Allowing Value
What is Generative AI Worth?
Erik Brynjolfsson et al.
Stanford Working Paper, April 2026
Abstract:
We estimate the consumer welfare gains from the rapid adoption of generative AI tools like ChatGPT, Gemini, Claude, or Copilot in the United States. Using online choice experiments, we elicit willingness to accept (WTA) compensation for giving up access to these AI chatbot tools for one month from representative samples of US adults, fielded in two waves, July 2025 and March 2026. We find that mean willingness to accept (WTA) increased from $98 in 2025 to $124.50 in 2026, a 27% increase, while the median value rose from $3.4 to $11.40. Combined with growth in the adult user base from 98 million to 115 million, these estimates imply that aggregate consumer surplus increased from $116 billion to $172 billion. This surplus substantially exceeds estimated revenues from generative AI in the United States, suggesting that consumers capture most of the welfare gains from these tools. We find substantial heterogeneity in valuations: usage frequency is the strongest predictor of WTA, followed by workplace use, and paid subscription status, with additional differences by gender, age, and ethnicity. Overall, the results suggest that generative AI is already generating substantial and rapidly growing welfare gains, even before its full effects on measured productivity and GDP are reflected in official statistics.
AI at the Wheel: The Effectiveness of Advanced Driver-Assistance Systems
Vikram Maheshri, Clifford Winston & Yidi Wu
Journal of Law and Economics, May 2026, Pages 387-411
Abstract:
Has automakers' use of artificial intelligence (AI) in advanced driver-assistance systems (ADASs) improved automobile safety? We address this question with a first-of-its-kind trim-level dataset of the universe of registered automobiles and accidents in Texas over a 9-year period. We find that ADASs reduce the risk of a motorist getting in any type of accident by 11 to 14 percent and reduce the risk of a motorist getting in a single-vehicle fatal accident by roughly one-third. Our finding that ADASs have improved automobile safety is especially important because it provides early evidence of the benefits of vehicle automation in actual travel environments. Hopefully, it will spur greater interest in the development and widespread adoption of fully autonomous vehicles and in the potential benefits of other transportation technologies using AI.
Corporate M&As and Labor Market Concentration: Efficiency Gains or Power Grabs?
David Cicero, Mo Shen & Jaideep Shenoy
Journal of Finance, June 2026, Pages 1437-1484
Abstract:
Mergers of firms that share labor markets increase labor market concentration which can lead to labor efficiency gains and/or create labor market power for the merged firms. Using a novel measure based on establishment-level employment data, we find that merger-induced increases in labor market concentration explain value creation in a sample of completed U.S. public firm mergers from 1991 to 2016. Analysis of the stock market reactions of rival, supplier, and customer firms, as well as firm- and establishment-level real effects in the merging firms, supports a labor efficiency explanation of these merger gains.
The regressive effects of worker protection: The role of financial constraints
Diego Huerta
Labour Economics, June 2026
Abstract:
I exploit the staggered adoption of U.S. state-level Employment Protection Legislation (EPL) to study its effects on firm profits and the wage bill. I find that EPL has unintended regressive consequences. EPL harms smaller firms and their workers, while only benefiting larger firms and their workers. These effects operate through a credit channel: EPL limits access to credit for smaller firms, forcing them to cut employment and even exit the market. Conversely, larger firms expand employment by issuing more debt while benefiting from lower competition. A model with heterogeneous firms and endogenous financial constraints guides the empirical analysis.
Income Mobility, Automation, and Occupational Licensing
Vincent Geloso, Alicia Plemmons & Pradyot Sharma
Southern Economic Journal, forthcoming
Abstract:
Technological change has long been tied with distributional concerns due to displacement against certain skills on labor markets. Short-run dislocations could create scarring in the long run. For example, shifts against less skilled workers with children could limit their ability to improve the inter-generational income mobility of their children. The existing literature rarely emphasizes the possibility that the ill effects of technological change are conditional on government regulations that limit the ability of workers to rapidly adjust-thus creating the scarring. We document the importance of these regulations by focusing on changes in licensing of low- and mid-income occupations, exposure to industrial automation in the United States since the 1980s, and patterns of intergenerational income mobility. We find that a significant share of the prediction of falling income mobility attributed to automation is actually tied to changes in occupational licensing. Areas that experienced labor market deregulation and high exposure to automation suffered far less than areas that did not engage in deregulation.
