the zip-code level. If zip codes differ substantially in their composition by preference
group, then they can determine which groups find the local paper’s targeting appealing.
They present direct evidence on the relationship between the distribution of con-
sumer types in the metropolitan area and the positioning of the newspapers. In particular,
they characterize the local product according to the percentage local coverage that is
“hard” news (news, business, government, etc.) as a function of local demographics.
They find that metropolitan areas with a higher black population share have a lower hard
news share.
To measure own-group and cross-group effects, they regress the share of a zip code’s
population purchasing the local paper on consuming at the zip-code level. To see their
approach, imagine that they had group-specific consumption data (e.g., the share of a zip
code’s white or black consumers choosing the local paper). Then they could regress
s
w
z
¼α
0
+ α
1
W
M
+ α
2
B
M
+ E
w
z
, (1.1)
s
B
z
¼β
0
+ β
1
W
M
+ β
2
B
M
+ E
B
z
, (1.2)
where s
z
w
shows the fraction of white population in zip code z consuming the paper, W
M
is metro area white population, B
M
is metro area black population, the αs and βs are coef-
ficients, and the E terms are errors. If targeting follows group preferences, then we would
see positive own-group effects through α
1
>0 and β
2
>0: e.g., whites would purchase
the paper more in markets with larger white population, all else constant. Negative cross
effects would emerge if, say, blacks purchased the paper less in markets with more whites,
all else constant (α
2
<0 and β
1
<0).
Extant data, which show total sales by zip code (s
z
), do not allow this approach. How-
ever, if we note that the share of a zip code consuming the paper is the weighted sum of
the unobserved shares of blacks consuming and the unobserved share of whites
consuming:
s
z
¼b
z
s
B
z
+1b
z
ðÞs
W
z
, (1.3)
then we can plug
(1.1) and (1.2) into (1.3) to yield
s
z
¼α
0
+ α
1
W
M
+ α
2
B
M
+ β
0
α
0
ðÞb
z
+ β
1
α
1
ðÞW
M
b
z
+ β
2
α
2
ðÞB
M
b
z
+ ν
z
:
That is, they estimate the coefficients of interest by regressing the zip-code consumption
share on the metro black and white population, as well as interactions of the metro area
group populations on the zip-code black share.
Using this approach
George and Waldfogel (2003) find positive own effects, partic-
ularly for blacks. That is, in markets with larger white populations, all else constant, a
higher share of whites are attracted to the newspaper, by about 5% per additional million
whites. In markets with more blacks, a higher share of blacks are attracted to consump-
tion, by about 40% per additional million blacks. Cross effects, by contrast, are negative.
Adding a million whites reduces black circulation by about 15%. Adding a million whites
33Preference Externalities in Media Markets
has a statistically insignificant impact on white newspaper consumption. Evidence on
Hispanics and non-Hispanics is quite similar in direction and relative magnitudes. Higher
non-Hispanic population raises non-Hispanic reading by a small amount. Higher His-
panic population raises Hispanic circulation substantially (by a substantial amount).
The cross effect running from non-Hispanics to Hispanics is negative: an additional
million non-Hispanics reduces Hispanic circulation by about 20%. The cross effect
operating in the other direction is not significant.
The negative cross effects running from majority population to minority consump-
tion tendencies are stark and constitute the fairly direct evidence of the tyranny of the
majority in a market context. A single product courting the widest possible audience
is driven by metropolitan area composition in a way such that larger population of
one group reduces the other group’s consumption.
Gentzkow and Shapiro (2010), while aiming at a different question, provide comple-
mentary evidence of an analogous tyranny of the majority (see Chapter 14). They show
that local newspapers are positioned politically according to the preferences of consumers
in the metropolitan area. Zip-code level demand, by contrast, depends on more local
preferences. Evidence for preference externalities operating through positioning is, for
example, that newspaper circulation is high in heavily Republican zip codes of metro-
politan areas that are also highly Republican.
1.4.2.2 Preference Externalities with Heterogeneous Consumers in Markets
with Many Products
We documented above that larger markets have more and more varied radio stations.
When we divide consumers into different groups according to their radio programming
preferences we see more specific evidence of targeting. A regression reported in Waldfogel
(1999) of the number of black-targeted radio stations in a metropolitan area on black pop-
ulation and white population shows that markets with a million more blacks have 7 more
black-targeted stations, while an additional million whites reduces black entry by 0.6 sta-
tions. Similarly, white-targeted entry is 4.6 stations higher in a market with an additional
million white consumers but 12.8 lower in a market with an additional million black con-
sumers. That is, own effects on entry are positive, while cross effects on entry are negative.
Similar own and cross effects appear among Hispanic and non-Hispanic stations. Positive
own effect is larger for the minority group. Negative cross effects are larger for the impact
of minority population on majority-targeted programming.
Group-targeted entry is particularly valuable to consumers. In a regression of white
AQH listening on the numbers of white- and black-targeted stations and controls (instru-
menting entry with the levels of group population), an additional white station raises lis-
tening by 0.15 percentage points while an additional black station raises white listening by
only 0.12 percentage points. An additional black station raised black listening by 0.78
percentage points while an additional white station raised white listening only 0.19
34 Handbook of Media Economics
percentage points. Patterns were similar for effects of Hispanic and non-Hispanic stations
on group listening.
The direct relationship between group listening and the populations of the respective
groups summarizes the preference externality. Using data on 100 US metro areas with
separate data on black and overall listening in 1997, Waldfogel (1999) documents that
own-group listening is higher in markets with more members of the own group. White
listening is 0.4 percentage points higher in metro areas with an additional million whites,
and black listening is 2.7 percentage points higher in markets with an additional million
blacks. Cross effects are insignificant, although the point estimate of black population on
white listening is negative. These estimates confirm the prediction of the two-group logit
model, of positive own-group and zero cross-group preference externalities.
1.4.3 Efficient Entry and Preference Externalities
Models of the efficiency of entry patterns, such as the logit model articulated above and
that of
Mankiw and Whinston (1986), have implications related to preference external-
ities. With fixed costs, with one group of consumers and one type of symmetrically dif-
ferentiated product, marginal entry reveals the market’s implicit welfare weight on the
marginal consumer. That is, suppose that the last entrant costs $1 million and raises overall
consumption by 10 units. For the sake of discussion, assume that marginal entry has no
effect on prices. This reveals that the market values consumers at $100,000. This char-
acterization is a slight over-simplification in that, while marginal entry raises consump-
tion by 10 on net, the gross consumption of the marginal entrant will typically exceed 10.
Say it’s 50. Then while 10 consumers are now getting some product rather than no prod-
uct, the other 40 are getting a product better than a product they were already
consuming.
Berry and Waldfogel (1999) study the efficiency of entry into US radio broadcasting,
treating welfare as the value of advertising produced, less the fixed costs of station oper-
ation. That is, they examine the efficiency of the market from the standpoint of direct
market participants, the buyers and sellers of advertising. They find that US radio markets
had about three times too many stations than the number that would maximize the wel-
fare of market participants. Understanding that radio programming has value to listeners,
they also inferred the value that a marginal listener would need to have attached to
programming to render observed entry patterns efficient. They found this to be
$893 per year, while the ad revenue was $ 277 (both in 1993 dollars). Note that the
implicit value is larger in larger markets, as they have more entry and a smaller net impact
of the marginal station on total listening, as well as higher costs of station operation.
The market’s welfare weight arises from a mechanism related to the preference exter-
nality: it shows how much firms in the market expend to deliver consumption to one
additional individual (again, putting aside the benefit experienced by consumers finding
35Preference Externalities in Media Markets
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.