How do you reconcile the trend in vacancies over the 20 years? If you try to predict vacancies using quits or unemployment, you get a significant time trend. And rising mismatch between 2001-2019 seems to be very unlikely during a period of falling unemployment and falling wage growth.
In principle, I agree with you that if vacancies do not mean what they used to, we should be careful about using them to guide policy. As we discuss in "u* = √uv", what we would ideally like to measure is the amount of labor (number of man-hours or FTE) devoted to recruiting each month. It would be easy to measure it by adding one extra question to JOLTS, asking firms to report the number of FTEs that they devote to recruiting each month. Then, comparing that number to the number of unemployed workers would allow us to determine whether the labor market is operating efficiently or not. In the absence of that, the best we can do is using vacancies as a proxy for recruiting effort.
I am not sure that I agree with your assessment that vacancies have a time trend. The second graph in the post above shows vacancies and unemployment on a log scale for 1950–2019. If anything it looks like vacancies are slightly trending down from 1970–2010. It is true that there is a large increase in vacancies over 2010–2019, but that is just the recovery from the Great Recession.
If you look at the 1999–2009 Beveridge curve in the first graph in the post, there is absolutely no trend: unemployment and vacancies move up and down along a straight line (down for the dot-com recession, up for the recovery, and down for the Great Recession).
Of course it's true that the Beveridge curve shifted out in an almost unprecedented way after the pandemic. (If you look at the last graph at https://pmichaillat.substack.com/p/us-labor-market-tightness-unchanged, you see that a shift of the same magnitude occurred at the onset of the Great Depression.) But I do think it reflects a real shock to the labor market. Indeed, if you break down the Beveridge curve by industry, you will see that it is only industries that were strongly affected by the pandemic that experience a shift. Industries in which people can easily work from home did not see any shift. This observation is hard to reconcile with a change in the meaning of vacancies.
Lastly, theory does not predict clear relationships between wage growth and vacancies, or quits and vacancies, so I am not sure what to do with the observation that these relationships have changed. How to link these relationships to welfare is also unclear to me. The Beveridge curve has direct implications for welfare and well-known structural interpretations, which is why I always focus on it.
Hi Pascal,
How do you reconcile the trend in vacancies over the 20 years? If you try to predict vacancies using quits or unemployment, you get a significant time trend. And rising mismatch between 2001-2019 seems to be very unlikely during a period of falling unemployment and falling wage growth.
Hi Justin,
In principle, I agree with you that if vacancies do not mean what they used to, we should be careful about using them to guide policy. As we discuss in "u* = √uv", what we would ideally like to measure is the amount of labor (number of man-hours or FTE) devoted to recruiting each month. It would be easy to measure it by adding one extra question to JOLTS, asking firms to report the number of FTEs that they devote to recruiting each month. Then, comparing that number to the number of unemployed workers would allow us to determine whether the labor market is operating efficiently or not. In the absence of that, the best we can do is using vacancies as a proxy for recruiting effort.
I am not sure that I agree with your assessment that vacancies have a time trend. The second graph in the post above shows vacancies and unemployment on a log scale for 1950–2019. If anything it looks like vacancies are slightly trending down from 1970–2010. It is true that there is a large increase in vacancies over 2010–2019, but that is just the recovery from the Great Recession.
If you look at the 1999–2009 Beveridge curve in the first graph in the post, there is absolutely no trend: unemployment and vacancies move up and down along a straight line (down for the dot-com recession, up for the recovery, and down for the Great Recession).
Of course it's true that the Beveridge curve shifted out in an almost unprecedented way after the pandemic. (If you look at the last graph at https://pmichaillat.substack.com/p/us-labor-market-tightness-unchanged, you see that a shift of the same magnitude occurred at the onset of the Great Depression.) But I do think it reflects a real shock to the labor market. Indeed, if you break down the Beveridge curve by industry, you will see that it is only industries that were strongly affected by the pandemic that experience a shift. Industries in which people can easily work from home did not see any shift. This observation is hard to reconcile with a change in the meaning of vacancies.
Lastly, theory does not predict clear relationships between wage growth and vacancies, or quits and vacancies, so I am not sure what to do with the observation that these relationships have changed. How to link these relationships to welfare is also unclear to me. The Beveridge curve has direct implications for welfare and well-known structural interpretations, which is why I always focus on it.