Robust prizes in all-pay contests with loss aversion
summary
The gist
Robust prize design in all-pay contests with loss aversion when the principal lacks knowledge of agents' convex effort costs is studied to determine how to allocate rewards optimally under…
In short
This study investigates how to design fair prize structures in all-pay contests when participants are risk-averse regarding their effort costs. The main finding is that the optimal prize structure depends heavily on loss aversion: it can have a zero bottom prize under moderate aversion but requires a positive bottom prize and partial participation under strong loss aversion.
Key concepts
- Robust Prize Structure
- This refers to the specific way prizes are allocated across ranks (e.g., equal prizes for top ranks). The paper finds that the best structure changes based on how much agents dislike losses, suggesting a flexible design is needed when effort costs are uncertain.
- Loss Aversion ($ heta$)
- Loss aversion measures how much an agent dislikes losing something compared to gaining something of equal value. When this aversion is strong (greater than 1), it significantly changes the optimal prize design, leading to different participation strategies and prize amounts.
- Cost Ambiguity
- The analysis handles uncertainty about the exact cost of effort by assuming costs are convex but unknown. By analyzing the worst-case linear cost function, the paper finds a robust prize schedule that works well regardless of which specific convex cost function is true.
Terminology used across episodes
This episode discusses
- Robust prizes in all-pay contests with loss aversion · Paper Radio
- Robust Tournaments · Paper Radio
- The psychology of prizes: Loss aversion and optimal tournament rewards
The paper
Robust prizes in all-pay contests with loss aversion · Read on arXiv
Dmitry Ryvkin, Qin Wu
RMIT University
We study robust prize design in all-pay contests with loss-averse agents when the principal does not know the agents' convex cost of effort, subject to an aggregate productivity constraint. The robustly optimal prize structure awards equal prizes to all contestants except the one ranked last. The optimal bottom prize is zero under moderate loss aversion, but becomes positive and increases with loss aversion when loss aversion is sufficiently strong. Under strong loss aversion, the robust solution involves partial participation in the contest.
Transcript
Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.
Ruth: We're looking at the paper: "Robust prizes in all-pay contests with loss aversion".
Warren: Here's the gist of the paper. Robust prize design in all-pay contests with loss aversion when the principal lacks knowledge of agents' convex effort costs is studied to determine how to allocate rewards optimally under uncertainty about effort costs The Core Finding The robustly optimal prize structure awards equal prizes to all contestants except the one ranked last, and the optimal bottom prize is zero under moderate loss aversion but becomes positive and increases with loss aversion when loss aversion is sufficiently strong Robust Prize Structure The robustly optimal prize schedule has a “punish-the-bottom” structure, with a common prize for ranks 1,..., n−1 and a lower prize for rank n Under moderate loss aversion, the bottom prize is zero and the budget is divided equally among the remaining ranks Under strong loss aversion, the optimal bottom prize becomes strictly positive and increases with loss aversion, while the common top prize falls The Role of Loss Aversion A positive bottom prize can be useful because an agent who receives the top prize obtains an expected benefit of b+
P −θP(1−P): d, where θ > 0 measures the strength of loss aversion When θ > 1, the expression in brackets is negative for sufficiently small P > 0, implying the agent would prefer to receive b with certainty rather than face the prize lottery Robustness via Cost Ambiguity The analysis separates equilibrium behavior from ambiguity about effort costs by transforming effort into cost expenditure, which allows identifying a linear cost function that is least favorable at the optimal schedule The resulting robust prize schedule requires only limited information about effort costs Key Results on Robust Design The robust design problem is solved by maximizing W(v) = inf q∈QA sup σ∈E(v) E
q(Xσ): The robust value of v to the principal is W(v) = 2A E∗(θ), where E∗(θ) is the maximal expected equilibrium cost expenditure in the normalized linear cost problem For θ > 1, the robust effort guarantee is attained in a partial-entry equilibrium, and the entry probability decreases with loss aversion Behavior under Strong Loss Aversion Under strong loss aversion, the robustly optimal prize schedule involves a positive bottom prize and partial participation This suggests that maximum participation in competition for awards should not always be a design goal Related Literature The closest benchmark for the analysis is Fang, Noe and Strack (2020), who show that, with identical agents and convex effort costs, awarding equal prizes to all agents except the one ranked last maximizes expected effort The previous studies are restricted to relatively weak loss aversion and do not consider the