The New Yorker - November 27, 2000THE PRODUCTIVITY MIRAGE
Are computers really that important?
John Cassidy
It is hard to pick up a business newspaper or magazine these days without coming across a reference to the "productivity miracle" that has reportedly taken place in the American economy during the past few years. Chief among the boosters, surprisingly enough, is Alan Greenspan, the normally lugubrious chairman of the Federal Reserve. In a recent speech to the National Governors' Association, Greenspan hailed the "remarkable changes" that are occurring in American business and traced them to the invention of the transistor, more than fifty years ago, which led to "the microprocessor, the computer, satellites, and the joining of laser and fiber-optic technologies." Innovations like these have created "an enormous new cape city to capture, analyze, and disseminate information," Greenspan went on "Indeed, it is the proliferation of in formation technology throughout the economy that makes the current period appear so different from preceding decades." Other commentators have been equally rhapsodic about the wonders of the microchip, but Greenspan was more specific than most of them. "The major contribution of advances in information technology and their incorporation into the capital stock has been to reduce the number of worker hours required to produce the nation's output, our proxy for productivity growth," he argued.
If Greenspan is right, and productivity growth has taken a quantum leap, it is a truly historic development. Productivity is not just another humdrum financial statistic, like the inflation rate, yesterday's Nasdaq close, or the trade deficit. It is the key to raising living standards, paying for the aging of the baby boomers, and much else besides. When productivity grows steadily, as it did in the quarter century after the Second World War, workers' wages also increase steadily, and most people end up much better off than their parents. But when productivity growth stalls, as it did during the nineteen-seventies and nineteen-eighties, living standards stagnate, and upward mobility disappears.
At the moment, the future looks as if it will the Eisenhower years more than it will the Carter era, but things always look good at the peak of a boom. In America during the nineteen-twenties, in Japan and Britain during the nineteen-eighties, and in Korea and Thailand during the nineteen-nineties, economists spied a New Economy, which, on each occasion, turned out to be a mirage. Our economic miracle may turn out to be equally chimerical. While Bill Gates, Steve Jobs, and Larry Ellison have changed the way many people work, the scale of the transformation has been exaggerated. The fact that a typical office worker now has a computer linked to the Internet on his desk instead of an ink blotter, doesn't necessarily mean he is any more productive.
Productivity growth may sound like an arcane economic concept, but it is actually pretty simple to grasp. If a baker works an eight-hour day and makes four hundred loaves of bread, which he sells to the public for a dollar each, his productivity is fifty dollars an hour. If he somehow manages to bake eight hundred loaves in the same workday, his productivity will double to a hundred dollars an hour. The same principles apply to business in general. The economy's productivity is calculated by tallying the value of the final goods and services produced in the country--the gross domestic product--and dividing this figure by the total number of hours worked in the business sector. If output grows faster than hours worked, productivity rises. If hours worked grow faster than output, productivity falls.
Greenspan's story comes in two parts: American workers are much more productive than they used to be; and the reason is the proliferation of information technology. The productivity figures from the Department of Labor's Bureau of Labor Statistics seem to support the first part: they show strong growth in recent years. Between the second quarter of 1995 and the second quarter of this year, output per hour in the non-farm business sector of the economy (the sector that economists usually look at) rose at an annual rate of 2.9 per cent, which is more than double the 1.4-per-cent annual growth between 1973 and 1995 and equal to the 2.9-per-cent annual growth between 1947 and 1973, a period that economists often refer to as a golden era in economic history. Now, 2.9 per cent may look like a small number, but the productivity debate is all about small numbers, which, thanks to the magic of compounding, translate into big differences over time. If productivity grows at an annual rate of 2.9 per cent, living standards double in twenty-four years, or one generation; at 1.4-per-cent productivitygrowth, living standards take almost fifty years, or two generations, to double.
That is the good news, but it shouldn't be taken at face value. In every economic upswing, firms work their employees and machines more intensely; output surges; and productivity growth surges with it. When the economy slows, as it inevitably does at some point, workers and machines maybe used less intensely; and productivity growth slips back. In the mid-nineteen-eighties, for example, output per hour rose just as rapidly as it has over the past few years, but the surge was cyclical, and it didn't last. At least some of the current productivity revival will prove to be equally temporary.
One way to get beyond short-term ups and downs is to look at trends across entire cycles, which include recessions as well as upswings. The last recession began in January, 1991, the month that the Gulf War started. Between then and now, output per hour has grown by about two per cent a year (a good deal less than the 2.9-per-cent annual growth since the middle of 1995). During the previous cycle, which began with the 1982 recession and ended in 1990, output per hour grew by 1.5 per cent a year. In comparing the two cycles, it appears that underlying productivity growth has risen by about 0.5 per cent a year. This is a significant increase, but it is hardly "remarkable."