Licensing Enforcement and the Supply of Childcare
Luis Faundez & Manisha Padi
University of California Working Paper, March 2026
Abstract:
Licensing authorities enforce regulations through inspections, which can worsen supply restrictions caused by licensing but may improve quality. We study the equilibrium impacts of licensing enforcement in the context of early childhood education. Using the childcare licensing program analysts' stringency as an instrument for violations, we study the causal impact of violations on facility closure. We find that violations lead to an increase of 4 percentage points in the probability of facility closure. Closures are not driven by involuntary administrative penalties like license revocation, but instead result from equilibrium shifts in both supply and demand. Overall, closure is more likely to result from administrative violations, and have greater impact on home daycares and underserved areas. Our results suggest that low quality facilities are more likely to exit after a violation, but that facilities in low supply areas are more likely to exit due to regulatory stringency. This implies licensing authorities facing a tradeoff between quality and adequacy of supply should increase stringency in detecting care-related, rather than administrative, violations.
Why Bans Fail: Tipping Points and Australia's Social Media Ban
Leonardo Bursztyn et al.
NBER Working Paper, April 2026
Abstract:
In December 2025, Australia became the first country to ban youth under 16 years old from holding accounts on major social media platforms, a policy now under consideration in more than a dozen countries and in numerous states. Because social media use is inherently social, the effectiveness of a ban that is easy to circumvent may depend on whether compliance reaches a tipping point: a share of compliant peers high enough to make it optimal for individuals to comply themselves. We surveyed 835 Australian teenagers four months after the ban took effect and find that only about one in four 14-15-year-olds comply. The social environment around use has barely moved: most banned teens believe that their peers are still using banned platforms and cite social reasons for continuing use. Sustaining high compliance requires two ingredients: the share of compliers must be high enough and those who comply must find it preferable to continue complying. The current ban achieves neither. Teenagers report that they require roughly two-thirds of peers to stop using social media to stop themselves, far above the share currently complying. They also perceive compliers as less popular than non-compliers, so the more influential teens disproportionately stay on the platforms. Together, these patterns suggest that compliance is more likely to diminish than to rise. Sustaining higher compliance will likely require pairing the ban with instruments that act on social norms and individual incentives directly.
Social trust and markets: Does regulation undermine the social foundation of exchange?
Peter Calcagno & Jeremy Jackson
Journal of Regulatory Economics, 2026
Abstract:
We contribute to the literature on the overlap of social trust and regulation. The literature on trust and regulation uses cross-country samples. We analyze the United States from 1972 to 2018 using time series data. Examining a single country allows us to focus on trends, which literature on social trust and regulation has not done. The literature suggests that market exchange facilitates and discovers trust amongst trading parties. When trust breaks down, so does the economy. Buchanan argues that if moral order exists, communities can interact and flourish with trust. However, if the moral order breaks down, it creates moral anarchy. Environments devoid of social trust may substitute regulation to facilitate exchange. Low social trust may necessitate high regulation. Thus, in highly regulated economies, exchange occurs not by social trust but by regulation. Regulation removes the social feedback loop that leads to generalized social trust. Thus, high regulation erodes social trust over time. A negative correlation between social trust and regulation exists, but an issue of simultaneity remains. We examine whether a lower social trust is to blame for the proliferation of regulations in the U.S. or if the increase in regulation contributes to declining social trust. Consistent with the literature, we find a negative correlation between trust and regulation, with increases in regulation leading to future decreases in social trust, for the cohort of individuals under the age of 42. This exposes a vicious cycle between regulation and trust for the younger cohort. However, for the full sample, increases in social trust growth rates are associated with a small increase in regulation growth.
Data Neutrality, Data Supply, and Market Competition
Hanming Fang & Soo Jin Kim
NBER Working Paper, April 2026
Abstract:
We analyze the effects of data neutrality regulations on downstream market competition, the incentive of the platform to produce data, and consumer welfare. In our framework, data neutrality requires that firms seeking access to the platform's data be treated equally, irrespective of whether they are affiliated with the platform. We consider two forms of regulation. Under weak data neutrality, the platform must provide the same amount of data to affiliated and unaffiliated sellers; under strong data neutrality, it must also charge the same price. We show that weak data neutrality can be largely ineffective, as the platform may restore exclusion through discriminatory pricing. Strong data neutrality is more consequential, but it does not necessarily raise welfare. Although it broadens access and intensifies downstream competition, it also reduces the incentive of the platform to refine and produce data. Consequently, data neutrality can reduce the equilibrium amount of data available in the market, and this data-reduction effect can dominate its benefits, which enhance competition. These findings suggest that regulating access to platform data requires balancing fair competition against the incentive to generate valuable data inputs.
