Consumer search with observational learning

• Author: Daniel Garcia,Sandro Shelegia
• Published date: 01 March 2018
• Date: 01 March 2018
• DOI: http://doi.org/10.1111/1756-2171.12224

RAND Journal of Economics

Vol.49, No. 1, Spring 2018

pp. 224–253

Consumer search with observational

learning

Daniel Garcia∗

and

Sandro Shelegia∗∗

This article studies observational learning in a consumer searchenvironment. Consumers observe

the purchasing decision of a predecessor with similar preferences. Consumers rationallyemulate

by initiating their search at the firm fromwhich their predecessor purchased, free-ridingon search

effort, and reacting less to price changes. Prices are nonmonotone in search costs and may be

as low as marginal costs. We discuss several extensions and show that the effect of emulation

on prices is stronger when (i) the number of firms increases, (ii) consumers’ first visits are more

elastic with respect to market shares, and (iii) prices are adjusted more frequently.

1. Introduction

Observational learning has been the focus of a large and important literature in economics

since the seminal contributions of Banerjee (1992) and Bikhchandani, Hirshleifer, and Welch

(1992). In the classical model, a sequence of individuals faces a simple decision problem under

uncertainty, and each individual observes the history of the decisions of her predecessors. As

argued by Banerjee (1992), this simple environment resembles the problem faced by consumers

in markets where previous consumers’ choices may be informative about the relative value

of different products or services, such as restaurants and consumer goods.1Search markets,

where consumers actively engage in costlyinfor mation gatheringabout different alternatives, are

prominent examples of such environments. Intuitively, consumers who are aware of each others’

∗University of Vienna; daniel.garcia@univie.ac.at.

∗∗Universitat Pompeu Fabra; sandro.shelegia@upf.edu.

We thank the Editor and three anonymous referees for their comments and suggestions. We also thank Jan Eeckhout,

Maarten Janssen, Marco Haan, Jose Luis Moraga-Gonz´

alez, Stephan Lauermann, and Andrew Rhodes for their comments,

as well as participants and colleagues at the 6th Consumer Search and Switching Workshopat Indiana University, JIE in

Barcelona, the MACCI Conference in Consumer Search in Bad Homburg, SAEe in Palma de Mallorca, RES Conference

in Brighton, UPF, University of Bern, PUC Chile, and Universidad Alberto Hurtado. Garcia gratefully acknowledges

financial support from the Hardegg Foundation. Shelegia acknowledges financial support from the Spanish Ministry of

Science and Innovation (MINECO-ECO2014-59225-P) and AGAUR (2014 SGR 694). All errors are to be attributed to

the authors only.

1These theoretical models have found extraordinary support in empirical work. Fora recent discussion on the vast

literature on social learning, see M¨

obius and Rosenblat (2014).

224 C2018, The RAND Corporation.

GARCIA AND SHELEGIA / 225

purchases are likely to have similar preferences and may use this information when they embark

on their own search for the best product. Suggestive of such behavior isa large body of evidence

in economics and marketing, to be detailed later, which shows that consumers’ decisions are

affected by the purchasing behavior of others. To the extent that such learning takes place in a

search market, it is natural to wonder how consumers search and make purchases with learning.

Consider, for example, a consumer searching for a new smartphone, who learns that a friend

purchased an iPhone. She is now more likely to begin her search at an Apple Store. Provided

that she is satisfied with the product she finds there, she may well reason that other alternatives

were considered by her friend and subsequently rejected, and as a result, she can safely buy an

iPhone. How does Apple and its competitors take this process into account when pricing their

smartphones? Are prices and market outcomes affected? Do consumers and/or firms benefit?

To the best of our knowledge,no prior work has addressed these issues. We attempt to do so

with a simple oligopoly model of search with heterogeneous products, in the spirit of Wolinsky

(1986) and Anderson and Renault (1999) (henceforth, ARW). In the simplest version of their

model, consumers are initially uninformed about their valuation of different varieties sold in the

market or the price charged for them but may learn these after engaging in costly sequential

search. Consumers search firms randomly and stop once their utility draw for a firm’s variety is

sufficiently high.

We depart from this model in two ways. First, we allow valuations for each variety to be

positively correlated across consumers. Second, we inform each consumer of the purchasing de-

cision of a single predecessor. Importantly, consumers do not observe whether their predecessors

visited other firms or the price they paid (i.e., learning is purely observational). In line with

our intuitions, observational learning in our model dramatically alters consumer search and firm

pricing. Because of the positive correlation, it is always optimal for consumers to begin their

search in the firm from which their predecessor bought. We refer to such behavior as emulation.

Once at the first firm, consumers continue to use the knowledge of the predecessor’s purchasing

behavior when evaluating whether or not to continue their search. This has two main effects.

First, consumers free-ride on the predecessor’s search effort in that they are willing to buy fewer

valuable products at the first firm, and therefore they search less than they would have without

observational learning. Second, consumers become less responsiveto price changes because these

affect market shares and, therefore, the relative informativeness of the predecessor’s purchase.

As we shall see, these two effects—collectively referred to as the learning effect—along with

emulation, crucially determine firms’ pricing, which is the focus of our study.2

We take a static approach to pricing and assume that firms set prices once and for all.

Although dynamic pricing is clearly an important consideration in the presence of observational

learning, we view our contribution as a first step towardthis wider research goal. Our static model

is particularly useful when considering introductory pricing or environments in which learning

occurs continuously over time.3

It is possible to neatly separate the effects of learning and emulation on equilibrium pricing.

Emulation unambiguously reduces prices through a social multiplier of demand. The mechanism

is as follows. In a search model, consumers are more likely to buy from the firm they visit first.

When consumers emulate others, they are more likely to start their search in those firms that

have a higher market share, linking individual demand to aggregate demand. Because of this

multiplier, firms lower prices not only to retain incoming consumers (as in the standard model),

but also to increase the number of consumers they attract through emulation.4We show that the

resulting equilibrium price is a fraction of the price in the standard model. We further show that

this fraction depends on the magnitude of the social multiplier, decreases in search costs and

2In Section 3, we review the empirical evidenceon emulation and lear ning.

3See the third subsection of Section 5 and the online web Appendix C for extensions to an infinite-horizon model

with dynamic pricing.

4A similar consideration arises in models of switching costs, further discussed in the literature review.

C

The RAND Corporation 2018.