Issues in Science and Technology, Online A Publication of National Academy of Sciences National Academy of Engineering Institute of Medicine University of Texas at Dallas
  • Home
  • Back Issues
  • Search

Is Information Technology creating a productivity boom?

Stephen S. Roach

No Productivity Boom for Workers

Information technology has yet to deliver on its promise of faster productivity growth.

America's love affair with the new technologies of the Information Age has never been more intense, but nagging questions remain about whether this passion is delivering on its promise to accelerate growth in productivity. Corporate spending on information technology hardware is now running in excess of $220 billion per year, easily the largest line item in business capital spending budgets. And that's just the tip of the cost iceberg, which has been estimated at three to four times that amount if the figure includes software, support staff, networking, and R&D--to say nothing of the unrelenting requirements of an increasingly short product-replacement cycle.

Many believe there are signs that the long-awaited payback from this technology binge must now be at hand. They look no further than the economic miracle of 1997, a year of surging growth without inflation. How could the U.S. economy have entered the fabled land of this "new paradigm" were it not for a technology-led renaissance in productivity?

The wisdom of corporate America's enormous bet on information technology has never been tested by a cyclical downturn in the real economy.

The technology-related miracles of 1997 go well beyond the seeming disappearance of inflation. The explosion of the Internet, the related birth of electronic commerce, and the advent of fully networked global business are widely viewed as mere hints of the raw power of America's emerging technology-led recovery. The most comprehensive statement of this belief was unveiled in a legendary article in Wired by Peter Schwartz and Peter Leyden. It argues that we "are riding the early waves of a 25-year run of a greatly expanding economy." It's a tale that promises something for everyone, including the disappearance of poverty and geopolitical tensions. But in the end, it's all about the miracles of a technology-led resurgence in productivity growth. This futuristic saga has become the manifesto of the digital age.

Against this backdrop, the "technology paradox"--the belief that the paybacks from new information technologies are vastly overblown--seems hopelessly outdated or just plain wrong. Could it be that the hype of the Information Age is supported by economic data. Ultimately, the debate boils down to productivity, which is the benchmark of any economy's ability to create wealth, sustain competitiveness, and generate improved standards of living. Have the new technologies and their associated novel applications now reached a critical mass that is introducing a new era of improved and sustained productivity growth that benefits the nation as a whole? Or does the boom of the 1990s have more to do with an entirely different force: namely, the tenacious corporate cost-cutting that has benefited a surprisingly small proportion of the actors in the U.S. economy?

A glacial process

For starters, we should remember that shifts in national productivity trends are typically slow to emerge. That shouldn't be surprising; aggregate productivity growth represents the synergy between labor and capital, bringing into play not only the new technologies that are embedded in a nation's capital stock but also the skills of workers in using them to boost their productivity.

The paradox begins on the capital stock side of the productivity equation, long viewed as a key driver of any nation's aggregate productivity potential. Ironically, although surging demands for new information technologies have boosted overall capital spending growth to an 8.5 percent average annual pace over the period from 1993 to 1996 (a four-year surge unmatched since the mid-1960s), there has been no concomitant follow-through in the rate of expansion of the nation's capital stock. Indeed, the growth of the total stock of business's capital over the 1990-1996 interval has averaged only 2 percent, the slowest pace of capital accumulation in the post-World War II era and only half the 4 percent average gains recorded in the heyday of the productivity-led recovery in the 1960s.

There is no inherent inconsistency between information technology's large capital-spending share and small capital-stock share. The disparity reflects a very short product-replacement cycle and the related implication that about 60 percent of annual corporate information technology budgets goes toward replacement of outdated equipment and increasingly frequent product upgrades. In other words, there is little evidence of a resurgence in overall capital accumulation that would normally be associated with an acceleration in productivity growth.

At the same time, the news on the human capital front is hardly encouraging. In particular, there is little evidence that the educational attainment of U.S. workers has moved to a higher level, which should also be a feature of an economy that is moving to higher productivity growth. The nationwide aptitude test results of graduating high school seniors remain well below the levels of the 1960s. Companies may be working smarter, but there are few signs that this result can be traced to the new brilliance of well-educated and increasingly talented workers.

Productivity is all about delivering more output per unit of work time. It is not about putting in more time on the job.

It is important to understand the historical record of shifts in aggregate productivity growth trends. Acceleration was slow to emerge in the 1960s, with the five-year trend moving from 1.75 percent in the early part of the decade to 2.25 percent at its end. Similarly, the great slowdown that began in the late 1970s saw a downshift in productivity growth, from 2 percent to 1 percent, unfold over 5 to 10 years. Even the 1960s changes fell well short of the heroic claims of the New Paradigmers, who steadfastly insist that U.S. productivity growth has gone from 1 percent in the 1980s to 3 to 4 percent in the latter half of the 1990s. Such an explosive acceleration in national productivity growth would outstrip any historical experience (see Figure 1).

