7.3 Education as an exogenous, continuous determinant of hourly earnings
We next re-estimate the regressions on the hourly earnings of workers at age 21 using ‘Effective Years of Education and Training’ as a regressor in place of the five indicator variables for formal qualifications. These OLS estimates in the first two columns of Table 19 can be compared to those under this somewhat more flexible specification reported in Table 16. We found earlier that the receipt of both School Certificate and University Bursary had positive and significant effects on hourly earnings. Yet, the estimated coefficients on this continuous measure of education or training are not statistically different from zero in both the short and long regressions. There is no evidence from these regressions that the investment in education or training raises potential wage rates in the labour market. It must be said, however, that these results do not rule out the possibility that these same investments raise hourly earnings beyond age 21.
| OLS | 2SLS | |||
|---|---|---|---|---|
| Independent Variables | WithoutBackgroundFactors | WithBackgroundFactors | WithoutBackgroundFactors | WithBackgroundFactors |
| Constant | 1.848**(0.184) | 1.816**(0.251) | 1.358**(0.488) | 1.351(2.338) |
| Effective Years of Education and Training | 0.008(0.013) | 0.001(0.014) | 0.046(0.032) | 0.020(0.152) |
| Actual Years of Work Experience | 0.063**(0.016) | 0.062**(0.017) | 0.067(0.059) | 0.134(0.152) |
| Maori | 0.081(0.050) | 0.089*(0.051) | 0.081(0.053) | 0.109(0.069) |
| Female | -0.017(0.028) | -0.023(0.030) | -0.026(0.029) | -0.019(0.031) |
| Male · Number Children Born to Respondent | -0.079(0.072) | -0.085(0.073) | -0.063(0.077) | -0.109(0.098) |
| Female · Number Children Born to Respondent | 0.105(0.100) | 0.098(0.102) | 0.170(0.111) | 0.121(0.243) |
| Part-Time Employment (<30 Hours per Week) | 0.079*(0.047) | 0.081*(0.047) | 0.060(0.072) | 0.144(0.134) |
| Enrolled in Education | 0.034(0.050) | 0.036(0.051) | -0.029(0.070) | -0.001(0.234) |
| Mother had School Qualification | --- | 0.021(0.033) | --- | 0.033(0.041) |
| Mother had Post-School Qualification | --- | 0.065(0.041) | --- | 0.065(0.060) |
| Mother had University Degree | --- | 0.075(0.074) | --- | 0.112(0.113) |
| Father had School Qualification | --- | 0.011(0.033) | --- | 0.003(0.048) |
| Father had Post-School Qualification | --- | -0.042(0.047) | --- | -0.043(0.059) |
| Father had University Degree | --- | -0.044(0.058) | --- | -0.019(0.079) |
| Years in Single-Adult Family | --- | 0.048(0.110) | --- | 0.047(0.110) |
| Maximum Number of Children in Family | --- | -0.010(0.012) | --- | -0.012(0.016) |
| Years Family Received Benefit | --- | 0.039(0.108) | --- | 0.029(0.143) |
| Real Family Income (in $10,000 units) | --- | 0.017(0.014) | --- | 0.013(0.020) |
| Mean Conduct Problems Score | --- | 0.001(0.002) | --- | 0.002(0.004) |
| Years Truant, Suspended or Expelled | --- | 0.027(0.250) | --- | 0.094(0.397) |
| Convicted of Criminal Offence | --- | -0.092*(0.051) | --- | -0.095(0.074) |
| Alcohol/Drug Abuse or Dependence | --- | -0.013(0.031) | --- | -0.002(0.050) |
| R2 | 0.035 | 0.052 | 0.020 | 0.025 |
| Adjusted R2 | 0.023 | 0.020 | 0.008 | -0.008 |
| Number of Observations | 671 | |||
| Mean of Dependent Variable | 2.159 | |||
** Significantly different from zero at 1% level.
* Significantly different from zero at 10% level.
Notes: These data are taken from respondents in the CHDS who provided valid information for the purposes of this study. The dependent variable is the natural logarithm of hourly earnings for the 671 individuals who were working at age 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 variables ‘Actual Years of Work Experience’ and ‘Effective Years of Education and Training’ are both treated as endogenous variables under Two-Stage Least-Squares (2SLS) estimation in the last two columns. The instrumental variables include an indicator for the CHDS child being an only child, and the number of older and younger male and female siblings in the family. These five variables are also interacted with the gender of the CHDS child to allow for the impact of sex composition of siblings on educational attainment to vary by the gender of the CHDS child. See Butcher and Case (1994) for the use of a similar approach with panel data in the U.S. Using this flexible specifications, the R2 statistic from the regression of ‘Effective Years of Education and Training’ against these instrumental variables is only 0.030.
At the same time, the estimated rates of return on a year of work experience are positive and statistically significant at better than a 1% level in these OLS regressions. These estimated effects are nearly identical in the short (0.063) and long (0.062) specifications. With or without measured family backgrounds, the average rate of return on a year of work experience is slightly more than 6%.
The change in the way education is measured in these regressions has little impact on the estimated coefficients on ethnicity. The estimated coefficients on Maori in Table 16 were 0.088 and 0.083 in the short and long regressions, respectively. The estimated coefficients on the same variable in Table 19 are 0.081 and 0.089 in the short and long regressions, respectively. All four estimated coefficients are significantly different from zero at between a 9% and 11% level.
7.4 Education as an endogenous determinant of hourly earnings
Instrumental variables are used to endogenise both ‘Effective Years of Education and Training’ and ‘Actual Years of Work Experience’ in these wage regressions. These 2SLS results are reported in the last two columns of Table 19. Again, the number, gender and birth order of siblings are used as instruments. Like the 2SLS findings on the accumulation of work experience, these results are disappointing. There is no evidence of any effects of years of education and training on hourly earnings. Although the estimated coefficients increase in magnitude when we move from the OLS to 2SLS results, they are still statistically insignificant. In addition, the estimated coefficients on work experience are now also statistically insignificant. The fact that the instrumental variables capture a small proportion of the variation in actual work experience has caused the standard errors on these estimated coefficients ‘blow up’.
The key for our purposes is whether or not the endogenising of the variables on education and training and work experience has any impact on the estimated effects of ethnicity on hourly earnings. The answer to this question is ‘no’. Both estimated coefficients on Maori continue to be positive, but are not statistically significant at a 10% level. Observationally equivalent Maori face higher, and not lower, wages in the labour market.[25]
Notes
- [25]Several attempts were made in an earlier version of this report to correct for sample selection bias for contemporaneous working behaviour in the wage regression by including the usual Heckman term. These procedures yielded estimated coefficients on the indicator variables for Maori that were always positive, and generally insignificantly different from zero at even a 10% level. The problem is that the lack of valid instruments for the IV procedure reported in this section also hamper efforts to identify this sample selection term.
