Sharing Audience Data: Strategic Participation in Behavioral Advertising Networks


I consider the incentives of special interest websites to participate in behavioral advertising intermediaries. Participation in the intermediary reveals valuable audience data and allows the intermediary to use those data to target the site’s audience on general interest websites—thus expanding the supply of impressions and decreasing average revenue per impression. I explore monopoly and duopoly settings and highlight the trade-off between sharing audience data and displaying higher-value ads, as well as the strategic interaction between sites serving the same advertising market. The model generates empirical predictions about the choice of intermediary technologies within advertising markets. I also find that higher concentration among special interest websites benefits consumer privacy.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3


  1. 1.

    PricewaterhouseCoopers IAB Internet advertising revenue report. 2014 full year results.

  2. 2.

    The online advertising industry uses the term “network” for these entities. Here, I use the term intermediary to avoid confusion with network externalities.

  3. 3.

    Many other tracking technologies exist, such as Flash locally stored objects, browser fingerprinting, and “supercookies.”

  4. 4.

    Here, there is no interaction between advertising markets, which thus allows the normalization. High-value and low-value markets may differ in other parameters, however, such as the number of ads that are shown to each person, or the premium that is offered by the behavioral intermediary.

  5. 5.

    I do not model the interaction between the number of ads that are shown per user and the number of users that are willing to browse the site, as this trade-off is an empirical question that is outside the scope of the present paper.

  6. 6.

    If consumers have strong preferences for privacy, there are tools to prevent their data from being shared with behavioral intermediaries. If consumers are unnerved by seeing targeted ads on general sites, they may be less likely to click on them. In setting where impressions are measured by click, this would be represented by reducing \(\gamma \).


  1. Athey, S., Calvano, E., & Gans, J. (2014). The impact of the internet on advertising markets for news media, Working paper. National Bureau of Economic Research.

  2. Athey, S., & Gans, J. S. (2010). The impact of targeting technology on advertising markets and media competition. The American Economic Review, 100(2), 608–613.

    Article  Google Scholar 

  3. Bergemann, D., & Bonatti, A. (2011). Targeting in advertising markets: Implications for offline versus online media. The Rand Journal of Economics, 42(3), 417–443.

    Article  Google Scholar 

  4. Bergemann, D., & Bonatti, A. (2015). Selling cookies. American Economic Journal: Microeconomics, 7(3), 259–294.

    Google Scholar 

  5. Berman, R. (2013). Beyond the last touch: Attribution in online advertising (November 16, 2017). Available at SSRN:

  6. Budak, C., Goel, S., Rao, J. M., & Zervas, G. (2014). Do-not-track and the economics of third-party advertising. Boston University School of Management Research Paper No. 2505643.

  7. Butters, G. R. (1977). Equilibrium distributions of sales and advertising prices. The Review of Economic Studies, 44(3), 465–491.

    Article  Google Scholar 

  8. Chen, J., & Stallert, J. (2014). An economic analysis of online advertising using behavioral targeting. MIS Quarterly, 38(2), 429–449.

    Article  Google Scholar 

  9. Englehardt, S., Reisman, D., Eubank, C., Zimmerman, P., Mayer, J., Narayanan, A., & Felten, E. W. (2015). Cookies that give you away: The surveillance implications of web tracking. In Proceedings of the 24th international conference on world wide web, WWW ’15 (pp. 289–299), New York, NY: ACM.

  10. Ghosh, A., Mahdian, M., McAfee, P., & Vassilvitskii, S. (2012). To match or not to match: Economics of cookie matching in online advertising. In ACM EC’12 (pp. 741–753).

  11. Goldfarb, A., & Tucker, C. E. (2011). Privacy regulation and online advertising. Management Science, 57(1), 57–71.

    Article  Google Scholar 

  12. Johnson, J. P. (2013). Targeted advertising and advertising avoidance. The RAND Journal of Economics, 44(1), 128–144.

