Last March, the federal government estimated that economic growth increased at a 2.1 percent annual rate in the first quarter.

By April, the government revised the estimate to 1.3 percent.

A month later, the government announced it was still a little bit off in its estimate and that the economy had grown at only a 0.7 percent annual rate in the first quarter of the year.

Then, this month, the government released its latest revision of growth in gross national product for the first three months of the year -- a 0.3 percent increase.

Why can't the the government get it right the first time?

The first estimate the government makes is called the "flash" number and is based on very little real data and a lot of projections. The flash uses data generally for the first month and second months of the quarter and projections for the remaining month. As more comprehensive and updated information becomes available, additional GNP estimates are released by the Commerce Department.

The preliminary estimate and subsequent first and second revisions use more and more information as it becomes available.

For many years, the flash number was not made public, and some economists recently have suggested the government return to that practice. Since its public release, the flash number, despite its frailty, has become a major statistic for economic and financial fortune tellers as they search for clues to future economic trends. It has become one of the most-sought-after statistics in Washington.

Over the years, the GNP figures are steadily revised, and estimates of the quarterly rates of change may be altered substantially by the revisions. For example, two-thirds of the revisions of the flash estimate of quarterly changes in constant-dollar GNP have been between -0.9 and +2.8 percentage points, Commerce said. The flash for the second quarter issued last week was 3.1 percent. So it is likely that the second-quarter change in real GNP will not be revised below a 2.2 percent increase or above 5.9 percent as subsequent changes are made in the estimate of real GNP.

Ninety percent of the revisions of the flash estimate of quarterly changes in constant-dollar GNP have been between -2.4 and +4.5 percentage points so that nine-tenths of the time the 3.1 percent estimate of the flash would not be revised below a 0.7 percent increase or above 7.6 percent.

The government prepares the flash estimate 15 days before the end of the quarter using only one or two months of data for the quarter. The government collects information in certain categories of spending and investment.

For personal consumption expenditures, the flash issued this month used April and May retail sales, unit sales of new cars through the first 10 days of June and sales of new trucks for April and May.

In the nonresidential fixed-investment category, the government collects the same data for automobiles and trucks as it did for personal consumption expenditures in addition to construction put into place in April and April manufacturers' shipments of equipment.

Under residential investment, government economists look at data for April construction and April housing starts.

The government measures changes in business inventories by examining April book values for manufacturing and trade and auto inventories for April and May.

One category that has been blamed for the sluggish growth in GNP in the last year has been net exports of goods and services -- basically, the difference between U.S. exports and imports. In collecting that data, the government looks at April merchandise trade.

The government also purchases goods and services, which are included in GNP. The economists study federal unified budget outlays for April, state and local construction in April and state and local employment for April and May.