Evaluation of the Australian Wage Subsidy Special Youth ...

Evaluation of the Australian Wage Subsidy Special Youth ... Evaluation of the Australian Wage Subsidy Special Youth ...

02.06.2014 Views

114 Table 4.1 Difference between treatment group and comparison group SYETP mean s.d. Non- SYETP mean Female 43.3 0.50 41.0 0.49 2.3* Average age 1984 19.0 1.97 20.1 2.42 1.1* Aboriginal/Torres Strait Islander 1.0 0.10 3.1 0.17 2.1* Other ethnic minority 7.7 0.27 7.9 0.27 0.2 Married 1984 2.9 0.17 12.5 0.33 9.6* Spouse employed 1984 1.9 0.14 6.3 0.24 4.4* Children 1984 1.9 0.14 5.8 0.23 3.9* Highest qualification in 1984 Degree/diploma 7.7 0.27 12.2 0.33 4.5* Apprenticeship 2.9 0.17 8.6 0.28 5.7* Other post-school qualification 6.7 0.25 7.1 0.26 0.4 Year 12 of school 23.1 0.42 13.9 0.35 9.2* Year 11 of school 17.3 0.38 13.6 0.34 3.7* Year 10 of school 31.7 0.47 31.5 0.46 0.2 Year 9 of school 10.6 0.31 12.6 0.33 2.0* Parental background when resp. aged 14 Father postschool qualification 26.0 0.44 34.6 0.48 8.6* Mother postschool qualification 20.2 0.40 18.2 0.39 2.0* Father manager, professional, para-professional 25.0 0.44 25.8 0.44 0.8 Father not employed 3.8 0.19 5.6 0.23 1.8* Father not present 19.2 0.40 15.4 0.36 3.8* Mother manager, professional, para-professional 6.7 0.25 9.8 0.30 3.1* Mother not employed 48.1 0.50 55.3 0.50 7.2* Mother not present 8.7 0.28 5.0 0.22 3.7* Longest job ever held by 1984 Never held a job 11.5 0.32 11.6 0.32 0.1 < 1 year 55.8 0.49 40.1 0.49 15.7* 1 year 13.5 0.34 13.5 0.34 0.0 2 years 13.5 0.34 14.1 0.35 0.6 3 years or more 5.8 0.23 19.8 0.40 14* Average pre-programme unemployment 69 19.0 1.97 20.1 2.42 7.5* Ever employed in 1986 70 86.5 0.34 72.9 0.44 13.6* Ever government programme 1986 71 14.4 0.35 10.7 0.31 3.7* Number of cases 104 1179 NOTE: Column 5 shows the t statistic for hypothesis that difference of mean for SYETP and comparison is zero, where * indicates is significant at the 1 percent level of significance. s.d. SYETP versus comparison absolute difference in means 68 68 The statistic for any variable is the absolute value of the difference in means for SYETP and the control groups. 69 Proportion of 1984 reference period to 3 June spent unemployed. 70 Ever held a non-subsidised, non-government program job in the 1986 reference period, after the first 17 weeks. 71 Ever go on a government program, including SYETP, in the 1986 reference period.

115 4.2 Propensity score matching methods The propensity score matching methods, expounded in Rosenbaum and Rubin (1983), Dehijia and Wahba (1998), Heckman, Ichimura and Todd (1997), Heckman, Ichimura and Todd (1998) and Imbens (2000) have more recently been applied in the evaluation literature. Lechner (2000, 2001) extended the propensity score matching methods to the multi-treatment case. Using propensity score matching moves the emphasis away from specifying the selection bias towards more careful construction of the comparison group. It has been suggested that the key enhancement for evaluation allowed by this method is the comparison of comparable people (Heckman, Lalonde and Smith (1999): 2083). In order to do this, irrelevant comparison cases, which are not similar to the treated, are removed from the analysis. Matching methods find for each individual in the treated, at least one comparison group member with very similar pre-treatment characteristics. The differences in outcomes after the treatment are then attributed to the programme. Recalling the evaluation problem discussed earlier, matching is subject initially to the same difficulty of all nonexperimental methods where assignment to treatment is non-random. However Rubin (1974) showed that matching balances the distributions of all pre-treatment characteristics that influence assignment to the treatment, and so gives an unbiased estimate of treatment on the treated, as long as all relevant similar pre-treatment characteristics (X) are controlled for, and the Conditional Independence Assumption (CIA) is invoked (further explained below). Propensity score matching uses the propensity score to provide a single measure of the set of characteristics (X) that influence the probability of participating and employment. Rosenbaum and Rubin (1983) established that if matching on a set of observed characteristics is valid, then matching on the probability of selection into the programme conditional on these characteristics, the propensity score, is also valid. Whereas matching on each characteristic leads to problems with dimensions, the propensity score reduces the problem to a single dimension.

