Labor Market Frictions and the Persistence of Disequilibrium

When assessing the state of the labor market, policy makers, analysts, and the media typically focus on the unemployment rate and the number of jobs being created or eliminated over a given period of time. Yet job growth and the unemployment rate fluctuate widely over time. As illustrated in Table 7.1, job growth in the previous cycle averaged just 0.2 percent per year, compared to 1.8 percent in the business cycle before that and more than 2 percent in the 1982–1991 business cycle. The unemployment rate has also varied widely throughout different business cycles, indicating that the labor market has not behaved in a consistent manner over the past three cycles.

The boom-and-bust dynamic of the business cycle leads to jobs being cut as economic activity contracts. Businesses are quick to cut employment, leading to payrolls falling and an excess supply of labor that drives the unemployment rate above equilibrium levels. Even accounting for trends in the business cycle using a Hodrick-Prescott (H-P) filter, payroll levels and the unemployment rate are often out of sync with their underlying trend and what could be considered full employment (Figures 7.2 and 7.3).¹

Through the cyclical variation in joblessness, the unemployment rate tends to revert back to its mean over time. Since 1982, the unemployment rate has averaged 6.4 percent, and an Augmented Dickey-Fuller (ADF) test confirms econometrically a single mean of the series.² It is worth noting that the mean is also above what is typically considered full employment by Federal Reserve officials, which is currently just below 5.0 percent.

In contrast, the U-6 rate of unemployment, the broadest measure, which includes workers employed part time but who would like full-time work and those marginally attached to the labor force, is not mean reverting. This is consistent with the view that in some ways, the labor market of the twenty-first century is behaving differently than in prior periods. Although the data only begin in 1994, an ADF test, along with the plot shown in Figure 7.4, suggests that the mean of this series has actually moved up over time. Therefore, policy makers and analysts should be cautioned against using past levels of U-6 unemployment as a benchmark for what may be considered “normal” levels.

Unemployment rates and topline payroll figures also mask the churn going on underneath the surface of the labor market. While the net number of jobs being created in an expansion or cut during a downturn is in the low hundreds of thousands each month, millions of workers actually leave a job and/or start a new one each month. For example, when 201,000 jobs were created in January 2015, five million jobs had a new employee in the position, either the product of a new employee filling an existing position or an employee being hired for a new position in a growing business (Figure 7.5).³

Similarly, millions of workers separate from their job in a given month. This can be due to voluntary separations, that is, workers quitting their position to go to another job, retire, or temporarily leave the workforce (Figure 7.6). A high rate of workers quitting their jobs is viewed as a positive trend for the economy. Quitting is a sign of improved job prospects, either realized if a job is already lined up or confidence that one won’t be hard to find. Therefore, quits exhibit a cyclical pattern over the business cycle. Quits can also be positive for underlying productivity, as workers are likely to switch to jobs that better suit their skills and interests.

Of course, turnover can also be due to involuntary separations, such as a worker being fired or laid off as demand for a company’s product falls during a downturn or the company’s production becomes out of date in the current economy. Total hiring will outstrip separations during the expansion phase of the business cycle, leading to an increase in the total number of jobs in the economy (Figure 7.5). In downturns, separations rise ahead of hiring and indicate an overall decline in the number of jobs in the economy.

Many workers quitting a job may do so because they found a better job elsewhere and start the new position within a matter of days or only a few short weeks. For others voluntarily leaving their job, they may find themselves unemployed for a spell as they hold out for a well-fitting job or look for a job in a new location following a job relocation of their partner. Similarly, finding a new job for a worker laid off that suits their skills may take some time as their industry may be in decline or demand conditions in general may be weak.

Search Costs: The Value of Imperfect Information

Finding a new job can take time for a number of reasons. First, information about what jobs are available, and at what wage, are neither instantly available nor widely distributed. It takes workers time to search out this information, whether through job postings online or through talking to people in their networks.⁴ Even when workers learn of new job opportunities, the available jobs may require skills that the unemployed worker does not have, leading the worker to continue looking for jobs that match his skills. Furthermore, location may be an issue. The United States is a fairly homogenous society, making it easy in a cultural sense to move from one location to another if a job that matches the right skills and offers an acceptable wage is available. However, in practice, moving is costly, whether in terms of the financial costs of selling a home, breaking a lease, and the physical transportation of household items, or through the less measurable emotional costs of cutting community ties.

