• No se han encontrado resultados

CBV (TM/HA) = BVT*Q.45 Dónde:

2.5.3 FLUJOGRAMA DE ACTIVIDADES

Other studies exploring the link between aid and education outcomes have selected countries into their research on the basis of income - accounting for low-, lower-middle- and upper-middle-income countries (Michaelowa and Weber 2007b; Dreher, Nunnenkamp et al. 2008; Christensen, Homer et al. 2011; Birchler and Michaelowa 2015). As stated above upper-middle income countries are unlikely to receive much aid for primary enrolment growth and whilst GNI per capita is a popular means of country selection in econometric analyses of aid, it is certainly not the only criteria used for allocating aid.

Moreover, well-recognised associations of aid with richer rather than poorer developing countries may confound attempts to establish the independent effect of aid on educational participation. The approach adopted here that recognises both a country’s ‘need’ for aid as well as its ability to allocate funds domestically to the education sector, is proposed by the researcher as a more accurate reflection of education aid in practice.

3.4.3. Model Specification

The conceptual framework guiding the model specification supposes that additional investment in education spending, in the form of aid to education, will result in an improved impact upon education outcomes. Additional resources to finance education resulting in the building of schools, the hiring and training of teachers, the provision of free textbooks and other supplies for pupils, and so forth, ought to improve both the quantity and quality of education. Indeed, anecdotal evidence from a number of countries indicates that education aid results in reduced levels of absenteeism in addition to improved

enrolment and retention rates (Asiedu and Nandwa 2007). The positive association between education aid and enrolment rates is also consistent with some early empirical findings (Michaelowa and Weber 2006; Dreher, Nunnenkamp et al. 2008), although Christensen, Homer et al. (2010) find there to be no effect of education aid on enrolment and only when education aid is decomposed by donor type – with aid from bilateral donors shown to have an effect (2011).

The research questions are addressed by looking beyond aggregate assessments of the effectiveness of education aid found in these few existing empirical studies on this topic, to explore whether aid might be more effective in particular development settings than others - contexts of good political and economic governance as well as in countries experiencing or emerging from conflict. It assumes that the effect of aid is not homogenous and will be influenced by these factors.

Dependent Variable

As a starting point for the model employed here it is stipulated that the allocation of aid to education is meant to support the local education system by providing goods and services not efficiently delivered by the existing education system. Enrolment rates are an effective method for assessing the delivery of services by a local education system, as low enrolment rates suggest that students and families either do not have the opportunity to enrol in primary school, or are persuaded not to do so by the poor quality of education that results from an inadequate supply of school buildings, teachers, supplies etc. (Lewin 2007).

Presumably, aid to education should address both of these problems. Education aid programmes are generally intended to reduce the cost of education for the end user by providing financial support to the national education system; through the building of new infrastructure; and by providing improved teacher training, as well as better curricula and learning materials. If aid improves access to, and the quality of, education in recipient countries, then enrolment rates would be expected to increase as a result of these programmes.

It could be argued that it would be preferable for a study of education aid effectiveness to focus on the quality of schooling rather than on schooling ‘quantity’: on these grounds, Hanushek and Wößmann (2008) question whether EFA and the MDGs may be misguided because of their focus on the quantity of education as opposed to its quality. However,

measure. Breton (2011) finds this to be the case, presenting empirical evidence finding the quantity of schooling statistically superior to measures of education quality when explaining differences in GDP per capita across countries. A further measure of educational

outcomes, gender parity, is included in some of the regressions for primary education run on the short-term annual panel as this is an area in which it would be expected that donors concentrate their aid for primary education given the importance that the MDGs place upon the education of girls.

