Actions and Signals

Featured

Tags

, , , , ,

In the canonical signaling model, a sender holding private information about her type chooses an action to change the beliefs of uninformed receivers about her type, e.g., a (future) worker chooses a higher education level to signal her ability. In other words, the chosen action serves as a signal of her privately known type. There are numerous examples of real-world settings that feature signaling, however, that work differently.

Suppose that, instead, the sender privately observes her type, and choses unobservable actions that together with her type determine an observable outcome, which acts then as a signal of both the sender’s type and her chosen actions. For example, consider the college attended by a worker. A more prestigious institution serves as a signal of both the innate ability or skill of the worker as well as of her effort in high-school and in studying for standardized tests. The worker, however, clearly does not choose her college, but her effort in high-school, volunteering, test prep, etc.

In “Actions and Signals,” co-authored with my colleague Mike Waldman, we introduce a generalized model of signaling that captures strategic incentives in these environments. We show that in equilibrium, a fundamentally different behavior than in the canonical signaling model, a special case of our framework, can arise.

If an action and a signal are one and the same, receivers know the sender’s action, and infer her type from the signal. In this scenario, over-investment, i.e. the sender choosing an action (=signal) that exceeds its efficient level, ensues (independent of whether the action is productive or not). This is because the sender gets rewarded for her action (as observed by receivers) but chooses an even higher level of the action to communicate that she is a higher type sender.

On the other hand, if an unobservable action and the sender’s type combine to generate an observable outcome that serves as a signal of both the action and her type, a different logic applies. When the sender chooses a higher level of the action, the signal increases. Receivers attribute some of this increase to a higher action and some of the increase to a higher type. As a result, equilibrium behavior of the sender depends on how the action and the sender’s type contribute to the observable signal vs. how they contribute to the sender’s output (for which receivers pay). In fact, if the action chosen by the sender is less important (relative to her type) for the signal than her output, under-investment results.

In terms of education signaling, this means that in cases where effort in high-school is of importance for future job-performance, the worker may choose an inefficiently low level of effort if it does not equally increase the quality of the college she will attend. We discuss education as well as third-party reviews such as rating systems or certification as two of the (many) applications of our model, and argue why the specific determinants in a signaling environment will be the drivers of efficiency in terms of over- or under-investment by the sender (or potentially both in the case of multiple actions).

We furthermore introduce incomplete sender information that allows us to span an environment with signaling (sender perfectly knows her type) and signal jamming (sender is uninformed) as the extreme cases, and show that uninformed sender behavior may be more efficient than that of a well-informed one. Finally, we discuss the tradeoffs between multiple actions and deal with productive signals, i.e., when the signal realization itself increases sender output.

Search Platforms: Big Data and Sponsored Positions

Featured

Tags

, ,

Search platforms (that rely on different business models) such as Amazon, Google or Yelp possess an abundance of consumer-specific data from past searches, purchases, and the behavior of ‘similar’ consumers. When a consumer searches for a product (or service) using a platform’s interface, the platform can invoke this knowledge when listing search results. In creating this list, it has become customary for platforms to sell prominently featured slots, so-called sponsored slots or positions, to firms hoping to steer consumers towards buying their products. Google founders Sergey Brin and Larry Page were on record, as early as 1998, warning of the implications of search advertising: “We expect that advertising funded search engines will be inherently biased towards the advertisers and away from the need of the consumers.” In fact, some of today’s consumers are reluctant to click on sponsored search results to circumvent steering by firms, as they believe non-sponsored (organic) search results promise to be a better match for their preferences.

In “Search Platforms: Big Data and Sponsored Positions,” co-authored by Maarten Janssen, Thomas Jungbauer, Marcel Preuss and Cole Williams, we argue that the issue at hand is more complicated. In fact, we introduce a game-theoretic model analyzing the interaction between a search platform, consumers and firms to show that, when the number of firms is large, it is in the best interest of the consumer to click on sponsored positions, while the introduction of sponsored positions in the first place may either benefit or harm consumers, depending on the platform’s competing objectives. For example, if the platform (also) maximizes sales commission revenueas is common practicesponsored slots are beneficial for the consumer. This is because platforms face an incentive to allocate sponsored positions to firms that constitute good matches for consumers in order to increase the revenue from auctions for sponsored positions. On the other hand, obfuscation of organic slots plays a big role in optimal search platform behavior. First, it reduces the consumer’s option value from searching beyond sponsored positions, and, second, it increases the number of positions a consumer inspects if they continue to search among organic slots, thereby increasing the probability of a sale being made.