Unemployment during the time needed for workers to find a new job that suits their skills, in a location they are already in or willing to move to and at an acceptable wage, is known as frictional unemployment. Frictional unemployment is not necessarily a bad thing. It indicates that workers are taking some time to find a job that best matches their skills, rather than the first job available, which should be good for the workers’ productivity and the economy’s productivity more broadly. Although a benign type of unemployment, estimating this component of the unemployment rate is important for policy makers when assessing the amount of slack in the labor market or whether the labor market is tightening to the point where it would begin to put upward pressure on wages.

Making policy makers’ jobs difficult in this arena is that the rate of frictional unemployment can change over time. Financial costs to moving were particularly steep in the wake of the Great Recession, when home prices collapsed and pushed many households into negative home equity positions. This likely raised the rate of frictional unemployment for at least a time. Evolving trends in the economy can also change the rate of frictional unemployment, as different sectors take on more or less importance. These sector shifts can raise frictional unemployment as the jobs demanding certain skills may be shrinking, making it more difficult for unemployed workers to find a job matching their skills. This may also be considered a form of structural unemployment, where the unemployed workers do not have the right skills for the jobs that are available.

Structural unemployment should dissipate over time as unemployed workers update their skills and retrain. However, in reality, not all workers will retrain. The cost-benefit trade-off of going back to school or undergoing a retraining program looks very different for older workers versus younger workers, the latter of which have longer time horizons to recoup their investment of updating their skills. The trade-off also would vary by location, depending on whether the new-skilled jobs are even available in a particular area. Therefore, structural unemployment may lead some workers to drop out of the labor force altogether.

We see evidence of frictional unemployment by looking at the unemployment rate and length of unemployment over time. Even during strong periods of economic growth and labor demand, the unemployment rate does not fall to zero. In fact, the lowest it has ever been was 2.5 percent in 1953, during a very different economy and labor market. Similarly, the median duration of unemployment fluctuates over the business cycle, but in recent decades has only fallen as low as five weeks, indicating the time it typically takes for an unemployed worker to find new suitable employment (Figure 7.7). The median duration of unemployment has drifted upward over time, or, in other words, is nonstationary. This comes as the series experienced two structural breaks following the financial crisis and Great Recession. The first was a shift upward in mid-2009, while the other was a shift lower in mid-2012, when the median duration began to fall more rapidly (both significant at the 1 percent level). The upward shift, however, was to a greater magnitude, implying the trend in the duration of unemployment has gravitated up since earlier business cycles.

The more severe structural type of frictional unemployment can be illustrated by the rate of long-term unemployment, or unemployment lasting longer than six months (Figure 7.8). The long-term rate of unemployment spiked to 4.0 percent in mid-2011, nearly four times its average since 1960 of 1.1 percent. Yet even the short-term unemployment rate, which better accounts for workers whose skills are up to date, never falls below 3 percent. The higher rate of long-term unemployment has persisted to where the series experienced multiple structural breaks upward in 2009.⁵ While long-term unemployment is nonstationary and has multiple means, the short-term unemployment rate is stationary.

The persistence of unemployment can also be seen by looking at the unemployment rate against the rate of job openings. Even in recessions, labor market turnover or secular growth in certain industries leads to companies needing to hire. However, when the economy weakens, fewer companies are looking to expand, and therefore the job openings rate tends to fall as the unemployment rate rises. This negative relationship is illustrated by the downward slope of the Beveridge curve, which plots the unemployment rate against the rate of job vacancies (Figure 7.9).⁶ Frictional unemployment is evident in that no matter what the rate of unemployment, there are always some job openings that remain unfilled.

Insights into structural unemployment are also offered by the Beveridge curve. A higher rate of unemployment for a given level of job openings would suggest that workers do not have the right skills or are not in the right location for jobs that are currently available. A higher rate of structural unemployment would therefore be indicated by an outward shift in the Beveridge curve. It is typical for the Beveridge curve to move outward in the recovery phase of the labor market, when job openings begin to rise but employers take their time filling them.⁷ However, a persistent shift outward, lasting through the expansionary phase of the labor market, would indicate that the needs of employers are changing more quickly than the skills of workers. This appears to be the case in the most recent labor market expansion. Even as the unemployment rate has fallen rapidly, the rate of job openings remains relatively high when compared to the labor market expansion of the 2000s, indicating less efficiency in matching available job opportunities to workers.

