Financial and Monetary Systems

The Great Recession and long-term unemployment: What happened?

Unemployed Belgian Mohamed Sammar (R) answers questions during a simulated job interview, which is recorded to help him get feedback afterwards in Brussels July 2, 2013. Sammar, 27, has been looking for a job in the construction sector for 2 years. "Fit for a job" is the initiative of former Belgian boxing champion Bea Diallo, whose goal was to restore the confidence of unemployed people and help them find a job through their participation in sports. Picture taken July 2, 2013.  REUTERS/Francois Lenoir (BELGIUM - Tags: SPORT BOXING SOCIETY BUSINESS EMPLOYMENT) - RTX11DQK

6.3 million unemployed individuals reported that they’d been looking for work for more than six months. Image: REUTERS/Francois Lenoir

Hie Joo Ahn
Economist, Board of Governors of the Federal Reserve System

In June of 2011 – two years after the Great Recession officially ended – the unemployment rate in the US stood at 9.1%, higher than the peak reached in all but one earlier postwar recession. And 6.3 million of those unemployed individuals reported that they’d been looking for work for more than six months, more than twice the number of long-term unemployed observed at the peak of any earlier recession.

Nevertheless, if you look at only the people who were unemployed in June 2011 who reported that they had just started looking for work, only 57% were still unemployed in July. By contrast, of those who said they’d been looking for longer than six months, 93% were still unemployed the following month. Understanding why the long-term unemployed have so much more trouble finding work is fundamental for characterising what happens during economic recessions.

One possibility is that the process of being unemployed for a longer period directly changes the individual, causing the person to lose human capital or become less attractive to potential employers, an effect that is sometimes described as ‘genuine duration dependence’. Experiments by Kroft et al. (2013) found that potential employers were less likely to call back on fictitious resumes with longer reported unemployment spells. But an audit study by Farber et al. (2016) failed to confirm those findings once more detailed individual characteristics were controlled for.

Another possibility is that the individuals who had been unemployed for six months in June 2011 had started out in January of that year with different characteristics from most of the other newly unemployed people in January, and were less likely to find work from the very beginning. If most of the newly unemployed can find jobs relatively quickly, when you condition on those still looking for work after six months, you have selected out a very different group of individuals. You can see how this could show up in the observed numbers with a simple example (Figure 1). Suppose that in a given group of 100 newly unemployed individuals, 20 of them are ‘type L’, who have an 85% chance of still being unemployed next month, while the other 80 are ‘type H’, who only have a 35% chance of remaining unemployed. After one month of searching, 17 of the type L (85% of the original 20) will still be looking for work, compared with 28 of type H (35% of the original 80). Of the people who are still searching after two months, 60% (15/25) will be type L, even though they were only 20% of the original 100. After six months, 9 of the original 100 may still be looking for work, and they are almost certain all to be type L.

Figure 1

Image: Ahn and Hamilton (2016)

In a recent paper, we show that the numbers used in the simple example above can do a pretty good job at describing the average numbers of people we observe at different durations of unemployment (Ahn and Hamilton 2016). The black circles in Figure 2 below correspond to the average numbers of people who have been looking for work at different durations. The red curve is the predicted numbers of type-L individuals if we assumed that they have an 85% probability of remaining unemployed in any given month, while the blue dashed curve is the sum of these plus type-H individuals who have only a 36% chance of remaining unemployed. Eighty percent of the newly unemployed are type H. But essentially all of the type-H individuals will have found a job after six months, so that the long-term unemployed are all type L.

Figure 2 Predicted (smooth curves) and actual (black circles) numbers of unemployed, full sample

 Predicted (smooth curves) and actual (black circles) numbers of unemployed, full sample
Image: Ahn and Hamilton (2016)

Notes: Horizontal axis shows duration of unemployment in months and vertical axis shows number of unemployed for that duration in thousands of individuals. Dots correspond to average observed numbers for selected durations over the period Jan 1976 to Dec 2013.
Source: Ahn and Hamilton (2016).

So what changed in the Great Recession? Suppose we repeat the above exercise using only data since December 2007. The results are shown in Figure 3 below. The probability of type-L individuals remaining unemployed is 89%, only a little higher than the historical average. But those individuals made up a higher fraction of the newly unemployed. There were one million type-L individuals who became newly unemployed in a typical month since 2007, compared with an average number of 680,000 over the full sample. The feature of the data that gives rise to this conclusion that the key change was an increase in the proportion of type L rather than a change in their job-finding probability is that a curve that would fit the observed totals (represented by the black dots) falls off after six months at a similar rate to what it used to, but starts at a somewhat higher level.

Figure 3 Predicted (smooth curves) and actual (black circles) numbers of unemployed, Great Recession

 Predicted (smooth curves) and actual (black circles) numbers of unemployed, Great Recession
Image: Ahn and Hamilton (2016)

Note: Dots correspond to average observed numbers for selected durations over the period Dec 2007 to Dec 2013.
Source: Ahn and Hamilton (2016).

In our paper, we develop a generalisation of this exercise in which we allow for magnitudes such as the number of newly unemployed type-L workers and their probability of remaining unemployed to change each month. Our framework also allows for the possibility of quite general genuine duration effects, as long as these do not change very much over time. One common assumption in the previous literature (e.g. Alvarez et al. 2016) is that the distribution of unobserved differences across workers is constant. By contrast, our hypothesis is that compositional shifts in the characteristics of the newly unemployed, arising for example from mismatch between idiosyncratic worker characteristics and available jobs, could be the key to understanding the dynamics of long-term unemployment, which is inconsistent with a time-invariant distribution. We show that the assumption of relatively stable genuine duration dependence allows us to characterise the contribution of changes in the kinds of individuals who are looking for jobs to the observed unemployment figures using a nonlinear state-space framework.

