CIAO DATE: 03/02

Critical Review

Winter–Spring 1998 (Vol.12 Nos.1–2)

It Doesn’t Work

By J. Bradford De Long _*

Abstract

Vedder and Gallaway are mistaken in their attempted demonstration that government policies to raise real wages have been the source of most or all U.S. unemployment in the twentieth century. Their case depends on a presumed correlation between high unemployment and high real wages that has not existed since World War II, and on a naive confusion between correlation and causation: just because real wages and unemployment were both relatively high during the Great Depression does not mean that high real wages were the cause of high unemployment. More likely, the reverse is true. The ascription of high unemployment to higher-than-equilibrium real wages is not always in error. But there is no reason whatsoever to believe Vedder and Gallaway’s claim that the government’s policies to raise wages have been “the major cause of high unemployment in the United States.”

There are two things wrong with Richard Vedder and Lowell Gallaway’s Out of Work: Unemployment and Government in Twentieth-Century America (New York: Holmes and Meier, 1993). The first is that it confuses correlation with causation. It jumps from its authors’ observation of high unemployment correlated with high real wages—the “real” purchasing power of what workers earn—to the conclusion that any (government) action that increases real wages increases unemployment.

As the back jacket reports: “Out of Work shows...that such policies as minimum wages, legal privileges for unions, civil rights legislation, unemployment compensation, and welfare have all played significant roles in generating joblessness.” The conclusion of the book, allegedly backed by “relentless and devastating evidence,” is “that the major cause of high unemployment in the United States...is government itself.”

But when we find high real wages and high unemployment occurring together, which is cause and which is effect? Is unemployment high because workers have relatively high purchasing power? Or are real wages high because unemployment is high? Economists have argued this back and forth for nearly the entire century, and the only sane answer is that sometimes it is one, sometimes it is the other, and sometimes some third factor is causing both.

In the United States in the 1930s, for example, the balance of the evidence is that real wages were high because unemployment was high—and that steps to reduce wage levels would have deepened, not alleviated, the Great Depression. In Europe in the late 1970s, on the other hand, the balance of the evidence is that unemployment was indeed high because real wages were higher than “warranted,” as a result of the impact on relative prices of the 1973 tripling of world oil prices (see Bruno and Sachs 1983)—and steps to reduce wage levels probably would have reduced unemployment. Had the authors taken a more careful look at the direction of causation in their presumed correlation between high real wages and high unemployment, they would have been led to a much more measured conclusion than the thesis that any (government) action that increases real wages increases unemployment.

The first flaw in the book is bad, but the second is worse. There is no doubt that there are times, places, and historical episodes—the Great Depression in the United States or Europe in the late 1970s—in which high unemployment and high real wages do go together. But there is strong reason to doubt that this is a general pattern. When the business cycle turns downward, the jobs that disappear are disproportionately low-wage jobs held by the less skilled. The jobs that remain pay, on average, more. So the shifting cyclical composition of employment generates the appearance of a countercyclical pattern in real wages: the average real wage rises in recessions because more low-paid workers have lost their jobs.

When one corrects for this “composition of employment” effect, the balance of the evidence is that in the average business cycle real wages are weakly procyclical: a particular job with particular skill requirements tends to carry higher real wages in a boom than in a recession (see Barsky and Solon 1993). Real wages are high in booms because demand for investment goods is high in booms—and thus the derived demand for labor is high as well.

Moreover, Vedder and Gallaway’s analysis handles long-run productivity growth incorrectly. Their theory is not that unemployment is high when real wages are high in absolute value, but when real wages are high relative to the trend of productivity growth. But the productivity series that they use to scale their real wage measure is wrong: it contains cyclical movements in productivity that are the result, not the cause, of low production and high unemployment.

So not only do Vedder and Gallaway confuse correlation with causation in those historical episodes in which high unemployment and high real wages go together, but their presumed general correlation of high wages and high unemployment is—most likely—an artifact of the way their data were constructed. High real wages (relative to productivity) cannot possibly be part of a general explanation for high unemployment, because real wages are more often than not low when unemployment is high.

Thus, there is virtually nothing left of Vedder and Gallaway’s grand attempt to “redefine the way we think about one of the most explosive issues of the twentieth century.” A lot of hard work has led them to false conclusions because it was based on flawed foundations. I do not pretend that I have the key to the riddle of the business cycle. But I do not think that Vedder and Gallaway have it, either.

Let me expand on why I have reached my negative judgment about this book. I will first discuss the pattern of real wages over the business cycle, and explain why I believe that the central correlation on which Vedder and Gallaway have based their argument is in all likelihood an artifact of how our data are constructed, not how the world really works. I will then go on to explain in more detail how it is that when high real wages and high unemployment go together, sometimes one is the cause and sometimes the other is, but the chain of causation is not clear and direct enough to justify laying responsibility for unemployment at the door of government policies to raise wages.

