NUMBERS CAN be powerful weapons, especially against those who don't understand them. And no administration has made more effective use of numbers to promote its social policies than the current one. Budget director David Stockman, the architect of many of those policies, is an avid consumer of federal statistics. You may question the quality of some of his data or disagree violently with his conclusions, but you cannot deny his numerical ability. His sources are the federal statistical agencies that produce data on income, employment and other measures of economic and social well-being.

In recent years, pressure on these agencies has increased for more and better data. As the economy moves through a period of rapid structural change, it is increasingly important to look behind the old familiar statistics to see what is really going on. The unemployment rate that receives so much attention each month, for example, masks important differences, not just in the age, sex and racial distribution of jobs, but also among areas and industries. With the economy now moving along two separate tracks--the booming path of energy and high technology industries and the deteriorating path of construction, automobiles and other basic industries--the overall unemployment rate is no longer a reliable guide to economic policy.

Average income measures are also misleading. You get quite a different picture of how the economy has been performing--and what policies are best--depending upon your data. Simple average weekly earnings give a rather bleak picture. But if you adjust for important differences in the age and family status of workers, things look much brighter. This is especially true if you add the growing importance of fringe benefits and better Social Security and medical coverage.

Because the administration's policies can be expected to have different consequences for different areas and different types of people, measuring their progress will require more detailed data than is now available. These new demands will come at the same time that statistical agencies will be taking substantial cuts in budget and personnel. Some important surveys have already been cancelled, and the Bureau of Labor Statistics has announced delays in future releases on employment, earnings, prices and other indicators.

Agencies will face a difficult choice between the quality and the quantity of their products--and the heavy pressure is likely to be on the side of quantity. There is a saying among seasoned players in the policy game that "some numbers beat no numbers every time." That's quite true. But there is an important corollary, which is that bad numbers are much more dangerous than no numbers. As the nation enters a period of major change, the greatest care must be taken to ensure not only the availability but the quality of the numbers that guide and measure government policy.