Report | Budget, Taxes, and Public Investment


Briefing Paper #87

Download PDF


by Jared Bernstein, Edie Rasell, John Schmitt, and Robert Scott

The strong economy of the past few years has lifted the fortunes of most working families. At the bottom of the wage and income scales, workers whose living standards declined steadily over the past 25 years have benefited from both the tight labor market and the increase in the minimum wage. At the top of the income ladder, the wealthy have continued to enjoy wage and income growth and have reaped huge gains from the growth in the stock market.

But what about the middle? Neither the minimum wage nor the stock market does much for workers in middle-income families.1 The tight labor market has helped those in the middle class, but the gains have been modest compared to the benefits received by those at the bottom.

From 1989 – the year the economy peaked before slipping into the early 1990s recession – to 1993, the income of the median family fell. It began to rise in 1994, and by 19972 it stood at $44,468 – about $300 (in 1997 dollars) above its 1989 level. Yet middle-class hourly wages have been stagnant over this period, and so the gains that have accrued to middle-class families have come mostly from working more hours. While the 1998 data will surely reveal further income growth for the median family, living standards have not improved as much as the highly touted nature of recent economic growth might suggest.

Three key features of the labor market have continued to buffet the progress of middle-class families even as the economic tide rises: wage trends for workers in the middle of the wage distribution, jobs in the manufacturing sector – an important generator of middle-class jobs – and the availability of employer-provided health insurance. Trends in these three areas have rendered the following consequences for families in the middle:

  • Despite gains over the past two and a half years, the inflation-adjusted hourly wages of middle-wage men were lower in 1999 than in 1989, the year of the previous business cycle peak.
  • Middle-wage women workers fared better than men over the business cycle. Nevertheless, by the middle of 1999, the median female worker’s wage was only 3.4% above its 1989 level.
  • The 62% of the workforce with just a high school degree or some college in 1998 has yet to reap the benefits of the high productivity growth that many think heralds a new economy. Real wages for high-school-educated men in this group were 2.7% lower in 1999 than in 1989, while men with some college had wages that were 0.5% lower in 1999 than in 1989.
  • Women with a high school degree or some college experienced little real wage growth until 1996. By mid-1999, their wages were 3.9% above their 1989 level for those with high school degrees and 2.2% higher for those with some college.
  • Middle-income families are working more hours than ever before. The typical married-couple family with children, for example, worked 256 more hours per year in 1997 than in 1989.
  • The share of high-school-educated workers with employer-sponsored health insurance coverage fell from 72.1% in 1989 to 69.5% in 1997. For those working full time, year round, the share with health insurance through their own employer fell by 5.3 percentage points between 1989 and 1997, from 66.0% to 60.7%.
  • The manufacturing sector remains an important source of middle-class jobs for non-college-educated workers (especially men), but the Asian crisis and the ensuing growth in the U.S. trade deficit have contributed to the rapid loss of manufacturing jobs – 491,000 since March 1998, or 2.6% of total manufacturing employment.
  • Productivity – a broad measure of growing economic efficiency – grew by 12% from 1989 to 1998. Despite their contribution to productivity growth, many middle-class workers continue to experience wage growth that lags behind the overall economy.
  • Since the labor market problems facing middle-class families are felt before any taxes are taken out of the paycheck, cutting the federal income tax is not a sufficient solution. Moreover, the share of income the Congressional Budget Office estimates that middle-class families pay in federal income taxes – 5.4% – is too small to make tax cuts an effective policy tool.

Economic policies can help to remedy these negative trends afflicting middle class families. First, the Federal Reserve should continue to resist raising interest rates so that the current pattern of positive, noninflationary wage growth can continue. Second, the Fed and the U.S. Treasury should work to lower the value of the dollar, thereby restoring the competitiveness of U.S. goods on world markets. Third, the U.S. should strive to provide health insurance coverage to all Americans through a publicly financed insurance system.

Middle-wage males lose ground on wages
The real wages of middle-wage male workers fell through most of the current business cycle since the last peak in 1989 (Figure A; see the data appendix for actual wage levels, in 1998 dollars, and a description of data sources and methods). By 1996, the 40th-percentile male wage was 8.6% below its 1989 level, while the median and 60th-percentile wages were down more than 7%.3 Since 1996, real wages reversed course for middle-wage males, but through the first half of this year they still remain below their 1989 level. Thus, eight years into the current recovery, the middle core of the male wage distribution has yet to recover from the early 1990s recession.