There is another reason that the recent figures demand scrutiny: the government may be undercounting the number of hours people are working. For a baker's productivity to increase, he must male more loaves during his eight-hour shift, or the same number of loaves in a shorter workday. It doesn't count if he bakes more bread by working longer hours. The federal government says that the workweek has barely changed in ten years, but that seems dubious, especially as it applies try the two-thirds of the workforce that is employed in the service sector. Many of these workers, such as investment bankers, editors, and insurance salesmen are not paid on an hourly basis; they receive an annual salary, or commission, and they organize their own schedules. With the aid of faxes, cell phones, and laptop computers, they can work at home, or on to road, as well as in the office.
"No one really knows how much work time is being put in the service sector," Stephen Roach, the chief economist at Morgan Stanley Dean Witter, who has studied this issue, told me. "The Labor Department researchers ring up a firm's human-resources department and ask how long a typical white-collar worker is working. The clerk checks the computerized time sheet, and it says thirty-five hours, forty hours, or whatever. I suspect it's a good deal more than that, as the workplace has expanded to home offices, planes, trains, and foreign hotel rooms in the middle of the night." If Roach is right--and Dora Costa, an M.I.T. economist, has unearthed some United States Census data that confirm that well-educated, upper-income Americans are working longer hours than they used to--then information technology is creating more work, not less; hundreds of millions of hours of work are going unrecorded by the government; and the productivity figures are biased upward.
Even if we assume that productivity growth has increased, how much credit should go to computers and the Internet? The economic importance of on-line retailing, satellite communications, and other digital wonders is often exaggerated. In 1998, the last year for which figures are available, the total value of the Internet economy, including investments in the network's infrastructure, subscriptions to Internet service providers, and on-line commerce of all descriptions, came to less than a fiftieth of the gross domestic product. It may be almost 2001, but the majority of businesses still have little or nothing to do with the Internet.
What the boom has seen is a big increase in capital spending. Consider the baker in his shop. One way for him to raise his output per hour is to come up with a recipe for bread that bakes more quickly, but that isn't easy. Another way is to buy a bigger oven so that he can bake more loaves at the same time. In the past decade, American firms have been following the second tack, buying machinery at a rate rarely seen before. During the nineteen-nineties, business investment in plant and equipment of all kinds almost doubled--from $626 billion in 1990 to $1.2 trillion in 1999-- with much of the increase going to information technology. Last year alone, businesses and government agencies spent a total of $433 billion on computers and information processing. Two recent studies--one by Dale W. Jorgenson, of Harvard, and Kevin J. Stiroh, of the Federal Reserve Bank of New York, the other by Stephen D. Oliner and Daniel E. Sichel, who work at the Federal Reserve Board, in Washington-- attribute about half of the post-1995 acceleration in productivity growth to the fact that firms have provided their employees with better equipment, much of it information-processing equipment.
These studies imply that all the money spent on information technology was not wasted, but they offer no encouragement for those who believe that we are witnessing something radically new. It turns out that computers and fibre-optic cables are just like mechanical diggers and bakers' ovens. When firms buy a lot of them a process economists call "capital deepening," their workers' productivity increases. But when firms cut back on investment spending, as they have begun to do in the past few months, this form of productivity growth will eventually decline.
Greenspan doesn't agree that information technology is just another form oi capital. He believes that, by reducing the amount of uncertainty in the economy. it has revolutionized the way firms do business. "Information technology raises output per hour in the total economy principally by reducing hours worked on activities needed to guard productive processes against the unknown and the unanticipated," he argued in a recent speech.
To translate Greenspan's theory into the world of baking, think of a large bread company that owns many individual bakeries. Before the dawn of the information age, such a business would have had only a vague idea of how many loaves its customers might demand em any given day. To protect against unexpected surges in demand, it would have kept extra bakeries open, with extra ovens, flour, and workers on hand. Some of the bread produced in these bakeries would have gone to waste, but that was the price of keeping customers satisfied. The microchip did away with this world. Thanks to retail bar codes, the bread company can track every loaf it sells: by monitoring its retail Web site, it can see how much bread its biggest customers, such as restaurants and supermarkets will need tomorrow; and by installing; global-positioning satellites in its delivery trucks it can insure that its loaves go to the right place at the right time. With all this timely information, the firm n longer needs much spare capacity. It can close down surplus bakeries and lay off employees without worrying that customers will be caught short. As a result its productivity will rise.
Greenspan's theory is clever, but it doesn't fit all the facts. The strongest evidence in its favor is a surge in something called "multifactor productivity," which is a measure of workers' efficiency after allowing for improvement n the equipment they are using. According to the Bureau of Labor Statistics, between 1995 and 1998 the annual rate of multifactor-productivity growth more than doubled compared with the growth rate during the previous five years. This was a sharp increase, undoubtedly, but multifactor productivity always rises in a boom, and it is unclear how much of the upturn is permanent.