Our results persist independently of the allocation of consumer-specific knowledge between the search platforms and firms and the variation in prices among firms. Among the major complications to overcome when characterizing optimal platform behavior is that consumers, when inspecting a firm in a given slot, learn about firms in slots they have not yet inspected when they understand the search platform’s algorithm. As a result, we can show that in environments with few firms only, the optimal algorithm of the search platform always depends on the intricacies of the specific problem, and no general rule for the ranking of search results can exist. However, when the number of firms is largeas is customary when dealing with search platformsthe platform behavior as described above is optimal almost surely. In order to prove our results with many firms, we adapt a result from the literature on social learning, the so-called mixing property of stochastic processes, for consumer search problems.

George Santos and the ambiguous effects of resume padding: the costs and benefits of lying and misrepresentation in the job market

Featured

Tags

, , , ,

Elected congressman George Santos has recently been subject of public scrutiny after it became known that he has repeatedly lied about his life achievements and personal background. His case, however, is hardly a one off. Resume padding, the misrepresentation of one’s personal history to increase job market attractiveness, is a common place phenomenon. While the detection of resume padding almost always leads to a breakdown of relationships due to an irrevocable loss of trust, the social effect of resume padding is more complex as explained in our paper “Self-Reported Signaling” (w Michael Waldman).

New York Congressman elect George Santos, preparing to take his seat in January, is facing strong headwinds amid calls to resign before even taking office. These demands came after it was revealed by the New York Times on December 19 that Mr. Santos has repeatedly and blatantly lied about his education and work credentials, charitable undertakings  and even his personal background. Journalists were neither able to verify his self-proclaimed working experience on Wall Street for Citigroup and Goldman Sachs, nor is there any record of him ever attending Baruch College as claimed on his biography. There is also hardly any evidence for his involvement in a dog rescue charity organization, Friends of Pets United, an activity he heavily leveraged on the campaign trail. Even claims in his online biography (now taken down) that his grandparents fled Jewish persecution in Europe have since been called in question.

After initially accusing the New York Times through his lawyer of an unsubstantiated vendetta against his persona, Mr. Santos has since apologized for “embellishing his resume,” and “a poor choice of words” in multiple interviews (New York Post, City and State New York) without taking responsibility for misrepresenting his life accomplishments and even his heritage. It is without question that Mr. Santos’ actions show a grave lack of respect for his constituents as well as at least an indifference towards others, such as people personally affected by the Holocaust.

The willingness to lie so blatantly for his own benefit without any regard for consequences is rightfully interpreted by many as a major character flaw for a public servant. Many raise questions how voters and Mr. Santos’ peers alike would ever be able to take his word for granted, and others ask whether his actions may even warrant criminal prosecution (NBC).

While I personally support these viewpoints and believe that Mr. Santos’ actions do indeed necessitate a legal sequel, particularly as it can be argued that his lying directly affected donations towards his candidacy, Mr. Santos’ story is blatant but hardly unique. In fact, he is only one among many who helped themselves to a position of power through misrepresentation of background and achievements. Resume padding is a common phenomenon employed as tactics by Chief Financial Officers, College Football Coaches, and even Prime Ministers. The detection of such a lie frequently triggers resignation or termination and even lawsuits. These are understandable consequences of the loss of trust in a person having catapulted herself into a position of power and decision making, and often, wealth.

Social consequences of resume padding, however, are much more involved, and potentially ambiguous. If lying about achievements and background is a common phenomenon, decision makers such as employers or even voters in turn will put less emphasis on these credentials when making hiring or promotion decisions. In turn, it becomes less attractive for a job-seeker or political candidate to engage in amassing these costly credentials, especially for those who face a harder prospect of doing so in the first place.

The standard theory of signaling teaches us that whenever engagement in costly activities such as education allows for inferences about personal ability, those who are vying for opportunities will overinvest in these activities/credentials. In other words, job seekers and political candidates will over-educate, build an overly packed working resume or engage in too many extra-curricular or charitable activities. By the logic above, the presence of resume padding, i.e., lying about these credentials, then lowers this overinvestment.

