Monday, May 23, 2016

What Does the Unemployment Rate Measure? Labor Market Slack or the Social Stress of Joblessness? Why Does it Matter?

The unemployment rate published monthly by the Bureau of Labor Statistics is one of the most widely watched of all economic indicators. But why? What does it really measure?

The news media, politicians, and voters tend to see the unemployment rate as an index of the social stress of joblessness. There is ample evidence to support that view. Researchers have linked high unemployment rates to increased mortality and impaired mental health. A recent cross-sectional analysis shows a strong relationship between unemployment  and suicide. There is evidence that unemployment undermines personal relationships, although the impact on divorce rates is ambiguous: A weak job market can break up some couples while leaving others stuck in bad marriages because they can’t afford divorce. Rising unemployment is also associated with higher crime rates, for both violent and property crimes. There is even something called the misery index, which is the sum of the unemployment rate and the inflation rate.

Economists, on the other hand, more often view the unemployment rate as an indicator of economic slack. People who want to work but are not working are a wasted resource. Finding jobs for them would add to GDP. Monetary and fiscal policymakers watch employment indicators closely because the more slack there is in the labor market, the more room there is for economic stimulus without risk of inflation.
The questions we need to address, then, are, first, can the standard unemployment rate do double duty as an indicator of both macroeconomic slack and social stress? And if not, what might be better?

Limitations of the standard unemployment rate

The standard unemployment rate, known as U-3, has well known limitations both as a measure of slack and of social stress.

U-3 is an inadequate measure of slack because it counts people as “unemployed” only if they are not working but have actively looked for work in the previous four weeks. That leaves out two important categories of willing and available labor.

The first consists of people who are not working and say they want a job, but do not count as unemployed because they have not looked for work recently enough. As of April 2016, 5.6 million people, or 2.2 percent of the adult population, fell into this category. Of these, 1.7 million who had looked for work within the past year, but not within the past month, were counted as marginally attached to the labor force. Among the marginally attached, 695,000 people, known as discouraged workers, reported that they had not looked for work because they thought no jobs were available.

A second category of labor that is unused but willing and available to work consists of people working part-time for economic reasons (PTER). These people, sometimes called “involuntary part-time workers,” work less than 34 hours per week because they can’t find full-time jobs or because their employers have cut their hours. As of April 2016, there were 5.8 million PTER workers. (They are distinct from the 21 million people who voluntarily worked part-time because of school, family responsibilities, retirement, or other “noneconomic” reasons.)

At the same time, the official unemployment rate in some ways overstates the social stress associated with the inability to find work. In part that is because a third of workers are unemployed for less than 5 weeks. Those include many who have voluntarily quit their previous jobs in the expectation of finding something better.

Other unemployed workers have adequate economic resources from savings, retirement income, or other family members. A 2014 poll conducted by the Kaiser Family Foundation, together with the New York Times and CBS found that of those who considered themselves unemployed but able to work, 36 percent reported no stress and another 20 percent only minor stress. That suggests that at least some of the unemployed want a job not for pressing economic reasons, but because they seek social interaction or professional fulfillment of some kind.

About a quarter of the nonemployed in the Kaiser survey classified themselves as “homemaker.” Of those, 43 percent said they want a full- or part-time job now and another 38 percent said they wanted a job in the future. However, 77 percent of nonemployed homemakers said that their lack of employment was not a source of stress, while another 15 percent said it was only a minor source of stress.

Alternative measures of joblessness

The shortcomings of the official unemployment rate have spurred numerous attempts to develop alternatives. The BLS itself publishes five alternative measures that attempt to meet various criticisms of the standard U-3 unemployment rate.

Two of these address the criticism that joblessness is not a source of economic stress for all unemployed workers. A measure called U-1 excludes people who have been out of work for less than 15 weeks. The measure U-2 takes includes people who have lost their jobs or been laid off, but not those who have voluntarily quit or are newly entering the job market.

