Journal Issue

Wage Dispersion and Job Growth in the United States

International Monetary Fund. External Relations Dept.
Published Date:
January 1995
  • ShareShare
Show Summary Details

IN THE United States, the gap between the highest and the lowest pay has been getting wider. But job creation has been dynamic. What does this mean for the quality of employment opportunities?

Widening wage differentials in the United States have raised questions about labor market performance and the quality of employment opportunities. Much of the attention stems from concern about the growing phenomenon of the working poor, whereby families with at least one full-time income fall below the poverty line. The number of families in this category rose from 1.6 million in 1980 to 2.1 million in 1990.

Real wages for workers near the bottom of the wage range have dropped, and the incidence of low pay is higher in the United States than in other industrial countries. At the same time, the US labor market has performed well relative to labor markets in other member countries of the Organization for Economic Cooperation and Development (OECD) in terms of rapid job creation in the private sector, and unemployment has recovered to trend levels following the recent recession, despite downsizing by many large companies.

What do these trends mean? There are three issues:

Low wages and new jobs. Rapid private sector job creation in the United States has been accompanied by an increase in wage dispersion to the highest level among industrial countries in the late 1980s. In particular, the relative position of unskilled workers has deteriorated. Skill-based wage differentials have widened less in Western Europe, but the region has experienced little employment growth for more than a decade. These developments suggest a positive link between employment growth and the share of the work force with low-wage jobs.

Regulation and collective bargaining. Downward wage adjustment may have been impeded in some European countries by legal minimum wages or by centralized bargaining practices. This may have ruled out the creation of jobs that were viable only at low wages.

Raising skill levels. If training can transform an unskilled worker into a more productive worker with marketable skills, then training can qualify current low-wage workers for higher-wage jobs, thereby raising and narrowing the lower end of the wage dis-tribution. However, training programs are more likely to provide a long-run solution for income inequality than a short-run remedy for low incomes or unemployment. Public training programs for the unemployed are very expensive and have had, at best, a mixed record of success.

Elaine Buckberg,

a US citizen, is an Economist in the IMF’s Western Hemisphere Department.

Alun Thomas,

a UK citizen, is an Economist in the IMF’s Western Hemisphere Department.

To address wage inequality in the longer run, one option for the United States and other countries would be to focus on improving the quality of basic education. The widely held belief that training is more effective for individuals with a stronger academic foundation suggests that the key to improving job market opportunities for high school graduates could lie in upgrading education at the primary and secondary levels.

The wage gap

Wage dispersion in the United States has been growing in recent decades, with the increase becoming particularly pronounced in the 1980s. Rising wage premiums for education and experience, the decline in the real minimum wage, and the growing share of workers in low-wage jobs have combined to increase the difference between the highest and lowest pay ranges.

Wage dispersion among men, whose labor supply patterns have been more stable than those for women and are therefore more amenable to study, has risen steadily since 1949 as low-wage earners lost ground to higher-income earners. From 1979 to 1989, the wages of the highest-paid men (top decile) grew 10 percent faster in nominal terms than the wages of workers in the middle deciles, while the wages of middle-decile workers grew 16 percent faster than the wages of the lowest-paid men (Chart 1).

Chart 1Widening wage gap Indexed weekly wage of white males1


Source: C. Huhn, “Wage Inequality and Industrial Change: Evidence from Five Decades,” NBER Working Paper No. 4684, National Bureau of Economic Research, March 1994.

1 The indexes are derived from the change in the nominal weekly wages of white men in the United States over ten-year periods.

2 Percentiles refer to percentiles of the wage distribution (nominal weekly wage of white men) where workers in the 99th percentile have the highest earnings.

Workers near the bottom of the income distribution also experienced a loss of real income during the 1970s and 1980s, with the drop in the 1980s amounting to 23 percent for the lowest decile of earners and 17 percent for the second decile. The real hourly wage of recent male high-school graduates and dropouts was 20 percent less in the early 1990s than in the early 1970s. Over the same period, the financial advantage associated with higher education increased. The wage premium for college-educated men relative to male high-school graduates rose by 13 percent during 1979–89.

