A Characterization of the n-th Degree Bounded Stochastic Dominance
Pith reviewed 2026-05-24 10:34 UTC · model grok-4.3
The pith
The n-th degree bounded stochastic dominance order holds precisely when all utilities with globally bounded Arrow-Pratt risk aversion on [a,b] rank the dominating lottery higher.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
The n-th degree BSD order on lotteries with support in [a,b] is equivalent to the requirement that every utility function on [a,b] whose Arrow-Pratt absolute risk aversion is globally bounded above (or that satisfies the n-convexity condition) ranks the dominating lottery at least as high as the dominated one. The order therefore encodes a specific class of risk preferences whose tolerance is controlled by the endpoints of the interval.
What carries the argument
n-th degree bounded stochastic dominance order, equivalently characterized by the class of utilities on [a,b] with globally bounded Arrow-Pratt risk aversion or by n-convexity on that interval.
If this is right
- Comparative statics for choice under uncertainty hold for all agents whose risk aversion is globally bounded on the relevant interval.
- Savings decisions under globally bounded prudence measures obey the same ordering.
- The lower-partial-moment order cleanly separates the largest payoff from the upper bound on risk aversion.
- New inequalities hold for all n-convex functions on the interval.
Where Pith is reading between the lines
- The dependence on a fixed [a,b] suggests that applications should choose the interval from observable economic bounds rather than treat it as arbitrary.
- The lower-partial-moment order may be easier to apply in portfolio problems where only downside protection matters.
- The characterization could be used to test whether observed choices are consistent with bounded rather than unbounded risk aversion.
Load-bearing premise
Lotteries have bounded support contained in a fixed interval [a,b] and the relevant utility functions are defined on that same interval.
What would settle it
Find a pair of lotteries on [a,b] and a decision maker whose Arrow-Pratt risk aversion exceeds the bound implied by the interval yet still ranks the BSD-dominant lottery higher; this would refute the claimed equivalence.
read the original abstract
We provide a novel characterization of the $n$-th degree bounded stochastic dominance (BSD) order, linking it to the risk tolerance of decision-makers and providing a decision-theoretic foundation for these stochastic orders. Our results reveal that BSD reflects specific risk preferences through the choice of the interval $[a,b]$, by characterizing it in terms of utility functions with globally bounded Arrow--Pratt risk aversion or that satisfy an $n$-convexity condition. They also highlight limitations of BSD, including its dependence on the chosen support interval and the resulting peculiar risk aversion behavior of decision-makers included in the generator of BSD. To partially address this issue, we use our characterization to separate two roles that are combined in BSD: the largest payoff in the lotteries and the upper endpoint of the interval that determines the Arrow--Pratt lower bound. We then introduce a related lower-partial-moment order that provides a clean trade-off between expected value and downside-risk protection. Using our characterization, we present comparative statics results for decision-making under uncertainty with globally bounded risk aversion measures and savings decisions under globally bounded prudence measures, and derive inequalities for $n$-convex functions.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper claims to deliver a novel characterization of the n-th degree bounded stochastic dominance (BSD) order, showing that it corresponds to expected-utility maximization over the class of utilities with globally bounded Arrow-Pratt absolute risk aversion on a fixed interval [a,b] (or, equivalently, n-convex utilities on that interval). It explicitly flags the resulting dependence of the order on the exogenous choice of [a,b] and the associated “peculiar risk aversion behavior,” partially addresses the issue by separating the role of the largest payoff from the upper endpoint that sets the risk-aversion bound, introduces a related lower-partial-moment order, and derives comparative-statics results for savings and decision problems under globally bounded risk aversion or prudence together with inequalities for n-convex functions.
Significance. If the stated characterization theorems are correct, the work supplies a clean decision-theoretic foundation for an existing family of stochastic orders and usefully isolates their dependence on the support interval. The lower-partial-moment order offers a potentially cleaner alternative that trades expected value against downside-risk protection without the same interval arbitrariness. The comparative-statics applications and the n-convex-function inequalities are of independent interest for economic theory.
major comments (2)
- [main characterization theorem (likely §3)] The characterization (abstract and the main theorem linking BSD to bounded Arrow-Pratt risk aversion on [a,b]) is load-bearing for the paper’s central claim yet leaves the choice of [a,b] exogenous; while the manuscript acknowledges that altering [a,b] can change whether X ≽_n Y holds for fixed lotteries, it supplies no canonical rule for selecting the interval from the supports alone. This is the precise limitation the paper itself flags, and it directly affects the practical applicability of the order.
