I have been writing some manuscript comments about labor market capacity constraints and inflation, which I hinted at in my previous article. One tangent that came up that will not fit the manuscript is the behavior of the employment-to-population ratio. Although the argument that the “labor market overheating is a major component of sustainable domestic inflation” is quite plausible, the problem is defining “overheating.” If we want to tell stories about the back history, we can pick and choose data as we wish. But if we want to make quantitative forecasts — which is what you need for a falsifiable theory of inflation — you need some variables to feed into your model.

The most popular model input variable historically was the unemployment rate, and it was part of the original “Phillips Curve” of Bill Phillips (although “Phillips Curve” has transmogrified into meaning almost any linkage between labor markets and inflation). The employment-to-population ratio (which I am mainly focussing on herein) was not popular historically because it moved a lot as women entered the workforce. As seen in the top panel of the chart above, we see a pronounced upward trend from the 1960s to the 1990s. In the post-2000 period, the aging population means that the older cohorts with early retirees drag down the workforce as a percentage of the total 18-65 year old population (which is what the chart above is based on).

The unemployment rate appears to correct for these structural changes — it is the percentage of the population that is looking for work. People who are not looking for work — stay-at-home spouses, people on disability payments, students, retirees, rich people — are not counted as part of the workforce. This explains why the unemployment rate is much smaller than 100% minus the employment-to-population ratio.

The problem with the unemployment rate is that is also affected by structural changes in the economy. (The count of claimants for unemployment insurance was affected by the tightening of unemployment insurance policies, but that theoretically should not affect the unemployment rate determined by the BLS survey.) The argument made in the 2010s (which I agreed with) was that the labor market was stagnant, and people drifted out of the official “looking for work” status. They either stopped looking (because they knew there was no chance of being hired), entered into educational schemes (of varying quality), or else ended up taking jobs that offered fewer hours than they desired. Thus, there was increased interest in alternative labor slack measures — other than the people who were convinced that the economy was going to overheat “any minute now” throughout the entire 2012-2020 period. As jobs were created, people drifted into the workforce at roughly the same pace, and so the number of unemployed did not go to zero.

If we just look at the “prime age” (25-54 years old) cohort — which lops off the university and early retirement ages — we get a better picture of the state of the labor market. We just need to compare to post-1990 levels, since the effects of “Women’s Liberation” had largely made their way through the prime age cohort by then. Using this measure, the ratio is near the pre-pandemic level, but below the 2000 boom level.

Since I am not offering a forecast for inflation, I will not comment further on the implications for what is happening next (are we truly running out of available workers?). Instead, I just want to point out that this measure made a lot more sense in explaining the post-2000 dynamics than the unemployment rate, which misled a lot of people in the previous cycle. From an inflation theory point of view, the breakdown of various capacity metrics in response to structural changes in the economy makes it difficult to test quantitative models. An indicator might work in one cycle, but it might break down 1-2 cycles later, which is not that surprising given that recent business cycles are about a decade long.

More By This Author: