if consumer 2 is the marginal consumer, then the equilibrium price is r
2
(0). By using adver-
tising, the firm may increase its revenue by 2 r
2
1ðÞr
2
0ðÞðÞand it chooses to advertise if this
is larger than 1. Because this has no impact on consumer surplus or on the producer surplus
of competitors, this rule is also optimal from a social surplus perspective.
Nichols (1985)
shows that if the commodity market is perfectly competitive, then the equilibrium adver-
tising expenditure is socially optimal.
Although the complementary view seems like the natural approach that is consistent
with the standard microeconomics of consumer behavior, many economists consider that
much of advertising induces some consumer response that can only be attributed to some
form of “irrationality.” This is what I have described in the introduction, following
Bagwell (2007), as the persuasive view. Still, there are two ways of interpreting the
increase in willingness to pay that advertising triggers. A first interpretation is that it is
pure manipulation that changes consumer behavior but not consumer tastes. This is,
for instance, explicitly the case in
Glaeser and Ujhelyi (2010), where advertising is pure
misinformation. This is also coherent with the analysis of advertising by
Braithwaite
(1928)
or Kaldor (1950). Social welfare should then be evaluated using the
no-advertising valuations, r
i
(0), i ¼1, 2, in my two-consumer example.
But the persuasive view has often been portrayed as considering advertising as a taste
shifter (see
Bagwell, 2007, p. 1720). For instance, Stigler and Becker (1977, p. 83) cite
Galbraith (1958), who writes that advertising’s “… central function is to create desires—
to bring into being wants that previously did not exist”. In a very influential and contro-
versial article,
Dixit and Norman (1978) undertake a welfare analysis of advertising that is
consistent with this point of view. Because tastes change as a result of the firm’s adver-
tising, it is unclear which preferences should be taken into account to establish the welfare
benchmark. Either the new preferences are treated as irrelevant to welfare and then the
analysis should proceed assuming advertising only induces a change in behavior, so the
original valuations should be used to compute social welfare. Or we should acknowledge
that this change in tastes defines a new welfare criterion and the post-advertising valu-
ations provide the proper welfare measure to compare the market situation with and
without advertising.
Dixit and Norman (1978) consider both possibilities and establish
a remarkable over-provision result, which holds independent of which welfare measure
is selected.
I now present the gist of Dixit and Norman’s argument using the two-consumer
monopoly setting. Assume that, without advertising, the firm only serves consumer 1,
whereas if it advertises it serves both. This increase in quantity ensures that there is some
social benefit associated with advertising. Otherwise, advertising would clearly be exces-
sive if the firm chose to resort to it. Further assume that r
2
1ðÞr
1
1ðÞso that consumer 2 is
the marginal consumer if the firm advertises. The firm’s surplus is therefore r
1
0
ðÞ
c if it
does not advertise and 2 r
2
1ðÞcðÞ1 if it does. Advertising thus generates the incre-
mental producer surplus 2r
2
1ðÞr
1
0ðÞc 1. Using the post-advertising tastes, the
187Advertising in Markets