Monday, Oct. 12, 1998
Do Computers Really Save Money?
By GEORGE J. CHURCH
New York Presbyterian Healthcare System, product of the newly merged Columbia Presbyterian and New York-Cornell hospitals in Manhattan, is the very model of a modern medical establishment. Formed in January 1998, it brought together more than 25 health institutions and merged their operations to achieve economies of scale. The company has spent a cool $100 million just installing new computers in its facilities to get state-of-the-art performance--and cost-saving efficiencies. Has the network got its money's worth? That, says senior vice president Guy Scalzi, is not always easy to calculate.
Take an automated laboratory at New York Presbyterian Hospital. Conveyor belts transport blood or urine specimens in containers that resemble toy railroad cars from a collection point to a computerized analyzer. The machine takes a sample with a dipstick; the computer reads the results and flashes them to the monitor of the doctor in charge of the case. The lab will save the salaries of dozens of people who "used to move the specimens around by hand, read the test results on a screen and then telephone the doctor," says Scalzi. The lab cost $7 million to set up but is expected to save $2 million to $3 million a year.
O.K., that was easy. But the benefits of computerizing some nurse's duties are literally incalculable. Instead of writing charts by hand, a nurse can now type data into a computer. Or a nurse can press a button on a computer connected to the heart monitor of a patient down the hall and get up-to-the-second readings on heart rate, blood pressure and temperature without leaving her station. Yet the hospital has not reduced its nursing staff. Instead, nurses who once spent 60% of their time doing paperwork now spend that 60% at bedsides, giving patients personal attention. Sick people are better cared for, most nurses are happier in their jobs--but the hospital's saving is harder to measure.
As computers push their way further into every nook and cranny of America's complex economy, these experiences underline a paradox that has long puzzled almost everybody who comes into contact with thinking machines. Computers help all sorts of people do their jobs faster and more efficiently. Many enthusiasts expect the machines to transform the American economy and society as completely as the internal-combustion engine and electric power did, beginning roughly a century ago. But why is there so little hard numerical evidence that this is happening? In particular, if computers are sparking a new industrial revolution, why have the numbers that measure the growth in output per labor-hour of the U.S. economy been so persistently anemic? As Nobel prizewinner Robert Solow summarized in what economists now call the Solow Paradox, computers are everywhere--except in the productivity statistics.
This is not merely an academic issue. It cuts to the heart of a decision faced by businesses large and small: whether to go digital or not. No country has gone further than the U.S. in embracing the computer revolution; no country has a bigger stake in the productivity question. The issue: Are computers worth the billions we lavish on them?
There is much debate and no single explanation for the paradox. To begin with, economists generally agree that official figures understate productivity, though they quarrel sharply about how much and why. It is clear too that many buyers fail to get the most out of their computers--some because they try to make the computers perform functions to which they are not really adaptable; others because they buy computers more powerful and more expensive than they truly need; others still because they fail to appreciate how hard it will be to train their employees to use the machines effectively.
But one big explanation is that, as in the case of New York Presbyterian's nurses, computers often improve the efficiency and quality of work--indeed, of life--in ways difficult to express in numbers. That is especially true now that computers have moved heavily into service industries--health care, finance, law, advertising--where productivity was notoriously hard to measure even in precomputer days.
In manufacturing, the effects of computerization are easy to quantify. Take the computerized quality-control systems that the Measurex division of Honeywell Inc. sells to papermakers. The systems continuously measure such factors as the thickness of paper speeding through machines at up to 60 m.p.h. That enables operators to make adjustments that prevent breaks in the paper, avoiding shutdowns that can cost tens of thousands of dollars each. Rick Rowe, vice president of global operations for Honeywell-Measurex, claims that within 12 to 24 months, papermakers recover the $350,000 to $5 million they spend to buy the control systems. If you're in manufacturing, figuring out the cost-benefit ratio for computers is no sweat.
But in service fields the improvements are sometimes measurable, sometimes not--often both. Doug Kinzley, vice chairman of MGA Communications of Denver, figures that his advertising, public relations and marketing firm can put together a business proposal for a client in "maybe half the time it used to take" before the firm brought in a staff computer expert and installed individual computers for its 28 employees. That, he says, "gives us more thinking time. Our time is better spent on the actual work." The time saved is measurable, the benefits of proposing better-conceived ad campaigns to clients less so.
Dixie Cleaners, which operates three dry-cleaning shops in Pensacola, Fla., was losing $100,000 a year because clerks were failing to add "up charges" (surcharges for hard-to-clean clothes such as linen or silk garments) on sales tickets. Now a computer enters the extras automatically. Dixie also mails monthly computer-prepared bills to customers who open charge accounts. That pleases customers, who save a few minutes on every trip to the store. But how does one measure their increased satisfaction?
Cops in Blue Earth County, Minn., like those in many other places, carry laptops that enable them to tap into a statewide database and find out if a driver they've stopped for a traffic violation has any arrest warrants outstanding. The officers make 10 to 12 arrests a week, vs. two or three in pre-laptop days. That might count as measurable productivity, but how to quantify the benefit to society of haling lawbreakers into court rather than letting them roam free?
