Pith. sign in

REVIEW 2 major objections 4 minor 151 references

Ex-ante minimax rules often prescribe actions the researcher would reject after seeing the data; two new criteria keep decisions optimal both before and after the sample arrives.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-14 11:05 UTC pith:ASV2NOK2

load-bearing objection Clean formalization of interim credibility for frequentist minimax rules, plus solid axiomatizations of two natural fixes that nest known special cases. the 2 major comments →

arxiv 2607.10519 v1 pith:ASV2NOK2 submitted 2026-07-12 econ.EM econ.THmath.STstat.TH

Dynamically Consistent Statistical Decisions

Cheaheon Lim , Yechan Park This is my paper
classification econ.EM econ.THmath.STstat.TH
keywords dynamic consistencyminimax regretstatistical decision theorytreatment choiceleast favorable prioras-if optimizationaxiomatic decision theory
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Many statistical decision rules used in econometrics are justified by ex-ante guarantees such as minimax regret. After the data are observed, a researcher may find those same prescriptions no longer make sense. The paper shows this dynamic inconsistency is not rare: in treatment choice and evidence-aggregation problems, least-favorable-prior updates that rationalize the ex-ante rule concentrate on parameter values the realized sample makes highly implausible. The authors therefore introduce and axiomatize two classes of criteria—dynamically consistent minimax loss and dynamically consistent minimax regret—that force the decision maker to anticipate her interim procedure and choose an ex-ante rule that remains optimal after every data realization. These criteria nest familiar plug-in and robust-Bayesian procedures while remaining semi-Bayesian: they need only a subjective marginal over samples, not a full prior over a high-dimensional state space. Real and simulated applications illustrate that the new rules avoid both the excessive conservatism and the occasional aggressiveness of classical minimax regret.

Core claim

Ex-ante minimax-loss and minimax-regret rules are dynamically consistent precisely when their interim actions coincide with the conditional Bayes problem induced by a least-favorable prior. Two new optimality criteria—dynamically consistent minimax loss and dynamically consistent minimax regret—aggregate interim worst-case payoffs with a subjective marginal over samples and are the unique preference and choice representations that remain optimal after every data realization while nesting Manski’s as-if optimization and Gamma*-minimax.

What carries the argument

The least-favorable-prior representation of minimax optimality (Proposition 1) together with the two axiomatized criteria DC-MML and DC-MMR, which evaluate a rule by the expected interim worst-case payoff or regret under a subjective marginal µ over the sample and a correspondence of interim belief sets Q_z.

Load-bearing premise

The decision maker must already possess a well-defined subjective distribution over possible samples even when she cannot form a prior over the high-dimensional state space.

What would settle it

In a binary treatment-choice experiment with known arm sizes, compute the least-favorable distance d_N; if after a large positive estimated treatment effect the researcher still finds the least-favorable posterior (supported only on |µ1−µ0|=d_N) a credible interim belief, the paper’s claim that classical minimax lacks interim credibility is false.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper studies dynamic consistency of frequentist statistical decision rules justified by ex ante criteria such as minimax loss and minimax regret. It shows that any interim criterion dynamically consistent with an ex ante minimax rule must coincide with the conditional Bayes problem induced by a least-favorable prior (Proposition 1), and uses this fact as a diagnostic: in treatment choice (Stoye 2009) and evidence aggregation (Yata 2021), least-favorable posteriors concentrate on effect sizes that the realized data make implausible, so a researcher may wish to deviate from the ex ante rule after seeing the data. The authors then define and axiomatize two dynamically consistent criteria (DC-MML and DC-MMR) that aggregate interim worst-case objectives with a subjective marginal µ over the sample space and an interim belief correspondence z ↦ Q_z; these nest Manski’s as-if optimization and Lim’s Gamma*-minimax. Applications to Bursztyn et al. (2020) and a simulated evidence-aggregation design illustrate that as-if MMR avoids the interim pathologies of pure minimax regret.

