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

90 An advantage over earlier SYETP analyses is that the comparison group was nonparticipants rather than participants in other programmes. This allows the more pertinent question for evaluation to be asked: instead of whether the programme was effective relative to another programme to whether the programme was effective in improving post-programme employment in an absolute sense compared to non-participants. Retention in the post-programme job is addressed through sample selection for those where continuous employment to the end of 1986 was not the case. This has potential to introduce the need to further model the selection process of retention or tenure in the one job, as it may be endogenous. Essentially, it selects a sample with short tenure jobs, and durations are censored at 1986. This might introduce other difficulties related to duration modelling and censoring of spells too. Richardson (1998), in relation to concerns for the sample reduction, examines the means of the retained amongst the treated and comparisons, and the effect on size of the treated sample against that of the comparisons. The conclusion of no bias problems relies on the observation that the retained are a similar share of treated and of comparisons, and that the proportion of SYETP before and after sample reduction is the same. This examination is more related to balance of the retention characteristic in the treated and comparison groups than whether retention is related to participation in SYETP.

91 2.3.5 General discussion Some general observations are apparent in comparing the SYETP analyses. All analyses of SYETP proceeded with survey data to enable the recording of post-programme employment. The earlier analyses of SYETP did not have a non-participant comparison group, whereas the Richardson (1998) analysis did. The results enable a gauge of whether participation in the SYETP programme was effective in improving post-programme employment. All analyses found positive effects. Three key themes emerge from the past analyses of SYETP. These relate to accounting for selection into programme participation, non-response in the observational data used, and controlling for differences in individual characteristics between the comparison and treatment groups. These themes are further developed in the later chapters of this study. In the following discussion, the past SYETP literature is assessed with regard to these points. Richardson (1998) is the evaluation of SYETP motivating this current study. The format of the Richardson (1998) evaluation forms a key basis for our analyses, which are contained in the later chapters starting with the Replication Study. In brief, Richardson used a Heckman bivariate probit which controlled for programme selection to analyse the SYETP using ALS panel data. A strong positive effect on post-programme employment was found. In this review, this evaluation is considered to contain the most sophisticated analysis of SYETP to date, allowing for both programme selection and individual characteristics. This evaluation also showed evidence that accounting for selection into SYETP participation in the modelling of employment was important in two ways: by testing the correlation coefficient and finding it statistically significant; and by showing the comparison to the unadjusted univariate employment results, which were also very different despite the employment equation being otherwise specified identically. The modelling of the effects on employment was the best developed of those evaluations reviewed, since it took into account selection into programme participation.

91<br />

2.3.5 General discussion<br />

Some general observations are apparent in comparing <strong>the</strong> SYETP analyses. All analyses<br />

<strong>of</strong> SYETP proceeded with survey data to enable <strong>the</strong> recording <strong>of</strong> post-programme<br />

employment. The earlier analyses <strong>of</strong> SYETP did not have a non-participant comparison<br />

group, whereas <strong>the</strong> Richardson (1998) analysis did. The results enable a gauge <strong>of</strong> whe<strong>the</strong>r<br />

participation in <strong>the</strong> SYETP programme was effective in improving post-programme<br />

employment. All analyses found positive effects.<br />

Three key <strong>the</strong>mes emerge from <strong>the</strong> past analyses <strong>of</strong> SYETP. These relate to accounting<br />

for selection into programme participation, non-response in <strong>the</strong> observational data used,<br />

and controlling for differences in individual characteristics between <strong>the</strong> comparison and<br />

treatment groups. These <strong>the</strong>mes are fur<strong>the</strong>r developed in <strong>the</strong> later chapters <strong>of</strong> this study.<br />

In <strong>the</strong> following discussion, <strong>the</strong> past SYETP literature is assessed with regard to <strong>the</strong>se<br />

points.<br />

Richardson (1998) is <strong>the</strong> evaluation <strong>of</strong> SYETP motivating this current study. The format<br />

<strong>of</strong> <strong>the</strong> Richardson (1998) evaluation forms a key basis for our analyses, which are<br />

contained in <strong>the</strong> later chapters starting with <strong>the</strong> Replication Study. In brief, Richardson<br />

used a Heckman bivariate probit which controlled for programme selection to analyse <strong>the</strong><br />

SYETP using ALS panel data. A strong positive effect on post-programme employment<br />

was found. In this review, this evaluation is considered to contain <strong>the</strong> most sophisticated<br />

analysis <strong>of</strong> SYETP to date, allowing for both programme selection and individual<br />

characteristics. This evaluation also showed evidence that accounting for selection into<br />

SYETP participation in <strong>the</strong> modelling <strong>of</strong> employment was important in two ways: by<br />

testing <strong>the</strong> correlation coefficient and finding it statistically significant; and by showing<br />

<strong>the</strong> comparison to <strong>the</strong> unadjusted univariate employment results, which were also very<br />

different despite <strong>the</strong> employment equation being o<strong>the</strong>rwise specified identically. The<br />

modelling <strong>of</strong> <strong>the</strong> effects on employment was <strong>the</strong> best developed <strong>of</strong> those evaluations<br />

reviewed, since it took into account selection into programme participation.

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

Saved successfully!

Ooh no, something went wrong!