Report | Retirement

The rising stakes of job loss: Stubborn long-term joblessness amid falling unemployment rates

Briefing Paper #162

Download PDF

The rising stakes of job loss

Stubborn long-term joblessness amid falling unemployment rates

by Andrew Stettner and Sylvia A. Allegretto

No one wants to lose his or her job. Families face grave difficulties when a worker is jobless, especially for an extended period. Workers receiving unemployment insurance (UI) benefits receive less than 40% of their prior wages.1 Typically, after six months out of work, the worker has exhausted unemployment benefits and has significantly or completely depleted savings. It is at this point that unemployment can have lasting effects such as elevated levels of debt, diminished retirement and savings accounts (tapped to meet daily expenses), or relocation from secure housing and communities to unfamiliar places in order to find employment.

Recent research has examined how unrelenting high rates of long-term unemployment were spawned by the lack of job creation that followed the 2001 recession. In this report, we examine this unprecedented period of long-term unemployment and compare it with the most recent economic downturn of the 1990s. We conclude that a different picture of long unemployment spells has emerged.

  • Three and a half years into the recovery, one in five of the unemployed have been out of work for six months or more. Never before has the overall unemployment rate (ranging from 5.2% to 6.3% from October 2002 to March 2005) been this low while so many of the jobless have been out of work for such long periods of time. Languid employment expansion has simply not provided those who lost their jobs during the downturn with opportunities to become reemployed during the recovery.
  • The patterns of job creation following the last two recessions have raised the stakes of job loss for a broadening segment of American families.
    • Women represented 43% of long-term jobless workers, on average, from 2001-04, up from 35% compared to the 1990-93 period. Such long-term unemployment has a direct impact on children and families, especially families with single mothers.
    • After experiencing historic labor market gains during the late 1990s, African Americans represented a greater share of the long-term jobless in this economic cycle.
    • Long-term unemployment is expanding beyond blue collar workers: higher levels of education and white collar jobs are no longer providing insulation against severe joblessness.

These consequences make the recent and persistent problem of long-term unemployment a critical labor market problem requiring policymakers’ attention, and assistance for the long-term unemployment was a major issue in the 108th Congress. Changing dynamics should cause law makers to rethink how policy can more effectively support family income while helping those who experience long-term joblessness return to work.

Long-term unemployment over economic cycles

Jobless recoveries followed each of the last two recessionary periods. A recovery is deemed jobless when a recession is officially declared over by the National Bureau Economic Research (NBER) but the economy continues to shed jobs as if in recession. NBER uses several economic indicators to assess peaks and troughs of business cycles, including gross domestic product, real income, industrial production, wholesale-retail trade, and employment. While employment is one of those indicators, it apparently weights less heavily in the calculation of cycles.2 Well into the last two recessionary periods, unrelenting weak employment growth contributed to rising unemployment rates. The extended period of contracted and anemic job growth increased the number of unemployed and the length of unemployment spells. To understand this pattern, we first look at unemployment rates and long-term unemployment in a historical context, comparing the current and previous downturns. We also compare the demographic composition of long-term unemployment that accompanied the recoveries following the 1990 and 2001 recessions. We conclude this report with a discussion of policy implications and suggestions.

Historical context of unemployment rates and long-term unemployment

Historically, as unemployment rates increased during economic downturns, so did, with a short lag, the share of long-term unemployment. Long-termers are the share of the unemployed who have been out of work for 27 or more weeks. Intuitively, it makes sense that as the unemployment rate ticks up it becomes harder for the unemployed to find a job, which, consequently, increases long-term unemployment. As recoveries get underway, the unemployment rate and shares of long-term unemployment fall as employment growth strengthens. The recovery that followed the 2001 recession has proven to be the worst jobs recovery on record. Hence, long-term unemployment problems are symptomatic of the contracted and weak job numbers that persisted long after the recession ended.