Figure A

Middle-wage females, on the other hand, have experienced some wage growth over this business cycle. As Figure B shows, real hourly wage growth was flat for female workers through 1996, when their wages began to grow. By the middle of 1999, the 40th, 50th, and 60th percentiles were 3-5% above their 1989 levels. While real wages grew sharply over this latter period, the rate slowed in 1999. It is too soon to tell whether this is a temporary pause in the pattern of increasing real wages or a return to the stagnant pattern that prevailed throughout the earlier period of the recovery.
Figure B
Slower growth in men’s wages than in women’s since 1989 led the gender wage gap to shrink over the period. Measured at the median, the female/male ratio grew from 73.1% in 1989 to 77.0% in the first half of 1999. About a third of this 3.9 percentage-point decline in the gap is attributable to median male wage decline; the rest is accounted for by gains among women.

One reason that the incomes of working families have held their own in the 1990s despite these wage trends is that middle-income families are working more hours than ever (Table 1). In 1997, the typical married-couple family with children put in 256 more hours of work – six additional full-time weeks – than it did in 1989. In fact, families in each of
the middle three income fifths increased their hours significantly between 1989 and 1997.

Table 1

Median wages by state: Table 2 shows inflation-adjusted median wages for male and female workers combined, by state, from 1989 to 1998; for the 10 largest states, the data go through the first half of 1999.4 Real wages were lower for mid-level earners through the first half of 1999 in the Northeast (down 2.7%) and West (down 2.8%). In the Northeast, median wages were flat or falling in every state except Rhode Island (up 3.2%). Despite fairly strong growth over the first half of this year, the real median wage in New York was 3.3% lower in 1999 than 10 years earlier.

Table 2

The Midwest has seen the fastest median wage (and income) growth in the country in recent years. The median wage grew 6% there through the first half of 1999, and wage growth was positive in almost every state in the region. Sample sizes in most Southern states were too small to derive first-half estimates, but through 1998 median wages had risen strongly in the Carolinas (5.5% and 7.4%), but less quickly in other mid-Atlantic states. Most states in the deeper South experienced fairly large gains in median wages, but from wage levels significantly lower than those in other states.

Relatively large median wage losses in California (-7.2% through the first half of 1999), Oregon, and Arizona (-7.1% and -6.9% through 1998) drove the 2.8% decline in median wages in the West. Median wages grew significantly through 1998 in Idaho, Colorado, and New Mexico.

Wages by education: Most workers – 62% in 1998 – have just a high school degree or some college but lack a four-year degree (27% had a college or advanced degree and 11% did not complete high school). Not surprisingly perhaps, the pattern of wages for this group resembles those of median workers. For males with a high school degree or some college (Figure C), real wages fell through 1996 and then rose. By mid-1999, however, the real wages of high-school-educated males remained 2.7% below their 1989 level, while those for males with some college were 0.5% lower.

Non-college-educated females (Figure D) experienced no real wage growth until 1996, when their wages began to rise. By mid-1999, wages were 3.9% above their 1989 level for women with high school degrees and 2.2% higher for those with some college. Gains over the business cycle have been modest.
Figure C and Figure D

Entry-level wages: Entry-level workers have done particularly well in the last few years, but, for the most part, these increases have not entirely made up for ground lost in the early and mid-1990s. Table 3 shows the wages for young high school and college graduates from 1989 through 1998 (we do not have data for 1999 for this group). From 1996 to 1998, real wages for these entry-level workers grew at an annual rate of 2.8% per year (high school) and 4.2% per year (college). Nevertheless, over the full decade, the wages of entry-level high school workers have fallen an average of -0.6% per year, even after the recent growth. For entry-level college workers, the recent strong growth in wages was sufficient only to return wages in 1998 to where they were in 1989. Despite doing the right thing – going to college – this group has not shared in national productivity growth over the current business cycle. (Wage trends for entry-level workers differ little between men and women. The data in Table 3 do show, however, that even among entry-level workers with similar educational backgrounds, men continue to earn substantially more than their female counterparts.)

Table 3

Wages and productivity: In 1996, for the first time in the 1990s recovery, median wages began to move in tandem with productivity growth (Figure E). The gap between the two is still striking: from 1989 through the first half of 1999, productivity grew about 14%, but the median wage (for men and women) had risen by less than 2%.

Figure E

Falling insurance coverage lowers
middle-class job quality

Employer-provided health insurance coverage is a key component of living standards for American workers and an important indicator of job quality.

Health insurance: Employer-sponsored health insurance coverage (received through either a worker’s own employer or the employer of another family member) has declined since 1989 (Table 4). Among 18- to 64-year-old workers with a high school diploma, coverage – whether through a worker’s own employer or that of the worker’s spouse – fell from 72.1% in 1989 to 69.5% in 1997 (the most recent year for which data are available). The fall-off in coverage has actually been even greater than these numbers indicate, since changes in the survey in the mid-1990s increased the share of workers reporting employer-sponsored coverage. While the improved survey more accurately reflects the reality for American workers, it also understates the decline in coverage that occurred over the 1990s. If the survey questions had been unchanged, the 1990s decline would be roughly 4.6 percentage points, not the 2.6 points shown in the table. Among workers with some college and for those with a college degree, the survey shows little change in coverage.

