Word‐of‐mouth communication and search

• Author: Arthur Campbell,Yves Zenou,C. Matthew Leister
• Date: 01 September 2020
• DOI: http://doi.org/10.1111/1756-2171.12337
• Published date: 01 September 2020

RAND Journal of Economics

Vol.51, No. 3, Fall 2020

pp. 676–712

Word-of-mouth communication and search

Arthur Campbell∗

C. Matthew Leister∗

and

Yves Zenou∗,∗∗

We develop a word-of-mouth search model where information flows from the old to the new gen-

eration for an experience good with unknown quality. We study the features of the social network

that determine product quality and welfare and characterize the demand-side (under provision of

search effort) and supply-side (inefficient entry by firms) factors that result in inefficiencies. One

striking result is to show that the more connected but also the less-unequal a friendship network

is (in the sense of second order stochastic dominance of the degree distribution), the less can

disreputable (low-quality) firms thrive in equilibrium.

1. Introduction

A central question for economists is the efficiency of markets in the presence of asymmet-

ric information. In particular, a voluminous literature has considered markets where the quality

of a product is known to the seller but is not observable to buyers prior to purchase.1Advice

from friends is frequently the most important and credible source of information for consumers

about the quality of available products, especially experience goods, for which the quality is

unobservable prior to purchase but observable after purchase.2

Wepresent the first study examining the role of social networks in determining the provision

of quality of an experience good in a market. Our main contributions are to (i) determine the fea-

tures of the network (described by a distribution of friendships) that affect the expected quality

∗Monash University; arthur.campbell@monash.edu, matthew.leister@monash.edu,yves.zenou@monash.edu.

∗∗Center for Economic Policy Research (CEPR).

The authors thank the editor, two anonymous referees, Ivan Balbuzanov, Maciej Kotowski, Sephorah Mangin, Nicola

Persico, Chengsi Wangand seminar participants at University of Melbourne, Monash University, National University of

Singapore, the University of Technology Sydney, the Australian National University,the University of Auckland, and the

Symposium on the Economics of Networks for their helpful comments.

1In some instances a market may fail to exist, or only low-quality products may trade (the market for lemons

example in Akerlof (1970) is one of the most prominent of many in economics).

2In the three most recent Nielson surveys of the “Global Trust in Advertising and Brand Messages” (Nielsen,

2012, 2013, 2015), a word-of-mouth recommendation is the most trusted source of information and the one that most

frequently leads to an action.

676 © 2020, The RAND Corporation.

CAMPBELL, LEISTER, AND ZENOU / 677

in the market; (ii) show that inefficiencies arise due to both demand side factors, as an under pro-

vision of search effort by consumers, and supply side factors, due to an inefficient allocation of

firms across markets with different social networks; and (iii) develop a tractablemodeling frame-

work for this setting that we demonstrate through considering a number of practical extensions.

We consider a market where consumers search via communication with their friends about

the quality (high or low) of the availableproducts/firms. A mass of consumers i sbor n in each pe-

riod and each consumer chooses a product to purchase. The product is an experience good so its

quality is unobservable prior to purchase but observable after purchase. Prior to buying the prod-

uct, individuals ask their connections, who made purchases during the previous period, whether

they had purchased high-quality products. In the event that one or more of their connections

buy and identify a high-quality product, the individual randomly selects one of those products

to purchase. If no such person is found, the individual randomly selects a product from the mar-

ket at large. High- and low-quality firms can freely enter. High-quality firms have better outside

opportunities (equivalently higher costs) than low-quality firms, so they must anticipate greater

profits in order to enter the market. Word-of-mouth search gives high-quality firms an advantage

over low-qualityfir ms. Thus, high-quality firms maysell to more consumers and may earn higher

profits in the market. We study the fraction and market share of high- and low-quality firms.

In our baseline model, we find that there is a unique steady-state equilibrium in which the

social network structure and the outside options of firms determine whether there is a mixture

of high- and low-quality firms or the market is entirely occupied by either high- or low-quality

firms. Three key variables matter in determining which equilibrium prevails: the percentage of

individuals with no friends, the expected number of friends in the network,and the ratio of outside

options for high-/low-quality firms. In particular, when the expected number of friends is small

relative to the ratio of outside options, only low-quality firms enter the market. Conversely, when

the fraction of people with no friends is small relative to the ratio of outside options, only high-

quality firms enter. In between these extremes, a mixture of low- and high-quality firms enter.

Furthermore, in the mixed-quality case, first-order stochastic dominance (FOSD) and second-

order stochastic dominance (SOSD) changes to the distribution of friendships, and decreases to

the ratio of outside options (thereby making it relatively cheaper for high-quality firms to enter)

all reduce the market share and the fraction of low-quality firms.

