The Use of Algorithms in Society
Harvard Working Paper, December 2022
The judgments of human beings can be biased; they can also be noisy. Across a wide range of settings, use of algorithms is likely to improve accuracy, because algorithms will reduce both bias and noise. Indeed, algorithms can help identify the role of human biases; they might even identify biases that have not been named before. As compared to algorithms, for example, human judges, deciding whether to give bail to criminal defendants, show Current Offense Bias and Mugshot Bias; as compared to algorithms, human doctors, deciding whether to test people for heart attacks, show Current Symptom Bias and Demographic Bias. But in important cases, algorithms struggle to make accurate predictions, not because they are algorithms but because they do not have necessary data. (1) Algorithms might not be able to identify people's preferences, which might be concealed or falsified, and which might be revealed at an unexpected time. (2) Algorithms might not be able to foresee the effects of social interactions, which can lead in unanticipated and unpredictable directions. (3) Algorithms might not be able to anticipate sudden or unprecedented leaps or shocks (a technological breakthrough, a successful terrorist attack, a pandemic, a black swan). (4) Algorithms might not have "local knowledge," or private information, which human beings might have. (5) Algorithms might not be able to foresee the effects of context, timing, serendipity, or mood. Predictions about romantic attraction, about the success of cultural products, and about coming revolutions are cases in point. The limitations of algorithms are analogous to the limitations of planners, emphasized by Hayek in his famous critique of central planning. It is an unresolved question whether and to what extent some of the limitations of algorithms might be reduced or overcome over time, with more data or various improvements; in the relevant contexts, there is no equivalent to the price system to elicit and aggregate dispersed knowledge.
Regulatory Intensity and Firm-Specific Exposure
Review of Financial Studies, forthcoming
Building on administrative data and machine-learning models, I develop a firm-specific measure of regulatory intensity: cost of compliance with all federal paperwork regulations. Regulatory intensity increases the cost of goods sold and overhead spending (SGA). It also incentivizes companies to reduce capital investment, hire fewer employees, and lobby more. The effects are particularly strong among financially constrained firms and those with irreversible investment opportunities, suggesting that regulation affects companies through budgetary pressures and heightened uncertainty. The findings highlight the real effects of regulation and the underlying mechanisms.
Do Firms Bunch at EEOC Kink Points? Implications for Firm Value Under Exogenous Risk
University of Texas Working Paper, January 2023
On July 14, 1992, the U.S. Equal Employment Opportunity Commission (EEOC) implemented enforcement guidance for compensatory and punitive damages available under section 102 of the Civil Rights Act of 1991. In this policy, the EEOC caps litigation payouts to $50,000, $100,000, $200,000, and $300,000 for firms with 15 to 100 employees, 101 to 200 employees, 201 to 500 employees, and 501 employees or more, respectively. I implement a threshold design around these EEOC enforcement kink points to examine (i) if firms bunch at these thresholds, (ii) the impact of bunching on firm value, and (iii) why firm value changes for firms that bunch. I find evidence that firms do indeed bunch at EEOC kink points. Bunching firms have negative annualized alphas of 14.88%, negative annualized raw stock returns of 7.26%, and negative Tobin's Q values of 9.44%. Workforce rigidity around exogenous risk explains why bunching firms have negative market values.
Discovering and forecasting extreme events via active learning in neural operators
Ethan Pickering et al.
Nature Computational Science, December 2022, Pages 823-833
Extreme events in society and nature, such as pandemic spikes, rogue waves or structural failures, can have catastrophic consequences. Characterizing extremes is difficult, as they occur rarely, arise from seemingly benign conditions, and belong to complex and often unknown infinite-dimensional systems. Such challenges render attempts at characterizing them moot. We address each of these difficulties by combining output-weighted training schemes in Bayesian experimental design (BED) with an ensemble of deep neural operators. This model-agnostic framework pairs a BED scheme that actively selects data for quantifying extreme events with an ensemble of deep neural operators that approximate infinite-dimensional nonlinear operators. We show that not only does this framework outperform Gaussian processes, but that (1) shallow ensembles of just two members perform best; (2) extremes are uncovered regardless of the state of the initial data (that is, with or without extremes); (3) our method eliminates 'double-descent' phenomena; (4) the use of batches of suboptimal acquisition samples compared to step-by-step global optima does not hinder BED performance; and (5) Monte Carlo acquisition outperforms standard optimizers in high dimensions. Together, these conclusions form a scalable artificial intelligence (AI)-assisted experimental infrastructure that can efficiently infer and pinpoint critical situations across many domains, from physical to societal systems.
Innovation and profitability following antitrust intervention against a dominant platform: The wild, wild west?