Experience Matters: The Different Labor Market Outcomes and Work Patterns across the Age Spectrum

Workers’ experience level is one of three major factors that drive labor market outcomes. Measuring workers’ effective experience across the economy would be difficult, to say the least, but age offers a good initial proxy, where older workers are presumed to be more experienced. Unemployment varies widely among different age groups (Figure 7.10). While the unemployment rate for 25- to 34-year-olds closely tracks that of the overall unemployment rate, there are significant differences among other age groups (Table 7.2).

Not surprisingly, more experienced workers—in other words, older workers—have stronger chances of employment. Older workers’ years of experience in a specific field or company and longer track record of their ability to perform are clearly valued when looking at the unemployment rates for different age groups. Among the working-age population not of retirement age (ages 16 to 65), the unemployment rate tends to be lowest for workers aged 55 to 64, and only slightly higher for workers aged 45 to 54.

In contrast, unemployment among teenagers is significantly higher than the broad population. Over the past 50 years, the overall unemployment rate has averaged 6.1 percent, while the rate for 16- to 19-year-olds has averaged 17.6. Similarly, although not as pronounced, this pattern holds true when looking at the unemployment rate for 20- to 24-year-olds, which has averaged 10 percent over the past five decades. The unemployment rates for both 16- to 19- and 20- to 24-year-olds have experienced a number of positive structural breaks, which suggests the unemployment rate has shifted upward over time. While unemployment is much less prevalent among 25- to 34-year-olds compared to teenagers and workers in their early 20s, it is notably higher than other prime-age workers (defined as workers aged 25 to 54) as these workers may possess some general work experience but less experience relevant to their career field.⁸

For young workers, the higher rate of unemployment comes as they tend to have higher rates of job separation. This may be due to the fact that when firms are under pressure, they first lay off workers with the least experience and tenure with the firm. However, it is also likely influenced by the voluntary separations of young workers as they seek out jobs in different industries or only work during certain times of the year.

Higher separation rates for younger workers are evident in the average length of employee tenure. While in 2014 the average length a worker had been with their current employer was 4.6 years, workers in their early 20s had been at the same employer for only 1.3 years and workers aged 25 to 34 an average of only 3.0 years.⁹ In contrast, workers aged 55 to 64 had been at their current employer for an average of 10.6 years. The longer tenure and lower separation rates come as older workers have gained more specialized skills and knowledge in a particular field and are less likely to switch occupations or industries. Longer tenure is also associated with the institutional knowledge a worker who has been with the firm for a long time possesses, which can make the worker more productive and valuable to the firm.

In addition to facing higher unemployment, younger workers are also more likely to work part time. The high rate of part-time employment among 16- to 24-year-olds is to be expected, given the greater likelihood that these workers are also in school. However, underemployment is also highest among young career-age workers (Figure 7.11). The share of 25- to 34-year-olds who are working part time but would like to be working full time is about 10 percentage points higher than for workers older than age 35, as their more limited experience makes employers more reluctant to hire them full time.

In addition to employment outcomes, age also plays an important role in the decision to participate in the labor market. The Bureau of Labor Statistics defines the working-age population as the population aged 16 and older, but participation rates vary widely across the age spectrum (Figure 7.12). The vast majority of teenagers are in some type of schooling, and therefore only around 30 percent participate in the labor market each year, most of whom engage only in the summer months. With the better income prospects afforded by a college degree, the participation rate among 20- to 24-year-olds is also lower than the general population since many members of this cohort are also in school. Between ages 25 and 54, however, participation rises dramatically to where about 80 percent of this group is involved in the labor market.

By age 55, however, the participation rate begins to fall noticeably. Even before workers hit the standard retirement age of 65, some have the income and wealth needed to retire. Others may not have the means, but either find themselves too old to retrain if unemployed or with disabilities that make it difficult to work. As workers hit 65, the traditional standard age of retirement, and are able to collect their full Social Security benefits, labor force participation falls even more, to around just 30 percent.

