Arnold Kling, "Arguing in My Spare Time", No. 13

"The latest reports show the economy continued to expand. . .At the same time, average wages and benefits are rising at a slower pace than last year. And prices of things produced here are rising at an annual rate of less than 1 percent.

"How can this be? In a world where nearly every American who wants a job has one, why aren't workers getting bigger raises and forcing companies to raise their prices, as happened so many times in the past?

"In a word, the answer is productivity."

--Steven Pearlstine, "Analysis," The Washington Post, May 2, 1998.

In a word, this analysis is hogwash.

In spite of numerous anecdotes, the statistical evidence for the hypothesis that productivity growth is high in the United States is surprisingly flimsy. An alternative hypothesis, which I call the "Junta hypothesis," is on much stronger ground.

This essay soon will descend into economic jargon. Before doing so, let me express the two competing hypotheses in layman's terms:

The productivity hypothesis is that the strong economy is a result of corporate management in America coming up with innovations and improvements.

The junta hypothesis is that corporate management has succeeded primarily in repressing wages and increasing the share of income going to profits.

In a nutshell, the productivity hypothesis says that management has made the pie larger; the junta hypothesis says that management has gotten better at capturing more of the pie for itself and for company owners.

Economic indicators

When we say that the economy is strong, we usually mean the following indicators:

low inflation;

high corporate profits;

high stock prices;

low unemployment;

This combination of indicators is the result of slow growth in unit labor costs. Unit labor costs are the costs paid to workers for producing one widget.

Labor costs can be thought of as W times L, where W is the wage rate, and L is the amount of labor employed. Unit labor costs are labor costs divided by total output. If output is Q, then unit labor cost = WL/Q.

When unit labor costs rise slowly, prices tend to rise slowly (low inflation). Even so, there is room for prices to rise faster than unit labor costs, leading to higher corporate profits, and thereby helping to support higher stock prices. With profitability high, firms want to hire more workers to earn more profits, and unemployment falls.

Productivity, by definition, is output per unit of labor. In our notation, productivity is Q/L. Another way to think of unit labor cost is that it is the wage rate, W, divided by productivity.

Returning to our competing hypotheses: the productivity hypothesis is that unit labor costs have grown slowly because of rapid growth in productivity; the junta hypothesis is that unit labor costs have grown slowly because wages have grown slowly.

The official statistics for the U.S. economy support the hypothesis that the share of the pie going to workers has fallen. In 1982, compensation of employees was 74.1 percent of national income. In 1997, preliminary figures suggest that the labor share was about 70.7 percent.

A sharp reader might point out that 1982 was a recession year, so that the trend in labor's share from 1982 to 1997 could be a cyclical phenomenon. That may be true, but if so it reinforces the main point of this essay, which is that the evidence for a secular improvement in productivity growth is lacking.

The direct evidence about productivity growth is controversial. According to the offical data on productivity maintained by the U.S. Bureau of Labor Statistics (BLS), productivity growth in the 1990's has not been unusually good. In fact, the 1990's has been the worst decade on record (the data go back almost to World War II) for productivity.

How bad has productivity growth been? So bad that the best year for productivity growth in the 1990's, 1992, saw a rate of productivity growth of 3.4 percent, which is below the average rate in the 1950's. So bad that for the 10 years ending in 1996, average productivity growth was 0.87 percent, less than one percent per year! In the 1950's and 1960's, productivity growth exceeded 3 percent per year on average. Even in the 1970's, when the economy was hampered by two oil price spikes, productivity growth averaged over 2 percent per year.

If the official statistics are credible, then it is difficult to understand how someone could even dare to propose that recent economic performance reflects high productivity growth. The numbers seem to show quite the opposite.

The official statistics have an Achilles heel, however. The measurement of productivity growth is only as good as the measurement of inflation. What we can observe reasonably accurately is the growth in the dollar value of total output. We derive the growth in real output only after adjusting for inflation. If the adjustment for inflation is biased upward, then the measure of productivity growth is biased downward.

It is possible that what the BLS reports as low productivity growth is a result of a biased measure of inflation. If inflation had been measured in an unbiased way in the 1970's, but in the 1990's inflation estimates were biased up by 1 percent per year, then productivity in the 1990's was about the same as in the 1970's. On the other hand, if inflation estimates always have been biased up by 1 percent per year, then the correct estimate of productivity growth in the 1990's will still be below that in the 1970's. To argue that productivity growth has accelerated in the 1990's relative to earlier decades, one would have to develop and prove a hypothesis that structural change in the economy has caused increased biases in the measurement of inflation over time.

The Official Verdict

The official economic statistics run strongly counter to the hypothesis that we have enjoyed unusually rapid productivity growth in recent years. The alternative hypothesis, that wages are repressed, is consistent with the official productivity statistics. The alternative also is supported by other evidence, including the drop in the share of labor income, as well as the current complaints of a "labor shortage," which suggests that wages are below the level needed to clear the market.

The Junta Hypothesis

"Redesigning products to make them cheaper to produce, reengineering processes to take out unnecessary costs--such activities have become a permanent focus of Corporate America. . ."

--Pearlstein, "Analysis"

To my knowledge, no one has ever done a time-and-motion study of top corporate managers in order to determine what is the "permanent focus of Corporate America." My impression is that, like members of a 1950's Latin American military junta, top managers focus on retaining power.

Using this metaphor (and it is nothing but a metaphor), the ruling generals want middle-management colonels to be effective in pacifying the peasants employed underneath them. If the peasants are more docile and less demanding because of an ideology or religion (such as "Total Quality Management" or "Teamwork" or "Learning Organization"), so much the better. However, colonels who are too charismatic or successful are a problem, because they might threaten to stage a coup.

The metaphor aside, the Junta hypothesis is that the behavior of economic indicators such as unemployment and corporate profits reflects the ability of large corporations to hold down labor compensation. The issue of what this means for the future of large corporations is a topic for a future essay.