Following yesterday’s low first quarter GDP growth figure, the New York financial community — traders dealing in bonds and stocks, investors seeking profitable opportunities — will be watching closely this morning as the Bureau of Labor Statistics releases its initial estimate of the number of new jobs created by employers in May. This figure is the first major monthly indicator of our economic strength.

But few know that the final payroll numbers for May 2007 won’t be available until February 2009, when the last adjustments for the period April 2007 to March 2008 are completed.

Despite the scrutiny facing today’s numbers, we will only know last month’s exact jobs level in 20 months. By then the number will be primarily of historical interest.

To judge from recent experience, May’s early snapshot may be followed by upward revisions in the months ahead as more data come in.

America issues monthly employment statistics faster than other countries, which wait weeks to issue the same information. Today we have data for May from two sources: a sample of 400,000 employers reporting on their payrolls, and the Census Bureau’s household survey, a telephone survey of 60,000 families.

BLS could improve the accuracy of the initial jobs figures by delaying publication for two weeks, allowing more information to be collected from employers. Such a change might, initially, displease financial markets but would yield more accurate statistics and smaller revisions.

We’d get our information on employment two weeks after the end of the month, about the same time that we get readings on retail sales, housing starts, and inflation. No one complains that these data are late.

Deputy commissioner of the Bureau of Labor Statistics, Philip Rones, says that he would be willing to postpone releasing employment data for two weeks if this was Americans’ preference, but analysts want to get their employment data early — if only because they always have. “We have not found a single data user who thinks that delaying the data release is a good idea,” Mr. Rones said. “Astute data users understand the limitations of using survey data.”

Payroll figures are revised in the following two months as more data are collected. Today BLS will publish revisions to April’s initial estimate of 88,000 new jobs and March’s second estimate of 177,000.

Much later, data are adjusted, or “benchmarked,” to fit a total employment count based on employers’ unemployment insurance filings. Annual benchmark adjustments for the 12-month periods ending every March are released the following February.

In contrast, household data, from which BLS derives state and national unemployment rates as well as a separate employment count that includes self-employed, farm jobs, and household workers, are never revised, leading to the false sense that they’re more accurate.

Over the past two years, as has previously been the case in a growing economy, monthly revisions to initial payroll estimates have been substantial.

In February 2007, BLS announced that an unusually high level of 752,000 additional jobs had been created over the period April 2005 to March 2006, 45% more than the initial estimated change.

Of course, revisions to the monthly job creation numbers have been small in relation to total payroll employment. In an economy with more than 137 million payroll jobs, the monthly revision is a change of one-tenth of 1%; the latest benchmarking was six-tenths of 1%. If you recognize that BLS’s target is total payroll employment, then revisions appear less problematic.

But changes rather than totals matter to traders and investors. Monthly changes influence market guesses of whether the Federal Reserve will raise or lower short-term interest rates. Because of intense market interest, radio, TV, and wire service reporters rush out stories right after the data release at 8:30 a.m. Even though initial estimates may be overstated or understated, today’s markets will react if payroll employment growth is substantially more or less than forecast amounts of at least 135,000.

When markets react, consumers and investors can gain or lose. Securities prices may rise of fall. Monthly jobs numbers can affect long-term interest rates.

As prices rise, some firms defer expansion and some home-buyers are priced out of the market. If borrowing costs fall, demand for credit rises. When the employment data are more accurate, then more precise signals are given to the financial and business community and to consumers.

Another way to improve data accuracy with early release dates would be for BLS to expand its use of telephone calls to firms who file employment reports through the mail or by touch tone phone service, rather than by computer. This would include 8% more of the sample in the first estimate, at an extra cost of $3 million annually.

Alternatively, data could be made more accurate by making the annual benchmark quarterly. Since BLS receives administrative counts every calendar quarter, it could do quarterly benchmarking and get the final number to the public more quickly.

This closely watched monthly series is the first broad-gauge picture of the strength or weakness of the economy. With a few small changes we can make it even more precise. Americans deserve better.

Ms. Furchtgott-Roth, former chief economist at the U.S. Department of Labor, is a senior fellow at the Hudson Institute.