Methods and Problems in Business Cycle Theory

Lucas opens by proposing that a theory should be understood not as a set of assertions about the behavior of the actual economy but as an explicit set of instructions for building a parallel, mechanical, “analogue” imitation economy, whose usefulness is judged by how well its constructed behavior answers the particular questions put to it rather than by its surface realism (Section 1, p. 697). On this view, insistence on “realism” in a model actually undermines its usefulness, because any model well-articulated enough to give clear answers will necessarily be abstract and, in Lucas’s words, “patently unreal,” and progress comes not from more verbal, “realistic” descriptions of the economy but from better abstract analogue models (Section 1, p. 697).

Lucas identifies two main outside sources of theoretical development: technical advances (in mathematical methods and computational capacity) that expand what kinds of analogue economies can be built, and changes in the substantive questions economists and the public want theory to answer (Section 1, pp. 697-698). He emphasizes the first as the most important force in economics generally, arguing that neglect of this technical dimension in the history of economic thought contributes to the mistaken impression that all economic ideas were anticipated by earlier writers such as Marshall, when in fact the capacity for detailed, explicit modeling of markets Marshall could only informally conjecture about is itself a recent achievement (Section 1, p. 697).

Mitchell’s early twentieth-century research sought an empirical definition of the business cycle by systematically excluding movements in economic time series that seemed explicable by existing theory (long-run growth trends and market-specific supply and demand shifts), using a wide variety of data-summarizing techniques across a broad collection of series (Section 2, p. 698). Lucas credits Mitchell with documenting a durable central finding – the broad similarity of peacetime business cycles to one another, once differences in duration are controlled for, in the sense that each cycle exhibits roughly the same pattern of co-movement among variables – a regularity so long taken for granted that its remarkable, non-obvious character is easy to forget (Section 2, pp. 698-699).

Lucas reads Keynes’s Treatise on Money as attempting to reconcile a quantity-theoretic account of nominal prices with a neoclassical, real-side account of output and employment along a secular trend, and judges its technical apparatus – built around a set of accounting identities Keynes calls “fundamental equations” – as intelligent in its underlying ideas but insufficiently equipped, by Keynes or his contemporaries, to move the analysis to a sharper level (Section 2, p. 699). He argues that the onset of the Great Depression did not improve Keynes’s analytical tools but instead let him reformulate the problem itself, from explaining a recurring pattern of booms and depressions to explaining a single, persistent shortfall of output and employment – a simpler problem that could be attacked, in the General Theory, by dropping the difficult task of modeling the labor-supply side altogether (Section 2, p. 700).

Lucas credits Paul Samuelson’s Foundations of Economic Analysis with supplying the main technical ingredient of the “neoclassical synthesis” – a mathematically explicit static general-equilibrium system in which households and firms jointly solve explicit maximization problems taking prices as parametrically given – and Samuelson’s separate dynamic model of price adjustment, in which prices change in each market according to that market’s “excess demand,” as the second ingredient that let a wide variety of out-of-equilibrium paths be made consistent with eventual return to that static equilibrium (Section 3, pp. 701-702). It was precisely the introduction of these additional, free parameters describing the speed of disequilibrium adjustment – parameters not needed to describe tastes and technology alone – that let the synthesis reproduce Keynesian-looking cyclical behavior around a Walrasian equilibrium core, and that made the resulting framework flexible enough to accommodate a wide variety of policy views as special cases (Section 3, p. 702).

Lucas argues that when James Tobin and Franco Modigliani summarize the “central propositions” of Keynesian economics – Tobin’s claim that prices and wages respond only slowly to excess demand or supply, Modigliani’s characterization of a rival econometric model of declining employment as attributable only to an implausible “attack of laziness” by workers – both economists take the correctness of their preferred neoclassical-synthesis framework as a given starting point and use it as a yardstick for dismissing alternative frameworks, rather than subjecting that framework itself to comparably serious scrutiny (Section 3, pp. 703-704). He suggests this kind of argument, in which competing econometric models are judged chiefly by whether they reproduce the incumbent model’s simulated results, does not actually advance the underlying debate about which framework is correct (Section 3, p. 704).