The red line in Figure 4 below plots our estimated probability for type-L workers to remain unemployed each month in the sample. These average 81% over the sample and show only a modest tendency to increase during recessions and remain higher after the Great Recession, consistent with the quick calculations above. Type-H workers only have a 34% probability on average of remaining unemployed, and this is actually lower today than in the 1980s.

Figure 4 Probability that a newly unemployed individual of each type will still be unemployed the following month

 Probability that a newly unemployed individual of each type will still be unemployed the following month
Image: Ahn and Hamilton (2016)

Source: Ahn and Hamilton (2016).

Figure 5 shows our estimates of the new inflows into unemployment of each type. Type-L individuals represent only 21% of the newly unemployed on average, again consistent with the above quick calculations. There is a big increase in this magnitude during each recession, and particularly dramatically so during the Great Recession. It was a change in the composition of people newly flowing into unemployment, rather than a change in the probability of finding a job for any given unemployed individual, that was the key reason unemployment went so high and took so long to come down.

Figure 5 Number of newly unemployed individuals of each type

  Number of newly unemployed individuals of each type
Image: Ahn and Hamilton (2016)

Source: Ahn and Hamilton (2016).

Other economists – such as Shimer (2012) and Krueger et al. (2014) – have been sceptical of such an interpretation because we do not see significant changes in observable characteristics of the long-term unemployed during recessions. But Ahn (2016) demonstrated that within any group of individuals with the same coarse observable characteristics, we find a similar pattern to that in the figures above. Conditional on any observed characteristics, someone in that group who has been unemployed for longer than six months is much more likely to still be unemployed the following month than someone with the same observable characteristics who has only been unemployed for one month. Ongoing research by Morchio (2016) and Fujita (2016) further supports that conclusion.

Although there are those we would characterise as type-L individuals within any observable demographic group, we find some important empirical regularities based on observable characteristics. Type-L individuals are more common among those who involuntarily lose permanent jobs, and cyclical variation in the number of type-L permanent job losers seems to be a key factor in the variation in long-term unemployment. Again this suggests that certain industrial or occupational sectors undergo restructuring during recessions, through which workers lose their jobs permanently with no hope of getting recalled to the previous employers. These workers are likely to experience mismatches between their skills and available jobs, and thus to have a longer duration of unemployment than those unemployed for other reasons.

Our conclusion is that a compositional shift in the inflows into unemployment is the key to understanding the dynamics of long-term unemployment. Any theoretical or empirical analysis that assumes that all unemployed individuals within a given observed class have the same characteristics will fail badly in understanding what really happens during an economic recession.

Authors’ note: The views in this column are solely the responsibility of the authors and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System or of any other person associated with the Federal Reserve System.

References

Ahn, H J (2016), “Heterogeneity in the Dynamics of Disaggregate Unemployment,” Finance and Economics Discussion Series 2016-063. Washington: Board of Governors of the Federal Reserve System.

Ahn, H J, and J D Hamilton (2016). “Heterogeneity and Unemployment Dynamics,” NBER Working Paper 22451.

Alvarez, F, K Borovičková, and R Shimer (2016). “Decomposing Duration Dependence in a Stopping Time Model,” NBER Working Paper 22188.

Farber, H S, D Silverman and T von Wachter (2016). “Factors Determining Callbacks to Job Applications by the Unemployed: An Audit Study,” American Economic Review: Papers and Proceedings 2016, 106(5):314-318.

Fujita, S (2016). “Access to Jobs and Duration Dependence,” Presentation at the North American Econometric Society Summer Meeting.

Kroft, K, F Lange, and M J Notowidigdo (2013). “Duration Dependence and Labor Market Conditions: Evidence from a Field Experiment”, Quarterly Journal of Economics, 128(3): 1123-1167.

Krueger, A B, J Cramer, and D Cho (2014). “Are the Long-Term Unemployed on the Margins of the Labor Market?,” Brookings Papers on Economic Activity, Spring 2014: 229-280.

Morchio, I (2016). “Work Histories and Lifetime Unemployment,” Working Paper, University of Vienna.

Shimer, R (2012). “Reassessing the Ins and Outs of Unemployment,” Review of Economic Dynamics, 15(2):127-148.

Don't miss any update on this topic

Create a free account and access your personalized content collection with our latest publications and analyses.

Sign up for free

License and Republishing

World Economic Forum articles may be republished in accordance with the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International Public License, and in accordance with our Terms of Use.

The views expressed in this article are those of the author alone and not the World Economic Forum.

Stay up to date:

Future of Work

Related topics:
Financial and Monetary SystemsJobs and the Future of Work
Share:
The Big Picture
Explore and monitor how Future of Work is affecting economies, industries and global issues
World Economic Forum logo

Forum Stories newsletter

Bringing you weekly curated insights and analysis on the global issues that matter.

Subscribe today

How digital platforms and AI are empowering individual investors

Rishi Kapoor

December 20, 2024

How investing in connectivity and digital infrastructure can be a catalyst for inclusion and empowering people

About us

Engage with us

  • Sign in
  • Partner with us
  • Become a member
  • Sign up for our press releases
  • Subscribe to our newsletters
  • Contact us

Quick links

Language editions

Privacy Policy & Terms of Service

Sitemap

© 2024 World Economic Forum