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.

61<br />

restriction was used. If <strong>the</strong>re was no exclusion restriction applied, and identification was<br />

based solely on functional form, <strong>the</strong>n Monte Carlo evidence in <strong>the</strong> literature indicates that<br />

<strong>the</strong> modelling suffers from poor performance. The estimation procedure also lacks fur<strong>the</strong>r<br />

detail, and so it is not clear whe<strong>the</strong>r <strong>the</strong> two equations needed for <strong>the</strong> Heckman sample<br />

selection model were estimated simultaneously, but it appears that a two step procedure<br />

was applied. If both equations are estimated as probits, <strong>the</strong>n this is inappropriate due to<br />

<strong>the</strong> nonlinearity <strong>of</strong> <strong>the</strong> probit. It was found that <strong>the</strong> factors affecting referrals did not have<br />

an independent affect on placement. Duration <strong>of</strong> <strong>the</strong> current unemployment spell also had<br />

an additional negative effect on placement if selection bias was not controlled for.<br />

Wielgosz found that tests for significance <strong>of</strong> <strong>the</strong> selection correction factor indicated no<br />

selection bias. But it was concluded that <strong>the</strong> very high correlation between <strong>the</strong> selection<br />

correction factor and <strong>the</strong> duration <strong>of</strong> unemployment, coupled with <strong>the</strong> significance <strong>of</strong> this<br />

variable when selection bias was not controlled for, suggested that that <strong>the</strong> sample<br />

selection bias was “…closely and solely related to duration <strong>of</strong> unemployment” (Wielgosz<br />

(1984): 19). It was fur<strong>the</strong>r noted that this was because to be referred, and so included in<br />

<strong>the</strong> sub-sample, was almost completely dominated by <strong>the</strong> length <strong>of</strong> unemployment spell.<br />

As a result, due to <strong>the</strong>ir correlation <strong>the</strong> coefficients for selection correction and<br />

unemployment duration were both insignificant in <strong>the</strong> placement equation because <strong>the</strong>y<br />

reflected <strong>the</strong> same phenomenon.<br />

Aungles and Stewart (1986) used <strong>the</strong> same referral and placement data as Wielgosz (1984)<br />

to examine referrals and <strong>the</strong> duration <strong>of</strong> unemployment, toge<strong>the</strong>r with exits from<br />

unemployment. Aungles and Stewart (1986) modelled durations <strong>of</strong> unemployment, and<br />

also found that <strong>the</strong> number <strong>of</strong> referrals jobseekers receive is related to <strong>the</strong> probability <strong>of</strong><br />

leaving unemployment. They found that jobseekers receipt <strong>of</strong> CES referrals were most<br />

likely to occur early in <strong>the</strong>ir unemployment spell, but once referred by <strong>the</strong> CES,<br />

differences did not exist in <strong>the</strong> probability <strong>of</strong> leaving unemployment. The hazard function<br />

for leaving unemployment initially rose to peak at 11 days, less than two weeks, after<br />

which it fell.

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

Saved successfully!

Ooh no, something went wrong!