Three other indicators focus on  reserves of willing but unused labor that are missed by U-3. U-4 adds discouraged workers. U-5 more broadly adds all marginally attached workers. U-6 expands U-5 by adding people working part-time for economic reasons. The following chart shows the complete family of BLS measures. 

Of these alternative measures, the only one that gets much publicity is U-6. That measure is frequently referred to as the “broad” unemployment rate. Because it shows the largest number of unemployed, writers who operate on the principle that bad news attracts readers sometimes like to call it the “true” or “real” rate.

Even U-6 is not bad enough for some people. John Williams of ShadowStats promotes his own alternative unemployment rate, which stood at 22.9 percent as of April 2016. To arrive at that number, he augments U-6 with “long-term discouraged,” whom, he says, have been “defined out of existence” by the BLS. Even though I find Williams’ estimates to be implausible high (see this earlier post ), he is right to think that there are untapped reserves of labor that lie outside the “marginally attached” group that enter into U-5 and U-6. That brings us to the next alternative measure.

The Hornstein-Kudlyak-Lange Non-Employment Index

One of the most useful alternative measures of joblessness is the non-employment index (NEI), based on research first published by Richmond Fed economist Andreas Hornstein, San Francisco Fed economist Marianna Kudlyak, and McGill University economist Fabian Lange. The Richmond Fed now provides monthly updates of the NEI.

The creators of the NEI maintain that U-3 defines unemployment too narrowly by excluding some reserves of willing but unused labor. At the same time, they argue that alternative measures like U-6 and the ShadowStats index overstate labor reserves by giving equal treatment to everyone who wants a job, but does not have one. Their solution is a weighted index of labor reserves that recognizes that not all nonemployed workers are equally attached to the labor force.

Specifically, the NEI uses weights based on transition rates, that is, the probability that workers in each labor force category will make the transition from nonemployment to employment in any given month. For example, in a good year like 2007, about 28 percent of people who have been out of work for 26 weeks or less find a job each month. Even in a bad year, such as 2010, the transition rate for the short-term unemployed is 22 percent.

At the other end of the spectrum, the transition rate for people who classify themselves as retired and say they do not want a job is only 1.5 percent a month, and does not vary much from good years to bad.
For people between the extremes of the short-term unemployed and the retired, transition rates vary less than one might think. Consider these four categories of workers:
  1. The long-term unemployed are those who have been out of work for more than 26 weeks, but who have actively looked for work within the last month. They are included in U-3. Their monthly transition rate is 14.4 percent.
  2. Discouraged workers (those who have looked for work within the past year but not the past four weeks and who give lack of job prospects as the reason for not working) are included in U-4, U-5, and U-6, but not U-3. Their transition rate is 13.1 percent.
  3. Other marginally attached workers, who are included in U-5 and U-6, but not U-4, have a transition rate of 12.7 percent.
  4. People who say they want a job, but are not counted as marginally attached because they have not looked for one in more than a year, have a transition rate that is close to that of the long-term unemployed—14.5 percent per month. Yet they are not included in any of the BLS measures.
In short, when it comes to the strength of labor force attachment, as measured by transition rates, the categories of “marginally attached” and “discouraged” provide little useful information. The NEI recognizes this by assigning weights to all four of these groups in proportion to their (nearly equal) transition rates.
The Richmond Fed also publishes a variant of the NEI that includes people working part-time for economic reasons. The next chart compares the NEI, with and without PTER workers, to the official U-3 unemployment rate and to the roughly comparable BLS alternative rates, U-5 (which excludes PTER) and U-6 (which includes PTER).

Looking first at the two series that exclude PTER workers, we see that the NEI lies above U-5 in most years. That is not surprising, since the NEI includes people who want a job but who have not looked for work for more than a year. Such workers do not fit the BLS definition of “marginally attached," so they are left out of U-5. The NEI, in contrast, gives them approximately the same weight as marginally attached workers because they have about the same average transition rate.

