02.06.2014 Views

Evaluation of the Australian Wage Subsidy Special Youth ...

Evaluation of the Australian Wage Subsidy Special Youth ...

Evaluation of the Australian Wage Subsidy Special Youth ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

247<br />

maintained, <strong>the</strong>n <strong>the</strong> Heckman bivariate probit modelling results give a modelling result<br />

in harmony with <strong>the</strong> role <strong>of</strong> unobservables in SYETP participation. But while this may be<br />

true, <strong>the</strong> form which <strong>the</strong> unobservables take is also highly restricted in <strong>the</strong> Heckman<br />

bivariate probit modelling, and <strong>the</strong> reasonableness <strong>of</strong> this must in turn be considered.<br />

Rosenbaum and Rubin (1985b) p 109 also note that <strong>the</strong> assumption <strong>of</strong> strongly ignorable<br />

treatment (CIA) needs to be consistent with <strong>the</strong> data and <strong>the</strong> causal mechanism, through<br />

which <strong>the</strong> SYETP treatment is thought to operate to produce employment effects, and<br />

without which <strong>the</strong> matching results are biased. CEP referrals is questioned as potentially<br />

violating <strong>the</strong> CIA requirements, because <strong>of</strong> <strong>the</strong> use <strong>of</strong> this variable for <strong>the</strong> exclusion in<br />

Richardson (1998). 166 It was argued earlier that CEP referrals credibly did not affect<br />

employment independently once SYETP participation was accounted for. The effects <strong>of</strong><br />

<strong>the</strong> CEP referrals for <strong>the</strong> replicated Heckman modelling in <strong>the</strong> same data have been<br />

shown to be strongly diminished by accounting for attrition with weighting. The<br />

<strong>the</strong>oretical view against <strong>the</strong> inclusion <strong>of</strong> CEP referrals in <strong>the</strong> weighted model for <strong>the</strong><br />

propensity can however also be argued on <strong>the</strong> basis <strong>of</strong> <strong>the</strong> evidence <strong>of</strong> Wielgosz (1984)<br />

and Aungles and Stewart (1986), discussed in section 2.2.6. If this conviction is upheld,<br />

<strong>the</strong>n <strong>the</strong> sensitivity results where CEP referrals are excluded, or <strong>the</strong> reduced specification,<br />

provide matching results consistent with this view.<br />

In conclusion, <strong>the</strong> sensitivity analysis shows that <strong>the</strong> employment effects found are<br />

subject to <strong>the</strong> specifications and modelling assumptions adopted. This was true for both<br />

<strong>the</strong> Heckman and PSM modelling approaches. In an overview <strong>of</strong> <strong>the</strong> various results<br />

gained which disregards <strong>the</strong> magnitude <strong>of</strong> point estimates, it can be determined that no<br />

negative employment effects were found for SYETP, even in sensitivity analysis.<br />

Friedlander et al. (1997) p1819 note that in debates about non-experimental evaluations<br />

<strong>the</strong>re is no clear-cut way <strong>of</strong> determining which modelling assumption is valid. Within <strong>the</strong><br />

range <strong>of</strong> alternatives explored, and as can be determined within <strong>the</strong> limits <strong>of</strong> <strong>the</strong> ALS data,<br />

<strong>the</strong> direction <strong>of</strong> employment effect is robustly positive. Statistical significance in this<br />

166 This was checked by modelling <strong>the</strong> univariate probit for employment, where <strong>the</strong>se variables were not<br />

found significant.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!