Skip to Main Content
Speaking of the Economy
A man in an apron holds a "Hiring" sign in a shop window
Speaking of the Economy
June 22, 2022

When Will Labor Markets Return to "Normal?"

Audiences: Economists, General Public, Policymakers

Andreas Hornstein and Thomas Lubik discuss their research on changes in the labor market since the pandemic. Hornstein and Lubik are senior advisors at the Richmond Fed.



Tim Sablik: Hello, and welcome to Speaking of the Economy. I'm your host, Tim Sablik, a senior economics writer at the Richmond Fed. My guests today are Andreas Hornstein and Thomas Lubik. Andreas and Thomas are both senior advisors in the Research department of the Richmond Fed and they join me today to talk about some of their recent work trying to understand the pandemic's effects on the labor market.

Andreas and Thomas, thanks for being here.

Andreas Hornstein and Thomas Lubik: Thanks for having us.

Sablik: The labor market has gone through some extraordinary swings since the pandemic hit the U.S. in March 2020. The unemployment rate soared to a record high of 14.7 percent in April 2020, largely driven by temporary layoffs and shutdowns. Since then, the unemployment rate has fallen rapidly to 3.6 percent.

Labor market dynamics is something that the researchers here at Richmond and all the Federal Reserve Banks follow very closely, since one of the Fed's mandates is to achieve maximum employment. Andreas, what measures can Fed policymakers use to determine whether the economy has reached that point?

Hornstein: As you just mentioned, attainment of maximum employment is one of the two main mandates, but the Federal Reserve is tasked with price stability. Of these two mandates, maximum employment is probably the most difficult one to define. For the longest time, the Fed did not put numbers to either of the two mandates.

Only about 10 years ago, the monetary policy committee of the Federal Reserve, the FOMC, adopted a Statement on Longer-Run Goals and Monetary Policy Strategy that defined price stability as a 2 percent inflation target. But it refrained from being specific about what constitutes maximum employment, mainly because it asserted that whatever maximum employment is, it changes over time owing largely to non-monetary factors.

That being said, what are some of the measures the Fed is looking at when it evaluates the labor market? The main measures are derived from monthly survey of representative sample of U.S. households, the Current Population Survey. From an employment perspective, people in the survey are classified into three groups: those who are employed, those who are not employed but are looking for work (that is the unemployed), and finally those who are neither employed nor unemployed. So, the unemployed and the employed constitute the labor force.

From these concepts we derive three statistics: first, the unemployment rate, that is the share of the unemployed in the labor force; second, the labor force participation rate, that is the labor force as a share of the population; and third, the employment-population ratio, that is the employed as a share of the population. From the latest data, the unemployment rate is 3.6 percent, the labor force participation rate is at 62 percent, and the employment population ratio is at about 60 percent.

So, how do these measures relate to maximum employment? The most frequently mentioned indicator of maximum employment is probably the unemployment rate. If the share of people that want to work but cannot find work is high, we are probably not at maximum employment. However, given that it takes time for workers and employers to find mutually beneficial job matches, it's unlikely that zero unemployment is a feasible outcome, or for that matter constitutes maximum employment.

In this context, people have been talking about the natural rate of unemployment as representing the level of unemployment that can be maintained absent shocks. Every other meeting of the FOMC, its members submit projections on how they expect the economy to perform over the next few years, and that includes a statement about the long-run unemployment rate. You can think of that as FOMC members estimates of a natural rate of unemployment.

The latest median projection for the long-run unemployment rate as of March is 4 percent. That is, about half of the FOMC members believe that over the longer run, the sustainable unemployment rate is higher than the current unemployment rate.

Given the low unemployment rate, policy discussion has evolved and people are now also considering the labor force participation rate. Is it unusually low, and can further sustainable increases in employment come from higher participation rates? I don't believe there is any consensus on that yet.

Sablik: Thanks for that overview.

As you mentioned, the unemployment rate changes during downturns and recoveries. During a downturn, we'd expect it to be higher. During a recovery, we'd expect it to be lower. How about labor force participation and the employment to population ratio? How do those change during downturns and recoveries?

Hornstein: The unemployment rate is most cyclical one, you're right. You can take the position that the unemployment rate more or less defines the business cycle.

For the labor force participation rate, it is less cyclical. If you look at the graph of labor force participation rate over time, you see long swings. You get an increase until 2000 and then a decrease since. There is limited variation in the labor force participation rate over business cycles. To the extent that there is cyclicality, the labor force tends to decline when the unemployment rate increases, but usually with a lag of maybe about two years.

With respect to the employment rate, it's kind of a mix of the two. You have fluctuations in the unemployment rate overlaid over the trend in the labor force participation rate.

Sablik: Okay, so those tend to be driven by longer term factors.

In a recent Economic Brief that you wrote with Marianna Kudlyak of the San Francisco Fed, you estimated the trend of each of these metrics over time to get a sense of what a normal labor market might look like at any point. Based on these estimates, do you have any sense of how we might have expected the labor market to evolve if the pandemic hadn't happened?

