Critical reflections on a realist interpretation of Friedman’s ‘Methodology of Positive Economics’

Abstract

Uskali Mäki has offered an innovative scientific realist account of Milton Friedman’s 1953 essay, ‘The Methodology of Positive Economics’, which directly challenges the dominant instrumentalist interpretation. This paper offers critical reflections on Mäki’s approach and interpretation. It is argued that Mäki’s method of rereading-rewriting the text is problematic; that an unforced instrumentalist account of unrealistic assumptions can be extracted from the text itself; and that seemingly realist passages can be plausibly read as expressing an instrumentalist stance.

Keywords:

• realism
• instrumentalism
• unrealistic assumptions
• as-if
• F53
• Friedman

Acknowledgements

Thanks are due to Dr Margaret Moussa and two anonymous referees for their valuable insights and critical comments on drafts of this article.

Notes

1. Actually, Mäki (1986) was perhaps the first to argue (in English) that F53 contained a significant realist element. His early verdicts in this regard were, however, highly equivocal (Mäki, 1989, 1992) – for example: ‘Although Friedman is an ontological, referential, representational, and veristic realist, I still think that it is mostly correct to characterise him as an instrumentalist’ (Mäki, 1989; p. 185). The realist interpretation of F53 was made more confidently in Mäki (2003), of which Mäki (2009a) was an apparent outgrowth.

2. It is acknowledged that exegetical questions about seminal texts may not always lie at the centre of things when it comes to making methodological progress in economics. That, however, does not render such interpretive efforts completely irrelevant to questions of the legitimacy of contemporary work of the economics discipline. After all, when texts such as F53 are charged with being complicit in, or even lying at the core of, the most recent economic crises (see e.g. Crotty, 2013; Mosini, 2012), economists should, at the very least, be able to identify what the accused seminal texts actually say.

3. All unattributed page numbers refer to Mäki (2009a).

4. Lehtinen (2012), who provides a comprehensive overview of Mäki’s work, points out that Mäki has been somewhat coy in explicating the kind of scientific realism he personally believes (but see Mäki, 2000; Mäki, 2009d, 2011). The dimensions of scientific realism that Lehtinen is able to divine from Mäki’s oeuvre broadly correspond to the ones outlined above.

5. Mäki (2009b, p. 57) means by this the following: ‘the strong consumptionist is only interested in the reader’s reception rather than the author’s intention. The goal is to read a methodological text from the point of view of its (actual or possible) reception by the relevant audiences, including its (actual or possible) interpretations and influences – while completely ignoring questions about the author’s beliefs and goals. This is what I call ‘reception methodology’ in analogy with reception aesthetics’. Incidentally, Mäki’s approach of focusing on the text alone (sans Friedman’s biography or applied work) is one that he has fairly consistently followed in most of his work on F53.

6. For example, no-one would regard the following syllogism as epistemically virtuous just because its conclusion is true: (1) All bananas are economists. (2) Friedman is a banana. (3) … Friedman is an economist.

7. See Sen’s (1980) distinction between ‘the whole truth’ on the one hand and ‘nothing but the truth’ on the other.

8. Letting C stand for ‘a complete, true description of reality’ and P stand for ‘a particular, true claim about an element of reality’, in modal terms we have the following propositions: C → □P; ~C → ◊P; P → ◊C; ~P → ~◊C.

9. Mäki cites Wong (1973), Boland (1979) and Caldwell (1992) as such commentators.

10. In this quote, Mäki refers to a statement about an ‘important causal factor’. In another place, he says that ‘core assumptions’ (p. 101) should be true.

11. These are not identical to Mäki’s examples. I have used them in order to provide a more parsimonious discussion of the point Mäki is making.

12. These two approaches need not be mutually exclusive, although, depending on the text in question, one may be deemed superior to the other for various contingent reasons. For example, one may advocate the former approach if the author’s thoughts at the time of writing are inaccessible to the reader – the author may have left no ‘clues’ (or may have left only contradictory ‘clues’) as to his or her thoughts at the time, or one may be suspicious of the author’s intentions. In the extreme, one may argue that the author’s true intentions, being veiled by subjectivity, are never available to the reader. On the other hand, the conventional meanings of terms may be misleading if the author’s word usage is idiosyncratic or his or her ideas are simply poorly expressed.

13. Also, to claim that F53 does not advocate prediction because the text does not appeal to predictions when ruling out monopolistic competition, is similarly evidence of misreading conventionally understood words (cf. p. 93). F53 quite unequivocally states that imperfect-monopolistic competition models contain inherently ambiguous terms, thus rendering them meaningless. ‘The deficiencies of the theory are revealed most clearly in its treatment of, or inability to treat, problems involving groups of firms – Marshallian ‘industries’’. Particularly, ‘fuzziness and undefinable terms’ such as ‘close’ substitutes and ‘substantial’ gaps in cross elasticities, are added to Marshall’s ‘abstract model where they have no place, and serves only to make the theory analytically meaningless – ‘close’ and ‘substantial’ are in the same category as a ‘small’ air pressure’ (Friedman, 1953, pp. 38–39).

