Working Paper

Piracy or Fair Use? Evaluating the Welfare of Software Copyright Takedowns: Theory and Evidence from GitHub

Göran Skogh Award for the best paper by a young scholar at the 2021 European Association of Law and Economics Conference

conference presentation: CEA 2021, CES 2021, SIOE 2021, AMES 2021, IEA 2021, GAMES 2021, YES 2021, EALE 2021; my 20-min video presentation in Oct. 2021

  • This paper presents a revealed preference approach to evaluate the welfare of the widely adopted content takedown policy that secures copyright in cyberspace. I build an analytical model to distinguish two hypotheses-the piracy hypothesis versus the fair use hypothesis. The piracy hypothesis predicts that copying is an unproductive investment to steal content; it reduces original contributions and copyright owners' revenue. The fair use hypothesis predicts that copying is a cost-saving method to expand and transform content; it invites original contributions and has little effect on copyright owners' revenue. The takedown policy is only socially beneficial in removing the piracy type of reproduction, but it is welfare reducing in removing the fair use type of reproduction. An event study design using data from GitHub provides evidence that reproduction on this platform is most likely driven by the motive for fair use, and enforcing the takedown policy is thus inefficient. I conclude by discussing possible optimal policies.

  • This paper presents a continuous-time dynamic model of market adoption of cybersecurity. Individual firms choose whether and when to make a precautionary investment in self-protection against the evolving security risk of direct attack and indirect contagion. The equilibrium adoption path has a ``tipping point'': firms will invest to get protected all at once when a critical mass of data breach has been reached. Perhaps surprisingly, in the presence of externalities, the market equilibrium path is found to be exactly the socially optimal path. This welfare theorem is robust under several extensions of the model, including general demand for security, in-house security team, weak targets, security upgrade, and ex-post recovery. The findings cast doubt on the government interventions that are justified by externalities in the market for information security.

Decentralizing Content Moderation (with Adrian Segura and Dongkyu Chang)

  • In this paper, we apply bargaining game and mechanism design to the study of platform content moderation. We start by analyzing and comparing four popular moderation procedures: centralization, flagging, counter-notice, and voting. Among the four, counter-notice minimizes the over-removal error, voting minimizes the under-removal error, and centralization minimizes the procedural cost. We generalize the welfare comparison to a mechanism design setup where the platform specifies the assignment and cost rules. The error-free mechanism that implements the efficient assignment rule is a repeated counter-notice procedure similar to Wikipedia's Talk Page. The consumer optimal mechanism that trades off errors with costs is instead centralized platform adjudication, which is common in spam filters.

Democracy and Internet Control: Theory and Evidence from Transparency Reports (with Shreyas Meher)

  • Internet control has long been considered a feature of authoritarian regimes alone. Drawing data from Google and Twitter transparency reports, we observe that democratic countries remove an equal amount of content as their authoritarian counterparts. The distinction between the two regimes lies not in the quantity but in the method of content removal. Democracy refrains from government takedown and instead delegates the removal right to the users. This paper conjectures that politicians' reputation concern is the key to understanding this phenomenon. To that end, we develop a political agency model that explains the stylized facts and derives testable hypotheses. Using the timing of elections as a natural experiment, we provide supporting evidence that the takedown requests from democratic governments decreased significantly as the election approached. This reputation effect is not observed in authoritarian regimes or other types of requests.

  • Multinational corporations must overcome a number of risks for long-term investments to be worthwhile. The holdup problem and the obsolescing bargain are sometimes conflated but can be reduced through several international institutions, including diplomatic protection, international investment arbitration, and political risk insurance. But which institutions best reduce which risks? We analyze and compare all three investment institutions with a formal model that distinguishes the holdup and bargaining problems. Our models demonstrate that under the broadest range of political risks, insurance yields a consistently higher total investment and higher bargaining outcome for MNCs. In contrast, arbitration generates more investment in low-risk environments but can lead to overinvestment. Our model suggests that greater use of insurance could improve the investment effects of either of the other institutions with greater consistency than diplomatic protection and without some of the pernicious effects of arbitration.

When Do People Fact Check? An Experimental Study (with Dongfang Gaozhao and Valeria Bodishtianu)

  • This paper investigates the cause and consequence of fact-checking. In an online experiment, we asked subjects to evaluate news veracity and varied two experimental conditions: (1) the opportunity to receive fact-checking results and (2) bonus payment for accuracy. We test three competing theories for fact-checking behavior: value of information (VoI), limited attention (LA), and motivated reasoning (MR). We find that monetary incentives do not promote fact-checking. Prior awareness of the news and perceived easiness in determining news authenticity significantly reduce fact-checking. Democrats are more likely to fact-check on the news aligning with Republicans' ideology, suggesting a tendency to seek information when there is a need to defend one's pre-existing belief. Overall, our results contradict VoI, show mixed evidence for LA, and support MR. When available, fact-checking consistently improves subjects' accuracy in evaluating news veracity by over 40%.

