WHY THE ECONOMIC DATA MISLEAD US Outmoded methods of reporting and analysis produce many shaky statistics. It's not just what we don't know that hurts; it's what we think we know that really isn't so.
By Louis S. Richman REPORTER ASSOCIATE John Labate

(FORTUNE Magazine) – IT COULD BE the biggest infrastructure problem of all. Once the envy of the world, America's system for gathering and interpreting economic statistics has fallen into disrepair. The resulting mismeasurements make it harder than ever to understand what's happening to the economy, both in the short run and over the long haul. Over the past decade, the statistics may have understated annual economic growth and productivity gains by a percentage point or more, argues Michael Darby, a former Treasury official who teaches at UCLA's graduate school of business. Annual consumer price inflation may have been overstated by as much as 1.8%. Unemployment? It's impossible to say what the rate really is. Such knowledge gaps create endless opportunities for bad policy decisions. If growth is faster than generally assumed, should the Clinton Administration back away from its plans to stimulate the economy? Then again, if inflation is only 1% after all, America has probably reached Federal Reserve Chairman Alan Greenspan's goal of effective zero inflation, leaving room for fiscal and monetary stimulus that could spur small-business growth. One can only guess. How did we reach this sorry state? Lay part of the blame on penny-pinching. Real government spending on data gathering and analysis, some $2 billion in 1992, has barely grown since the late 1970s, though the economy today is nearly 30% larger and infinitely more complex. Washington's otherwise laudable efforts to reduce the burden of paperwork on business has made it more difficult to gather data detailed enough to ensure better accuracy. And shrunken companies can't spare the people to feed data to government statisticians. But there's a more important problem: The statisticians are analyzing yesterday's news. The loosely coordinated efforts of the half-dozen or so agencies that crank out most of the data haven't kept pace with enormous technological and structural changes that have transformed the economy. The federal reporting system was developed and refined over decades to detect the cyclical swings of a manufacturing-based economy largely isolated from the rest of the world. Now companies routinely turn out new goods tailored to specific market niches, greatly expanding consumer choice and shortening product life cycles. Much activity has shifted to harder-to-measure small businesses. And globalization has blurred the boundaries between the American economy and the rest of the world. Says Harvard University economist Zvi Griliches: ''Our data are weakest in precisely those areas where economic change has been most dynamic, such as technological innovation, the service sector, and trade.'' The divergence between the two economies -- the one the government measures and the rapidly evolving one economists and business people are struggling to understand -- is widening. Whole industries from biotechnology to telemarketing hatched and flourished in the 1980s. But you wouldn't know it from the government's Standard Industrial Classification (SIC) system, the business categorization scheme that planners, marketers, and economists use to track their industries' performance. Though the Office of Management and Budget (OMB) revised the 1,005 classifications in 1987 for the first time in a decade, they remain hopelessly out of date. OMB finally created a separate category for semiconductors, but it lumps advanced microprocessors together with photovoltaic cells and simple memory chips. Electronic vacuum tubes, meanwhile, get a category of their own. Anyone interested in following the plodding progress of the ''nonrubber ^ footwear'' industry can find details on no fewer than four different varieties. But telemarketing is grouped together with 134 other business services such as wig styling and window trimming. OMB has started a project to completely overhaul the code, but results aren't likely until 1997 at the earliest. Meantime, the putative data the government collects for industry groups are almost certainly misleading. MISMEASUREMENT of inflation is the biggest source of economic distortion; the errors affect productivity calculations, long-term wage data, and the GDP growth rate itself. If measurements of price changes by the Department of Labor's Bureau of Labor Statistics (BLS) overstate the inflation rate, real growth of output will be understated except in the few cases, such as autos, where the agency actually analyzes in detail the units produced. Result: Productivity and incomes will seem to stagnate. This overstatement of inflation is precisely what appears to be happening, particularly in the measurement of the consumer prices that make up about two-thirds of the GDP deflator. (Government purchases and payrolls, capital equipment, and construction account for the balance.) While few experts think the distortions are as large as UCLA's Darby estimates, he's not totally alone. A Federal Reserve Board working paper suggests that in recent years inflation may have been exaggerated by about one percentage point. And many other economists, such as Harvard's Griliches and Robert Gordon of Northwestern, have no doubt that the official data misrepresent the economy's true strength to some degree. Says Gordon: ''We have been preoccupied with problems such as slow productivity growth, the competitiveness of U.S. industry, and stagnating real wages. But the trends in income and productivity are not nearly as bad as the numbers make them look.'' EACH MONTH the BLS dispatches 450 inspectors, most of whom work part time, throughout the land to gather standardized price data on everything Americans buy or rent, from apples and apparel to apartments, appliances, and appendectomies. But today's consumers change their buying patterns too rapidly for the BLS shoppers to keep up. Because the bureau adjusts the products and retail outlets it surveys only gradually over a five-year period, it is slow to detect how swiftly those preferences shift. Major case in point: BLS has largely missed the huge swing from full-price food retailers to grocery wholesalers and discount outlet clubs such as Cub Foods, Sam's Club, and Costco. Looking at food sales alone, for example, the bargain outlets increased their share of the total from 3.2% in 1980 to 16.2% in 1992, according to Progressive Grocer, a food trade publication. As