Education Pays: Labor Market Outcomes across the Skill Spectrum

Education is the second clear link between labor market outcomes and worker characteristics. Education levels provide a good approximation of workers’ skill levels. Not surprisingly, more skilled workers have an easier time of finding employment and obtaining higher wages for their work. More broadly, a more skilled workforce is one of the factors that improves productivity in the economy. Stronger productivity growth is a key ingredient in boosting real economic growth and in turn raising the living standards of a society. This helps to explain policy makers’ focus on improving the quality and access to education among the population.

Perhaps the most obvious way in which education impacts workers’ experience in the labor market is through earnings. Workers with more skills can command higher compensation as the supply of labor in which to fulfill those jobs is smaller. Presumably, all college graduates could work a cash register, but few retail clerks have the training needed to work as an engineer. This leads incomes to rise alongside education. In 2013, a household headed by someone with a college degree earned $80,100, 72 percent more than the median household income and 120 percent more than the median household headed by someone with only a high school diploma. Furthermore, the premium of a college degree has held steady in recent years despite rising costs of attendance as income for non–college graduates has followed a similar trend over the past decade.¹⁰

Not only do workers with more education receive higher wages, but they have an easier time finding employment. As workers’ education levels go up, unemployment rates go down (Figure 7.13). The differences in unemployment rates by education are statistically significant compared to the overall unemployment rate across all education levels (Table 7.3). Over the past two decades, the unemployment rate for college graduates has run three percentage points lower than the unemployment rate for workers with only a high school diploma, on average. The differences in unemployment rates become even more pronounced during downturns in the labor market, which suggests that the impact of recession hits workers of different education levels in disproportionate ways. For example, in 2009, when the overall U.S. unemployment rate peaked at 10 percent, the unemployment rate for college-educated workers climbed only as high as 5 percent.

As employers had to make tough decisions about firing and hiring during the past recession, their preference for more educated workers was made clear. Bachelor’s degree holders now account for 38 percent of employed workers compared to 34 percent at the start of the recession. As employers continue to seek workers who are able to add value in a more knowledge-based economy, relatively fewer opportunities have become available for low-skilled/semiskilled workers—often associated with those without a high school diploma. Younger generations have a higher share of college graduates, but with only 36 percent of 25- to 34-year-olds having college degrees, there is still great variation in educational attainment.

With higher wages and an easier time finding employment, participation in the labor force is positively associated with education. Labor force attachment rises with education as the opportunity cost of not working, that is, forgone wages, increases. The labor force participation rate for college-educated workers stood at 75 percent in 2014 compared to 58 percent among those with only a high school diploma. This participation gap has also widened in recent years, having stood at less than 15 percent in the mid-2000s. Not only do workers with less education earn less income, but a significantly higher share is not engaged in the labor market. The growing differences in participation rates, combined with higher unemployment and lower earnings prospects, have led to a wider divergence in income among educational groups.

Gender’s Role in Labor Market Outcomes: Persistent Differences between Men and Women

Gender is a third important way in which workers’ labor market experiences can vary drastically. Cultural norms that have dominated for centuries are still apparent in the modern labor market. Since in many households the role of primary breadwinner falls to the man, men have higher rates of labor force participation than women, who still bear the bulk of household and child-rearing responsibilities. In 2014, the participation rate among prime-age working males was 88 percent, 15 percentage points above the participation rate of women. For women who do work outside the home, they are much more likely to work part time. For most women, the decision to work part time is a choice, as they balance the same household responsibilities that keep some women out of the labor force altogether. Only around 20 percent of prime-age women working part time are underemployed in terms of hours worked, compared to about two-thirds of prime age men working part time (Figure 7.14).

Different labor market patterns between men and women are also evident when looking at the types of jobs held. Men still tend to dominate employment in a number of industries, most notably the construction and mining sectors (Figure 7.15). On the flip side, women tend to comprise a relatively high share of employment in education and health services. While unemployment rates between men and women have tracked fairly closely since 1980, these industry differences can lead to shorter-term variations. For example, the unemployment rate for men rose to a high of 11.1 percent following the Great Recession versus 9 percent for women. This came as male-dominated industries were hit particularly hard by the downturn. Employment in the construction industry fell 30 percent from peak to trough and 14 percent in the manufacturing sector. Meanwhile, government employment, which is composed of a relatively high share of women, fell only 4 percent, and education and health services employment continued to grow throughout the recession and recovery.