114<br />

Table 4.1 Difference between treatment group and comparison group<br />

SYETP<br />

mean<br />

s.d.<br />

Non- SYETP<br />

mean<br />

Female 43.3 0.50 41.0 0.49 2.3*<br />

Average age 1984 19.0 1.97 20.1 2.42 1.1*<br />

Aboriginal/Torres Strait Islander 1.0 0.10 3.1 0.17 2.1*<br />

O<strong>the</strong>r ethnic minority 7.7 0.27 7.9 0.27 0.2<br />

Married 1984 2.9 0.17 12.5 0.33 9.6*<br />

Spouse employed 1984 1.9 0.14 6.3 0.24 4.4*<br />

Children 1984 1.9 0.14 5.8 0.23 3.9*<br />

Highest qualification in 1984<br />

Degree/diploma 7.7 0.27 12.2 0.33 4.5*<br />

Apprenticeship 2.9 0.17 8.6 0.28 5.7*<br />

O<strong>the</strong>r post-school qualification 6.7 0.25 7.1 0.26 0.4<br />

Year 12 <strong>of</strong> school 23.1 0.42 13.9 0.35 9.2*<br />

Year 11 <strong>of</strong> school 17.3 0.38 13.6 0.34 3.7*<br />

Year 10 <strong>of</strong> school 31.7 0.47 31.5 0.46 0.2<br />

Year 9 <strong>of</strong> school 10.6 0.31 12.6 0.33 2.0*<br />

Parental background when resp. aged 14<br />

Fa<strong>the</strong>r postschool qualification 26.0 0.44 34.6 0.48 8.6*<br />

Mo<strong>the</strong>r postschool qualification 20.2 0.40 18.2 0.39 2.0*<br />

Fa<strong>the</strong>r manager, pr<strong>of</strong>essional, para-pr<strong>of</strong>essional 25.0 0.44 25.8 0.44 0.8<br />

Fa<strong>the</strong>r not employed 3.8 0.19 5.6 0.23 1.8*<br />

Fa<strong>the</strong>r not present 19.2 0.40 15.4 0.36 3.8*<br />

Mo<strong>the</strong>r manager, pr<strong>of</strong>essional, para-pr<strong>of</strong>essional 6.7 0.25 9.8 0.30 3.1*<br />

Mo<strong>the</strong>r not employed 48.1 0.50 55.3 0.50 7.2*<br />

Mo<strong>the</strong>r not present 8.7 0.28 5.0 0.22 3.7*<br />

Longest job ever held by 1984<br />

Never held a job 11.5 0.32 11.6 0.32 0.1<br />

< 1 year 55.8 0.49 40.1 0.49 15.7*<br />

1 year 13.5 0.34 13.5 0.34 0.0<br />

2 years 13.5 0.34 14.1 0.35 0.6<br />

3 years or more 5.8 0.23 19.8 0.40 14*<br />

Average pre-programme unemployment 69 19.0 1.97 20.1 2.42 7.5*<br />

Ever employed in 1986 70 86.5 0.34 72.9 0.44 13.6*<br />

Ever government programme 1986 71 14.4 0.35 10.7 0.31 3.7*<br />

Number <strong>of</strong> cases 104 1179<br />

NOTE: Column 5 shows <strong>the</strong> t statistic for hypo<strong>the</strong>sis that difference <strong>of</strong> mean for SYETP and comparison<br />

is zero, where * indicates is significant at <strong>the</strong> 1 percent level <strong>of</strong> significance.<br />

s.d.<br />

SYETP<br />

versus<br />

comparison<br />

absolute<br />

difference in<br />

means 68<br />

68 The statistic for any variable is <strong>the</strong> absolute value <strong>of</strong> <strong>the</strong> difference in means for SYETP and <strong>the</strong> control<br />

groups.<br />

69 Proportion <strong>of</strong> 1984 reference period to 3 June spent unemployed.<br />

70 Ever held a non-subsidised, non-government program job in <strong>the</strong> 1986 reference period, after <strong>the</strong> first 17<br />

weeks.<br />

71 Ever go on a government program, including SYETP, in <strong>the</strong> 1986 reference period.

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

Saved successfully!

Ooh no, something went wrong!