- [lower-partial-moment order section] § on the lower-partial-moment order: the new order is introduced to separate the largest payoff from the upper endpoint that determines the Arrow-Pratt bound, but the manuscript does not state whether the new order coincides with, is weaker than, or is stronger than standard n-th degree stochastic dominance when the interval is chosen to contain the supports; without this comparison the claimed “clean trade-off” remains informal.
minor comments (2)
- [preliminaries] Notation for the n-convexity condition and the precise statement of the generator set of utilities should be displayed in a single definition box for easy reference.
- [applications section] The comparative-statics propositions would benefit from an explicit statement of the maintained assumptions on the utility class (bounded risk aversion or bounded prudence) before the results are stated.
Simulated Author's Rebuttal
We thank the referee for the detailed and constructive report. We address each major comment below. The manuscript already emphasizes the exogenous nature of the interval [a,b] as an inherent feature of bounded stochastic dominance rather than a shortcoming of the characterization.
read point-by-point responses
-
Referee: [main characterization theorem (likely §3)] The characterization (abstract and the main theorem linking BSD to bounded Arrow-Pratt risk aversion on [a,b]) is load-bearing for the paper’s central claim yet leaves the choice of [a,b] exogenous; while the manuscript acknowledges that altering [a,b] can change whether X ≽_n Y holds for fixed lotteries, it supplies no canonical rule for selecting the interval from the supports alone. This is the precise limitation the paper itself flags, and it directly affects the practical applicability of the order.
Authors: We agree that the interval [a,b] is exogenous and that its choice can alter the ranking for fixed lotteries; the paper explicitly flags this dependence and the associated risk-aversion behavior as a limitation of BSD. The characterization is deliberately stated for a fixed interval because that is how the order is defined in the literature. No canonical selection rule from the supports alone is supplied because none exists: the upper endpoint encodes the decision maker’s global risk-aversion bound, which is a modeling primitive separate from the lotteries’ supports. This is not a defect in the theorem but a substantive feature the characterization makes transparent. revision: no
-
Referee: [lower-partial-moment order section] § on the lower-partial-moment order: the new order is introduced to separate the largest payoff from the upper endpoint that determines the Arrow-Pratt bound, but the manuscript does not state whether the new order coincides with, is weaker than, or is stronger than standard n-th degree stochastic dominance when the interval is chosen to contain the supports; without this comparison the claimed “clean trade-off” remains informal.
Authors: We accept the point. When the interval is chosen to contain the supports, the lower-partial-moment order is strictly weaker than n-th degree stochastic dominance (it coincides with the latter only in the limit as the upper endpoint tends to infinity). We will add a short proposition and accompanying discussion making this comparison explicit, thereby clarifying the sense in which the new order isolates downside-risk protection from interval effects. revision: yes
Circularity Check
No circularity: standard mathematical characterization of BSD via utility classes
full rationale
The paper derives a characterization theorem equating n-th degree BSD (defined on lotteries with support in [a,b]) to dominance under the class of utilities with globally bounded Arrow-Pratt risk aversion (or n-convexity) on that same interval. This is an equivalence proof resting on standard expected-utility axioms and integral representations of stochastic orders; it does not reduce the claimed result to a fitted parameter, a self-citation chain, or a definition that presupposes the target order. The explicit dependence on the exogenous interval [a,b] is flagged by the authors as a limitation rather than smuggled in via ansatz or self-reference. No load-bearing step matches any of the enumerated circularity patterns; the central claim remains independent of its own outputs.
Axiom & Free-Parameter Ledger
axioms (2)
- domain assumption Decision makers maximize expected utility with sufficiently smooth (n-times differentiable) utility functions.
- domain assumption Lotteries have bounded support contained in a fixed interval [a,b].
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.