Few economic indicators are of more concern to Americans than unemployment statistics. Reports that unemployment rates are dropping make us happy; reports to the contrary make us anxious. But just what do unemployment figures tell us? Are they reliable measures? What influences joblessness?

How Is Unemployment Defined and Measured?

Each month, the federal government’s Bureau of Labor Statistics randomly surveys sixty thousand individuals around the nation. If respondents say they are both out of work and seeking employment, they are counted as unemployed members of the labor force. Jobless respondents who have chosen not to continue looking for work are considered out of the labor force and therefore are not counted as unemployed. Almost half of all unemployment spells end because people leave the labor force. Ironically, those who drop out of the labor force—because they are discouraged, have household responsibilities, or are sick—actually make unemployment rates look better; the unemployment rate includes only people within the labor force who are out of work.

Not all unemployment is the same. Unemployment can be long term or short term. It can be frictional, meaning someone is between jobs; or it may be structural, as when someone’s skills are no longer demanded because of a change in technology or an industry downturn.

Is Unemployment a Big Problem?

Some say there are reasons to think that unemployment in the United States is not a big problem. In June 2005, for example, 33.5 percent of all unemployed people were under the age of twenty-four, and presumably few of them were the main source of income for their families. One out of six of the unemployed are teenagers. Moreover, the average duration of a spell of unemployment is short. In June 2005 it was 16.3 weeks. And the median spell of unemployment is even shorter. In June 2005 it was 7.0 weeks, meaning that half of all spells last 7.0 weeks or less.

On the basis of numbers like the above, many economists have thought that unemployment is not a very large problem. A few weeks of unemployment seems to them like just enough time for people to move from one job to another. Yet these numbers, though accurate, are misleading. Much of the reason why unemployment spells appear short is that many workers drop out of the labor force at least temporarily because they cannot find attractive jobs. Often two short spells of unemployment mean a long spell of joblessness because the person was unemployed for a short time, withdrew from the labor force, and then reentered the labor force.

And even if most unemployment spells are short, most weeks of unemployment are experienced by people who are out of work for a long time. To see why, consider the following example. Suppose that each week, twenty spells of unemployment lasting 1 week begin, and only one begins that lasts 20 weeks. Then the average duration of a completed spell of unemployment would be only 1.05 weeks. But half of all unemployment (half of the total of 40 weeks that the twenty-one people are out of work) would be accounted for by spells lasting 20 weeks.

Something like this example applies in the real world. In June 2005, for example, 42.9 percent of the unemployed had been unemployed for less than five weeks, but 16.9 percent had been unemployed for six or more months.

What Causes Long-Term Unemployment?

To fully understand unemployment, we must consider the causes of recorded long-term unemployment. Empirical evidence shows that two causes are welfare payments and unemployment insurance. These government assistance programs contribute to long-term unemployment in two ways.

First, government assistance increases the measure of unemployment by prompting people who are not working to claim that they are looking for work even when they are not. The work-registration requirement for welfare recipients, for example, compels people who otherwise would not be considered part of the labor force to register as if they were a part of it. This requirement effectively increases the measure of unemployed in the labor force even though these people are better described as nonemployed—that is, not actively looking for work.

In a study using state data on registrants in Aid to Families with Dependent Children and food stamp programs, my colleague Kim Clark and I found that the work-registration requirement actually increased measured unemployment by about 0.5 to 0.8 percentage points. If this same relationship holds in 2005, this requirement increases the measure of unemployment by 750,000 to 1.2 million people. Without the condition that they look for work, many of these people would not be counted as unemployed. Similarly, unemployment insurance increases the measure of unemployment by inducing people to say that they are job hunting in order to collect benefits.

The second way government assistance programs contribute to long-term unemployment is by providing an incentive, and the means, not to work. Each unemployed person has a “reservation wage”—the minimum wage he or she insists on getting before accepting a job. Unemployment insurance and other social assistance programs increase that reservation wage, causing an unemployed person to remain unemployed longer.

Consider, for example, an unemployed person who is accustomed to making $15.00 an hour. On unemployment insurance this person receives about 55 percent of normal earnings, or $8.25 per lost work hour. If that person is in a 15 percent federal tax bracket and a 3 percent state tax bracket, he or she pays $1.49 in taxes per hour not worked and nets $6.76 per hour after taxes as compensation for not working. If that person took a job that paid $15.00 per hour, governments would take 18 percent for income taxes and 7.65 percent for Social Security taxes, netting him or her $11.15 per hour of work. Comparing the two payments, this person may decide that an hour of leisure is worth more than the extra $4.39 the job would pay. If so, this means that the unemployment insurance raises the person’s reservation wage to above $15.00 per hour.

