After almost three decades of rising incomes, average real earnings have fallen relentlessly since 1973. In 1982 dollars, the weekly wage for full-time workers was $327 in 1973, $303 in 1979, $277 in 1982, and just $265 in 1990. During the same period, the income distribution became steadily more unequal. The most dramatic earnings collapse was for poorly educated men, a pattern that accelerated in the 1980s.
The single most widely accepted explanation for the earnings crisis is "skill mismatch." According to this view, there has been a fundamental shift in industry's demand for skills, leading to a collapse in opportunities and wages for low-skill workers. This shift in the demand for skills is widely attributed to techno logical changes. The beauty of this account is its apparent consistency with both the textbook labor market, in which relative wages reflect relative skills, and a wealth of anecdotal evidence on the skill-upgrading effects of computer-based technologies. The Clinton administration, accordingly, has placed heavy emphasis on skill mismatch in promoting its job training initiatives.
However, a review of the statistical evidence casts considerable doubt on the skill-mismatch hypothesis. There is little direct evidence that the rate of skill upgrading was substantially greater in the 1980s than in earlier decades, or that technological change was the main source of the skill upgrading that we can measure. Nor is there evidence that changes in the mix of skills can explain much of the recent growth in either earnings inequality or the share of low-wage jobs.
There is another explanation, one that does not fit neatly into a standard supply-and-demand diagram but that does offer a better fit with the facts. First, there was a marked increase in competitive pressures (globalization and deregulation) and an ideological shift in favor of markets in the late 1970s. In this climate, business strategies and government policies severely undermined traditional wage-setting institutions (collective bargaining, internal labor market norms, minimum wages) that had protected low-skilled workers from the full force of labor market competition. Second, the effective supply of labor competing for low-skill jobs substantially increased; and, consequently, labor lost relative bargaining power.
Several factors contributed to the crowding at the hiring gate. Employment opportunities in the middle of the earnings distribu tion, notably high-wage blue-collar jobs and moderately skilled white-collar jobs, disappeared. Fewer good job opportunities, rising numbers of displaced workers, and sharply rising numbers of low-skill immigrant workers fed the pool of workers competing for moderate and low-skill jobs paying low wages. Blue-collar labor markets also became more competitive than at any time since the early 1930s because of the decline in union membership and the growing capacity of firms to relocate to and buy supplies from low-wage areas.
The result was a collapse in the wage paid for low-skill work. In this alternative view, the main restructuring trend among nonsupe rvisory male workers in the 1980s was not a massive shift away from lower skill jobs. Indeed, the share of low-skill jobs was remark ably stable from 1983 into the 1990s. Rather, the real shift was away from higher wage jobs.
In short, employers in the 1980s responded to increased competitive pressures by taking a "low-road" human resource strategy, one aimed above all at reducing current labor costs. In direct contrast to the technology-induced skill mismatch story, the fundamental problem in the 1980s was that most employers did not follow a high-tech, high-skills path. In a great many industries, workers learned new skills to work with more advanced production technologies -- but their higher productivity was not reflected in higher wages. (See graph above).
Note that technology plays a key role in both accounts. In the skills-mismatch story, workers were not keeping up with the rising skill requirements demanded by new computer-based technologies. In the dissenting account, technology facilitated both globalization and displacement, as well as strategies of contracting out to small, low-wage suppliers. In a context of high unemployment, hostile management and government policies, and weaker unions, these trends all eroded wages. So technology did contribute to wage collapse, but not primarily through skill restructuring.
These contrasting stories, not surprisingly, have political overtones.
If the skill mismatch explanation is correct, the market is only paying workers what they are truly "worth" and is signalling workers to upgrade their skills. More competition in the labor market, as in the market for shoes or computers, is better because competition ensures that resource allocation will be efficient. Since technological change is held to be the main cause of the restructuring, and since few would favor inhibiting the use of new technologies, declining wages become inevitable for those without, say, a good college education or at least a competent technical one.
