Previous articleNext article FreeCommentEnrico SpolaoreEnrico SpolaoreTufts University and NBER Search for more articles by this author PDFPDF PLUSFull Text Add to favoritesDownload CitationTrack CitationsPermissionsReprints Share onFacebookTwitterLinked InRedditEmailQR Code SectionsMoreGilles has done it again. He has written a very interesting, provocative paper on an important topic. The analysis is elegant, making clever use of a simple, crisp model. The paper also includes a brave attempt to bring the ideas to the data, using surveys of professional forecasters. The general question addressed in this paper is: How do macroeconomists’ preferences (ideology) affect what models they provide to policymakers and the private sector? The paper approaches this broader issue by investigating a more specific question: May an ideologically motivated macroeconomist willingly provide a biased model in order to affect macroeconomic policy? In sum, we could say that the question boils down to: How much can a lying, manipulating macroeconomist get away with, and laugh all his way to the (central) bank?Gilles does not allow the macroeconomist to provide any model he or she may like. Rather, Gilles assumes that a macroeconomist can only provide an “autocoherent” model. A model is autocoherent if it correctly predicts the moments of the observables when everybody uses the model in equilibrium. Autocoherence is closely connected to the concept of “self-confirming equilibrium” (Fudenberg and Levine 1993), but it’s a property of the model, not of the equilibrium. In this paper the lying macroeconomist “wins,” even though he faces such autocoherence constraint. That is, the macroeconomist can obtain his first-best policy outcome by providing an incorrect autocoherent model. Using Gilles’s terminology, the economist can act as a quasi-dictator. Sure, the economist may have to make “ideological” concessions about some parameter values, but those have no impact on the economist’s own utility. At the end, in this paper the lying macroeconomist gets what he wants, in terms of government policies. How come? Why can the macroeconomist achieve such first-best while providing an incorrect model? The reason is that he lives in a Macroeconomist’s Paradise. In such paradise, the macroeconomist is:1. omniscient—he is the only one who knows the true structural model2. inscrutable—nobody (neither the government nor the private sector) knows his true preferences 3. unique—the macroeconomist has full intellectual monopoly over models 4. necessary—people need the macroeconomist’s model in order to form expectations and to make policy decisions5. trusted—everybody believes the macroeconomist’s model is correct, even when it is not.6. free—the macroeconomist has enough degrees of freedom because the data do not allow observers to distinguish between alternative models (underidentification) The assumption of underidentification is key for the result. If the parameters of the model were uniquely identified from the moments of the observables, autocoherence would compel the omniscient macroeconomist to reveal their true value. When the parameters of the structural model are not identified, the macroeconomist can “lie” by telling the government false values of the structural parameters that still satisfy autocoherence. For this, it is essential that the optimal decision rules of the agents depend on the structural parameters, not just on the reduced form.Gilles’s analysis has several interesting implications. For instance, a “conservative” macroeconomist—defined as somebody who attaches a high cost to fiscal activism—will provide a model that underestimates the Keynesian multiplier but also overestimates the long-term inflationary impact of output expansion. The opposite holds for a “progressive” macroeconomist. In an extension, Gilles considers the case of a macroeconomist who shares the preferences of a government with credibility problems, and shows that such a macroeconomist will report the total Keynesian multiplier as if it were just the impact Keynesian multiplier. This intriguing and provocative paper raises a large number of interesting questions and issues. In the rest of my discussion, I will focus on four issues:1. The macroeconomist’s “influence through lying” is derived under extreme assumptions, which are somewhat in tension with each other. Is this a “possibility” result or an “impossibility” result?2. Is autocoherence a necessary and sufficient condition for credible modeling?3. What do we learn from the empirical evidence?4. What are the deeper mechanisms through which biases may affect economists’ advice? Are they purely political-economy mechanisms (conscious lies) or should we move toward a political psychology of unconscious biases?I. Possibility or Impossibility Result?Several assumptions behind “the Macroeconomist’s Paradise” are extreme. For example, Gilles recognizes that the assumption of an omniscient macroeconomist who knows the correct model “is to be taken as a simplifying assumption and a metaphor for the much more subtle ways in which ideological biases affect the design of theories in practice.” However, a key question remains: to what extent are these simplifying assumptions logically consistent with each other, and necessary for the results?It seems that there is some logical tension between different assumptions. For instance, how can the macroeconomist know the true model even though the