Bradley J. Ruffle

We develop a model of monopolistic competition in the rental housing market for lowincome households with endogenous tenancy default. Identical suppliers choose the number of rental units to supply and the rental price to charge to maximize expected profits. Potential tenants who differ in their incomes and face an uninsurable income risk choose whether to engage in a costly search for rental housing. If they search and find a rental unit, then they must commit to a rental agreement before their income uncertainty is resolved. Consequently, some tenants may default on their rental payments. We show that tenancy default can explain persistent excess demand in the low-income rental housing market without any government price regulations, and that such excess demand can lead to nonstandard effects of government regulations. We also test whether the excess-demand equilibrium is behaviorally plausible by designing a laboratory experiment. Our experimental results reveal that, with feedback and repetition, the excess-demand equilibrium that requires landlord participants to restrict supply is no more difficult for participants to reach than a market-clearing equilibrium.

As artificial intelligence (AI) increasingly acts autonomously in economic and social environments, understanding how humans and AI interact strategically has become a central scientific and societal question. We report controlled experimental evidence on how humans and autonomous AI agents behave toward one another in repeated conflict and coordination settings. Unlike prior work using pre-programmed strategies or hypothetical judgments, we embed autonomous LLM agents directly into an incentivized laboratory experiment as strategic decision-makers interacting with both humans and other AI agents. Thus, AI agents in our study independently interpreted instructions, formed expectations, made gameplay decisions, generated round-by-round reasoning for their decisions and adapted dynamically to feedback. Specifically, humans and AI agents play either a 40-round repeated Prisoner's Dilemma or Battle of the Sexes, 20 rounds against each counterpart type. We elicit both humans' and AI's earnings beliefs before and after an initial 20-round block of play against each counterpart type, and observe dynamic behavior across human-human, human-AI, and AI-AI dyads under identical laboratory conditions. Humans show no strong ex ante preference between human and AI counterparts, but exhibit more reciprocity, cooperation and accommodation toward fellow humans. In contrast, AI agents display a systematic in-group bias toward AI. AI's thought processes highlight its aim to maximize joint payoffs, establish mutual cooperation and achieve coordination. Consequently, AI-AI pairs achieve the highest payoffs in both games. Our findings show that social preferences are endogenous to counterpart identity in mixed agent systems. These counterpart-contingent social preferences generate markedly different equilibrium dynamics across dyad types.

We study whether pay-to-quit incentives (i.e., payments offered to employees who resign) can help organizations maintain a workforce of employees who highly identify with the organization. Across a laboratory experiment and an online vignette experiment, we find that pay-to-quit incentives induce employees with lower levels of organizational identification (OI) to leave their organization. Employees with low (high) OI are generally likely (unlikely) to leave the organization regardless of whether a pay-to-quit incentive is offered. As a result, relative to no incentive, the effect of a pay-to-quit incentive on employees’ likelihood of leaving is strongest among employees with moderate OI. We also find that employees who forgo the pay-to-quit incentive subsequently exert higher work effort than before. This research contributes to the rising accounting literature on the sorting effects of incentives by showing that pay-to-quit incentives both help retain highly identified employees and motivate those who remain. Data Availability: Data are available from the authors upon request.

An underexplored class of social dilemmas revolves around a shared resource pool with privately observed entitlements, making the resource susceptible to dishonest overclaiming. To study these dilemmas, we introduce a novel four-player game that permits individuals to claim more than their entitlement. Two fundamental alternatives that often arise involve participants submitting claims simultaneously (e.g., municipal department budgets, public-housing and university-dorm allocations), or sequentially (e.g., food pantries, community fridges, player negotiations on salary-capped sports teams). Motivated by such cases, our experimental design contrasts simultaneous and sequential reporting as well as plentiful and limited budgets. We find that simultaneous reporting reduces dishonesty when resources are plentiful and promotes greater equity when resources are limited. Under sequential moves with limited resources, Player 4 frequently protests the inequity by choosing to earn zero rather than the remaining budget. Our results demonstrate that these social dilemmas can be better managed by employing simultaneous reporting structures, which obscure information about individual claims.

We investigate whether the price paid for insurance explains dishonesty in reporting an insurance claim. In our laboratory experiment, participants earn money in a real‐effort task but risk losing some of this income through one of four randomly assigned, privately observed loss amounts. Before observing their loss, participants indicate their reservation price for insurance that pays an indemnity equal to their stated loss. Participants are insured if their randomly assigned premium is less than their stated reservation price. This mechanism provides data on each participant's consumer surplus from insurance. After receiving their cash earnings minus their assigned loss in private, participants report their loss. We find that the insured report modestly but statistically insignificant larger losses than the uninsured. Among the insured, we find no clear evidence that their reporting of excess losses increases in the randomly assigned price of insurance or decreases in the consumer surplus from insurance.

We study the factors associated with compliance with social‐distancing regulations using a unique data set on the behaviour of Ontarians during the COVID‐19 pandemic. To start, we build a simple theoretical model of social distancing in order to understand how some individual and community‐level factors influence compliance. We test our model's predictions by designing and conducting a survey on Ontarians in June and July 2020 in which we elicit their degree of compliance with existing distancing regulations as well as alternative policies such as fines for non‐compliance and wage subsidies for staying at home. Consistent with our model, we find that variables related to one's risk of infection (e.g., pre‐existing health condition, age, necessity of working outside the home and regional COVID‐19 cases) are significant predictors of compliance as are gender, political beliefs, risk and time preferences. To boost compliance among those who are young, healthy or apartment dwellers, we demonstrate the effectiveness of fines and wage subsidies.

We design a field experiment to test for price discrimination at seemingly highly competitive Israeli produce markets. We trained 90 buyers and sent them to produce markets across Israel. After verifying a product’s posted price, they asked for a discount on a one-kilogram or one-unit purchase. Vendors employ third-degree price discrimination: women are offered larger and more frequent discounts than men, and the more attractive the female buyer, the larger and more frequent the discount offered. Male buyers do not benefit from this beauty discount. No other buyer characteristic is a significant predictor of the likelihood or size of a discount. To understand our findings, we provide a more nuanced view of these markets that includes search costs and considerable vendor price-setting discretion.