Are job vacancies still as plentiful as they appear? Implications for the “soft landing”

One year ago, we posted an analysis on MinneapolisFed.org that was far from a consensus view: A soft landing for the U.S. economy was not only possible, but likely. Prominent economists, including Larry Summers and Olivier Blanchard, had argued otherwise; commentators such as Bloomberg Economics had recently declared a recession inevitable.

At the heart of our assessment in late 2022 was an alternative interpretation of the health of the labor market. At the time, for instance, there were 6.8 job openings for every 100 filled and unfilled jobs, despite uncommonly low unemployment. This job vacancy rate was down from a peak of 7.4 percent earlier in 2022, but still historically high.¹ Summers and Blanchard interpreted these conditions as a structural problem with the job market: Vacancies abounded because employers and potential workers were making fewer good “matches.” If the labor market had somehow “deteriorated” in this way, unemployment would likely need to sharply increase for monetary policy to bring down inflation.

We offered a simpler theory. The economy was in an unprecedented “corner” of the labor market: Employers had many jobs to fill in the aftermath of the pandemic, but unemployment was near historic lows. With fewer job seekers, many jobs were simply going unfilled. This meant the reduction in activity and labor demand necessary to reduce inflation would, most likely, not require a large increase in unemployment. Instead, job vacancies would fall as employers in a cooling economy lost their appetite for new hiring.

Support for this view came from a measure of “match efficiency” developed in prior published research.² If the labor market had truly deteriorated, hiring would be unusually weak and we would find a low level of match efficiency. In contrast, our estimates indicated that match efficiency had already returned to pre-pandemic levels by summer 2022.³

The labor market hadn’t deteriorated. As labor demand cooled and vacancies fell, we did not expect unemployment to experience large increases—and it has not. One year from our prior post, core annual inflation has fallen from 5.1 percent to 3.5 percent.⁴ The official job vacancy rate has fallen more than a full percentage point to 5.7 percent, with little change in the unemployment rate.

Another look at vacancy rates

These data look like the progress toward the soft landing we had anticipated. However, in this article we consider a new complication. As we move beyond the extreme economic disruptions of the pandemic, we can document that the relationship between job vacancies and other labor market indicators has become unstable over a period of many years.

If the vacancy rate is not sending a clear signal—or sending a different signal than in the past—predictions based on old relationships may be fundamentally wrong. With inflation still above target, this alternate view suggests that monetary policy will face more pronounced trade-offs between inflation and unemployment going forward. The remaining runway for the soft landing may be shorter than it appears in the official data.

Figure 1 shows the official U.S. job vacancy rate. It has generally declined from its high in March 2022 but remains far above any level we had experienced before 2021. Taken at face value, this observation suggests a substantial amount of cooling off in vacancies can still occur before unemployment starts increasing.

Figure 1 also plots the “quits rate,” that is, the percent of workers who voluntarily leave their job each month. Before 2010, the relationship between vacancies and quits was relatively stable. Between 2010 and 2019, however, the vacancy rate experienced a large increase relative to quits that grew even wider after the pandemic. The relationship between the two indicators has changed.

We illustrate this point by comparing vacancies with the quits rate. But we could use the hiring rate, the job-finding rate of the unemployed, or another labor market series where we don’t expect to have long-run trends. The vacancy rate is the clear exception, while other relationships between labor market indicators have remained relatively stable.

We do not take a stand here on what has caused this shift in the vacancy data. One trend to consider, however, is that digital technologies have dramatically changed the cost to employers of job posting, recruiting, and evaluating candidates. These changes, over time, might have contributed to a steady increase in measured vacancies.

Where is the labor market heading?

If the relationship between data on vacancies and other labor market indicators has become unstable, how does this affect our assessment of labor market conditions? To address this question, we compute an adjusted vacancy rate. We first find a vacancy trend using data from 2010 to 2019, purposefully leaving off the pandemic.⁵ We then use the estimated trend to adjust the vacancy data over the years 2010 to 2023, including the pandemic years. The dashed line in Figure 2 plots this adjusted vacancy rate series. Vacancies increased to a historic high during the pandemic years, but this high is muted relative to the raw vacancy data and similar to vacancy rates observed in the early 2000s.

Our earlier assessment that job-matching efficiency had not permanently deteriorated during the pandemic is largely unaffected by the adjusted vacancy data (Figure 3). If anything, match efficiency returns to 2019 levels somewhat earlier.

In contrast, implications for the “Beveridge curve“—the relationship between unemployment and job vacancies over time—are materially different. Figure 4a shows the traditional curve using the raw, unadjusted vacancy data. It suggests that vacancies are still quite high relative to unemployment. Looking at this official data suggests an ongoing reduction in vacancies might still be achieved without much of an increase in unemployment. That is, we remain on the near-vertical portion of the Beveridge curve.

In Figure 4b, however, we get a different picture when we plot the Beveridge curve with our adjusted vacancy series. In this view, the U.S. economy is already back on the “old” Beveridge curve, at the point where the curve starts to slope again (increasing unemployment).

This observation is consistent with what we’ve seen in the most recent months of data: Vacancies have continued to decline, but unemployment has ticked up modestly from 3.4 percent in April to 3.9 percent in October.

From this point, we would predict that further cooling of vacancies would move the economy down and to the right along the Beveridge curve. In contrast to our near-vertical progress since early 2022, smaller decreases in vacancies would be associated with relatively larger increases in unemployment. Given that we still need to travel the “last mile” to the Fed’s 2 percent inflation target, going forward we might expect more substantial job losses.

Endnotes

1 Monthly vacancy and unemployment data referenced in the article text may differ slightly from three-month rolling averages in figures.

2 “Aggregate Recruiting Intensity” by Alessandro Gavazza, Simon Mongey, and Giovanni L. Violante, published in 2018 in the American Economic Review. Crucially, the approach developed in this paper uses data not only on the ratio of vacant jobs to unemployed workers but also on how much hiring is going on in the economy to make inference about the health of the labor market.

3 This finding was robust to whether the pool of job seekers included employed workers searching for a new job, or those out of the labor force potentially finding a job opportunity and skipping a spell of unemployment.

4 Personal Consumption Expenditures Price Index, Excluding Food and Energy (“core PCE price index”) November 2022 to October 2023 (latest as of this writing).

5 Specifically, we proceed in five steps: (1) We fit a constant elasticity relationship between the (log of the) quits rate and the (log of the) vacancy rate over 2001 to 2010, (2) We compute a counterfactual vacancy rate over 2011 to 2019 that adheres to this relationship, (3) We compute the gap between the vacancy rate in the data and this counterfactual vacancy rate, (4) We fit a linear trend to this deviation, and (5) We subtract this linear trend from the 2019 to 2023 (September) data. The results we describe are remarkably robust to the period over which we fit the relationship in Step 1, the period over which we fit the trend in Step 2, or whether we use other data in place of the quits rate (for example, the gross hires rate).