    Article  Google Scholar 

  13. Mayer, J. R., & Mitchell, J. C. (2012). Third-party web tracking: Policy and technology. In 2012 IEEE symposium on security and privacy.

  14. Tucker, C. E. (2012). The economics of advertising and privacy. International Journal of Industrial Organization, 30(3), 326–329.

    Article  Google Scholar 

  15. Zhang, K., & Katona, Z. (2012). Contextual advertising. Marketing Science, 31(6), 980–994.

    Article  Google Scholar 

Download references


This paper benefited greatly from discussions with Lou Silversin and comments by the editor and anonymous referees.

Author information



Corresponding author

Correspondence to Steven Schmeiser.


Appendix 1: Endogenous P

Here, I examine the behavioral intermediary’s choice of P. As discussed in the main text, P accounts for two things: the premium that is generated by the behavioral intermediary’s additional information about a consumer that a contextual intermediary cannot obtain, and any difference in revenue shares between behavioral and contextual intermediaries. The latter effect is a direct choice of the behavioral intermediary, while the former is driven by how additional consumer information drives advertising demand. To isolate the behavioral intermediary’s choice of revenue sharing, I let \(P=\delta p\) where \(\delta \) is the fraction of revenue that the intermediary passes on to the website, and p is the premium that is generated by extra behavioral data.

Consider an advertising market with a monopoly special interest site that chooses the behavioral intermediary: The total revenue that is generated in the market is given by \(pV(N,\alpha +\gamma ;\lambda )\). \(\delta \) of this total revenue goes to the publisher, and the ad intermediary retains \((1-\delta )pV(N,\alpha +\gamma ;\lambda )\).

The behavioral intermediary chooses \(\delta \). Increasing \(\delta \) decreases the amount of revenue that the intermediary keeps. However, increasing \(\delta \) also attracts marginal advertisers that switch from contextual to behavioral intermediaries as P increases. Recall that the cutoff premium for a monopoly special interest site is given by \(P_M\) in Eq. 6 and \(P_M\) is strictly decreasing in \(\lambda \). Therefore we can invert the relationship and for any P obtain \(\lambda (P)\): the minimum \(\lambda \) that will induce transactions with the behavioral intermediary given P. Suppose that advertising markets are indexed by their \(\lambda \) and that \(\lambda \) is distributed according to \(F(\cdot )\) with density function \(f(\cdot )\). For expositional clarity, I assume that \(N,\alpha \), and \(\gamma \) are the same across advertising markets. The optimal \(\delta \) is then given by the solution to the behavioral intermediary’s problem

$$\begin{aligned} \max _{\delta \in [0,1]} \int _{\lambda (\delta p)}^1 p(1-\delta )V(N,\alpha +\gamma ;\lambda )f(\lambda )d\lambda . \end{aligned}$$

The marginal benefit of increasing \(\delta \) is given by \(-\lambda _\delta (\delta p)(1-\delta )p^2V(N,\alpha +\gamma ;\lambda )f(\lambda (\delta p))\). \(\lambda _\delta (\delta p)\) is negative, which makes the marginal benefit positive. The marginal cost of increasing \(\delta \) is given by \(-\int _{\lambda (\delta p)}^1 pV(N,\alpha +\gamma ;\lambda )f(\lambda )d\lambda \).