Significance. The contribution is substantial for econometric decision theory. The literature has produced many finite-sample and asymptotic minimax-regret rules for treatment choice and policy learning, yet has largely ignored whether those rules remain optimal after the data realize. Formalizing the interim problem via least-favorable priors, documenting empirically relevant inconsistencies, and supplying complete axiomatizations of dynamically consistent alternatives that nest existing proposals (Manski as-if, Gamma*) fills a clear gap. The proofs of Propositions 1–3 and Theorems 1–2 are standard game-theoretic and Anscombe–Aumann arguments and appear carefully checked; the empirical illustrations use public data and transparent Monte-Carlo designs. The framework also clarifies when a researcher can use a subjective marginal over data without a full prior over a high-dimensional state space—an attractive semi-Bayesian middle ground.

major comments (2)
  1. Section 6.2 and Axioms 1/5: the DC criteria require a well-defined full-support subjective marginal µ over Z. When Θ is high-dimensional this is presented as weaker than a prior over Θ, yet the paper never discusses how a practitioner should elicit or validate µ, nor what happens under misspecification of µ. Because the ex-ante ranking is defined by aggregation under µ, this is load-bearing for the claim that DC rules are operationally usable; a short discussion of elicitation or robustness would strengthen the contribution.
  2. Section 7.2 / Table 1 and Figure 2: the empirical illustration enumerates overlapping covariate cells (Appendix B.1) rather than a single partition. While each cell is a valid randomized experiment, the reported counts of disagreements (e.g., 18 subgroups at α=0.10) are therefore not independent observations. The qualitative message is unaffected, but the paper should state more clearly that the exercise is illustrative rather than a formal frequency claim about the population of cells.
minor comments (4)
  1. Page 12, Remark 1: the distinction between “too conservative” and “not conservative enough” forms of dynamic inconsistency is useful; a one-sentence pointer back to Figure 1 would help the reader locate the quantitative illustration.
  2. Figure 3 caption: the pink/beige/blue regions are described in the text but the color legend in the figure itself is dense; a short key in the caption would improve readability.
  3. Appendix A.2, Example A.1: the numerical value d_1100 ≈ 0.023 is given without stating the optimization routine or tolerance; a one-line note would aid replication.
  4. References: a few recent related papers on pre-analysis plans and robust Bayes (already cited in the introduction) could be cross-referenced again in Section 7 when the interim credibility of pre-committed rules is discussed.

Circularity Check

0 steps flagged

No significant circularity: DC criteria are intentionally defined to aggregate interim objectives, and the axiomatizations and least-favorable-prior rationalization are independent representation arguments.

full rationale

The paper’s load-bearing claims do not reduce to their own inputs by construction in the sense of the circularity patterns. Proposition 1 is a standard iterated-expectation / pasting argument: once a least-favorable prior π* supports an ex ante minimax rule, the rule’s continuation is posterior-Bayes almost surely under P_π*; π* is an endogenous saddle-point object of the zero-sum game, not a free parameter fitted to force interim agreement. Definitions 2–3 deliberately build dynamic consistency by aggregating interim Q_z objectives with a marginal µ; that is design of a criterion class, not a claimed independent derivation of consistency from something else. Theorems 1–2 recover (µ,{Q_z}) from independent axioms (Z-SEU / IIA on Z-measurable acts, Θ-MEU or Stoye’s endogenous-prior regret axioms plus Z-separability); the representations are standard and do not smuggle the target optimality into the primitives. Nesting of Manski as-if (Q_z = Δ(Θ̂(z))) and Lim’s Gamma*-minimax (Q_z = Π_z) is exact specialization of the same objects, not a prediction forced by a fit. The Lim (2026) self-citation is only used to exhibit nesting and is not a uniqueness theorem that forbids alternatives for the main results. Empirical diagnostics compare least-favorable posteriors to conventional evidence without fitting parameters that are then re-labeled as predictions. No equation equates a claimed first-principles result to a quantity defined by that same result.