If the alternative story is right, however, the erosion in wages is mainly the result of an asymmetry in bargaining power, reflecting low-wage strategies by employers, a failure of government to help maintain traditional labor market institutions, and increased wage competition in the external labor market. Growing shares of workers with low wages reflect neither major skill shifts nor the use of new technologies. There is nothing inevitable or "natural" about the growing earnings problems of the two-thirds of the workforce without college degrees.
These contrasting explanations of the wage collapse suggest dramatically different policy strategies. If the earnings problems of low-skill workers are mainly the result of a shift in the demand for skills due to increasing use of information technology, public policy must substantially upgrade the cognitive skills of large portions of the current and future workforce. On the other hand, if the real mismatch in the 1980s was between skills provided and wages paid, improvements in education and training programs per se are unlikely to have much effect on the problem of earnings or earnings-distribution. There is little doubt that American workers need better skills, but a purely human capital approach will do little to reverse the downward shifts in wage norms in the last decade.
Wages and Skills
If the wage reflects the value a worker produces for the employer, and this value reflects the worker's skills, a collapse of wages for a particular skill group reflects a drop in the demand for these skills. But much empirical research since at least the 1940s indicates that the productivity of workers provides only a partial explanation of wages and wage trends.
Relative wages reflect not only skills and education, but a variety of other factors, such as gender, race, union coverage, firm size and industry of employment. One only has to ask, for example, why most jobs dominated by women (child care workers, secretaries, teachers, librarians, etc.) pay substantially less than those domi nated by men despite higher average educational attainment required in the "female jobs." Moreover, "female" jobs in small firms in competitive industries will tend to pay even less over time, even if their relative skill levels have risen. In addition to these easily measured determinants, wage setting is also affected by such hard to measure factors as changes in collective bargaining laws, changes in enforcement of government regulations affecting earnings and conditions of work, and changes in labor-management relations.
Consistent with many earlier studies, recent scholarly work by Lawrence Summers and Lawrence Katz found that similar workers in the same occupations are paid substantially different wages in different industries. My own research with Edward Wolff, using data from the Dictionary of Occupational Titles, found that the wage distribution was not a good proxy for the skill distribution at the occupation or industry level, whether skills are measured by cognitive skills, motor skills or interactive ("people") skills. The link was particularly weak for nonsupervisory occupations. In short, skills help determine individual wage levels, but so do a lot of other factors.
Economists agree that raising real earnings requires raising the rate of productivity growth. As the 1994 Economic Report of the President states, "Periods of rapid productivity growth have been accompanied by increases in real wages." What the report doesn't say is that this relationship did not hold in the 1980s. As the accompanying figure demonstrates, manufacturing wages grew with productivity between 1967 and 1973, continuing a trend that began with the postwar boom. But manufacturing wages have stagnated since 1973, despite productivity growth in the 1980s that compares favorably to anything we've seen since the 1950s. Despite the introduction of productivity-enhancing technology such as optical scanners, the wages of retail sector workers have dropped even more dramatically.
Sources of Inequality
Additional evidence on the imperfect link between wages and skills can be found in the earnings inequality literature. Attempts to account for the unprecedented increase in male earnings inequality throughout the 1980's have naturally turned to measures of skill. If the wage distribution mirrors the skill distribution, persistent growth in earnings inequality within industries should be associat ed with widening skill differentials. But the data show little support for this expectation. Confirming earlier research by economist Lynn Karoly, Susan Wieler, a research associate at the Center for Education and the Economy at Columbia University, found that changes in the variation of educational attainment and work experience do a poor job of accounting for the steady increase in earnings inequality within industries observed in the 1980s. Wieler went one step further and examined the effect of changes in the mix of cognitive, interactive and motor-skill requirements on the job from the Dictionary of Occupational Titles on earnings inequality. Examining 33 industries between 1973 and 1990, she found no statistical effect at the industry level in either decade. Unlike the substantial widening of the wage distribution in the 1980's, the dispersion of skill requirements was unchanged in the 1980's, even among technologically advanced manufacturing industries.