In order to measure the effect that aid has upon education outcomes, a structural enrolment equation is estimated that includes aid for education and other explanatory variables specifying the education system. The selection of control variables is based upon the literature relating to education outcomes. Schultz (1988) models country education systems using a production-demand framework, an approach followed by Roberts (2003) and Baldacci, Guin-Siu et al. (2003), among others. The concept of a social production function has been used in the literature exploring the relationship between government expenditure and social outcomes. The concept is considered well suited to measuring the relation between education outcomes and measurable educational inputs (Hanushek 1995;

Birchler and Michaelowa 2015).

Explanatory Variables

Roberts (2003) discusses the expected predictors of education enrolment in terms of supply and demand factors. The supply factors most commonly cited in the education production-demand framework are domestic spending on education and the pupil-teacher ratio. Education aid is included as an additional supply-side factor as it is assumed that additional resources allocated for the purpose of improving the provision of education ought to result in improved educational outcomes.

The main explanatory variable of interest to this is study is education aid. It is assumed that increased aid spending ought to increase participation in education by reducing the cost, and increasing the quality, of education through the provision of additional financing in support of the national education system. The direction of the relationship between education aid and education outcomes might not be altogether clear, however, as donors are likely to allocate aid to countries demonstrating greater need (i.e. with lower rates of enrolment in education) meaning that the possibility of an endogenous relationship between aid and enrolment needs to be addressed, an issue that is discussed later in this

section. Aid may not necessarily have an impact upon education outcomes in the short term and its impact might, therefore, only be perceived in the medium to long-term - prompting the inclusion of a lagged aid variable. A lag of aid over the previous period is included in models run on the short-term annual panel, to explain the fact that the impact of education disbursements may not be automatic. A non-linear specification of aid (aid squared) is also accounted for in order to depict the potentially decreasing returns to aid investment and to assess individual countries’ capacity to absorb additional amounts of aid.

This is a common feature in the literature on aid and growth (see, for example, Dalgaard, Hansen, and Tarp 2004).

It seems logical that domestic expenditure on education likewise ought to boost enrolment by affording greater access to education. Increased public spending on basic education, it is argued, should allow not only a greater proportion of the population to complete primary and secondary education, but also improve each student’s preparation and ability to complete tertiary education (Bergh and Fink 2006). However, Dreher, Nunnenkamp et al.

(2008) conclude from their findings that domestic spending on education has virtually no effect on education outcomes. Rajkumar and Swaroop (2008), on the other hand, who specifically analyse the effects of government spending on health and education outcomes, deduce that public spending improves education outcomes in well-governed nations, but has no impact in poorly-governed countries. In spite of the inconclusive findings in the econometric literature on this topic, the inclusion of public education expenditure is a standard feature of education production functions due to it being the most significant source of funding for national education systems (Lewin 2012), and is thus accounted for in the present study.

Student-teacher ratios are a much-discussed aspect of education; with academics querying why in developed countries reduced class size has been shown to increase education outcomes (Krueger and Whitmore 2002), while the effect of pupil-teacher ratio is

consistently found to be insignificant in low-income countries (Banerjee, Cole et al. 2007;

Duflo, Dupas et al. 2007). Duflo, Dupas et al. (2007) suppose that the insignificant effect of smaller class sizes in poorer countries might be due to weak governance reducing the impact of additional education expenditure. In contrast to much of the rest of the literature on education outcomes which proposes there to be no correlation between class size and education outcomes, Michaelowa and Weber (2006) find high pupil-teacher-ratio to exert a

suggest that parents’ perception is clearly that crowded classrooms are problematic and that demand is affected by such perceptions independently of whether these perceptions are justified or not.

On the demand side, factors regularly cited in the literature pertaining to enrolment are: per capita income, percentage of the population that is ‘youth’, the extent of urbanisation, and adult literacy (a proxy for parental level of education) (see, for example, Huisman and Smits 2009; Kazeem, Jensen et al. 2010).

The per capita income indicator is often used as a proxy for household poverty and reflects demand for schooling (Mingat and Tan 1998; Gupta, Verhoeven et al. 1999; Baldacci, Clements et al. 2004). The Education Policy and Data Center (2008) find, across four studies of education growth, that inequality in enrolment is the product of disparity in pupil income; in almost all cases, the poorer the pupils, the smaller the enrolment rates.