Even though the unemployment rate remained relatively low in the most recent recession, it masked, and continues to mask, weakness in the labor market. Since the 2001 recession, the unemployed rate has not provided an accurate measure of labor market weakness because it excludes discouraged and other workers who are no longer looking for work. That is why other economic indicators—such as long-term unemployment and falling labor force participation rates—can be informative gauges of the condition of the job market. As will be shown, the last three peaks in unemployment rates were accompanied by long-term unemployment shares that were well over 20%.

It is problematic when the share of long-term unemployed is at or above 20%. The labor market has considerable slack when long-term unemployment shares are above 20%, which makes it difficult for the jobless to move quickly through spells of unemployment and leads long-termers to exhaust standard methods of assistance like unemployment insurance.

As Figure A shows, unemployment rates peaked right at the end of the “double-dip” recessions of the 1980s, which is consistent with all post-war recessions. However, this was not the case for the last two economic cycles. In addition, shares of long-term unemployment peaked much sooner in the 1980s and all other post-war recoveries compared to the two most recent ones. A closer look reveals how the last two cycles have differed from historical measures.

Figure A

Excluding the last two cycles, since 1948 it took, on average, 1.6 months into an economic recovery for unemployment rates to peak and 8.3 months for long-term unemployment shares to peak.3 In other words, unemployment peaked very near the end of recessions and long-term unemployment peaked shortly thereafter. The last two cycles have seen a different pattern emerge. Following the 1990-91 recession it took 15 months for unemployment and 19 months for long-term unemployment to peak. The lag was even longer following the 2001 recession, when it took the unemployment rate 19 months and long-term unemployment 29 months to peak.

Following the 1990-91 recession, the unemployment rate increased more gradually, reaching a maximum of 7.8%. As the unemployment rate fell from 7.8% to 5.8%, long-term joblessness remained above 20% for 22 of those 29 months. This is the first instance of high shares of long-term unemployment persisting despite relatively low and falling unemployment rates.

For the most recent economic cycle, the share of long-termers reached 20% in October 2002—11 months into recovery. The share of long-term unemployed has stayed above 20% ever since—an unprecedented streak of 31 conse
cutive months and counting. During the past 31 months, the unemployment rate varied between 5.2% and 6.3%. The divergence of these two indicators is unmistakable in Figure A. The enormous gap that materialized after the 2001 recession represents an unambiguous break with precedent.

Since October 2002, the unemployment rate varied between 5.2% and 6.3%—with an average of 5.7%. For this same time frame, the share of long-term unemployed averaged 21.7%.4 Comparatively, for any month since January 1948 that had an unemployment rate between 5.2% and 6.3%, the unemployment rate averaged 5.7%, but the share of long-term unemployed averaged just 12.3%.5

Jobless recoveries and long-term unemployment

The 1990 and 2001 recessions were followed by jobless recoveries. Several important features distinguish the last two economic cycles. First, unemployment and long-term share peaks were reached well after the recoveries began. Second, the relatively low unemployment rate that prevailed following the 2001 recession was accompanied by persistent and uncharacteristically high shares of long-term unemployment.

As a recession gets underway, employment is expected to decrease. At some point employment contraction hits a low point and job creation returns to the economy. Figure B illustrates this employment pattern for the past two downturns (1990 and 2001) and an average of all other post-war recessions. The horizontal line at 0% indicates employment equal to the pre-recessionary level. Hence, when a trend line crosses the 0% line in the graph it indicates that employment is back to where it was prior to the recession. Figure B shows that recessions prior to 1990 tended to be more severe but not as prolonged. On average, it took 21 months to recoup the jobs lost during economic downturns. A very different picture emerged in the last two cycles. It took 31 and 46 months, respectively, to reclaim peak level employment following the 1990 and 2001 recessions.

Figure B

Weak employment growth continued after the 1990 recession, and prolonged, persistent job loss followed the 2001 recession. Lackluster employment, in conjunction with rising unemployment rates, made it difficult for unemployed workers to secure jobs. Consequently, unemployment spells lengthened and the share of long-term unemployment increased. The information in Figure B represents a measure of monthly net employment numbers. However, a relatively new data series called Business Employment Dynamics is also collected by the Bureau of Labor Statistics and permits analysis of gross job gains and job losses over the last two cycles.

