Why macroeconomics does not supervene on microeconomics

Abstract

In recent years, the project of providing microeconomic foundations for macroeconomics has taken on new urgency. Some philosophers and economists have challenged the project, both for the way economists actually approach microfoundations and for more general antireductionist reasons. Reductionists and antireductionists alike, however, have taken it to be trivial that the macroeconomic facts are exhaustively determined by microeconomic ones. In this paper, I challenge this supposed triviality. I argue that macroeconomic properties do not even globally supervene on microeconomic ones. This is simply a consequence of the difference in the explanatory goals of the respective fields, which implicitly carve out the microeconomic property set in such a way that it underdetermines macroeconomic properties. It means, however, that microeconomics-based foundations for macroeconomics are inadequate in principle.

Keywords::

• methodological individualism
• macroeconomics
• methodology of economics
• economic models
• ontology

Notes

^1. General equilibrium theory had historically been so idealized that few lessons could be drawn about the relations among macrovariables in a real economy. Even Arrow and Hahn (1971) suggest that the key lesson of investigating the properties of equilibria is to provide assurance of the possibility of arriving at equilibrium in a complex system, rather than being useful for macroeconomic policy (cf. Hausman, 1992).

^2. This involves treating agents as an aggregate of homogeneous ‘representative’ individuals, which has properties averaged over the population. Without idealizing agents as a representative agent, the problem of aggregating individual preferences and consumption is often analytically intractable.

^3. The self-conscious employer of the representative agent model holds out the hope that even though it is possible for aggregated preferences to be ill-behaved, they will not be. A critic will regard representative agent models to be poor idealizations, but different skeptics may draw diametrically opposed conclusions from this. Some hold that it militates for a more sophisticated approach to microfoundations, and indeed, some recent work on DSGE models has begun to incorporate heterogeneous agents. Others hold that it is evidence that microfoundational projects are a pipe dream.

^4. In some cases this claim has been discussed at length, such as in teleofunctional accounts of biological properties, which deny that certain biological properties supervene on synchronic supervenience bases.

^5. Not everyone assumes it is trivial. Hoover explicitly argues for supervenience and against reducibility (Hoover, 1995, 2001a, 2001b, 2006). And as I discuss in Section 2, Reiss challenges it.

^6. Supervenience is not generally regarded as sufficient to capture the full ‘ontological dependence’ relation, as supervenience involves a modal covariation, while ontological dependence arguably involves essentiality rather than only necessity (cf. Fine, 1994). Still, modal covariation seems to be a necessary even if insufficient condition for ontological dependence.

^7. According to Kim's definition, A-properties supervene locally on B-properties just in case for every pair of objects x and y, if x and y are B-indiscernible then they are A-indiscernible (cf. Kim, 1984). This is also known as ‘individual supervenience,’ but that term is potentially confusing in the present context.

^8. There is a variety of interpretations of global supervenience, turning on how we are to understand the indiscernibility of worlds (cf. Sider, 1999; Bennett, 2004; Shagrir, 2002). But discriminating among these complicates the discussion, and all of them will fail for the same reason, in the present context.

^9. Some audiences encountering this material have been unclear about this point. The most widely used definitions of supervenience are in Kim (1984) and Kim (1987). McLaughlin and Bennett (2005) give an excellent overview of the variants of supervenience. Some people have wondered whether it is not events, rather than properties, that are typically taken to as the sets of supervening and subvening entities. In his early discussions of supervenience, Donald Davidson does in fact speak of the supervenience of events; but he glosses it in terms of property supervenience. Kim has likewise spoken of fact supervenience, but again glosses it in terms of property supervenience. The supervenience of objects is also typically cast in terms of property supervenience, for a given object N employing the identity property being N. Other people have worried about talk of properties, and are more comfortable speaking of economic predicates rather than properties. Since predicates are linguistic items, however, there are two ways of understanding what we might mean by the supervenience of predicates. The reasonable way again is as property supervenience, where we take predicates to refer to properties, and then understand the supervenience of a predicate set on another predicate set as the supervenience of the set of properties denoted by the first predicate set on the set of properties referred to by the second. If someone is genuinely interested in the supervenience of predicates themselves, then since predicates are linguistic items this will be a question for linguistics, rather than the relation between two domains of economics.

