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.

101<br />

Heckman, Lalonde and Smith (1999) p1912 clearly point out that different choices about<br />

data handling such as variable choice and decisions about whom to include in <strong>the</strong> sample<br />

can strongly influence <strong>the</strong> measurement <strong>of</strong> treatment impacts. To avoid differences<br />

arising from this source, <strong>the</strong> initial replication is made as close as possible to that <strong>of</strong> <strong>the</strong><br />

original. Using <strong>the</strong> reported bivariate probit equation in Richardson (1998) Tables 4 and 5,<br />

<strong>the</strong> variables matching those described are selected for use in <strong>the</strong> estimation. This means<br />

that <strong>the</strong> same identifying restrictions are applied. Thus, CEP referrals and age are<br />

excluded from <strong>the</strong> employment equation but included in <strong>the</strong> selection equation,<br />

additionally in <strong>the</strong> employment equation qualifications, children, marital status and health<br />

affecting work are 1985 or 1986 dated and in <strong>the</strong> selection equation <strong>the</strong>y are 1984 dated.<br />

The same set <strong>of</strong> base variables are used for <strong>the</strong> equations as that <strong>of</strong> Richardson (1998) by<br />

reference to <strong>the</strong> reported bivariate probit results. The variables are described in detail in<br />

Appendix 1: Data Appendix, Table 1. Analysis was carried out in STATA 7.0.<br />

3.4 Replication results<br />

The results <strong>of</strong> replication for <strong>the</strong> bivariate probit are shown in Table 3.1, parts a and b. To<br />

facilitate comparison <strong>of</strong> <strong>the</strong> replication result with <strong>the</strong> original results, <strong>the</strong> estimates as<br />

shown in Richardson (1998) Tables 4 and 5 are repeated in column 1 <strong>of</strong> Table 3.1 part a<br />

and part b. Table 3.1 part a shows <strong>the</strong> employment equation <strong>of</strong> <strong>the</strong> bivariate probit, while<br />

Table 3.1 part b shows <strong>the</strong> selection equation <strong>of</strong> <strong>the</strong> bivariate probit which models <strong>the</strong><br />

entry into SYETP. The second column <strong>of</strong> Table 3.1 gives <strong>the</strong> replication estimates.<br />

The replication is found to be generally successful. The replicated results match closely<br />

those <strong>of</strong> Richardson (1998). There is some minor variation, but at <strong>the</strong> 2 decimal places<br />

level, most estimates <strong>of</strong> <strong>the</strong> coefficients are identical. There is slightly more disparity in<br />

<strong>the</strong> t-statistics. The coefficient <strong>of</strong> key interest is <strong>the</strong> SYETP variable in <strong>the</strong> employment<br />

equation, which measures <strong>the</strong> treatment effect. In this case, <strong>the</strong> replication result is<br />

slightly larger, and <strong>the</strong> t-statistic is also larger. However, in no instance does <strong>the</strong><br />

replication report results that are not closely comparable to those reported in Richardson<br />

(1998). The appendix Tables A2.0a and A2.0b show <strong>the</strong> replicated univariate probits, for

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

Saved successfully!

Ooh no, something went wrong!