Musical Chairs in Britain, Revisited
Economist writer H.C. has an interesting Free Exchange post discussing Scott’s “musical chairs” model and British unemployment. Scott also comments. Here is my curve-fitting exercise:
One of the difficulties in working with the UK macro data is that the ONS don’t produce a good quality high frequency time series for hourly wages. Of the available data we have low frequency, high accuracy data in the Annual Survey of Hours and Earnings (ASHE) based on very large sample of employers’ payroll data held by HMRC. At the other end we have high frequency, low accuracy data in the Average Weekly Earnings (AWE) data, produced from a small survey which deliberately excludes a number of workers (notably, I believe, high earners and small employers). I compared the range of various sources at the beginning of last year.
In this post I’ll use the lower quality AWE data, from which we have time series for both regular (ex bonuses) and total weekly pay, and also for average actual hours worked. I’ll also assume that bonus pay is not very sticky. From either pay series we can then calculate average hourly wages by simple division.
The musical chairs model says there is a strong positive correlation between the unemployment rate and the ratio of hourly wages to nominal GDP. I’ll cheat slightly (I did warn you this was a curve-fitting exercise) and use nominal GVA at basic prices instead of NGDP. An increase in nominal GDP which is accounted for entirely by an increase in indirect taxes would not be expected to increase the resources available to hire workers; since we’ve had big VAT shocks in the UK I think this is a reasonable “cheat”.
Here’s the first version of the musical chairs graph, using regular wages to derive hourly wages:
I would say this has done quite well since 2008: though the recovery is predicted slightly too fast, and the correlation is absent for the period prior to 2008.
A second interpretation is to use aggregate employee income instead of nominal GDP/GVA, so changes in the labour share of GDP don’t affect the results:
This is a better fit.
One thing that either of these models will struggle with, if only because of data limitations, is the shift between employment and self-employment. There have been two very large swings in UK self-employment in the period covered above: a 9% rise in 2003 and a 10% rise in 2013/4. We cannot really measure the wages of the self-employed, and as far as I know those workers are entirely excluded from the AWE data – I would not find it surprising that the musical chairs model appears to breaks down for that kind of shock. If there are studies on self-employment and wage stickiness I’d be interested to hear in the comments.