Empirical properties of diversion ratios

The diversion ratio for products j and k is the fraction of consumers who leave product j after a price increase and switch to product k. Theoretically, it is expressed as the ratio of demand derivatives from a multi-product firm's Bertrand-Nash first-order condition. In practice, diversion rat...

Full description

Saved in:
Bibliographic Details
Published inThe Rand journal of economics Vol. 52; no. 4; pp. 693 - 726
Main Authors Conlon, Christopher, Mortimer, Julie Holland
Format Journal Article
LanguageEnglish
Published Santa Monica Wiley Subscription Services, Inc 01.12.2021
Rand Corporation
Subjects
Consumers
Online AccessGet full text

Cover

More Information
Summary:The diversion ratio for products j and k is the fraction of consumers who leave product j after a price increase and switch to product k. Theoretically, it is expressed as the ratio of demand derivatives from a multi-product firm's Bertrand-Nash first-order condition. In practice, diversion ratios are also measured from second-choice data or customer-switching surveys. We establish a LATE interpretation of diversion ratios, and show how diversion ratios are obtained from different interventions (price, quality, or assortment changes) and how those measures relate to one another and to underlying properties of demand.
Bibliography:We thank Joe Farrell, Phil Haile, Dennis Carlton, Ken Hendricks, Fiona Scott Morton, Aviv Nevo, Bill Rogerson, Ralph Winter, Dan O'Brien, and seminar participants at: Charles River Associates, Columbia University, the Department of Justice, NYU Stern, the University of Western Ontario, the University of Mannheim, the Bates White Antitrust Conference, the Behavioral Game Theory Workshop, the FTC Microeconomics Conference, ‘I.O. Fest’ at UC Berkeley, the Dartmouth Winter IO Conference, the NBER Summer Institute, the Penn State ‐ Cornell Econometrics and IO Conference, and the antitrust conference at the University of British Columbia, for comments. Chitra Marti provided exceptional research assistance. Special thanks to Jeff Gortmaker for custom updates to PyBLP. Any remaining errors are our own.
ISSN:0741-6261
1756-2171
DOI:10.1111/1756-2171.12388