Skip to main content
SpringerLink
Log in

Repeated Budgeted Second Price Ad Auction

  • Published:
Theory of Computing Systems Aims and scope Submit manuscript

Abstract

Our main goal is to abstract existing repeated sponsored search ad auction mechanisms which incorporate budgets, and study their equilibrium and dynamics. Our abstraction has multiple agents bidding repeatedly for multiple identical items (such as impressions in an ad auction). The agents are budget limited and have a value per item. We abstract this repeated interaction as a one-shot game, which we call budget auction, where agents submit a bid and a budget, and then items are sold by a sequential second price auction. Once an agent exhausts its budget it does not participate in the proceeding auctions.

Our main result shows that if agents bid conservatively (never bid above their value) then there always exists a pure Nash equilibrium. We also study simple dynamics of repeated budget auctions, showing their convergence to a Nash equilibrium for two agents and for multiple agents with identical budgets.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2

Similar content being viewed by others

Notes

  1. In real applications, even the second price auction does not have truth-telling as a dominant strategy, due to their repeated nature, bidders’ budgets, externalities, and more.

  2. A similar conceptual abstraction to a one-shot game was done for studying the truthfulness of click through rates [5, 10].

  3. While technically, advertisement impressions are clearly not divisible, due to the large volume of impressions, this is a very reasonable abstraction. Although using a single divisible item is equivalent, we found it less ‘natural’ for modeling ad auctions.

  4. This hard budget approach is used also in [11]. Another way to ensure agents do not exceed their budget is if agents deposit their budget to the auctioneer (instead of just reporting it), and the auctioneer returns any unused budget at the end of the auction.

  5. Theoretically, agent i might profit by overbidding his value, since it increases the price of the agent i−1 who ranked above him and therefore, decreases agent i−1 allocation, i.e., x i−1. This will result in more items left for agent i and might increase agent i allocation, i.e., x i . Nevertheless, such bidding might expose the agent to negative utility, since he might pay more than his value.

  6. We assume that agent 2 slightly overbids c 1, and we ignore this small perturbation.

  7. We assume that agent j slightly underbids c j , and we ignore this small perturbation.

  8. Note that there is a small difference in the starting price. At this point not necessarily all agents starting price is the new reserved price: agents 1,…,j−1 bid p min =c j , while agents j=1,…,k bid p min +ϵ=c j +ϵ. We can now recalculate the new critical bids for the remaining agents, taking under consideration that agent j is not participating, and the new reserved price is c j . From Lemma 13 we know the new lowest critical bid is at least ϵ higher than c j , so agents 1,…,j−1 increase their bid to p min +ϵ when it is their turn to move, and all ‘starting prices’ are equal again.

References

  1. Andelman, N., Mansour, Y.: Auctions with budget constraints. In: SWAT, pp. 26–38 (2004)

    Google Scholar 

  2. Asdemir, K.: Bidding patterns in search engine auctions. In: Second Workshop on Sponsored Search Auctions, ACM Conference on Electronic Commerce (2006)

    Google Scholar 

  3. Azar, Y., Birnbaum, B.E., Karlin, A.R., Mathieu, C., Nguyen, C.T.: Improved approximation algorithms for budgeted allocations. In: ICALP, pp. 186–197 (2008)

    Google Scholar 

  4. Azar, Y., Birnbaum, B.E., Karlin, A.R., Nguyen, C.T.: On revenue maximization in second-price ad auctions. In: ESA, pp. 155–166 (2009)

    Google Scholar 

  5. Babaioff, M., Sharma, Y., Slivkins, A.: Characterizing truthful multi-armed bandit mechanisms. In: ACM Conference on Electronic Commerce, pp. 79–88 (2009)

    Google Scholar 

  6. Bhawalkar, K., Roughgarden, T.: Welfare guarantees for combinatorial auctions with item bidding. In: SODA (2011)

    Google Scholar 

  7. Borgs, C., Chayes, J., Immorlica, N., Mahdian, M., Saberi, A.: Multi-unit auctions with budget-constrained bidders. In: ACM Conference on Electronic Commerce, pp. 44–51 (2005)

    Google Scholar 

  8. Borgs, C., Immorlica, N., Chayes, J., Jain, K.: Dynamics of bid optimization in online advertisement auctions. In: Proceedings of the 16th International World Wide Web Conference, pp. 13–723 (2007)