Technological change. Among the most frequently cited explanations for the widening gap between low- and mid-wage workers, and the explanation that best fits with the rising return to education, is skill-biased technological change. The change in technology would raise the productivity of, and demand for, skilled workers relative to unskilled workers. As a result, assuming unchanged supplies of skilled and unskilled workers, the relative wage of skilled workers would rise.

The skill-biased technological change hypothesis faces two counterarguments. First, because the number of unskilled workers declined and the number of college graduates rose during the 1980s, supply factors alone would suggest declining rewards to education. To have caused the education premium to rise, the technology-induced increase in relative demand for skilled workers must have exceeded the shift in supply. Second, because technological change was not skill-biased in the past, one must explain why its impact on wages was skill-biased in the 1980s.

The simple supply story can be countered by evidence that the composition of employment has changed, even in traditional industries. If one assumed that industries employed college and high-school graduates in the same proportions in 1990 as in 1980, and then used employment growth by industry to predict labor demand, one would predict a decline in the education premium from 1980 to 1990, because the supply of college graduates increased more rapidly than the predicted increase in jobs for them. In the 1980s, however, technological change altered production techniques and the composition of employment economy-wide enough to more than offset changes in labor supply. For example, employment in traditional high-wage, blue- collar industries (manufacturing, transportation, and public utilities) has shifted toward educated nonproduction workers.

Technological progress in the 1980s—the development of information technology and advances in computers—differed from that of other decades in its impact on the wage distribution, in that it affected white-collar workers. Previous technological developments primarily affected production processes and production workers. One indicator of the kind of worker that has been affected by new technology in the 1980s is the use of computers in the workplace, which is highly correlated with education: 67 percent of college graduates worked with computers in 1989, compared with fewer than 10 percent of workers without high-school diplomas. Moreover, many newly created jobs require computer literacy: of 1.7 million new jobs created in 1993, 1 million were connected with information technologies.

Trends outside the United States. Other industrial countries have experienced similar tendencies toward increased wage differentials, although both their levels of dispersion and rates of increase were lower than in the United States. A comparison of dispersion between wages at the median and at the bottom end of the wage scale seems to explain the larger variance of wages in the United States relative to other countries. The median US worker earns a wage equal to 2.8 times the wage of workers in the bottom decile. In contrast, in the Group of Seven (G-7) countries and Australia, all other countries, with the exception of Australia, have ratios below 2 (Australia’s ratio is 2.15). But if we look at inequality between workers at the top and in the middle of the wage scale, the United States is similar to the other G-7 countries—the ratio of incomes between the 90th percentile and the median is in the middle of the range across countries.

During the 1980s, the wage premium for higher education rose in other industrial countries as it did in the United States. The return to education rose in four of the five other OECD countries for which comparable data are available: Australia, Canada, Sweden, and the United Kingdom. However, the education premium rose significantly faster in the United States than in any of the other countries—by 11 percent over five years; the next highest increase in the education premium during this period was 8 percent in the United Kingdom.

One reason why wage differentials have widened less in other economies may be the reliance of the European countries and Australia on collective bargaining, which tends to limit wage divergence within firms and occupations. Australia, Germany, and the Nordic countries (among others) all use centralized bargaining, which sets industry or economy-wide wages for each job description. Minimum wages for each pay group (which apply to both union and non-union members) are determined as part of the collective bargaining process, and the wage floors typically are higher, and thus the wage range is more concentrated, than in the United States. Moreover, the size of the union sector in other countries makes it difficult for non-union employers to attract workers at wages much below union levels.