^10. Similarly, if we are interested, for instance, in assessing whether a biological property holding at time t supervenes on chemistry, then it becomes important to determine whether the facts about chemistry or the chemical properties in question include only the chemical properties of the world at t, or the chemical properties at the times preceding t as well. If biological properties are historical, as many philosophers of biology have taken them to be, then they clearly will require historical properties to be included in the supervenience base as well. Clarifying just what the chemical properties are is the greatest part of ascertaining whether biological properties do or do not supervene on them.

^11. Hoover has also responded to Reiss in a recent paper (Hoover, 2009), pointing out some different issues than I do here.

^12. It may be thought that the ‘causal exclusion’ argument, advanced by Kim (e.g., Kim, 1998, 2001) provides some reason to believe in this incompatibility, but this is not correct. The causal exclusion argument involves the incompatibility of five intuitive theses, including both a supervenience and an exclusion thesis. However, the four theses even apart from the supervenience thesis are already mutually inconsistent (cf. Kallestrup, 2006). Moreover, proponents of the argument diagnose it as implying the failure of one of the other four theses, since it is generally discussed in connection with mental causation, and few would deny the supervenience of the mental on the physical.

^13. Hoover points out that certain macroeconomic properties, such as price index levels, may be interpreted in multiple equally useful ways (Hoover, 2001a, 2009). This, however, is compatible with any precisification of such properties still supervening on microeconomic ones.

^14. Virchow was aware of this problem, but was committed to his cell theory. So he argued, implausibly, that the extracellular parts of the body were divided into ‘cell territories,’ over which cells had superintendence, much as Lake Superior is apportioned in part to the USA and in part to Canada.

^15. Udehn (2001, p. 247) criticizes Janssen along these lines, for his view that ‘households or firms may be regarded as “individuals” relative to the aggregate level of the economy as a whole’ (Janssen, 1993, p. 7).

^16. The idea of expanding the microeconomic properties to a bigger set S is just so as not to artificially restrict the individualistic properties to the intrinsic properties of individual people. But instead, to expand it to a set of properties that still count as individualistic but are not intrinsic. The problem with this generosity, however, is avoiding a kind of circularity. Despite wanting to admit nonintrinsic properties, we cannot define the superset S of ‘individualistic’ properties in terms of the close contact of individuals with ‘individualistic’ things. At an intuitive level, one would want to avoid institutions and other ‘macroeconomic entities.’ At the end of the day, these worries are a bit tangential to the argument against supervenience on S, since the counterexamples for supervenience remain counterexamples even if institutions are included in S. Still, it is a difficult issue. I am grateful to an anonymous referee for pressing this point.

^17. See, for instance, Erturk and Ozgur (2008).

^18. This property is not entirely independent of properties of individuals, but does not have to be, for it to fail to supervene globally on individualistic properties. An analogous case is facts about human anatomy: they are not entirely independent of facts about dermatology, and yet anatomical facts fail to supervene globally on dermatological ones.

^19. Cf. Fodor (1980) and the large subsequent literature.

^20. Christian List has pointed out to me that certain versions of game theory would admit ‘moves’ of nature as part of an account of these situations. What is admissible in game theory in general, however, is not necessarily part of an individual-centered microeconomic supervenience base. For instance, game theory models the strategic interaction among nations, governments, firms, and so on as freely as it does of individuals. In the case of the ‘natural move,’ the properties of game theoretic agents can involve extrinsic properties that are not direct causes affecting individuals. For the theorist wanting to use game theory as a way of delineating a supervenience base for microeconomics, it will not do to sanction just any extraindividualistic factor. Otherwise, she risks trivializing the distinction between macroeconomics and microeconomics. The game theorist must delineate an appropriate supervenience base, just as the other microfoundations theorists must. It may be possible to expand microeconomics to include such factors. My thanks to an anonymous referee for stressing this point.