Adding people who are working part-time for economic reasons transforms U-5 into U-6 and the NEI into the NEI+PTER. In U-6, PTER workers count the same as those in other categories. In NEI+PTER, they get a weight of 30 percent, which is the average monthly transition rate for PTER workers from part- to full-time work. That happens to be about the same as the transition rate for short-term unemployed workers.

One feature that stands out in the chart is that both NEI variants are less volatile than ay of the BLS measures. As a reference point to highlight the difference in volatility, we can use June 2005, which was the month when the official unemployment rate first fell to 5.0 percent as the economy recovered from the 2001 recession. In April 2016, U-3 again stood at 5.0 percent. U-5 and the NEI were also back at their 2005 levels in that month, while U-6 and the NEI+PTER were only fractionally higher than they were 11 years earlier. In between, however, the various series behaved differently. At their peaks for the Great Recession, U-3 hit a level that was double its June 2005 level, and U-5 and U-6 rose to 190 percent of their June 2005 values. In contrast, even at its peak, the NEI reached only 130 percent of its June 2005 level, and the NEI+PTER peaked at just 136 percent.

That lower volatility is one of the factors that leads the developers of the NEI to see it as a better measure of labor market slack than either the standard or alternative BLS indexes. Their research suggests that the BLS measures, including U-3, overstated the weakness of the labor market from 2008 through 2013. Seen through the lens of the NEI, the Great Recession was still the worst downturn in recent US history, but not as bad as official labor market data indicate.

Toward a better measure of the social stress of joblessness

The NEI was designed to be a measure of labor market slack. As such, it arguably does a better job than either the broad and narrow BLS unemployment measures. However, that does not automatically make the NEI a good measure of the social stress of joblessness.

One reason is that the use of transition rates as weights gives more importance to short-term than to long-term unemployment. That makes sense if we want to measure how quickly additional macroeconomic stimulus could draw unemployed workers back into the labor market. Among other things, the higher transition rates of the short-term unemployed presumably mean, that they have skills that more closely match the needs of employers. However, in terms of human suffering, anecdotal evidence suggests that the long-term unemployed, including those whose skills poorly aligned with employer needs, are among those who were hit hardest during the Great Recession.

While the NEI has recovered to its prerecession level, and while the percentage of the labor force unemployed for 5 weeks or less is now lower than before the recession, measures of long-term unemployment remain elevated. The median duration of unemployment reached a prerecession low of 16.3 weeks in 2007 and peaked at 40.5 weeks in 2011. It has fallen as the economy has recovered, but in April 2016, it still stood at 27.7 weeks, well above its prerecession low. That suggests that a significant amount of the social stress of long-term unemployment persists.

A second reason why the NEI may be a better measure of labor market slack than of social stress is that it assigns the same weight to job losers as it does to job leavers and new entrants. Job loss is classically among the top five sources of personal stress and anxiety, along with death of a loved one, divorce, major illness, and moving. In contrast, newly entering the labor market or quitting a job in the expectation of finding a better one, far from being a source of stress, is likely to be the prelude to an improvement in one’s personal economic situation.

The rarely noticed U-2 unemployment indicator, which excludes job leavers and new entrants, is helpful here. In April 2016, U-2 stood at 2.4 percent, almost, but not quite back to its prerecession low of 2.1 percent. However, while U-3 doubled at its peak and the NEI rose just 30 percent, U-2 was three times higher at its peak (2009) than it had been in 2006. In that sense, both U-3 and the NEI understated the increase in social stress incurred by job losers during the Great Recession.