But perhaps it is most relevant to examine that slice of activity where the new synergies are presumed to be occurring: the white-collar services sector. According to U.S. Department of Commerce statistics, fully 82 percent of the nation's total stock of information technology is installed there, in retailers, wholesalers, telecommunications, transportation, financial services, and a wide array of other business and personal service establishments. Not by coincidence, around 85 percent of the U.S. white-collar work force is employed in the same services sector. Thus, the U.S. productivity debate is all about the synergy, or lack thereof, between information technology and white-collar workers.

Where the rubber meets the road

A look at the shifting mix of U.S. white-collar employment provides some preliminary hints about what lies at the heart of the U.S. productivity puzzle. In recent years, employment growth has slowed most sharply in the back-office (that is, processing) categories of information-support workers who make up 29 percent of the service sector's white-collar work force. In contrast, job creation has remained relatively vigorous in the so-called knowledge-worker categories--the managers, executives, professionals, and sales workers that account for 71 percent of U.S. white-collar employment.

Increasing the productivity of knowledge workers is going to be far more difficult to achieve than previous productivity breakthroughs for blue-collar and farm workers.

The dichotomy between job compression in low value-added support functions and job growth in high value-added knowledge-worker categories is an unmistakable and important byproduct of the Information Age. Capital-labor substitution works at the low end of the value chain, as evidenced by an unrelenting wave of back-office consolidation, but it is not a viable strategy at the high end of the value chain, where labor input tends to be cerebral and much more difficult to replace with a machine. Consequently, barring near-miraculous breakthroughs in artificial intelligence or biogenetic reprogramming of the human brain, productivity breakthroughs in knowledge-based applications should be inherently slow to occur in the labor-intensive white-collar service industry.

Debunking the measurement critique

There are many, of course, who have long maintained that the U.S. productivity puzzle is a statistical illusion. Usually this argument rests on the presumed understatement of service sector output (the numerator in the productivity equation). This understatement reflects Consumer Price Index (CPI) biases that deflate a current-dollar measure of output with what is believed to be an overstated price level. But there's also a sense that statisticians are simply unable to capture that amorphous construct, the service sector "product." That may well be the case, although I note that last summer's multiyear (benchmark) revisions to the Gross Domestic Product (GDP) accounts, widely expected to uncover a chunk of the "missing" output long hinted at by the income side of the national accounts, left average GDP (and productivity) growth essentially unaltered over the past four years.

I worry more about accuracy in measuring the denominator in the productivity equation: hours worked. Existing labor-market surveys do a reasonably good job of measuring the number of employed workers in the United States, but I do not believe the same can be said for the work schedule of the typical employee. I maintain that working time has lengthened significantly over the past decade and could well reduce the accuracy of the labor input number used to derive national productivity.

Ironically, this lengthening of work schedules appears to be closely tied to an increase in work away from the office that is being facilitated by the new portable technologies of the Information Age: laptops, cellular telephones, fax machines, and beepers. Many white-collar workers are now on the job much longer than the official data suggest. Productivity is all about delivering more output per unit of work time. It is not about putting in more (unmeasured ) time on the job. If work time is underreported, then productivity will be overstated no matter what problems exist in the output measurement.

According to a recent Harris Poll, the median number of hours worked per week in the United States rose from 40.6 in 1973 to 50.8 in 1997. This stands in sharp contrast to the 35-hour weekly work schedule assumed in the government's official estimates of productivity. U.S. workers obviously feel they are working considerably longer hours than Washington's statisticians seem to believe. The government's companion survey of U.S. households hints at the same conclusion; it estimates the average 1996 work week in the nonfarm economy at close to 40 hours. That's far short of the 51 hours reported in the Harris Poll but still considerably longer than the work week used in determining official productivity figures.

Analysis suggests that underreporting of work schedules since the late 1970s has been concentrated in the services sector. The discrepancy is particularly large in the finance, insurance, and real estate (FIRE) component. Similar discrepancies are evident in wholesale and retail trade and in a more narrow category that includes a variety of business and professional services. By contrast, recent trends in both establishment- and household-based measures of work schedules in manufacturing, mining, and construction--segments of the U.S. economy that also have the most reliable output figures--tend to conform with each other.