If demand for skills is not increasing as fast as the standard account suggests, perhaps the problem is that workers are getting dumber. However, according to a 1990 report from the Educational Testing Service, 83.5 percent of white 17 year old high school students in 1971 read at "intermediate" or higher levels; this figure rose to 87.3 percent in 1980 and 89.3 percent in 1988. Even more impressive were the results for black students, who increased from just 39.7 percent in 1971 to 76 percent in 1988. Yet, real wages for young black men -- even those with a college degree -- declined over this period.
In the 1980s, higher skills have simply not led to higher wages. In industry after industry, average educational attainment rose while wages fell. Indeed, wages fell more rapidly for young workers, despite their higher educational attainment. Young workers (16-39) in the stone, clay, glass and primary metals industry showed a 9 percent decline in workers with a high school degree or less, but a 76 percent increase in those paid an hourly wage below that necessary to keep a family of three 50 percent above the poverty line. The communications industry saw its low-skill share decline by 33 percent and its low-wage share increase by 33 percent. Even more dramatically, the automobile industry's low-skill employment share declined by 6 percent despite its low wage share growth of 142 percent. Goods industries with high-wage, low-skill workforces appear to have restructured in the 1980s by radically lowering wages and gradually raising skill requirements -- in short, by moving in the direction of the typical service sector workplace.
Massive Skill Restructuring?
Much of the attraction of the skills-mismatch story lies in its consistency with both the observed declines in the relative wages of low-skill workers in the 1980s and the popular vision of the effects of computer-based mechanization in the workplace. In the new "high performance" workplace, workers must possess the cognitive and diagnostic skills necessary to perform a broad range of frequently changing tasks. Computers and related technologies require higher skills and workers with obsolete or insufficient skills inevitably get paid less and ultimately lose their jobs, leaving behind a more skilled workforce. This transformation is un doubtedly underway. But was the demand for higher skills in the 1980s substantially greater than in earlier decades? And does it explain the wage collapse?
In a recent study of changing employment shares in manufacturing, Eli Berman, John Bound, and Zvi Griliches call attention to a sharp increase in the ratio of nonproduction to production workers and attribute this "skill" restructuring to computer-based technological change. Interestingly, the change in this ratio over time shows that virtually all of the "skill upgrading" took place in just 3 years -- the recession years 1980, 1981 and 1982. Although the authors note that employers use recessions to restructure, there is no evidence of a substantial shift in skill mix (as they measure it) in the 1990-91 downturn. If the technology-induced skill mismatch hypothesis story is right, why would new workplace technologies cause a sharp skill restructuring between 1980 and 1982 but not between 1983 and 1992, when the latter period was characterized by a far higher rate of investment in computer-based equipment?
In fact, there was great stability of the skill mix in manufactur ing after the recessions of 1980-82. Between 1983 and 1988, the ratios of craft to semi-skilled workers, technicians to clerical workers, and professionals to managers remained virtually unchanged. And the ratio of craft workers to laborers actually declined steadily from about 4 to 1 in 1983 to to 3.4 (it then rose slightly to 3.5 in 1988). These figures do not suggest the kind of skill restructuring that could explain the enormous earnings decline suffered by low-skill workers or the considerable widening of the earnings gap between low- and high-skilled workers observed in the last decade.
In recent research with Maury Gittleman, an economist at the Bureau of Labor Statistics, I allocated private sector workers to 621 jobs and grouped these into six job contours and three labor market segments based on a variety of measures of job quality using a statistical technique known as cluster analysis. We found that the worst jobs, labelled the "secondary segment," included 39 percent of the workforce in 1979. This share remained almost constant in the 1980's, falling slightly to 37.9 percent in 1990. The major restructuring that we found was not at the bottom, away from secondary jobs, but away from the middle -- good nonsupervisory jobs -- and towards the best white collar jobs. Most of these shifts took place between 1979 and 1983, well before most of the investment in computer-based technologies.
Over the long term, there has doubtless been a gradual increase in the demand for more highly skilled workers. As the renowned economist Fritz Machlup pointed out back in the early 1960s, the increase in the demand for information workers -- those with relatively high cognitive skills -- dates to at least the turn of the century. Indeed, as my own research with Edward Wolff indicates, the share of semi-skill and low-skill manual workers in total employment declined substantially in each decade from 1950 to 1980.
