So yesterday Steve asked the question about our unemployment rate, “Is it really 5.3% or is it something closer to 6.4%”? He made the important point that unemployment by itself isn’t as informative as we’d like to think. The participation rate is important information too. So what is the unemployment rate given the participation rate?
So what Steve did was add back the change in participation rate to the published unemployment rate to get an estimate of a participation-adjusted unemployment rate. But that assumes that everyone at the higher participation rate would be unemployed. That isn’t necessarily a bad assumption if you think that the participation rate has declined entirely because people have become discouraged and dropped out of the labour market. That may well happen from time to time. But in general it is too strong an assumption. Part of the change in the participation rate must to attributed to changes in the unemployment rate and part of it to employment.
So first thing I downloaded the seasonally adjusted unemployment rate and seasonally adjusted participation rate from the ABS and graphed them.
Over that period the participation rate has slowly trended up from about 60% to about 65%. Similarly, unemployment has trended down from the early 90s with an untick during the GFC. I don’t expect unemployment to return to pre-GFC levels due to the current labour market regulations – but that is another story.
So then I do something really naughty – I run an OLS regression with the unemployment rate as the dependent variable and the participation rate as the independent variable. Naughty because that regression suffers from serial correlation and the variables themselves have unit-roots. I’m just interested in the fitted values. This month (actually November 2012) the seasonally adjusted unemployment rate was reported to be 5.2% and the participation rate was reported as being 65.1%. The fitted value for November 2012 is 5.9%. So not 6.4%, but somewhat higher than 5.2%.
Econometricians will be horrified by the regression and others can argue that the various changes to the IR regime over the period 1978 – 2012 means that there might not be a stable relationship between unemployment and participation and so on.
Update 1: GrantB asks the question: “Would a plot of both normalised to their Feb 1978 values show the trends better?” Here it is.
Update 2: Grey trolls, “If you actually had an confidence in it having any meaning at all you would show the line of the “true” unemployment rates since 1978 so we could have a good laugh.”
Not a bad story there – actual unemployment rises above participation-adjusted unemployment during recessions. Got driven very low during the Howard era and then reverts in 2008 (GFC effect? or Fair Work?).
Update 3: Seems our friend Grey can’t tell the difference between fitted values and residual values. So I’m posting a picture of the residuals.