. WIDER Discussion Paper//World Institute for Development Economics (UNU-WIDER) No. 2001/112, ISBN 9291900478. Available at: https://www.econstor.eu/dspace/bitstream/10419/52912/1/336656432.pdf (accessed 26 June 2013).