To make high-quality research more accessible and easier to explore.

Fields:
6 results ✕ Clear filters

Inductive Inference: An Axiomatic Approach

Econometrica 2003 71(1), 1-26 open access
A predictor is asked to rank eventualities according to their plausibility, based on past cases. We assume that she can form a ranking given any memory that consists of repetitions of past cases. Mild consistency requirements on these rankings imply that they have a numerical representation via a matrix assigning numbers to eventualitycase pairs, as follows. A memory is identified with a vector, counting the number of repetitions of each case. Multiplication of the matrix by a memory vector yields a numerical representation of the ordinal plausibility ranking given that memory. Interpreting this result for the ranking of theories or hypotheses, rather than of specific eventualities, it is shown that one may ascribe to the predictor subjective conditional probabilities of cases given theories, such that her rankings of theories agree with their likelihood functions. 1 Introduction It is well known that inductive inference is not logically valid. As David Hume (1748) put it, "... The co...

Empirical Similarity

The Review of Economics and Statistics 2006 88(3), 433-444 open access
An agent is asked to assess a real-valued variable Yp based on certain characteristics Xp = (Xp1, …, Xpm), and on a database consisting of Xi1, … Xim, Yi) for i = 1, …, n. A possible approach to combine past observations of X and Y with the current values of X to generate an assessment of Y is similarity-weighted averaging. It suggests that the predicted value of Y, Ȳps, be the weighted average of all previously observed values Yi, where the weight of Yi for every i = 1, …, n, is the similarity between the vector Xp1, …, Xpm, associated with Yp, and the previously observed vector, Xi1, …, Xim. We axiomatize this rule. We assume that, given every database, a predictor has a ranking over possible values, and we show that certain reasonable conditions on these rankings imply that they are determined by the proximity to a similarity-weighted average for a certain similarity function. The axiomatization does not suggest a particular similarity function, or even a particular form of this function. We therefore proceed to suggest that the similarity function be estimated from past observations.We develop tools of statistical inference for parametric estimation of the similarity function, for the case of a continuous as well as a discrete variable. Finally, we discuss the relationship of the proposed method to other methods of estimation and prediction.

Utilitarian Aggregation of Beliefs and Tastes

Journal of Political Economy 2004 112(4), 932-938 open access
Fifty years ago, Harsanyi published the first of his seminal two papers on utilitarianism. His results were derived within the von Neumann Morgenstern expected utility theory. A year later, Savage incorporated subjective probability into expected utility theory in his famous book. In this note we extend Harsanyi’s utilitarianism to Savage’s framework. We show that a Pareto condition implies utilitarian aggregation: both society’s utility function and its probability measure are linear combinations of those of the individuals. This conclusion contrasts the impossibility of reconciling a Pareto condition and linear aggregation of beliefs and tastes, that was noted by several authors. We argue that the indiscriminate Pareto condition considered by these authors is not compelling. Society should not necessarily endorse a unanimous choice when it is based on contradictory beliefs. Restricting the Pareto condition to choices that only involve identical beliefs allows the extension of Harsanyi’s result to Savage’s framework.

Fact-Free Learning

American Economic Review 2005 95(5), 1355-1368 open access
People may be surprised to notice certain regularities that hold in existing knowledge they have had for some time. That is, they may learn without getting new factual information. We argue that this can be partly explained by computational complexity. We show that, given a knowledge base, finding a small set of variables that obtain a certain value of R 2 is computationally hard, in the sense that this term is used in computer science. We discuss some of the implications of this result and of fact-free learning in general.

Objective and Subjective Rationality in a Multiple Prior Model

Econometrica 2010 78(2), 755-770 open access
A decision maker (DM) is characterized by two binary relations. The first reflects choices that are rational in an “objective” sense: the DM can convince others that she is right in making them. The second relation models choices that are rational in a “subjective” sense: the DM cannot be convinced that she is wrong in making them. In the context of decision under uncertainty, we propose axioms that the two notions of rationality might satisfy. These axioms allow a joint representation by a single set of prior probabilities and a single utility index. It is “objectively rational” to choose f in the presence of g if and only if the expected utility of f is at least as high as that of g given each and every prior in the set. It is “subjectively rational” to choose f rather than g if and only if the minimal expected utility of f (with respect to all priors in the set) is at least as high as that of g. In other words, the objective and subjective rationality relations admit, respectively, a representation à la Bewley (2002) and à la Gilboa and Schmeidler (1989). Our results thus provide a bridge between these two classic models, as well as a novel foundation for the latter.