Discussion Paper

No. 2007-48 | October 25, 2007
Global Factors, Unemployment Adjustment and the Natural Rate


OECD unemployment rates show long swings which dominate shorter business cycle components and these long swings show a range of common patterns. Using a panel of 21 OECD countries 1960-2002, we estimate the common factor that drives unemployment by the first principal component. This factor has a natural interpretation as a measure of global expected returns, which is given added plausibility by the fact that it is almost identical to the common factor driving investment shares. We estimate a model of unemployment adjustment, which allows for the influence both of the global factor and of labour market institutions and we examine whether the global factor can act as a proxy for the natural rate in a Phillips Curve. In 15 out of the 21 countries one cannot reject that the same natural rate, as a function of the global factor, appears in both the unemployment and inflation equations. In explaining both unemployment and inflation, the global factor is highly significant, suggesting that models which ignore the global dimension are likely to be deficient.


JEL Classification:

E2, J1



Cite As

Ron Smith and Gylfi Zoega (2007). Global Factors, Unemployment Adjustment and the Natural Rate. Economics Discussion Papers, No 2007-48, Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2007-48

Comments and Questions

Anonymous - Japan's unemployment
November 03, 2007 - 12:05

This is a very important contribution to the area of unemployment and inflation adjustment, showing persuasively that global factors play an important role. Theory and econometric methods are orthodox and sound.
The basic unenployment equation as shown in Appendix A2 is not significantly estimated at all only for ...[more]

... Japan. The life-long employment system was traditional there. Labour market institutional indicators of Nickel and Nunziata may not reflect this traditional institution of Japan well. However, this traditionnal system has been gradually disappearing after 1990's mainly because "global factors". It may be interesting to treat Japan's data differently.

Anonymous - Invited Reader Comment
November 13, 2007 - 10:40

see attached file

Anonymous - Specification issues
December 17, 2007 - 10:57

see attached file

Anonymous - insightful paper
January 23, 2008 - 13:36

This paper provides an extremely useful and insightful look at the impact of global factors on country-specific unemployment rates.

It would be interesting to see the full version of Table 4 (ie before the least significant coefficients are sequentially dropped).

Amongst the variables used by NNO is ...[more]

... owner-occupation rates. I would like to see this included as a variable used in (at least the initial estimates underpinning the results reported in) Table 4.

Some theoretical motivation should be given for the use of (i) prices rather than wages, and (ii) second rather than first differencing on the lhs of the Phillips curve.

The RCM Phillips curve estimate, in which u is a function of f, is supported by the significance of the coefficients on u and f in Table 5. However, f is significant in very few of the countries for which results are reported in Table A3, and indeed the point estimate of this coefficient is negative in a good many of these countries. The authors should perhaps be modest about their Phillips curve results.

Ultimately if the global factor is what determines country-specific unemployment, then there is a question about what drives the global factor. How do the equations reported in Table A2 and A3 work out if, for example, unemployment in the USA is used in place of f?

Anonymous - Referee Report
January 30, 2008 - 10:32

see attached file

Anonymous - Response to referee report posted January 30, 2008.
February 07, 2008 - 07:03

We are sorry the referee found it dense. We would be loath to drop section 4, since it establishes that the attractor for unemployment also determines inflation and the cross-equation restrictions are satisfied, thus providing evidence that it can be interpreted as a “natural rate.”

Our interpretation ...[more]

... of the common component as an expected rate of return seemed natural since the same common component drove both investment and unemployment, labour market supply side factors, such as demographic seem less likely to drive investment.

We did also experiment with using the first PC of investment but with very similar results, which is not surprising given how highly they are correlated. Although the first PC of unemployment will be correlated with the error term, the covariance between it and the error falls very rapidly with N, the number of countries used to construct the PC. With 20 countries the endogeneity bias would be very small, which is also confirmed by the fact that the results using investment are so similar.

The issue as to why the first PC loads so lightly on the US is an interesting one. It may reflect that the US unemployment pattern was very different from other countries and the pattern in Japan, Canada and Australia was much more like the European pattern than the US pattern: lower unemployment than the US in the early part of the period and higher in the later part. There are a variety of explanations for this difference, e.g. Ljunqvist and Sargent, Econometrica Jan 2008, explain it in terms of the difference in the systems of employment protection and unemployment insurance.

We have some sympathy with the Holm argument, but adjusting the size of the individual tests is not widely done. We feel more comfortable with the usual procedure.

The Swamy estimates calculate the standard errors non-parametrically from the distribution of the coefficient estimates and are robust to serial correlation or heteroscedasticity in the individual equations. There did not seem to be serial correlation in the individual equations once the lagged change in unemployment was included, but this may be a problem with the fixed effect estimates, where the standard errors are corrected for heteroskedasticity but not serial correlation.

Anonymous - Referee Report
February 13, 2008 - 08:15

See attached file

Gylfi Zoega - Response to referee
February 26, 2008 - 19:58

See pdf file attached below.