Hornstein: We tried to get at the demographic drivers of the trends in the labor force participation rate and the unemployment rate. These drivers are basically the age, gender, and education composition of a population, not only the composition but also how the unemployment rate and participation rate changes over time within these groups. … There are systematic differences across groups. More educated people tend to have higher labor force participation rates and lower unemployment rates. And, the relative size of the population groups has been changing over time. The U.S. population has become more educated but also older.

What you find is that these demographic changes can account for a substantial part of the decline in the trend of the unemployment rate and labor force participation rate. Prior to the pandemic, if you estimate the trends using data up to 2019, we would have concluded that after this long decline from the Great Recession in 2008, finally the labor force participation rate started exceeding its trend in 2019. Going forward, the population dynamics still point to further declines in the participation rate. As for the unemployment rate, we estimated it to be substantially below trend in 2019.

Sablik: Okay. When you include the data after the pandemic, how does the picture look now? Where does the labor market appear to be in relation to what normal trends would be?

Hornstein: The pandemic has changed the level of our trend estimates quite a bit, especially for the participation rate. As I mentioned before, in the past the participation rate has not been very cyclical and, to extent that it was cyclical, it was lagging the unemployment rate. But during the pandemic, the labor force participation dropped sharply, coincident with the increase in the unemployment rate. Since the participation rate is recovering in 2021, our estimation method takes a lot of signal from that behavior and shifts down the path of the trend estimate of the participation rate relative to our previous estimates. But going forward, we still expect a continued decline in the trend of the participation rate.

Following the pandemic, we also revised our trend estimate of the unemployment rate upwards. We don't feel very confident about our estimates of the trend unemployment. There's quite a bit of uncertainty associated with it.

In any case, the current labor force participation rate appears to be somewhat above its trend and the current unemployment rate is now quite a bit below trend.

Sablik: So, in terms of an overall picture of labor force recovery, what would be your verdict?

Hornstein: It's hard to say what you think about the estimated trends.

One thing which gives me more confidence that we get the trends right relative to actual is the current conditions in the labor market. People did not question whether the labor market was hot prior to the pandemic in 2019. But I think people do agree now that the labor market is hot. You see wage gains, you see vacancy postings high, all kinds of signs of a very strong labor market. So, from that perspective, our assessment that the unemployment rate is substantially below trend seems a reasonable assessment, at least from my point of view.

Sablik: That's actually a perfect segue to our next part of the discussion about some of Thomas's research.

You mentioned this news story that I think many people are familiar with, which is that employers lately are having a tough time finding workers. The BLS releases a survey called the JOLTS survey that measures job openings. It recently indicated that there were roughly two job openings for every job seeker in the economy, which is a record high.

In economics, the relationship between job openings and unemployment is captured by a measure known as the Beveridge curve, which Thomas is something that you've researched a lot. I wonder if you could first explain what the Beveridge curve is and what it tells us.

Lubik: Thanks for this question, Tim.

At a very basic level, the Beveridge curve is a graphical description of the relationship between unemployment and vacancies, as you point out. It's named after a British economist who had talked about these issues surrounding these two variables already in the 1940s.

Think of the Beveridge curve as a scatterplot. The BLS, Bureau of Labor Statistics, collects monthly data on the unemployment rate and vacancies. When you plot the combinations of unemployment and vacancies, you see a downward sloping relationship emerging. That is precisely the Beveridge curve.

Intuitively, what the Beveridge curve says is that when times are bad, [the] unemployment rate is high and vacancy postings are low. When times are good and the unemployment rate is low, vacancy postings are high. It's this negative relationship. So, what the Beveridge curve tells us is essentially the state of the labor market. It shows how labor demand [and] labor supply are interacting in one convenient graph. In a sense, it's a tool for labor economists to try to understand the mechanics of the labor market.

What we typically see over a business cycle, as described by the Beveridge curve, is that the economy or the labor market would move along the curve. Think of it this way – at the start of a downturn or at the peak of the expansion, at some point labor demand would weaken. There would be fewer vacancy postings because firms want to hire fewer workers. There will be less hiring, so the unemployment rate would rise. This process then goes on and the economy, the labor market moves along the Beveridge curve. Similarly, at the beginning of an expansion, right at the end of the recession when unemployment is high vacancy postings are low, labor demand would rise. Firms would want to hire more workers, vacancies would be posted, and the unemployment rate would fall.

Sablik: As you were mentioning, the economy moves along the Beveridge curve during these business cycles. But you've also been looking at how the Beveridge curve has maybe changed or shifted over time, right?

Lubik: That's a fascinating aspect of the Beveridge curve. I described … the normal behavior of the Beveridge curve. But occasionally we see shifts in the location of the Beveridge curve in this unemployment-vacancy graph.

In a paper I wrote 10 years ago at the end of the Great Recession, we documented this sideways shift outwards. Inwards shifts are typically associated with very long and deep recessions, or long expansions – for instance, the recession of the 1970s, the Volcker disinflation of the 1980s, the Great Recession following the financial crisis, and now, most dramatically, the pandemic.