  • Online intermediary liability has become a topic of increasing importance as the landscape of the Internet becomes more complex and omnipresent. This paper provides a political economy explanation for why American intermediary liability rules are different for online speech and online copyright. Section 230 of the Communications Decency Act famously grants immunity to platforms for hosting third-party content, whereas Section 512 of the Digital Millennium Copyright Act notably holds platforms liable if they do not respond expeditiously to notices of alleged infringement. We examine the origin and the development of the two pieces of legislation, and how the special interest politics affect the law-making process. By analyzing public records of congressional lobbying spending, we compare the lobbying efforts expended by key stakeholders in both areas and find a sharp contrast between the spending for the status quo and the spending against it. We also use Google as a case study to illustrate how Internet intermediaries’ political activism shaped the two areas of regulation differently. We show a strong co-movement of Google’s lobby spending and the frequency of its litigated cases. Our study draws on key economics concepts such as opportunity cost, free-riding, and the logic of collective action. Through the lens of political economy, our findings provide actionable suggestions for Section 230 and Section 512 reformers.

  • This paper presents an empirical test of disputants' settlement behavior using online copyright notices. The Section 512(c) notice-and-takedown regime provides a natural setting to study the signaling aspect of pretrial bargaining. We apply text analysis to quantify the attributes of notice as a pretrial signal, and we use the text data to evaluate how different factors help to close the information gap and improve the settlement rate. The three primary determinants that help settlement are found to be text features of the complaints, legal representation, and platform mediation. A strong signal is short, easy to read, and more specific. Legal representation improves the credibility of the signal. Platform mediation, on the other hand, adds commitment to the signal. Interestingly, how the lawyers draft a notice compromises the positive effect of legal representation. Lawyers prefer long sentences, big words, and more terminology, whereas an effective notice is much more concise. Most of our empirical findings support theoretical predictions, but we also discuss some discrepancies between the two.

Mediation and Costly Evidence (with Yi Chen)

conference presentation: AEA 2021, ALEA 2021, LET 2022

  • This paper studies the welfare properties of mediation design when disputants are asymmetrically informed and hard evidence can be acquired or presented with a cost. The model encompasses both facilitative mediation where the mediator only transmits information, and evaluative mediation where the mediator bases recommendation on verifiable information. We consider both the adversarial procedure where the disputants present evidence, and the inquisitorial procedure where the mediator acquires evidence.We identify two primary sources of inefficiency in mediation design. The first one is the settlement failure: profit maximization might lead to an inefficiency low settlement rate for both procedures. The second one is the procedural bias: a profit-maximizing mediator prefers the inquisitorial procedure, whereas a welfare-maximizing mediator prefers the adversarial procedure. Positive predictions of the model match the practice of professional mediators on mediation style, settlement rate, settlement terms, and satisfaction ratings.

  • COVID-19 has reinvigorated the policy debate for a universal healthcare system, attracting much attention on social media. In this paper, we study the online discourse of Medicare-For-All before and after COVID-19 by examining the Twitter feeds of two opposing health advocacy groups -- Physicians for a National Health Program (PNHP) and Partnership for America’s Healthcare Future (P4AHCF). Our empirical results show a sharp contrast between the two interest groups’ communication strategies. PNHP showed a consistent narrative before and after the onset of COVID-19 on March 11th, 2020, marked by personalized stories, references to diverse demographic groups, and a growing number of Medicare-for-All tweets. In contrast, P4AHCF showed more scientific terminology and data-centric tweets and had an inconsistent narrative with a sudden surge in positive sentiments and a complete silence on Medicare-for-All right after March 11th. The difference in communication strategies is consequential. PNHP has higher engagement with Twitter users and is more adaptive to a pandemic narrative than P4AHCF. We argue that the distinctive social media strategies can be explained by the groups’ different audiences and objectives. The findings add to our understanding of American’s activism on social media and the implication of a pandemic for health policy reform.