Net jobs are a combination of gross job gains and losses in the economy (see Figure C). These data are only available since 1990, but they do provide insight into the last two cycles. As expected, gross job gains initially fell during both recessionary periods. The contraction in the 1990s recession was much less severe and less pronounced than in the 2001 recession. Following the 1990s recession, gross job gains rebounded and stayed on a rather constant flat trend. This was not the case following the 2001 recession. Gross job gains headed downward at the end of 1999, and they have yet to rebound significantly. A strong rebound in gross job gains is what is needed for economic expansion to take hold.

Figure C

Gross job losses peaked during the recessionary periods, the 1990 peak was much more severe than the one in 2001. Post-recession gross job losses fell off during both periods. Hence, the 2001 post-recessionary economy saw a fall off in gross job losses—but because this period was marred by an unrelenting decline in gross job gains, a weak labor market resulted.

This recent pattern of gross job gains and losses is particularly relevant to long-term unemployment. In the recovery of the early 1990s, both job gains and losses were higher because there was more churning in the economy—each quarter saw greater numbers of workers become unemployed, but long-termers who had lost their jobs in earlier periods had more chances to get back to work. Hence, long-term unemployment peaked earlier.

A different pattern emerged after the 2001 recession. With far fewer job gains in the most recent recovery, long-termers found their pathway out of unemployment blocked. The fact that job losses declined meant that the overall jobs picture started improving, but this did little to help those who had already lost their jobs during the heart of the slump. It was these workers who became and remained long-term unemployed as job creation languished many months into the recovery.

The decline in job growth stemmed from many factors: employer reluctance to hire because of continued instability fostered by weak demand, escalating fixed costs of hiring (especially health care costs), and the escalated use of just-in-time employment practices.

Several options enable employers to avoid hiring the more traditional full-time permanent worker and instead resort to just-in-time hiring, which is likened to inventory-adjustment practices used by firms to respond to demand. To meet cycling demand, “just-in-time” hiring practices include more traditional options, such as the use of overtime, but it also includes newer practices such as the use of contingent workers, temporary workers hired (and fired) through temp agencies, and contract workers (many of whom were once employees of the firms they contract for). While such strategies may raise profits, they prolong the lack of job creation, and as such likely help explain the unusually weak job creation in this recovery.

The demographic composition of the long-term unemployed following the 1990 and 2001 recessions

As already noted, laid off workers, even when the overall unemployment rate is low, now face the prospect of an extended period of income loss and difficult job search. What makes this phenomenon more compelling are the changes in exactly which workers are facing periods of long-term joblessness. By comparing the demographic composition of the unemployed over these two periods, distinct changes emerge in the gender, race, educational and occupational make up of the long-term jobless.

Table 1 compares the long-term jobless in the last two jobs slumps, illustrating the average composition over the four-year period from 1990 to 1993 and the comparable period from 2001 to 2004.6 This table clearly indicates that severe long-term unemployment has become much more broadly based demographically.

Table 1

Gender: Long-term joblessness approaches parity 

Conventional wisdom suggests that long-term unemployment was almost exclusively the province of men. The 2001 cycle, however, saw long-term unemployment shares approach gender parity, as women represented a roughly equal share of all unemployed in each of the last two recessions (43% in the 1990s and 45% in the more recent slump). Yet women’s share of long-term unemployment expanded from just under 35% to 43%. It wasn’t that women were more likely to have been unemployed from 2001 to 2004, but that they were more likely to experience longer spells of unemployment.

