The cost of job loss, long-term unemployment, and wage growth

Abstract

Since the end of the Great Recession in June 2009, and despite the current low unemployment rate, wage growth in the United States has been substantially slower than in previous recoveries. This paper argues that the cost of job loss—the one year income loss associated with job loss—and the long-term share of unemployment can account for slow wage growth relative to other labor market variables. Using a wage-Phillips curve for three sample periods, empirical evidence suggests that the cost of job loss and the long-term share of unemployment better explain and forecast wage growth in the United States, particularly during the recovery from the Great Recession relative to conventional measures. This finding better highlights and captures deeper trends in the U.S. labor market, specifically the declining bargaining power of labor.

Keywords:

• Cost of job loss
• great recession
• long-term unemployment
• wage growth

JEL CLASSIFICATIONS:

• E24
• J3
• E02

Acknowledgements

The author thanks an anonymous referee for exceptionally helpful comments and suggestions.

Disclosure statement

The author declares that he has no relevant or material financial interests that relate to the research described in this paper.

Notes

1 Wage growth began to accelerate modestly in early 2016; however, due to data limitations, this analysis ends in 2014. But this recent increase in wage growth—3.1 percent on an annual basis as of this writing—is still low relative to previous expansions, suggesting that continued elevated levels of unemployment duration and thus the cost of job loss, are still likely to be exerting downward pressure on wage growth.

2 Figure 1, which uses a weighted average of five wage and compensation series further described in section 2, begins with the recovery from the 1981-82 recession due to data availability.

3 Since this is an aggregate analysis, it could be argued that sectoral or compositional effects might cast doubt on the findings. Future research, which requires different and more detailed data, will examine the cost of job loss, long-term unemployment, and wage growth by industry, occupation, skill level, education, age, gender, and other characteristics.

4 Although the alternative measures of labor market slack offer marginal improvements in explaining and forecasting wage growth, it can be argued that this highlight deeper institutional shifts that suggest neoliberal policy has transformed labor-capital relations in such a way that slow wage growth, even in tight labor markets, might represent a new norm in labor market outcomes.

5 Unemployed workers might be willing to accept positions that do not fully utilize their skills and pay less than their previous job due to economic need. During a slow recovery with depressed labor demand, workers could remain in these positions for an extended period of time, further depressing wage growth deep into a recovery (Mazerolle and Singh 2004).

6 Kroft et al. (2014) found that short-term unemployment rates returned to their average level in 2013, while long-term unemployment rates continue to hover above their pre-recession average. This will lead to a composition effect where the reduction in short-term unemployment will increase average unemployment duration and the share of the long-term unemployed (Valetta 2013).

7 The weighting program for EViews was generously provided by Ben Spielberg, research assistant at the Center for Budget and Policy Priorities. Whereas the original index uses real compensation per hour, this analysis uses nominal compensation per hour for consistency with other data.

8 The index in Figure 5 is for the annual sample 1982–2014. As will be discussed in Section 3, two additional indexes are calculated for different sample periods: quarterly for 1998.4–2014.4 and annually for 1967–2014. For the latter sample, only two measures of wage growth are available—average hourly earnings and compensation per hour.

9 Because of the one-year estimation of the cost of job loss, the measure presented in this paper is a conservative estimate. Additional analyses of income losses due to job loss, using different data sets and methodologies, find substantial income losses well after initial displacement. Couch and Placzek (2010) found earnings losses of 15 percent for six years; Rothstein (2014), Davis and von Wachter (2011), and von Wachter, Song, and Manchester (2009) found losses of approximately 20 percent for as long as 20 years, with losses higher for displacement that occurs during a recession; and Jacobson, Lalaonde, and Sullivan (1993) find losses averaging 25 percent per year.

10 This formulation of the cost of job loss assumes that unemployment duration cannot exceed one year and that job loss results in an inflow into unemployment, despite the possibility of a worker voluntarily quitting and experiencing an employer-to-employer transition.

11 Despite unemployment duration reaching a record high during the recent recovery, it might actually understate the true depth of long-term unemployment because workers who drop out of the labor force are no longer counted as unemployed, despite being jobless.

12 It is possible that the increase in the cost of job loss—and its effect on wage growth—is concentrated within a few particularly affected industries. In future research, I intend to examine the cost of job loss and wage growth by sector, but that is beyond the scope of the aggregate analysis used in this paper, which simply seeks to identify broad trends in this relationship.

13 Obviously this will vary by education, skill, industry, age, gender, etc., but for an aggregate analysis, it is a reasonable approximation.

14 The rise in long-term unemployment also only accounts for those classified as unemployed, by definition those actively looking for work. Schmitt and Jones (2012) documented that “long-term hardship” in the labor market, which they measure as long-term unemployment plus discouraged and marginally-attached workers. Their findings show that long-term hardship is twice the level of the official long-term unemployment rates since marginally attached and discouraged workers are likely to have been classified as long-term unemployed prior to slowing their job search or dropping out of the labor force altogether.

15 The underemployment rate U-6 and the two Federal Reserve Bank of Kansas City labor market indexes are not available for the latter two samples, and the Federal Reserve Bank’s labor market condition index is not available for the longest sample.

16 Dynamic pseudo out-of-sample forecasts truncate the model at a specified date, then use endogenously-generated values of the lagged dependent variable to forecast for the out-of-sample period.

Additional information

Notes on contributors

Aaron Pacitti

Aaron Pacitti is an Associate Professor, Economics Department , Siena College, Loudonville, NY, USA.