Abstract

Newly-developed large language models (LLM)—because of how they are trained and designed—are implicit computational models of humans—a homo silicus. LLMs can be used like economists use homo economicus: they can be given endowments, information, preferences, and so on, and then their behavior can be explored in scenarios via simulation. Experiments using this approach, derived from Charness and Rabin (2002), Kahneman, Knetsch and Thaler (1986), and Samuelson and Zeckhauser (1988) show qualitatively similar results to the original, but it is also easy to try variations for fresh insights. LLMs could allow researchers to pilot studies via simulation first, searching for novel social science insights to test in the real world.

Abstract

This paper studies the impact of generative AI technologies on the demand for online freelancers, using a large dataset from a leading global freelancing platform. We focus on how the demand effect of the release of ChatGPT differs across various jobs that require different skills or software. Our findings indicate a 14 percent decrease following the ChatGPT introduction in the number of job posts in jobs associated with writing, statistical analysis, engineering, accounting research, and web development when compared to jobs associated with data entry, video services, and audio services, which require more manual tasks. Furthermore, we utilize prior evidence on AI exposure to different occupations and Google Trends to demonstrate that the more pronounced decline in freelancer demand within those specific occupations is linked with their heightened exposure to AI technology, as well as higher general public awareness of ChatGPT's substitutability.

Abstract

In this paper we study the market effects of common ownership in a setting where any ownership structure and any shareholder size is allowed. We depart from the standard reduced form approach of assuming that firms maximize a weighted average of shareholders' portfolios, and instead study the collective choice problem of shareholders head-on. In our model shareholder meetings elect firm managers by one-share one-vote majority rule. Managers differ in their degree of aversion to the negative externality of production. Voting for socially concerned managers therefore provides a mechanism for common owners to direct away the firm from own profit towards industry profit maximization. We show that allowing shareholders of any size to freely diversify their portfolio leads to monopolistic outcomes. Our results have the novel policy implication that the anticompetitive effects of common ownership can emerge even when blockholders are undiversified, but the majority of shares belongs to small diversified shareholders, indicating that small diversified portfolios may also be a threat.

Abstract

The development of blockchains technologies, including smart contracts and cryptographic tokens, have a potential to change the competition between platforms. In this presentation, based on a couple of projects, I will discuss how utility tokens can help new platforms enter the market, and how governance tokens can help platforms to increase social welfare. I will also discuss limitations of these technologies in improving platform strategy and competition.

Abstract

Harassment can be harmful to mental health and reduce creativity and productivity. This paper provides evidence on the productivity effects of online harassment by examining a sudden and unanticipated change on YouTube where dislike counts were hidden from public view. Using detailed channel-level data, I show that this design change had several effects on content creators. First, harassment in form of “dislike attacks”, where users drive up dislike counts unrelated to video quality, decreases following this platform design change. Prior to this design change, women seem to have been more affected by this type of harassment. Second, this design change leads to a persistent increase in productivity for female, relative to male, content creators. Third, there is some evidence of redistribution of demand among content creators.

Abstract

We empirically investigate the relevance of multi-homing in two-sided markets. We build a structural econometric model that allows for multihoming. We then estimate readers’ and advertisers’ demand using an original dataset on the Italian daily newspaper market that includes information on double-homing by readers. The results show that a model that does not allow for multi-homing produces biased estimates on both sides of the market. On the reader side, accounting for multi-homing helps to recognize complementarity between products; on the advertising side, it allows to measure to what extent advertising demand depends on the shares of exclusive and overlapping readers.

Abstract

Online platforms often face challenges being both fair (i.e., non-discriminatory) and efficient (i.e., maximizing revenue). Using computer vision algorithms and observational data from a microlending marketplace, we find that choices made by borrowers creating online profiles impact both of these objectives. We further support this conclusion with a web-based randomized survey experiment. In the experiment, we create profile images using Generative Adversarial Networks that differ in a specific feature and estimate it’s impact on lender demand. We then counterfactually evaluate alternative platform policies and identify particular approaches to influencing the changeable profile photo features that can ameliorate the fairness-efficiency tension.

Abstract

In this paper, we consider competition across two markets where data collected in each market adds value in the other market. A multiproduct firm (1) is a monopolist in market A, and is also active in a second market B where it competes with an established (specialist) firm 2. Firm 1 leverages its presence in both markets to gain a data advantage. In such a market structure, we study the incentive of the specialist firm to share data with its dominant competitor and its implications on profitability, competition and consumer welfare. Our main contribution is to show that the smaller firm has incentives to share data with the dominant firm even for free. This data altruism acts as a strategic device employed by the small firm to lower the intensity of competition by transforming a competitor into a co-opetitor by creating "co-dependence". Specifically, by sharing value-enhancing data, the specialist firm makes the dominant firm a stakeholder in the valuable data collected by the specialist and hence also in its market share. This lowers the competitive intensity of the dominant firm in the competitive market B which enhances (demand and hence also) data collection by the smaller firm. This value creation arising from data altruism substitutes value addition through costly investments made by the dominant firm. A direct managerial implication of this result is that data altruism by the specialist firm can be a win-win outcome for both firms. While data altruism increases market share of the smaller firm, it is not consumer welfare enhancing as it lowers competition (in the secondary market). This result has clear policy implications.