What explains this shift? The establishment survey tracks the number of women on the payrolls of surveyed firms (it is different than the household survey that collects unemployment numbers by surveying 60,000 households a month).7 During the recessions of the 1970s, 1980s, and 1990s, women’s employment continued to grow as more women entered the lab
or force, and their jobs were concentrated in service occupations that were less sensitive to economic cycles. Women who lost their jobs were more likely to be able to become reemployed before the six month mark. However, in 2001-03, an increasingly service-based economy saw job losses in sectors dominated by women that had previously appeared largely recession-proof.8 As these sectors failed to bounce back quickly, jobless women faced a harder road back to work. The recent jobs slump was the first sustained period in the last 40 years when the number of women on employer payrolls declined, falling in both 2002 and 2003. Simply put, in today’s economy women are more vulnerable to recessions and long-term joblessness.

The increase in long-term unemployment among women has detrimental implications for children and families. Long-term unemployment represents a major income loss to those families depending on women’s work for a greater share of their income. These problems are most dramatic for single mothers, who have neither spousal income nor welfare to fall back on.

Race: Historic gains have been lost for African Americans   
The near full employment economy of the late 1990s brought the African American unemployment rate to a 30-year low of 7.6% in 2000. However, the jobs slump sharply reversed this trend. Not only did the African American unemployment rate quickly increase to over 10%, African American long-term unemployment failed to improve between the two recessions. In fact, African Americans in 2001-04 represented a slightly greater share of the long-term jobless compared to the 1990s period. One would have hoped that the relative strong position of African Americans in 2000 would have braced them against the impact of a recession, but this was not the case.

Elevated African American long-term unemployment defies a simple explanation. As described below, long-term unemployment has become less of a problem among the less-educated population and blue collar occupations, where African Americans are over-represented. A more insidious failure to recover a hold in the labor market across sectors, education levels and occupations appears to be at work. For example, the employment rates of college-educated African Americans decreased by 3.7 percentage points over the slump—nearly double the decrease of white college graduates.

Age: Older and younger workers affected along with prime-age workers 

Long-term unemployment was more widely distributed by age as less than half (43.9%) of the long-term jobless were in their prime working years (25-44) compared to 53.4% during the 1990s. Long-term unemployment shares increased greatly among younger workers—many of whom struggled to find stable jobs upon graduating from high school or college. Employment and labor force participation rates among this age group were also low during this period.

Long-term unemployment is also more prevalent among workers over 55, but for different reasons. Older Americans are working longer into their careers—sometimes in order to maintain employer-provided health care coverage—and their employment rates are reaching record levels. However, those who lose their jobs are also staying unemployed longer, as more appear to be foregoing an early retirement and continuing to look (often unsuccessfully over a six-month job search) for new work.

Education: Better-educated workers represent a larger share of the long-term jobless  

Lesser-educated workers still bear the brunt of economic downturns in terms of overall unemployment. Table 1, for example, indicates that the least-educated workers (with less than a high school degree) represented a steady portion of long-term unemployment in both periods. At the same time, there has been a significant shift of long-term unemployment over the two periods by educational attainment. From the 1990s to the 2000s, an increasing percentage of the long-term jobless had completed some college or more, whereas high school graduates represented a decreasing share. In other words, education did not provide the same degree of insulation as it had in prior economic downturns.

Better-educated workers who lose their jobs have always experienced particular difficulty in finding suitable work to match their prior pay and specific training, lengthening their unemployment spells. As the educated workforce expands, these issues are becoming a greater phenomenon within the overall dynamics of unemployment. The educated unemployed become especially frustrated during the early stage of economic recoveries when a disproportionate share of new jobs is often in lower-paid positions. This pattern is likely to increase in future downturns: seven of the 10 largest growing occupations over the next decade will be entry level jobs like cashiers, janitors, and retail clerks that do not require education beyond high school.9 There is a distinct possibility—given technological advancement, the willingness of other countries to enter world markets, and an oversupply of educated workers in other countries—that the future offshoring of white collar jobs will exacerbate long-term unemployment problems faced by the United State’s college educated workforce.

Occupation: Downturns more broadly based along occupational lines10   
The recoveries examined differ significantly along occupational lines. Unemployment and long-term unemployment shares shifted heavily from blue collar occupations—traditionally male and less educated—to service occupations and, to a greater extent, white collar occupations. This shift indicates that the 2001 recession and weak recovery affected white collar workers more than usual.

