Butterfly Economics: A New General Theory of Social and Economic Behavior
Michael J. Stutzer - Professor of Finance, University of Iowa
Published September 1, 2000 | September 2000 issue
By Paul Ormerod
Anyone browsing the shelves at book superstores has seen them: books with provocative titles, whose covers promise to revolutionize our understanding of important phenomena. Often, these books are written by authors who aren't famous academicians. As a result, readers must be especially careful to avoid being swept away by persuasive arguments that won't really stand up to proper scientific scrutiny.
Butterfly Economics was written by Paul Ormerod, who has been a visiting professor at the universities of London and Manchester, as well as a research head on the staff of the respected British weekly The Economist. As expected, the cover of the book promises that it contains "A New General Theory of Social and Economic Behavior." Of course, I read it with more than a modicum of skepticism. Yet the book is better than most in this genre, particularly at inducing the reader to think more deeply about the work of researchers and its practical significance.
The author discusses some social phenomena, including the fluctuations over time in the fractions of people who vote for a specific political party, who are criminals or who are married. But the most detailed analysis in the book is devoted to theories attempting to explain business cycle fluctuations in U.S. output.1 Hence I will devote the rest of my precious column-inches to the author's treatment of this subject.
Ormerod first examines the ability of a currently popular academic tool, called real business cycle theory (RBC), to explain fluctuations in output.2 This theory explains the fluctuations in output by utilizing mathematical optimization models for the dynamics of worker/consumers' labor, spending and saving decisions, and for the dynamics of firms' hiring, investment and output decisions. Like most theories in economics, RBC contains unobservable parameters, whose values must be specified (either by some estimation or calibration process) before it can be used to make quantitative predictions. The author justly credits this theory's ability to be parametrized in a way that correctly predicts the variability of U.S. output and the correlations of output fluctuations with other macro variables. But the author also cites studies of other researchers claiming that extant RBC models cannot properly account for some other statistical properties of U.S. output.3 Of course, continuing modifications of RBC theory will likely fix the problems mentioned in the book, but there will probably always be some other relevant statistical findings that cannot be jointly explained by any parametrization of the RBC models existing when they are found.
In place of RBC, the author proposes a much simpler statistical model of U.S. output. The model first equates each firm's future output growth to a weighted average of its own current output growth and a (firm-specific randomly shocked) index of all firms' separate sentiments about future growth prospects. Second, each firm's own sentiment about future growth is posited to be a different weighted average of its own past sentiment and -1 times the (firm-specific randomly shocked) average of firms' output growth. During an economic expansion, average output grows when the individual firms' outputs grow, so this negative effect eventually causes firms' sentiments to grow more pessimistic. The first equation then eventually causes firms' outputs to decline, and (loosely paraphrasing a former president of the United States) a recession results. Weighting and shock parameter values can be found so that total firm output fluctuations will match all the statistical findings described in the previous paragraph, in contrast to the RBC theory's ability to do so. But again, there will probably always be some other relevant statistical findings that the author's theory will not be able to explain.
So is this a better theory than RBC? This depends on the purpose of theory. If the sole purpose is to predict a bunch of previously known statistical findings, then a theory is successful when (its parameter values can be specified so that) it does so. For this purpose, it is certainly plausible that simpler theories, such as that put forth by the author, might outperform more elaborate theories like RBC.4
But another useful of purpose of theory is to provide a detailed explanation of the workings of the economy that causes those statistical findings, in the hope that this explanation will lead to additional useful predictions that would otherwise be missed. The explanations provided by the author, for the social and economic phenomena he discusses, are all based on two premises. One premise is that decision-makers are directly influenced by the behavior of their peers, for example, a firm's output growth "decision" was directly influenced by other firms' sentiments about future output growth. The other premise is that decision-makers employ relatively simple "rules of thumb" when making decisions based on these influences; for example, a firm's output growth decision was a simple weighted average of its own previous growth and a (randomly shocked) index of all firms' sentiments. In contrast, the explanations provided by RBC are based on two different premises. The first premise is that decision-makers are indirectly influenced by others, via their joint interaction that determines the directly influential market prices, wages and interest rates. The second premise is that decision-makers act as if they employed sophisticated optimization methods when making decisions, which may or may not produce the simple behavioral "rules of thumb" advocated by the author. The better theory is the one whose premises will eventually lead to the derivation of currently unknown additional predictions that will eventually be found to be in accord with statistical findings about them.5
More detailed structural theories, like RBC, are able to make more predictions about the effects of productivity-improving technological change on output growth. If these are found to be in accord with statistical findings, it will (in this sense) be a better theory than the simpler equation system cited by Ormerod. I venture the opinion that it is too early to say whether RBC or other analogous theories6 will withstand the coming onslaught of attacks made by those who will vary from either or both of its premises. But the ability to fund, publish and promote this alternative research is important to the continuing advancement of economics as a science, rather than a religious faith devoted to protecting a particular set of premises from outside challenges. Butterfly Economics succeeds in this dimension, if not in the more ambitious endeavor stated on its cover.
1 The author also discusses cross-country differences in economic growth and prosperity, and the centuries-old trend toward industry domination by large firms.
2 This theory is closely associated with several of my former colleagues in the Research Department at the Federal Reserve Bank of Minneapolis, but I will try to be fair in assessing the author's claims to have found a better alternative to it.
3 In particular, he cites studies claiming that RBC models do not account for the observed autocorrelation of output, that is, the fact that periods of relatively high or low output growth tend to be followed by periods of similar output growth, nor account for the tendency of output growth to irregularly cycle at intervals between two and seven years.
4 This possibility should not surprise Federal Reserve Bank of Minneapolis researchers, who throughout the 1980s argued that a theoretical vector autoregressive models made better statistical forecasts than did detailed structural economic models.
5 A seminal example of this occurred in physics, when Einstein's general theory of relativity challenged the reigning Newtonian theory of gravity. Much was made of the ability of Einstein's theory to explain the observed data about the perihelion of Mercurya problem that had stumped the extant Newtonian theory. But others (for example, Robert Dicke) succeeded in modifying Newton's theory to do the same thing. The current widespread acceptance of Einstein's theory was earned by theorists' subsequent use of it to derive additional (often counterintuitive) predictions about nature that were eventually tested and confirmed. Suitably modified Newtonian theories can account for some, but not all, of these findings. Nor are they as likely to lead to further predictions that will be confirmed by data.
6 An unstated third premise, shared by most theories, is that unpredictable random shocks are present. Much of the debate about RBC to date has focused on whether or not the source of these shocks should be attributed to the actions of monetary and fiscal authorities, rather than whether or not its market-oriented and optimization-based premises are wrong.