So what does all this mean for aggregate productivity growth? To answer this question, I have performed two sets of calculations. The first is a reestimation of productivity growth under the work-week assumptions of the Labor Department's household survey. On this basis, productivity gains in the broad services sector (a nonmanufacturing category that also includes mining and construction) averaged just 0.1 percent annually from 1964 through 1996, about 0.2 percentage points below the anemic 0.3 percent trend derived from the establishment survey. In light of the results of the Harris Poll, this is undoubtedly a conservative estimate of the hours-worked distortion in productivity figures. Indeed, presuming that work schedules in services move in tandem with the results implied by the Harris Poll, our calculations suggest that service sector productivity growth is actually lower, by 0.8 percentage points per year, than the government's official estimates.

A final measurement critique of the productivity results also bears mentioning: the belief that statistical pitfalls can be traced to those sectors of the economy (such as services) where the data are the fuzziest. This point of view has been argued by Alan Greenspan and detailed in a supporting paper by the Federal Reserve's research staff. In brief, this study examines productivity results on a detailed industry basis and concludes that because the figures are generally accurate in the goods-producing segment of the economy, there is reason to be suspicious of results in the service sector, especially in light of well-known CPI biases in this segment of the economy. But it may simply be inappropriate to divide national productivity into its industry-specific components. Distinctions between sectors and industries are increasingly blurred by phenomena such as outsourcing, horizontal integration, and the globalization of multinational corporations.

We have performed some simple calculations that suggest that productivity growth would be lower in manufacturing and higher in services if a portion of the employment growth in the surging temporary staffing industry were correctly allocated to the manufacturing sector rather than completely allocated to the services sector as is presently the case. (We start with the assumption that about 50 percent of the hours worked by the help supply industry provides support for manufacturing activities, which is broadly consistent with anecdotal reports from temporary help companies. That 50 percent can then be subtracted from the services sector, where it currently resides in accordance with establishment-based employment accounting metrics, and added back into existing estimates of hours worked in manufacturing. This knocks about 0.5 percentage points off average productivity growth in manufacturing over the past six years and boosts productivity growth in the much larger service sector by about 0.1 percentage point over this same period.)

All this is another way of saying that there are two sides to the productivity measurement debate. Those focusing on the output side of the story have argued that productivity gains in services may have been consistently understated in the 1990s. Our work suggests that the biases stemming from under-reported work schedules could be more than offsetting, leaving productivity trends even more sluggish than the official data suggest.

A new cost structure

Yet another element of the productivity paradox is the link between America's open-ended commitment to technology and the flexibility of corporate America's cost structure, especially in the information-intensive services sector. For most of their long history, U.S. service companies were quintessential variable-cost producers. Their main assets were workers, whose compensation costs could readily be altered by hiring, firing, and a relatively flexible wage-setting mechanism.

Now, courtesy of the Information Age and a heavy investment in computers and other information hardware, service companies have unwittingly transformed themselves from variable- to fixed-cost producers, which denies this vast segment of the U.S. economy the very flexibility it needs to boost productivity in an era of heightened competition. Moreover, the burdens of fixed costs are about to become even weightier thanks to the outsized price tag on the Great Year 2000 Fix--perhaps $600 billion--yet another example of dead weight in the Information Age.

A few numbers illustrate the magnitude of the new technology bet and its impact on business cost structures. Between 1990 and 1997, corporate America spent $1.1 trillion (current dollars) on information technology hardware alone, an 80 percent faster rate of investment than in the first seven years of the 1980s. At the same time, the information technology share of business's total capital stock (expressed in real terms) has soared from 12.7 percent in 1990 to an estimated 19.1 percent in 1996. The recent surge in this ratio is a good approximation of the ever-expanding increases in fixed technology costs that are now viewed as essential in order to keep transaction-intensive and increasingly global service companies in business.

To be sure, a large portion of these outlays is written off quickly. Nevertheless, with their tax-based service lives typically clustered over three to five years, about $460 billion still remains on the books, which is a little over 40 percent of cumulative information technology spending since 1990. This is hardly an insignificant element of overall corporate costs; by way of comparison, total U.S. corporate interest expenses are presently running at about $400 billion annually.

Let me also stress that the wisdom of corporate America's enormous bet on information technology has never been tested by a cyclical downturn in the economy. Under such circumstances, it is highly unlikely that U.S. businesses will prune those costs aggressively. After all, information technology is now widely viewed as a critical element of the business infrastructure, essential to operations and therefore not exactly amenable to the standard cost-cutting typically employed to sustain profit margins. Lacking the discretion to pare the technology, managers will be under all the more pressure to slash labor costs. Yet that strategy may also be quite difficult to implement in the aftermath of the massive head-count reductions made earlier in the 1990s.

