Thursday 6 May 2010

Youth unemployment

The March Household Labour Force Survey is out, so time to update the prior results on youth unemployment. Recall that, back in February, I put up this little graph showing the residuals from a very very simple regression:

youth unemployment rate = constant + b1 * adult unemployment rate + residual

The residual tells us how far out any particular observation is from the trend line. So, if something weird starts happening to the youth unemployment rate RELATIVE to the adult rate, it'll show up in the residual of the regression. The data series goes back to 1986 and so includes a fair few prior recessions. Here's the updated series (noting of course that all points move a bit with the additional observation, attenuating results).

Again, the regression residual really starts spiking up at December quarter 2008. In all prior recessions, the residual never went past 2.7: in recessions, the youth unemployment rate gets hit harder than the adult rate. But, the adult unemployment rate this time round hasn't come close to touching what it had hit in the early 1990s, so we're not getting oddities solely due to a worse recession than usual. The residual spiked at 9.8 in December quarter 09; it's now at 8.5. If we take the prior period's maximum as being the biggest possible effect of a normal recession, then we have excess youth unemployment of 5.8 percentage points. As there were 161,600 workers in the 15-19 age bracket, we have about 9400 young workers currently unemployed who would have been employed had the pre-2008 relationship continued. In December quarter, that figure was about 12,000.

I still say the best first cut explanation is the elimination of the differential youth minimum wage. It's encouraging that more kids are finding work and that the residual is dropping (as we'd expect in the recovery). But excess youth unemployment due to high minimum wages has negative effects on lifetime earnings. I expect that few of those adversely affected will figure out that they ought to pin the blame on Labour for eliminating the youth minimum wage and on National for voting down Sir Roger Douglas's attempt to fix things.

Again, this is just a very simple regression. But it's a shame that National didn't at least ask Treasury (best I'm aware) to do something a bit more substantive rather than just asking the pollsters.


  1. It's a shame the Household Labour Force Survey doesn't split youth unemployment into 15-17 and 18-19 brackets as I wonder if the actual youth unemployment rate is being damped by the fact that a portion of the group have been classed as adults since 2001. (i.e. the spike in the 15-17 rate might be even more pronounced over the adult rate). Would that potentially make any difference to the results of your analysis?

  2. It can be done by requesting more disaggregated data. Prior work by Hyslop and Stillman found no effect of exposing 18-19 year olds to the adult rate earlier, but the period of their analysis had some of the strongest adult employment outcomes in the last 25 years; it would be rather interesting to have Treasury (or someone) continue with the disaggregated data.

    My expectation would be that it's coming from both groups: raising the 18-19 year olds to the adult rate wasn't a binding constraint when folks couldn't find workers and had to pay higher than minimum; now, that constraint has become binding on the 18-19 year olds at the same time as 15-17 year olds become subject to it.