Skip to main content

Editor's note: This is part of a three-day series on the changing economy. To see more, click here.

Nov. 5, 1998: The

Bureau of Labor Statistics

reports a 52,000-job drop in manufacturing payrolls, marking the seventh decline in eight months. The expected economic slowdown seems to be materializing. Economists predict the

Federal Reserve

will cut rates several more times. The service sector adds 154,000 jobs.

February 1999: Heads spinning, economists find they have been pummeled into submission by data showing continued economic strength. The decline in manufacturing employment has only dented overall economic growth.

What happened? Forty years ago, manufacturing was paramount to the success of the economy. In 1999, it constitutes just 18% of nominal

gross domestic product

. But economic measurement tools haven't kept pace with the economy's growing service orientation.

Forecasters still lean heavily on manufacturing data to determine the direction of the economy. But if manufacturing led the economy, this year the U.S. would have slipped into a recession or come pretty close. With a dearth of service-sector data, the source of the lion's share of GDP is sorely underrepresented in monthly economic surveys.

"The relationship started to fall apart in the last year or two," says Mike Niemira, vice president at

Bank of Tokyo-Mitsubishi

. "Until recently, even with the long-term trend to smaller total output,

manufacturing output maintained a good relationship with the national economy."

Where Manufacturing Went Wrong

Manufacturing output as a share of nominal GDP has declined steadily over 40 years. Manufacturing payrolls as a share of total employment have remained relatively stable while the rest of the economy has ballooned. In 1960, about 64 million people were employed in the U.S., with 20 million in manufacturing jobs, according to the

Conference Board

. Now about 18.5 million people work in manufacturing out of a workforce of 133 million.

Economists say that even as the sector's importance has waned, manufacturing-sector data remained a useful proxy for the entire economy. When cyclical forces forced the manufacturing economy to contract or expand, the service sector would follow.

But recently, the link between the service and manufacturing economies seems to have weakened, owing to several factors. As manufacturing's share of this nation's output shrinks and economic globalization takes hold, consumers depend less on domestic manufacturers. And the runaway popularity of personal investing has recently raised financial services' economic profile. Economists also say government adjustments to economic data fail to accurately reflect business activity within specific industries.

Add these factors together, and the relationship between various manufacturing indicators and the state of the economy erodes significantly.

"I find

the purchasing managers index very useful, and I think it is one of the best indicators for the manufacturing sector," says Paul L. Kasriel, chief economist at

Northern Trust

. "But I think this year more than in the past, there's been a real disconnect between manufacturing and the rest of the economy."

According to Niemira, the move toward quality-adjusted data within the government's GDP,

producer price index


consumer price index

measures has served to disconnect the data from the true state of particular industries. This is most prevalent in the computer industry, where adjustments are made for the value of purchasing a newer, more powerful product, even though overall sales may be softer.

"A company complains that at different times business is soft, but with what you see from the government, I haven't seen soft data in computer output in years," says Niemira. "There is a disconnect in that sense."

Finding a Middle Path

Then there's the one about a drunken man looking for his keys, which could represent some economists. The tipsy fellow is stumbling around near a streetlight. Another man approaches and asks, "What are you doing?" "Looking for my keys," the drunk says. "Where did you drop them?" the man asks. "Over there," the drunk answers, pointing to a dark alley. "Why don't you look over there?" the man asks. "Because," the drunk says, "there's more light over here."

Despite consumer spending and housing's strong performance last year vis-a-vis manufacturing, many economists insisted consumer spending would slow this year, simply because it had to.

But data on industries driven by consumer spending might have struck a balance between manufacturing and the indicator most commonly linked to robust consumer spending: the stock market. The financial services, utilities, entertainment, transportation and health-care sectors of the economy are relatively underutilized in terms of timely information, with the exception of the employment report and a few other indicators.

"A lot of the strength in 1998 was related to the financial services sector; included in that sector could be commissions for real estate transactions," says Kasriel. "We need an index of financial activity or something to try to measure that."

Other data that could fill some of the gaps left by the narrowly focused manufacturing series include monthly indices that complement the current retail sales, inventories, wholesale trade and factory orders data. The problem, some say, is that production in a service industry is less tangible, harder to quantify. Retail sales data devoted to services could simply examine a common figure, such as hours booked, tickets punched or total commissions, depending on the industry.

"Go from the sales side," Niemira, a member of the

National Association of Purchasing Management's

advisory board, contends. "Money is the common denominator."

Early returns on the PMI survey "have not been real good," says Michael Moran, chief economist at

Daiwa Securities

. "I would emphasize that it is early and not even seasonally adjusted because they don't have the data. In the fourth quarter we saw remarkable growth, and it didn't show much movement. That made you leery about using that measure."

The nascent survey has been tabulating data since the beginning of 1997, so it hasn't made it through even one business cycle. In addition, this measure attempts to determine the business conditions in manufacturinglike categories such as imports, inventories and supplier deliveries in industries where those factors may not be important -- somewhat explaining its erratic performance.

"What we're concerned with was, this is so big, is this too broad?" asks Ralph Kauffman, chair of the NAPM's survey committee. "We're now looking at a subindex that's just services, which would exclude mining and agriculture."

Shades of Gray

The service economy has become more prominent in ways more easily quantifiable: Outsourcing of former in-house employees and relying on temporary workers have increased service-sector employment in this economy. But also blurring the lines between manufacturing and service producers is this: Some manufacturers already do service-related work, such as retail trade and shipping, while some service providers buy products from manufacturers or purchase enough capital equipment that their activities are similar to those of manufacturers.

"As companies have diversified, they will put a company in a certain sector based on their predominant activity," says Josh Feinman, global markets economist at

BT Alex. Brown

. "Somehow we can get into the data that way."

Kauffman believes that eventually the service sector will be represented with more specific surveys as the economy continues to warrant it. Its growing influence is evident in the NAPM's membership itself. Twenty years ago, he says, NAPM members were mostly purchasing managers in manufacturing industries. Now the ratio is 50-50.

"Eventually we'll probably see more information," Kauffman says. "To some extent, it's a little difficult. One of the questions is, 'What is the capacity to produce services?' It's a little more difficult for service industries to answer that. At this point, we don't have enough data to assess what this does mean -- we need a few more years of data."

See Results

See Results