Inexactness, Ceteris Paribus Clauses, and “Unrealistic Assumptions”

As mentioned in the previous section, the most important methodological issue concerning economics involves the simplification, idealization, and abstraction that characterize economic theory and resulting doubts about whether economics is well supported.

3.1 Classical economics and the method a priori

The first extended reflections on economic methodology appear in the work of Nassau Senior (1836) and John Stuart Mill (1836). Their essays must be understood against the background of the prevailing economic theory. Like Adam Smith's economics (1776), to which it owed a great deal, and modern economics, the “classical” economics of the middle decades of the nineteenth century traced economic regularities to the choices of individuals facing social and natural constraints. But, as compared to Smith, more reliance was placed on severely simplified models. In David Ricardo's Principles of Political Economy (1817), a portrait is drawn in which wages above the subsistence level lead to increases in the population, which in turn require more intensive agriculture or cultivation of inferior land. The extension of cultivation leads to lower profits and higher rents; and the whole tale of economic development leads to a gloomy stationary state in which profits are too low to command any net investment, population growth drives wages down to subsistence levels, and only the landlords are affluent.

Fortunately for the world, but unfortunately for classical economic theory, the data consistently contradicted these gloomy predictions (de Marchi 1970). Yet the theory continued to hold sway for more than half a century, and the consistently unfavorable data were explained away as due to various “disturbing causes.” It is consequently not surprising that Senior's and Mill's accounts of the method of economics emphasize the relative autonomy of theory.

Mill distinguishes between two inductive methods. The method a posteriori is a method of direct experience. In his view, it is only suitable for phenomena in which few causal factors are operating or in which experimental controls are possible. Mill's famous methods of induction (1843, Book III) provide an articulation of the method a posteriori. In his method of difference, for example, one holds fixed every causal factor except one and checks to see whether the effect ceases to obtain when that one factor is removed.

Mill maintains that direct inductive methods cannot be used to study phenomena in which many causal factors are in play. For example, if one attempts to investigate whether tariffs enhance prosperity by comparing nations with high tariffs and nations without high tariffs, the results will be worthless because the prosperity of the countries studied depends on so many other causal factors. So, Mill argues, one needs instead to employ the method a priori. Despite its name, Mill emphasizes that this too is an inductive method. The difference between the method a priori and the method a posteriori is that the method a priori is an indirect inductive method. One first determines the laws governing individual causal factors in domains in which Mill's methods of induction are applicable. Having then determined the laws of the individual causes, one investigates their combined consequences deductively. Finally, there is a role for “verification” of the combined consequences, but owing to the causal complications, this testing has comparatively little weight.

As an example of the method a priori, Mill discusses the science of the tides. Newton determined the law of gravitation by studying planetary motion, in which gravity is the only significant causal factor. Then he and others developed the theory of tides deductively from that law and infor­mation concerning the positions and motions of the moon and sun. The implications of the theory will be inexact and sometimes badly mistaken, because many subsidiary causal factors influence tides. By testing the the­ory one can uncover mistakes in one's deductions and evidence concern­ing the role of the subsidiary factors. Because of the causal complexity, such testing does little to confirm or disconfirm the law of gravitation, which has already been established. Although Mill does not often use the language of “ceteris paribus” his view that the principles or “laws” of economics hold in the absence of “interferences” or “disturbing causes” provides an account of how the principles of economics can be true ceteris paribus (Hausman 1992b, ch. 8, 12).

Because economic theory includes only the most important causes, its claims, like claims concerning tides, are inexact. Its predictions will be imprecise and sometimes completely wrong. Mill maintains that it is nevertheless possible to develop economic theory by studying in simpler domains the laws governing the major causal factors and then deducing their consequences in more complicated circumstances. For example, the statistical data are ambiguous concerning the relationship between minimum wages and unemployment of unskilled workers; and since the minimum wage has never been extremely high, there are no data about what un­employment would be in those circumstances. On the other hand, every­day experience teaches one that firms can choose among more or less labor-intensive processes and that a high minimum wage will make more labor-intensive processes more expensive.

In defending a view of economics as in this way inexact and employing the method a priori, Mill was able to reconcile his empiricism and his commitment to Ricardo's economics. Although Mill's views on economic methodology were questioned later in the nineteenth century by some economists who believed that the theory was too remote from the contingencies of policy and history (Roscher 1874, Schmoller 1888, 1898), Mill's methodological views dominated the mainstream of economic theory for well over a century (for example, Cairnes 1875). Mill's vision survived the so-called neoclassical revolution in economics beginning in the 1870s, and it is clearly discernible in the most important methodological treatises concerning neoclassical economics, such as John Neville Keynes' The Scope and Method of Political Economy (1891) or Lionel Robbins' An Essay on the Nature and Significance of Economic Science (1932). Hausman (1992b) argues that current methodological practice closely resembles Mill's methodology, despite the fact that few economists would explicitly defend it.

