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.

141<br />

Table 4.7 Matching results, All-Within- caliper/Radius with replacement<br />

Match<br />

with<br />

caliper<br />

width<br />

0.001<br />

Match<br />

with<br />

caliper<br />

width<br />

0.005<br />

Match<br />

with<br />

caliper<br />

width<br />

0.01<br />

Match<br />

with<br />

caliper<br />

width<br />

0.02<br />

Match<br />

with<br />

caliper<br />

width<br />

0.05<br />

Difference in employment 90 for 0.14 0.17 0.17 0.16 0.15<br />

matched treated and comparisons<br />

F statistic 91 13.3 13.5 16.7 15.3 14.1<br />

df<br />

Number <strong>of</strong> SYETP matched 87 101 102 103 104<br />

Number <strong>of</strong> comparison which satisfy 462 886 917 918 920<br />

<strong>the</strong> caliper rule<br />

Number <strong>of</strong> comparisons unmatched 458 34 3 2 0<br />

Mean Standardized bias calculation 8.18 6.89 7.72 7.84 7.77<br />

after matching 92<br />

4.10 Discussion<br />

The PSM results are now compared to those found for <strong>the</strong> Heckman bivariate normal<br />

selection model in <strong>the</strong> earlier replication chapter. Earlier, <strong>the</strong> difficulty <strong>of</strong> which PSM<br />

result to choose was addressed. The 0.001 caliper result was chosen as <strong>the</strong> result for<br />

comparison upon <strong>the</strong> basis that it had no great deterioration in performance but covered<br />

all SYETP cases. This facilitates similarity <strong>of</strong> conditions with <strong>the</strong> Heckman result.<br />

We need to compare <strong>the</strong> marginal effect <strong>of</strong> SYETP in <strong>the</strong> Heckman probit to <strong>the</strong><br />

matching measure to make a common measure <strong>of</strong> <strong>the</strong> mean effect <strong>of</strong> treatment on <strong>the</strong><br />

treated. The Table 4.8 shown below contains <strong>the</strong>se estimates and <strong>the</strong>ir statistical<br />

significance. Both methods find a positive statistically significant effect on postprogramme<br />

employment from participation in SYETP. The difference between <strong>the</strong><br />

employment gain estimated from <strong>the</strong> two different methods is however quite large.<br />

Whereas for <strong>the</strong> Heckman bivariate probit <strong>the</strong> employment gain is closer to 30 percentage<br />

points, for PSM <strong>the</strong> employment gain is estimated as closer to 20 percentage points.<br />

90 Ever employed in 1986 survey.<br />

91 Degrees <strong>of</strong> freedom df=n-1, n=number <strong>of</strong> matched treated and untreated. The t statistic is not produced<br />

for this type <strong>of</strong> PSM.<br />

92 For all variables entering <strong>the</strong> probit <strong>of</strong> SYETP participation.

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

Saved successfully!

Ooh no, something went wrong!