Appendix 2: Proofs

Proof of Proposition 2

$$\begin{aligned} \frac{\partial P_M}{\partial \lambda }&= \frac{\alpha + \gamma }{\alpha } \times \frac{\alpha (e^{-(\alpha +\gamma )(1-\lambda )}-e^{-\alpha (1-\lambda )})+\gamma (e^{-(2\alpha +\gamma )(1-\lambda )}-e^{-(\alpha +\gamma )(1-\lambda )})}{\alpha (e^{-(\alpha +\gamma )(1-\gamma )})^2}\,<\, 0 \\ \frac{\partial P_M}{\partial \alpha }&= \frac{e^{-\alpha (2-\lambda )-\gamma } \left( -\gamma e^{\lambda (\alpha +\gamma )}-(e^{\alpha +\gamma \lambda } - e^{\alpha +\gamma })(\alpha (1-\lambda ) (\alpha +\gamma )+\gamma ) \right) -\gamma }{a^2 \left( e^{-(1-\lambda ) (\alpha +\gamma )}-1\right) ^2} \,<\, 0 \\ \frac{\partial P_M}{\partial \gamma }&= \frac{\left( 1-e^{a (\lambda -1)}\right) \left( e^{-(1-\lambda ) (\alpha +\gamma )} (-(\alpha +\gamma ) (1-\lambda )-1)+1\right) }{\alpha \left( e^{-(1-\lambda ) (\alpha +\gamma )}-1\right) ^2} \,>\, 0 \end{aligned}$$

Proof of Proposition 3

The following comparative statics show how V varies with the parameters of interest holding intermediary choice constant. If a change in the parameters results in a change in equilibrium intermediary, the results hold as revenue is the upper envelope of R(C) and R(B), which have the (weakly) the same sign slope in each of the parameters. As \(\lambda \) increases, V also increases. As \(\alpha \) increases, both \(V(N,\alpha )\) and \(\frac{\alpha }{\alpha +\gamma }V(N,\alpha +\gamma )\) increase. Last, as \(\gamma \) increases, there is no effect on revenue if the site uses the contextual intermediary, and revenue is decreasing if it uses the behavioral intermediary, as \(\frac{\alpha }{\alpha +\gamma }V(N,\alpha +\gamma )\) is decreasing in \(\gamma \). Next, V is increasing in N. Results for P follow immediately from the definition of R(C) and R(B).

$$\begin{aligned} \frac{\partial }{\partial \lambda }V(N,\alpha ;\lambda )&= \frac{e^{1/N} \left( e^{-\alpha (1-\lambda )} \left( (\alpha N-1) e^{\lambda /N}-\alpha e^{1/N} N\right) +e^{\lambda /N}\right) }{N \left( e^{1/N}-e^{\lambda /N}\right) ^2}\,>\, 0\\ \frac{\partial }{\partial \alpha }V(N,\alpha )&= (1-\lambda )\frac{e^{-\alpha (1-\lambda )+1/N}}{e^{1/N}-e^{\lambda /N}}\,>\,0 \\ \frac{\partial }{\partial \alpha }\left( \frac{\alpha }{\alpha +\gamma }V(N,\alpha +\gamma )\right)&= \frac{e^{1/N}\left( \gamma -e^{-(\alpha +\gamma )(1-\lambda )}\left( \gamma -\alpha (\alpha +\gamma )(1-\gamma )\right) \right) }{(e^{1/N}-e^{\lambda /N})(\alpha +\gamma )^2}\,>\,0 \\ \frac{\partial }{\partial \gamma }\left( \frac{\alpha }{\alpha +\gamma }V(N,\alpha +\gamma )\right)&= -\frac{\alpha e^{1/N} \left( e^{-(1-\lambda ) (\alpha +\gamma )} (1-\lambda (\alpha +\gamma )+\alpha +\gamma +1)\right) }{(\alpha +\gamma )^2 \left( e^{1/N}-e^{\lambda /N}\right) }\,<\,0 \\ \frac{\partial }{\partial N}V(N,\alpha )&= \frac{(1-\lambda ) \left( 1-e^{-\alpha (1-\lambda )}\right) e^{(1+\lambda )/N}}{N^2 \left( e^{1/N}-e^{\lambda /N}\right) ^2} \,>\, 0 \end{aligned}$$

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Schmeiser, S. Sharing Audience Data: Strategic Participation in Behavioral Advertising Networks. Rev Ind Organ 52, 429–450 (2018).

Download citation


  • Online advertising
  • Online privacy
  • Behavioral advertising