implies that the continuation value of searching is higher for consumers, inducing a
downward pressure on prices. By contrast, without targeting, a firm cannot adjust its
advertising strategy D along with its equilibrium price P . The per-click-price is therefore
considered a fixed cost and is not passed through into the final good price.
60
As a con-
sequence, the overall effect is ambiguous, and
de Cornie
`
re (2013) shows that targeting
can lead to a welfare loss.
Another important question regards the incentive of the search engine to choose the
most relevant ads after a consumer has entered a keyword. To this end, suppose that the
search engine can choose the value of accuracy of its search results by choosing D itself.
So, the search engine has two choices, D and p.
De Cornie
`
re (2013) shows that, in this
case, the search engine can extract the whole profit from firms with the per-click fee.
Consequently, the search engine chooses D in order to maximize firms’ profits. The opti-
mal matching accuracy for the search engine is, then, D
SE
D
. The intuition behind the
result is as follows. If the search engine sets D < D
, the distance is strictly smaller than the
consumers’ reservation distance. Then, a price increase in the final good market does not
lead to reduced demand. This implies that firms have an incentive to increase the price up
to the reservation price of the marginal consumer, leading to a negative utility of the mar-
ginal consumer (remember that the search costs are sunk). As a consequence, consumers
will not participate. Therefore, a high level of matching accuracy lowers product market
competition, which dissuades consumers from using the search engine. Instead, when
D > D
, some consumers search more than once. This leads to lower firm prices and
ensures higher consumer participation. Therefore, the search engine may find it optimal
to set D
SE
D
but never D
SE
< D
. Therefore, the search engine does not have an
incentive to choose a more accurate matching than advertising firms themselves.
To sum up,
de Cornie
`
re (2013) demonstrates that keyword advertising induces better
targeting by advertisers and thereby reduces the search frictions of consumers. However,
it also changes product market competition. Because advertisers pay per click, they face
higher marginal costs, which can lead to higher final consumer prices. In addition, a
search engine may not want to offer the highest level of matching accuracy because this
leads to a small number of results, reduces product market competition, and makes the
search engine less attractive for consumers.
In the empirical literature, keyword advertising and targeting technologies on the
Internet have also attracted considerable attention. For example, to determine the effec-
tiveness of keyword advertising,
Ghose and Yang (2009) use a panel data set of several
hundred keywords from a nationwide retailer that advertises on Google. They find that
click-through and conversion rates fall in the keyword rank. However, this is not
60
Dellarocas (2012) provides an in-depth analysis of the implications of pay-per-click pricing on final con-
sumer prices. He shows that performance-based advertising, such as pay-per-click pricing, leads to double
marginalization. As a result, consumer prices are higher than with per-impression pricing.
509
The Economics of Internet Media
necessarily true for profitability. In particular, keywords in middle positions are often
more profitable than those at the very top of a search engine’s results page. Interestingly,
Ghose and Yang (2009) find that the effect of retailer-specific information in a keyword is
very different from brand-specific information. Whereas retailer-specific information
leads to an increase in conversion rates of up to 50.6%, brand-specific information leads
to a decrease of 44.2%. Similar patterns are observed for CTRs. It could be of interest to
extend the model outlined above (or related frameworks) to allow for many search results
with different ranks to explain this empirical evidence.
Other studies also analyze the influence of the position (or rank) on click-through and
conversion rates and provide an analysis distinguishing between different advertising
effectiveness measures. For example,
Agarwal et al. (2011) also find a positive effect
on the CTR but demonstrate that the conversion rate is often higher for middle-ranked
positions.
Rutz and Trusov (2011) provide a model and an empirical analysis of the rela-
tion between CTRs and conversion rates, whereas
Rutz and Bucklin (2011) demonstrate
positive spillovers between generic and sponsored search. That is, a generic search often
generates a subsequent sponsored search.
In a recent study,
Blake et al. (2015) also measure the effectiveness of keyword adver-
tising, explicitly distinguishing between brand and non-brand keywords. When a user
types in a brand such as “Macys” or “eBay” as a query in a search engine, it is very likely
that the user is already familiar with the brand. In response to the user’s query, the search
engine displays paid ads at the top of the search results, and the brand pays the PCP given
that the user clicks on this query. However, the user would have found the brand’s site
almost surely through organic search.
Blake et al. (2015) test this with data from eBay.
They halted advertising for eBay-brand-related queries on the search engines Yahoo! and
MSN for some time and found that 99.5% of traffic from the paid link was immediately
captured by traffic from the organic link. Hence, substitution between paid and unpaid
traffic for eBay is almost complete.
61
Is this also true for non-brand keywords? To answer
this question,
Blake et al. (2015) conducted another natural experiment by stopping eBay
advertising via paid links in designated areas for 60 days. In addition, they segmented con-
sumer groups into those who are frequent eBay visitors and those who are not. They
found that paid links did not have a statistically significant effect on the first group since
users of this group are already familiar with eBay. However, there was a significant
increase in newly registered users and purchases in the second group due to exposure
of paid links. This supports the informative view of advertising.
Several empirical studies focus on how targeted advertising interacts with other forms
of advertising.
Goldfarb and Tucker (2011b) conduct a large-scale field experiment
exposing individuals to two different forms of online advertisements. The first is a
61
A natural explanation is that eBay is a well-known brand and is therefore highly listed in the organic search
results.
510
Handbook of Media Economics
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.