Faculty and Research

Widely used online “trust” authorities issue certifications without substantial verification of recipients’ actual trustworthiness. This lax approach gives rise to adverse selection: the sites that seek and obtain trust certifications are actually less trustworthy than others. Using a new dataset on web site safety, I demonstrate that sites certified by the best-known authority, TRUSTe, are more than twice as likely to be untrustworthy as uncertified sites. This difference remains statistically and economically significant when restricted to “complex” commercial sites. In contrast, competing certification system BBBOnline imposes somewhat stricter requirements and appears to provide a certification of positive, albeit limited, value.

Online advertisers frequently fall victim to dishonest, tech-savvy publishers. Here's a sampling of common scams with some advice on how to outwit their perpetrators.

This paper studies the adult online entertainment industry, particularly the consumption side of the market. In particular, it focuses on the demographics and consumption patterns of those who subscribe to adult entertainment websites. On the surface, this business would seem to face a number of obstacles. Regulatory and legal barriers have already been mentioned. In addition, those charging for access to adult entertainment face competition from similar content available without a fee. In the context of adult entertainment, free access offers consumers an extra benefit: online payments tend to create records documenting the fact of a customer's purchase; consumers of free content may feel more confident that their purchases will remain confidential. More broadly, measured levels of religiosity in American are high. On the other hand, social critics often argue that the rise of Internet pornography is contributing to a coarsening of American culture. Do consumption patterns of online adult entertainment reveal two separate Americas? Or is the consumption of online adult entertainment widespread, regardless of legal barriers, potential for embarrassment, and even religious conviction?

The Internet is sold to advertisers as a highly measurable medium that is the most efficient way to target exactly the right customers. But online advertising is also easily subverted-letting fraudsters claim advertising fees for work they did not actually do. The trickiest frauds deceive advertisers so effectively that measurements of ad effectiveness report the fraudsters as exceptionally productive and high quality, rather than revealing that their traffic was actually worthless. This is a quiet scandal. In a time of tightening ad budgets, losses to advertising fraud come straight from the bottom line-but savings can be equally dramatic.

Here's a look behind the veil-an explanation of ad practices that have cheated even the Web's largest advertisers. Advertising scams take plenty of victims, both witting and not, but I offer strategies to help determined marketers protect themselves.

I offer five rights to protect advertisers from increasingly powerful ad networks-avoiding fraudulent charges for services not rendered, guaranteeing data portability so advertisers get the best possible value, and assuring price transparency so advertisers know what they're buying. I explain the need for these rights by presenting specific practices causing particular concern.

Online advertisers face substantial difficulty in selecting and supervising small advertising partners. Fraud can be well hidden, and limited reputation systems reduce accountability. But partners are not paid until after their work is complete, and advertisers can extend this delay both to improve detection of improper partner practices and to punish partners who turn out to be rule-breakers. I capture these relationships in a screening model with delayed payments and probabilistic delayed observation of agents' types. I derive conditions in which an advertising principal can set its payment delay to deter rogue agents and to attract solely or primarily good-type agents. Through the savings from excluding rogue agents, the principal can increase its profits while offering increased payments to good-type agents. I estimate that a leading affiliate network could have invoked an optimal payment delay to eliminate 71% of fraud without decreasing profit.

With forces both supporting and opposing zero prices, typical Internet-related activities—like surfing the web, web searches, and e-mail, along with behind-the-scenes practices like domain names and the allocation of IP (Internet Protocol) addresses—present a natural context to reevaluate our sense of the tradeoffs that arise between free and a positive price. In this piece, I offer a series of specific examples of resources offered without charge, for a positive price, or for a flat fee ("all-you-can-eat"). I conclude by assessing the characteristics that shape pricing structure for these resources.

At first glance, online advertising seems to be as measurable a medium as any ever invented.  Advertisers can count how many times an ad was sent, then measure sales – yielding an analysis that seems to report the value of an online ad campaign.  But the reality is considerably more complicated. 

