Friday, May 24, 2019

Reconsidering Okun's Law

A recent paper by Lim, Dixon and van Ours (2018) (note - the link is to a pdf of their working paper) has an interesting discussion of issues pertaining to estimating Okun’s Law relationships.  There are a couple of versions of Okun’s Law:  broadly it deals with the relation between aggregate output and the unemployment rate but it is sometimes written as a relation between changes in output and changes in the unemployment rate and other times as a relation between two gaps – the gap between actual unemployment and the natural rate, as a function of the gap between actual and potential output.  Lim et al. (2018) deal with the changes version – what must the rate of growth of real output be if unemployment is to be held constant, or to fall.

Okun’s Law dates from the 1960s, and fell into abeyance for quite a while, but there’s been something of a revival in interest in it in recent years, connected with the revival in interest in the concept of the output gap.  The gaps version is a more natural fit for that literature, but implementing it requires that we be able to estimate both the potential level of output and the natural rate of unemployment.  For practical purposes the changes version is easier to implement, but runs into the problem that the estimated relation, as judged by the coefficient on the rate of growth in real output in the unemployment equation, seems to change over time.    One objective of Lim et al. (2018) is to investigate why that might be the case.


Their basic argument rests on the fact that, while we tend to think in terms of two labour force conditions – working or unemployed – there are in fact three: working, unemployed or not in the labour force.  We start with the concept of the working age population, and define the labour force as that paer of the working age population which is either employed or actively looking for work.  Anyone who is in the working age population but is neither employed nor actively looking for work is regarded as not being in the labour force.  Clearly the line between being unemployed (not working but actively looking for work) and not being in the labour force (not working and not actively looking for work) is pretty permeable.  The unemployment rate is the proportion of the labour force which is not employed (but is actively looking for work).  So the unemployment rate can change if the number unemployed changes but also if the size of the labour force changes.  That’s why the unemployment rate frequently rises at the beginning of a recovery – more people might have jobs, but if improved job prospects prompt some others to start actively looking for work, the proportion of the labour force which is unemployed might rise.  It also relates to the argument which has been made quite frequently in the US recently, that the reason wages haven’t risen much as the recovery picked up strength was because people who had been out of the labour force were being drawn back into it, causing the labour supply curve to shift to the right along with the labour demand curve.

What Lim et al. do is look at flows among the three states – employed, unemployed and not in the labour force – and at the way that changes in the unemployment rate can be decomposed into combinations of those flows.  They show how variations over time in the estimated Okun coefficient (the one relating changes in the unemployment rate to growth in real output) could be due to a tendency for flows between (say) Unemployed and Not in the Labour Force differ depending on whether the economy is growing or contracting.  They then use a recently released US data set which treats flows among the three states consistently to estimate the three transition equations as a set, and from that set of three estimated equations they derive the coefficients of an Okun’s Law equation.

Lim et al.’s particular multi-equation approach rests on having a specific type of data set, and I’m not sure whether Canadian data would be consistent with their approach, but fortunately they show how the three transition equations can be combined with the definition of the unemployment rate to find a reduced form version of the Okun’s Law equation, which can then be estimated using current and lagged levels and changes in unemployment and real output as explanatory variables.  While that equation, if estimated by itself, wouldn’t permit us to separate out the contribution of the three flow equations to the overall relation, it should be quite possible to estimate it using Canadian data, to see whether the Lim et al. conclusions hold up and to see how Canadian unemployment responds to upswings and downturns in real output.  If the Canadian relation, like the American one, is time invariant, it would be a useful element in the process of developing a monetary policy recommendation.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.