From the FRBSF:
Improvements in labor market indicators such as job growth and the unemployment rate are strong signals that the U.S. economy is returning to health. One puzzling exception has been the sluggish rise in wages. While wage growth typically rises as unemployment falls, this relationship has been muted in the current recovery. In this Economic Letter, we show that changes in the composition of the workforce propped up wages during the recession, despite a significant increase in labor market slack. As the labor market has recovered, this pattern has reversed. We find that cyclical components, such as the entry of low-wage workers to full-time jobs, have combined with secular components, specifically the exit of higher-wage retirees, to hold down recent measures of overall wage growth.
The recent wage growth conundrum
economic theory tells us that wage growth and unemployment are intimately linked. Wage growth slows when the unemployment rate rises and increases when the unemployment rate falls. The experience since the Great Recession has been very different. Figure 1 shows overall wage growth averaged across four standard measures: average hourly earnings, the employment cost index (ECI) for wages and salaries, median weekly earnings, and compensation per hour. This average wage growth slowed much less than expected during the recession and has stayed relatively flat during the recovery. Even now when most measures of the labor market signal full employment, wage growth has lagged. Average wage growth across the four measures has been hovering around 2¼% for the past two years, significantly below the 3¼% average rate of wage growth from 1983 to 2015. The picture is similar for each measure of wage growth separately (not shown).
Disentangling the puzzling patterns
So what explains these patterns? Aggregate measures of wage growth like the ones used in Figure 1 are computed by taking the average wage in the economy in one period and comparing it with the average wage in the economy in a later period. By construction these measures include information on actual wage changes among individuals who are employed in both periods and information about wage differences between those who moved into versus those who moved out of employment between the two periods. Since these two components have different cyclical patterns, combining them into a single aggregate measure can disrupt the expected correlation between wage growth and unemployment (Bils 1985 and Solon, Barsky, and Parker 1994). Even the ECI, which is designed to minimize these issues by holding constant the composition of jobs in the economy, is susceptible to this problem (Ruser 2001).
To see how much of the wage growth puzzle can be traced back to compositional issues, Daly, Hobijn, and Wiles (2012) and Daly and Hobijn (2016) directly link movements in individual wages to changes in the median of the aggregate earnings distribution. The Daly-Hobijn approach uses a percentile decomposition method to analyze how the change in the aggregate median log-wage over a period relates to (1) wage changes among continuously full-time employed workers that cause them to cross the median and (2) entry to and exit from full-time employment for workers below versus above the median. The measure of median earnings used in the decomposition is closely related to the usual median weekly earnings series reported quarterly by the Bureau of Labor Statistics. By focusing on movements that affect the aggregate median, this method is able to account for how much compositional changes and wage changes of full-time employed contribute to the published aggregate used by policymakers.
Figure 2 shows the results of the Daly and Hobijn decomposition from 2002 through 2015. The gray line is the log change in median weekly earnings growth over a 12-month period. The blue and red lines show the contributions to this aggregate from changes in the earnings of the continuously full-time employed (blue) and from the net effect of worker entry and exit (red).