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

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

167<br />

5.5.4 Attrition: natural attrition or analytical sample reduction?<br />

Table 5.5 draws a distinction between <strong>the</strong> sources <strong>of</strong> sample reduction. Sample reduction<br />

resulting from attrition, where <strong>the</strong> respondent did not take part in later surveys, is here<br />

termed natural attrition. As outlined earlier, fur<strong>the</strong>r sample reduction through analytical<br />

selection also takes place before <strong>the</strong> final sample used for analysis is reached. Zabel<br />

(1998) points out such analytical selection can in <strong>the</strong>ory introduce biases <strong>of</strong> a similar<br />

nature to attrition bias.<br />

Indeed, Table 5.5 shows that both forms <strong>of</strong> sample reduction result in statistically<br />

significant differences to <strong>the</strong> means observed <strong>of</strong> those remaining in <strong>the</strong> final sample. This<br />

hints that both sources <strong>of</strong> sample loss may be suspected <strong>of</strong> altering <strong>the</strong> sample<br />

characteristics. Also, <strong>the</strong> pattern <strong>of</strong> differences varies by <strong>the</strong> form <strong>of</strong> sample reduction<br />

being natural attrition or analytical selection. This signals <strong>the</strong> added value <strong>of</strong> treating each<br />

source separately in order to allow for differential impacts.<br />

The first two columns <strong>of</strong> Table 5.5 show <strong>the</strong> mean and standard deviation in <strong>the</strong> final<br />

sample, to which <strong>the</strong> two sources <strong>of</strong> sample reduction are compared. The next four<br />

columns make up <strong>the</strong> first panel <strong>of</strong> <strong>the</strong> table which relates to <strong>the</strong> sample lost due to<br />

natural attrition, and <strong>the</strong>y show <strong>the</strong> mean, standard deviation, absolute difference in <strong>the</strong><br />

means and <strong>the</strong> t statistic, in a similar fashion to <strong>the</strong> former tables. The second panel gives<br />

<strong>the</strong> statistics for those lost though analytical selection. Only differences that are<br />

statistically significant at normally adopted confidence levels are discussed.<br />

The outcome variables ‘ever employed in 1986’ and ‘ever in a government programme in<br />

1986’ are not observable for those lost through natural attrition. It is useful to compare<br />

<strong>the</strong> statistically significant differences in <strong>the</strong> pr<strong>of</strong>ile <strong>of</strong> those remaining in <strong>the</strong> sample to<br />

those lost to natural attrition. Of those not interviewed again: fewer had taken part in<br />

SYETP, more had year nine <strong>of</strong> school but less had year 11, more had held jobs <strong>of</strong> 1-2<br />

years while less had held jobs <strong>of</strong> 2-3 years, <strong>the</strong> proportion <strong>of</strong> unemployment in 1984 had

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

Saved successfully!

Ooh no, something went wrong!