Gauging Output: It's Not Just Counting Widgets Anymore

It's easy to measure the output of a steel mill. Steel does not change much from year to year, so gauging production means toting up how much was made of the different grades. But what about a computer factory, where every year the power of the products improve and prices drop? Or a hospital, where no one is even sure what output means?

The U.S. is using a Model A system to calculate the performance of an Intel 486 economy. The government uses a simple definition of productivity: inflation-adjusted output per hour of work. The problem is the way output is gauged--essentially, the value of goods sold less the cost of goods purchased. When the statistical system was devised in the 1930s, goods-producing industries accounted for 57% of output and 40% of the nonfarm jobs. But by 1992, goods accounted for just 40% of output and 22% of employment. In some fields, such as financial services, government statisticians simply can't define just what is being sold.

None of this would matter much if investors, business planners, and policymakers didn't depend on productivity numbers. As the Clinton Administration dabbles in industrial policy, it will inevitably look to productivity numbers to pick winners and losers.

Three separate problems distort the data: Government lumps industries in groups that make no economic sense, so the same amount of information is provided for buggy whips (SIC 3199) as for microprocessors (SIC 3674). The Bureau of Labor Statistics (BLS) gives data for 17 categories of apparel manufacturing but only one for computers. Also, the measurement of output, especially in services, often amounts to little more than an informed guess. And statistical agencies' budgets have been squeezed so hard that they have trouble producing data on schedule.

GO FIGURE. A major revision of the data classifications is under way, but it will take years to implement. Even then, fundamental conceptual problems will need to be tackled. For instance, in an information economy, intellectual property is a key output. You'd never know it by how we evaluate software: Exports are valued at the cost of the disks and the manuals, so a $500 program might be entered at just a few dollars. And the understanding of service output is so poor that despite the financial-services boom of the early and mid-1980s, the government data show the finance sector grew more slowly than the overall economy.

"If we're going to measure output, we first have to figure out what it means," says former BLS head Janet L. Norwood. One of the biggest problems, she notes, is adjusting data for qualitative changes. In the late 1980s, a major government effort resulted in a new way to measure the value of computers that accounts for their increasing power. That helped, but until other improvements are devised, investors and government planners may be flying blind into the 21st century.

Before it's here, it's on the Bloomberg Terminal.