Testability and the Methodology of Positive Economics

Friedman also made an influential contribution to economic methodology in arguing that an economic theory should be tested by its predictive accuracy alone, without regard to the realism or unrealism of its assumptions (Friedman 1953). The argument is questionable because, owing to the difficulty of assessing such consequences (reliable experiments concerning economic behavior are very difficult to conduct), economists have tended to rely heavily—certainly Friedman did—on assumptions, often assumptions based on priors rooted in ideology, personality, or other subjective factors.

Richard A. Posner, "Why Is There No Milton Friedman Today?", Econ Journal Watch 10: 210-213 (p. 210)