 

Real Wages Are Not Countercyclical

In 1936, John Maynard Keynes’s General Theory of Employment, Interest and Money inaugurated the intellectual tradition of modern macroeconomics. One feature of the business cycle Keynes described—a feature he went to significant pains to analyze—was a presumed countercyclical nature of real wage movements: when production was high relative to trend, Keynes argued, unemployment and real wages were low; when production was low relative to trend—when the economy was in recession or depression—both unemployment and real wages were high.

This part of Keynes’s book drew sharp and immediate dissent from economists specializing in the study of the labor market and labor institutions. Forty pages of the March 1939 Economic Journal were devoted to the question of cyclical movements in real wages. John Dunlop—then a young economist; later, chairman of the War Labor Board during World War II and Secretary of Labor under President Ford—contributed a substantial article, arguing that real wages were not countercyclical: that high unemployment and high real wages relative to trend did not go together. Lori Tarshis contributed a short comment to the same effect.

Keynes wrote a substantial reply, essentially conceding the point: Keynes said that he had tried to minimize points of difference between his approach and the previous “classical” approach; that countercyclical real wages had been an important part of the “classical” approach; that nothing important to his argument hinged on the countercyclical nature of real wages; and that he deferred to the superior statistical, institutional, and historical expertise of the labor economists.

Since then, economists have often revisited the question of the cyclical behavior of wages and prices. Recently, valuable contributions have come from Mark Bils (1985), Michael Keane, Robert Moffitt, and David Runkle (1988), and Paul Beaudry and John DiNardo (1991) in the Journal of Political Economy; Gary Solon, Robert Barsky, and Jonathan Parker (1994) in the Quarterly Journal of Economics; and Michael Bruno and Jeffrey Sachs (1982) in their book, The Economics of Stagflation.

For the most part, subsequent revisitations of this question have backed Dunlop: the real wage is weakly procyclical, with higher wages associated with low unemployment; or it is, at most, acyclical.

Now come Vedder and Gallaway to claim not only that 60 years of economic research has been wrong—that real wages are countercyclical, as Keynes believed, such that high unemployment is associated with high real wages—but that the association between high unemployment and high real wages is so strong as to account for practically all fluctuations in American unemployment over the past century.

Why this massive dissonance between the consensus of contributors to the University of Chicago’s Journal of Political Economy and what is reported by Vedder and Gallaway? This is the question that first interested me in the book. And I quickly found that Vedder and Gallaway would be of little help in trying to reconcile these two lines of thought: neither the Bruno and Sachs study of Europe in the 1970s nor the Journal of Political Economy articles that contain the latest word on real wages make it into Vedder and Gallaway’s bibliography.

One difficulty plaguing the topic is “composition bias” in measurements of economy-wide real wages. The jobs that vanish in depressions are disproportionately low-wage jobs held by the relatively lesser-skilled. The average wage is calculated by taking total payroll and dividing it by total work hours. Thus, anything that shifts the relative proportions of high-wage and low-wage jobs will change the calculated average wage—and a recession or a depression shifts the relative proportions of high-wage and low-wage jobs.

Note that this “composition bias” is all effect and no cause: in its pure form the wages paid to any one individual doing any one job have not changed, so there has been nothing to alter employers’ decisions to hire or retain workers: yet the economy-wide average real wage does change, because the average contains fewer low-wage workers as a result of recession- or depression-induced unemployment. Vedder and Gallaway, however, take no account of these complications.

A second difficulty is that Vedder and Gallaway have committed an error in how they scale their real-wage variable. Real wages are high or low relative to the productive potential of the average worker, so Vedder and Gallaway divide real wages by a measure of productivity. Unfortunately, they divide real wages not by potential productivity—not by what output per hour would have been if factories had been running at normal levels of operation—but by actual productivity. And actual productivity varies greatly over the business cycle. Whenever production is depressed, measured productivity is relatively low because the American economy possesses substantial increasing returns to scale: the “overhead” workers in a factory are necessary to keep it from rusting into oblivion, whether the factory is running three, two, one, or zero shifts.

The “composition bias” and the “productivity bias” serve to associate Vedder and Gallaway’s measure of real wages more strongly with high unemployment than a properly constructed measure would.

A third—and perhaps the most important—difficulty is that the American economy may well not have the same structure today that it had before World War II. During the Great Depression, real wages rose, while prices of commodities in the consumer price index fell faster than did the wages paid to workers. During the Great Depression there was a positive correlation between high real wages and high unemployment.