Sruthi Thatchenkery & Riitta Katila
Strategic Management Journal, forthcoming
This study examines whether "unblocking" competition through antitrust intervention against a dominant platform can spur complementor innovation in platform ecosystems. Using a novel dataset on enterprise infrastructure software and a difference-in-differences design, we examine the relation between the U.S. antitrust intervention against Microsoft (dominant enterprise platform) and subsequent innovation and profitability of infrastructure applications firms (complementors). The data show that innovation among complementors -- particularly ones with low market share -- soared when the competitive pressure on the dominant platform amplified. However, the profitability of these complementors dropped. Our results contribute to understanding links between competition and innovation in platform ecosystems, as well as the opportunities and threats related to dominant platforms in those ecosystems.
The costs of insecurity: Pay volatility and health outcomes
Journal of Applied Psychology, forthcoming
Every day, millions of individuals rely on fluctuating financial rewards in the form of customer tips, commissions, piece-rate, and performance-based pay. While these compensation systems are increasingly common, the volatility in pay that they create may harm employee health. Based on conservation of resource theory assumptions that money is a valued resource, I propose that volatility in pay represents resource insecurity, with costs to health. Across an experience sampling study of tipped workers (Study 1) and longitudinal studies of gig workers (Study 2) and those in sales, marketing, and finance (Study 3), findings demonstrate the harmful effects of pay volatility. Specifically, pay volatility had direct or indirect effects on physical symptoms, insomnia, sleep quality, and sleep quantity. Volatile pay was found to induce a scarcity mindset, where individuals ruminate and direct cognitive resources toward remedying the source of scarcity, with worse health outcomes as a result. Neither mindfulness nor savings rate moderated the effect. Exploratory analyses in Studies 2 and 3 revealed that one's dependence on volatile pay acted as a moderator that strengthened effects. Overall, performance-based pay creates pay volatility, which is linked to psychological threat and poor physical health for employees in a broad range of industries.
Are the home values and property tax burdens of permanent homeowners affected by growth in housing rentals and second homes: Evidence based on big data from Florida
Keith Ihlanfeldt & Cynthia Fan Yang
Journal of Regional Science, forthcoming
Homeowners who make their homes their primary residence have resisted the entry of rentals into their neighborhoods and cities. Possible reasons underlying this resistance are that rentals reduce the property values, increase the property tax burdens, and raise the price of public services for these homeowners. We relate the market values of single-family homes occupied by permanent homeowners, the effective property tax rate of these homeowners, and the tax price they pay for public services to shifts in their city's housing units toward a variety of different types of rentals and second homes. Our analysis is based on large panels of Florida homes containing hundreds of thousands and millions of observations. Our results show that increases in the share of a city's housing units used as rentals or second homes reduce the home values, increase the property tax burdens, and raise the public services prices of homeowners who permanently reside in single-family homes. Impacts vary in magnitude among single-family, condominium, and mobile homes used as rentals and second homes. Estimated impacts are the strongest for share increases in single-family rentals and second homes.
Man vs. Machine: Technological Promise and Political Limits of Automated Regulation Enforcement
Oliver Browne et al.
NBER Working Paper, January 2023
New technologies allow perfect detection of environmental violations at near-zero marginal cost, but take-up is low. We conducted a field experiment to evaluate enforcement of water conservation rules with smart meters in Fresno, CA. Households were randomly assigned combinations of enforcement method (automated or in-person inspections) and fines. Automated enforcement increased households' punishment rates from 0.1 to 14%, decreased summer water use by 3%, and reduced violations by 17%, while higher fine levels had little effect. However, automated enforcement also increased customer complaints by 1,102%, ultimately causing its cancellation and highlighting that political considerations limit technological solutions to enforcement challenges.
Is Walmart the same as ten years ago? A non-parametric difference-in-differences analysis of Walmart development
Regional Science and Urban Economics, forthcoming
Walmart experienced tremendous growth and changes from 1962 when the first store opened. This study examines the temporal variation in Walmart's effects on neighboring house prices and local labor markets employing a new non-parametric difference-in-differences method to address concerns that areas immediately around a Walmart and comparison areas nearby may have different trends prior to a store opening. The results show that the prices of homes located within 0.5 miles of Walmart increased at a maximum rate of 2.97% due to store openings from 1990 to 2000. Conversely, store openings in 2002-2007 resulted in approximately zero effects on neighboring house prices, and stores that opened in 2008-2013 experienced house price declines of roughly 3% within 0.5 miles of the store. Explorations of the underlying mechanisms suggest that Walmart openings had led to deteriorating shopping convenience but increasing negative spillovers from the early to the late 2000s. Government subsidies and zoning policies targeting these issues might help mitigate the negative spillovers from Walmart.
Mutual Fund Revenue Sharing in 401(k) Plans
Veronika Pool, Clemens Sialm & Irina Stefanescu
NBER Working Paper, December 2022
Recordkeepers in DC pension plans are often paid indirectly in the form of revenue sharing from third-party funds on the menu. We show that these arrangements affect the investment menu of 401(k) plans. Revenue-sharing funds are more likely to be added to the menu and are less likely to be deleted. Overall, revenue-sharing plans are more expensive as higher expense ratios are not offset by lower direct fees or by superior performance. Rebates increase with the market power of the recordkeeper suggesting that third-party funds may revenue share to gain access to retirement assets.