Gottfried Haberler, Current Research in Business Cycles—Discussion (1949)

Gottfried Haberler’s contribution to this collective discussion assesses the state of business-cycle research through responses to papers by Gordon and Koopmans. It moves from a critical survey of quantitative methods to a diagnosis of theoretical proliferation, then proposes a more modest basis for cooperation between theory and empirical investigation. Its central contention is that the failure to establish a comprehensive explanation of business cycles should not obscure substantial agreement about their characteristics, immediate causes, and possible mitigation. Haberler challenges both excessive pessimism about existing knowledge and excessive confidence in particular research techniques.

Gordon’s inventory of unanswered questions about the interwar period makes ignorance appear almost total. Haberler calls this partly an “optical illusion”: more precise questions, accompanied by an inventory of established findings, would produce a less discouraging picture. Koopmans’s account of the ongoing revision of the Cowles Commission approach offers little reassurance. Haberler doubts whether sophisticated econometric machinery has yet yielded results commensurate with its methodological ambitions; its deployment against the simplest acceleration principle seems disproportionate to the intellectual task.

Appropriately enough, research in the field of business cycles seems to move in cycles—or should we say in fashions?

This observation organizes a compressed history of research programmes. Business barometers lost credibility but left valuable data and methods of time-series analysis. The consumption-function approach, founded on oversimplified Keynesianism, produced failed forecasts of immediate postwar depression, yet also contributed something durable to economic knowledge. Haberler thus distinguishes the failure of a programme’s claims from the value of its residual findings. The National Bureau’s work under Mitchell and Burns has greater staying power, although partly through caution and silence on urgent issues. Frickey supplies firm descriptive results without answering many policy questions. Leontief’s fashionable input-output analysis is primarily structural and long-run, with possibilities for cyclical analysis that remain unproven. Gordon’s quantitative-historical proposal receives a warmer response because it continues a practice already exemplified by major cycle theorists: combining statistical evidence with historical knowledge that cannot yet be quantified.

The theoretical difficulty is different but related. Constructing models that generate oscillations has become easy; selecting models that illuminate actual fluctuations has not.

However, the more models we have, the less we seem to know of the real business cycle.

The paradox concerns explanatory discrimination, not the uselessness of models. Lags and initial conditions can produce regular, damped, or explosive movements, but this mathematical versatility does not determine which mechanism operated historically. Expectations complicate matters further, ending the prospect of a straightforwardly mechanical theory. Haberler rejects any return to an earlier simplicity: acquired knowledge of possible mechanisms cannot simply be discarded.

We can only sift them, and reject many as unrealistic, on the basis of empirical (which is by no means purely statistical) evidence.

The parenthesis is crucial. Empirical testing must draw on historical interpretation as well as statistical series, and it promises laborious discrimination rather than rapid, decisive answers. Haberler’s position combines respect for theoretical advances with insistence on their evidential selection.

He then asks whether agreement about depression remedies implies some knowledge of causes. Medicine offers an apparent counterexample: a cure can be discovered experimentally before its operation is understood. But economists cannot repeatedly subject whole economies to controlled trials. Historical events and government policies supply imperfect experiments, while expectations may make an economy under deliberate experimentation behave differently from one left alone. His footnotes qualify this distinction: experimental and nonexperimental sciences differ in degree, and the absence of economic laboratory experiments does not justify an anti-empirical methodology. Policy agreement nevertheless suggests shared causal hypotheses rather than remedies discovered independently of explanation.

The constructive conclusion identifies the appropriate scale of those hypotheses. Haberler doubts that even institutionally similar historical cycles must admit one encompassing theory. Investigators should not begin by presuming that a single model explains them all. They should instead develop the comparatively ordinary propositions on which researchers already converge: closely matching chronologies of short cycles, common descriptions of movements in output and prices, and aggregate expenditure fluctuations as a generally accepted proximate cause of changes in output, employment, and prices. This last claim is empirical rather than tautological, with possible exceptions. Investment fluctuates more than consumption, and durable goods more than perishables. The directions of effects associated with saving, budgets, credit, interest rates, and foreign trade are also reasonably understood, although their magnitudes remain disputed.

These types of hypotheses (theories) ought to be more fully exploited and utilized by empirical business cycle investigators.

This distinction between comprehensive cycle theories and limited causal propositions resolves Haberler’s qualified defence of the National Bureau. Mitchell and his collaborators were right not to organize inquiry around one complete cycle model, but they could use modest theoretical findings more fully. Replacing the elusive “reference cycle” with income and employment would sharpen their focus and facilitate constructive exchange with other researchers. Haberler explicitly grounds that proposal in pre-Keynesian work as well as later developments. The contribution’s enduring methodological point is that empirical research need neither embrace a universal model nor dispense with theory: cumulative progress depends on historically tested, limited explanations and clearly defined objects of inquiry.