The long term employment data tells a chilling tale of a decade already lost in the USA. There is a fair argument that the economy has never recovered from the bust of the dot.com bubble – which challenges my prior conclusion that equity-busts do not hurt like debt-busts.
Up close, this looks to be a decent enough recovery (well, at least as good as the prior recovery).
Zooming out, however, reveals the full extent of the damage – and the lack of recovery. I have added a trend line to show that there was a reliable growth path prior to the dot.com crash.
To better show the output gap, i’ve taken the difference between the pre-2001 trend line and the actual data. This measure suggests the output gap reached ~17% of potential in Q3’09, and recovered to ~15% of potential in Q1’12 — and has since slouched back to ~16% of potential.
This just seemed too bad to be real – however it does not seem to be down to demographics. The above chart shows a normalised ratio of hours worked to the population. As you can see, the index of hours worked per person fell to a little under 1 standard deviation below average during the most recent recession, recovered to around -0.6 sigma by Q1’12, and is now around -0.65 sigma below average.
Given Women’s lib and improved medicine, I’d expect an uptrend in this ratio over time – indeed, these were the explanations posited to 2000 when hours worked per head were substantially outside of the 1964–1994 range.
When you consider how much larger the ‘eligible’ population has become (due to the above-mentioned factors), it’s shocking to see the employment to population ratio around the high end of the range traced in the boomer generation – when married women often did not work and mothers seldom did (and when children were more common).
The failure of aggregate hours worked and the employment to population ratio to recover suggests that the US has been in a long slump that began at the end of 2000.
It is only in the unemployment duration data that I find evidence for the conclusion that equity-busts do not hurt as much as debt-busts.
The plight of the unemployed this recession has been unusually difficult. The median duration of unemployment is presently ~18 weeks, which is down from a peak of ~25 weeks in mid 2010, but is up from ~17 weeks the prior month. The old range was 5 weeks to 10 weeks.
Taken together, this data leads me to two conclusions:
1/ that the dot.com bust did more damage that i had realised (especially if one considers that the housing boom was partly a policy response to the bust); and
2/ that those who argue that the current labour market slack is structural are drawing a pretty long bow.
The social and medical changes over the prior four decades will have pushed up potential labour force attachment. We should have seen an uptrend in labour force attachment as a result. That participation in employment has declined so far so fast strongly suggests that there is a problem on the demand side – and that there has been a problem for some time.