My co-author Michael Waldman and I detail this argument in our paper “Self-Reported Signaling,” forthcoming in the American Economic Journal: Microeconomics. Note that our theory relies on the (realistic) assumption that fact-checking a resume is costly, as otherwise the truth would be readily available to everyone. (Mr. Santos story strongly supports this assumption as it took investigative journalism by the New York Times to uncover inconsistencies in his story.) It follows that the overall effect of resume padding depends on the trade-off between the cost of mismatch, auditing and the breakdown of relationships with the benefit of a reduction in the over-investment in costly activities. While blunt misrepresentation such as in Mr. Santos’ case likely leads to welfare loss due to the irrevocable loss of trust, the social effect of more moderate but systematic resume padding is not necessarily negative.

South-Korea’s new regulation on in-app purchases

Tags

, , , ,

I am grateful to BeFM, South Korean public radio, to give me the chance to comment on their show “Morning Wave in Busan” on South Korea’s new legislation trying to rein in the control of tech giants such as Apple and Google over payments for content in their respective app stores. We talked about increasing competition, and thereby potentially increasing quality and/or decreasing prices for consumers, while increased fraud and privacy protection issues will likely bring about a challenge. Find the full interview HERE!

Online Advertising, Data Sharing, and Consumer Control

Tags

, ,

When advertisers share consumer-relevant data (e.g., that they have visited the advertiser’s website) with ad exchanges, the facilitators of targeting consumers across the web and online advertisement auctions, these ad exchanges do not offer to share this information with other rival advertisers in the same product category. This is even true if all parties in the market (ad exchanges, advertisers, consumers, publishers) were be better off in case more information was shared.

We identify the strong property rights of advertisers (website owners) as the culprit of this undersharing of information (ad exchanges cater to advertisers with restrictive data sharing policies), and show that small tweaks such as endowing consumers with easier ways not to be tracked can even worsen that situation. Instead what it takes is a system that weakens advertisers’ property rights over consumer-generated information. When consumers are for example allowed to directly share purchase intent in a product category with ad exchanges, advertisers in equilibrium share more information themselves enabling very efficient “cross-targeting” of consumers. We show that even highly criticized initiatives such as those by Google and Apple to abandon third-party cookies may improve consumer welfare by altering the current system.

The Strategic Decentralization of Recruiting

Tags

, ,

Why do firms decentralize recruiting? Larger firms are impersonal, slow to act, and struggle to leverage expertise. Thus, we expect larger firms to decentralize hiring more often. This intuition, however, is flawed (thomas-jungbauer.com/research/).

A 2017 study by Mercer indicates that, for the large part, the opposite is true. In “The Strategic Decentralization of Recruiting,” co-authored with my colleague Yi Chen, we show how market power in professional labor markets may justify this behavior.

The bigger a firm, the larger its market power, i.e., its ability to dampen wages across the market. On the other hand, when a firm delegates hiring to divisions, it increases wages through competition among its divisions.

Outside competitors, however, foresee that this increased level of competition makes aggressive bidding less worthwhile, and reduce their own bids for workers. As a consequence, the delegating firm may hire more skilled workers without overly raising wages.

We describe this tradeoff between market and commitment power, and the resulting delegation patterns as a function of firm size and its productivity. We find that both highly productive as well as non-productive firms never decentralize recruiting in a given labor market.

Branding Vertical Product Line Extensions

Tags

, , ,

“Branding Vertical Product Line Extensions,” a paper with Christian Schmid from U Vienna. We build a stylized model to analyze the optimal branding decision of a firm expanding its product line as a function of the vertical direction of the extension and the level of competition.

We assume that consumers are not only affected by the quality of the product they consume, but every product sold under the same brand. This relation may arise due to product confusion, conspicuous consumption, etc. The main result of the paper is that firms may employ branding as a commitment device to soften quality competition.

Under competition, firms do not recoup their full investment in R&D, advertisement, factories, etc. Therefore, they opt for the branding regime that minimizes investment. Interestingly, that means that firms often choose the branding regime preferred by consumers when expanding downwards but not when expanding upwards.

Our paper provides a potential explanation why in some markets we see firms successfully expanding downwards under an umbrella brand and upwards with individual brands (think Mercedes vs. Toyota/Lexus) and vice versa in other markets (GAP/Old Navy vs. Adidas).        