measured by the BLS, food prices rose at an annual 4.2% during the Eighties. But the food inflation index doesn't pick up the savings consumers are enjoying. The result, says Marshall Reinsdorf, a BLS economist, is that the actual inflation rate was 1.7 percentage points lower. More recently the bureau has overlooked the trend by consumers to substitute far less costly private-label store brands for virtually identical big-name-brand products; the generics have grown from 12.5% to 14% of total supermarket dollar sales volume since 1989. Government economists are also too slow to recognize how new technologies and better product quality pump up consumers' effective buying power. Penicillin was left out until 1951, though it had been in common use for several years, and its price had dropped by 99%. The common pocket calculator, introduced in 1970, didn't show up in BLS's market basket until eight years later, when it had become 90% cheaper than the earliest models. Says Harvard's Griliches: ''Whole generations of new goods come and go before the data collectors can measure their impact on prices, productivity, and growth.'' Thus BLS misses years of price deflation even after fast-growing new products have become pervasive throughout the economy. The distortions compounded manyfold in the 1980s with the information and communications technologies that revolutionized so many factories and offices. Says Henry Kelly, a senior associate at the Office of Technology Assessment in Washington: ''Information technology is now 40% of all capital goods investment. But if the price deflators don't reflect its decreasing true cost as the equipment becomes more powerful, the real value of that investment is badly understated.'' Economists at the Commerce Department's Bureau of Economic Analysis (BEA) are slowly beginning to adjust. In 1985 the bureau, with help from IBM, revised its methods for calculating the computer price indexes by taking into account increases in computing speed. As a result, computer prices adjusted for quality changes are plunging at a rate of 12% a year. But in most other areas, the task of adjusting price indexes is monumentally incomplete. The producer price index, which contains thousands of values for such pedestrian products as bolts, flanges, valves, and cleats, still has no accurate measure for semiconductors or for communications equipment, a $58- billion-a-year industry and the biggest category of producer durables. Some economists, like Martin Baily of the University of Maryland and Northwestern's Robert Gordon, finger mismeasurement as one of the main suspects in the great productivity mystery. The rapidly expanding service sector has created almost all of the economy's new jobs. Measuring productivity in services is treacherously difficult -- particularly so for such large sectors as financial services, whose output is hard to define. The government's methods are painfully primitive -- and in some cases useless. Commerce Department statisticians estimate output in financial service businesses, for example, as the value of the labor employed to create it. But labor productivity is by definition total output divided by total labor inputs. So the crude official calculations make it mathematically impossible for productivity to grow at all. When Baily and Gordon took a close look at the securities industry, they found a lot of productivity growth. In 1973 the average volume of shares traded by the industry's 209,000 employees totaled 5.7 million daily. By 1987 trading volume had grown to an average 63.8 million shares a day -- a better than 11-fold increase though industry employment had only doubled. ''Instead of zero productivity growth,'' they conclude, ''growth was rapid and accelerating in the 1980s.'' Baily and Gordon also found serious measurement problems for the supposedly unproductive real estate and insurance industries. Such mismeasurement helps explain the puzzle of why heavy investments in computers have not visibly lifted productivity. Over the past decade, for example, grocery wholesalers and retailers purchased more information technology -- computers, scanners, and the like -- than all aircraft manufacturers, with virtually no measured productivity payoff. But the official measurements don't account for the vast proliferation in services that computer power made possible. Example: Electronic tracking of inventories, which allows retailers to offer consumers a far richer choice of products while shrinking their inventory/sales ratios. Indeed, a new study by the McKinsey Global Institute, the research unit of the McKinsey & Co. management consulting firm, found that productivity in major service industries is far higher in the U.S. than it is in Europe and Japan. In banking, for example, it's nearly 50% higher than in Europe, largely because American banks were quicker during the 1980s to apply computer technology to their back-office clearing operations, install automated teller machines, and adopt new, more efficient organizational structures. PERHAPS the biggest lacuna is the poor quality of data on America's trade flows and international financial transactions. Until the early 1970s, imports and exports were so small a part of total activity that economists could almost regard the current account balance as a rounding error. No more. Over the past two decades the value of U.S. imports and exports has more than doubled as a share of GDP, to 11% last year. The very factors that expanded trade are making the job of measuring it far harder. Falling tariffs and other barriers, cross-border industrial alliances, and cheap, fast communications foil the auditing systems that were set up in a time when trade was much more tightly controlled. Dun & Bradstreet chief economist Joseph Duncan headed the statistics branch at the Office of Management and Budget during the mid-1970s. Says he: ''Our methods of gathering and reporting trade data were designed for a mercantilist economy.'' Here again the result is almost certainly an understatement of America's strength. Because most imported goods arrive in U.S. ports accompanied by detailed shipping and customs documents, the Commerce Department captures their value reasonably well. But a commission of the National Academy of Sciences finds that exporters have strong incentives to lowball the value of the products they send abroad, minimizing the tax bite on reported profits or reducing insurance costs. Software makers, for example, quite logically insure for the value of the plastic in their diskettes rather than for the knowledge inscribed