4.2  Outcome Trends over the Period

We now turn our attention to describing the trends over the sample period in a variety of outcomes of interest to this analysis for the three age groups used above.[24] The two main outcomes that we focus on are being employed which captures the extensive margin of labour market adjustment, and the number of hours worked conditional on being employed which captures the intensive margin. Figure 4 graphs the quarterly trends in the fractions of each age group that is employed as a wage or salary worker, and the average actual weekly hours worked by wage or salary workers from the first quarter 1997 until the third quarter 2003.[25] Simple models for the impact of the minimum wage changes predict a change in outcomes for teenage workers would occur after the first quarter 2001.

The employment and hours-worked graphs in figure 4 display the importance of both age effects (young adults are more likely to work, and also to work more hours, than teenagers), and business cycle (and other) effects over time. Employment levels declined from 1997 until 1999 (2000 for 17-18 year-olds), before increasing over the later part of the period. The minimum wage changes that occurred after 2001 do not appear to have had a substantial effect on employment rates of teenage workers relative to young adults, although the trends for the different groups do vary somewhat. The employment rates of 18-19 year-olds picked up later than 16-17 and 20-25 year-olds, and continued to increase from 2001, while the employment of these other groups stabilized or fell a little. A similar story is observed for hours worked. Average hours worked for both 18-19 and 20-25 year-olds remained fairly stable over the sample period, while the hours for 16-17 year-olds were fairly constant or weakly increasing over the sample period, until a noticeable increase in 2002. These two graphs suggest that, except perhaps hours worked by 16-17 year-olds in 2002, there is little to suggest that employment or hours worked by teenage workers was affected by the minimum wage increases.

In figure 5 we present the quarterly patterns for other outcomes that may be affected by the youth minimum wage changes: namely study, unemployment, and inactivity (neither working nor studying) rates. Over the sample period, study rates fell slightly for 16-17 year-olds, and were reasonably stable or slightly increasing for 18-19 and 20-25 year-olds. Unemployment and inactivity rates tended to fall for 18-19 and 20-25 year-olds but were relatively stable for 16-17 year-olds. However, as with the employment and hours worked in figure 4, there is little, if any, systematically apparent change in the trends around 2001 for teenagers relative to young adults to suggest an impact of the minimum wage changes.

Figure 6 graphs the annual trends in the fraction receiving non-student benefits, the average actual weekly earnings for wage and salary workers, and the average actual weekly income for all individuals in each age group over the same sample period, derived from the June quarter HLFS-IS. The trends in the fractions receiving benefits suggest there may have been an increase for teenagers compared to young adults after 2001, although the age-group trends differ somewhat. Changes in average weekly earnings conditional on employment reflect the combination of changes in hours worked and changes in wage rates over time. This outcome allows us to evaluate the effect of the minimum wage reform on the average wage and salary worker. On the other hand, as changes in average weekly income also factor in the possible displacement effects that an increase in the minimum wage may have (i.e. some workers lose their jobs and receive either lower or zero income), it provides a better overall measure of the welfare effects of the minimum wage reform. Both of these graphs show similar patterns. The average earnings and incomes of 20-25 year-olds were roughly constant over the sample period, while 18-19 year-olds’ earnings increased until 2001 before falling in 2002, and their incomes increased slightly. Both the average earnings and incomes of 16-17 year-olds show a distinct increase after 2000 that is consistent with, and suggestive of, the increase in minimum wages of this group in 2001 and 2002.

The descriptive evidence presented here is suggestive of, at best, weak effects of the large increases in the minimum wages for young workers. We next formalise the analysis to control for these and other factors, and quantify the impact of the youth minimum wage policy changes on these outcomes. We begin with a very simple difference-in-differences framework and then control for various factors in a sequential manner in order to help understand their impact on the estimated minimum wage effects.

4.3  Difference-in-Differences Estimates of Employment and Hours Worked

Table 3 summarises the levels and changes in the employment rates and hours worked conditional on employment before and after the 2001 youth minimum wage policy changes for 16-17, 18-19, and 20-25 year-olds. We present the data by age group in columns (1) – (3), and the differences between the 16-17 and the 20-25 year-olds’ outcomes, and between the 18-19 and the 20-25 year-olds’ outcomes, in columns (4) and (5) respectively. Panel A pertains to the employment rate, while panel B pertains to hours worked. Within each panel, rows 1 and 2 show the average outcomes in the period before (1997q1-2001q1) and after (2001q2-2003q3) the policy change respectively, and row 3 presents the change in these outcomes. The numbers in bold at the bottom right of each panel are the simple difference-in-differences estimates of the impact of the minimum wage changes on the respective outcomes.

Panel A shows that the fractions of 16-17, 18-19 and 20-25 year-olds employed as wage or salary workers increased by 1.2, 0.6 and 0.5 percentage points, respectively, after the policy reform. Comparing the changes for 16-17 and 18-19 year-olds to that of 20-25 year-olds gives the simple difference-in-differences estimates of the effect of the minimum wage increases was to increase employment for 16-17 year-olds by 0.7 percentage points, and for 18-19 year-olds by 0.1 percentage points. However, neither of these estimates is statistically significantly different from zero.

Similarly, panel B shows that weekly hours worked increased by 1.31 hours for 16-17 year-olds, and decreased by 0.03 and 0.36 hours for 18-19 and 20-25 year-olds, respectively, after 2001. The resulting difference-in-differences estimates of the policy change are 1.67 hours for 16-17 year-olds, and 0.33 hours for 18-19 year-olds, respectively. The estimated increase for 16-17 year-olds is statistically significant and represents a large relative increase in hours worked (from a base of 17.4 hours per week, the 1.67 hours estimated increase represents a 10 percent increase).