However, even as unemployment outcomes are similar between men and women over the course of the business cycle and more women are now receiving college degrees than men, income among men remains higher. Some of this can be explained by the industry distribution of female workers, where women hold a relatively higher share of jobs in lower-paying industries. Yet even within individual industries and occupations, women almost always earn less than men. Part of this is due to different levels of experience, where women are more likely to take time off in their careers for child rearing or work part time. Another factor is that even among “full-time” workers, women still tend to work fewer hours than males. In 2014, women in their prime working years classified as full time (working 35 hours or more) worked an average of 41.0 hours per week compared to 43.9 hours for men working full time. Yet even when controlling for industry, education, experience, and hours worked, evidence suggests an unexplained earnings gap persists.¹¹

Why Does Worker Heterogeneity Matter? Cyclical vs. Secular Trends in the Labor Market

Imperfect information and heterogeneity among workers make it difficult for the labor market to achieve and maintain the theoretical steady state. Fluctuations in the business cycle make it even more difficult as labor needs for employers change and wages adjust. Beyond the business cycle, the supply and demand for labor is constantly evolving as the economy grows. This creates structural trends that can outlast the business cycle and make the current position of the economy in the business cycle more difficult to determine. Many of these trends stem from the changing composition of the workforce. Given the different labor market outcomes and work patterns among different workers, shifts in the composition of the labor force can have meaningful effects on aggregate measures of the labor market. As a result, the movement of some indicators must be interpreted more carefully, accounting for the longer-term secular trends in the economy.

Labor market differences among age groups and the changing age distribution of the workforce are two ways in which secular trends can obfuscate the current state of the labor market. Some analysts believe that the low rate of job turnover in recent years not only stems from the weak labor market but is also being held down by the relatively large share of the workforce in the later stages of their careers. As noted previously, older workers switch jobs at lower frequencies, and therefore the low rate of job turnover may not be as far from normal levels as it first appears. Similarly, the rising share of workers at the younger end of the wage spectrum, which tend to work in lower-paying industries and obtain lower wages in general, may be keeping average hourly earnings growth muted.¹²

Perhaps nowhere has the impact of the changing age composition of the adult population been more apparent and in need of more scrutinized interpretation than the labor force participation rate in recent years. In 2008, the labor force participation rate began to fall sharply as the Great Recession took hold and job opportunities for the most part disappeared (Figure 7.16). Since then, the labor force participation rate has fallen more than 3 percentage points. However, the drop in labor force participation also coincided with a major demographic milestone: the first of the Baby Boomers turned age 65 in 2011. Moreover, the link between education and employment/earnings outcomes has led to more young adults obtaining a college education. While some students work part time, many do not, which has led to labor force participation rates for the population under age 25 continuously sliding since the early 1990s as young adults join the workforce later in life.

Therefore, analysts and policy makers have sought to determine how much of the decline since the Great Recession is due to cyclical factors and how much is due to secular trends. In addition to cyclical factors weighing on the participation rate as workers of all ages were more discouraged over their job prospects, a historically large share of the working-age population was reaching the age in which participation declines dramatically even in a strong labor market. Understanding the cyclical component is an important facet in determining how much slack is in the labor market and in turn the most prudent stance of monetary policy.

Secular trends such as the aging of the population and young workers staying out of the labor force longer to obtain more education have made it difficult to determine precisely how much of the decline in the labor force participation rate since the Great Recession has been due to cyclical weakness. Therefore, it has been difficult to determine how much the labor force participation rate may bounce back once the labor market strengthens. To assess the degree to which demographics have weighed on the labor force participation rate since 2007, we can examine how the distribution of the working-age population has shifted in recent years. As the Baby Boomers have aged, the share of the population aged 55 and older has grown (Figure 7.17). Holding constant the participation rate for each detailed cohort since the recession began, demographics alone would have lowered the participation rate to 64.6 percent between 2007 and 2013, or by 1.5 percentage points.¹³ Therefore, demographics appear to have accounted for a little over half of the 2.8 percentage point drop in participation between 2007 and 2013. At face value, the decline in the labor force participation rate, without taking account of the different work patterns of older workers, would lead to an overestimation of the amount of slack in the labor force.

