Testability in Economic Science

• Date: 01 March 1977
• Author: L. A. BOLAND
• Published date: 01 March 1977
• DOI: http://doi.org/10.1111/j.1813-6982.1977.tb00510.x

Testability in Economic Science

L. A. BOLAND*(1)

PHILOSOPHERS of science have drawn fine distinctions between the ideas of testability, falsifiability, verifiability, and 'the like, but

little of their literature seems relevant for economists. Yet, the profession seems to have adopted a particular view of the

philosophy of science. That view says that economic theories are "scientific" if and only if they are testable. Certain aspects of this

popular view are still not clear.*(2) What is meant by "testability" is not obvious. Does testability of a theory mean that it is

conceivably false (thus, not a tautology) or that it is conceivably falsifiable (thus, it is possible for the real world to be so

constructed that some possible observation would contradict the theory as a whole) or does it mean something else?

In this paper I will examine the basis for testability where testability is to be viewed as the possibility of empirical refutation: the

possibility of observing conceivable facts which when conjoined with some theory in question forms a logical contradiction. This

paper is intended to be a challenge to those economists who would have us utter nothing but testable (empirically refutable)

statements and who also put a premium on logical arguments. The examination is concerned with two basic questions: (1) When is

a statement (e.g. prediction or assumption) considered to be false? And (2) when is a theory considered to be false? My argument

will be an outline of a proof that on the basis of rather standard views of testing (it is relative to testing conventions), of the nature

of logic (it is ambiguous regarding which assumption "caused" a false prediction), and of the role of models (to add more

assumptions in order to test a theory), it is impossible to test any economic theory convincingly even when it is not tautological.

Thus, the requirement that economic theories must be testable has to be rejected as an

1977 SAJE v45(1) p94

impossible goal, or those who consider economic science as consisting of necessarily testable theories must concede defeat. I

stress that in what follows, I am only reporting.

As a matter of logic, the logician Karl Popper argues that testability cannot be verifiability.*(3) Verification, the demonstration of

the truth of a statement or theory, is impossible any time we have to prove that a strictly universal statement is true - even if it is

true.*(4) (E.g.,in proving that "all swans are white" we must guarantee that no future swan will be non-white.) This

consideration does not preclude falsifiability, as we can prove that such a statement is false with the demonstration of at least one

counter-example.*(5) There is an asymmetry between the testability of statements which use the quantifier "all" and the testability

of those which use the quantifiers "some" or "at least one". With this quantificational asymmetry in mind, the view to be examined

here is the view that in economics we must be concerned to put forth statements or theories which, as Samuelson says, "could

conceivably be refuted, if only under ideal conditions".*(6) As it turns out, Popper's logic sword cuts both ways when we build

models of our theories.

Economics today is alleged to be "scientific" mainly because economic theorists now have accepted the need to express their ideas

in terms of specific models. Mathematical models of economic theories supposedly have enabled us to put our economic theory to

test. The question of this paper is: Does refuting a specific model of a given theory necessarily refute the theory represented by

that model? If not, then the popular methodological requirement that our theories be testable would seem to be rather puzzling. It

will be argued here that consistency demands that the requirement of testability (as "empirical refutability") be abandoned for the

same reasons that we traditionally drop the requirement of verifiability - even when there are indisputable facts available to

1977 SAJE v45(1) p95

formulate a would-be refutation or verification. In short, for fundamental logical reasons it is impossible to test an economic theory

by constructing specific models ofthat theory which are more empirically restrictive than the theory itself. The argument presented

here is based mostly on the recognition ofaccepted views ofthese matters which can be found in the standard literature ofeconomic

theory.

Tautology V. Testability

In 1955 Samuelson warned us that "to test, refute, or `verify' a meaningful proposition is a tough empirical job and no amount of

flossy deduction can obviate this". Despite Samuelson's warning, little has been said in our literature that examines the logic

oftesting in economics. The exception is Bear and Orr,*(7) but even their examination was limited to the discussion surrounding

Friedman's instrumentalistic view concerning the need to test assumptions.*(8) The current positive concern for testability stems

instead from the traditional methodological prescription that we should strive to verify our theories, assumptions or predictions.

The problem that might arise in this regard is that a tautology could be verified even though we might think it is not meaningful or

informative.*(9) That is, a successful verification might not be meaningful. To assure that any verification is meaningful we should

make sure our theories are not tautologies.*(10) On the other hand, if a theory is testable then any verification would be

51