Unemployment, therefore, may not be as costly for the jobless person as previously imagined. But as Harvard economist Martin Feldstein pointed out in the 1970s, the costs of unemployment to taxpayers are very great indeed. Take the example above of the individual who could work for $15.00 an hour or collect unemployment insurance of $8.25 per hour. The cost of unemployment to this unemployed person was only $4.39 per hour, the difference between the net income from working and the net income from not working. And as compensation for this cost, the unemployed person gained leisure, whose value could well be above $4.39 per hour. But other taxpayers as a group paid $8.25 in unemployment benefits for every hour the person was unemployed, and got back in taxes only $1.49 on this benefit. Moreover, they gave up $3.85 in lost tax and Social Security revenue that this person would have paid per hour employed at a $15.00 wage. Net loss to other taxpayers: $10.61 ($8.25 − $1.49 + $3.85) per hour. Multiply this by millions of people collecting unemployment, each missing hundreds of hours of work, and you get a cost to taxpayers in the billions.

Unemployment insurance also extends the time a person stays off the job. Clark and I estimated that the existence of unemployment insurance almost doubles the number of unemployment spells lasting more than three months. If unemployment insurance were eliminated, the unemployment rate would drop by more than half a percentage point, which means that the number of unemployed people would fall by about 750,000. This is all the more significant in light of the fact that less than half of the unemployed receive insurance benefits, largely because many have not worked enough to qualify.

Another cause of long-term unemployment is unionization. High union wages that exceed the competitive market rate are likely to cause job losses in the unionized sector of the economy. Also, those who lose high-wage union jobs are often reluctant to accept alternative low-wage employment. Between 1970 and 1985, for example, a state with a 20 percent unionization rate, approximately the average for the fifty states and the District of Columbia, experienced an unemployment rate that was 1.2 percentage points higher than that of a hypothetical state that had no unions. To put this in perspective, 1.2 percentage points is about 60 percent of the increase in normal unemployment between 1970 and 1985.

There is no question that some long-term unemployment is caused by government intervention and unions that interfere with the supply of labor. It is, however, a great mistake (made by some conservative economists) to attribute most unemployment to government interventions in the economy or to any lack of desire to work on the part of the unemployed. Unemployment was a serious economic problem in the late nineteenth and early twentieth centuries prior to the welfare state and widespread unionization. Unemployment then, as now, was closely linked to general macroeconomic conditions. The great depression, when unemployment in the United States reached 25 percent, is the classic example of the damage that collapses in credit can do. Since then, most economists have agreed that cyclical fluctuations in unemployment are caused by changes in the demand for labor, not by changes in workers’ desires to work, and that unemployment in recessions is involuntary.

Even leaving aside cyclical fluctuations, a large part of unemployment is due to demand factors rather than supply. High unemployment in New England in the early 1990s, for example, was due to declines in computer and other industries in which New England specialized. High unemployment in northern California in the early 2000s was caused by the dot-com bust. The process of adjustment following shocks is long and painful, and recent research suggests that even temporary declines in demand can have permanent effects on unemployment, as workers who lose jobs are unable to sell their labor due to a loss of skills or for other reasons. Therefore, most economists who study unemployment support an active government role in training and retraining workers and in maintaining stable demand for labor.

The Natural Rate of Unemployment

Long before Milton Friedman and Edmund Phelps advanced the notion of the natural rate of unemployment (the lowest rate of unemployment tolerable without pushing up inflation), policymakers had contented themselves with striving for low, not zero, unemployment. Just what constitutes an acceptably low level of unemployment has been redefined over the decades. In the early 1960s an unemployment rate of 4 percent was both desirable and achievable. Over time, the unemployment rate drifted upward and, for the most part, has hovered around 7 percent. Lately, it has fallen to 5 percent. I suspect that some of the reduction in the apparent natural rate of unemployment in recent years has to do with reduced transitional unemployment, both because fewer people are between jobs and because they are between jobs for shorter periods. Union power has been eroded by domestic regulatory action and inaction, as well as by international competition. More generally, international competition has restrained wage increases in high-wage industries. Another factor making unemployment lower is a decline in the fraction of the unemployed who are supported by unemployment insurance.


About the Author

Lawrence H. Summers is Charles W. Eliot University Professor at Harvard University. He was previously the president of Harvard University. Before that, he was secretary of the U.S. Treasury.


Further Reading

Feldstein, Martin. “The Economics of the New Unemployment.” Public Interest 33 (Fall 1973): 3–42.
Feldstein, Martin. “Why Is Productivity Growing Faster?” NBER Working Paper no. 9530. National Bureau of Economic Research, Cambridge, Mass., 2003.
Friedman, Milton. “The Role of Monetary Policy.” American Economic Review 58 (March 1968): 1–17.
Hall, Robert. “Employment Fluctuations and Wage Rigidity.” Brookings Papers on Economic Activity 1 (1980): 91–141.
Summers, Lawrence H. Understanding Unemployment. Cambridge: MIT Press, 1990.
Summers, Lawrence H. “Why Is the Unemployment Rate So Very High Near Full Employment?” Brookings Papers on Economic Activity 2 (1986): 339–383.
Summers, Lawrence H., and Kim B. Clark. “Labor Market Dynamics and Unemployment: A Reconsideration.” Brookings Papers on Economic Activity 1 (1979): 13–60.