Generalizing heuristic switching models

• Verfasst von: Galanis, Giorgos [VerfasserIn]
• Verfasst von: Kollias, Iraklis [VerfasserIn]
• Verfasst von: Leventidis, Ioanis [VerfasserIn]
• Verfasst von: Lustenhouwer, Joep [VerfasserIn]
• Titel: Generalizing heuristic switching models
• Verf.angabe: Giorgos Galanis, Iraklis Kollias, Ioanis Leventidis, Joep Lustenhouwer
• Verlagsort: Heidelberg
• Verlag: Universitätsbibliothek Heidelberg
• E-Jahr: 2022
• Jahr: 11 Mai 2022
• Umfang: 1 Online-Ressource (51 Seiten)
• Illustrationen: Illustrationen
• Gesamttitel/Reihe: AWI discussion paper series ; no. 715 (April 2022)
• DOI: doi:10.11588/heidok.00031617
• URN: urn:nbn:de:bsz:16-heidok-316179
• Datenträger: Online-Ressource
• Sprache: eng
• Form-SW: Graue Literatur
• K10plus-PPN: 180166854X
• K10plus-PPN:
  Universitätsbibliothek
• Lokale URL UB: Zum Volltext

Abstract

The growing literature in behavioral finance and macroeconomics that uses dynamic discrete choice models has overwhelmingly assumed that individual choices are made on the basis of a logit framework. While this assumption allows for analytical tractability, it comes with a number of restrictions with regards to the economic environments it can represent. These restrictions are lifted if a probit framework is used instead. In this paper we compare the two approaches and show that, due to its ability to allow for correlations between the random part of different choice alternatives as well as random taste variation, the probit-based model can better fit actual choice data from an existing laboratory experiment, especially if there are more choice alternatives. On the other hand, for the case of two choice alternatives without random taste variation, the probit-based and logit-based models result in very similar dynamics. But even in that case, we find that important qualitative differences arise - in terms of an additional region of chaos - in the cobweb model of the seminal work of Brock and Hommes (1997). Our work highlights the usefulness of using the probit framework for extensions of existing theoretical models and as a way to better fit dynamic experimental or real world choice data.