4.3.5.1 Cheap Talk
I start by considering Nelson’s claim that statements regarding search attributes are inher-
ently more credible than those regarding experience attributes. Consider a monopoly
firm selling a product for which the consumer has a price-sensitive demand
dp, s, tðÞ¼Θ
s,t
p, where s is the product’s quality, which is either high, s ¼h,or
low, s ¼‘, and t is the consumer’s type, which is either high, t ¼h, or low, t ¼‘. Further
assume that Θ
h,h
¼6, Θ
h,‘
¼Θ
‘,h
¼2, and Θ
‘,‘
¼1. All types are independent and have
equal probability 1/2. Production costs are zero. If the consumer has perfect information
about quality while the firm only knows the probability distribution of t, then it is readily
verified that the high-quality firm only serves the high-type consumer at a price of 3 and
the low-quality firm charges a price of 3/4 and serves both types. Corresponding profits
are 9/2 for firm type h and 9/16 for firm type ‘. If the product was an experience good
and the consumer did not know the quality as in
Section 4.3.4, the firm could convey no
credible information to the consumer by advertising an unsubstantiated claim about its
quality: a low-quality firm would always claim its quality is high.
Now assume that the consumer must incur a visit cost γ > 0 in order to purchase the
product. If she does, she learns s as well as the price charged by the firm. Furthermore, in
case of a visit, the firm incurs a processing cost ρ > 0, whether the product is sold or not
(
Nelson, 1974 mentions such processing costs as deterrents to mislead consumers for
advertisers of search goods). The firm may post a cheap-talk message in an ad that the
consumer can observe before she decides whether to visit or not. This cheap-talk message
is a claim that s has some value ^s 2 ‘, h
fg
. Now, because the consumer observes quality
before buying, a firm’s profit gross of processing costs is its full information profit. Also
note that no matter how small γ is, a type ‘ consumer never visits the firm if it expects it to
have high quality because she then expects zero surplus. Now the high-quality firm gains
nothing from inducing a visit from a type ‘ consumer but it incurs the processing cost ρ.
By contrast, if ρ is not too large then a low-quality firm prefers to have both types of
consumer visiting and if γ is not too large, a type ‘ consumer visits the firm if it expects
it is low quality, since she expects a price of 3=4ðÞ< 1. Hence there is an equilibrium
where the firm truthfully and credibly reveals its quality in its ad.
39
I now turn to Nelson’s insight concerning the credibility of cheap-talk messages
pertaining to experience goods. He gives the example of a medicine that either cures
athlete’s foot or indigestion. Provided that the potential demand for the two types of
medicine is similar, the firm benefits nothing from making a false claim about its product
(if there are potential repeat purchases as Nelson assumes there are, then it is actually
39
This simple model is adapted from that of Gardete (2013). He also finds that cheap-talk revelation of
quality is possible for a search good but uses a different specification of preferences, such that it is profitable
for a type ‘ firm to be matched with a type ‘ consumer and for a type h firm to be matched with a type h
consumer. He has no processing costs.
162
Handbook of Media Economics