model cannot be identified? Why do the government and the private agents trust the macroeconomist even though they do not know his preferences and are unable to verify whether he may lie? Insofar as one needs to live in a Macroeconomist’s Paradise in order to affect policy outcomes with biased models, the paper could be interpreted as a parable illustrating the limits of a macroeconomist’s potential intellectual influence via ideological bias. II. The Role of AutocoherenceAn important issue is whether autocoherence is a necessary and sufficient condition for credible macroeconomic modeling. For example, as briefly discussed in the paper, autocoherence may not be sufficient when people are willing to pay the (small) cost of experimenting away from the self-confirming equilibrium. Perhaps more importantly, autocoherence may not be necessary. That is, the government and the private sector may be willing to use the macroeconomist’s model even though it is not autocoherent. This is related to the broader epistemological issue of the intellectuals’ ability to influence people’s beliefs (a topic on which Gilles has done much interesting work). In this paper, Gilles writes: “Intellectuals cannot force people to believe anything they want. A model that predicts that 2 + 2 = 5 will soon be discredited and abandoned.” Okay, but wouldn’t the equilibrium outcome depend on the alternatives? Gilles assumes that people need a model to make decisions. But what if economists only supply nonautocoherent models? Some nonautocoherent models may be better than others, and could be used as proxies. If model A predicts 2 + 2 = –100, while model B predicts 2 + 2 = 5, model B may work pretty well against the evidence, and survive. Hence, is autocoherence really a necessary property for a viable model? This line of reasoning brings us to broader question whether autocoherence can be derived as an endogenous property in an intellectual market. Such an issue is especially important for this paper, where the macroeconomist holds an intellectual monopoly. If the macroeconomist is the only one who can provide any model, why does he need to provide an autocoherent model? It might be useful to model the alternative—the “default” set of beliefs and policies that would be followed if the government were to abandon the macroeconomist’s model. More broadly, a natural extension of this paper would be to model explicitly both demand and supply of models in an “intellectual market,” and obtain equilibrium properties of the supplied models, such as autocoherence, endogenously from market conditions.III. Empirical Evidence The paper discusses some intriguing empirical evidence. The discussion of the empirics includes comments on findings by Fuchs, Krueger, and Poterba (1998), and an original empirical analysis of professional forecasters’ surveys. In his discussion of Fuchs, Krueger, and Poterba (1998), Gilles points out that there is a correlation between economists’ values and their beliefs about structural parameters. In Gilles’s interpretation, ideological values may have affected the economists’ “official beliefs” about parameters. However, the opposite may also be true: if I have certain beliefs about the economy, I am also more likely to adopt certain ideological views consistent with my beliefs. In general, values and beliefs about parameters are likely to be determined simultaneously.The second part of the empirical analysis is a creative attempt to bring the ideas of this paper to the data by estimating a “pseudo-model” linking forecasts for GDP growth, inflation, and federal government expenditure growth. This exercise shows the conceptual difficulty of “testing” for autocoherence (as Gilles writes: “the proper autocoherence conditions involve variables that are not observed”). Overall, the indirect evidence is not very supportive of the “autocoherence” implications from the theory. In particular, forecasters who believe in a larger Keynesian multiplier also believe in a larger (more positive) response of output to inflation. This analysis raises intriguing questions about the goals of professional forecasters. Should we interpret the findings as implying that forecasters want to influence policy using biased models? Or wouldn’t it be more plausible to think that they may just make errors that tend to go in the same direction? One is reminded of Napoleon’s razor (also known as Heinlein’s razor): “Never attribute to malice that which is adequately explained by incompetence.” In addition to the empirical evidence presented in this paper, there are several other possible areas of application of the paper’s ideas and insights. A promising area could be the study of fiscal multipliers using structural vector autoregressions (SVARs). As recently pointed out by Dario Caldara (2011): “The appeal of SVARs is that they control for endogenous movements in fiscal policies by only imposing a minimal set of assumptions, known as identification schemes. Yet, despite their simple structure and the use of similar data, studies employing SVARs document fiscal multipliers that are spread over a broad range of values.” In particular, as shown by Caldara (2011, 2), Blanchard and Perotti (2002), and Mountford and Uhlig (2009), identification schemes imply output elasticities of tax revenue equal to 2.3 and 3.3, respectively. Sign restrictions on impulse response functions imply output elasticities of tax revenue between 0 and 16. In