on them. The overworked International Trade Division at Commerce lacks the manpower to perform more than a cursory audit on exports. The most underreported part of trade is fast-growing U.S. service exports. ''Despite significant recent improvements by the Commerce Department, the data are still poor,'' says University of Wisconsin economist Robert Baldwin, who last year headed the National Academy of Sciences panel that examined the quality of U.S. international data. What value do American exporters place on their services? How to get service companies, most of them small, to report their foreign sales? How many globe-hopping consultants, lawyers, or architects leave their fees in overseas banks? Government statisticians have only the dimmest idea. Official estimates for 1991, the last full year available, showed the U.S. as having a $45 billion surplus in service exports, but the true number is undoubtedly larger -- perhaps far larger. SO IF INFLATION is cooler, productivity growing more briskly, and foreign trade healthier than the official measurements show, the economy must be in much better shape than we think? Not necessarily. Some mismeasurements have overstated growth, partly offsetting the damping effects of the errors based on inflation. While it's often noted that manufacturing continues to account for 22% of real GDP as it did in the 1950s, today's manufacturing needs a lot more service support. The Office of Technology Assessment figures that the share of activities such as advertising, legal services, insurance, security, and communications that go into final output has nearly doubled. These intermediate services -- OTA calls them ''transactional businesses'' -- are now as large a proportion of GDP as goods production. Much of them amount to input that eludes measurement. University of Nebraska economists Scott M. Fuess Jr. and Hendrik van den Berg have calculated the economic impact of this transactional activity by subtracting the measured growth in output that can be attributed to it. The fast growth of the transactions, they say, has sapped productivity. Thus, measured economic growth is less robust that it appears. Their conclusion: Real per capita GDP grew a quarter percentage point a year less between 1983 and 1989 than the average 2.9% computed by the Commerce Department. The growth figure may be further inflated because the government is measuring output from the new economy that used to go unrecorded. Fuess and van den Berg estimated that many married women who joined the labor force in the 1980s turned to the market for services such as child care and household maintenance, formerly produced as unmeasured housework (although as Zoe Baird showed us, not all of this output gets into the statistics). That added nearly another three-tenths of a percentage point to measured per capita economic growth. SINCE JOBS are Topic A today, the unemployment rate is among the most critical of statistics. The world's largest ongoing monthly data-gathering exercises are the BLS's surveys of businesses and households on the number of jobs and the unemployment rate. Yet the accuracy of the official numbers is problematic -- and especially so during hard times. The employment picture has been particularly murky. Last November the survey of businesses reported that total employment grew by 417,000 in the first ten months of 1992. But separate data gathered by local BLS offices from state departments of employment showed the job base shrinking by 149,000. Since then, the state offices have been registering higher growth than the national figures. Both estimates are suspect. The nationwide figures are adjusted to include finger-in-the-wind estimates of the net number of jobs created by new business formations minus those lost to closings. The state data, meanwhile, have been compromised by budget cuts. Only the year-end reconciliation of federal tax withholding forms filed by all employers will determine which estimates were closer to the mark. They will be available this spring. Attempts by BLS to measure unemployment by monthly surveys of some 60,000 representative households have been bedeviled by demographic and structural shifts. Among other questions the survey cannot answer: How many of the ''unemployed'' work in the underground economy? What proportion of the 84 million Americans putting in 35 or more hours a week and counted as full-time workers actually hold two or more part-time jobs? Might not many laid-off managers adopt the less ego-bruising label ''self-employed'' to mask the harsh reality of joblessness? BLS has refined its survey methods for the first time since 1967 and will begin probing these weak spots next year. Until it does, the true unemployment rate could be higher -- or lower -- than the official data report. What to do about the crumbling statistical infrastructure? More money is needed, but any change has to begin with some basic reengineering -- starting with the abolition of unneeded work. In 1991 the Department of Agriculture spent as much to gather numbers on farm production, which accounts for just 3% of GDP, as the BLS, the BEA, and the Treasury Department combined spent on data gathering. The statistical agencies could also stretch the resources they do have by making better use of technology. Instead of sending its agents scurrying out to price oranges at fruit stands, for example, BLS could get timely readings of changes in prices and preferences by tapping into the electronic scanning data now widely used by retailers and their suppliers. AS ALWAYS, change will have to come from the top. Michael Boskin, chairman of George Bush's Council of Economic Advisers, laid out a promising proposal. The principal goals: Harmonize U.S. accounts with those of other countries, introduce into the price indexes more quality adjustments for new products and technologies, improve the measurement of service sector output, and refine the Labor Department's surveys of changes in employment. The cost: a modest $200 million or so extra by 1997. Instead, Congress balked -- for example, it cut the $30 million increase in outlays requested for 1992 to $18 million. Action on the Boskin initiative would help. So would continuous research into new measurement methods and closer coordination among the agencies. The U.S. could learn much from the Canadians, whose centralized data office, Statistics Canada, is the industrial world's acknowledged champ at economic measurement. Americans need to know what's happening in their real economy -- not in the statistical artifact that's being monitored today.