Since World War II there has been no such correlation. Figure 1 plots my calculations of productivity-adjusted real wages and the unemployment rate. Those postwar years when unemployment was the highest were not the years in which adjusted real wages were the highest. And the postwar years, when unemployment was the lowest, were not the years in which adjusted real wages were the lowest.

Thus, Vedder and Gallaway’s book is fatally flawed. The presumed correlation between high adjusted real wages and high unemployment is the foundation on which they build their argument. This foundation is made out of mud. Since World War II, there is no sign that high adjusted real wages go with high unemployment; real wages appear procyclical and not countercyclical. Before the Great Depression, real wages may have been countercyclical, but because no one has unscrambled the effects of “composition bias” and “productivity bias” on Vedder and Gallaway’s adjusted real wage measure, we really do not know.

 

High Real Wages Sometimes Are the Result of Unemployment

During the Great Depression, real wages were certainly countercyclical: prices of commodities in the consumer price index fell faster than did the wages paid to workers. There was therefore a positive correlation between high real wages and high unemployment. But there is no good reason to think that high real wages in the Depression were a cause, rather than an effect, of high unemployment.

When Keynes wrote his General Theory (1936), he was convinced (i) that real wages were higher in recessions than in booms, and (ii) that the cause of the business cycle lay in fluctuations in aggregate demand: fluctuations both in the marginal propensity of households to consume and in the “animal spirits” of investors, whose actions determined the demand for capital goods. In Keynes’s vision, there was correlation between high real wages and high unemployment, but no causation: if the government, workers, firms, or unions took steps to reduce nominal wages that had no impact on real aggregate demand, the likely consequence would not be a fall, but a rise, in unemployment.

How could this be? Keynes’s vision of the economy was one in which the nominal wages paid to workers are relatively slow to respond to changing demand and supply conditions. In times of recession, firms find themselves able to hire additional workers from the unemployed at the prevailing wage. They have every incentive to do so. But to sell the output additional workers produce, firms have to cut their prices a little bit below those of their competitors in order to expand their market share.

When every firm cuts its price relative to the prevailing nominal wage, the level of prices falls relative to the prevailing nominal wage. This process of mutual price-cutting continues until the price level has fallen so far that the extra product of an additional worker to a representative firm can no longer be sold at a profit. At that price level (relative to the nominal wage), it doesn’t pay firms to hire more workers. Thus, during a recession, real wages are relatively high because the price level is relatively low.

What did Keynes think would happen if nominal wages were not relatively inflexible in a recession? What if nominal wages fell sharply as unemployment rose? The first-order effect in Keynes’s model is that a fall in nominal wages would have no effect on employment: with a lower wage level, firms would find it worthwhile to continue the downward spiral of price-cutting even further. The end result? An economy with the same real wage, the same level of unemployment, and lower nominal wages and prices. The second-order effect in Keynes’s model is that a fall in nominal wages (and the induced further fall in product prices) raises unemployment: falling prices discourage investment because they generate (i) extremely high real interest rates and (ii) large-scale bankruptcies and financial collapses that destroy the web of financial intermediation. The third-order effect in Keynes’s model is that a fall in nominal wages (and the induced further fall in product prices) increases the purchasing power of households’ real money balances, stimulates consumption, raises aggregate demand, and boosts employment.

Which effect is dominant? De Long and Summers (1988, extending Tobin 1975) argue that in economies with a Keynesian structure and with parameter values like those of the United States today, the second effect dominates: increasing nominal wage flexibility would increase the volatility of the business cycle.

What do Vedder and Gallaway have to say in response to Keynes’s extended argument that causation runs from unemployment to high real wages, and that policies to lower nominal wages in a depression would be unhelpful?

They have little to say.

On page 47 they assert that any belief that “higher unemployment causes higher adjusted real wages” is “implausible.... It makes sense that higher wages would price workers out of the market, causing increased unemployment, but it makes no sense that higher unemployment would cause wages to rise (if anything, higher unemployment might induce wage-cutting).” But this is to misunderstand Keynes and the subsequent literature. The argument was never that higher unemployment would cause (nominal) wages to rise. The argument was, and is, that excess inventories and deficient demand for goods would cause the prices of goods to fall; that in modern economies the prices of goods would fall farther and faster than nominal wages—all nominal wages and prices are falling in a depression, but at different speeds—and that the net result would be higher real wages, because nominal wage levels would be relatively slow to fall.

This argument Vedder and Gallaway do not acknowledge: they never even try to come to grips with their principal intellectual opponent. They have offered no reasons to accept their assumption that a high unemployment/high real wage correlation—when it exists—is the result of high real wages causing high unemployment. They have given no reasons to reject Keynes’s theory that, when such a correlation exists, it is the result of deficient aggregate demand causing high unemployment and also pushing down product prices, leading to a relatively high real wage.