interaction of prize design with endogenous participation The closest studies of prize design with expectation-based loss aversion are Fu, Wang and Zhu (2021) and Ryvkin and Wu (2024) The present paper is different because there is no noise, rankings are determined directly by effort, and there is ambiguity about convex effort costs Model Setup The model considers n ⩾ 2 identical agents who decide whether to participate, exerting effort ei at cost c(ei), where c(·) is continuous, strictly increasing, and convex with c(0) = 0 Effort can be recovered from cost expenditure as ei = q(xi), where q(·) = c−1(·) is the inverse cost function The agent’s utility from cost expenditure x in the form u(x; F) is given by equation (3) Equilibrium Concept The equilibrium concept adopted is a (modified) version of the choice-acclimating personal equilibrium (CPE), where agents have correct beliefs about how their efforts affect the probability of future outcomes A symmetric realization-acclimating mixed CPE is defined as a distribution F such that u(x; F) = ¯u for all x ∈ supp(F) and u(x; F) ⩽ u¯ for all x /∈ supp(F), for some u¯ Robustness Proof Strategy The robust design argument establishes an upper bound across all existing symmetric equilibria and constructs an equilibrium that attains it The linear cost analysis provides a uniform upper bound, E∗(θ), on cost expenditure in every symmetric mixed CPE under every feasible prize schedule Optimal Prize Schedule Theorem 1 states that the prize-schedule v∗(θ) given by v∗ r(θ) = 1 + m∗(θ)n − 1 + nm∗(θ), v∗ n(θ) = m∗(θ)n − 1 + nm∗(theta), solves the robust design problem (9) For θ ∈
0, 1: , the optimal schedule supports a unique equilibrium with full participation The robustly optimal prize allocation becomes weakly more equal as θ increases, and strictly more equal for θ > 1 The optimal prize differential d∗ n−1(θ) = v∗ n−1(θ) − v∗ n(θ) is constant for θ ⩽ 1 and strictly decreasing for θ > 1 Importantly, partial participation is optimal under strong loss aversion The results suggest that encouraging more agents to enter can be costly even when entry itself is costless References Akerlof, Robert J., and Richard T.
Ruth: Holden.
Ruth: “The nature of tournaments.” Economic Theory, 51(2): 289–313 Barut, Yasar, and Dan Kovenock.
Title and authors: Ruth: So we're starting by looking at the title and who wrote this paper, "Robust prizes in all-pay contests with loss aversion." It immediately tells us the core tension is between designing a fair prize structure and dealing with agents who are loss-averse.
Warren: The authors are Dmitry Ryvkin and Qin Wu. They’re bringing in expertise from economics and maybe some of the quantitative modeling side, which makes sense given how they're tackling this uncertainty.
Iris: I remember reading earlier work on prize design, but this paper adds the layer of ambiguity about the effort costs, which is a significant addition to the standard models we see in contests.
Ruth: That’s right. The main implication here is that if you’re designing a contest and you can’t know exactly how much it costs an agent to put in effort, your prize schedule needs to be robust enough that it works even if the cost function is something completely different than what you initially guessed.
Warren: It suggests that simply picking a prize structure based on one specific cost estimate isn't enough; you have to design for the worst-case scenario within that set of possible costs.
Iris: And the authors explicitly state their goal is to maximize expected expenditure in a way that guarantees an equilibrium exists, which is a strong identification strategy because it doesn't depend on knowing the exact cost function beforehand.
Ruth: So what they’re doing is finding this robust prize structure, and they find that under moderate loss aversion, the result looks pretty standard—basically equal prizes for everyone except the last person.
Warren: But as you ramp up that loss aversion parameter, things get interesting; the bottom prize starts to move from zero to something positive and even increases with more loss aversion.
Iris: That shift in the optimal bottom prize suggests that for very strong loss aversion, a simple setup where everyone gets something or nothing isn't the best outcome.
Ruth: It points toward a situation where you have to consider how participation itself changes when agents are really averse to losing out on an opportunity.
Warren: And that brings us right into the heart of the paper’s main contribution—how loss aversion directly influences participation and prize allocation simultaneously, especially under strong conditions.
The paper's summary: Iris: Moving into the summary, they outline how they set up their model using an all-pay contest with agents who have expectation-based loss aversion, which is a specific way agents feel like losing when outcomes aren't what they expected.
Ruth: They frame the problem as a manager trying to allocate rewards while facing two major hurdles: knowing that agents are loss-averse, and not knowing the exact effort costs of those agents.
Warren: The key finding they highlight is that under moderate loss aversion, the robustly optimal prize schedule awards equal prizes to everyone except for the one ranked last contestant.
