Online advertising and privacy

• Date: 01 February 2016
• DOI: http://doi.org/10.1111/1756-2171.12118
• Published date: 01 February 2016

RAND Journal of Economics

Vol.47, No. 1, Spring 2016

pp. 48–72

Online advertising and privacy

Alexandre de Corni`

ere∗

and

Romain de Nijs∗∗

An online platform auctions an advertising slot. Several advertisers compete in the auction,

and consumers differ in their preferences. Prior to the auction, the platform decides whether to

allow advertisers to access information aboutconsumers (disclosure) or not (privacy). Disclosure

improvesthe match between advertisers and consumers but increases product prices, even without

price-discrimination. Weprovide conditions under which disclosureor privacy is privately and/or

socially optimal. When advertisers compete on the downstream market,disclosure can lead to an

increase or a decrease in product prices depending on the nature of the information.

1. Introduction

The online advertising industry has grown rapidly in the last decade.1One of the two main

forms of online advertising, along with search, is display advertising. The basic organization of the

display advertising market is the following:while surfing the Inter net, users visit variouswebsites

(also called publishers) whohave an inventory of advertising slots to sell. This can be done directly

by reaching to potential advertisers, or indirectly, through intermediaries whose main role is to

aggregate supply and demand of advertising space and to act as matchmakers. Advertising

networks (such as Google AdSense, Adblade) and advertising exchanges (DoubleClick, OpenX)

are such intermediaries.

Intermediaries and large publishers (such as Facebook or Google), which we will designate

as platforms, have access to technologies that enable them to gather and analyze a considerable

amount of data at a very high speed, making it possible to customize advertising using real-time

auctions (see, e.g., “Getting to Know You,” 2014). Advertisers may submit bids that depend, for

∗Toulouse School of Economics; adecorniere@gmail.com.

∗∗Ecole Polytechnique and Ecole des Ponts ParisTech; romain.de-nijs@polytechnique.edu.

We thank Bernard Caillaud, Chris Dellarocas, Gabrielle Demange, Peter Eso, Philippe F´

evrier, Benjamin Hermalin,

Nabil Kazi-Tani,Fr ´

ed´

eric Koessler,Philippe J ´

ehiel, Bruno Jullien, John Morgan, R´

egis Renault, Jean Tirole, and seminar

participants at Crest, the Paris School of Economics, the sixth biannual Conference on the Economics of Intellectual

Property, Software, and the Internet in Toulouse, the second Workshop on the Economics of ICT in Universidade

de Evora, the 2011 IIOC conference, the workshop Communication and Beliefs Manipulation at PSE, and the Third

Annual Conference on Internet Search and Innovation. Corni`

ere author benefited from funding by the ANR (grant no.

2010-BLANC-1801 01).

1See Evans (2008, 2009) for insightful discussions about this industry.

48 C2016, The RAND Corporation.

CORNI`

ERE AND NIJS /49

FIGURE 1

MARKET STRUCTURE

n

instance, on the correspondence between the website’s content and the advertisement, but also

on data about the location of the consumer (obtained through the Internet Protocol [IP] address),

his past browsing history (obtained through cookies), or whatever information he gave to the

platform or its partners (through subscription questionnaires for instance, or any information

posted on his Facebook wall). These newopportunities give firms additional incentives to acquire

and use personal information about consumers, which has led regulators and consumers to express

worries, or at least to acknowledge some potential pitfalls. Among these are privacy breaches or

fraudulent use of personal information, but also practices of behavioral targeting and pricing.

In this article, we study the decision by a platform of whether to use the information it has

gathered about consumers to increase its revenue from advertising. Do such practices have social

value? Who benefits most from them?

The situation we have in mind is the following (see Figure 1): consumers visit a platform,

which auctions an advertising slot among nadvertisers.2Consumers are heterogeneous, in the

sense that they do not derive the same value from consuming advertisers’ products. Thanks to its

technology, the platform gathers, for each consumer, information correlated with the consumer’s

willingness to pay for any product. The platform does not know how to interpret the information

in terms of implied willingness to pay for different products, but advertisers are able to do it. For

instance, the platform knows that the consumer is a young man living in a metropolitan area, but

it is not able to infer his willingness to pay for good A or B. On the other hand, firm A knows

that young men living in a metropolitan area are especially likely to have a high willingness to

pay for its product, whereas firm B offers a product that is less likely to be a good match for such

consumers.

In the economic literature, privacy has been defined as “the restriction of the collection or

use of information about a person or corporation” (Stigler, 1980).3In our setup, we define privacy

as the platform’spolicy under which it does not disclose consumer information to advertisers. To

make the analysis as transparent as possible, we assume that consumers do not exhibit intrinsic

preferences over their privacy but care about it insofar as it has economic effects.

We are thus concerned with two main questions: (i) What are the effects of the disclosure

policy on market outcomes, that is, on the interactions between consumers and advertisers? and

(ii) When does the platform provide the efficient amount of privacy?

Regarding (i), we show that disclosure has positive and negative consequences. On the pos-

itive side, when advertisers can condition their bids on information about consumers, the highest

bidder, in equilibrium, is the firm that offers the best match, which is efficient. However, when

good matches correspond to higher marginal revenues for advertisers, we show that disclosure

2The case of several slots is analyzed in Section 5.

3See Png and Hui (2006) for a survey of the economics of privacy.

C

The RAND Corporation 2016.