VoxEU Column Labour Markets

It’s not a skill mismatch: Disaggregate evidence on the US unemployment-vacancy relationship

US unemployment seems stuck at an unusually high level of 8%, prompting some to suggest a widespread skills mismatch. This column argues that a skills mismatch is not supported by the evidence. Rather, out of the possible explanations, it seems that any shift in the ratio between unemployment and vacancies is driven by either lower search efforts by the long-term unemployed or by a reduction in their employability.

The Beveridge curve – the empirical relationship between unemployment and vacancies – is thought to be an indicator of the efficiency of the functioning of the labour market. Normally when vacancies rise, unemployment falls following a curved path that typically remains stable over long periods of time. When vacancies rise and unemployment does not fall (or falls too slowly) this may be an indication of problems of structural mismatch in the labour market leading to an increase in the lowest unemployment rate that can be maintained without increasing inflation (the Non-Accelerating Inflation Rate of Unemployment (NAIRU). Key contributions to this strand of work were progressively made by Dow and Dicks‐Mireaux (1958), Lipsey (1960), Holt and David (1966), Hansen (1970), and Bowden (1980).

More jobs, more jobless

The unemployment-vacancy relationship has received much attention among economists and policymakers over the past few years. Since the end of the Great Recession, we started seeing the Beveridge curve shifting out toward the upper right, reflecting a decrease in labour market efficiency. The outward shift means that firms can’t fill their available job openings as readily as we would have expected in light of the high unemployment rate (Kocherlakota 2010).

Controversial interpretations of the data

The basic fact that recent ‘job vacancy, unemployment’ points lie outside the locus of points that seemed to define the Beveridge curve in the 2000s is not in dispute, but its interpretation has been controversial. Interpretations of the recent data range from a temporary cycling around a stable Beveridge curve due to the prolonged slow recovery from the Great Recession to a quasi-permanent shift of the Beveridge curve due to pervasive mismatch between the qualifications of job applicants demanded by employers and the qualifications offered by unemployed job searchers.

Figure 1 plots vacancy-unemployment points from 2001 on. The curved relationship between unemployment and vacancies depicted in Figure 1 is often called the Beveridge curve. Over the recession of the early 2000s, and the recovery from that recession, the vacancy-unemployment relationship remained remarkably constant, as it has for long periods of time in the past. However, in the recovery from the most recent recession we see that vacancies have grown considerably without producing the normal decline in unemployment. It looks as if the Beveridge curve may have shifted out. The figure displays an empirical relationship combining data on vacancies from the Job Openings and Labour Turnover Survey (JOLTS) with the aggregate unemployment rate obtained from the Bureau of Labor Statistics’ monthly household survey. The data span the period January 2001 through June 2012 and are seasonally adjusted. The solid line in Figure 1 reflects a stylised Beveridge curve that was estimated using data on unemployment and vacancy rates for the period prior to the start of the recession. The plot reveals that by September 2009, the vacancy-unemployment points started to deviate away from the fitted curve in an anti-clockwise direction indicating a higher unemployment rate at any given level of job openings.

Figure 1. Total vacancies and unemployment rates by unemployment duration

Source: CPS and JOLTS. Data are monthly rates, span the period 2001m1-2012m6, and are seasonally adjusted.

The seasonally adjusted BLS series of monthly unemployment rates for all employees, 16 years and over is disaggregated to examine the vacancy-unemployment relationship at different durations of unemployment. We do this to see if the unemployed with different durations benefit differently from the recent increase in the vacancy rate. We use data from the BLS’s Job Openings and Labour Turnover Survey (JOLTS) for the aggregate vacancy rate and plot that against the fraction of the labour force unemployed at different durations. Figure 2 presents that relationship for those unemployed for less than five weeks.

Figure 2. Monthly vacancy and unemployment rates, using unemployed persons with duration less than five weeks

Source: CPS and JOLTS. Data are monthly rates, span the period 2001m1-2012m6, and are seasonally adjusted.

The relationship between the vacancy rate and the very short run (less than five weeks) unemployment rate is essentially vertical: around 2% of the labour force is in the first five weeks of a spell of unemployment regardless of the level of job vacancies. There does not appear to have been any change in the relationship in 2009. Hence, individuals with unemployment spells less than five weeks do not explain what we see in the aggregate plot.

Figures 3 and 4 illustrate the vacancy-unemployment relationships using unemployed persons with duration 5-14 weeks and 15 -26 weeks respectively.

Figure 3. Monthly vacancy and unemployment rates using unemployed persons with duration of 5-14 weeks

Source: CPS and JOLTS. Data are monthly rates, span the period 2001m1-2012m6, and are seasonally adjusted.