Even as the participation rate has been influenced by population trends, additional secular trends have weighed on the participation rate in recent years. The participation rate for prime-age workers—ages 25 to 54—had started to fall well ahead of the 2007–2009 recession. Between 2000 and 2007, the prime labor force participation rate declined 1 percentage point. The drop in the prime-age participation rate ahead of the Great Recession suggests a structural decline in the total participation rate beyond the widely discussed demographics and educational trends.

The labor force participation rate in the second half of the twentieth century was dominated by secular trends. Female labor participation rate for prime-age women steadily increased beginning in the early 1960s through the mid-1990s (Figure 7.18). Between 1950 and its peak in 1999, the labor force participation rate among women nearly doubled. The steady secular trend of women increasingly working outside of the home often masked the cyclical variations in labor force participation. By looking at the aggregate data, there would be little reason to believe that labor force participation varied with the business cycle, increasing when demand for employment and wage growth was strong, and declining when unemployment rose and job opportunities became relatively scare.

By the early 2000s, the trend in female labor force began to level out and could no longer mask the trends in male participation rates. The rate for prime-age men has been declining since the mid-1950s. There is no widely agreed-upon reason for the ongoing decline in prime-age male labor force participation. However, one possible reason is increased social insurance—particularly disability insurance.¹⁴ Another possible explanation for the long-term decline in prime-age male participation is declining real wages for lower-skilled jobs.¹⁵ In other words, the opportunity cost of leisure has declined for lower-skilled workers. This may also account for the plateau in female participation rates, as the income earned from working outside the home for many women is not enough to offset its costs, such as child care or higher marginal tax rates.

The rise in female labor force participation in the second half of the twentieth century was an important factor in raising the supply of labor available in the economy. As labor is an input into production, the increased labor supply helped to spur stronger growth in the economy, as well as raise the real income of households, as many now included a second breadwinner. However, now it seems that female labor force participation has hit a ceiling, which will limit the future rate at which the labor supply, and therefore the economy, can grow. In other words, the leveling off of the participation rate among women, combined with the continued decline in the labor force participation rate for prime-age men, has limited the potential gross domestic product (GDP) growth of the economy.

Implications for Private-Sector Hiring and Policy Makers

Although job gains have picked up notably in 2014, the subpar pace of job growth compared to prior expansions should be taken within the context of a slower trend in payroll gains. Much has been made of the unemployment rate not fully capturing remaining slack in the labor market, but if the trend in payroll growth has downshifted as appears to be the case, it should be less surprising to analysts and policy makers that the unemployment rate has fallen so swiftly.

Accurately assessing the trend in job growth will have important implications for businesses and policy makers. First, with labor force growth set to slow in the coming years, it may not be a “buyers’ market” for labor as long as some employers currently expect. Instead, employers may find themselves facing talent shortages, particularly for specialized professions, and may have a tougher time filling jobs at prevailing wage rates. If hiring continued to proceed well above its trend level, workers should finally see the long-awaited pickup in wages and salaries, which could also lure some nonparticipants back into the labor force.

Second, for governments that rely heavily on income taxes, a lower trend in job growth would suggest slower growth in tax revenues, particularly as it implies weaker economic growth more broadly. The weaker rate of growth in tax revenues would make it more difficult to meet entitlement expenditures promised under more optimistic scenarios. A persistently lower rate of labor force participation would also challenge tax revenues, as a smaller share of the population is engaged in the economy and earning income that would generate tax revenues. In turn, it would lead to a higher dependency ratio, not just for the population of typical working age, and suggest a greater share of working-age adults are dependent on the earnings of family, charity, or government support programs. This in turn could lead to higher taxes for businesses and individuals to fund entitlement programs and other government spending.²¹

Third, establishing appropriate monetary policy would also hinge on understanding how trends in job growth can change over time. If Federal Reserve policy makers overestimate the trend in payroll growth, they risk ignoring signs that the labor market may be overheating. In contrast, underestimating the trend in payroll growth may lead to the Fed’s tightening too early in a business cycle. With the labor market constantly fluctuating and made up of a heterogeneous workforce, finding this balance remains no easy task. Fluctuations around the unemployment rate and the dynamic estimates of what constitutes full employment make it difficult for policy makers to determine when the labor market is in equilibrium.