Related pages

The theory of the Beveridge curve is in pretty good shape. (The best discussion is by Olivier Blanchard and Peter Diamond in "The Beveridge Curve", Brookings Papers on Economic Activity, 1989, No. 1, pp. 1-60 and 74-76; they have a handful of other papers, including a brief introduction in "The Flow Approach to Labour Markets", American Economic Review (Papers and Proceedings), vol. 82, May 1992, pp. 354-9.) The real problem is empirical. The very concept of a count of job vacancies is vague around the edges, and the same can be said of a count of unemployed workers. Employers can be more or less serious about filling a job, just as people can be more or less serious about finding a job. But national unemployment counts exist almost everywhere, whereas vacancy statistics are quite rare.

In the United States, official vacancy statistics were collected only for a brief interval in the 1960s. Instead the custom is to use a privately-collected time series of the volume of help-wanted advertising in newspapers. This is obviously an imperfect surrogate; for example, a vacancy may be advertised several times or not at all, and in any case the role of the newspaper as an advertising medium has been changing. But it is a lot better than nothing. In Europe the availability of vacancy statistics differs from country to country, and there are occasional changes in definitions and methods. But there is enough to get on with.

The sort of labour-market model encouraged by the use of the Beveridge curve allows one to talk of labour-market flexibility as distinct from simple wage-flexibility. I have taken that opportunity for a couple of reasons. Those who talk about the need for more flexibility in European labour markets are presumably not just asking for more wage-flexibility; if they were, they could say so directly. Besides, the macroeconomics of wages is a very long story, going back at least to Pigou's Theory of Unemployment; there is nothing to be gained by bringing it up here. There is, however, one relevant empirical point that I would like to underline.

There was a time, in the early and middle 1980s, when the "wage-gap hypothesis" was a leading candidate explanation for what was even then seen as unusually high unemployment for Europe. (The main reference is The Economics of Worldwide Stagflation, 1985, by Michael Bruno and Jeffrey Sachs.) The hypothesis was that real wages in Europe had outrun labour productivity. Among the consequences were low profitability, low investment and a lot of unemployment. It is easy to see how this could have happened. Real wages typically move with labour productivity. An unexpected productivity slowdown began some time around 1970 (and continues still, though no longer unexpected); slow adaptation to this change could account for the opening of a wage-gap if real wages continued to reflect inertia induced by older expectations. One common extension was the idea that in Europe real wages were sticky; in the U.S. nominal wages were sticky so the real wage could be "inflated down."

To say that the real wage has outrun the productivity of labour is to say that the share of wages in aggregate output has risen; and in fact the profit share in the major Continental economies was unusually low from about 1975 to the early 1980s.

Beginning in the early 1980s, however, there was a remarkable distributional shift to profits. The wage share in Europe began to fall, and may not yet have stopped falling. By now the wage share on the Continent is substantially lower than in North America. The wage-gap has disappeared, more than disappeared so to say, but the unemployment lingers on. The significance of this fact is that one can not build a really convincing story about current unemployment that rests primarily on wage-rigidity that holds the economy-wide real wage at too high a level.

It is worth a reminder that the unemployment rate in the U.K. climbed to very high levels in the years 1981-1987 and has since receded, not to where it was in the 1970s but to a figure substantially lower than in France and Germany. During this period the profit share fluctuated around an essentially horizontal trend; here too the profit share has been rising in the 1990s, but it is still ever so slightly lower than in the mid-1980s. There is no particular comfort for the wage-gap story here either.

The Beveridge curves provide a somewhat heterodox perspective on the role of labour-market rigidities. (Here and elsewhere I am deeply indebted to Professor James Medoff of Harvard for his pioneering empirical work and his generosity in helping out with data and analysis. The data plotted in the European Beveridge-curve diagrams are all extracted from official sources. The figures are not really comparable from country to country, not even the unemployment rates. They are useful primarily for comparisons over time for each country separately.) I begin with the U.S. because it is the natural benchmark for comparison with France, Germany and the U.K.

Figure 1 plots the "vacancy rate" vertically and the unemployment rate horizontally, on an annual basis. Remember that the "vacancy rate" is really an index of help-wanted advertising normalized by the labour force. It would be possible to doctor the data: the unemployment rate could be corrected for demographic changes, as George Perry suggested long ago; and the help-wanted index could be keyed to the brief period when vacancies were actually measured, and could be adjusted for the change in the importance of newspapers as an advertising medium, as Katharine Abraham showed a decade ago. I have omitted such refinements because it is only the qualitative picture that matters for now.

That qualitative picture stands out clearly. From 1958 through 1971, the U.S. seemed to move along a well-defined Beveridge curve. During 1972-1974 the curve shifted adversely, and settled for 1975-1986 about three percentage points of unemployment to the right of its earlier location. Then, in 1987 and 1988, the curve seemed to return to its initial position, and has remained there for the past decade. One can interpret this as saying that the U.S. experienced an episode of acute labour-market rigidity between the early 1970s and the early 1980s, and has now reverted to form. (Blanchard and Diamond, in the 1989 paper already cited, produce a monthly Beveridge curve for a different period, 1952-1988, using slightly different data. The general evolution is entirely consistent with what I have just described.) Does this make sense?

I can invent a libretto to go along with that melody. The story line includes the pronounced productivity slowdown, leading first to an unrealistic reservation wage, and then to an eventual adjustment to reality. One could also make something of increasing segmentation of the labour market as older manufacturing industries decayed, and the economic structure shifted in favor of the service sector and the Sunbelt. I called this a libretto precisely to underline the lightweight character of the exercise. One can always invent a plausible story to cover a single episode; in this case the episode lasted for 40 years, at least, and had three acts. It is worth noting, however, that the sorts of scenery emphasized in the usual version of the European opera do not seem to have been on stage in the U.S.