Lucas identifies Milton Friedman’s presidential address articulating the natural-rate hypothesis, and Edmund Phelps’s related search for microeconomic foundations of wage and price setting, as jointly implying that “excess demand” was neither necessary nor sufficient for price or wage inflation, and that essentially any average inflation rate was compatible with any level of unemployment – a conclusion that conflicted with the output-inflation trade-off central to neoclassical-synthesis models (Section 4, p. 705). Formalizing this hypothesis exposed that conventional ways of modeling how expectations are formed were inadequate, and John Muth’s rational-expectations hypothesis – originally developed for a different purpose – turned out to be the natural way to complete the Friedman-Phelps argument, with subsequent research revealing how far-reaching, and subversive of the neoclassical synthesis’s core presumptions, that hypothesis actually is (Section 4, p. 705).

Building on Hicks’s reinterpretation of dynamic choice as choice among dated goods, Lucas describes how Kenneth Arrow and Gerard Debreu extended static general-equilibrium theory to incorporate uncertainty by indexing commodities not just by delivery date but also by the state of nature in which delivery occurs, turning a competitive equilibrium into a “contingent-claim” equilibrium over all such state-contingent goods (Section 5, p. 707). Lucas connects this formalism to rational expectations by noting that Muth’s hypothesis follows from applying the general principle that competitive equilibrium leaves no systematic, unexploited rents to the particular case of price expectations, so that in a contingent-claim equilibrium, expectations cannot systematically differ from the probability distributions implied by that same equilibrium (Section 5, p. 707).

Lucas uses a thought experiment in which a single hungry monkey’s reaction to a thrown banana is fairly predictable from prior knowledge of monkey behavior, but the reaction of five hungry monkeys to a single banana thrown into their cage is a qualitatively different, much harder problem, because it also requires a theory of how the group interacts, not just of individual preferences and technology (Section 6, pp. 710-711). Adding competition – letting the monkeys trade pieces of banana for a fixed amount of mutual back-scratching – restores predictability of the group outcome using only the hypothesis of competitive equilibrium, with no additional free parameters describing the interaction itself; Lucas argues the same logic applies to modeling aggregate wage and employment outcomes, where a competitive-equilibrium model needs only parameters describing preferences and technology, in contrast to a Phillips-curve-style model that needs an additional, separately unobservable parameter describing the speed of an implicit auctioneer’s wage adjustment (Section 6, pp. 710-712).

Lucas argues the equilibrium model’s key advantage is that its parameters, describing the degree of intertemporal substitutability of labor and leisure, can in principle be estimated independently from individual- and household-level panel and census data, not only from aggregate time series, whereas the Phillips-curve model’s auctioneer-adjustment-speed parameter can be inferred only by observing the whole system in operation and therefore can never be understood in the same disaggregated, “microeconomic-foundations” sense (Section 6, pp. 711-712). He is careful to note this is a claim about hopes for future research rather than accomplished fact, and to flag that whether one should ultimately prefer a model with no free adjustment parameters over one with one or two such parameters is not yet settled by the evidence available at the time of writing (Section 6, p. 712).

Lucas states the task, restated more bluntly and operationally, as writing a “FORTRAN program” that accepts specific economic policy rules as input and generates as output statistics describing the operating characteristics of the resulting time series – for example, predicting what average unemployment rate would have prevailed in the postwar United States under a fixed 4-percent annual money-growth rule – while cautioning that confidence in such a program’s component parts must come from independently documented evidence on individual behavior, not merely from the program’s ability to fit the very aggregate data it is meant to explain (Section 6, pp. 709-710).

In his concluding remarks, Lucas frames the neoclassical synthesis itself as a reasonable, historically situated compromise between what economists wanted to know about business cycles and what the analytical methods available at the time made possible to know, rather than as a body of doctrine to be defended or attacked as such (Section 7, p. 713). He argues that the same standard should be applied going forward: newer equilibrium models of the business cycle should be judged by whether they expand the questions that can be productively asked, and that nothing would be more damaging to the constructive use of these newer methods than treating the categories and constructs produced by the older compromise as fixed constraints on how business cycles can be thought about today (Section 7, p. 713).