This paper studies whether experiencing unequal opportunities changes redistributive preferences. We design a spectator experiment in which third-party spectators first complete one of two versions of a real-effort task that differ in difficulty, and then decide how to redistribute a bonus between two workers, one of whom completed the easy version of the task and one of whom completed the hard version. This design allows us to test whether own experience changes how spectators interpret merit, fairness, and redistribution. We find that spectators who experience the disadvantaged task revise their beliefs about the role of opportunities, view the initial allocation as less fair, and redistribute more toward the disadvantaged worker. This increase persists even when redistribution requires spectators to pay a personal cost, suggesting that experience affects not only fairness views but also the willingness to implement them. In an extension, we show that directly experiencing both the advantaged and disadvantaged tasks generates even stronger redistributive responses, while receiving comparable information through reports or statistics produces weaker effects. The findings suggest that redistributive preferences are shaped by how unequal opportunities are encountered.

Abstract: 

Meritocratic fairness justifies inequality when it stems from performance. Yet performance is influenced by one's genetics. I investigate whether individuals' redistribution preferences are affected by their beliefs about  genetics' role in generating performance inequality. In an incentivized online experiment, impartial spectators can redistribute the earnings that two workers earned based on their performance in a mathematical task. Across two treatments, I modify beliefs about the importance of genetics in performing in the task by means of an information provision treatment. I find that spectators for whom genetics is framed to play a larger role compensate the worse performer more, compared to those for whom genetics is framed to play a smaller role. When comparing the spectators' decisions before and after the provision of information, I find that about 23% of spectators compensate the worse performer more whereas the majority does not change their allocation. This study highlights that individuals' redistribution preferences are affected by their beliefs about genetics' role in generating performance inequality.

Find it here

We study whether biased responses to random sequences persist when deviations from expected-value maximization become costly. In two online experiments, participants play repeated binary lotteries with independent draws. Study 1 varies costliness by changing the probability composition of the lottery while keeping payoffs fixed: one color is drawn with probability 50%, 51%, or 61%. Study 2 holds probabilities fixed at 50–50 but varies the payoff attached to each color, creating the same expected-value differences through payoffs rather than probabilities. Across both studies, participants respond to incentives by choosing the higher-EV option more often when the expected-value gap is larger. However, only a minority fully adopt the payoff-maximizing strategy, even when doing so is straightforward. We document several sources of deviation, including probability matching, history-dependent choice, and negative recency consistent with the gambler’s fallacy. Belief elicitations and incentivized recommendations show that many participants attend to past outcomes and perceive patterns even in explicitly random environments. The results suggest that costliness reduces, but does not eliminate, biased decision-making under randomness. 

This study examines whether online panels can be used to collect reliable device-based sleep duration data from respondents who already own activity trackers. We conducted a three-wave survey experiment with UK-based Prolific respondents, comparing an unrestricted sample with a sample restricted to device owners. Device owners were asked to upload screenshots of sleep data from their trackers, while all respondents self-reported their sleep duration. We find that collecting device-reported sleep data through online panels is feasible: most uploaded screenshots allowed sleep duration to be extracted. However, the added value of these data is limited for measuring sleep duration alone, as self-reported and device-reported sleep were closely aligned among device owners. At the same time, the approach introduces important trade-offs. Restricting recruitment to device owners changes sample composition, and making uploads mandatory substantially increases attrition. These results suggest that collecting device-based sleep data through online panels can be useful and cost-efficient, especially when richer sleep measures are needed, but researchers should carefully weigh these benefits against losses in representativeness and retention. 

Health utilities, which reflect how people value health conditions, are often elicited using the composite time trade-off (cTTO) method. This asks people how many years they would give up to avoid living in impaired health states. However, cTTO performs inadequately for states considered worse than being dead (WTD): more severe states do not necessarily receive lower utility. This study (i) explores whether this low sensitivity can be addressed by using ‘unified’ methods—methods that do not require separate approaches for eliciting utilities for states considered WTD, and (ii) investigates if the low sensitivity problem extends to risk-based methods such as the standard gamble (SG)—a method where participants choose between living with a health condition or taking a risky treatment. We ran an experiment with 198 participants randomized to one of four arms. One arm used cTTO followed by lead-time TTO (LT-TTO), which applies the same method for WTD and BTD (better than dead) states. The other three arms used variations of SG including a unified variant of the SG to explore the performance of risk-based and unified methods. Our results show that for both SG and TTO, unified procedures have improved sensitivity, i.e. they display a negative correlation between severity and utility for WTD states. However, the LT-TTO, compared to the cTTO, increases the number of WTD states and leads to lower utilities.

Find it here