general, different restrictions on the output elasticity of tax revenue bring about a large dispersion in the estimates of tax multipliers. In principle, consistently with the logic of Gilles Saint-Paul’s analysis, different economists could choose different restrictions on the output elasticity of tax revenues in order to obtain results that they consciously or unconsciously prefer for ideological reasons. Of course, this is only a logic possibility, and I am not implying that any of the specific results mentioned earlier are the outcome of explicit or implicit ideological biases! These observations lead us to the final, crucial issue, explained in the next section.IV. Ideological Biases: Motivation and MechanismsIf macroeconomists are indeed ideologically biased, as suggested in this paper, is that in order to consciously influence others? Or do macroeconomists simply “deceive” themselves (well, ourselves) in good faith? Macroeconomists, like everybody else, may suffer from “excessive certitude” and confirmation bias—the tendency to pick out evidence that support their views and ignore the rest, holding on to beliefs in the face of contrary evidence. Therefore, an alternative approach could be a framework where macroeconomists do not know the true model, but have strong priors about some parameters of the model, which may be (a) consistent with some ideological preferences or objectives, and (b) pretty resistant to contrary evidence.Unconscious biases may indeed be there in order to win political arguments, as stressed by a growing literature in psychology. This view is sometimes called the “argumentative theory of reasoning.” For instance, Mercier and Sperber (2011, 57) write: [M]uch evidence shows that reasoning often leads to epistemic distortions and poor decisions. . . . Our hypothesis is that the function of reasoning is argumentative. It is to devise and evaluate arguments intended to persuade. Reasoning so conceived is adaptive given the exceptional dependence of humans on communication and their vulnerability to misinformation. A wide range of evidence in the psychology of reasoning and decision making can be reinterpreted and better explained in the light of this hypothesis. They add: Skilled arguers . . . are not after the truth but after arguments supporting their views. This explains the notorious confirmation bias. This bias is apparent not only when people are actually arguing, but also when they are reasoning proactively from the perspective of having to defend their opinions. Reasoning so motivated can distort evaluations and attitudes and allow erroneous beliefs to persist. (57)A recent article in the New York Times (Cohen 2011) summarized these ideas as follows: “Lack of logic and other supposed flaws that pollute the stream of reason are instead social adaptations that enable one group to persuade (and defeat) another. Certitude works, however sharply it may depart from the truth. The idea is the brainchild of French cognitive social scientists.” The same New York Times article also reported about some scholars’ strong disagreement with this view: “Darcia Narvaez, an associate professor of psychology at the University of Notre Dame . . . said this theory ‘fits into evolutionary psychology mainstream thinking at the moment, that everything we do is motivated by selfishness and manipulating others, which is, in my view, crazy.’” In contrast, we economists do not usually find pervasive selfishness and manipulation so crazy. Gilles’s excellent paper is a very useful and thought-provoking step toward applying the implications of economists’ more cynical views about human nature and motivation to our own production of macroeconomic thinking.NotesFor acknowledgments, sources of research support, and disclosure of the author’s material financial relationships, if any, please see http: // www.nber.org / chapters / c12497.ack.ReferencesBlanchard, Olivier, and Roberto Perotti. 2002. “An Empirical Characterization of the Dynamic Effects of Changes in Government Spending and Taxes on Output.” Quarterly Journal of Economics 117:1329–68.First citation in articleGoogle ScholarCaldara, Dario. 2011. “The Analytics of SVARs: A Unified Framework to Measure Fiscal Multipliers.” Institute for International Economic Studies, Stockholm University.First citation in articleGoogle ScholarCohen, Patricia. 2011. “Reason Seen More as Weapon Than Path to Truth.” New York Times, June 14.First citation in articleGoogle ScholarFuchs, Victor, Alan Krueger, and James M. Porterba. 1998. “Economists’ Views About Parameters, Values, and Policies: Survey Results in Labor and Public Economics.” Journal of Economic Literature 36 (3): 1387–425.First citation in articleGoogle ScholarFudenberg, Drew, and David K. Levine. 1993. “Self-Confirming Equilibrium.” Econometrica 61:523–45.First citation in articleGoogle ScholarMercier, Hugo, and Dan Sperber. 2011. “Why Do Humans Reason? Arguments for an Argumentative Theory.” Behavioral and Brain Sciences 34:57–111.First citation in articleGoogle ScholarMountford, Andrew, and Harald Uhlig. 2009. “What Are the Effects of Fiscal Policy Shocks?” Journal of Applied Econometrics 24: 960–92.First citation in articleGoogle Scholar Previous articleNext article DetailsFiguresReferencesCited by Volume 8, Number 12012 Article DOIhttps://doi.org/10.1086/663663 Views: 99 © 2012 by the National Bureau of Economic ResearchPDF download Crossref reports no articles citing this article.