Whenever it comes, the next recession will be the first cyclical downturn of the Information Age. And it will find corporate America with a far more rigid cost structure than has been the case in past recessions. This suggests that the next recession might also take a far greater toll on corporate earnings than has been the case in past recessions, a possibility that is completely at odds with the optimistic profit expectations that are currently being discounted by an ever-exuberant stock market. In short, the next shift in the business cycle could well provide an acid test of the two competing scenarios of the productivity-led renaissance and the technology paradox. Stay tuned.

Cost cutting vs. productivity

Let me propose an alternative explanation for the so-called earnings miracles of the 1990s. I am not one of those who believes that explosive gains in the stock market over the past three years are a direct confirmation of the (unmeasured) productivity-led successes in boosting corporate profit margins. A better explanation might be an extraordinary bout of good old-fashioned slash-and-burn cost cutting. Consider the unrelenting surge of downsizing that were a hallmark of the 1990s. Whether such strategies took the form of layoffs, plant closings, or outsourcing, the result was basically the same--companies were making do with less. Sustained productivity growth, by contrast, hinges on getting more out of more--by realizing new synergies between rapidly growing employment and the stock of capital. That outcome has simply not been evident in the lean and mean 1990s. As can be seen in Figure 2, recent trends in both hiring and capital accumulation in the industrial sector have been markedly deficient when compared with the long sweep of historical experience.

How can this be? Doesn't the confluence of improved competitiveness, upside earnings surprises, and low inflation speak of a nation that is now realizing the fruits of corporate productivity? Not necessarily. In my view, it is impossible to discern whether such results have been driven by intense cost cutting or by sustained productivity growth. The evidence, however, weighs heavily in favor of cost cutting. Not only is there a notable lack of improvement in official productivity results for the U.S. economy, there is also persuasive evidence that corporate fixation on cost control has never been greater.

This conclusion should not be surprising. It simply reflects the extreme difficulty of raising white-collar productivity. This intrinsically slow process may be slowed even further if the challenge is to boost the cerebral efficiencies of knowledge workers. And slow improvement may not be enough for corporate managers (and shareholders) confronting the competitive imperatives of the 1990s. As a result, businesses may have few options other than more cost cutting. If that's so, then the endgame is far more worrisome than the one implied in a productivity-led recovery. In the cost-cutting scenario, companies will become increasingly hollow, lacking both the capital and the labor needed to maintain market share in an ever-expanding domestic and global economy.

Indeed, there are already scattered signs that corporate America may have gone too far down that road in order to boost profits. Recent production bottlenecks at Boeing and Union Pacific are traceable to the excesses of cost cutting and downsizing that occurred in the late 1980s and early 1990s. In a period of sustained growth in productivity, corporate growth is the antidote to such occurrences. But in a world of unrelenting cost cutting, bottlenecks will become far more prevalent, particularly with the rapid expansion of global markets. And then all the heavy lifting associated with a decade of corporate restructuring could quickly be squandered.

The fallacy of historical precedent

Yet another flaw in the productivity revivalist script is the steadfast belief that we have been there before. The New Paradigm proponents argue that the Agricultural Revolution and the Industrial Revolution were part of a continuum that now includes the Information Age. It took a generation for those earlier technologies to begin bearing fruit, and the same can be expected of the long-awaited technology payback of the late 20th century. Dating the advent of new computer technologies to the early 1970s, many are quick to argue that the payback must finally be at hand.

This is where the parable of the productivity-led recovery really falls apart. The breakthroughs of the Agricultural and Industrial Revolutions were all about sustained productivity growth in the creation of tangible products by improving the efficiency of tangible production techniques. By contrast, the supposed breakthroughs of the Information Age hinge more on an intangible knowledge-based product that is largely the result of an equally intangible human thought process.

It may well be that white-collar productivity improvements are simply much harder to come by than blue-collar ones. That's particularly true in the the new global village. It's a cross-border operating environment that also crosses multiple time zones and involves new complexities in service-based transactions. That's certainly the case in the financial services industry, where increasingly elaborate products with multidimensional attributes of risk (such as currencies, credit quality, and a host of systemic factors) are now traded 24 hours a day. In the Information Age, much is made of the exponential growth of computational power. I would argue that the complexity curve of the tasks to be performed has a similar trajectory, suggesting that there might be something close to a standoff between these new technological breakthroughs and the problems they are designed to solve.

The issue of task complexity is undoubtedly a key to understanding the white-collar productivity paradox. The escalating intricacy of knowledge-based work demands longer schedules, facilitated by the portable technologies that make remote problem-solving feasible and, in many cases, mandatory. Whether the time is spent surfing the Web, performing after-hours banking, or hooking up to the office network from home, hotel, or airport waiting lounge, there can be no mistaking the increasingly large time commitment now required of white-collar workers.