Their third sentence is that the “Keynesian Revolution led the economics profession down an unproductive, destructive path for decades.” In saying this, Vedder and Gallaway assume the obligation to give reasons why the Keynesian approach is incorrect, unproductive, and destructive. Yet they do not; they make a contract with their readers, and then casually break it.

None of this means that high real wages are always the consequence and never the cause of high unemployment. Edmond Malinvaud (1977) advances a more complex—and much more reasonable—position in his attempt to extend the scope of Keynesian analysis. Malinvaud argues that the market economy is subject to many possible macroeconomic maladies, and that not all business cycles—not all booms, not all depressions—are alike. Malinvaud distinguishes between “classical” and “Keynesian” unemployment. In both cases, the real wage paid to employed workers is higher than in full-employment equilibrium. But in the second case, high real wages are the result of a recession, while in the first case they are the cause.

In the same vein, Michael Bruno and Jeffrey Sachs (1982) provide a powerful and detailed analysis attributing high European unemployment in the wake of the 1973 oil shock and productivity slowdown to real wages that exceeded “warranted” levels. The French economist Daniel Cohen (1996) believes that high European unemployment today exists because “the war on unemployment is in the hands of governments which represent first and foremost the point of view of the people who have jobs and fear losing them.” Thus, unemployment has been “feared less than the remedies that would have been necessary to contain it.”

In Cohen’s view, European unemployment could have been reduced by freeing up the labor market, but then the wages of the currently employed would have fallen as “those who have jobs are forced to compete with those who do not.” European unemployment could have been reduced if the government had been willing to boost public employment and spend large sums on reentry programs for the jobless, thus increasing the value of the work that the currently unemployed could do. But these costly reentry programs must be paid for by higher taxation, and the fear is either that higher taxation will slow economic growth, or that programs for the unemployed will crowd out other social benefits. The European Union has been stuck in the middle.

So the ascription of high unemployment to higher-than-equilibrium real wages is not always wrong. There are times and places when it is right.

But Vedder and Gallaway are mistaken when they attempt to demonstrate that government policies to raise real wages have been the source of most, or all, U.S. unemployment in the twentieth century. They are mistaken because at the foundations of their argument are a presumed correlation between high unemployment and high real wages that does not exist today, and a confusion about how to distinguish cause from effect in macroeconomic analysis.

 

References

Beaudry, Paul, and John DiNardo. 1991. “The Effect of Implicit Contracts on the Movement of Wages Over the Business Cycle: Evidence from Micro Data.” Journal of Political Economy 99 (August): 665–88.

Bils, Mark. 1985. “Real Wages Over the Business Cycle: Estimates from Panel Data.” Journal of Political Economy 93 (August): 666–89.

Bruno, Michael, and Jeffrey Sachs. 1982. The Economics of Stagflation. Cambridge, Mass.: Harvard University Press.

Cohen, Daniel. 1996. The Misfortunes of Prosperity. Cambridge, Mass.: mit Press.

De Long, J. Bradford, and Lawrence H. Summers. 1986. “Is Increasing Price Flexibility Stabilizing?” American Economic Review 76 (December).

Dunlop, John. 1938. “Movement in Real and Money Wage Rates.” Economic Journal 48 (September): 349–66.

Keane, Michael, Robert Moffitt, and David Runkle. 1988. “Real Wages Over the Business Cycle: Estimating the Effect of Heterogeneity with Micro Data.” Journal of Political Economy 96 (December): 1232–66.

Keynes, John Maynard. 1936. The General Theory of Employment, Interest and Money. London: Macmillan.

Keynes, John Maynard. 1939. “Relative Movements of Real Wages and Output.” Economic Journal 49 (March): 34–51.

Malinvaud, Edmond. 1977. The Theory of Unemployment Reconsidered. Oxford: Basil Blackwell.

Solon, Gary, Robert Barsky, and Jonathan Parker. 1994. “Measuring the Cyclicality of Real Wages: How Important Is Composition Bias?” Quarterly Journal of Economics 109 (February): 1–25.

Tarshis, Lori. 1939. “Changes in Real and Money Wage Rates.” Economic Journal 49 (March): 150–54.

Vedder, Richard K., and Lowell E. Gallaway. 1993. Out of Work: Unemployment and Government in Twentieth-Century America. New York: Holmes and Meier.

 

---

Endotes

*: J. Bradford De Long, Associate Professor of Economics, University of California at Berkeley, Berkeley, CA 94720, a research associate of the National Bureau of Economic Research and a visiting scholar at the Federal Reserve Bank of San Francisco, would like to thank the National Science Foundation and the Alfred P. Sloan Foundation for financial support, and Bob Barsky, Barry Eichengreen, Jeffrey Friedman, Chris Hanes, and Christie Romer for helpful comments and discussions.  Back.