What Is a Science?
Should we accept claims of economists who say they are scientists? To decide, we must first know what science is.
Philosopher Karl Popper's widely accepted definition of science says that a statement is scientific only if it is open to the logical possibility of being found false. This definition means that we evaluate scientific statements by testing them, by comparing them to the world about us. A statement is nonscientific if it takes no risk of being found false; that is, if there can be no way to test the statement against observable facts or events. Popper called this distinction the "line of demarcation."
An implication of Popper's definition is that one can never be completely sure that any scientific theory is true. Accepted scientific theory is only theory that has not yet been contradicted by evidence, though the future may bring a contradiction. For example, we cannot be absolutely sure that the statement, "The sun will rise in the east tomorrow" is true because it is a scientific statement. We can easily think of a logical possibility that would refute it—a sunrise in the west. We have great confidence that such an event will not happen because the sun has always risen in the east. However, the fact that all previous experience has been consistent with the statement does not prove that the statement will never be refuted.
Popper saw the growth of scientific knowledge as a process of conjecture and refutation. Someone originally comes up with a way of explaining a set of facts: a conjecture or guess or theory about how the facts are related. If further observation is inconsistent with the theory, the theory is considered refuted and a new theory or conjecture must be found. In contrast, if the original explanation is nonscientific, it will never be refuted and there will never be any need to change beliefs.
Most economists see their discipline as scientific in Popper's sense of the word. Economic theory makes statements about how facts fit together, and there are constantly new sets of facts arising that allow one to test the theory to see whether the facts are as theory predicts. However, this process is more difficult for economists than it is for most physical scientists.
Unlike physical scientists, economists can almost never use controlled experiments to gather facts with which to test theories. Rather they must use whatever facts the world gives them and rely on statistical procedures to draw conclusions. Although statistical procedures let economists hold some variables constant to see the effect of other variables, just as a controlled experiment does, they are subject to serious limitations. If there are variables that the theory says are important but they cannot be measured or they can be measured only imperfectly, statistical procedures may give misleading results. Or the procedures may fail if the theory is uncertain exactly which of the many possible variables that may be involved must be controlled. One strength of a properly done controlled experiment is that there is no need to list all the factors that are controlled. The procedure is such that only one factor, or a small and known group of factors, is different between the control and experimental groups. Given these difficulties, it is not surprising that controversy about whether a theory is supported or rejected by the facts can last for many years in economics.1
There is a minority of economists, however, who do not see economics as scientific in Popper's sense. A group of economists called the Austrian school, for example, has argued that economics starts with assumptions and that economic theory is the logically deduced results of those assumptions. If the theory does not fit the facts, one cannot conclude that the theory is wrong, but only that it is inappropriate to apply the theory in that particular situation because the initial conditions do not agree with the assumptions of the theory.
Besides distinguishing between scientific and nonscientific statements, one can make a positive/normative distinction.
1 Statistical procedures are used in controlled experiments, but the controlled environment leads to simpler procedures with results less open to lengthy dispute.