    Google Scholar 

  9. Cary, M., Das, A., Edelman, B., Giotis, I., Heimerl, K., Karlin, A.R., Mathieu, C., Schwarz, M.: Greedy bidding strategies for keyword auctions. In: ACM Conference on Electronic Commerce, pp. 262–271 (2007)

    Google Scholar 

  10. Devanur, N.R., Kakade, S.M.: The price of truthfulness for pay-per-click auctions. In: ACM Conference on Electronic Commerce, pp. 99–106 (2009)

    Google Scholar 

  11. Dobzinski, S., Nisan, N., Lavi, R.: Multi-unit auctions with budget limits. In: FOCS (2008)

    Google Scholar 

  12. Edelman, B., Ostrovsky, M.: Strategic bidder behavior in sponsored search auctions. In: Workshop on Sponsored Search Auctions, ACM Conference on Electronic Commerce, pp. 192–198 (2005)

    Google Scholar 

  13. Edelman, B., Ostrovsky, M., Schwarz, M.: Internet advertising and the generalized second price auction: selling billions of dollars worth of keywords. Am. Econ. Rev. 97 (2005)

  14. Fiat, A., Leonardi, S., Saia, J., Sankowski, P.: Single valued combinatorial auctions with budgets. In: ACM Conference on Electronic Commerce, pp. 223–232 (2011)

    Google Scholar 

  15. Kitts, B., LeBlanc, B.J.: Optimal bidding on keyword auctions. EM 14(3), 186–201 (2004)

    Google Scholar 

  16. Lehmann, B., Lehmann, D.J., Nisan, N.: Combinatorial auctions with decreasing marginal utilities. Games Econ. Behav. 55(2), 270–296 (2006)

    Article  MATH  MathSciNet  Google Scholar 

  17. Leme, R.P., Tardos, É.: Pure and Bayes-Nash price of anarchy for generalized second price auction. In: FOCS, pp. 735–744 (2010)

    Google Scholar 

  18. Mehta, A., Saberi, A., Vazirani, U., Vazirani, V.: Adwords and generalized online matching. J. ACM 54(5), 22 (2007)

    Article  MathSciNet  Google Scholar 

  19. Muthukrishnan, S.: Ad exchanges: research issues. In: WINE, pp. 1–12. Springer, Berlin (2009)

    Google Scholar 

  20. Nisan, N.: Google’s auction for TV ads. In: ESA, p. 553 (2009)

    Google Scholar 

  21. Nisan, N., Rougarden, T., Tardos, E., Vazirani, V. (eds.): Algorithmic Game Theory. Cambridge University Press, Cambridge (2007)

    MATH  Google Scholar 

  22. Orda, A., Rom, R., Shimkin, N.: Competitive routing in multiuser communication networks. IEEE/ACM Trans. Netw. 1(5), 510–521 (1993)

    Article  Google Scholar 

  23. Varian, H.R.: Position auctions. Int. J. Ind. Organ. (2006)

Download references

Author information

Authors and Affiliations

  1. Tel Aviv University, Tel Aviv, Israel

    Asaph Arnon & Yishay Mansour

  2. Google Tel Aviv, Tel Aviv, Israel

    Asaph Arnon

Authors
  1. Asaph Arnon
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Yishay Mansour
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Yishay Mansour.

Additional information

This research was supported in part by the Google Inter-university center for Electronic Markets and Auctions, by a grant from the Israel Science Foundation, by a grant from United States-Israel Binational Science Foundation (BSF), by a grant from the Israeli Ministry of Science (MoS) and by the Israeli Centers of Research Excellence (I-CORE) program, (Center No. 4/11).

Appendix: Characterization of PNE for Two Agents

Appendix: Characterization of PNE for Two Agents

In the following we give a characterization of the PNE for two agents. First, we can derive from Lemma 6 the following property of an equilibrium.

Lemma 14

If b=(b 1,b 2) is a PNE and b 1≥b 2 then b 2≤c 1 and b 1≥c 2.

In the following we discuss a few cases, based on the relationship between c 1 and c 2, and the budgets versus the values.

