Economists spend a lot of their time trying to understand how the world works.

Their advice to politicians and government officials, their forecasts for the future, and their predictions about how certain policies will affect the economy, all depend on notions about what makes people behave the way they do. The more accurate these notions, the better the economists' advice.

It's no secret that the reputation of the economics profession is now very low, with the United States and much of the rest of the world mired in the worst recession since World War II. The experts apparently cannot agree about the main cause of recession and, more important, what is the best way out of it.

What is needed, says economist Michael Hartley, is a much better model of how the world works. If economists only knew more precisely just why people save or spend; or under what conditions companies will hire more workers, build new plants, raise or lower prices; then they could assess much more accurately what the impact of different government policies would be, he says.

Hartley, with the World Bank, thinks that he has a "key idea" that could eventually produce the needed improvement in modeling, and could solve a number of problems that now baffle theoretical economists. In a series of papers available as World Bank discussion papers, he describes how it may be possible to "build up" models of the economy from detailed information on how households, companies, farmers, banks and other individual participants behave to give a more accurate picture of the interrelationships within the whole economy.

At present, most economic models are constructed "top down" by using data for the whole economy. In the United States, for example, they are based on "national" incomes, "national" consumption or "national" investment.

The "models" consist of a series of equations that define the relationships of these different economic aggregates to each other and to outside factors, such as changes in tax rates.

Nobel prize winner Lawrence Klein, under whom Hartley has studied, pioneered this kind of "top down" modeling. However, Klein says that "most people agree that it is better in theory to go 'bottom up' than 'top down.' " Changes in total consumer spending, for example, depend on how millions of individuals make their spending decisions. Each one may have a slightly different reaction to a cut in taxes, for example, or a rise in the stock market. When economists work out the interrelationships in the whole economy just by looking at summary data, they lose much of the detailed information about how each individual acts.

However, for "practical reasons," Klein says, he and others have concentrated on refining and improving the more traditional macro-models, rather than building up macro-models from micro-models.

These practical problems are now potentially soluble, Hartley believes, largely because of the tremendous advances in computer technology of recent years.

Hartley's "key idea" involves using computers to work out the relationships that determine spending or saving, for example, when these relationships are too complex to be described by mathematicians without the aid of a computer. At present, models of the macro economy can only be built up from individual decisions when mathematical economists can describe, in an equation, how the decisions are determined and how to add them up to macro-economic concepts, such as national income, or gross national product. However, there are many cases when this cannot be done. The way in which the individual--whether consumer, company, farmer or bank--reacts to economic circumstances is too complicated for the mathematician to work it out with pen and paper.

By contrast, a computer could work out the best way of describing the economic interactions, Hartley says, if it is supplied with enough data about how people, companies and so on actually behave. The computer would have fed into it a wealth of data, gathered from surveys or censuses, about individual decisions. It would then try out over and over again different ways of explaining just how the decisions were made, and how different factors affected the eventual outcome. It would test many different explanations--or descriptions of the relationships--against the information that it had been given about what actually happened. Eventually, it would determine which one was the best, or most likely.

Once a "micro-model" of the decisions of individuals, or groups of individuals, is developed, it can be used to build a "macro" model of the whole economy by adding together the relationships between things such as spending, income and so on, rather than by adding together the individual totals for spending, income, profits, investment, and so forth, and attempting to work out the overall relationships from the aggregated data, Hartley says.

Klein agrees that Hartley's work is a "fresh approach" that "has promise." However, he believes that considerable problems remain, and said that he believed "all scholarly research proceeds one step at a time . . . this represents an important step." But it could take several years of further work and research by a team of people to develop the idea and build a Hartley-style model, he says.

One difficulty is that the proposed method of model building would be "very demanding of data, and especially of data that may be hard to come by in the developing countries" in which Hartley, as a researcher at the World Bank, is particularly interested, Klein said.

The method would also require a lot of expensive computer time. Rapid advances in computer technology make this a less daunting problem than it might be, Hartley says. He has discussed his work with the eminent applied mathematician Richard Bellman, of the University of Southern California, who says he believes the computing and mathematical problems can be solved, although 10 years ago the necessary computing would have been too costly.