Axiom & Free-Parameter Ledger

0 free parameters · 6 axioms · 2 invented entities

The central claims rest on standard decision-theoretic axioms plus two domain-specific separability conditions that isolate beliefs about data from interim ambiguity about states. No free parameters are fitted; the invented objects are the two new criteria and the interim belief correspondence, both defined explicitly and characterized.

axioms (6)
  • standard math Z-SEU: preference restricted to Z-measurable acts is full-support subjective expected utility (Axiom 1)
    Recovers the unique marginal µ over data; standard Anscombe–Aumann SEU.
  • standard math Θ-MEU: every conditional preference ≿_z is maxmin expected utility (Axiom 3)
    Gilboa–Schmeidler representation of interim ambiguity sets Q_z.
  • standard math Stoye (2011a) axioms for endogenous-prior minimax regret (non-triviality, monotonicity, independence, INA, mixture continuity, ambiguity aversion, C-betweenness) (Axiom 4)
    Imported wholesale to obtain the regret representation before imposing dynamic consistency.
  • domain assumption Z-Separability / Z-Separable Choice: ranking or choice of acts that differ only at a single data realization z is independent of the continuation outside z (Axioms 2 and 6)
    Encodes the requirement that unrealized data values are counterfactual and should not affect interim comparisons; the key new behavioral restriction.
  • standard math IIA on Z-measurable acts (Axiom 5)
    Strengthens Stoye’s constant-act IIA so that the marginal µ can be recovered uniquely.
  • domain assumption D closed under measurable pasting
    Technical feasibility condition used in the proof of Proposition 1 to construct improving deviations.
invented entities (2)
  • Dynamically Consistent Minimax Loss (DC-MML) and Dynamically Consistent Minimax Regret (DC-MMR) no independent evidence
    purpose: Ex-ante criteria that average interim worst-case payoffs under a subjective marginal µ and an arbitrary interim belief correspondence z ↦ Q_z
    Defined in Definitions 2–3; shown to be the unique representations of the stated axioms and to nest existing procedures.
  • Interim belief correspondence z ↦ Q_z no independent evidence
    purpose: Allows any statistical procedure (confidence region, identified set, posterior set, …) after data realization without requiring Bayesian updating of a prior on Θ
    Primitive of the DC criteria; recovered uniquely from the conditional MEU preferences.

pith-pipeline@v1.1.0-grok45 · 33623 in / 2750 out tokens · 31298 ms · 2026-07-14T11:05:46.596388+00:00 · methodology

0 comments
read the original abstract

A large literature in econometrics proposes decision rules with optimality guarantees based on ex ante criteria, such as minimax regret. We develop a framework for analyzing the dynamic consistency of such rules and show that, in many empirically relevant settings, the researcher may wish to deviate from the interim prescription of ex ante optimal rules after observing the data realization. To address this problem, we propose and axiomatize two classes of optimality criteria that yield dynamically consistent decision rules.

Figures

Figures reproduced from arXiv: 2607.10519 by Cheaheon Lim, Yechan Park.

Figure 1
Figure 1. Figure 1: Probability that the minimax-regret rule assigns [PITH_FULL_IMAGE:figures/full_fig_p013_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Every covariate cell of the twelve-characteristic partition (at least five [PITH_FULL_IMAGE:figures/full_fig_p027_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Color-coded regions over the (z1, z2) plane based on the interim pre￾scriptions of the δ ∗ , Tα, and as-if MMR rules (σ = 1, C1 = 2, C2 = 1). vative: it introduces the policy only when the lower endpoint L is positive. By contrast, the as-if MMR rule introduces the policy whenever the confidence set is centered above zero. Hence, whenever the hypothesis testing rule adopts the policy, the as-if MMR rule do… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

151 extracted references · 1 canonical work pages

  1. [1]

    A Smooth Model of Decision Making under Ambiguity , urldate =

    Peter Klibanoff and Massimo Marinacci and Sujoy Mukerji , journal =. A Smooth Model of Decision Making under Ambiguity , urldate =

  2. [2]

    2007 , author =

    A subjective model of experimentation , journal =. 2007 , author =

    2007

  3. [3]

    Subjective Probability and Expected Utility without Additivity , urldate =

    David Schmeidler , journal =. Subjective Probability and Expected Utility without Additivity , urldate =