Erik Brynjolfsson, a professor of management at M.I.T. who has done extensive surveys of individual companies on this issue, asked senior executives what they hoped to get out of their investments in computer power. Their four top goals: to improve service to customers, target new customers, improve quality of products or services, and reduce total costs. Only the fourth objective would have much effect on productivity statistics as they are currently measured, which reflects an old-fashioned bias toward things that can be readily counted. Nowadays, simply counting widgets doesn't tell us what's going on. Brynjolfsson likens that approach to the reasoning of the drunk in an ancient joke: he looks for his lost keys under a streetlight not because that's where he dropped them but because the light is brighter there.
Robert Gordon, an economics professor at Northwestern University, argues that computer-driven productivity gains may be undermeasured, but so what? They are no more wrongly estimated, he says, than in precomputer decades. Besides, computers are not exactly brand-new: commercial mainframes date back more than 40 years; PCs, 15 or 20 years. In all that time, they should have had greater impact, even on measured productivity.
Why haven't they? In Gordon's view, "what keeps computers from being truly productive is that these damn human beings keep getting in the way." Many jobs cannot be fully automated: "Planes will always need two pilots and trucks a driver." Computers cannot replace beauticians, gardeners or restaurant chefs. Moreover, there is the law of diminishing returns: the latest PCs do not represent as great an advance over earlier computers as the first did over typewriters, or as typewriters did over writing by hand. Says Gordon: "I cannot type or think any faster than I did with my first personal computer in 1983, although it contained only one-fiftieth of the memory and operated at one-thirtieth the speed of my present model."
Experts also point out that much investment in computers does not even aim at increasing productivity. It is intended to expand or protect market share. Computers can be misused too, especially by businessmen who buy them to keep up with the corporate Joneses. "We have clients who buy the most expensive laptops and systems and then just run an ordinary word-processing program," says Alex Reppen, a New York computer consultant, who notes, "People try to automate things that really have no business being automated."
Ben Jarvis, a California businessman, gives an example. A clerk at a tire shop once told him he would have to pay a bill in cash. "I told him I had an account there," says Jarvis, "but the guy said he had already punched the order up as a cash sale and it would take 10 minutes to redo the whole thing in the computer. Funny thing is, if the system were mechanical, you'd simply cross out cash and write in charge. The guy said, 'Yeah, that's what we used to do.'" Such experiences have persuaded Jarvis to keep computers out of his own company, Accelerated Business Forms, a 14-employee manufacturer of business documents. "If you've got any kind of repetitive operations, a computer is perfect," he says, "but if you handle custom orders, a computer does more harm than good."
Another problem is the endless updating of electronic technology, which makes many a computer supposedly obsolete just when its user has finally learned how to handle it. "I think the technology gets churned too frequently," says Alan Blinder, former vice chairman of the Federal Reserve. "Every time we have to learn a new system, there is a decline in productivity."
Stanford economist Timothy Bresnahan argues that this is more a problem with PCs than with large business computers, where upgrades are handled by professional managers. But changing systems can be a serious problem for medium-size businesses too. Insurance Management Associates, a commercial insurance brokerage firm, has just laid out more than $1 million to install a new computer operating system. In the Denver office alone, says president Robert Cohen, "we had 2,500 hours of training for 70 employees and kept the business running while handling the usual glitches and two-hour breakdowns, as well as the three days the system was down entirely while it was being installed." Cohen nonetheless expects to see an increase in productivity--in about six months.
Some employers fall victim to an electronic version of Parkinson's Law: Work expands to fill the digitized time available for its completion. In many graphics departments, says New York consultant Reppen, employees use their computers to try so many different images that "they spend more and more money to get things done at the last minute." And, of course, employees often spend some of their work time idly surfing the Net or playing games. Clifford Stoll, an astronomer and systems manager in California and author of Silicon Snake Oil, draws this analogy: "Suppose a business said everyone on the sales force was getting a free deck of cards so that when they get bored they can play solitaire. Not going to happen, right? But if you give everyone on the sales force a $2,000 computer, you know they're going to play some solitaire because it's the second or third most common program run."
In short, you need to think before you spend. Before going back to the abacus, however, know that the skeptics are still outnumbered by economists and executives who insist that business on the whole is more productive. Economist Allen Sinai of Primark Decision Economics points out that the U.S. has lately enjoyed "superstrong growth, superlow inflation and a superlow unemployment rate." That could not happen if productivity were really as low as the official figures indicate, he says; the numbers--er, do not compute. So productivity must be increasing faster than calculated, and one likely reason is computerization. Maybe the experts need somebody who can design a computer program to measure that elusive productivity increase once and for all.
--Reported by Dan Cray/Los Angeles, William Dowell/New York, Maureen Harrington/Denver, Marc Hequet/St. Paul and Timothy Roche/Pensacola
With reporting by Dan Cray/Los Angeles, William Dowell/New York, Maureen Harrington/Denver, Marc Hequet/St. Paul and Timothy Roche/Pensacola