The U.S. Bureau of Labor Statistics (BLS) recently snuck out a telling confession beneath everyone's radar: Its flagship payroll survey is likely undercounting hundreds of thousands of jobs. Most economic observers were too busy fretting over the lackluster gain of 32,000 payroll jobs in July to take notice of the other positive indicators, let alone the quiet little study that acknowledges payrolls have a problem.

The study describes how job-changing can inflate the payroll survey's numbers artificially. When worker turnover is brisk, as in the late 1990s, millions of workers are counted twice when they switch jobs. About 3.9 million people changed employers during a typical month during the 1990s, but only 3.1 million do so now.

Why is job-changing dropping? Maybe stability is preferred since 9/11. Perhaps lower turnover is a reflection of the aging workforce and low participation rate of current teens. Or maybe more workers are becoming self-employed. The reason doesn't matter, but the effect on payrolls does.

For months, the debate has been raging over how to measure jobs. Being that we're in a presidential election year, the issue has been magnified. But why should the average person care? Because only an accurate reading can gauge the country's true economic health and affect everything from interest rates to consumer confidence.

A survey we can't count on

The payroll survey has long been seen as the best measure because its larger sample dwarfs those of other methods. It surveys 160,000 businesses and government agencies. But now, BLS is admitting that the sample is muddy.

There's an alternative. The U.S. Labor Department's household survey, which is used to calculate the unemployment rate, is not subject to the job-changing effect. It surveys 60,000 households and counts self-employed consultants, real estate agents, farmers and other non-traditional workers who aren't on old-style payrolls.

Yet because the household survey shows 2 million more working Americans under President Bush than ever before, it has been attacked for partisan reasons. Those who favoring the payroll survey are quick to quote Federal Reserve Chairman Alan Greenspan's congressional testimony in February favoring conventional numbers.

In March, we at the Heritage Foundation released a critique of the payroll survey, putting into numbers what many had long suspected: It doesn't accurately reflect today's economy.

Greenspan gets it

In his July 20 testimony to Congress, Greenspan cited measures from the payroll and household surveys. Then the Federal Reserve, led by Greenspan, voted unanimously to raise interest rates. It said the economy is "poised to resume a stronger pace of expansion" and noted that labor-market conditions continue to improve. It's no secret which survey would lead to that conclusion.

The Fed's actions helped everyone, including Wall Street, remember the good news. Claims for unemployment benefits, for example, are 10% below their 30-year average, while the unemployment rate has fallen to its lowest level since 2001. Best of all, the household survey showed a gain of more than half a million jobs in July alone.

Everything adds up - except the payroll survey.

In today's economy, workers can punch in from home, specialize, freelance and be their own bosses. Working moms can better balance a career with family by consulting 20 hours a week. Some people would contend that this is a crisis and that every person not on the job for 40-plus hours a week is suffering. That's an insult to the progress our country has made.

The definition of a job has changed, but the payroll survey hasn't. It is a huge mistake to focus on an illusory problem of economic weakness. Instead, policymakers should update business laws to reflect the new reality - the rapid pace of change in the workplace.

American workers need health care that's portable between jobs. They need pensions and 401(k) rules that are as flexible to move between employers as they are. What they don't need is more hot air about flawed statistics.

Timothy Kane is a research fellow in macroeconomics in the Center for Data Analysis at the Heritage Foundation (www.heritage.org), where Andrew Grossman is an editor.