Nor is it clear that information technologies have led to dramatic improvements in time management; witness information overload in this era of explosive growth in Web-based publishing, a phenomenon that far outstrips the filtering capabilities of even the most powerful search engines. The futuristic saga of the productivity-led recovery fails to address the obvious question: Where does this incremental time come from? The answer is that it comes increasingly out of leisure time, reflecting an emerging conflict between corporate and personal productivity.

This is consistent with the previous critique of productivity measurement. Productivity enhancement, along with its associated improvements in living standards, is not about working longer but about adding value per unit of work time. This is precisely what's lacking in the Information Age.

Paradigm lost?

There can be no mistaking the extraordinary breakthroughs of the new technologies of the Information Age. The faster, sleeker, smaller, and more interconnected information appliances of the late 1990s are widely presumed to offer a new vision of work, leisure, and economic and social hierarchies. But is this truly the key to faster productivity growth for the nation?

My answer continues to be "no"; or possibly, if I don my rose-colored glasses, "not yet." Improvements in underlying productivity growth are one of the most difficult challenges that any nation must confront. And increasing the productivity of knowledge workers in particular is going to be far more difficult to achieve than previous productivity breakthroughs for blue-collar and farm workers.

That takes us to the dark side of America's technology paradox. Rushing to embrace the New Paradigm entails a real risk of overlooking the most basic and powerful benefit of an improvement in overall productivity: an increase in the national standard of living. On this, the evidence is hardly circumstantial: more than 15 years of virtual stagnation in real wages, an unprecedented widening of inequalities in income distribution, and a dramatic shift in the work-leisure tradeoff that puts increasing stress on family and personal priorities. At the same time, there can be no mistaking the windfalls that have accrued to a small slice of the U.S. population, mainly those fortunate managers, executives, and investors who have benefited from the corporate earnings and stock market bonanza of the 1990s.

In the end, I continue to fear that much of the debate over the fruits of the Information Age boils down to the classic power struggle between capital and labor. I find it difficult to believe that corporate America can cut costs forever; there really is a limit to how far managers can take the credo of "lean and mean," and there are signs that the limit is now in sight. I find it equally difficult to believe that workers will continue to acquiesce in a system that rewards few for the efforts of many, especially in view of the dramatic cyclical tightening of the labor market that has taken the national unemployment rate to its lowest level in 24 years. A recent upturn in the wage cycle suggests that the forces of supply and demand are now beginning to weigh in with the same cyclical verdict. All this implies that the pendulum of economic power may be starting a long-overdue swing from capital back to labor, repeating the timeworn patterns of power struggles past.

Like it or not, the New Paradigm perception of a technology-led productivity renaissance is about to meet its sternest test. That test should reflect not only the social and economic pressures of worker backlash but also a classic confrontation between cost-cutting tactics and the pressures of the business cycle. Moreover, to the extent that the technology paradox is alive and well (and that remains my view) the days of ever-expanding profit margins, subdued inflation, and low interest rates could well be numbered. Needless to say, such an outcome would come as a rude awakening for those ever-exuberant financial markets that are now priced for the perfection of the Long Boom.

Recommended reading

Alan S. Blinder and Richard E. Quandt, "The Computer and the Economy," The Atlantic Monthly, December 1997.

Carol Corrado and Lawrence Slifman, Decomposition of Productivity and Unit Costs. Washington, D.C.: Federal Reserve Board, Occasional Staff Studies, November 18, 1996.

Stephen D. Oliner and Daniel E. Sichel, "Computers and Output Growth Revisited: How Big is the Puzzle?" Brookings Papers on Economic Activity, 1994: 2.

Stephen S. Roach, "Technology and the Services Sector: America's Hidden Competitive Challenge," in B.R. Guile and J.B. Quinn, Eds., Technology and Services: Policies for Growth, Trade, and Employment. Washington, D.C.: National Academy Press, 1988.

Stephen S. Roach, "Services Under Siege--The Restructuring Imperative," Harvard Business Review, September-October 1991.

Stephen S. Roach, "The Hollow Ring of the Productivity Revival," Harvard Business Review, November-December 1996.

Stephen S. Roach, "Outsourcing and Productivity Accounting." Morgan Stanley Global Economic Forum,, February 28, 1997.

Stephen S. Roach, "The Worker Backlash," The New York Times, August 24, 1997.

Peter Schwartz and Peter Leyden, "The Long Boom," Wired, July 1997.