Iris: But then they show a crucial transition point: when loss aversion becomes sufficiently strong, that bottom prize doesn't stay at zero; it becomes strictly positive and keeps growing as the strength of that loss aversion increases.
Ruth: That means the optimal prize differential between the best and worst performers changes as the agents get more risk-averse about not winning anything at all.
Warren: And they connect this to a psychological idea—the agent prefers getting a guaranteed outcome, even if it's less than what they might win in a lottery, especially when loss aversion is high enough that the expected benefit calculation becomes negative for small prize differences.
Iris: The methodology involves separating the equilibrium behavior from the ambiguity about costs by using linear cost functions as a representative of all possible costs, which gives them a uniform upper bound on effort expenditure across every possible symmetric equilibrium.
Ruth: That upper bound is what they use to construct the optimal schedule, showing that even with cost ambiguity, you can still define a robust design that guarantees an equilibrium exists.
Warren: So in short, they show how loss aversion dictates a shift from giving nothing to the last person when conditions are severe, and this happens because of those psychological factors interacting with the uncertainty about effort costs.
The paper's improvements: Iris: One major improvement they point out is that their approach separates equilibrium behavior from the ambiguity about effort costs, which allows them to use a linear cost function as a representative of the entire ambiguity set.
Ruth: That separation is vital because it means they don't have to worry about every single possible cost function; they just need to find one linear cost function that’s least favorable for their prize schedule.
Warren: This leads directly to the robust prize structure, which they describe as having a "punish-the-bottom" shape: a common prize for ranks one through n minus one, and a lower prize specifically for the last rank.
Iris: They also find that under strong loss aversion, the optimal bottom prize becomes strictly positive and increases with loss aversion, which contrasts with moderate loss aversion where it stays zero.
Ruth: This is a big piece of information for designers because it tells you exactly when you should expect to see a positive minimum payout for the lowest rank.
Warren: They also found that under strong loss aversion, the robustly optimal solution involves partial participation in the contest, meaning some agents might choose not to enter even if they are willing to compete.
Iris: That suggests that maximizing total participation isn't always the goal when loss aversion is high; sometimes it's better to have fewer people enter if it leads to a more favorable prize structure for those who do participate.
Ruth: So, the improvement here is moving beyond just finding *an* equilibrium and instead constructing a schedule that remains optimal across *all* possible cost functions within the ambiguity set.
Warren: It’s about building resilience into the design itself rather than assuming you know all the underlying dynamics perfectly.
Conclusion: Ruth: Wrapping up, what this paper by Ryvkin and Wu does is show how to robustly design prizes in contests when you don't know exactly how much effort agents will put in, factoring in loss aversion.
Warren: The main implication for practitioners is that if you’re designing a contest, especially one involving voluntary effort, you need to account for the possibility that the optimal structure changes based on how risk-averse your participants are.
Iris: Specifically, the finding that under strong loss aversion, partial participation becomes optimal means we shouldn't always push for maximum entry numbers in these types of settings.
Ruth: That’s right. The paper shows that as loss aversion gets stronger, the prize structure shifts from a zero bottom prize to one that is strictly positive and even increases with the strength of that loss aversion.
Warren: So, to summarize the robust design problem in "Robust prizes in all-pay contests with loss aversion," they found a structure where equal prizes are given for everyone except the last rank under moderate conditions, but that flips when loss aversion is strong.
Iris: And they show that this robustness holds even though the manager only has a linear idea of how effort costs scale.
Ruth: It’s a careful look at how psychological factors like loss aversion interact with economic constraints like budget limits and uncertainty about agent behavior.
Warren: It gives us a concrete way to think about setting up these contests when you can't have perfect information on the underlying cost functions.
More episodes
- Daily Summary for 2026-11-19
- 2608.21274-Recommendation Quality and the Concentration of Consumption: Experimental Evidence from Netflix
- 2408.02757-A nonparametric test for diurnal variation in spot correlation processes
- 2306.16393-High-Dimensional Canonical Correlation Analysis
- 2605.17299-Geometric Brownian motion with intermittent entries and exits
- 2607.04753-Fooling Yourself: how narratives shape beliefs
- 2408.14671-Inference on Treatment Effects with Repeatedly Mismeasured High-Dimensional Controls
- 2201.06898-Difference-in-Differences Estimators for Treatments Continuously Distributed at Every Period
- 2501.15307-Influence Functions: A Unified Differentiation-Based Framework
- The day in conversation for 2026-11-19