Two Types of Censorship (with Shreyas Meher)

  • Not all autocracies are doing the same kind of censorship. Countries like China build a closed border and a policing workforce for their internet, whereas countries like Russia compete in their internet with pro-government messages and requests. This paper employs a data-driven approach to study the variety of censorship among autocratic countries. Internet controls are measured by the panel data from Freedom House, V-Dem, OONI, and Google Transparency Report. Using an unsupervised learning technique of cluster analysis, we group the countries' censorship behaviors based on multi-dimensional indicators of internet access, content restriction, and technological barriers. We discover two distinct types of censorship: pervasive control regime (e.g., China) and influence operation regime (e.g., Russia). The two types are supported by country-specific studies and are shown to predict the country's content restriction strategies. We also show that differences in national IT capacity explain a country's distinct censorship style. Sending takedown requests is a cost-saving alternative to a state-run monitoring workforce. A one-unit increase in the country's IT capacity leads to 9,206 fewer requests and 11,398 more incidents of blocking the internet annually.

The Local Economic Effects of Increased Copyright Protection: An Event Study of the Film Industry (with Afif Mazhar)

  • Coming Soon !

Did the Budapest Convention Deter Cybercrime? An Event Study Using a Panel of Countries (with Andrew Fowler and Humza Khan)

  • Coming Soon !

Publication

An Economic Model of Intermediary Liability (with James Grimmelmann), forthcoming in Berkeley Technology Law Journal

  • Scholars have debated the costs and benefits of Internet intermediary liability for decades. Many of their arguments rest on informal economic arguments about the effects of different liability rules. Some scholars argue that broad immunity is necessary to prevent overmoderation; others argue that liability is necessary to prevent undermoderation. These are economic questions, but they rarely receive economic answers.

    In this paper, we seek to illuminate these debates by giving a formal economic model of intermediary liability. The key features of our model are externalities, imperfect information, and investigation costs. A platform hosts user-submitted content, but it does not know which of that content is harmful to society and which is beneficial. Instead, the platform observes only the probability that each item is harmful. Based on that knowledge, it can choose to take the content down, leave the content up, or incur a cost to determine with certainty whether it is harmful. The platform's choice reflects the tradeoffs inherent in content moderation: between false positives and false negatives, and between scalable but more error-prone processes and more intensive but costly human review.

    We analyze various plausible legal regimes, including strict liability, negligence, blanket immunity, conditional immunity, liability on notice, subsidies, and must carry, and we use the results of this analysis to describe current and proposed laws in the United States and European Union.

A Reference Dependent Regret Theory (with Yuval Erez and Felipe A. Araujo), Decision 9(1), 1

  • We propose an extension of the classical regret theory model (Loomes and Sugden, 1982; henceforth LS) incorporating the notion of a reference point. As in LS, the model can account for a number of documented deviations from expected utility theory. Additionally, we show that our model is consistent with a class of behaviors known as omission bias, for example a reluctance to exchange lottery tickets, and generates predictions which are consistent with recent empirical evidence on the common-ratio effect with correlated outcomes. The model also provides a novel interpretation for risk aversion in small-stakes, equiprobable gambles. The predictive power of the theory, as well as its relative shortcomings and advantages, are examined and compared to that of other extensions of regret theory and three alternative reference-dependent models.

The Economics of Profit-Cap Policy: Big Pharma, Big Tech, and the Duopoly Rule (with Kaushik Basu and Fikri Pitsuwan) forthcoming in Journal of Economic Behavior and Organization

  • A known policy dilemma occurs between the need to curb extra-large profits by some industries, like pharmaceuticals, and the need to ensure the incentive to produce is not damaged. This paper shows that a profit cap, imposed via taxation on a group of firms, can simultaneously eliminate inefficiency and excess profit by intensifying competition among oligopolistic firms. The result has a direct bearing on policy debates on COVID-19 vaccine sharing and the use of vaccine donation as a “humanitarian obligation,” and, more generally, on the regulatory institutions needed for industries that rely on R&D.

Herd Immunity and a Vaccination Game: An Experimental Study (with Wooyoung Lim), PLOS ONE 15(5): e0232652.

media coverage: PLOS COVID-19 Pandemic Collection, Brookings Global Economy and Development

theory-and-evidence working paper version

  • Would the affected communities voluntarily obtain herd immunity if a cure for COVID-19 was available? This paper experimentally investigates people’s vaccination choices in the context of a nonlinear public good game. A “vaccination game” is defined in which costly commitments (vaccination) are required of a fraction of the population to reach the critical level needed for herd immunity, without which defectors are punished by the natural contagion of epidemics. Our experimental implementation of a vaccination game in a controlled laboratory setting reveals that endogenous epidemic punishment is a credible threat, resulting in voluntary vaccination to obtain herd immunity, for which the orthodox principle of positive externalities fails to account. The concave nature of the infection probability plays a key role in facilitating the elimination of an epidemic.