5.2  Interacting two-way split in youth ethnicity with other covariates

Results from the short and long regressions, estimated separately for Maori and non-Maori, are reported in Table 11. The reason for interacting ethnicity with all other regressors is to see whether or not the estimated marginal effects vary systematically between the two ethnic groups. The estimated effect of years not enrolled in education or training is significantly smaller for Maori (0.273) than non-Maori (0.434) in the short regression, but the estimated parameters are not statistically different from one another in the long regression. There is no consistent evidence in these regressions that the rates of accumulation of actual work experience relative to potential work experience vary between the ethnic groups.

** Significantly different from zero at 1% level.

* Significantly different from zero at 10% level.

Notes: These data are taken from the 973 respondents in the CHDS who provided valid information for the purposes of this study. The dependent variable is the effective years of work experience accumulated by the individual between the ages of 16 and 21. Youth are defined as ‘Maori’ in these regressions if they identify Maori as at least one of their ethnicities at age 21, and had at least one parental figure claiming Maori ethnicity by age 14 of the CHDS child. All other youth are considered to be ‘non-Maori’. Standard errors are in parentheses.

The birth of a child to female respondents significantly reduces accumulated work experience for both ethnic groups. Yet, the marginal effects are substantially greater in magnitude for Maori in both the short and long regressions. The number of children born by age 21 has a larger detrimental impact on work experience for Maori relative to non-Maori women.

The influences of formal qualifications on work experience are fairly similar between the ethnic groups. Holding years not enrolled in education and training constant, the joint effects of a School Certificate or 6^th Form or Higher School Certificate are positive for both Maori and non-Maori in both the short and long regressions. The same can generally be said of personal and family background factors. However, small sample sizes, especially for Maori, make it difficult to distinguish any ethnic differences in the effects of many of these variables. We can say, however, that family background factors as a group are relatively more important for Maori than non-Maori at explaining the observed variation in work experience. The R² statistic is lower for Maori (0.198) than non-Maori (0.320) in the short regression, but higher for Maori (0.417) than non-Maori (0.356) in the long regression. Thus, the gain in explanatory power in adding these family background measures is substantially greater among Maori (110.6%) than non-Maori (11.3%).

5.3  Quantile regressions: Two-way split in youth ethnicity

Figure 1 shows that overall differences in work experience between non-Maori and Maori are concentrated in the lower tails of these distributions. Ordinary Least-Squares regressions estimate the effects of determinants on the mean values of work experience, once relevant factors are held constant. Quantile regressions estimate these effects at various points in the conditional distribution.[21] In particular, this regression technique indicates the extent to which our explanatory variables can capture differences between non-Maori and Maori in the lower tail of these distributions.

Our quantile regression results are summarised in Table 12 for the two-way split in youth ethnicity. Each column heading under the five percentile categories represents the results from two separate multiple regression models. Only the parameter estimates on the Maori indicator variable are reported. The results in the first row correspond to the short regression specification, where the additional covariates are those reported in the first column of Table 10 (ie, without background factors). The results in the second row correspond to the long regression specification, where the additional covariates are those reported in the second column of Table 10 (ie, with background factors).

** Significantly different from zero at 1% level.

* Significantly different from zero at 10% level.

Notes: These data are taken from the 973 respondents in the CHDS who provided valid information for the purposes of this study. The dependent variable is the effective years of work experience accumulated by the individual between the ages of 16 and 21. Youth are defined as ‘Maori’ in these regressions if they identify Maori as at least one of their ethnicities at age 21, and had at least one parental figure claiming Maori ethnicity by age 14 of the CHDS child. All other youth are considered to be ‘non-Maori’. Additional covariates are included in these regressions, but these results are not reported. The other explanatory variables included in these short and long specifications are listed in Table 10. Each column under a ‘Quantile’ lists the results from two separate regressions estimated in STATA. Standard errors are in parentheses.

An estimated coefficient in this regression model can be interpreted as a partial derivative of that conditional quantile with respect to the relevant regressor. For example, the estimated coefficient of –0.500 on Maori at the 0.10 quantile says that the estimated effect of being Maori lowers cumulative work experience by one-half year at the 10^th percentile in the experience distribution, once other factors have been held constant in this short regression. This estimated coefficient is significantly different from zero at better than a 2.5% level. As we move to higher points in the experience distribution, the effects of ethnicity appear to weaken. It falls to –0.266 at the 25^th percentile, and is significant at a 5.5% level. It is negative, but insignificant at higher points in the distribution. When we control only for personal characteristics, ethnic differences in the accumulation of work experience appear to be concentrated in the lower tail of the distribution.

When we control for both personal and family background factors (second row in Table 12), the disparities between non-Maori and Maori in the lower tail of the experience distribution disappear. The estimated coefficients on Maori are still negative, but insignificant in the long regressions at the 10^th and 25^th percentiles. Only the estimated coefficient at the 75^th percentile is significantly different from zero. It appears that our family background measures capture at least some of the ethnic disparities in the accumulation of work experience at the lowest levels.