vacancy rate

[camilla-adams]

Hi Pascal,

As I began to write, I noticed a mistake in my code. I am calculating vacancy rate as job openings/employment in the industry. However, it should be job openings/(labor force).

There is not a way to calculate the total number of unemployed in an industry. The CPS calculates unemployment rate as unemp/(emp+unemp), but I don't think this would work for calculating vacancy rate since the surveys are different.

I was reading about what the BLS does in their vacancy rates tables (they do not survey the number of unemployed per industry). They calculate vacancy rate to be job openings/(unemp + job openings). Does that seem reasonable to you? I think it is at least a step in the right direction.

CA

[pmichaillat]

Unemployment and vacancy have to be measured in the same unit to be consistent with the theory. So it makes perfect sense to divide the JOLTS vacancies by (employed + unemployed) in that industry—where you use the CPS value of (employed + unemployed) in the industry. That way the number of unemployed workers and number of vacant jobs in an industry are divided by the same number. This is the way to go.

The JOLTS definition of vacancy rate is not consistent with the definition of unemployment rate from CPS—since the denominators are not the same—so you should not use that.

[camilla-adams]

Hi Pascal, thanks for your thoughts on this. Do you think this makes sense as a solution?

Idea: Figure out total number of unemployed people (fred time series). Use the CPS to figure out percent of unemployed people who belong to a given industry. (e.g. there are 100 unemployed people, from the CPS we know that 5% of the unemployed people report finance as their industry. there are 5 unemployed in finance.)

I agree that in theory the denominator should be employed + unemployed for each industry. It seems that JOLTS does not collect information on unemployment (i.e. number of unemployed individuals), but does ask firms the number of employees when surveying about the number of job openings at the firms.

I think the denominator from the CPS does not work. it is (unemp)/(unemp+emp) but the unemp measure reflects the CPS sampling (not the full population). Hence, the ratio makes sense for measuring unemployment rate, but the raw measures of unemployment and unemployment are not useful since they are at the level of the sample. Therefore, I don't think it makes sense to use the denominator from the CPS for the JOLTS.

If you look at BLS tables of how "job opening rates," what I was thinking of as vacancy rates are calculated it is (job openings)/(job openings + employed)
https://www.bls.gov/news.release/jolts.t01.htm

Good thing to know and watch out for. The other paper on industry beveridge curves does not mention this issue. I am not sure what they are doing for vacancy rate calculation.

https://www.bostonfed.org/publications/public-policy-brief/2012/what-can-we-learn-by-disaggregating-the-unemployment-vacancy-relationship.aspx

[pmichaillat]

Your idea is great. (I was not aware of the sampling issues you mention.) This way you will get the number of unemployed workers by industry and can then compute the labor force by industry (employed + unemployed) and then the vacancy rate (vacancies/labor force). This will work perfectly for this paper. And in the future it will be very helpful to have these time series for your research.

Let's see how much the results change with the new measure of vacancy rate. Hopefully it should not be too bad. My guess is that the covid Beveridge curves will be slightly steeper, since the vacancy rate was overestimated during covid.

Yes, you always need to check how rates vacancy rates are calculated. The BLS concept of (job openings)/(job openings + employed) does not make any sense. There is no model in which this is a useful measurement. I am not sure why they compute it this way.

[camilla-adams]

Okay. Made the change. Your guess was right! Still looking at how it changes efficient unemployment etc. Thanks for your thoughts!

[camilla-adams]

This paper uses the ``job opening'' rate as the vacancy rate. See Figure 2.
https://www.frbsf.org/economic-research/wp-content/uploads/sites/4/wp10-32bk.pdf