But in contrast to the skill mismatch story, the evidence does not indicate an acceleration in skill upgrading in the 1980s. Indeed, my own studies with Wolff, as well as research by Lawrence Mishel and Ruy Teixera using detailed occupational employment data matched to Dictionary of Occupational Titles skill measures show a deceleration in skill growth between the 1960s and 1980s. Using different data and methods, economists Kevin Murphy and Finis Welch, writing in the American Economic Review, concluded that "we do not find that the demand for skill grew particularly rapidly during the 1970s and 1980s, a period when wage inequality expanded in comparison to the three earlier decades...." Similarly, Lawrence Katz, chief economist of the U.S. Department of Labor, has noted that "estimates of within- and between-industry demand shifts indicate little or no acceleration in the 1980s" and suggests that "differences in the 1970s and the 1980s depend substantially on differences in relative supply shifts."
The 1990 Report of the Commission on the Skills of the American Workforce, "America's Choice: High Skills or Low Wages!," despite its vivid title, found that only 5 percent of firms surveyed were actually concerned about a skill shortage. This is consistent with the key finding of a new study on international competitiveness by a prominent commission, assembled by the McKinsey consulting firm and headed by Professor Robert Solow of MIT, that it was not primarily skills or technology that distinguished productivity rates among nations, but rather management and labor relations policies. Based on the recent practices of employers, a more apt description might be "High Skills and Low Wages."
Taking the Low Road
The more convincing explanation for the wage collapse is the intensifying product market competition and the ideological shift away from government intervention and towards laissez faire. These developments undermined traditional wage setting institutions, leading to a declining willingness to pay low- and moderate-skill workers "living-wage" paychecks. At the same time a slower growth and globalization led to an oversupply of workers competing to perform low- and moderate-skill jobs.
At the end of the 1970s, firms began to fundamentally reassess their employment and wage-setting practices. Large integrated (high-wage) firms began to downsize and rely more heavily on low-wage suppliers. Advances in telecommunications and transporta tion facilitated the relocation of lower-skilled operations to low-wage sites, leaving behind a core of permanent, relatively skilled employees supplemented increasingly by part-time and temporary workers.
Consistent with this alternative account, George J. Borjas and Valerie A. Ramey have argued that much of the increase in earnings inequality in the 1980s can be traced to the erosion of "rents" earned by low-skilled workers in concentrated industries -- an erosion they attribute to increased foreign competition in durable goods manufacturing. They show that trade competitiveness in this sector has closely mirrored trends in wage inequality since the 1970s. A 1993 study by Maria Papadakis underscores the role of durable goods industries in the worsening of the trade balance in the early 1980's. She reports that the trade balance shifted from a $3.6 billion surplus in 1982 to a $67.8 billion deficit just two years later. By 1987 the manufacturing trade deficit rose to $125 billion. She attributes about 55 percent of the increase in the trade deficit to four durable goods industries that pay relatively high wages: motor vehicles (24.5 percent, nonelectrical machinery (14.7 percent), electronic equipment (9.7 percent) and electrical machinery (5.5 percent).
Other economists, including Paul Krugman and Robert Lawrence, argue that trade shifts do not account for the bulk of the wage collapse. These studies, however, focus exclusively on various indicators of import penetration and export shares, and not on how trade patterns affect labor's bargaining power.
By the 1980s, concession bargaining had become widespread. According to labor economist Daniel Mitchell, the proportion of workers with private-sector union contracts whose wages were frozen or cut ranged from 0 to 5 from 1964 through 1980, rose to 8 percent in 1981, jumped to 44 percent in 1982 and 37 percent in 1983. While dropping to 23 percent in 1984 and 26 percent in 1985, by historical standards these rates have remained at remarkably high levels. In the eminently mainstream Brookings Papers on Economic Activity, Mitchell describes the increasingly confrontational approach of employers:
"[T]he longevity of the (wage) concession movement and its spread to less-than-dire situations suggest that the initial concessions have encouraged other employers to try their luck in demanding similar settlements. . . . Management, cheered by what is perceived as a shift in the balance of power, has changed its bargaining goals..."