If Greenspan is right, moreover, the industries that make the greatest use of information technology should have enjoyed the most rapid rises in productivity. This hasn't happened. As Jorgenson and Stiroh report, "the empirical record provides little support for the 'new economy' picture of spillovers cascading from information technology producers onto users of this technology." About four-fifths of total business spending on information technology takes place in just three industries: wholesale and retail trade; finance, insurance, and real estate; and business services. These are all industries in which productivity growth has lagged for years. In commercial banking, for example, the rate of productivity growth actually fell between 1995 and 1998 compared with the previous five years, according to the Bureau of Labor Statistics.
Facts like these raise a tricky question: Just how useful are computers? Undeniably, it is a marvel that, for less than two thousand dollars, I can buy a box containing more computing power than the I.B.M mainframe that NASA used to land Apollo 11 on the moon. The issue is, what can I use all this memory and processor speed to accomplish, since I don't have a spacecraft to navigate? Write this article? I did that. But I could have done it fifteen years ago on my first computer, an I.B.M. XT that ran at about five megahertz and had a flashing green cursor. In some ways, the old XT was faster than my spiffy new Apple Cube, because the software it used was much simpler. Surf the Internet? I did that, too, to research some of the material for this article. It was much quicker than going to the library or calling people and asking them to send me their articles. While I was on the Net, though, I also spent a good deal of time checking out the English soccer scores and E-mailing my friends and family, neither of which helped me finish the piece faster.
True, many people are less easily distracted than I am, but, even if they confine their computer use to job-related tasks, they don't necessarily become more productive, as a host of studies has shown. Take the simple task of writing business letters. A study of I.B.M. workers found that letters that were handwritten or typed were modified, on average, eight times. Letters produced on a word processor were modified, on average, forty-one times, but with no discernible difference in quality. Or consider the wealth of business information available at the touch of a mouse, something that Greenspan identifies as a major benefit of the computer revolution. Does it really allow firms to operate more efficiently? Thomas K. Landauer, a former researcher at Bell Labs who has spent twenty-years studying the impact of computers, hardly thinks so. "Once a computer is in place and all of a company's inventory, sales and expense data are available for electronic access and manipulation, the opportunities for analysis are dizzying," he wrote in his book "The Trouble with Computers." He continued, "Every manager now gets reports measured in stack-feet per month. Nobody can read it all or digest what they've read." Landauer's grim conclusion: "We expected computers to bring across-the-board productivity help, work efficiency improvements for small and large alike. This they have not delivered."
The most dramatic increases in productivity have been in manufacturing, but Greenspan's theory doesn't really hold up there, either. In aircraft assembly, the car industry, and many other businesses, the "just in time" production techniques that he is referring to were widely adopted before productivity growth took off in 1995. They were invented by Japanese firms, such as Toyota and Mitsubishi, during the nineteen-seventies. Starting in the early eighties, American firms, such as General Motors and Boeing, copied their Japanese rivals slavishly. The resulting changes certainly boosted the productivity of American industry, but they had as much to do with encouraging teamwork and flexibility among the workers as with information technology.
Some manufacturers have invested heavily in information technology but have seen little or no increase in productivity. The newspaper industry is a good example. In the past decade or so, laptop computers, faxes, and the Internet have revolutionized news gathering, and satellite technology has transformed the printing and distribution side of the business. It's hard to believe that these changes wouldn't have made people working in newspapers more efficient, but they haven't. Between 1987 and 1997, the latest period for which the Bureau of Labor Statistics has figures, productivity in the newspaper industry fell by 2.3 per cent a year.
There is one manufacturing business in which digital technology has certainly boosted the reported rate of productivity growth: the computer industry itself. The most startling fact about the apparent productivity revival, rarely mentioned by Greenspan and other optimists, is the extent to which it has been concentrated in firms that produce information technology, rather than in firms that use information technology. Between 1995 and 1997, for example, workers at computer manufacturers, such as Dell and Compaq, raised their productivity at an annual rate of 41.3 per cent. Workers at firms that make electronic components and accessory such as Cisco Systems, increased their productivity by 23.3 per cent a year.
The computer sector, which includes computers, peripherals, and software, generates about a thirtieth of the gross domestic product, but it has accounted for roughly a third of economy-wide productivity growth in recent years, according to Michael Harper, an economist at the Bureau of Labor Statistics. Some mischievous economists go further than Harper and attribute virtually all productivity revival to the computer industry. Robert J. Gordon, of Northwestern University, reckons that, after allowing for the effects of the economic cycle, almost nine-tenths of the economy has seen no increase in multifactor-productivity growth since 1995. Other economists have challenged how Gordon arrived at this conclusion, but his main point stands: productivity growth has been heavily based in the technology sector.