The size of the school population is deemed to be reflective of the relative demand for education. This variable is included with the purpose of holding constant the degree of strain that the composition of the national population places on the education system.

Countries with a greater percentage of the population aged less than 15 have the potential for more students to be enrolled in education and a smaller percentage of adults to provide and pay for schooling. Gupta, Verhoeven et al. (1999) report that the share of the

population under 15 exerts a strong influence on enrolment. Michaelowa and Weber (2006) also find that a relatively high share of youth significantly increases the difficulties in

reaching high completion rates.

The extent of urbanisation is also supposed to effect enrolment rates, although the evidence for this is mixed with Dreher, Nunnenkamp et al. (2008) finding the variable to be insignificant, whilst Fafchamps and Wahba (2006) find that in the case of Nepal, children living in urban areas are more likely to attend school than those living in rural areas.

Adult literacy is invariably included in enrolment equations to account for the large effect that parental education is likely to have on education enrolment rates. Though previous literature uses this data as a key control variable (see Dreher, Nunnenkamp et al. 2008) it is excluded from the present study on the grounds of insufficient data and the high potential for collinearity between literacy and enrolment that could affect other coefficients in the

model. Inclusion of a variable to denote adult literacy was considered problematic due to the large number of missing variables. The measure employed by Dreher, Nunnenkamp et al. (2008), held values from 1975 onward only with the majority (64 per cent) of the countries having just 1 to 4 values over the time span. 64 out of the 234 countries for which the WDI publishes data hold no information on adult literacy rates for the entire time period. On this basis, dropping the variable from the model was deemed preferable.

An additional explanatory variable is included in the modelling of education outcomes. A dummy variable - Period - is included to allow for assessments of whether enrolment has been greater in particular periods as compared to others. Inclusion of this variable is, to the author’s knowledge, unique to this study.

The basic equation employed in the modelling of the education aid/enrolment relationship takes the following form:

Equation 1: The Relationship Between Education Aid and Primary Enrolment

signifies enrolment at primary level in country in year ; and is education aid expressed per capita. X is the vector of control variables, denotes country fixed effects, and signifies the disturbance term.

Accounting for Endogenous Relationships

A serious problem with this basic regression model is that some explanatory variables may not be exogenous. Aid is not randomly assigned, with indicators of need having been shown to be related to aid allocations (McKinlay and Little 1977; Thiele, Nunnenkamp et al. 2007). In the case of education, it is plausible that donors make decisions about the allocation of education aid on the basis of prevailing enrolment rates in recipient countries.

It may be supposed that if enrolment rates are high, the recipient’s education sector is in less need of external educational assistance. In such a scenario, the effect of aid for

education on the enrolment rate would be offset by the effect of the enrolment rate on aid and, as such, the endogenous aid-enrolment relationship would lead to an underestimation of education aid’s true impact.



schooli,t 1schooli,t12aidi,t1BX1i,t

Schooli,t i t



aidi,t



1



i,t

It may also be expected that the relationship between domestic education expenditure and the pupil-teacher ratio with enrolment is also endogenous. Higher primary NERs are the result of more primary-aged children enrolling in primary school, lowering the amount of spending per student and increasing the number of students per teacher. In this instance the causal effect runs from the primary NER toward domestic expenditure and the pupil-teacher ratio as opposed to the other way round.

This potential for endogenous relationships prompts the use of instrumentation.By

construct, the system GMM dynamic panel model used to estimate the effects of education aid assumes the explanatory variables to be endogenous by using lags of each variable as their own instrument. On the whole, the option to use a lagged explanatory variable as its own instrument has the benefit of a strong correlation with the initial variable. However, as Michaelowa and Weber (2006) argue, there are certain instances in which it may be difficult to maintain that the instrument is strictly uncorrelated with the error term (a key

requirement for a valid instrument). This is particularly the case when endogeneity is attributable to reverse causation - where the dependent variable exerts influence on an explanatory variable. As it is not inconceivable to suppose that education aid donors intentionally allocate aid for education on the basis of educational need (as determined by low primary enrolment and completion rates), or that enrolment rates influence the amount of resources available for spending on education and the pupil-teacher ratio, using a lag of the explanatory variable may not be sufficient for addressing these relationships.