3.2 Milton Friedman and the defense of “unrealistic assumptions”

Although some contemporary philosophers have argued that Mill's method a priori is defensible (Bhaskar 1978, Cartwright 1989, Hausman 1992b), by the middle of the twentieth century Mill's views appeared to many economists out of step with contemporary philosophy of science. Without studying Mill's text carefully, it was easy for economists to misunderstand his terminology and to believe that his method a priori is opposed to empiricism. Others took seriously Mill's view that the basic principles of economics should be empirically established and found evidence to cast doubt on some of the basic principles, particularly the view that firms attempt to maximize profits (Hall and Hitch 1939, Lester 1946, 1947).

Philosophically reflective economists proposed several ways to replace the old-fashioned Millian view with a more up-to-date methodology that would continue to justify current practice (see particularly Machlup 1955, 1960 and Koopmans 1957). By far the most influential of these was Milton Friedman's contribution in his 1953 essay, “The Methodology of Positive Economics.” This essay has had an enormous influence, far more than any other work on methodology.

Friedman begins his essay by distinguishing between positive and normative economics and conjecturing that policy disputes turn largely on disagreements about the consequences of alternatives and thus are capable of being resolved by progress in positive economics. Turning to positive economics, Friedman asserts (without argument) that the ultimate goal of all positive sciences is correct prediction concerning phenomena not yet observed. He holds a practical view of science and looks to science for predictions that will guide policy.

Since experimentation is often impractical and since the uncontrolled phenomena economists observe are difficult to interpret (owing to the same causal complexity that bothered Mill), it is hard to judge whether a particu­lar theory is a good basis for predictions or not. Consequently, Friedman argues, economists have supposed that they could test theories by the “real­ism” of their “assumptions” rather than by the accuracy of their predictions. Friedman argues at length that this is a grave mistake. The realism of a theory's assumptions is, he maintains, irrelevant to its predictive value. Theories should be appraised exclusively in terms of the accuracy of their predictions. It does not matter whether the assumption that firms maxi­mize profits is realistic. What matters is whether the theory of the firm makes correct and significant predictions.

As critics have pointed out (and almost all commentators have been critical), Friedman refers to several different things as “assumptions” of a theory and means several different things by speaking of assumptions as “unrealistic” (Brunner 1969). Since Friedman aims his criticism at those who investigate empirically whether firms in fact attempt to maximize profits, he must take “assumptions” to include central explanatory generalizations, such as “Firms attempt to maximize profits,” and by “unrealistic,” he must mean, among other things, “false.” In arguing that it is a mistake to appraise theories in terms of the realism of assumptions, Friedman is arguing at least that it is a mistake to appraise theories by investigating whether their central explanatory generalizations are true or false.

It would seem that this interpretation makes Friedman's views incon­sistent, because in testing whether the assumption that firms attempt to maximize profits is realistic, one is checking whether predictions of theory concerning the behavior of firms are true or false. An “assumption” such as “firms maximize profits” is itself a prediction. But there is a further wrinkle. Friedman is not concerned with every prediction of economic theories. In Friedman's view, “theory is to be judged by its predictive power for the class of phenomena which it is intended to explain” (1953, 8; emphasis added). Economists are interested in only some of the implications of economic theories. Other predictions, such as those concerning the results of surveys, are irrelevant to policy. What matters is whether economic theories are successful at predicting the phenomena that economists are interested in. In other words, Friedman believes that economic theories should be appraised in terms of their predictions concerning prices and quantities exchanged on markets. In his view, what matters is this “narrow predictive success” (Hausman 2008a), not overall predictive adequacy.

So Friedman permits economists to ignore the disquieting findings of surveys. They can ignore the fact that people do not always prefer larger bundles of commodities to smaller bundles of commodities. They need not be troubled that some of their models suppose that all agents know the prices of all present and future commodities in all markets. All that matters is whether their predictions concerning market phenomena turn out to be correct or not. And since anomalous market outcomes could be due to any number of uncontrolled causal factors, while experiments are difficult to carry out, it turns out that economists need not worry about ever encountering evidence that would disconfirm fundamental theory. Detailed models may be disconfirmed, but fundamental theory is safe. In this way one can understand how Friedman’s methodology, which appears to justify the pragmatic view that economists should use any model that appears to “work” regardless of how unreasonable its assumptions might appear, has been put in service of theoretical orthodoxy. For other discussions of Friedman’s essay, see Bear and Orr (1967), Boland (1979), Hammond (1992), Hirsch and de Marchi (1990), Maki (1992), Melitz (1965), Rotwein (1959), and Samuelson (1963).

Over the last two decades there has been a surge of experimentation in economics, and Friedman’s methodological views do not command the same near unanimity that they used to. But they remain enormously influential, and they still serve as a way of avoiding awkward questions concerning simplifications, idealizations, and abstraction in economics, rather than providing a response to them.

4.