Table 4

Coverage received by workers through their own employers can serve as a measure of job quality, and, by this measure, job quality for middle-wage workers has suffered: the decline in health insurance received through a worker’s own employer has been even greater than the overall decline in coverage. Among 18- to 64-year-olds working full time, year round, those with a high school diploma saw a 5.3 percentage-point decline in coverage between 1989 and 1997, from 66.0% to 60.7%. (There was no change in these survey questions, so these data reflect workers’ actual experience.) The decline was 3.8 percentage points for those with some college.

The Asian crisis and the loss of manufacturing employment
Part of the explanation for the paradox of a robust economy and meager wage and benefit gains for middle-wage workers lies in the Asian crisis, and the resulting loss of manufacturing employment.

Manufacturing provides middle-class jobs and a channel of upward mobility for non-college-educated workers (especially men). Compensation is higher and fringe benefits (such as health insurance and pension coverage) are more common than in other industries that, like manufacturing, employ non-college graduates. Blue-collar workers in manufacturing are also more likely to be union members, and thus they have more bargaining power than do comparable workers in services.

While the share of jobs in manufacturing has been declining for many years, the Asian crisis and the ensuing growth in the U.S. trade deficit have accelerated this job loss. Despite the continued overall strength of the economy, the U.S. lost 491,000 manufacturing jobs from March 1998 to June 1999, or 2.6% of total manufacturing employment (Table 5). Job losses have been widespread across manufacturing industries, and a few secto
rs have been particularly hard hit. The textile and apparel industries have lost 151,000 workers, continuing a trend of escalating job losses over the last five years since the North American Free Trade Agreement and World Trade Organization agreements went into effect. Surprisingly, large job losses have also occurred in several core high-technology industries such as computers and electronic equipment (132,000 jobs lost), industrial machinery and equipment (93,000 jobs), and aerospace (33,000 jobs). Other manufacturing industries, such as steel (12,000 jobs lost), have also been hard hit.

Table 5

These recent losses result from sharp increase in U.S. imports in 1998 and 1999. U.S. imports of goods and services increased nearly $60 billion between 1997 and 1998, and through May 1999 the goods and service deficit was 52% higher than in the same period last year. These figures suggest that the full impact of the global financial crisis on the U.S. trade deficit could exceed $100 billion in 1999.

The International Monetary Fund recently forecast that the U.S. current account deficit (the broadest measure of the trade balance) would reach $310 billion in 1999 (IMF 1999).5 The U.S. can expect to lose another 300,000 to 500,000 manufacturing jobs as a result, even if the economy continues to expand at its current pace in 1999.

Can income tax cuts help the middle class?
Some policy makers argue that cutting the federal income taxes paid by middle-income families is the best way to deal with their stagnant living standards. This approach, however, has several problems.

First, all of the trends in wages and incomes discussed above have occurred before taxes. Federal income taxes, therefore, cannot be the cause of stagnating and deteriorating wages and incomes, declining health care coverage, or the loss of manufacturing jobs.

Second, as a share of total income, the federal income tax paid by the middle fifth of American families has fallen over the last two decades, from 6.9% in 1977 to just 5.4% in 1999 (according to the Congressional Budget Office). Further tax cuts are likely to be as ineffective in reversing trends today as they were in the past.

Third, the 5.4% share of income that middle-class families pay in federal income tax is already small relative to the problems documented above. Under current tax cut proposals, this share would fall slightly, yielding an average tax cut worth about $278 for middle-income families.6

Finally, a tax cut is, at best, a one-time adjustment that does nothing to change the long-term trends in wages and incomes. After the tax cut, middle-income families will continue to face the same long-term labor market difficulties outlined above.

By most economic indicators, current economic conditions are “as good as it gets.” Unemployment has been below 5% for over two years, and labor productivity grew by 12% between 1989 and 1998. Numerous economic analysts promulgate the virtues of the “new economy,” in which technological advances deliver broad-based growth, benefiting all but the least educated.

Yet, despite their substantial contribution to the growing economy, many middle-class workers are no further ahead now than they were in 1989. The real hourly earnings of middle-wage men and mid-level educated workers (with just a high school degree or some college), though rising since 1996, remain below their 1989 level. Mid-level wages for women have increased, but these wage increases are modest and fall far behind the economy’s rate of productivity growth. Manufacturing jobs are disappearing at an accelerated rate, and the share of non-college-educated workers with employer-provided health coverage – whether through their own or their spouse’s employer – has declined, at least through 1997 (the most recent year of available data).

The booming economy has thus far failed to lift the economic prospects of middle-class workers beyond where they were before the last recession. If an economic boom won’t do it, what are the policies that should be pursued to link middle-class workers to productivity growth?