14. Since ‘initial knowledge’ is taken to refer here to empirical knowledge, it does not refer to what might be called non-empirical knowledge, such as knowledge of the rules of logic or Euclidean geometry or calculus. As such, a theory could still be deemed ‘simple’ in terms of requiring very little prior empirical information to make predictions, whilst simultaneously requiring substantial prior knowledge of, say, the mathematical operations necessary to construct and solve equations that contain that small amount of empirical information.

15. Note that although Mäki has long argued that abstraction and idealisation can be understood in ways that cohere with ‘realistic’ assertions (e.g. Mäki, 1994, 2009c), the issue here is how F53 alone conceptualises ‘unrealistic’.

16. Aside: during the mating process, the male Australian redback spider (Latrodectus hasselti) puts itself in a position that virtually guarantees its death at the hands (the fangs) of its sexual partner (Andrade, 1996).

17. This includes variations such as ‘realistic “enough”’ and ‘sufficiently close’.

18. Letting S = predictively successful and G = judged sufficiently good approximation, if it were the case that F53 was stipulating only ‘S→G’, then ‘~S & G’ should be possible. If, however, the stipulation is intended to be ‘G↔S’, then ‘~S & G’ is not possible (as seems to be implied by the text).

19. That is, in cases where there is no conflict between the ‘virtues’, such as a simpler theory versus a more fruitful theory where both are equally predictively successful.

20. Although F53 is setting up an obviously uneven contest here, the example is not one of literally ‘no contest’. One could formulate, say, a racist eye colour theory of business behaviour that could generate predictions intended to rival a cost theory. F53, in addressing economists, is simply taking it as obvious that the cost theory (whatever the functional form for costs used) would be the more predictively successful one. From that, the judgement of ‘sufficiently good’ follows.

21. Again, this does not necessarily imply that one requires little knowledge per se. The fictional mechanism which does the work of linking the prior empirical knowledge to a prediction may indeed require a substantial amount of non-empirical knowledge. In each of F53’s examples, one may require, say, knowledge of Newton’s laws of motion, Euclidean geometry and the calculus for optimisation under constraints. What one apparently does not need much of is detailed knowledge of, say, actual cognitive processes, or actual biological processes. It is also possible that knowledge of pre-existing synthetic presuppositions determines whether an ‘as if’ formulation is deemed appropriate. As Lehtinen (2013, p. 10) points out, ‘One does not say, for example, that “stones fall as if swallows do not migrate to warmer territories during the autumn”, because one presupposes that the migration behaviour of these birds could not affect the stones in any way’. However, a presupposition or ‘commonplace’ does not necessarily count as empirical knowledge of the world. It may simply be a convention based on a pervasive ideology. For some people it is permissible to say that trees sway as if prostrating to God; for others, it is an impermissible formulation. In any case, for F53, pre-test presuppositions need not direct theorisation. Say an intrepid mathematician constructs an equation which aims at predicting the terminal velocity of falling stones using flight patterns of a class of birds when feeding. Despite the fact that it would (one assumes) violate our presuppositions about causal relations, F53 would not forbid one from saying: ‘stones S fall as if they were hungry birds B’. Furthermore, the formulation would be deemed acceptable (from F53’s perspective) if it were subsequently validated.

22. It is true that the passages in F53 citing the example of Galileo’s law of fall entailing an ‘as if’ assumption about vacuums can easily be interpreted in a realist manner as a negligibility assumption. But as Mäki himself points out, when it comes to making the analogical inference to profit maximising entrepreneurs, F53 fudges the analogy and opens the backdoor to instrumentalism (cf. Reiss, 2010).

23. The Shorter Oxford English Dictionary offers the following: comprehending or apprehending the meaning or importance of something; comprehending or realising a fact; knowing by information received; accepting as true or existent; accepting or believing as a fact without knowledge; believing or assuming on account of information received or by inference; interpreting or viewing something in a certain way.

24. It is worth recalling that the ‘Logical completeness and consistency’ (Friedman, 1953, p. 10) of theories and their implications is referred to immediately after F53 defines ‘fruitfulness’. It should also be recalled that with respect to testing the worth of a theory, F53 tells us that ‘further testing [of a theory] involves deducing from it new facts capable of being observed but not previously known and checking these deduced facts against additional empirical evidence’ (Friedman, 1953, p. 13, emphasis added). Clearly, the implications of a theory – its predictions – are logical derivations.