WASHINGTON - The U.S. government is having a tough time guesstimating how many small businesses failed in this recession, casting doubt on the reliability of vital data on employment and economic growth.

The formula the U.S. Labor Department designed to help it deliver timely, thorough monthly employment reports broke down in the heat of the financial crisis, miscounting the number of jobs by an estimated 824,000 in the year through March.

The most likely culprit is the so-called birth-death model, which the Labor Department uses to estimate how many companies were created or destroyed.

That model appears to have misjudged how many companies went out of business during the recession, meaning the labor market was even weaker than initially thought when President Barack Obama took office in January. More recent figures may still be underestimating job losses now, but it will be many months before the Labor Department is certain.

One characteristic of this recession is that it has hit small businesses especially hard, driving down demand and choking off vital sources of credit at the same time.

Obama's administration is scrambling to try to prop up small business -- it hosted a summit on that topic on Wednesday -- because those companies are essential to bringing the jobless rate down from its current 10.2 percent, having accounted for the lion's share of new job growth in recent years.

Government data has difficulty gauging the health of smaller firms because there are simply too many of them, leaving officials to rely on surveys and models that are hit and miss.

Jan Hatzius, an economist at Goldman Sachs in New York, thinks that is distorting not only the employment data, but also figures for retail sales, durable goods and even the biggest economic indicator of all -- gross domestic product.

Our conclusion is that if small firms aren't captured well in the advance GDP data, the economy may be growing less quickly than suggested by the recent official data, he wrote in a recent note to clients.

Recent data has suggested the economy grew at a somewhat slower pace in the third quarter than the 3.5 percent rate the government initially estimated, and many economists think the figure will be revised down later this month.

Because of problems tracking small firms, Hatzius thinks the figure will eventually be revised down much more sharply, perhaps by as much 2 percentage points. But this could take years.

Government statisticians always are in a tug-of-war between trying to provide economic data quickly enough to be useful for policy makers -- and investors -- and ensuring its accuracy.

The task became even trickier during this downturn because the economy suffered such a swift and severe slump that the data struggled to keep up.

BIRTHS, DEATHS AND BREAKDOWNS

Olsson's Books & Records is one casualty of the recession. It filed for Chapter 7 bankruptcy liquidation last year, closing its five stores in the Washington area. At its height, the chain boasted nine bookstores, with some 200 employees.

Terence McCann, who was a business manager at Olsson's for 20 years, said it just got harder and harder to compete every year.

In filing for liquidation, the company blamed low cash reserves, and an inability to renegotiate current leases, along with a continuing weak retail economy and plummeting music sales -- common ailments among small businesses.

Indeed, 43,546 businesses filed for bankruptcy in 2008, the highest tally since 1998, and the pace has picked up this year, according to data from the American Bankruptcy Institute.

In the second quarter of 2009, the most recent data available, 16,014 businesses filed for bankruptcy, up from 14,319 in the previous three-month period and the highest mark in 16 years.

The Labor Department simply can't catch all those failures fast enough to compile its monthly employment reports, which are normally released on the first Friday after the end of the month. So it must make an educated guess.

Each month, the department surveys about 160,000 firms to get a sense of how many jobs were added or cut. It also uses the birth-death model to try to estimate out how many companies opened or closed.

Once a year, the department looks at unemployment insurance tax records to get a more accurate picture of how many people were employed, and matches that up with its own data. Each February, it tries to reconcile these differences by releasing a benchmark revision.

Normally, the discrepancy is modest. This coming February, it is likely to be about 824,000, according to the Labor Department's preliminary estimate last month. That would mean instead of about 7.2 million jobs lost since the start of the recession in December 2007, there were more like 8 million.

Preliminary research indicated that a big portion of that was a result of a breakdown in the birth-death model, said Chris Manning, the department's benchmark branch chief.

Until this recession, the birth-death model had a track record of performing well regardless of whether the economy was growing or shrinking.

Manning said his department was still trying to figure out what went awry this time. One possibility is that the model was not sensitive enough to the credit crunch, which choked off borrowing and pushed many companies into bankruptcy.

We're researching ways to better understand the limitations of the model, in particular when it comes to responding to economic shocks, he said.

One change under consideration is plugging data into the model more often, perhaps quarterly instead of once a year, so that discrepancies would be apparent sooner.

If the department decides on any changes in the next couple of months, it will spell them out in February when it releases the annual benchmark revisions, Manning said.

(Editing by Chizu Nomiyama)