ISM Will Check All Data Manually After U.S. Factory Index FlubVictoria Stilwell
As it heads into its second high-profile release of the week, the Institute for Supply Management said it will manually vet all data because it hasn’t identified why its computer software introduced an error into yesterday’s May manufacturing index.
Analysts at the Tempe, Arizona-based purchasing managers association will verify the figures against what the computer spits out, Kristina Cahill, a spokeswoman for ISM said today. A glitch yesterday caused its manufacturing index to drop. That was later corrected to show factories expanded last month at the fastest pace of the year.
ISM will be “manually calculating each index, each factor, each aspect of the report,” and comparing it with the figures the system produces, she said.
“We definitely are very serious about making sure that doesn’t happen again,” said Cahill, who added that the group received “a number of inquiries” about yesterday’s botched manufacturing report. “We’re incredibly concerned about credibility, and that’s why we’re taking these steps.”
The ISM’s service industries gauge is scheduled to be published tomorrow at 10 a.m. New York time. The index probably rose to 55.5 in May from 55.2 the month before, according to the median estimate of economists surveyed by Bloomberg.
The manufacturing index is closely followed by economists and investors in part because of its long history, dating back to 1948, while the services report is of interest because it tracks almost 90 percent of the U.S. economy.
The May factory index was corrected twice yesterday, creating confusion and roiling markets. Economists crunched the numbers separately and noticed the data were wrong after adjusting for seasonal variations. The ISM first reported that its factory index dropped to 53.2 in May from 54.9 the prior month, signaling an unexpected slower pace of expansion that pushed stocks lower.
Later, ISM verbally corrected the figure through the survey’s chairman Bradley Holcomb, who said its gauge rose to 56, sending equities up. The group then again corrected the data through a second release, saying the index was 55.4. An ISM index reading of 50 is the dividing line between growth and contraction.
The software problem caused the initial error, while the second mistake came after the sub-indexes used to derive the headline number -- orders, production, employment, supplier deliveries and inventories -- were inaccurately averaged in haste, Holcomb said.
Kenneth Kim, an economist at Stone & McCarthy Research Associates Inc. in Princeton, New Jersey, was the first to question the accuracy of the data. On a hunch that there was a discrepancy, he applied the April seasonal-adjustment factors to the raw data for May and came up with the number initially released by ISM.
In the wake of the discovery, “people I haven’t heard from for 10 or 15 years have messaged me out of the blue, saying, ‘great job,’” Kim said today in an interview.
The ISM wasn’t one of them.
“I’m not expecting a call from them, which is fine,” he said. “I’ve caused them enough consternation.”
As word of Kim’s finding spread yesterday, Stephen Stanley, chief economist at Pierpont Securities LLC in Stamford, Connecticut, and others also ran to double check the data. “It was clear that the numbers were wrong,” Stanley said.
The mistakes will probably cause some economists to verify the ISM’s figures in the months ahead, he said. Still, yesterday’s error probably won’t do permanent damage to the manufacturing index’s credibility given its long history and the information it provides on economic activity.
“There’s nothing in what happened yesterday that would de-legitimize that data,” Stanley said. Economists “will respond as if the same exact thing is going to happen for a while, until it doesn’t.”
Gennadiy Goldberg, a U.S. strategist at TD Securities USA LLC in New York, said he’ll be creating a manual check, adjusting his spreadsheets to calculate some of the data so they can be spot-checked. The firm’s sales staff also asked about whether adjustments on other data could be off after the ISM corrections.
“Most of it was jokingly, but it shows you that the question was out there,” Goldberg said.