Lemma 15

Assume that we have two agents with c 2<c 1 and B 2/v 1<N. Then any bids (b 1,b 2) and \((\hat{B}_{1},\hat{B}_{2})\) such that c 1≥b 1≥b 2≥c 2 and b 2=min{v 2,b 1}, are a PNE, and in this case those are the only PNEs where agents submit their true budget and bid at most their value.

Proof

First, note that since B 2/v 1<N we have that u 1>0, since if agent 1 deviates and bids \(b'_{1}=v_{1}-\epsilon\) he is guaranteed to be allocated items, i.e., \(x'_{1}>0\).

We show that there is no PNE with b 1<b 2. For contradiction, assume there is such a PNE. By Lemma 6, for this to be a PNE, we need that b 2≥c 1 and c 2≥b 1. Since the lemma assumes that c 1>c 2, we have b 2≥c 1>c 2≥b 1. Since agent 1 has positive utility, u 1>0, he can deviate to \(b'_{1}=c_{1}\), strictly increasing his allocation \(x'_{1} > x_{1}\), paying p min , and therefore increasing his utility, \(u'_{1}>u_{1}\). This contradicts that agent 1 is best responding. We conclude that b 1≥b 2.

Since b 1≥b 2, by Lemma 6, for this to be a best response we need that c 1≥b 2 and b 1≥c 2. Therefore, c 1≥b 1≥b 2≥c 2. we consider two cases based on the magnitude of B 1/v 2: (i) If B 1/v 2≥N, it implies that c 2=v 2 (since the second agent, when bottom rank, has zero utility). Since b 2≥c 2 and v 2≥b 2, then b 2=c 2=v 2. (ii) If B 1/v 2<N then the second agent has positive utility, u 2>0, since the deviation \(b'_{2}=v_{2}-\epsilon\) would guarantee a positive utility. Therefore, agent 2 would maximize his utility by maximizing his bid while maintaining his bottom rank, i.e., b 2=min{v 2,b 1}. □

Lemma 16

Assume that we have two agents with c 2<c 1 and B 2/v 1≥N. Then:

  1. (1)

    if v 1>v 2 we have v 1=c 1≥v 2≥c 2 and b 1≥b 2 and b 2∈[c 2,v 2] and b 1∈[v 1,c 2].

  2. (2)

    If v 2≥v 1 we have v 2≥v 1=c 1>c 2 and two possible outcomes:

    1. (2a)

      b 2≥b 1 and b 1∈[p min ,c 2] and b 2∈(v 1,v 2], and

    2. (2b)

      b 1≥b 2 and b 1∈[c 2,c 1] and b 2∈[p min ,c 2].

In this case those are the only PNEs where agents submit their true budget and bid at most their value.

Proof

If B 2/v 1≥N then agent 1 critical bid is c 1=v 1, since for any z<v 1, we have B 2/z>N, and hence x 1=0 in the bid vector (z,z) when agent 1 is bottom rank.

For case (1), we have that v 1>v 2. This implies that c 1=v 1>v 2≥b 2, and by Lemma 6 agent 1 prefers the top rank. Therefore b 1≥b 2. In order for agent 1 best response to be top rank we need that b 2∈[c 2,v 2], since v 2<v 1=c 1. In order for agent 2 best response to be bottom rank we need that b 1∈[c 2,v 1].

For case (2), we have that v 1≤v 2. In (2a) we have that agent 1 allocation is x 1=0 and hence his utility is u 1=0. Agent 2 prefers the top rank since b 1≤c 2. Since b 2>v 1 agent 1 prefers the bottom rank. In (2b) agent 1 is top rank and may have positive utility. Agent 1, if he deviates to \(b_{1}'=b_{2}-\epsilon\) will have zero utility. Agent 2 prefers the bottom rank since c 2≤b 1. □

Lemma 17

Assume that we have two agents with c 2=c 1=c then b=(b 1,b 2)is a PNE if and only if either: (i) b 1∈[c,v 1] and b 2∈[p min ,c], or (ii) b 2∈[c,v 2] and b 1∈[p min ,c].

Combining Lemmas 15, 16 and 17 gives a characterization of the PNE for two agents.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Arnon, A., Mansour, Y. Repeated Budgeted Second Price Ad Auction. Theory Comput Syst 54, 453–478 (2014). https://doi.org/10.1007/s00224-013-9472-1

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00224-013-9472-1

Keywords

Search

Navigation