Skip links

Discovering Truths in Economics: Part I Cocktail Economics

Economists ordinarily employ a scientific method. This method can be described as a three-step process. The first step in economic analysis is for the economist to make some simplifying assumptions. That is, the economist says, “Suppose that certain things are true to start with. What follows from them? The idea of starting with statements we assume to be true may strike readers as odd, but it is not. We cannot know the truth directly. It is not usually obvious. And we have to have some starting points in our investigation, to avoid an endless chain of “whys”. When a father gets tired of answering his child’s interminable “whys”, he says, “Because I say so.” This statement can be likened to the starting points or assumptions of economists. They say, for example, “Let us assume that all people act entirely out of self-interest.” Now, of course, all people do not act entirely out of self-interest. But the economist supposes that they do and asks, “If all people do act solely out of self-, what do we expect will happen?”

To find out what we would expect to happen if everyone acted selfishly, we trace out the logic of its assumption. To do this, we employ the techniques of logic and mathematics. Once we do this, we have a set of hypotheses or predictions. For example, mainstream economists take the assumption of self-interested behavior and narrow it down to have meaning in an economic context. In capitalist economies, the production and distribution of output take place in markets. Markets are where buyers and sellers meet (not necessarily literally) and make exchanges. So the economists take the assumption that everyone acts out of self-interest to mean that the buyers and sellers in markets act only in their interests. To make this assumption amenable to the tools of logic and mathematics, they further refine it to get the most of something. For the sellers of automobiles, for example, the substance maximized is profits.

If we assume that each market participant is a maximiser, we can trace out the logic of this assumption in such a way that we end up with a large number of predictions or hypotheses. For instance, one such prediction is that, in a world of “maximisers”, if a government raises the minimum wage, employers will cut employment. That is, the economist predicts that the minimum wage and employment are related in an opposite or inverse way: an increase in the one (minimum wage) is associated with a decrease in the other (employment). There are thousands of other examples of predictions that derive from the maximizing assumption.

The first two steps—making assumptions, and deriving hypotheses from them—are sometimes called “building a model” of the economy or some part of it. However, we cannot get at the truth at this point. All we have is a hypothesis (or lots of them) derived from the initial assumptions. We need a further and crucial step. The predictions must somehow be tested against the evidence. When the government does, in fact, raise the minimum wage, does employment, in fact, fall? The test is the absolutely critical part of the scientific method. Without it, we have nothing except an exercise in logic.

Testing predictions against the evidence is a difficult undertaking. Problems abound. In the natural sciences, researchers can conduct ‘controlled’ experiments. They can set up their experiment in such a way that they can hold fixed any number of variables that might affect the outcome. With these variables fixed, they can then allow one other variable to change. Any change in the outcome, then, is likely to be due to the change in this one variable. It cannot be due to the other variables, because these have not been allowed to change. In the world of human interactions, however, it is not always or even normally possible to conduct controlled experiments. Therefore, there is always a certain element of uncertainty inherent in social science research. We can never be sure that the change in the result (employment in the minimum wage example) is the consequence of the change in the variable we are interested in (the minimum wage). It may be due to a variable we could not control. Economists have devised a large number of ingenious tests of predictions, using methods that operate as if we had conducted a true controlled experiment. Unfortunately, these seldom have the same power as test in the natural sciences.

There are many other testing problems. Measuring economic variables is often difficult or subject to differences of opinion. Economists have defined such commonplace entities as profits and incomes in entirely different ways. What this means is that a test of a prediction may give different results depending on how the variables are defined. An economist may predict that an increase in corporate profits will increase corporate investment, but the test results may hinge on exactly how profits are defined. Even if there is agreement on definitions and it is possible to precisely measure the variables, the results of any test may be due to chance. It may appear as though a higher minimum wage is correlated with lower employment, but this may just be the result of the particular sample of workers the economist chose to investigate. If we tool a sample of 5000 persons drawn randomly from a much larger population (random means that each person in the population has the same chance of getting chosen in the sample) and calculated the average height, we could almost always get an average very close to that of the entire population. But not always. By chance, we may have picked only the tallest persons for our sample. The same problem occurs in every test of an economic prediction.

All of this is not to say that we can never discover the truth in economics. If numerous studies are done by independent researchers and all of the results are pretty much the same, we can be fairly certain that the hypothesis is true. If very large numbers of the predictions derived from tracing out the logic of a particular set of assumptions are supported in repeated research by different researchers using different sets of data, we may also conclude that the assumptions themselves are true, or at least very useful in our search for the truth.

By Michael D. Yates
Economist, Labour Educator and Assistant Editor of Monthly Review, New York.



Reference

Extracted and adapted from Yates, Michael D. 2003. Naming the System: Inequality and Work in the Global Economy. Monthly Review Press. New York.

This website uses cookies to improve your web experience.