Wages and jobs

Statistics show that wage-dispersion levels across countries are positively related to job- creation rates. In the United States, downward real wage flexibility in the 1980s and early 1990s may have been important in generating new employment, particularly among the low-wage job categories. Australia, Canada, and the United States have all experienced strong private sector employment growth since the 1980s; at the same time, wage differentials widened and real wages declined. Newly created jobs in these countries were disproportionately low-wage: for example, of net new service and manufacturing jobs created in the United States during 1979–93, 63 percent were in sectors offering wages 20 percent or more below the average. In contrast, countries of the European Union maintained relatively stable wage distributions but experienced only weak employment growth, with most new jobs being created in the public sector.

Chart 2 illustrates the positive relationship between the standard deviation of earnings (a measure of how dispersed earnings are around the average) and job creation for six industrial countries. Germany, Italy, and the United Kingdom have had relatively low wage dispersion, with standard deviations of wages below 50 percent, and experienced low-to-negative job growth (net) from 1979 to 1993. Australia, Canada, and the United States have had substantially higher wage inequality and generated more new jobs.

Chart 2Wage dispersion and employment growth in six industrial countries1


Sources: IMF, World Economic Outlook, 1994; Australian Bureau of Statistics; and F. Blau and L. Kuhn, “International Differences in Male Wage Inequality: Institutions Versus Market Forces,” NBER Working Paper No. 4678, National Bureau of Economic Research, March 1994.

1 For Australia, nonfarm employment growth, 1983–93. Standard deviation of wages for various years over 1985–89.

2 Standard deviation of wages (percent).

While the European G-7 countries (France, Germany, Italy, and the United Kingdom) experienced only single-digit labor force growth during 1979–83, their employment growth equaled less than half of labor force growth. Germany and the United Kingdom actually experienced net reductions in employment. In Australia, Canada, and the United States, job creation more closely matched labor force growth, which was greater than in Europe owing to high immigration. Job creation in the United States exceeded 20 percent, and, even excluding low-wage jobs, the United States generated a 7 percent increase in jobs. Newly created premium-wage jobs in the service sector (1.7 million, net) made up for most of the premium-wage jobs lost in manufacturing (1.9 million, net).

As the incidence of low-wage employment has increased, the US Government has sought to address the problem of the working poor through policies to supplement the income of working families whose wage incomes leave them below the poverty line. In European countries, policies such as legal minimum wages and collective bargaining have served to limit the downward movement in wages. A few authors suggest that these policies have discouraged the creation of low-wage jobs and rationed employment, and thereby may have contributed to rising structural unemployment rates in Europe.

Does training help?

The US Government has advocated job training to address declining opportunity for workers at the bottom of the wage distribution. The rationale for these programs is that if wages are a function of a worker’s productivity, then the most direct way to address low- wage employment is to qualify workers for higher-wage jobs by enhancing their skills. Moreover, insofar as unemployment in the United States has become increasingly structural, reflecting an incompatibility between the characteristics of the unemployed (mostly unskilled workers) and the needs of employers, policies to reduce the structural unemployment rate would need to provide the unemployed with the skills that employers demand. Among the Government’s new initiatives are programs to provide high-school graduates with marketable skills, as well as programs designed to reorient federal unemployment assistance away from income support and toward training and job placement.

Although the Government’s strategy seems well suited to the problem, two questions arise: do active labor market policies succeed in raising workers’ wages and reducing unemployment; and can the Government finance enough training to address low-wage employment economy-wide? The evidence on the success of training programs is mixed. Estimates of the total cost of providing enough training to reverse wage dispersion trends are staggering.

The experience of other countries indicates that active labor market programs have achieved mixed success in controlling unemployment where the policies were targeted at those workers already unemployed. Of the five OECD countries (Austria, Finland, Germany, New Zealand, and Sweden) that spent the highest sums per unemployed worker on manpower programs (such as training, recruitment subsidies, and mobility assistance) in 1987, only three held unemployment below the OECD average of 8.2 percent over 1989–93. Four of the five did, however, succeed in holding the share of the long-term unemployed in the total unemployed below the OECD average of 32.9 percent (see table).