(1) Keep it going
First, as Figures A-E show, the tight labor market has finally changed the course of wage growth for mid-level earners. For many of these workers, males in particular, these are the first persistent real wage gains in over two decades. Thus, one of the most important lessons of this stage of the current recovery is that full employment is an essential condition for lifting the wages of middle-class workers.

The Federal Reserve has resisted raising interest rates as the unemployment rate has fallen from over 6% down to 4.3%, well below the level many economists thought would be inflationary. The Fed must continue to resist rate increases if middle-class workers are to sustain the wage gains of the past few years and to claim their fair share of growing productivity.

(2) Stop the bleeding
Avoiding unnecessary rate increases can also help to staunch the flow of manufacturing job losses. Instead of raising rates to fight an invisible inflationary threat, the Fed should hold rates steady in order to hold down the value of the dollar. A falling dollar would help to increase the competitiveness of U.S. goods on world markets and improve the trade deficit, thus restoring some of the jobs lost in manufacturing since last year. In addition, the U.S. should also initiate a coordinated effort with other members of the G7 group of leading industrial nations to reduce the value of the dollar.

(3) Cover everyone
Finally, the current employment-based health insurance system is clearly not meeting the health needs of the American middle class. The U.S. should institute a publicly financed health insurance system, like Medicare for all, that is uncoupled from employment and that provides universal care. Such a policy would do far more to improve the living standards of middle-class families than a tax cut.

September 1999

The Economic Policy Institute is grateful to The Ford Foundation and The John D. and Catherine T. MacArthur Foundation for their support of this project.

Data Appendix

The wage data for Figures A-E come from the Outgoing Rotation Group of the Current Population Survey (CPS). The sample includes all wage and salary workers, age 18-64, with positive hourly wages between $0.50 and $100 in 1989 dollars. For hourly paid workers, the reported hourly wage is used; for weekly workers, the hourly wage is constructed by dividing usual weekly earnings by usual weekly hours. Top-coded weekly earnings were replaced with the estimated value of the mean weekly salary above the top-code, using the assumption that the upper “tail” of the distribution follows a Pareto format. Percentile estimates, such as those shown in Figures A and B, use a smoothing procedure to accommodate “clumps” in the reported distribution of earnings. The construction of this wage series is discussed in greater detail in Mishel et al. (1999, Appendix B).

Wage data for the first half of 1999 in Figures A, B, and E and Table A1 are seasonally adjusted, as explained in Bernstein (1999). For state wage data and education wages (Figures C-D), 1999 first-half values reflect the growth rate of these variables between the first half of 1998 and that of 1999. Since this procedure measures the growth rate over the same period (the first six months of both years), seasonal effects are extracted.

Table A1

The CPS education question changed in the year 1992 from “years of school completed” to “highest degree attained.” In order to create a consistent series across this coding change, we use one of the methods suggested by Jaeger (1997). This calls for coding those who have attended but not completed a 13th year of schooling as “some college” in the pre-1992 data. Formerly, we included such cases with high school graduates.

Nominal values are converted to real dollars using the CPI-U.

Health insurance coverage data come from the March CPS.


1. The impact of increases in the minimum wage is small primarily because the median wage was already more than twice the minimum wage in 1998. As for stocks, the average middle-income household owned about $8,000 worth of stock (including stock held in pensions) in 1997, as compared with average holdings of $2.5 million for households in the top 1% (Mishel et al. 1999, p. 270).
2. Income data for 1998 will not be available until fall 1999.
3. The worker at the 40th percentile earns more the 40% and less than 60% of the workforce.
4. Since sample sizes are smaller at the state level, we are unable to derive reliable wage estimates by gender for any except the largest states.
5. The U.S. current account deficit in 1997 was $155 billion and will reach $310 billion in 1999, for a total increase of $155 billion in the wake of the crisis (See the International Monetary Fund’s IMF Outlook for May 1999).
6. Citizens for Tax Justice (August 4, 1999). Its distributional analysis of the House-Senate tax plan shows that the average gain for the top 1% is $46,389. The analysis is available online at


Bernstein, Jared. 1998. QWES (Quarterly Wage and Employment Series). Washington, D.C.: Economic Policy Institute, Vol. 1, No. 1.

Mishel, Lawrence, Jared Bernstein, and John Schmitt. 1999. The State of Working America 1998-99. Economic Policy Institute Series. Ithaca, N.Y.: Cornell University Press.

Jaeger, David A. 1997. “Reconciling the Old and New Census Bureau Education Questions: Recommendations for Researchers.” Journal of Business and Economic Statistics, Vol. 15, No. 3.

Research assistance by Danielle Gao and Lea Bonnecaze; helpful comments were provided by Eileen Appelbaum and Ruy Teixeira.