Abstract

The digital advertising industry is characterized by a proliferation of specialized intermediaries helping advertisers in their purchases of online ad space. This study contributes to the analysis of the vertical complementarities in this market by posing and estimating a structural econometric model of how advertisers match to intermediaries. Exploiting novel data and some recent methods in the estimation of many-to-many matching games for large markets, we quantify the value created by the matches, their driving forces and, counterfactually, evaluate the likely effects of the growing concentration among intermediaries. The estimates clearly show that competing advertisers benefit from dealing with a common intermediary and, moreover, offer a precise quantification of various forces including industry specialization, exclusive contracting and diversification.

Abstract

Maintaining good ratings increases the profits of sellers on online platforms. We analyze the role of strategic pricing for ratings management in a setting where a monopolist sells a good of unknown quality. Higher prices reduce the value for money, which on average worsens reviews. However, higher prices also induce only those consumers with a strong taste for the product to purchase, which on average improves reviews. Our model flexibly parametrizes the two effects. This parametrization can rationalize the observed heterogeneity in the relationship between reviews and prices and highlights the dependence of outcomes on the dominant effect. We analytically characterize a seller’s optimal dynamic pricing strategy, long-run profits and consumer surplus, as well as consumers’ speed of learning. Knowledge of the relative strength of price and selection effect is essential for managing ratings with prices. Our results have important implications for the design of rating systems.

Abstract

Online reviews significantly impact consumers' decision-making process and firms' economic outcomes and are widely seen as crucial to the success of online markets. Firms, therefore, have a strong incentive to manipulate ratings using fake reviews. This presents a problem that academic researchers have tried to solve over two decades and on which platforms expend a large amount of resources. Nevertheless, the prevalence of fake reviews is arguably higher than ever. To combat this, we collect a dataset of reviews for thousands of Amazon products and develop a general and highly accurate method for detecting fake reviews. A unique difference between previous datasets and ours is that we directly observe which sellers buy fake reviews. Thus, while prior research has trained models using lab-generated reviews or proxies for fake reviews, we are able to train a model using actual fake reviews. We show that products that buy fake reviews are highly clustered in the product-reviewer network. Therefore, features constructed from this network are highly predictive of which products buy fake reviews. We show that our network-based approach is also successful at detecting fake reviews even without ground truth data, as unsupervised clustering methods can accurately identify fake review buyers by identifying clusters of products that are closely connected in the network. While text or metadata can be manipulated to evade detection, network-based features are more costly to manipulate because these features result directly from the inherent limitations of buying reviews from online review marketplaces, making our detection approach more robust to manipulation.

Abstract

We conduct a large-scale online experiment, where we compare the short- and longer-term effects of fact checking to a brief media literacy intervention (ten tips to spot false information) as a means to debunk “fake news". We find that the effect of fact checking is limited to the specific “fake news" that are being targeted, whereas the media literacy intervention helps users to distinguish between false and correct information more generally, both immediately and two weeks after the intervention. Our results promote media literacy as an effective tool against “fake news", that is furthermore cheap, scalable, and easy-to-implement by social media platforms.

Abstract

We consider a merger between firms that operate on data-connected markets: products are neither substitutes nor complements, but  sales on market A generate data that can be potentially used on market B. We show that the effects of the merger depend on (i) whether data is pro- or anti-competitive on market B, (ii) the existence of frictions regarding the trade of data, (iii) the intensity of competition on market A.

Abstract

In this paper, we build a model of the news market where advertisers choose to allocate their ads between a social media platform and a news website that is the content creator. Our main objective is to evaluate a policy intervention that aims to foster news creation by redistributing revenues from social media to news websites. Such interventions, commonly referred to as news media bargaining codes, were first implemented in Australia in 2021 and have been recently debated worldwide. We focus on a novel trade-off between the higher advertising efficiency of social media and the value of content creation by news websites. We find that a parameter region always exists where the equilibrium level of news creation is socially sub-optimal. Moreover, we show that policy intervention mandated by the bargaining code is always welfare-increasing. We identify conditions when the policy even results in a Pareto improvement over laissez-faire. We show that in all other cases, the improved welfare comes at the detriment of the social media company. The main results hold under different news consumption behaviors and ad pricing mechanisms.