In view of this, it is surprising that more attention hasn't been paid to the unusual way in which the government values the output of the computer industry, and the impact this has on productivity statistics. In 1999, final sales of computers came to $92.5 billion. This is a straightforward figure--firms, households, and government departments paid this much money for computers. But the Commerce Department didn't use $92.5 billion when it was reporting real G.D.P.; it used $245.9 billion. As a result, final sales of computers increased by 47.2 per cent over the previous year instead of 6.3 per cent. The difference arose because of a controversial statistical technique, "hedonic pricing," which adjusted the raw dollar figure for the fact that computers are constantly improving in power and quality.
The details of hedonic pricing are complicated, but the idea behind it isn't: quality improvements are worth something even if they aren't reflected in a good's price, so they should be included in the official statistics. Consider, for example, the workers at Compaq whose job is to assemble a typical two-thousand dollar personal computer. Five years ago, such a computer might have featured a central processing chip running at a hundred megahertz, with a one-gigabyte hard drive. Today, a typical two-thousand dollar computer could well have a processor running at eight hundred megahertz, with a twenty-gigabyte hard drive. The statisticians at the Commerce Department argue that Compaq's workers are creating more value in the new computer, even though its price is the same as the old one. To allow for this, they increase the value of the new computer in the real G.D.P. (A similar method is used to adjust the output of the software industry upward.) This massaging makes the concept of "output" very slippery. It is as if the Commerce Department had decided that, because bread is more nourishing than it used to be, every two loaves the baker sells should be counted as three.
It isn't easy to ted how much impact hedonic pricing has on the over-all productivity statistics--Gene Epstein, of Barron's, and James Grant, of Grant's Interest Rate Observer, have been arguing about this for months--but, in a debate that hangs on one or two percentage points, the question isn't immaterial. Between 1995 and 1999, the high-technology sector accounted for about a quarter of the annual growth in G.D.P. (0.9 per cent of 3.8 per cent), according to J. Steven Landefeld, the director of the Commerce Department's Bureau of Economic Analysis. If a third of this contribution is attributable to hedonic pricing, which Landefeld says is a reasonable guess, hedonic pricing has boosted productivity growth by about 0.4 per cent a year, or about a seventh of the total.
It is also worth noting that other industrialized countries, such as Germany and Britain, don't use hedonic pricing for computers, and this is one reason that their economies appear to lag behind the United States' economy. The Bundesbank, Germany's central bank, recently recalculated German spending on information technology since 1991 using American methods. The old calculations showed spending growing by a niggardly six per cent a year. When the bank used hedonic pricing, the figure leaped to twenty-seven and a half per cent a year. Britain's Office for National Statistics carried out a similar exercise and found a similar result: introducing hedonic pricing for computers appeared to triple the growth rate of the British manufacturing industry during the past four years.
Carlyle, when he coined the term "dismal science," in the mid-nineteenth century, wasn't making a pejorative statement; he was merely reflecting the chronic pessimism of classical economists such as Thomas Malthus. The current boom has turned many of Malthus's successors into optimists, which is hardly surprising, given that virtually every economic indicator is at or near record levels. But, in the end, numbers, however detailed, can tell us only so much. If the economy were vastly more productive than it used to be, we would be able to tell simply by looking around. Fewer worker hours would be needed to produce the goods and services--from cars to jeans to haircuts--that make up the G.D.P., so most people would have more time on their hands. This is the future that Keynes, an economist who wasn't dismal in any sense, wrote about seventy years ago in his famous essay "The Economic Possibilities for Our Grandchildren." In Keynes's vision, technological progress would have advanced to such an extent that ordinary folk would be able to spend most of their time at leisure, reading books and cultivating their interest in the arts. (The Cambridge don, who married a Russian ballet dancer and spent his own spare time re-reading the Greek classics, had a quaint view of humanity, it must be said.) Instead, more people are working than ever before, and they are working longer hours, or so it seems.
In January of 1990, roughly a hundred and nineteen million people were working (legally) in the American economy. By October of this year, the number had risen to about a hundred and thirty-five million, Some of the sixteen million new workers are recent immigrants and youngsters, but many of them are people who used to be outside the labor force, such as retirees, disabled individuals, and former welfare recipients who had given up on trying to get jobs. About nine million of the new workers are women. More working-age women are employed than ever before--57.7 per cent of them, compared with 54.6 per cent a decade ago.
These statistics bear thinking about. Alan Greenspan, were he to present an accurate picture of the current boom, would have to add the advent of child care to the invention of the transistor as a major precipitating factor. The image of a tired woman rushing home from work to cook for her kids is not the one most people call to mind when they think about twenty-first-century capitalism, but it is just as much a part of the New Economy as any productivity miracle.