The new calculations by the Army, which had about 500,000 active-duty troops at the end of 2006, significantly alter the annual desertion totals since the 2000 fiscal year.

In 2005, for example, the Army now says 2,543 soldiers deserted, not the 2,011 it had reported. For some earlier years, the desertion numbers were revised downward.

National Public Radio first reported on Tuesday that the Army had been inaccurately reporting desertion figures.

A soldier is considered a deserter if he leaves his post without permission, quits his unit or fails to report for duty with the intent of staying away permanently. Soldiers who are absent without leave — or AWOL, a designation that assumes a soldier still intends to return to duty — are automatically classified as deserters and are dropped from a unit’s rolls if they remain away for more than 30 days.

Some Army officers link the recent uptick in annual desertion rates to the toll of wartime deployments and point to the increasing percentage of troops who are on their second or third tours in Iraq or Afghanistan.

But an Army spokeswoman, Maj. Anne Edgecomb, gave different reasons. Most soldiers desert because of personal, family or financial problems, Major Edgecomb said, adding, "We don’t have any facts to indicate that soldiers who desert now are doing so for reasons different from why soldiers deserted in the past."

Lt. Col. Bryan C. Hilferty, an Army spokesman, said the desertion data errors were caused by confusion among employees who tally them. "They were counting things wrong, and doing it inconsistently," Colonel Hilferty said in an interview.

He added, "We are looking at the rise in desertions, but the numbers remain below prewar levels, and retention remains high. So the force is healthy."

The failure to count deserters accurately is inexcusable, said Derek B. Stewart, director for Defense Department personnel issues for the Government Accountability Office.

"It is just unbelievable to the G.A.O. to hear that the Army does not know what that number is," Mr. Stewart said in an interview Thursday.

Noting that the problem with the desertion numbers arises when the service cannot find enough recruits to fill certain crucial specialties like medical experts and bomb defusers Mr. Stewart said, "In the context of their current recruiting problems for certain occupations, these desertion numbers are huge."

The new figures also show a faster acceleration in the rate of desertions over the previous two fiscal years than announced. In 2006, for instance, desertions rose by 27 percent, not 17 percent, as the Army had previously reported, a spokesman said.

The revised figures show 2,543 desertions in the fiscal year 2005, an 8 percent increase from the 2,357 the year before. Previously, the service said 2005 desertions dropped by 17 percent, to 2,011 from 2,432.

But from the fiscal year 2000 through 2003, there were hundreds fewer desertions than the Army had previously reported. The Army’s revised data, while reflecting significant errors in year-to-year desertions, showed a total of 22,468 desertions since the fiscal year 2000, nearly the same as the old count of 22,586.

Over all, desertions, a chronic problem in the Army but hardly pervasive, now account for less than 1 percent of active-duty soldiers. The current annual rates pale in comparison with the 33,094 soldiers — 3.41 percent of the total force — who deserted the Army in 1971, during the Vietnam War.

The Army’s data does not reflect deserters from the 63,000 currently activated National Guard and Reserve soldiers, and Colonel Hilferty said that data was not available yesterday. But he said few soldiers from those units deserted.

In an e-mail statement yesterday, Colonel Hilferty also said that the record keeping was damaged in the Sept. 11 attack on the Pentagon, which destroyed personnel records.

"Unfortunately, for the past several years," he said, "our methodology for tracking deserters at the macro level has been flawed."