The picture in the U.K., as shown in Figure 2, is more complicated. Perhaps bemused by what happened in the U.S., I am inclined to push my luck and suggest an analogous, though not quite similar, evolution here. A determined reader of tea leaves could certainly see an initial Beveridge curve for the years 1964-1972. Beginning in 1973, during a period of mostly rising unemployment, the whole curve seems to migrate to the right, settling down from 1983 to 1987, and then moving leftward again to what looks like a stable position — at least temporarily — in the 1990s. The initial rightward shift spans almost nine percentage points of unemployment at the extreme; and the reversion to the left takes about four percentage points back. A vacancy rate that would have corresponded to three percent unemployment in the 1960s is accompanied by roughly eight percent unemployment in the 1990s. This is obviously a much bigger change than can be inferred in the U.S.

There is, however, an underlying similarity in timing. In both cases the adverse shift of the Beveridge curve begins around 1972 or 1973. The temptation is strong to identify it in the U.K., as in the U.S., with the slow adjustment of wages to the productivity slowdown that began in those years. (It might once have been thought that the first OPEC oil shock was the source of the maladjustment of wages, or even of the productivity slowdown itself. But that idea has lost whatever plausibility it ever had, if only because the later fall in the real price of oil had no corresponding effect. In any case, the rightward shift of the Beveridge curve seems to have begun a bit too early to be explained in that way.)

The leftward migration of the Beveridge curve also dates from 1987 in both countries. That coincidence might offer a hint as to the underlying cause. But I would prefer to leave that inference to others who know more than I do about the timing of institutional, political and other changes in the U.K. and U.S. labour markets that might account for the stories told by the two Beveridge-curve diagrams. Apart from these similarities in timing, there are drastic differences. The two most noticeable are, first, that the adverse shift in the U.K. was larger and more drawn out in time than the corresponding shift in the U.S., and, second, that the U.K. Beveridge curve has reverted only about halfway back to its initial favourable location, whereas in the U.S. the 1960s and the 1990s seem to look alike. It will take a knowledgeable combination of formal analysis and local anecdote to account for those differences. My immediate interest lies elsewhere, and especially in the contrast with the corresponding developments in France and Germany.

For that we can look at the third and fourth Beveridge-curve diagrams (Figure 3; Figure 4), which are in fact very much like each other and very different from the preceding ones for the U.S. and the U.K. In both France and Germany there is a suggestion of a vertical portion of the curve at the extreme left. This is what one would expect to see if there were a minimal level of frictional unemployment necessary for the labour market to function at all; it would reflect entry and exit from the labour force, turnover from one job to another, and so on. The diagrams make it look as if that minimal unemployment rate were just under 3 per cent in France, achieved in the late 1960s and early 1970s, and 1 percent in Germany, achieved at exactly the same time.

Then the picture gets more interesting. Something may have happened beginning in 1975, in both countries. But the simple configuration of the data allows two interpretations. One is that there was a small rightward shift of the Beveridge curve in both countries, amounting to about one percentage point of unemployment in France and fractionally more in Germany. The other is that there was no shift at all, and the whole 28-year period traces out a single, more or less stable, Beveridge curve. In practice, this is a distinction without a difference, because the adverse shift, if there was one, was so small.

In the case of France, moreover, the years 1964-1969 are anomalous. The eye could make a case that a significant shift separates the years before and after 1970. But that would seem to have little to do with the period of endemic high unemployment in the 1980s and 1990s.

Unfortunately the interval described in these graphs ends in 1991, because the later data are for various reasons incomparable with the earlier observations. So we can not look at the 1990s through this particular lens. However the small reductions in unemployment that took place in France between 1986 and 1990 and in Germany between 1983 and 1991 do seem to be traversing much the same Beveridge curve as was traced out in the opposite direction in France between 1980 and 1986 and in Germany between 1983 and 1991. In saying this I am taking account of the normal presumption that evolving data would trace out counter- clockwise loops around the curve representing stationary equilibrium positions.

I have said that these observations are open to slightly different interpretations. But I also have to claim that the main message transmitted by the Beveridge curves transcends these alternatives. That message goes squarely against the cliché that high and persistent European unemployment is entirely or mainly a matter of "labour-market rigidities." It is precisely in France and Germany, where unemployment has been higher and more persistent, that there is no sign of a big adverse shift in the Beveridge curve. It is precisely in the U.S. and the U.K., where unemployment has been at least more variable and, in the case of the U.S., lower, that one can detect a substantial adverse shift, followed by a favourable one.

To the extent that the location of the Beveridge curve is a reasonable summary for the degree of labour-market rigidity, the large continental economies do not seem to have suffered from noticeably more rigid labour markets during the high-unemployment 1980s than they did in the low-unemployment 1970s. In fact what stands out from the pictures for France and Germany is the depressed level of the vacancy variable. It is a pity not to have comparable data for the last five years.

In the case of Germany, where the data now include the Ostländer, one can at least say that there is no indication of a rebound in vacancies. The case of France is even less clear. There is a new series of "new job vacancies" and it has risen smartly since 1991; but this sounds like a measure of "job creation" and it is impossible to interpret it in isolation from information about job destruction and pre-existing vacancies. It is a reasonable judgment that the major difference between France and Germany now and in the early 1970s is that the demand for labour is now much weaker. It is not reasonable to blame that large increase in unemployment on worsened labour-market rigidity.

|  MORE>  |  Return to title page  |  Publications HomePage  |  Academy HomePage  |