All surveys have both strengths and weaknesses, and the above discussion has highlighted aspects of the former with respect to the 2002 NATSISS. What now follows considers some deficiencies of these data, in particular the lack of useful information concerning labour market experience. Specifically, the data set has no measures of either the length of time individuals have spent in paid employment (general labour market experience) or how long employed individuals have been in their current place of work (tenure). The discussion now examines the potential significance of the omission from the data of measures of general labour market experience).

An important focus of modern labour economics concerns the role of skills or, to use the accepted parlance, human capital. Human capital is seen to be a major—even the major—contributor to individuals’ success or otherwise in the labour market. There are two important aspects of human capital: formal education and the skills acquired by individuals from on-the-job training. In both areas, there are significant issues associated with measurement, since the pure human capital aspects of both education and training are not directly observed.

Labour market experience is typically represented in surveys like the 2002 NATSISS by the length of time spent in paid employment. Unfortunately, this variable is unavailable in the survey, and this raises the possibility that labour market statistical analyses of the 2002 NATSISS will provide inadequate, even misleading, results concerning the true determinants of Indigenous labour market success or failure.

Not having information on labour market employment history can be seen to be a major weakness of the 2002 NATSISS. In part, this is because Indigenous Australians have much higher rates of movement between labour force states than non-Indigenous Australians (Gray & Hunter 2005) and have much more interrupted labour market histories. For example, using a longitudinal sample of Indigenous job-seekers, Hunter, Gray & Jones (2000) find that 33.6 per cent of Indigenous males and 37.6 per cent of Indigenous females had been employed for less than 25 per cent of the time since leaving school. Only 16.5 per cent and 18.0 per cent of Indigenous males and females respectively had been employed for more than 75 per cent of the time since leaving school.

In order to illustrate the extent of the potential problem associated with the omission of measures of labour market experience from the 2002 NATSISS, we have examined econometric modelling in two areas: wages and being in employment. Our aim is to demonstrate the likely empirical importance of having to use the wrong variable. Our approach is to use an alternative data set that contains both a poor and a better measure of labour market experience. The poor measure is the length of time individuals could have spent in the labour force after finishing formal education, and the better measure is the number of years an individual has actually spent in paid employment. The models are estimated using both labour market experience measures and the results compared. One such data set can be derived from the HILDA survey.

We have chosen the female sample, since the potential significance of not having the more correct experience measure will be greater for groups with less attachment to the paid labour force, such as women (and Indigenous individuals). The econometric models are now briefly described.

Wage determination exercises take many forms, with the most basic human capital approach being represented by the following equation:

Wage = a + b*EXP + c*EXP2 + d*YOS + e

Where wage is the log of the hourly wage received by the individual, EXP is the number of years of paid employment, and YOS is the number of years of formal education. EXP2 is the square of the experience term, which is included because it is believed that the wage-experience term is non-linear. Table 10.5 compares the coefficients from the estimation of this wage equation (with the log of wages as the dependent variable) for the 2002 NATSISS specification and the HILDA specification.

a. All coefficients are significant at the 1% level.

Source: Author’s calculations

While the results are apparently similar for the two specifications (certainly the coefficients on years of schooling are very close), closer inspection suggests that at low levels of measured experience there are significant differences in the wage relationships. This is illustrated in Table 10.6, which shows the percentage change in individuals’ hourly wages for additional years of experience at different levels of experience.

Source: Author’s calculations

The results of Table 10.6 suggest the following:

We then repeated the above exercise with respect to estimating the determinants of whether or not a person is employed. The typical econometric approach used in this area takes an equation of the following form:

EP = a + bEXP + cEXP2 + dEDUC + eDEMOGRAPHY + e

Where EP is the probability that an individual is employed, EDUC are measures of education and DEMOGRAPHY reflects demographic factors. In our exercise, DEMOGRAPHY includes measures of marital status, whether or not the person is an immigrant, and the presence and age of children. The major relationship sizes for both specifications are available from the authors, and the experience effects are now shown in Table 10.7.

Source: Author’s calculations

The data of Table 10.7 suggest strongly that the poor measure of labour market experience available from the 2002 NATSISS has a significant potential to be misleading with respect to the effects of labour market experience on employment. The following results can be highlighted:

These comparative exercises make it apparent that the statistical problem associated with the omission in the 2002 NATSISS of a good measure of labour market experience are potentially very important. By comparing the same modelling with results found with a data set which has available a better measure of experience, it is clear that the 2002 NATSISS understates the value of experience for wages, and that this understatement becomes less as the experience measure increases. Similarly, results on the determinants of employment using the 2002 NATSISS seem to importantly get the story wrong with respect to the true role of experience. And, as with wages, the extent of the problem seems to be greater at the lower levels of experience.

It is important to record that the interpretation difficulties associated with the 2002 NATSISS not having an accurate measure for labour market experience seem to be confined to estimation of the true role of experience. In other words, the modelling and data problem has not affected estimates of the role of variables such as education with respect to wages, and education and demography with respect to the determinants of employment. This suggests that even though researchers are unlikely to be able to show with accuracy the effect of experience on labour market success, there are no associated difficulties for determining the true role for Indigenous labour market performance of other critical variables.