What does an outward shift of the Beveridge curve mean relative to movement along the Beveridge curve? What it means is that for every given unemployment rate, when the Beveridge curve shifts outward, more vacancies are now being posted compared to what happened before the shift. More precisely, more vacancies need to be posted by companies to maintain the same level of hiring, the same level of the unemployment rate, that we saw before the shift.

A similar logic applies to an inward movement of the Beveridge curve which, as I said, are associated with long and strong expansion systems such as in the 1990s. For every given unemployment rate, if the Beveridge curve shifts inward, fewer vacancies need to be posted to maintain the same level of hiring.

Sablik: Mm-hmm.

Lubik: This captures the rich labor market dynamics that we see in the U.S. – movements along the Beveridge curve and then sideways shifts. This is in the data.

We can see this most dramatically during the pandemic. The first full data point for the effects of the pandemic – April 2020. We saw that the unemployment rate would shoot up, vacancy postings would also go up, and the Beveridge curve shifted outwards. The latest data that were released by the BLS were the April '22 data, so we now have two full years of a Beveridge curve shift and the same pattern still applies. Compared to the pre-pandemic Beveridge curve, the current location of the Beveridge curve is shifted far out, it is far off where it was before.

This can probably best be seen in the statistic that you described in your initial question. We now have for every unemployed job seeker two job openings, whereas before the pandemic it was closer to one.

Sablik: Mm-hmm. Do you have any sense from your research about what could be causing that dramatic shift?

Lubik: This is an issue that economists are still grappling with, particularly in the current situation.

As an economist, I'd like to think about these issues in terms of a theoretical model because it tries to put boundaries on potential causes for these shifts. Economists describe the Beveridge curve in terms of what we call a search-and-matching model. Firms search for workers, job seekers search for jobs, and then they get matched. When they get matched, they become hires of filled vacancies. When they are not matched, then the workers remain unemployed.

This model has a key parameter, what we call "match efficiency." Just as the name says, it describes the ease of how open positions, vacancies get turned into jobs. It's a little bit like the productivity of the labor market process. The higher match efficiency is, the more productive the matching process is.

One can easily describe the outward shift in the Beveridge curve as a decline in match efficiency. For every unemployed job seeker, more vacancies have to be posted to fill a given vacancy. In other words, firms have to work harder to find the workers that they need, also by recruiting more aggressively.

Match efficiency [is a] convenient concept in the model, but it also hides a lot of interesting things that may be going on below the surface.

For instance, what affects match efficiency are the outside options of the job seekers – such as unemployment benefits or, most prominently, the income support in the pandemic – which overall makes people less willing to search for jobs. It could be geographic mismatch between job seekers and open positions. This was a big factor in the Great Recession.

Often, we talk about changes in job search technology, such as online job postings, that have improved match efficiency. This is less of an issue in the current situation, I would argue, because now we have new technology, Zoom, and we work from home. All of these make it easier to search for jobs.

What stands out in the current situation is the extremely high number of open positions. They are at historically high levels. For one, this signals an extremely strong labor demand. But at the same time, one could also argue that firms tried to engage in labor hoarding. They realize that because the job market is so good, they might lose workers to other firms. So, they tried to preemptively post vacancies. But the pre-pandemic situation was also that lots of employers talked about labor shortages. So, employers may have seen the current situation as an opportunity to hire the workers that they want.

No matter the underlying reasons, labor demand is extremely hot, as Andreas said. Job vacancy postings are extremely high. This is part of the explanation of why the Beveridge curve has shifted outwards.

Sablik: So if these changes persist, what does it mean for the current recovery or for future recoveries? How do policymakers react to this information?

Lubik: What we know from the historic behavior of the Beveridge curve is that it indicates that the U.S. labor market is highly dynamic. I spoke about movements along the Beveridge curve and I spoke about these jumps in the Beveridge curve. It is highly likely that the Beveridge curve will shift back to where it used to be pre-pandemic, and possibly even further because the current sideways shift is truly unprecedented. We haven't seen anything like this in U.S. data.

But this implies one of two things -- vacancy postings would have to fall, which is associated potentially with a decline in labor demand, and/or an improvement in match efficiency. In the current situation, what we do know is match efficiency is lower than pre-pandemic in the sense that more vacancies need to be posted.

The match efficiency aspect of the Beveridge curve does have structural reasons, as economists like to say. These are not directly subject to stabilization policy. An alternative would be to think about reducing labor demand. In the current situation, monetary policy is set on a path that would reduce aggregate demand which would, in and of itself, mean a reduction in labor demand. Whether this leads to a movement along the Beveridge curve or another shift in the Beveridge curve, that is still up in the air.

Sablik: I suspect there'll be plenty more to talk about on this topic in the days and months ahead. I'll put it in another plug for our listeners, if they haven't already, to subscribe to this podcast to keep up on this and other issues. Also, consider rating us and leaving a review on your favorite podcast app if you enjoyed today's conversation.

That's all we have for today. Andreas and Thomas, thanks again very much for joining me today.