In another respect, though, both the NEI and U-6 may overstate the social stress arising from changes in labor market conditions. Both of those indicators give the same weight to PTER workers as to those who are fully unemployed. (U-6 does not use weighting at all, and the NEI puts the same weight on PTER workers as on the short-term unemployed because they have similar transition rates.) But in practice, people who say they are working part time for economic reasons put in an average of 22 hours a week, about half the average of full-time workers. Certainly, having one or more earners cut back to half time can put a stress on a household budget, but that stress is unlikely to equal to that of losing a job altogether.

What would be nice would be to have is a sound, data-based measure of the social stress of joblessness—an SSJ index, we might call it. This is not the place to construct such an index in detail, but the considerations we have listed suggest how we might approach the task. An SSJ index would, like the NEI, apply weights to the various BLS labor market categories, but they would be different weights. Short-term unemployment would receive a low weight—perhaps even zero for workers out of a job for 5 weeks or less, who currently account for a third of all unemployed. Job leavers also would be assigned low weights, and new entrants would not get much weight unless they failed to find jobs after a long period of search. PTER workers would receive weights in proportion to their average hours worked.

If other data were not available, transition rates might serve as a plausible proxy for social stress for some groups of workers, just as they serve as a proxy for labor market slack in the NEI. For example, we have seen that the transition rates to employment are lower for marginally attached workers than for those who have recently looked for work. The same is true for workers who say they want a job but have not looked within the past year. The fact that people in these categories look less actively and find work less often could, at least in some cases, indicate that joblessness is less stressful to them than it is for the long-term unemployed, who keep looking for work month after month even though they fail to find it.

Ideally, though, an SSJ index would use actual data, rather than proxies, weight to various labor market groups. More detailed survey data would be helpful. For example, the Kaiser Family Foundation survey cited earlier found that 56 percent of all people who were not working, but wanted a job, suffered no stress, or only moderate stress, because of their nonemployment. A similar study that more carefully distinguished among BLS categories (marginally attached, discouraged, PTER, long- vs. short-term, leavers vs. losers, and so on) could potentially provide weights that could be used in a broad index of the social stress of joblessness.

Policy implications

Some readers might ask what difference any of this makes. As the above charts make clear, all indexes of unemployment and nonemployment rise and fall together over the business cycle. Both survey data and anecdotal evidence suggest that subjective social stress goes up and down over the cycle as well. Why, then, isn’t any one number, such as U-3, as good as any other?

The answer, I think, depends on what kind of policy we are looking at. For macroeconomic policy, what matters is labor market slack. Those in charge of monetary and fiscal policy need a measure of labor market slack to know how much room there is to boost aggregate demand without setting off inflationary pressure on wages. For that purpose, it makes good sense to use a measure like the NEI, which has weights that reflect the ease with which different categories of nonemployed workers can be brought back to the labor market.

If we take microeconomic perspective, however, we need to recognize that structural changes in the labor market could cause the social stress of joblessness to vary independently of the business cycle. For example, rapid technological change could cause early obsolescence of job skills. We have seen some hints of this: Even though official unemployment is back down to its prerecession level, painful long-term joblessness and the ratio of job losers to job leavers remain elevated.

Suppose, then, that a hypothetical SSJ index stayed high when the NEI fell. What would be the appropriate policy response? Not, presumably, more macroeconomic stimulus. By assumption, that would only lead to inflation, which itself is a source of social stress. Instead, a rise in the SSJ index relative to U-3 or the NEI would require microeconomic policy remedies. Among other possibilities, those might include greater investment in education and job training, improvements in the social safety net, or even (following current political fashion) changes in trade policy.

The bottom line here is that joblessness is a multi-dimensional problem. The measures that are now most widely publicized—U-3 and U-6—are mixed indicators that do not measure either labor market slack or the social stress of joblessness as well as they might. With regard to labor market slack, the NEI looks like a solid step in the right direction. Hopefully, someone will soon undertake a similar effort to devise a sound, empirically based index of the social stress of joblessness.

First published at

Related post
What Does the Consumer Price Index Measure? Inflation or the Cost of Living?

No comments:

Post a Comment