The Impact of AI-Generated Text on the Internet
Jonas Dolezal et al.
Stanford Working Paper, April 2026
Abstract:
The proliferation of AI-generated and AI-assisted text on the internet is feared to contribute to a degradation in semantic and stylistic diversity, factual accuracy, and other negative developments (sometimes subsumed under the "Dead Internet Theory"). What has hindered answering these questions is that it has not been understood just how much of the internet is actually AI-generated or AI-edited. To this end, we construct a representative sample of websites published on the internet between 2022 and 2025 using the Internet Archive, and apply a state-of-the-art AI text detector on them. We find that by mid-2025, roughly 35% of newly published websites were classified as AI-generated or AI-assisted, up from zero before ChatGPT's launch in late 2022. We also find statistically significant evidence for some of the identified hypotheses; for example, that increases in AI-generated text on the internet correlate negatively with semantic diversity and positively with the prevalence of positive sentiment. We do not, however, find statistically significant evidence supporting the hypothesis that an increased rate of AI-generated text on the internet decreases factual accuracy or stylistic diversity. Notably, this diverges from public perception, which we measure in a user study, where the majority of US adults turned out to believe in all four of the above-mentioned hypotheses. Individuals who do not use AI or use it infrequently tend to believe in these negative impacts more than those who use it frequently; similarly, individuals who hold negative views of AI tend to believe in these hypotheses more than those with favorable views of the technology.
Proposed Mergers Where Efficiencies Are Needed Most Might Be the Least Likely to Deliver Them
Robert Metcalfe, Alexandre Sollaci & Chad Syverson
NBER Working Paper, May 2026
Abstract:
Mergers are commonly evaluated by weighing their expected market power effects against any efficiency gains they create. The larger the market power effect of a proposed merger, the larger must be any efficiencies for it to raise social welfare. We show selection into merger proposal distorts the observed relationship between market power and efficiency effects. Even if market power and efficiency gains are independent (or even positively correlated) across all potential mergers, they will generally be negatively related among proposed mergers. This is because parties propose to merge only if the merger's expected profitability exceeds a threshold, so the underlying components of profitability become substitutes in clearing that hurdle. It does not rely on managerial bias, behavioral frictions, or strategic misrepresentation. We demonstrate this negative correlation is present under very general conditions when the two effects are uncorrelated among all mergers. We also characterize conditions where this still holds even in the presence of positive underlying correlations and firms' uncertainty about their own merger's profitability. Policies that might raise the selection hurdle for proposed mergers do not alleviate the negative correlation; indeed, they would exacerbate it. Our analysis has direct implications for interpreting empirical merger retrospectives and for evaluating efficiency defenses in antitrust policy.
Hearing, not heeding: Procedural acknowledgment and substantive influence in rulemaking
Alexander Love
Journal of Public Administration Research and Theory, April 2026, Pages 219-233
Abstract:
Public participation processes promise that citizens will be heard, but rarely guarantee they will be heeded. This distinction between procedural acknowledgment and substantive influence lies at the heart of bureaucratic responsiveness, yet these two forms of responsiveness are often conflated in empirical research. I demonstrate that in federal rulemaking, procedural acknowledgment (being heard) is empirically distinct from substantive policy influence (being heeded). Drawing on theories of bureaucratic responsiveness, I argue that agencies strategically cite commenters not primarily to signal agreement but to build defensible administrative records that satisfy procedural requirements while preserving their policy autonomy. Analyzing 854 federal rules from 2017 to 2023, I use semantic text analysis to track changes in binding regulatory provisions distinct from the explanatory preamble. I show that agencies systematically cite comments they ultimately reject, particularly from well-resourced groups. Roughly two-thirds of comment citations are not accompanied by any responsive change to the regulatory text. This reveals that procedural responsiveness can function as a strategic substitute for substantive policy change. These findings suggest that procedural engagement and substantive influence operate as distinct modes of bureaucratic responsiveness, with agencies often prioritizing legal defensibility over policy adaptation when facing potential judicial review.