"A policy-maker I know would urge his staff into action with this reminder: "In this town, some numbers beat no numbers every time."

Any politician can get a certain amount of mileage from a heart-warming anecdote, patriotic appeal or hair- raising warning. But nowadays, when the time comes to divvy up the budget pie, policy-makers feel better if they can point to an impressive- looking pile of statistics to justify their actions.

The current virtuoso of the numbers trade is, of course, Office of Management and Budget Director David Stockman. In the early days of the Reagan administration, he wowed Congress with artfully constructed charts, intimidating computer printouts and a ready supply of authoritative-sounding numbers in support of the administration's plans.

So dazzled was Congress by his briefcase full of numbers that it overlooked a large failing in the Stockman plan: his numbers didn't add up. If you totaled the domestic cuts, defense increases and tax reductions, there remained a startling imbalance between federal revenues and expenses.

In Washington, there are three types of people who thrive on numbers: those who know the numbers, understand them and report them accurately; those who know the numbers, understand them and manipulate them to their advantage; and those who haven't a clue what the digits mean, but go about spouting numbers--often in billions and trillions--hoping to sound economically astute.

The number-crunchers are those who try to keep the numbers honest. They are the economists and policy analysts who mix basic program and statistical data together with various assumptions about the past and future, stir them up in a computer and cook up estimates of what this or that policy change will do.

Number-crunching started in the Pentagon and moved to the domestic agencies. In time Congress caught on to the fact that to keep up in the policy game it needed its own sets of numbers. Today the most capable group of estimators are in Congress. It is the staff of the Congressional Budget Office, which provides a steady stream of numbers on almost every program that Congress considers. It is also this staff that is often the first to point out that Stockman's numbers do not compute. Congress can also call upon the Congressional Research Service, or on the budget, tax and economic committees.

All these statisticians and number-crunchers and economic forecasters have helped to elevate both the quality and the quantity of congressional debate. But they have also added to the confusion. Computers have so transformed the numbers racket in Washington that it starts to resemble the latest space-age game at the video arcade. Consider the world of the number crunchers: they deal with the realm of the known, where there are "facts" that presumably describe the world as it is. But they are also fascinated with the realm of the imagined, where they spend their days with "factoids"--data produced by a computer's simulation of the world as it might be. Consider too a lawmaker who must try to separate the "facts" from the "factoids" to understand how a program is to work or a policy to be enacted.

Where do Stockman and other purveyors of quantified wisdom get their facts? Or, where do factoids come from? The federal statistical agencies are a major source: the Census Bureau, Bureau of Economic Analysis and the Bureau of Labor Statistics. There are also centers in other federal agencies that compile information on health, education, crime and other indicators of the nation's social well-being. Federal, state and local agencies put together information to monitor their operations and support their budget requests. Finally there are special-purpose surveys conducted by universities and other private research organizations at the behest of federal agencies.

Some of these numbers speak for themselves -- or at least appear to. For example, if the gross national product goes up and unemployment goes down, you can bet the economy is improving.

But numbers can send conflicting signals, or they may not tell you what you need to know. Economic indicators rarely give any information about those outside the mainstream: nursing home residents, for example, or ghetto youth. And data can be misleading, too, because they only tell us about the past and may lead us to think that the future will be more of the same.

But the numbers racket-- for all the opportunities it provides for misuse and misinterpretation--is an honorable trade. For the better part of two decades--in a variety of government agencies and private organizations--I was an avid maker and peddler of numbers. I have concocted estimates on such subjects as the impact of nuclear attack upon the economy (devastating) to the effects of changes in work and income-reporting rules on the behavior of welfare recipients (modest but not negligible.)

Most of the predictions that my colleagues and I generated have--happily for both us and the world--never been put to the test of experience. But when I've had occasion to compare what we predicted--about, say, program caseloads or budget costs-- with what actually happened, I've usually been surprised at how close we were.

And the exercise of constructing an estimate is a useful discipline. It makes you think about how programs really work and how the people they affect will react to them. One reason why the Great Society programs ended up costing so very much more than their founders predicted was that the program's planners didn't take the time to figure out what would happen when, for example, poor people were given free access to medical care and doctors could stop worrying about getting paid.

The danger, of course, is that numbers--especially when printed out by a computer--have a spurious accuracy and authenticity. Members of Congress have learned the hard way to be less gullible about the numbers that others feed them. So should we.