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.

149<br />

(1990) extended and improved <strong>the</strong> Hausman-Wise model. These models incorporate<br />

attrition into <strong>the</strong> behavioural model.<br />

Selection modelling is central to our model <strong>of</strong> intrinsic interest, bivariate modelling <strong>of</strong><br />

SYETP and employment outcomes. Accounting separately for attrition within this model,<br />

would require at least trivariate selection modelling. Recent work by Capellari (2001)<br />

uses <strong>the</strong> multivariate probability approach to account for endogenous panel attrition and<br />

selectivity in earnings, allowing estimation by application <strong>of</strong> simulated maximum<br />

likelihood techniques for <strong>the</strong> multi-dimensional integrals. However early examinations by<br />

Schmertmann (1994) found that <strong>the</strong> estimation approaches available for selectivity<br />

modelling <strong>of</strong> higher orders were not very good. The estimation suffers from <strong>the</strong> strict<br />

restrictions that need to be assumed for <strong>the</strong> bivariate correlations in <strong>the</strong> error distribution.<br />

It was concluded that in models with many selection criteria, or where sample sizes are<br />

small, <strong>the</strong> methods would not give good estimates. Fur<strong>the</strong>rmore, very strong assumptions<br />

are required about <strong>the</strong> correlations between response, SYETP participation and<br />

employment, and so on <strong>the</strong>se grounds this method <strong>of</strong> incorporating attrition with<br />

selection modelling is ruled out for this application. Fur<strong>the</strong>rmore, considering <strong>the</strong> small<br />

sample size we have available, combined with <strong>the</strong> strength <strong>of</strong> <strong>the</strong> assumptions required,<br />

this is beyond <strong>the</strong> scope <strong>of</strong> this paper.<br />

5.3 Empirical aspects <strong>of</strong> <strong>the</strong> effects <strong>of</strong> attrition on estimates<br />

Recent examinations <strong>of</strong> <strong>the</strong> empirical effects <strong>of</strong> attrition have raised <strong>the</strong> argument that <strong>the</strong><br />

<strong>the</strong>oretical potential for attrition bias does not always warrant concern. Fitzgerald et al.<br />

(1998a) concluded that <strong>the</strong> empirical existence and magnitude <strong>of</strong> attrition bias was not<br />

clearly related to <strong>the</strong> size <strong>of</strong> <strong>the</strong> attrition rate. They also concluded that attrition which is<br />

random in nature can result in no measurable bias arising in economic modelling, in<br />

accordance with earlier statistical literature (Fitzgerald et al. (1998a): 256).<br />

Fur<strong>the</strong>r recent studies on <strong>the</strong> evidence for <strong>the</strong> effects <strong>of</strong> attrition bias on estimates in<br />

labour market analysis make stronger claims. Lillard and Panis (1998) p456 argued that<br />

<strong>the</strong>ir results for <strong>the</strong> US panel data Michigan Panel Study on Income Dynamics (PSID)

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

Saved successfully!

Ooh no, something went wrong!