Clearly, if current educational aid is affected by current educational outcomes, lagged educational aid will be affected by lagged educational outcomes. It is therefore necessary to include an instrumental variable to address the endogeneity issue.

Addressing endogeneity – in this case caused by simultaneity (interdependence) between the education aid and primary net enrolment variables – is a critical aspect of measuring aid effectiveness. Instrumental variable estimation is used to address the endogeneity issue here. This requires that a variable is found that is correlated with the problem variable but which does not suffer from endogeneity – an instrumental variable (IV) that is correlated with education aid, but not with the error term. Michaelowa and Weber (2007a) show success with Energy Aid – which captures all assistance allocated to the production of energy, energy sector policy planning, institution building and distribution management (OECD DAC 2015a). Energy aid is both truly exogenous and found to be correlated with education aid, with the relationship between the two variables significant at the 1 per cent level, indicating that it is an appropriate choice as an instrumental variable for the modelling of

aid on education enrolments and addressing the potential for endogeneity.

Introducing Interactions to Explore the Differential Impact of Education Aid The heterogeneity of countries is likely to be a significant factor in the effectiveness of education aid - differing political, institutional and economic forces will inevitably impinge upon the absorption and application of aid and its outcomes in the education sector across developing countries. By including interactions, the model tests the extent to which factors related to the quality of governance and presence of conflict, work through aid with the intention of revealing the differential impact of aid for education.

Good Governance

At the turn of the millennium, World Bank research conducted by Collier and Dollar (1999) initiated serious discussion about aid effectiveness and its implications for aid allocations. The key point of debate was over the importance that good policy plays in determining the degree of aid effectiveness. The authors of the World Bank report argued that aid works best when government policies are good, and that pursuing a more selective allocation of aid to poor countries demonstrating sound policies would lead to larger reductions in poverty. In line with this argument, many bilateral and multilateral donors have reassessed their patterns of aid allocation over the course of the past decade, with a particular emphasis on making aid more performance-based (Benyon 2003; Colenso 2011).

Whether and to what extent the impact of foreign aid depends on the quality of policy and institutions has been heatedly debated and remains unresolved in the macroeconomic literature. By including interactions to show how the effect of various governance

indicators work through education aid, the question of whether aid for education is more conducive to an improvement in education outcomes when recipient countries are well governed can be addressed. Three variables - government stability, the extent of

democratic freedom, and economic openness - are used in interaction with education aid to explore this issue.

Although measures of government stability have been widely used in the literature on corruption and governance (see La Porta, Lopez-de-Silane et al. 1997; Armah 2010), its use has not been as widespread in the aid-growth literature. The popularity of the measure has increased, however, as Knack (2001) and Braütigam and Knack (2004) both employ data

Minoiua and Reddyb (2010) employ the measure in aid-growth regressions. These authors report the measure of government stability published by The PRS Group (2015), as

adopted in the present research, to provide meaningful and intuitive findings. Government stability, as defined by the ability of a government to stay in power and carry out its

programmes, is a sensible measure of governance for this research as it allows for analysis of how the effects of governance work through aid. This helps to test whether arguments found in the aid-growth literature that support the allocation of aid to countries

demonstrating ‘good’ governance, are relevant to aid for education.

That the effectiveness of education aid in determining enrolment might be dependent upon the degree of democratic freedom in recipient countries is also accounted for. It has been

That the effectiveness of education aid in determining enrolment might be dependent upon the degree of democratic freedom in recipient countries is also accounted for. It has been

Documento similar