The reported figure, which estimated that jobless claims had dropped to 292,000, about 31,000 fewer than the week before, seemingly suggested that the economy was finally entering a self-sustaining recovery on the back of a healing job market.
The number, however, is unreliable, the government said, skewed by upgrades on two state computer systems that caused those states to underreport claims. The total number of initial jobless claims is almost certainly higher than reported, though nobody knows the scope of the mismeasurement at this point.
The data malfunction has called into question the accuracy of a major leading indicator, one scrutinized by investors, economists and policy makers alike. It also shined a light on the imperfect and often outdated systems that states and the federal government use to provide benefits to workers and cull data on the labor market and the broader economy — a situation that some experts warn might become even worse because of the $1 trillion in budget cuts spread over 10 years known as sequestration.
The Labor Department would not confirm which two states had issues or guess as to the scope of the mismeasurement. But Nevada confirmed that it had not reported complete claims data to the federal government because of a computer upgrade.
“When we get data, we have an obligation to put it out there,” said Jason Kuruvilla of the Labor Department, explaining why the department did not wait rather than release incomplete data. He emphasized that the department did not recommend reading too much into any one week’s figure, at any rate.
“One week is not a trend,” he said. Mr. Kuruvilla said the two states that had misreported data would become more apparent after new state-level jobless claims data were released next week.
But some outside experts had scathing words for the Labor Department, and others described the data problem as not a one-time issue but a symptom of a chronic lack of money for one of the most critical functions of the government.
Rick McHugh of the National Employment Law Project, a nonprofit group in Washington, noted that unemployment insurance programs are partly financed by the federal government but administered by the states.
“This is a symptom of a longstanding problem with the unemployment insurance programs,” Mr. McHugh said. “These are state agencies, but they’re funded by federal dollars. And governors don’t see them as their agencies. They’re orphans.”
Many state unemployment agencies have struggled to keep up with the demands of their rapidly expanding and changing rolls through the recession and the tepid recovery.
Sequestration has only made matters worse, as benefit cuts this year for the long-term unemployed required state to retool their systems.
Nevada has struggled to carry out the across-the-board 5 percent cut to a federally financed program for the long-term jobless. States that put the cut in place partway through the fiscal year, which ends on Sept. 30, trimmed many recipients’ benefit checks about 11 percent. Nevada, which cut its checks only recently, did so by nearly 60 percent.
Problems with the computer system caused the mixup in reporting, said Mae Worthey, a public information officer with the Nevada government. “It wasn’t a glitch,” she said. “It was just implementing a new system.”
Economists said the reporting problems made it harder for experts and officials to keep track of the economy. That is especially critical now, as the Federal Reserve prepares for a crucial policy meeting next week.
“Officials should be completely transparent when they know that bureaucratic issues are distorting their data,” wrote Justin Wolfers of the Brookings Institution. “Such transparency would allow economists to provide simple statistical fixes, making the data more reliable and useful.”