Downsizing often works in tandem with wage concessions. For example, General Motors has recently agreed to sell most of its ownership in three plants employing some 2,000 employees. These workers, currently covered by a UAW contract that guarantees them the same wage as assembly line workers, will have to negotiate a separate agreement with the new company at the end of the contract, virtually guaranteeing wage concessions in the near future.
Another way to reduce labor costs is to substitute part-time and temporary workers for permanent full-time workers. The temporary help industry grew eight times faster than employment in all nonagricutural industries between 1978 and 1985 and increased from 620,500 workers in 1984 to 1,031,500 in 1989.
While the same technological advances have been occurring in Canada and Western Europe, there has not been a comparable collapse of wages. A key difference, of course, is that unions and other wage setting institutions remain much stronger in other industrial nations. Labor economists John DiNardo and Thomas Lemieux have compared wage inequality trends in Canada and the U.S. in the 1980s. Despite the similar labor markets, Dinardo and Lemieux find that "during this period, union density fell precipitously in the United States but declined very little in Canada. Similarly, the real minimum wage declined by 23 percent in the United States but by only 12 percent in Canada . . . We find that unions and the minimum wage accounted for 80 percent of the difference in the growth of inequality in the two countries." Another study, by Gary Loveman and David Blanchflower, concludes that unions and the minimum wage help to explain the different experiences of France (low inequality growth) and Great Britain (increasing wage inequality) during this period.
Moreover, our national social policies have undermined traditional wage-setting norms in the private sector. The U.S. continues to rely heavily on employers to provide health insurance, pensions, child care, and other fundamental benefits -- benefits that appear as labor costs to employers. These costs are assumed by the public sector in most other developed countries. As the costs of benefits rise, our "privatized" benefits system encourages employers to substitute part-time and temporary low-wage jobs. This magnifies the gap between primary and secondary labor market jobs, increasing both the share of low earners and wage inequality.
The undermining of traditional wage-setting institutions has lowered wages for those with the least bargaining power in the labor market, thus increasing inequality between skilled and unskilled workers. It may have also tended to increase wage inequality among workers in the same education, age, and gender group in the same industry. While the conventional view is that technological change has increased the demand for skill, leading to an increased premium for "unobserved skills" within these groups, it may be that the de-institutionalization of the labor market has had a greater effect. Wage norms appear to have broken down within firms (as internal labor markets are opened up to external competi tion), within industries (as increasing competition causes differ ences among firms to become a more critical factor in wage outcomes), and among communities (as transportation and telecommu nications facilitate the relocation of some, but not all, firms to lower wage areas). In short, the "law of one price" that supposedly characterizes free markets may have been undermined, not promoted, by the heightened competition in labor markets.
Crowded Low-Wage Job Markets
Today, more workers are crowding into a pool of "secondary" jobs that remained a fairly constant share of total jobs throughout the 1980s, tending to lower the wages of what were already the worst jobs in the labor market. One source of this crowding is displacement from high-wage blue-collar jobs. According to a recent Department of Labor study by Diane E. Herz, more than 4.3 million workers were displaced during the boom years of 1985-89. Only 72 percent had been re-employed by January 1990 and of these, about 10 percent worked part-time. Among those re-employed full-time, about 40 percent earned less in current dollars than on their previous job. Not surprisingly, those least successful in the labor market after displacement were high-wage blue-collar men.
There has also been a stream of highly educated workers forced to compete in the low-skill, secondary labor market. Despite the rising average premium for a college degree relative to a high school degree in the 1980s, a weak job market has forced many lower-level white-collar workers with college degrees to compete for relatively low-skilled jobs. This became particularly pronouncd at the end of the decade, during the "white-collar recession" of 1990-91. The same computer-based technologies and corporate restructuring have made large numbers of middle-level managers redundant. According to the U.S. Labor Department, throughout the 1980s about 20 percent of college graduates were working at jobs that don't normally require a degree, and this is expected to increase to 30 percent at the end this decade. The share of black and Hispanic college graduates with poverty-level wages rose dramatically in this decade, from about 9 percent to just under 15 percent. One consequence of declining opportunities for moderately skilled white-collar jobs has been to increase competition, contributing a downward pressure on their wages.