The relationship between spending on active labor market policies and unemployment rates needs to be interpreted with caution. From these statistics, one cannot distinguish the extent to which manpower programs control unemployment by qualifying workers for, and placing them in, private sector jobs and the extent to which unemployment is resolved by relying on the public sector as the employer of last resort.

Micro-level studies of training programs help clarify the relationship between spending on manpower programs and improved wages and employment prospects for the unemployed. One recent study of US retraining programs run by the states found that program graduates’ wages rose, on average, by 8 percent for men and 34 percent for women. A study of a federal retraining program targeted at displaced manufacturing workers, however, suggests that training does not necessarily deserve the credit for wage gains. After controlling for demographic and experience characteristics, the study found that trainees and nontrainees were equally likely to be employed and earned equal wages on average.

The outlook for improving labor market opportunities through active labor-market programs directed at young people is brighter. A recent study has found that on-the-job training and retraining programs for displaced workers aimed at young workers produce higher wage increases than programs that include workers at a later stage in their careers. Moreover, cross-country evidence shows that inequality has risen less in those countries that more closely integrate high-school education and the transition into the work force. Among OECD countries, wage inequality rose most in the United Kingdom and the United States, which have the most decentralized wage setting, provide few levers to help youths make the transition from school to the workplace, and experienced declines in union power during the 1980s. More government intervention in labor markets in France, Italy, and Sweden prevented increases in inequality during at least part of the 1980s, but may have caused higher unemployment instead.

Are labor-market programs cost-effective?



Long-term unemployed

as percentage

of total


Expenditure on “active”

labor-market programs

per unemployed person,

expressed as a percentage

of GDP per worker1

New Zealand9.231.913.1
United Kingdom8.128.14.6
United States6.311.22.4
Sources: IMF, World Economic Outlook, 1994; R. Layard, R. Jackman, and S. Nickell, Unemployment: Macroeconomic Performance and the Labor Market, 1990; and The OECD Jobs Study, 1994.

Per worker GDP is defined as GDP divided by the number of people in the labor force.

Sources: IMF, World Economic Outlook, 1994; R. Layard, R. Jackman, and S. Nickell, Unemployment: Macroeconomic Performance and the Labor Market, 1990; and The OECD Jobs Study, 1994.

Per worker GDP is defined as GDP divided by the number of people in the labor force.

The high cost of training may pose a greater threat to the US Government’s strategy. The cost of reversing recent wage-dispersion trends has been estimated by assessing the rate of return, or wage gain, to workers who receive training. Studies show that the average rate of return to investment in human capital is 10 percent, although the return on government training programs is estimated to be considerably lower. Therefore, to raise a worker’s wages by $1,000 per year, a $10,000 training investment is required. On the basis of a 10 percent return, providing training to restore the 4 percent real income loss of male high-school graduates from 1979 to 1989 would require an investment of $10,000 per graduate, or $212 billion in 1989 dollars; to restore the 13 percent real income loss of high- school dropouts would require an investment of $25,000 each, or $214 billion in 1989 dollars. To put these numbers in context, the fiscal year 1989 budget appropriated $36.7 billion for education and training.

Which strategy?

The problem at the source of rising wage inequality is declining job-market opportunities for workers without advanced education. These workers are increasingly limited to low- wage jobs that require unskilled labor. Improved opportunities for today’s low-wage workers and the unemployed, particularly for young people, might be achieved through vocational skills training programs, although the success of such programs has been mixed. Addressing wage dispersion in the long run, however, may require addressing skill levels at a much earlier stage. Rethinking curriculums in primary and secondary schools could potentially produce graduates with better labor force skills, irrespective of whether these might be trade skills or communication and analytical skills.

Buckberg Elaine and Alun Thomas

Other Resources Citing This Publication