From Mad Men to Maths Men: Concentration and Buyer Power in Online Advertising

This paper analyzes the impact of intermediary concentration on the allocation of revenue in online platforms. We study sponsored search documenting how advertisers increasingly bid through a handful of specialized intermediaries. This enhances automated bidding and data pooling, but lessens competition whenever the intermediary represents competing advertisers. Using data on nearly 40 million Google keyword auctions, we first apply machine learning algorithms to cluster keywords into thematic groups serving as relevant markets. Using an instrumental variable strategy, we estimate a decline in the platform’s revenue of approximately 11 percent due to the average rise in concentration associated with intermediary merger and acquisition activity. (JEL C45, D44, G34, L13, L81, M37)

Mixed Mechanisms for Auctioning Ranked Items

This paper deals with the problem of designing and choosing auctioning mechanisms for multiple commonly ranked objects as, for instance, keyword auctions in search engines on Internet. We shall adopt the point of view of the auctioneer who has to select the auction mechanism to be implemented not only considering its expected revenue, but also its associated risk. In order to do this, we consider a wide parametric family of auction mechanisms which contains the generalizations of discriminatory-price auction, uniform-price auction and Vickrey auction. For completeness, we also analyze the Generalized Second Price (GSP) auction which is not in the family. The main results are: (1) all members of the family satisfy the four basic properties of fairness, no over-payment, optimality and efficiency, (2) the Bayesian Nash equilibrium and the corresponding value at risk for the auctioneer are obtained for the considered auctions, (3) the GSP and all auctions in the family provide the same expected revenue, (4) there are new interesting auction mechanisms in the family which have a lower value at risk than the GSP and the classical auctions. Therefore, a window opens to apply new auction mechanisms that can reduce the risk to be assumed by auctioneers.

Reserve price of risk-averse search engine in keyword auctions with advertisers’endogenous investment

Motivated by vigorous development of keyword auctions, this paper analyzes the reserve price policies in keyword auction with advertisers’ endogenous investment and risk-averse search engine. We explore advertisers’ optimal investment and equilibrium bidding strategies , and derive the determination functions where utility-maximizing reserve price and efficient reserve price which maximizes the social welfare satisfy respectively. The results show that advertisers’ equilibrium bidding is monotonously increasing in bidders’ valuations, the number of advertisers, as well as the reserve price. Meanwhile, advertisers’ optimal investment is negatively correlated with reserve price and the number of advertisers. By numerical examples, the utility-maximizing reserve price decreases with the risk aversion parameter and the number of advertisers. Search engine’s expected utility increases with risk aversion parameter and decreases with the number of advertisers. Moreover, we declare that search engine can use reserve price as a regulatory tool to increase the utility. But there is an upper bound on search engine’s utility. It is interesting to find the efficient reserve price equals to zero. Hence there is a trade-off between total efficiency and search engine’s utility by search engine that has incentive to withhold reserve price that would benefit social welfare.

The Promise and Perils of Myopia in Dynamic Pricing With Censored Information

A seller with unlimited inventory of a digital good interacts with potential buyers with i.i.d. valuations. The seller can adaptively quote prices to each buyer to maximize long-term profits, but does not know the valuation distribution exactly. Under a linear demand model, we consider two information settings: partially censored, where agents who buy reveal their true valuations after the purchase is completed, and completely censored, where agents never reveal their valuations. In the partially censored case, we prove that myopic pricing with a Pareto prior is Bayes optimal and has finite regret. In both settings, we evaluate the myopic strategy against more sophisticated look-aheads using three valuation distributions generated from real data on auctions of physical goods, keyword auctions, and user ratings, where the linear demand assumption is clearly violated. For some datasets, complete censoring actually helps, because the restricted data acts as a "regularizer" on the posterior, preventing it from being affected too much by outliers.

Boost Keywords Conversion of Search Engine

more and more people are used to purchase online. And companies also want to pay for the keyword on big search engine, such as Baidu, Google. So the strength of the competition plays a significant role in the efficiency of an online advertising campaign. Models for keyword auctions usually assume that, the value of a click is fixed and independent of the other sponsored ads in the auctions result, for the advertiser. In china, the big Online shopping mallTmall. Consumers can search products and service through Tmall Search Engine, a search engine allows consumers search in the Tmall. And we employ a zero-inflated Poisson regression to model the conversion of keywords.

Boost Keywords Conversion of Search Engine

more and more people are used to purchase online. And companies also want to pay for the keyword on big search engine, such as Baidu, Google. So the strength of the competition plays a significant role in the efficiency of an online advertising campaign. Models for keyword auctions usually assume that, the value of a click is fixed and independent of the other sponsored ads in the auctions result, for the advertiser. In china, the big Online shopping mallTmall. Consumers can search products and service through Tmall Search Engine, a search engine allows consumers search in the Tmall. And we employ a zero-inflated Poisson regression to model the conversion of keywords.