Relocation of operations to low-wage, high unemployment regions has also contributed to a rising effective supply of labor. This shift to lower-wage labor markets, and not a search for more highly-skilled workers, is behind most recent relocation decisions. For example, according to spokespersons of Pratt and Whitney, the firm's decision to relocate as many as 9,000 high paying production jobs from a high-skill state (Connecticut) to lower-skill states (Maine and Georgia) was expressly designed to reduce labor costs.
Not long ago, it was an article of faith that new foreign workers were a net plus for the U.S. economy. They did unpleasant jobs at low wages, which native-born Americans were unwilling to perform, and had little downward effect on prevaiing wages. That story is no longer accurate. At a time of extraordinarily high levels of nonrecessionary unemployment and declining real wages for low-skill workers, the flow of low-skill (legal and illegal) foreign workers into the U.S. reached record levels in the 1980s. According to Vernon M. Briggs the counted foreign-born population was 9.6 million (4.7 percent) in 1970, 13.9 million (6.2 percent) in 1980, and 19.8 million (7.9 percent) in 1990. A recent study by Borjas, Freeman and Katz found that the combination of rising imports and growing numbers of low-skill foreign workers had substantial negative effects on the relative earnings of native low-skill workers. The authors concluded that "between 30 percent and 50 percent of the ... decline in the relative weekly wage of high school dropouts from 1980 to 1988 can be attributed to trade and immigration flows." Federal immigration policy, at least until 1990, helped to supply an already overflowing pool of low-skill workers.
Case study evidence also suggests that, particularly in large urban centers, increased competition from recent immigrants has had negative employment impacts on native-born workers. In his study of the Los Angeles restaurant and hotel industry, Roger Waldinger concluded that "the story of black displacement in restaurants and hotels can be traced not to skill upgrading, but rather to competition with a rapidly growing immigrant population." Similarly, a General Accounting Office study found that "Janitorial firms serving downtown Los Angeles have almost entirely replaced their unionized black workforce with non-unionized immigrants." Again, it appears to have been labor costs, not skill restructur ing, that explains this replacement of black with immigrant work ers. A study by Katherine Newman, an urban anthropologist at Columbia University, found that despite a 25 percent decline in the value of the minimum wage in the 1980s, there was intense competition for minimum wage jobs in two fast food restaurants in Harlem. Over the first six months of 1993, over 400 people applied for 50 jobs in these establishments. Only a handful of the job winners were high school dropouts. Half had high school degrees, about one-quarter were currently enrolled students, and the remainder were mainly foreign-born adults.
In sum, much of the low-wage problem appears to be just the reverse of that proposed by the technology-induced skill mismatch hypothesis: in the face of mounting competition, employers have reduced unit labor costs and increased flexibility in the produc tion process by following the "low road" -- lower wages, little training, and fewer permanent employees. This strategy has been facilitated both by the undermining of traditional wage-setting institutions and regulations, such as unions, labor relations laws and the minimum wage, and by an increasing supply of workers forced to compete for low-wage jobs. Since the late 1970s, federal and local government policies have played a central role in both of these developments, and in so doing have contributed to the adoption of low-wage competitive strategies by many private sector employers.
Few will, or should, oppose the Labor Department's "get smart" policies for workers, but the reality is that changes in the ability of workers to provide the skills needed in technologically advanced workplaces had little to do with the startling growth in poverty-wage jobs, the drop in real earnings, and the growth of earnings inequality in the 1980s. We do need to improve our education and training system, but making workers smarter will not, by itself, have much impact on the distribution of earnings. Training policies can only be relied upon to pay off for workers in a high-employment context. In addition to a better educated workforce, we need policies and strategies -- both macroeconomic and structural -- to restore labor bargaining power so that workers with rising skills in an economy with rising productivity will again command rising wages.