« Intelligently Designed Clues Point to Evolution | Main | Are Unemployment and Capacity Utilization Measures Irrelevant for Monetary Policy? »

Saturday, April 22, 2006

Revisiting the Beveridge Curve

This Economic Letter from the San Francisco Fed uses the Beveridge curve relating vacancy and unemployment rates to examine how the efficiency of job matching has changed through time. The research finds that "...the speed and effectiveness of the job-matching process deteriorated in the 1970s through the early 1980s and then improved" with the onset of the great moderation in the mid-1980s due, perhaps, to a decline in the geographic mismatch of jobs and workers1:

Job Matching: Evidence from the Beveridge Curve, by Rob Valletta and Jaclyn Hodges, Economic Letter, San Francisco Fed: Conditions in labor markets are largely reflected in the number of jobs employers want to fill (job vacancies) and the number of people seeking jobs (the unemployed). Over the business cycle, for example, job vacancy rates and unemployment rates generally exhibit negative co-movement, with high vacancies and low unemployment when the economy is growing and vice versa when the economy is contracting. Beyond that short-run relationship, ... positive co-movement of the vacancy and unemployment series over longer time periods reflects changes in the speed and effectiveness of job matching.... When the job-matching process is slow, perhaps due to changes in the amount of necessary job reallocation across geographic regions or industries, both unemployment and vacancies can coexist at high levels, representing underutilized labor resources.

This Economic Letter examines the evidence on long-term shifts in the speed and efficiency of job matching in U.S. labor markets by using the so-called Beveridge curve. The Beveridge curve is an empirical measure of the relationship between the job vacancy rate and the unemployment rate. Changes in the job-matching process suggested by movements in the Beveridge curve ... have not been extensively analyzed... In this Economic Letter, we utilize new data from the U.S. Bureau of Labor Statistics (BLS) to construct a long-term vacancy series and corresponding estimates of the Beveridge curve. We find that declining dispersion of economic growth across geographic regions helps explain improvements in the job-matching process since the mid-1980s and reinforces existing depictions of improved performance of the U.S. aggregate labor market in the 1990s (for example, Katz and Krueger 1999).

Forming the empirical Beveridge curve

...Figure 1 displays the U.S. Beveridge curve based on the ... vacancy rate series and age-adjusted unemployment rate series for selected periods between 1960 and 2005.

The Beveridge curves in this figure exhibit a typical counterclockwise adjustment pattern around recessions (1960-61, 1981-82, and 2001), as vacancies rise more quickly than unemployment falls during the recovery phase. The outward shift in the Beveridge curve between the periods 1960-69 and 1979-85, as identified by Abraham (1987), is clearly evident, as is a substantial inward shift between 1979-85 and 2000-05. This pattern suggests that the speed and effectiveness of the job-matching process deteriorated in the 1970s through the early 1980s and then improved.

Regional mismatch

One leading explanation for these movements in the Beveridge curve is changes in the dispersion of employment growth across regions.

If labor demand is growing in some parts of the country and shrinking in others, a "regional mismatch" can occur, whereby large numbers of unemployed individuals must move across geographic regions in order to be matched with available jobs. The need for such costly and time-consuming geographic reallocation slows down the job-matching process and increases the likelihood that unemployment and vacancies will both exist at high levels. Indeed, Abraham found that rising regional mismatch accounted for a large share of the outward shift in the Beveridge curve between the 1960s and early 1980s (as depicted in Figure 1).

As a graphical illustration of changes in the degree of regional mismatch, Figure 2 displays Beveridge curves for two of the nine census divisions: New England and the Pacific. Their Beveridge curves were far apart during the years 1979-85, reflecting substantial geographic mismatch in the strength of labor demand. Since then, the regional curves have largely converged, indicating a decline in mismatch. While we only display two regions in Figure 2, this pattern of convergence in labor demand and supply conditions is evident across U.S. regions and states more generally.

Figure 3 shows the overall degree to which patterns of growth in labor demand have converged across regions; it displays the dispersion (standard deviation) of yearly employment growth across census divisions. This series plays a key role in explaining the outward and inward movements of the U.S. Beveridge curve in recent decades. The dispersion of employment growth increased substantially between the late 1960s through about 1980, when the Beveridge curve shifted out, and then declined by an even greater amount since the early 1980s, when the Beveridge curve shifted back in.

On net, the magnitude of the Beveridge curve shifts and changes in regional growth dispersion are quite consistent over the period 1970-2005. Abraham (1987) found that about 1.4 percentage points of the 1.8 percentage point increase in the unemployment rate associated with Beveridge curve shifts between 1970 and 1985 was due to rising geographic dispersion of employment growth. By contrast, we find that geographic growth dispersion fell below its 1960s levels by the early 2000s, reversing the pattern identified by Abraham. The result was a 1.5 to 2 percentage point decline in the unemployment rate associated with Beveridge curve shifts between the mid-1980s and early 2000s, producing a Beveridge curve that is interior to any observed since the 1960s.


The adjusted vacancy and unemployment data that we use indicate that a pronounced inward shift in the position of the Beveridge curve has been evident since the mid-1980s, reversing the earlier pattern identified by Abraham (1987) and implying reduced job reallocation or increased efficiency for the job-matching process in the United States. Our analyses of regional Beveridge curves and the dispersion of labor demand across regions suggest that a decline in necessary job reallocation was responsible for this shift. In particular, the process identified by Abraham (1987), whereby increasing dispersion of labor demand growth across geographic areas caused the Beveridge curve to shift out, has been reversed, causing the Beveridge curve to shift back in.

More generally, our finding of inward shifts in the Beveridge curve over the past two decades reinforce Katz and Krueger's (1999) conclusions regarding improved U.S. labor market performance in the 1990s. The inward shift in the Beveridge curve may underlie the more favorable tradeoff between unemployment and wages that has been estimated for the 1990s. Moreover, we find that these favorable trends continued into 2005. These findings suggest that the Beveridge curve may merit renewed attention by researchers.

1There is a discussion in the paper about the problems with help wanted indexes and how the authors constructed of a consistent measure of the job turnover rate used in the analysis, an impediment to work in this area in the past, that is not included here.

    Posted by on Saturday, April 22, 2006 at 02:34 AM in Economics, Methodology, Unemployment | Permalink  TrackBack (0)  Comments (4)


    TrackBack URL for this entry:

    Listed below are links to weblogs that reference Revisiting the Beveridge Curve:


    Feed You can follow this conversation by subscribing to the comment feed for this post.