Online advertising presents remarkable efficiencies — better targeting, improved measurement and greater return on investment. Yet there are challenges, particularly when networks of intermediaries place ads through convoluted relationships, and all the more so when small advertisers cannot effectively negotiate terms dictated by advertising powerhouses. The result is a troubling mess of ads gone wrong — advertisers charged in ways they didn’t fairly agree to, and on terms they didn’t meaningfully accept. 

But online advertising doesn’t have to be a wild west. I propose five specific rights advertisers should demand as they buy online placements.

"Typosquatting" is the practice of registering domain names, identical to or confusingly similar to trademarks and famous names, in hopes that users will accidentally request these sites -- whereupon they will receive, typically, advertisements.  This piece presents the basic typosquatting business model, based on my analysis of more than 80,000 typosquatting domain names.  I analyze the advertising intermediaries that make typosquatting profitable, and I assess the legislation and litigation that are beginning to put a check on this practice.

We investigate the “generalized second-price” auction (GSP), a new mechanism used by search engines to sell online advertising. Although GSP looks similar to the Vickrey-Clarke-Groves (VCG) mechanism, its properties are very different. Unlike the VCG mechanism, GSP generally does not have an equilibrium in dominant strategies, and truth-telling is not an equilibrium of GSP. To analyze the properties of GSP, we describe the generalized English auction that corresponds to the GSP and show that it has a unique equilibrium. This is an ex post equilibrium, with the same payoffs to all players as the dominant strategy equilibrium of VCG.

We examine sponsored search auctions run by Overture (now part of Yahoo!) and Google and present evidence of strategic bidder behavior in these auctions. Between June 15, 2002, and June 14, 2003, we estimate that Overture's revenue from sponsored search might have been higher if it had been able to prevent this strategic behavior. We present a specific alternative mechanism that could reduce the amount of strategizing by bidders, raise search engines' revenue, and also increase the overall efficiency of the market. We conclude by showing that advertisers' strategic behavior has not disappeared over time; rather, such behavior remains present on both search engines.

How should players bid in keyword auctions such as those used by Google, Yahoo! and MSN? We consider greedy bidding strategies for a repeated auction on a single keyword, where in each round, each player chooses some optimal bid for the next round, assuming that the other players merely repeat their previous bid. We study the revenue, convergence and robustness properties of such strategies. Most interesting among these is a strategy we call the balanced bidding strategy (bb): it is known that bb has a unique fixed point with payments identical to those of the VCG mechanism. We show that if all players use the bb strategy and update each round, bb converges when the number of slots is at most 2, but does not always converge for 3 or more slots. On the other hand, we present a simple variant which is guaranteed to converge to the same fixed point for any number of slots. In a model in which only one randomly chosen player updates each round according to the bb strategy, we prove that convergence occurs with probability 1. We complement our theoretical results with empirical studies.

The Chinese government has made few official statements about its filtering of Internet content, but this report explores the scope, depth, and various methods used to selectively bar to Internet access through networks in China.

Read the news of recent computer security guffaws, and it's striking how many problems stem from online advertising. Advertising is the bedrock of web sites that are provided without charge to end users, so advertising is everywhere. But advertising security gaps are equally widespread: from "malvertisement" banner ads pushing rogue anti-spyware software, to click fraud, to spyware and adware, the security lapses of online advertising are striking.

During the past five years, I have uncovered hundreds of online advertising scams defrauding thousands of users—not to mention all the web's top merchants. This chapter summarizes some of what I've found—and what users and advertisers can do to protect themselves.

We model competing auctions for online advertising, with attention to the participation costs that limit advertisers' interest in using small ad platforms. When participation costs are large relative to the volume of traffic an ad platform can offer, an advertiser may forego use of an ad platform that the advertiser otherwise finds profitable. Mergers between ad platforms can increase advertiser welfare if the resulting click-through rate and volume of traffic are sufficiently improved relative to the offerings of the ad auctions when separate. When there is an insufficient improvement, such mergers can harm advertisers.

