Alex Domash, Lawrence H. Summers 17 March 2022
Since the start of the pandemic, labor market indicators have been sending different signals about the degree of slack in the US labor market. This column uses time series and cross-sectional data to show that firm-side unemployment, a measure that links the unemployment rate to the vacancy and quit rate, predicts wage inflation better than either the unemployment rate or the job vacancy rate. employment, and that the company The sideways unemployment currently experienced in the US corresponds to a degree of tightness previously associated with less than 2% unemployment. The findings suggest that labor markets in the US are extremely tight and will likely contribute to inflationary pressures for some time.
Since the start of the Covid-19 pandemic, labor market indicators that traditionally move together have been sending very different signals about the amount of slack in the US labor market. Supply-side indicators such as employment-to-working-age population ratios are still below pre-pandemic levels (79.5% in February 2022 vs. 80.5% in February 2020), suggesting a modest degree of slack in the labor market. On the other hand, demand-side indicators such as layoff and vacancy rates have hit record highs in recent months, indicating a very tight labor market.
The divergence between supply-side and demand-side indicators has sparked a debate about which measure should be used to assess labor market tightness. Some, such as Federal Reserve Chairman Jerome Powell (2021), have suggested looking at employment indicators such as the prime-age employment rate to measure labor market slack. Others have found that demand-side indicators, such as the vacancy-to-unemployment ratio (Barnichon and Shapiro, 2022) or the attrition rate (Furman and Powell, 2021), are better predictors of wage inflation.
In our recent paper (Domash and Summers 2022), we use cross-sectional and time-series data to compare alternative labor market indicators as predictors of wage inflation. We compare four different slack measures: the overall unemployment rate, the prime-age employment rate, the vacancy rate, and the quit rate, and find that unemployment is a better predictor of wage inflation than the employment rate, and that the vacancy and resignation rate are roughly equivalent to the unemployment rate in explanatory power. We then construct a new indicator, firm-side unemployment, that links the unemployment rate to vacancy and churn rates, and find that firm-side unemployment performs remarkably better than the unemployment rate in predicting salary inflation.
Labor market indicators have diverged significantly
Figure 1 shows Beveridge-type curves showing the relationship between supply-side and firm-side labor market indicators since 2001.
Figure 1 Relationship Between Firm-Side and Home-Side Slack Measures, January 2001 – December 2021
Historically, measures of slack on the supply side, such as the unemployment rate and the working-age (25-54) unemployment rate (one minus the employment-to-working-age population ratio), have moved along with the measures of slack on the demand side, such as vacancy and resignation rates, meaning that different indicators gave broadly corroborating signals of the tightness of the labor market. Figure 1 shows, however, that since the start of the Covid-19 pandemic, the supply-side indicators and the demand-side indicators have diverged significantly (represented in orange).
These shifts in the Beveridge curves imply a higher level of vacancies and resignations for a given level of unemployment or non-employment. This begs the question: Are supply-side or demand-side slack indicators more significant in predicting wage inflation?
Firm-side unemployment has dominant explanatory power for wage inflation
We use quarterly time series and cross-sectional data between 1990 and 2019 to compare the explanatory power of different slack measures for wage inflation. Since data on job vacancies and resignations from the Job Vacancies and Labor Turnover Survey (JOLTS) are only available from 2001, we used two alternative data sets to extend these series to 1990:
- For job vacancies, we use data constructed by Barnichon (2010), who uses the Help-Wanted Index published by the Conference Board to create a historical series of vacancy rates from 1960 to 2001.
- For quit rates, we use estimates of quarterly job quits from Davis et al. (2012), who construct a quarterly data set of worker resignations (DFH-JOLTS) by combining cross-sectional relationships of worker flows with data on the cross-sectional distribution of growth rates of establishments.
We first document that the U-3 unemployment rate is better than the prime-age employment rate in predicting wage inflation at the state and aggregate levels, and that the job vacancy rate and the quit rate are comparable to the unemployment rate. U-3 in its explanatory power. These findings hold across different wage series and time periods. We then estimate an equivalent unemployment rate on the firm side by examining what unemployment rate is consistent with current measures of the vacancy rate and the churn rate. We regress the unemployment rate on the log of the vacancy rate and the log of the quit rate, using JOLTS monthly data from January 2001 to December 2019. We run several different model specifications, including different lengths lag, a time trend, and a structural break in July 2009. Overall, all models fit the data from 2001 to 2019 very well, but show a clear break in the relationship after February 2020.
Figure 2 shows the relationship between the actual unemployment rate and the expected unemployment rate on the firm side, using a model with 12-month lags, a time trend, and a structural break. The projected business-side unemployment rate in January 2022 was between 1.3% and 1.7%.
Figure 2 Actual unemployment rate versus predicted unemployment rate on the firm side
Using a wage Phillips curve model that includes the actual unemployment rate and our forecast unemployment rate as regressors, we find that the firm-side forecast unemployment rate has essentially all the explanatory power to predict wage inflation over the period 2001 to 2019. These findings are robust across 12 different model specifications that vary the wage series used to calculate nominal wage growth and the lag length of our explanatory variables. Furthermore, the results also hold in the cross-sectional data: at the state level, declines in firm-side unemployment are more predictive of state-level wage growth than declines in actual unemployment.
Given the extremely low unemployment rate forecast today from the business side, these results provide strong evidence that the current job market is very tight. Projected unemployment on the firm side has fallen from an average of 3.6% in the fourth quarter of 2019 to an average of 1.5% in the fourth quarter of 2021, corresponding to an increase in wage inflation of 4.0% to 4.9% (using CPS-ORG median wages). Figure 3 shows estimates of nominal wage growth from a Phillips curve model of wages using forecasted firm-side unemployment as the slack variable and controlling for lagged inflation. The results indicate that the estimated salary inflation in the fourth quarter of 2021 is the highest in the last 20 years in the four salary measures.
figure 3 Projected YoY Nominal Wage Growth Using Firm-Side Unemployment as a Predictor Variable
Perspectives on the tightness of the labor market in the future
Some economists believe the tight labor market will ease over the next year through increases in labor supply. We conduct a cursory analysis of the current job deficit and estimate that labor force participation is likely to remain significantly depressed through at least the end of 2022, with a retirement glut, Covid-19 health issues, immigration restrictions, changes in the tastes of the workers. represented by reservation wages and changes in the demographic structure that explain most of the labor deficit.
Furthermore, if employment increased due to an increase in labor participation, it would be accompanied by increases in income and thus an increase in demand. Therefore, we believe that supply-side gains over the next year are unlikely to materially mitigate inflationary pressures in the labor market. Taken together, our research concludes that labor markets are likely to remain very tight unless there is a significant slowdown in labor demand. These findings suggest to us the need for substantial caution regarding the possibility of inflationary pressures in the labor market advancing.
Barnichon, R (2010), “Constructing a Composite Index of Aid Requested”, economy charts 109(3): 175-178.
Barnichon, R, and A.H. Shapiro (2022), “What is the best measure of economic slack?”, FRBSF Economic Letters 2022(4): 1-05.
Davis, SJ, RJ Faberman, and J Haltiwanger (2012), “Labor Market Flows Across the Cross Section and Over Time,” Journal of Monetary Economics 59(1): 1-18.
Domash, A and L Summers (2022), “How Tight Are US Labor Markets?”, NBER Working Paper 29739.
Furman, J and W Powell III (2021), “What is the best measure of labor market rigidity?”, Peterson Institute for International Economics blog post, November 22.
Powell, JH (2021), “Getting Back to a Strong Labor Market”, speech at the Economic Club of New York (via webcast), February 10.