4. WHEN DOES EDUCATION AID WORK?
4.2. MULTIVARIATE ANALYSIS
4.2.1. The Direct Effect of Per Capita Education Aid Commitments on Primary Net
rather than the other way round.
In order to address the potential for endogeneity and simultaneously to correct for the bias associated with the introduction of the lagged dependent variable in the dynamic panel model context, a system Generalised Method of Moments regression estimator is
employed. The estimator has several advantages that favour its use over alternatives such as the Ordinary Least Squares estimator. By design, the system GMM estimator presumes that the independent variables are endogenous and therefore employs lags of each of these variables to act as their own instrument. The possibility of including lagged explanatory variables as their own instrument offers the benefit of a strong association with the initial variable; nevertheless it becomes difficult to make the case that this is not correlated with the error term. The issue usually presents itself in situations where endogeneity is the product of reverse causation, with the outcome variable wielding influence over the regressor concerned.
In the association between education aid and education outcomes if the current level of education aid is affected by the current level of primary enrolment, the lag of education aid will likewise be affected by the lag of primary enrolment and using the lag of the education aid variable may be insufficient to address the reverse causality problem. In order to adequately address the endogeneity issue and potential for reverse causation, the final multivariate model regressions were estimated with Energy Aid as an instrumental variable as carried out in the bivariate analysis, described in section 4.1 of this chapter.
4.2.1. The Direct Effect of Per Capita Education Aid Commitments on Primary Net Enrolment
Table 9 displays the results for Model 1 in which an enrolment equation is modelled to measure the effect of per capita education aid commitments on the primary net enrolment
Table 9: The Effect of Per Capita Education Aid Commitments on Primary Enrolment 1970-2013
61 Developing Countries (Model 1)
NER
L1. ***0.308 (0.000)
Period (Reference: 1970-89) *-2.503 (0.068)
Education Aid Commitment ***0.348 (0.007)
Domestic Education Expenditure (per capita) 0.015 (0.147)
Pupil-Teacher Ratio -0.127 (0.350)
Youth Population *-0.528 (0.079)
Extent of Urbanisation ***0.318 (0.000)
Per Capita GDP (adjusted for PPP) **0.001 (0.027)
Education Aid Squared **-0.000 (0.021)
Education Aid*Government Stability ***0.007 (0.004)
Education Aid*Democratically Partly Free (Ref: Not Free) -0.064 (0.583)
Education Aid*Democratically Free (Ref: Not Free) -0.026 (0.813)
Education Aid*Conflict (Ref: No Conflict) *-0.310 (0.075)
Education Aid*Conflict at any Time (Ref: No Conflict) -0.190 (0.281)
Sources: OECD CRS (2015); World Bank (2015a); The PRS Group (2015); Freedom House (2015); UCDP/PRIO (2015); and Gwartney, Hall et al. (2015)
P-value in parentheses * significant at 10% ** significant at 5% *** significant at 1%
It is evident that the lag of the dependent variable - enrolment - holds considerable explanatory power [0.308 (0.000)]. Inclusion of a lag of the dependent variable in the right hand side of the explanatory equation has the advantage of accounting for the possible persistence in outcomes, with the initial level of education predicting future rates of
enrolment. It is theoretically appealing as the level of enrolment in primary education in the previous period may be assumed to be highly relevant to primary enrolment in the
subsequent period which appears to be the case as indicated by the positive coefficient result. Of course, although NER in year n-1 will always have a correlation with NER in year 1, it should not be assumed that this is a causal driver of enrolment rates in the long term.
A dummy variable ‘Period’ is included in order to allow comparison of enrolment between the period 1970-1989 and 1990-2013 - offering insight into enrolment outcomes during and following the Cold War. This is of interest because the post-Cold War period (here denoted as 1990-2013) witnessed increasing global consensus towards the objective of achieving universal primary education following the 1990 World Conference on Education for All that was later fully realised with the signing of the Millennium Declaration in 2000, and has also seen dramatic expansions in primary education enrolment.There has been a
particular emphasis on the global community coming together to work toward the attainment of universal primary education. The period variable demonstrates significant power in predicting the outcome of the dependent variable [-2.503 (0.068)]. The period 1990-2013 is correlated with a 2.5 per cent greater rise in primary net enrolment as
compared to the period 1970-89. That this period should be such a significant predictor of primary enrolment when controlling for all other covariates indicates that the importance of these global goals runs beyond raising additional resources for education - it suggests a ‘policy effect’. The findings indicate that, as the world order shifted from policy driven by political ideology - a politically competitive process with no universal ground in which developing countries sided with either Communist or Capitalist orthodoxy - toward a development-led consensus that culminated in an unprecedented global effort to align education policy in the pursuit of universal primary education, enrolment in education improved dramatically. This finding is a strong advocate for global education goals. Global goals may be effective in themselves as local, national and international policies - and actors at each level - converge and the intensified collaborative effort to create an environment conducive to improving enrolment in primary education does just that.
Education aid commitments expressed per capita have a positive and statistically significant effect on enrolment [0.348 (0.007)] - with a US$ 1 increase in average per capita education aid increasing school enrolment by around 0.3 per cent. A non-linear specification of education aid (Aid Squared) is employed to depict the potentially decreasing returns to aid investment in education. There is evidence of a diminishing effect of aid on primary enrolment with the negative coefficient for commitment per capita squared working in the expected direction, but the magnitude of this effect is negligible.