Education, Employment and Earnings of Secondary School-Leavers ...

5.3 The Estimated School EffectsOne motivation for the inclusion of the school fixed effects is to control, inter alia, forschooling quality in terms of its potential effect on education returns and on earningsmore generally. Their inclusion is also motivated by a sense that they might alsocapture labour market network effects. The inclusion of the school controls tends toreduce both the return to senior secondary schooling and to a university qualificationbut induces a sharper reduction in the former relative to the latter. 32 This is a relativelyintuitive result and highlights the importance of controlling for schooling quality incomputing rates of return to schooling qualifications. It is apparent that parentaleducational background is more important to the estimated university returns thansecondary schooling quality.The school effects can be interrogated more thoroughly by normalizing the tenestimated school effects as a deviation from an overall school weighted averageeffect. This transformation has intuitive appeal in that the estimated earningsdifferences are now relative to an overall average rather than an arbitrary base groupand are thus more easily interpretable. This approach follows the approach adoptedby Krueger and Summers (1988) in the literature on inter-industry wage differentials.If we define the school effect from the interval regression model for the k th school asγ k the deviation is expressed as:10γ k – j=1π j γ j [8]where π j is the proportion of the sample in the j th school. The corresponding samplingvariances are computed as per Zanchi (1998). 33 Table 6 reports the estimateddeviations for the ten schools for both wage employees and the self-employed.[Table 6]32 In Table 4 the estimated coefficient for the senior secondary qualification reduced by one-quarterwith the inclusion of school fixed effects, while the comparable estimate for the university qualificationreduces by just under one-tenth.33 See also Haisken-DeNew and Schmidt (1997) for a more detailed discussion.24