From the 1990s recovery to the 2001 recovery, the average share of long-term unemployed blue collar workers decreased by nine percentage points from 40.5% to 31.6%. This share decrease was largely shifted onto white collar workers, which increased from 38.5% to 44.4%—a 5.9 percentage-point increase.

Summary of the demographic dimensions of long-term unemployment  

Long-term unemployment is now reaching a greater cross-section of the population than in past economic downturns. Men and women are now just about equally apt to find themselves unemployed for longer periods. Education and a white collar job do not insulate workers from economic downturns as they have in the past. While joblessness hits less-educated workers the hardest, college-educated Americans increasingly find themselves to be among the long-term unemployed. Finally, despite great improvements during the economic recovery, African Americans represented a larger share of the long-term jobless in the most recent period.

Policy recommendations

Long-term unemployment is becoming a more prevalent and widespread problem that demands a comprehensive response. Longer spells of unemployment take a greater toll on the finances and health of families. States have created model programs to deal with these issues that could be replicated on the national level and in other states. Based on a successful program in Pennsylvania, proposed federal legislation would create a self-financing foreclosure prevention program. Massachusetts imposes a modes
t payroll tax to cover jobless workers’ COBRA costs or to place them in a state-funded insurance program.

For many, a primary need is effective reemployment and retraining services. Here, too, there are promising models. Sectoral strategies have helped jobless workers enter into growth industries. The Trade Adjustment Assistance (TAA) program provides extended income support to workers who can benefit from a two-year retraining program. However, these programs have been geared to blue collar workers (TAA, for example, excludes service workers who have lost their jobs due to trade) and overall spending on such strategies has represented a decreasing share of public and private resources. With the current increased need, the time has come to rethink policies affecting the unemployed.

Six months of benefits is insufficient  

Unemployment insurance remains the economy’s first responder to labor market weakness. Basic state unemployment benefits provide 26 weeks (six months) of income maintenance, with checks that average just under 40% of average weekly wages. The persistent long-term unemployment that followed the past two recessions demonstrates that the basic safety net failed before many were able to find jobs. Extended unemployment benefits are crucial for meeting the needs of families during lengthy periods of widely experienced long-term unemployment. Such benefit checks help families preserve their savings and meet basic expenses as workers continue prolonged searches for new jobs.

In each of the downturns since World War II, extended benefits have been paid out. For example, the Temporary Extended Unemployment Compensation program paid out more than $20 billion in benefits from March 2002 until December 2003. The problem is that Congress cut off the program too soon—largely because the overall unemployment rate appeared low by historical standards. Because long-term unemployment remained severe throughout 2004, many working families suffered because of the lack of available benefits. In 2004, more than three million Americans depleted their state unemployment benefits without access to an extension. This represented a 30-year high in the number of workers cut-off (and a record level since statistics have been kept). There were far more Americans denied extended benefits during the 2001-04 period than the 1990-93 period, when the Emergency Unemployment Compensation covered a greater portion of workers needing assistance. If temporary extensions continue to be the prefered method for addressing this problem, policymakers need to pay more careful attention to long-term unemployment trends in deciding when to cut off benefits.

An effective automatic program is needed  

With extended benefits serving an urgent need, it is not advisable to leave the program to the whims of Congress and political debate. In 1970, Congress created the federal Extended Benefits (EB) program, precisely for this reason. The goal of the program was to automatically lengthen the period during which an unemployed worker could receive unemployment insurance in extended periods of economic downturn. The automatic mechanism for this extension eliminated political debate in Congress. Under EB, benefit costs are split equally between the federal government and the states. Workers receive an extra 13 to 20 weeks of benefits if they qualify for EB.