A central issue in markets with asymmetric information is whether the outcomes in the mar-

kets are efficient and the nature of any inefficiencies. We develop two extensions to the baseline

model to consider these issues. First, we endogenize consumers’ search effort. In a costly search

model, each individual decides how actively she wants to search for friends. Focusing on sym-

metric equilibria, we show that a steady state with only low-quality firms always exists whereas

a steady state with only high-quality firms is never possible. Furthermore, when the search costs

are low enough, there are generically two more steady states with a mixture of high-/low-quality

firms. In the latter case, the utility of a representative consumer is increasing in the equilibrium

effort level, which is below the effort level that maximizes consumer welfare. The inefficiency

is composed of both demand and supply side factors. On the demand side, conditional on the

composition of firms in the market, consumers search too little. Furthermore, the composition of

firms in the market contains too high a fraction of low quality firms.

Second, we consider a setting where a fixed number of high- and low-quality firms choose

one of two markets to enter. We find that the market with the more connected social network will

contain a greater fraction and market share of high-quality firms. Less obviously,this market will

be more congested and/or will charge a higher price. In contrast, a social planner maximizing

welfare will allocate a greater fraction of low quality firms to the better connected market, where

low-quality firms are more effectively detected and avoided.

Our baseline model shuts down many forces. In considering the simplest possible model, we

are able to focus entirely on the impact of the efficacy of the word-of-mouthprocess on aggregate

outcomes and, more generally, to study the impact of word-of-mouth search in various markets

C

The RAND Corporation 2020.

678 / THE RAND JOURNAL OF ECONOMICS

with asymmetric information. To show that these results are quite general and the modeling

framework is quite flexible, we develop a number of extensions that are discussed in Section 6.

We believe that our model could serve as a useful benchmark for modeling experienced

goods where information is mainly diffused through word-of-mouth communication. Moreover,

the business operating in these markets seem sufficiently small that websites attempting to aggre-

gate opinions will be subject to manipulation and therefore perceived to be unreliable. In these

types of markets, the recommendation of a friend is often viewed as the most credible source

of information for consumers. In particular, most other sources of information come from enti-

ties with an interest in inducing the consumer to purchase a particular product. Consequently, in

markets for experience goods, consumer search via communication with friends is an important

determinant of demand for each firm. Moreover, it is reasonable to expect that the efficacyof this

word-of-mouth search process affects the average quality provided in the market.3

This last result, that denser or more connected networks lead to better quality products, car-

ries some empirical relevance in markets beyond local-services markets. Indeed, there is a large

empirical literature on technology adoption that shows that network effects matter.4Far mer s in

developing countries are asked to choose between a better-quality product (the new technology)

and a lower-quality one (the old technology) without knowing with certainty the quality of the

new technology. The results show that the individual adoption of the new technology crucially

depends on the adoption of the individual’s friends or acquaintances. This implies that, in par-

ticular, in a more connected society (or village), the adoption of a new technology (high-quality

good) is much higher than in a less-connected society.

2. Related literature

This article is most directly related to models where firms face a moral hazard problem

in the provision of quality or consumers face an adverse selection problem. This issue has been

approached in broadly two different ways. First, in static models, the focus has been on the ways

in which a firm may signal to consumers that a product is high quality through prices, advertising,

money burning activities and contracts, such as guarantees/warranties (see, for instance, Nelson,

1974; Schmalensee, 1978; Grossman, 1981; Milgrom and Roberts, 1986; Bagwell and Ramey,

1988). Second, in dynamic models, consumers observe the quality at some moment after their

initial purchase (this does not need to be immediate). Once a consumer has observed the quality

of the good, their behavior contingent on this information affects high- and low-quality firms

differently. This line of research has considered dynamic elements such as reputation (Shapiro,

1982; Rogerson, 1983), as well as quality and price cycles (Gale and Rosenthal, 1994).5

Our article is more closely related to the dynamic literature. In our model, high-quality firms

obtain greater demand (from positive word of mouth), and earn greater profits than low-quality

firms. An article closely related to ours is that of Galenianos, Pacula, and Persico (2012), which

also studies a steady-state distribution of quality (in their case, quality is defined in terms of

the purity of an illicit drug) and has the feature that firms with higher quality sell more in the

steady state. However, the mechanisms for this are very different in the two articles. Their article

focuses on how the nature of the ongoing relationship between a drug consumer and dealer allows

high-quality dealers to sell more.6In contrast, in the current article, a high-quality firm obtains

greater future demand for its product through the effect of positive word of mouth on future

3In Section 3, we discuss products and markets for which our model is a good fit.

4One of the important empirical question in the network literature has been the study of the adoption of agricultural

technologies (such as fertilizer or high-yielding seeds) and basic or preventative health technologies(such as deworming

drugs or bednets). Seminal articles include that of Foster and Rosenzweig (1995) and Conley and Udry (2010), and a

recent wave of field experiments further explores this question. See Breza (2016) for an overviewof this literature.

5See, also, Godes (2017) and Jiang and Yang(2019), for two recent contributions from the marketing literature.

6They consider how the number of sellers, consumer search costs, and temporary unavailability of a dealer affect

quality. The authors also consider how different enforcement policies may affect these quantities and examine their

subsequent effect on the market.

C

The RAND Corporation 2020.