« Back to Results

Identification and Inference in Limited Attention Models

Paper Session

Friday, Jan. 4, 2019 2:30 PM - 4:30 PM

Atlanta Marriott Marquis, International C
Hosted By: American Economic Association
  • Chair: Francesca Molinari, Cornell University

Discrete Choice Under Risk with Limited Consideration

Levon Barseghyan
,
Cornell University
Francesca Molinari
,
Cornell University
Matthew Thirkettle
,
Cornell University

Abstract

We propose an alternative to the additive-noise random utility model (e.g., conditional logit model, mixed logit model) that allows for decision makers with limited consideration of the feasible set of alternatives. The model we put forward is easy to implement and remains computationally tractable even when the feasible set is very large. We apply the proposed method to estimate household preferences over risky alternatives. We build a semi-parametric model that incorporates (A) a standard preference relation (e.g. expected utility theory) over the considered alternatives with non-parametric unobserved heterogeneity in preferences; and (B) a probabilistic mechanism which determines the likelihood that a decision maker considers a given option in the feasible set. Our model overcomes the shortcomings of the Luce-McFadden framework emphasized by Apesteguia and Ballester (2018, Journal of Political Economy) and can match several patterns in the data that the mixed logit and other related models cannot. We prove that our model is identified and we provide a maximum likelihood based estimator. We apply our method to estimate risk preferences from a large dataset on households’ property purchases.

​A Random Attention Model​

Yusufcan Masatlioglu
,
University of Maryland
Matias Cattaneo
,
University of Michigan
Elchin Suleymanov
,
University of Michigan
Xinwei Ma
,
University of Michigan

Abstract

We introduce a Random Attention Model (RAM) allowing for a large class of stochastic consideration maps in the context of an otherwise canonical limited attention model for decision theory. The model relies on a new restriction on the unobserved, possibly stochastic consideration map, termed Monotonic Attention​,​ which is intuitive and nests many recent contributions in the literature on limited attention. We develop revealed preference theory within RAM and obtain precise testable implications for observable choice probabilities. Using these results, we show that a set (possibly a singleton) of strict preference orderings compatible with RAM is identifiable from the decision maker's choice probabilities, and establish a representation of this identified set of unobserved preferences as a collection of inequality constrains on her choice probabilities. Given this nonparametric identication result, we develop uniformly valid inference methods for the (partially) identiable preferences. We showcase the performance of our proposed econometric methods using simulations, and provide general-purpose software implementation of our estimation and inference results in the R software package ramchoice . Our proposed econometric methods are computationally very fast to implement.

Inferring Cognitive Heterogeneity from Aggregate Choices

Paola Manzini
,
Univeristy of Sussex
Marco Mariotti
,
Queen Mary University of London
Valentino Dardanoni
,
University of Palermo
Chris Tyson
,
Queen Mary University of London

Abstract

We study the problem of identifying the distribution of cognitive characteristics in a population of agents when only aggregate choice behavior from a single menu is observable. Focusing on two models of limited attention, we demonstrate that both “consideration probability” and “consideration capacity” distributions are substantially identified by aggregate choice shares when tastes are homogeneous. We then show how our methodology can be extended to allow for heterogeneous tastes, and suggest how the attention models can be embedded in an econometric specification of the inference problem. Finally, we conduct Monte Carlo simulations of both models and use our results to recover the true parameters.

What Do Consumers Consider Before They Choose? Identification from Asymmetric Demand Responses

Abi Adams
,
University of Oxford
Jason Abaluck
,
Yale University

Abstract

Consideration set models relax the assumption that consumers are aware of all available options. Thus far, identification arguments for these models have relied either on auxiliary data on what options were considered or on instruments excluded from consideration or utility. In a discrete choice framework subsuming logit, probit and random coefficients models, we prove that utility and consideration set probabilities can be separately identified without these data intensive methods. In full-consideration models, choice probabilities satisfy a symmetry property analogous to Slutsky symmetry in continuous choice models. This symmetry breaks down in consideration set models when changes in characteristics perturb consideration, and we show that consideration probabilities are constructively identified from the resulting asymmetries. In a lab experiment, we recover preferences and consideration probabilities using only data on which items were ultimately chosen, and we apply the model to study hotel choices on Expedia.com and insurance choices in Medicare Part D.
Discussant(s)
Aluma Dembo
,
University of Oxford
Jason Abaluck
,
Yale University
Arthur Lewbel
,
Boston College
Richard Blundell
,
University College London
JEL Classifications
  • C4 - Econometric and Statistical Methods: Special Topics
  • D9 - Micro-Based Behavioral Economics