This automatic benefit program ought to be able to respond each time there is a period of persistently high long-term unemployment. However, EB has failed during the past two recessions, because of the rules used to activate benefits. Having abandoned a single national standard for when to pay out benefits, the law now requires that a state have a 5% insured unemployment rate to qualify for EB—Alaska was the only state to meet this requirement from 2001-04.11 Workers in another four states received extended benefits because their states had adopted an optional rule that activates benefits when a state’s overall unemployment rate exceeds 6.5%.

Two changes could be made to improve the EB program:

• A new national trigger is needed. We can expect future jobless recoveries to produce elevated levels of long-term joblessness, even when the overall unemployment rate is low. Therefore, a new national trigger could logically be based upon the total number of jobs in the economy. For example, extended benefits could be triggered automatically when payroll employment falls by 1.0% from the employment peak. When payrolls recover to their pre-recession level, extended benefits would be phased out. Such a measure would have put an extension in place from November 2001-December 2004.

• The total unemployment rate trigger needs to be expanded. If the system retains only state-by-state triggers, Congress should require that all states adopt the 6.5% total unemployment rate trigger. If this provision had been in place during the past recession, workers in large states like California, Illinois, and Texas would have qualified, but none of these states had the optional trigger. Workers in these three states alone represented more than 20% of all UI recipients.

Deeper questions about unemployment benefits also emerge from this long-term unemployment story. Even three months of additional benefits is not likely to suffice for many jobless workers when job growth is stalled. Many of today’s long-term unemployed have indicated that a full year of jobless assistance is needed, as a matter of fairness and practicality.12 The diversity of the unemployed population described above validates this approach. Long-term jobless workers are not only the limited set of blue collar factory or construction workers who frequently get unemployment checks. In fact, a recent GAO report found that only one out of five workers who received UI benefits between 1979 and 2002 had received benefits more than twice.13 Employers contributed billions to federal unemployment benefit accounts based on the wages of the many experienced workers who suffered their first long spell of joblessness during this most recent slump.


1. U.S. Department of Labor, “Unemployment Insurance Quarterly Data Summary,” 2nd Quarter Calendar Year 2003 available at

2. For information on the dating of business cycles, see

3. The range of months it took for unemployment to peak was zero to four. The range of months it took for long-term unemployment shares to peak was five to 17 months.

4. The standard deviation was one percentage-point.

5. The standard deviation was 3.6 percentage-points.

6. Each period starts from the official beginning of the recession and includes the next three years.

7. For information regarding the household versus the payroll survey see:

8. Heidi Hartmann, Institute for Women’s Policy Research, presentation to the National Press Club, October 6, 2004, available at

9. 2002 Bureau of Labor Statistics Employment Projections, 2002-2012, Table 3c available at

10. An occupational coding change was implemented by the Census Bureau in 2000, and broad occupational categories were cons
tructed for comparative purposes. Contact the authors for detailed occupational aggregation methodology.

11. This is known as the insured unemployment rate: the ratio between insured unemployment, the number of people receiving UI, and covered employment, the number of people working in the state. The total unemployment rate represents the percent of the total workforce without a job regardless of whether they receive UI. In some states, the total unemployment rate can be twice as much as the insured unemployment rate because many jobless workers do not receive UI checks.

12. See for example, Unemployed in America. Peter D. Hart Research Associates, May 2003.

13. General Accountability Office, “Unemployment Insurance: Information on Benefit Receipt,” GAO-05-291 (March 2005): 24.


Faberman, Jason R. 2004. “Gross Job Flows over the Past Two Business Cycles: Not all ‘Recoveries’ are Created Equal.” Bureau of Labor Statistics, Office of Employment and Unemployment Statistics. Working paper #372.

U.S. Department of Labor. Bureau of Labor Statistics (BLS). 2004a. Current Population Survey. Washington, D.C.: U.S. Government Printing Office.

U.S. Department of Labor. Bureau of Labor Statistics (BLS). 2004b. Current Employment Statistics. Washington, D.C.: U.S. Government Printing Office.

See related work on Retirement

See more work by Andrew Stettner and Sylvia Allegretto