Figure 4. Monthly vacancy and unemployment rates using unemployed persons with duration of 15-26 weeks

Source: CPS and JOLTS. Data are monthly rates, span the period 2001m1-2012m6, and are seasonally adjusted.

Figure 5 combines all those unemployed for less than 27 weeks and compares their fraction of the labour force to the aggregate vacancy rate.

Figure 5. Monthly vacancy and unemployment rates using unemployed persons with duration less than 27 weeks

Source: CPS and JOLTS. Data are monthly rates, span the period 2001m1-2012m6, and are seasonally adjusted.

Evidence of an outward shift

As with the graphs for each of these durations individually, there is no evidence of an outward shift. However, when the relationship is plotted using the fraction of the labour force that has been unemployed for more than 26 weeks, a number of interesting features are immediately apparent. First, the pattern in Figure 6 reveals a counter-clockwise outward shift that is consistent with what we see when we use the aggregate unemployment rate.

Figure 6. Monthly vacancy and unemployment rates using unemployed persons with duration greater than or equal to 27 weeks

Source: CPS and JOLTS. Data are monthly rates, span the period 2001m1-2012m6, and are seasonally adjusted.

In addition to the shift, the pattern in Figure 7 shows that the vacancy and unemployment points for the long-term unemployment group starts shifting out at the same time the aggregate vacancy-unemployment relationship breaks down.

Implications of an outward shift of the Beveridge curve

Disaggregating the vacancy-unemployment relationship reveals some interesting new facts that may shed light on the implications of what appears to be an outward shift of the Beveridge curve in recent years. While the Beveridge curve for all workers appears to be shifting out starting in 2009, data on vacancy and unemployment rates for individuals who have been unemployed for less than 27 weeks reveals the usual downward sloping relationship with no sign of any outward shift. Interestingly, a dynamic plot of the vacancy versus short-term unemployment rates shows clockwise cycling of the vacancy-unemployment points for those unemployed. In contrast, we see a large anti-clockwise movement when the vacancy rate is plotted versus the unemployment rate for those unemployed for more than 26 weeks. Taken together this suggests that the short-term unemployed are benefiting more than the long-term unemployed from increases in vacancies during the recovery.

Other than the contrast between the long- and short-term unemployed, we break down the vacancy-unemployment relationship by industry, education levels, age groups, and among both blue and white collar workers. In these decompositions we notice a similar pattern to what we see in Figure 1: there appears to be a breakdown in the vacancy unemployment relationship sometime at the trough of the recession.

We conclude that any explanation for the change in the vacancy unemployment relationship must account for its pervasiveness across different industry, blue-white collar occupations, age, and education groups, its concentration among the long-term unemployed and its absence from the short-term relationship.

  • One reason that the Beveridge curve relationship for the long-term unemployed shifted may be a shift in the desirability of the long-term unemployed to employers.

It is possible that the long-term unemployed are increasingly made up of workers whose skills are not suited to available jobs. However, if this were the case why wouldn’t we see some outward shift in the short-term relationship as well? Furthermore, the fact that the vacancy-unemployment relationship has shifted in all industries when only the workers who were previously employed in those industries are considered calls the mismatch hypothesis into question as well.

  • Another possibility is that the long-term unemployed in this recession may be searching less intensively, either because jobs are much harder to find or because of the availability of unprecedented amounts and durations of unemployment benefits.

This seems like a more likely explanation, though if a drop in search intensity is due only to difficulty finding jobs it again raises the question of why we wouldn’t see that at shorter durations as well.

References

Ghayad, Rand and William Dickens (2013),”What Can We Learn by Disaggregating the Unemployment-Vacancy Relationship?”, Prospective article, Federal Reserve Bank of Boston, Working Paper.

Bowden, R (1980), “On the existence and secular stability of the u‐v loci”, Economica, 47, 35–50.

Dow, J and Dicks‐Mireaux, L (1958), “The excess demand for labour. a study of conditions in Great Britain, 1946–56”, Oxford Economic Papers, 10, 1–33.

Hansen, B (1970), “Excess demand, unemployment, vacancies and wages”, Quarterly Journal of Economics, 84, 1–23.

Holt, C and David, M (1966), “The concept of vacancies in a dynamic theory of the labour market”, in ibid. (eds.) Measurement and Interpretation of Job Vacancies, NBER, New York, Columbia University Press.

Kocherlakota, Narayana (2010), “Inside the FOMC”, Speech delivered in Marquette, Michigan., 17 August, Federal Reserve Bank of Minneapolis.

Lipsey, R (1960), “The relation between unemployment and the rate of change of money wage rates in the United Kingdom, 1862–1957: a further analysis”, Economica, 27, 1–31.

3,150 Reads