4.4  Regression Analysis of Employment

We now extend the simple difference-in-differences analysis of the impact of the youth minimum wage changes using regression methods to control for other sources of variation that may confound the estimates of interest. In this subsection, we focus on the minimum wage impact on employment and present results from a variety of regressions, before considering the other outcomes in the next subsection. Ignoring individual and time subscripts, the basic regression specification we use is

(1)

where Y is the outcome of interest (Employment here), age16-17 and age18-19 are dummy variables for the respective teenage groups, Post-2001 is a dummy variable for the post-reform period, X is a vector of variables to control for other factors influencing the outcome, and u is an error term to capture unobserved effects. All specifications include single-age dummy variables to capture systematic age differences in the outcomes assumed constant over time, and quarter-specific dummy variables to flexibly capture time-varying effects due to aggregate business cycle and other factors assumed constant across age groups As the analysis essentially compares group averages of outcome variables to determine the effect of the policy reform, we present Huber-White robust standard errors that allow for arbitrary correlation in individual error terms within age-quarter cells. Our main focus of interest are the coefficients and , which represent the effects of the minimum wage reform on 16-17 and 18-19 year-olds controlling for other factors.

Table 4 contains the estimated minimum wage reform impacts on employment based on a variety of specifications. In the first column we present the difference-in-differences specification from table 3, except that it also includes the age- and quarter-specific dummy variables. The estimated impacts on teenage employment are 0.5 and 0.0 percent for 16-17 and 18-19 year-olds respectively, and as with the difference-in-differences estimates, are not statistically significantly different from zero. In column (2), we control for the demographic characteristics of the sample using dummy variables for gender, ethnicity, marital status, New Zealand born, urbanicity, and region of residence, and the relative size of the population of each age group (16-17, 18-19, 20-21, 22-25) in a particular year. The estimated employment effects in this specification are -0.6 and -1.0 percent for 16-17 and 18-19 year-olds but again are insignificantly different from zero.

As discussed above, if the policy reform has a direct effect on teenage workers it may also have an indirect effect on the outcomes of young adults. To examine this possibility, the specification in column (3) allows the policy change to have spillover effects on 20-21 year-olds, by including a dummy variable, age20-21*Post-2001, that is equal to 1 for 20-21 year-olds after 2001 and 0 otherwise.[26] To the extent that a spillover effect exists, the coefficients and will represent the direct effects of the minimum wage changes on teenagers, while the coefficient on this variable () will represent the indirect effect on 20-21 year-olds. The results for this specification continue to find modest and statistically insignificant impacts on teenage employment.

In column (4) we repeat the previous specification but, in addition, include an indicator variable for whether a proxy interview was used and interact this with each of the control variables. This allows the outcomes of proxy and non-proxy respondents to differ across the dimensions of the control variables but assumes that these outcomes are not differentially affected by the minimum wage reforms. The estimated impacts from this specification are only marginally different from those in column (3).[27]

The specification in column (5) allows the policy impact to vary across the three post-reform years by including separate dummy variables for the periods following each of the 2001, 2002 and 2003 changes: 2001 equals 1 for the 4 quarters from second quarter 2001 to first quarter 2002 and 0 otherwise, 2002 equals 1 for the 4 quarters from second quarter 2002 to first quarter 2003 and 0 otherwise, and 2003 equals 1 in the second and third quarters of 2003 and 0 otherwise. This might be important for 16-17 year-olds, given the phasing in of the reforms for that group. In fact, F-tests for the null hypotheses of constant year-effects for each of the age groups rejects the hypothesis for 16-17 (p-value=0.02) and 20-25 year-olds (p-value<0.01), but accepts the hypothesis for 18-19 year-olds (p-value=0.23). The estimated employment effects for 16-17 year-olds vary from a 2.2 percent increase in 2001 (significant at the 10% level), to essentially zero in 2002, and a 2.5 percent decline in 2003. A similar pattern is found for 18-19 and 20-25 year-olds: for 18-19 year-olds the estimates range from a 0.7 percent increase in employment in 2001 to a 1.9 percent decline in 2003, although with no statistical significance; for 20-25 year-olds, the estimates range from a 4.2 percent increase in 2001 (significant at the 1% level), to a 1.9 percent decline in 2003. Although the 2001 estimate for 20-21 year-olds is quite strong suggesting significant positive spillover effects for this group, the absence of any strong direct effects on 18-19 year-olds does draw into question the robustness of this interpretation, as well our identification strategy more generally.

The results in column (5) suggest that, although the individual estimates are generally insignificant, the small overall effects estimated in the previous specifications disguise quite different point estimates and patterns across the three years. In column (6), we allow for age-specific seasonal variation in employment by interacting quarterly dummy variables with each age dummy variable. The results in this specification are qualitatively the same as those in column (5).[28]

The final specification presented in table 4 also includes age-specific linear time trends to control for possible secular trends in the outcomes. The motivation for this is provide by suggestion in some of the descriptive graphs in figures 4-6 of age-group specific trends in the outcomes. For example, the study rate of 16-17 year-olds appears to be trending upwards prior to 2001, while the study rate of 20-25 year-olds is possibly trending down. The results in column (7) imply that while controlling for such trends has a limited effect on the point estimates of the impact of the minimum wage changes for 16-17 and 20-21 year-olds, it has a large effect on the estimates for 18-19 year-olds. The minimum wage reform is estimated to have increased employment for 18-19 year-olds by 3-3.5 percent in each of the post-reform years, and the estimates for 2001 and 2002 are significant at the 5% level. The F-statistic for the joint hypothesis that the time trend coefficients are zero is significant at the 1% level.