The Organization of Innovation: Property Rights and the Outsourcing Decision

Tags

, , ,

Why do firms outsource research and development for some products while they opt to conduct R&D in-house for similar ones? In our new paper, Sean Nicholson, June Pan, Michael Waldman and I argue that companies want to protect their existing product portfolio. If a firm already successfully operates in a given product category, it is more reluctant to relinquish control of the research and development of new products in order to limit cannibalization of their existing successful products.

We build a novel theoretical model and show that a firm is more likely to conduct R&D for a new product in-house if a.) the company already sells a product in the same product category, b.) the longer the patent on the existing product, and c.) the higher the market share of the existing product are. Data from the pharmaceutical industry strongly supports our findings. We control for various measures of competition and patent existence to exclude simple category specific expertise as an explanation.

Self-reported actions, signaling, and auditing

Tags

, , , , , , ,

Actions that affect the value of a service are often self-reported rather than publicly observable. The diligence of a contractor, the education level of job applicant, or the true mileage of a used car are typically reported by the seller. This opens the door for lying and misrepresentation.

In “Self-Reported Actions, Signaling, and Auditing,” my co-author Mike Waldman and I present a model in which multiple receivers bid for the service of a sender, the value of which depends on a action taken by the sender. Instead of the action itself, receivers only observe a message reported by the sender indicating which action was taken. Receivers may opt for costly auditing to verify that the message matches the action.

We find that lying may increase social welfare when the action serves as a signal of a desirable trait of the sender. A positive likelihood of misrepresentation lowers the value of the action as a signal, and therefore counteracts the well-known over-investment result in the signaling literature. Therefore, factors that promote misrepresentation, such as a lower disutility of lying or a higher auditing fee, may increase social welfare.

This result stands in stark contrast to cases in which the action does not signal the sender’s type. We also find that the level of auditing is inverse U-shaped in the probability of the sender being dishonest, and that receivers may audit more often if the action does not serve as a signal, despite gaining less information when auditing. We apply our insights to education signaling, college applications, and odometer fraud in the used car market.

Find the full text paper HERE . I will present it at this year’s virtual editions of the EEA and the ESWC.

It’s not about flattening the curve. Let’s get rid of it!

Tags

, , , ,

By now you have surely heard about “flattening the curve” and seen the pretty picture that typically comes with it. A variation even made it into the New York Times last week.

covid_curve

The original argument/idea behind this picture crudely goes as follows. The COVID-19 pandemic will be over once a critical fraction (“the point of herd-immunity”, estimations vary but converge typically around two thirds) of the population develops immunity against the virus. Comparing death rates from regions that were surprised and thus overwhelmed by the virus (e.g. the Wuhan region in China or Lombardy in Italy) with those who were well prepared and hence faced a smaller amount of cases per capita (e.g. South Korea or the remainder of China) teaches us that pushing the number of serious cases below the capacity of the local healthcare system saves a ton of lives.

In order to achieve herd-immunity, however, the fraction of the overall population that gets infected with the virus does not change. This is depicted by the fact that the areas below the curves above are approximately equal. While this sounds like an intriguing argument, it is not realistic. Crude calculations show that it could take years to decades until we reach that point.

What we are (or should be) after, is, in fact, a reduction of the area under the curve itself, not stretching the curve over years. Instead of achieving herd-immunity, our goal should be to eradicate the virus as quickly as possible. While we may hope that summer, a vaccination or medications will put an abrupt end to the pandemic, these are hypotheticals that are far from certain to manifest. China with its rigorous policy of social distancing is the prime example after which other countries should model their response to the virus. The Washington Post published a neat little simulation, misleadingly also referring to flattening of the curve , that exemplifies the argument and generates these neat gifs:

covid_sim Not only does the curve get stretched by means of social distancing but the area under the curve, i.e. the number of overall infected people, diminishes. Since I trust that you neither want to see exorbitant numbers of older people and the odd younger ones die, nor that you want to sit at home for the next fifteen years, let’s give it some effort for a change and stop spreading dubious conspiracy theories.

Waiting for a vaccination or effective medications is a dangerous game.  Social distancing does not only save lives, it also allows to get back to our lives significantly faster. If we want to eradicate the curve, however, governments need to take more decisive (and also painful) action than if we wanted to stretch the curve. In particular, it is insufficient to just provide recommendations and it is certainly counter-productive to propagate herd-immunity. No country should witness world-war like scenes in its hospitals due to the indecisiveness of world leaders.