We use a unique database of every SSRN paper download over the course of seven years, along with detailed resume data on a random sample of SSRN authors, to examine the role of demographic factors, career concerns, and social comparisons on the commission of a particular type of gaming: the self-downloading of an author's own SSRN working paper solely to inflate the paper's reported download count. We find significant evidence that authors are more likely to inflate their papers' download counts when a higher count greatly improves the visibility of a paper on the SSRN network. We also find limited evidence of gaming due to demographic factors and career concerns, and strong evidence of gaming driven by social comparisons with various peer groups. These results indicate the importance of including psychological factors in the study of deceptive behavior.

We develop a model of online advertising in which each advertiser chooses from multiple advertising measurement metrics - paying either for each click on its ads (CPC), or for each purchase that follows an ad‐click (CPA). Our analysis extends classic auction results by allowing players to make bids using two different pricing schemes, while the driving information for bidders' endogenous selection - the conversion rate - is hidden from the seller. We show that the advertisers with the most productive sites prefer to pay CPC, while advertisers with lower quality sites prefer to pay CPA - a result that may be viewed as counterintuitive since low quality sites cannot proudly tout their conversion rates. This result holds even if an ad platform's assessment of site quality is correct in expectation. We also show that by offering both CPC and CPA, an ad platform can weakly increase its revenues compared to offering either alternative alone.

How should players bid in keyword auctions such as those used by Google, Yahoo! and MSN? We model ad auctions as a dynamic game of incomplete information, so we can study the convergence and robustness properties of various strategies. In particular, we consider best-response bidding strategies for a repeated auction on a single keyword, where in each round, each player chooses some optimal bid for the next round, assuming that the other players merely repeat their previous bids. We focus on a strategy we call Balanced Bidding (BB). If all players use the BB strategy, we show that bids converge to a bid vector that obtains in a complete information static model proposed by Edelman, Ostrovsky, and Schwarz. We prove that convergence occurs with probability 1, and we compute the expected time until convergence.

I analyze all auctions since the inception of the Google Answers service, and I find notable trends in answerer behavior: More experienced answerers provide answers with the characteristics askers most value, receiving higher rankings as a result. Answerers’ rate of earnings increases in experience – showing both selection effects and learning on the job. Answerers who focus on particular question categories provide answers of higher quality (“specialization”) but earn lower pay per hour (perhaps “lack of versatility”). Answers provided during the business day receive higher payments per hour (a compensating differential for working when outside options are most attractive), but more experienced answerers tend to forego these opportunities.

A new dataset collected by the author allows estimation of the effect on book sales of promotional listing on Amazon’s editorial discretion pages. Following Goolsbee and Chevalier (2001), sales quantities are inferred from sales rank data freely available on Amazon’s web site, and an automated system tracks which books are promoted when, where, and how often. The results indicate that promotion of books on editorial discretion pages within Amazon’s web site yields increases in sales, and more frequent promotion of a book is associated with larger increases in sales. Increases in sales are greatest for newly-released hardcover books; increases are larger for childrens’ books, books in stock, and books more favorably priced at Amazon than at its foremost competitor, Barnes & Noble. Increases in sales are larger during the period between Thanksgiving and Christmas than before or after the holiday season, and promotion has a  larger effect when editorial discretion pages feature only a few books than when they feature many. Finally, the average short-run effect of promotion on one of Amazon’s editorial discretion pages is found to be roughly one third as large as the effect of an appearance in the New York Times Book Review, and the annual sum of Amazon’s editorial discretion promotional activities shows a total short-run impact on sales roughly three fifths as large as the totality of annual Times book reviews.

PRT vehicles -- often called "driverless taxis" -- sought to combine the best characteristics of cars, taxis, and trains, while adding features unavailable in any existing transportation system. Like cars and taxis, PRT vehicles carried small groups -- often just a single passenger -- with no need to wait for a shared vehicle to arrive or for others to board. Yet PRT followed train systems in using an exclusive right of way that avoided delays from other traffic.  Where would such systems be most useful?  Could system designers successfully compete with well-established networks of trains, buses, cars, and roads?