Abstract

We explore the impact of a feature on a large review platform that allows users to more easily observe the ethnicity of black-owned restaurants. We find that the addition of this feature increased demand for black owned businesses. The results suggest that the effect is larger in areas that are higher income, higher education, and in areas that are predominantly white.

Abstract

This paper analyzes the impact of the recent trade tensions between the United States of America and China on the prices of smartphones commercialized in uninvolved countries. The empirical analysis relies on a large dataset of daily prices of smartphones commercialized in Austria, France, Germany, Italy, Spain, and the United Kingdom. The events related to the ongoing trade war offer a natural experiment that can be used to test the causal relationship between trade tensions and the price development of smartphones. The estimation results suggest that the ban on Huawei’s products in the U.S. impacted negatively and significantly the prices of the smartphones sold by the Chinese firm in all of the countries studied. Also, I find that both Huawei and Android increased their market shares in the focal countries. This paper is the first to provide empirical evidence of the consequences of the ongoing trade war on the prices of durable goods in uninvolved countries.

Abstract

Recent literature has shown that the existence of supply and demand-side non-generic complementarities (“demand-side linkages") within ecosystems raises questions about the pertinence of defining a single relevant market comprising substitute products (“substitutability approach"). However, empirical methodologies to measure these linkages and asses the competitive dynamics underpinning them are lacking. Using recent data from internet traffic between the major 246 European digital platforms, we develop such a methodology and test some theoretical findings of the ecosystems literature with major implications for competition and regulatory analysis. We corroborate that demand-side linkages are a non-negligible phenomenon: 18% of these platforms show them. However, unlike what the ecosystems literature predicts, in roughly half of the cases they do not link complementors but platforms competing in at least one market. Finally, while, as expected, we observe demand-side linkages mostly within industry-defined ecosystems, we find evidence of industry-agnostic ecosystems. These could be instigated and orchestrated by platform users instead of by a firm. We conclude that the substitutability approach is not obsolete, but needs to be complemented with alternative approaches in order to i) take into account coopetition within the same relevant market and ii) analyze how the competitive process in one market can impact the welfare generated in another (industry's) market through non-generic complementarities.

Paper available here.

Abstract

Digitization has led to a dramatic increase in the number of new creative products brought to market, challenging traditional product discovery channels such as expert critics. While technological change has also brought alternative sources of information in the form of crowed-based reviews, their nature still raises important questions about the effects of digitization. On the one hand, crowd-based reviews can inform consumers about a larger set of products than the one reviewed by expert critics. On the other hand, crowd reviews – which are usually left anonymously - can potentially suffer from lack of objectivity or even from certain types of pernicious biases. This is in contrast with reviews from professional critics who have strong incentives to appear unbiased due to reputational costs. I rely on a difference-in-differences approach to explore whether crowd-based reviews suffer from gender bias in the context of the movie industry. Using various measures of on-screen female presence and controlling for movie quality, I provide evidence consistent with audience reviewers, males in particular, being biased against female movies. Back-of-the-envelope calculations suggest that such disparities in reviewing can have significant implications for the revenues of movies with larger female presence.

Abstract

In this paper, we build a theoretical model in which an online hosting platform can develop a copyright filter to screen content that contributors wish to upload. The technology is imprecise, since non-infringing material may be incorrectly filtered out. Once the content is hosted on the platform, a right-holder may send a take-down notice if its own automated notice system, also imprecise, finds it to be copyright infringing. We investigate the efficient design of regulation and liability and we find that (i) the right-holder should be given incentives to evaluate fair use when submitting a notice and (ii) the platform should be fined if the take-down to notice ratio is above some predetermined threshold. Such dual system would achieve efficient copyright enforcement without excluding fair-use material.

Abstract

The taxation regime to which online multinational platforms are subject to has been the centre of a fierce debate in recent years. Whereas international consensus has yet to be reached, different countries have taken unilateral actions by announcing versions of a “Digital Service Tax” (DST). Using a dataset that includes stock market prices for a sample of companies potentially subject to the DST, we conduct a financial event study to provide a first evaluation of three groups of proposed taxes (global, EU and US initiatives). We identify a negative and significant impact of the initial European Commission-OECD digital tax proposal on March 2018, driven mainly by the GAFAM equity returns. On the contrary, later announcements as the US digital advertising tax, had a more limited effect. This initial evidence is fundamental to understand the impact of the DST proposal, and it also calls for in-depth further research of the mechanisms and implications.

Abstract

Time-inconsistent internet users neglect future privacy costs and release too much data to digital platforms. We study how regulations that require user consent for data processing affect platform profits, welfare and user surplus, depending on business models and the degree of time inconsistency. Consent mechanisms increase user surplus or welfare only if their design facilitates consent refusal. If platforms can make it difficult to opt out, it may be better for society if the former choose the disclosure level. Voluntary caps on usage, as recently adopted by some platforms, can raise profits by making some users consent to more disclosure.

Paper available here.