Why do unemployment rates vary?

The other week the report Lönebildning bortom NAIRU (“Wage setting beyond the NAIRU”) was published by The Swedish Trade Union Confederation, LO, with a lot of interesting stuff written by Tony Johansson, Lennart Erixon, Servaas Storm, C.W.M. Naastepad, Marc Lavoie, Eckhard Hein and Josef Taalbi (report and seminar, both in Swedish, here and here).

I contribute with a chapter, which is mostly econometrical results on the correlation between the unemployment rate, labor market institutions, capital stock, and some other macro variables. My data covers at most 37 countries, for the years 1960-2013. My method is similiar to, among others, the one in Stockhammer and Klär (2010), but my analysis is a bit shorter and I have a somewhat different dataset.

All chapters in the report discusses which factors that might affect the equilibrium unemployment in the long run (NAIRU). The theories that have dominated both Economics and policy for the last decades, in most OECD countries, argue that supply-side factors are the only thing that is important for the unemployment rate in the long run, why econometric results should indicate a relatively distinct correlation between unemployment and unemployment benefits etcetera. However, I find the following two things among my econometric results (which is similar to at least some earlier studies), interesting:

  1. None of the factors that is usually described as important for the labor market supply-side, such as unemployment benefits, union density, etc., correlate in any statistically significant robust manner with unemployment. I have also compared other measures in different combination – none of these factors seems to be of importance, see below.
  2. Investments correlate with unemployment: when capital accumulation is larger, unemployment tends to be lower. In at least 34 of the 37 countries included in the study, this covariance may described as distinct. Overall, these econometric results seems to be stable in different econometric models, combined with different control variables etc. To measure investment I mainly use AMECOs data on net capital stock, per cent rate of change per year (i.e. capital accumulation / capital formation).

One interpretation of result (1) is that the connection between the unemployment rate, and the conditions for the employees and the unemployed is much more limited, or at least much more complex, than is so often claimed. One interpretation of (2) is that aggregate demand may affect the NAIRU, at least over the medium term. If I’ve put together my econometric models correctly, capital accumulation is even of importance in the longer run – contrary to dominant theories.

The correlation between the unemployment rate and capital accumulation (graph a) and capital formation as per cent of GDP (b) for Sweden 1960-2013


The most problematic aspect of my results regarding the capital stock, I think is that the size of the coefficient is smaller in some of the econometric models where I analyzed annual values (which can be seen in one of the results below). There are several ways to measure the capital stock and capital formation. Also, there is difficult to create long time series over several countries, that are correct and fully comparable, so the results should probably be interpreted cautiously regardless.

What does this prove?

I use various methods to try to measure as much of the long run relationship as possible (the long run NAIRU): five-year averages with one value per five-year period; inflation change; a measure of the output gap; and lagged values ​​of unemployment. But it is difficult to know to what extent I really manage to catch the reeeally long run.

These results do not mean that unemployment benefits, or other institutional factors, are unimportant for how the labor market functions. Of course, people are affected by labor market conditions, rules and regulations etc. However, the relationship may be more complex than is often claimed. In that case, it may indicate that the dominant NAIRU theory is inadequate. In fact, it might suggest that their is no long run NAIRU, which is structurally stable, in the sense that dominant theoretical framework suggest (see for instance Storm and Naastepads excellent book). Either way, capital stock and formation seems to deserve more attention when discussing unemployment. And unemployment benefits might deserve a bit different attention.

Results Table. All models are fixed effect (equivalent to dummie variables for countries and time). Results without stars = not statistically significant. See translation and explanation below. More econometrical results in this pdf.ta

Model 1-5 = five-years moving averages, one value per five-year period
Model 6-9 = annual values

NRR = unemployment benefits
BD = benefit duration
EP = employment protection
COV = per cent of the work force covered by collective aggrements
AUTH = trade union authority (from Visser, 2006)
TW = tax wedge
REGREF = regulation index for a selection of the largest industries
ACCU = capital accumulation (per cent change per year of the net capital stock)
ILRV = long term real interest rate
GAP = GDP gap, created by hp filter
INFL1DIFF = inflation first difference
UT5 = Unemployment , t-5. (First lag in model 1-5)
UT1 and UT2 = Unemployment, first and second lag in model 6-9

Antal observationer = no. of observations.
Justerade R2 = adjusted R2


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s