In an era in which the physical world is precisely calibrated in femtoseconds, gigaflops and trillionths of a gram, it is startling to wake up in late May to the news that China's economy is four times as large as the most skilled measurers at the International Monetary Fund had previously thought.

China is but the most recent example of the unreliability of unrevised national economic figures. Yet, when published, these figures are the stuff not only of headlines but of policy decisions as well.

The great physicist Werner Heisenberg understood the centrality of accurate measurement. "The transition from the 'possible' to the actual," wrote Heisenberg, "takes place during the act of measurement." By this new act of measurement, China's economy rose from number ten in the world to number three (behind the U.S. and Japan). In one fell measurement swoop, the share of the developing world in global output almost doubled. In the real world nothing changed; the same people produced the same output. But new measures produced a whole new ranking of world economies.

The economists do a poor job of measuring the U.S. economy, too. Headlines trumpet a fractional increase or decrease in GDP, and pundits fill the airways with sage policy recommendations. GDP is up--tighten the money supply! GDP is down, extend unemployment benefits!

But the pundits never mention the fact that the difference between the Commerce Department's first report on the GDP for a quarter and the final figures, which are not produced until three years later, show huge variations--so large in fact as to make Commerce's flash reports a flimsy guide to policy. If we read, for example, that GDP growth is 3%, about half the time the final figure will turn out to be either less than 1.5% or more than 4 .5%. Figures that move around that much are not a sound foundation upon which to base economic policy.

Since all calculations of budget deficits are based on the projected growth of GDP, the fragility of the deficit figures bandied about is apparent. Honest people's estimates of projected deficits differ by many billions of dollars. The Congressional Budget Office estimates the "baseline" deficit in fiscal 1992 at $314 billion, while equally skilled measurers at the Office of Management & Budget come up with $327 billion.

To further confuse the issue, there is no standard measure of the deficit. There are at least four methods for measuring the deficit, depending on whether Social Security and deposit insurance are included or excluded, to say nothing of the deficit in the national income and product accounts. In a world of sound bites, media commentators and politicians rarely specify which deficit number they are talking about.

Even if the specific deficit number is identified, there is a deeper problem. Government officials resist all prudent efforts to put their accounts on an accrual basis, and so our $6 trillion economy has all the bookkeeping sophistication of a child's lemonade stand. No private business could be run on such rudimentary accounting.

In addition to the lack of accrual accounting, everything from a 25- cent pencil to billions of dollars' worth of bridges, roads and other productive assets is expensed. There are no capital accounts, even though the General Accounting Office has for years been recommending that government accounting be brought into the 20th century.

Examples of the crudeness of our economic measurements are everywhere. The original Standard Industrial Codes that once told us how industry was organized are now hopelessly out of date. Of the 12 major SIC code divisions, only 2 reflect the service industry, yet about 80% of all employed Americans work in the service sector.

Our productivity figures are similarly flawed. Only about 40% of the companies in the service sector are even measured for productivity increases, while we continue to collect data on yesterday'S society. For example, government numbers reveal the number of railway brakemenonce an important figure--but fail to tell us the number of computer programmers.

Who would contest that the huge investment in software that runs our factories and businesses is an asset? And yet software is uncounted in our national investment figures, which still concentrate on bricks and mortar.

Since so much political and economic policy is built upon these uncertain numbers, it is time for a national effort, using modern computer power, to set up new national accounts and modern accounting. Bad numbers cannot support good policy.