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

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

84<br />

Table 2.19 Baker (1984) Estimated probabilities <strong>of</strong> labour market outcomes for<br />

participants from <strong>the</strong> model <strong>of</strong> employment<br />

Continuous fulltime Non-Continuous Out <strong>of</strong> work<br />

work<br />

fulltime work<br />

SYETP<br />

Commonwealth<br />

.414 .400 .186<br />

SYETP private .517 .358 .124<br />

2nd SYETP .418 .399 .183<br />

Extended SYETP .611 .307 .082<br />

Source: Baker (1984) p48 Table A10 May 1982 post-programme Survey <strong>of</strong> participants. Predicted<br />

probabilities were estimated for <strong>the</strong> eligibility period <strong>of</strong> 17 weeks unemployment in <strong>the</strong> past 52 weeks, for a<br />

participant who left school in year 10/11, did post-school training, took part in no o<strong>the</strong>r programme and<br />

completed <strong>the</strong> programme <strong>of</strong> participation.<br />

A chief difficulty with <strong>the</strong> results <strong>of</strong> modelling is <strong>the</strong> likely endogeneity <strong>of</strong> <strong>the</strong><br />

programmes. As <strong>the</strong> authors also point out, <strong>the</strong>y did no selection modelling for<br />

programme entry. They did however try to account for <strong>the</strong> survey response problem,<br />

which could introduce bias and inconsistency to <strong>the</strong> estimates if unaccounted for. The<br />

means <strong>of</strong> implementing <strong>the</strong> Heckman selection adjustment was to enter <strong>the</strong> selection<br />

correction term linearly into <strong>the</strong> ordered probit model, but this is problematic due to nonlinearity<br />

<strong>of</strong> <strong>the</strong> probit functional form. An appropriate model combining response and<br />

employment would have estimated <strong>the</strong> equations jointly and allowed for correlated errors.<br />

A fur<strong>the</strong>r difficulty with <strong>the</strong>ir nonresponse modelling is that it combines item nonresponse<br />

and survey non-response, which requires <strong>the</strong> assumption that <strong>the</strong> same model<br />

can account properly for <strong>the</strong>m both. There is no non-participant comparison group for<br />

programme participants, only participants <strong>of</strong> o<strong>the</strong>r programmes are available for<br />

comparison. This is because <strong>the</strong> data are from a survey <strong>of</strong> programme participants only.<br />

2.3.3 Rao and Jones (1986)<br />

Rao and Jones (1986) used <strong>the</strong> same survey data as used for Baker (1984), but with<br />

additional data from a second follow-up survey conducted in May 1983. The additional<br />

survey allowed employment outcomes at 18-20 months to be examined, and <strong>the</strong>y<br />

combined <strong>the</strong> survey data with <strong>the</strong> CES administrative data. They defined a quasi-control<br />

group <strong>of</strong> those who did not complete <strong>the</strong> programme, where <strong>the</strong>y used <strong>the</strong> cut<strong>of</strong>f <strong>of</strong> one<br />

third <strong>of</strong> programme period to separate <strong>the</strong> treatment from <strong>the</strong> quasi-control <strong>of</strong> non-

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

Saved successfully!

Ooh no, something went wrong!