Determinants of inflationary performance: Corporatist structures vs. central bank autonomy

Thanks go to John Aldrich, Robert Bates, William Bernhard, Richard Burdekin, Henry Chappel, Dudley Wallace and Thomas Willett for suggestions and to Ted Smith for computational assistance.     

Modeling and testing the diffusion of expectations: An EITM approach

This paper uses the empirical implications of theoretical models (EITM) framework to examine the consequences of the asymmetric diffusion of expectations. In the spirit of the traditional two-step flow model of communication, less-informed agents learn the expectations of more-informed agents. We find that when there is misinterpretation in the information acquisition process, a boomerang effect exists. In this equilibrium the less-informed agents’ forecasts confound those of more-informed agents. We apply the EITM approach to a key economic variable known to have a relation to economic fluctuations – inflation expectations. Using surveyed inflation expectations data for the period, 1978–2000, we find the boomerang effect exists. One implication of this finding pertains to economic policy and economic volatility: because policymakers have more information than the public, the boomerang effect can lead policymakers to make inaccurate forecasts of economic conditions and conduct erroneous policies which contribute to economic instability.     

A Framework for Unifying Formal and Empirical Analysis

An important disconnect exists between the current use of formal modeling and applied statistical analysis. In general, a lack of linkage between the two can produce statistically significant parameters of ambiguous origin that, in turn, fail to assist in falsifying theories and hypotheses. To address this scientific challenge, a framework for unification is proposed. Methodological unification leverages the mutually reinforcing properties of formal and applied statistical analysis to produce greater transparency in relating theory to test. This framework for methodological unification, or what has been referred to as the empirical implications of theoretical models (EITM), includes (1) connecting behavioral (formal) and applied statistical concepts, (2) developing behavioral (formal) and applied statistical analogues of these concepts, and (3) linking and evaluating the behavioral (formal) and applied statistical analogues. The elements of this EITM framework are illustrated with examples from voting behavior, macroeconomic policy and outcomes, and political turnout.     

Information diffusion within the electorate:: the asymmetric transmission of political–economic information

This paper examines the connection between information and expectations. Drawing upon Lazarsfeld et al.'s (Lazarsfeld, P.F., Berelson, B., Gaudet, H., 1944. The People's Choice. Duell, Sloan, and Pearce, New York) classic work on communication flows, we devise an asymmetric transmission model of expectation formation. This model assumes that the expectations of the less informed segments within the electorate are influenced by the expectations of more informed groups but that the opposite does not hold. We use educational differences as a proxy for information heterogeneity. A vector autoregression (VAR) analysis on monthly surveys of inflation expectations supports the asymmetric transmission hypothesis.     

Testing for Cointegrating Relationships with Near-Integrated Data

Testing theories about political change requires analysts tomake assumptions about the memory of their time series. Appliedanalyses are often based on inferences that time series areintegrated and cointegrated. Typically analyses rest on Dickey–Fullerpretests for unit roots and a test for cointegration based onthe Engle–Granger two-step method. We argue that thisapproach is not a good one and use Monte Carlo analysis to showthat these tests can lead analysts to conclude falsely thatthe data are cointegrated (or nearly cointegrated) when thedata are near-integrated and not cointegrating. Further, analystsare likely to conclude falsely that the relationship is notcointegrated when it is. We show how inferences are highly sensitiveto sample size and the signal-to-noise ratio in the data. Wesuggest three things. First, analysts should use the singleequation error correction test for cointegrating relationships;second, caution is in order in all cases where near-integrationis a reasonable alternative to unit roots; and third, analystsshould drop the language of cointegration in many cases andadopt single-equation error correction models when the theoryof error correction is relevant.