  4. [4]

    1989 , issn =

    Maxmin expected utility with non-unique prior , journal =. 1989 , issn =. doi:https://doi.org/10.1016/0304-4068(89)90018-9 , url =

    doi:10.1016/0304-4068(89)90018-9 1989

  5. [5]

    Part I , journal =

    Knightian decision theory. Part I , journal =. 2002 , author =

    2002

  6. [6]

    Objective and Subjective Rationality in a Multiple Prior Model , urldate =

    Itzhak Gilboa and Fabio Maccheroni and Massimo Marinacci and David Schmeidler , journal =. Objective and Subjective Rationality in a Multiple Prior Model , urldate =

  7. [7]

    Objective rationality foundations for (dynamic) -MEU , journal =

    Mira Frick and Ryota Iijima and Yves. Objective rationality foundations for (dynamic) -MEU , journal =. 2022 , issn =. doi:https://doi.org/10.1016/j.jet.2021.105394 , url =

    doi:10.1016/j.jet.2021.105394 2022

  8. [8]

    2004 , issn =

    Differentiating ambiguity and ambiguity attitude , journal =. 2004 , issn =. doi:https://doi.org/10.1016/j.jet.2003.12.004 , url =

    doi:10.1016/j.jet.2003.12.004 2004

  9. [9]

    Decision Making with Belief Functions: Compatibility and Incompatibility with the Sure-Thing Principle , urldate =

    Jean-Yves Jaffray and Peter Wakker , journal =. Decision Making with Belief Functions: Compatibility and Incompatibility with the Sure-Thing Principle , urldate =

  10. [10]

    2007 , issn =

    Choice under uncertainty with the best and worst in mind: Neo-additive capacities , journal =. 2007 , issn =. doi:https://doi.org/10.1016/j.jet.2007.01.017 , author =

    doi:10.1016/j.jet.2007.01.017 2007

  11. [11]

    Expected Uncertain Utility Theory , urldate =

    Faruk Gul and Wolfgang Pesendorfer , journal =. Expected Uncertain Utility Theory , urldate =

  12. [12]

    mimeo , year =

    Igor Kopylov , title =. mimeo , year =

  13. [13]

    Coping with Ignorance: Unforeseen Contingencies and Non-Additive Uncertainty , urldate =

    Paolo Ghirardato , journal =. Coping with Ignorance: Unforeseen Contingencies and Non-Additive Uncertainty , urldate =

  14. [14]

    Proceedings of the Eleventh International Symposium on Imprecise Probabilities: Theories and Applications , pages =

    An Axiomatic Utility Theory for Dempster-Shafer Belief Functions , author =. Proceedings of the Eleventh International Symposium on Imprecise Probabilities: Theories and Applications , pages =. 2019 , volume =

    2019

  15. [15]

    A. P. Dempster , journal =. A Generalization of Bayesian Inference , urldate =

  16. [16]

    A. P. Dempster , journal =. Upper and Lower Probabilities Generated by a Random Closed Interval , urldate =

  17. [17]

    A. P. Dempster , journal =. New Methods for Reasoning Towards Posterior Distributions Based on Sample Data , urldate =

  18. [18]

    A Mathematical Theory of Evidence , urldate =

    Glenn Shafer , publisher =. A Mathematical Theory of Evidence , urldate =

  19. [19]

    Additive Representations of Non-Additive Measures and the Choquet Integral , volume =

    Gilboa, Itzhak and Schmeidler, David , year =. Additive Representations of Non-Additive Measures and the Choquet Integral , volume =. Annals of Operations Research , doi =

  20. [20]

    Savage , editor =

    Leonard J. Savage , editor =. The Foundations of Statistics , year =

  21. [21]

    Risk, Ambiguity, and the Savage Axioms , urldate =

    Daniel Ellsberg , journal =. Risk, Ambiguity, and the Savage Axioms , urldate =

  22. [22]

    1990 , issn =

    Decision analysis using belief functions , journal =. 1990 , issn =. doi:https://doi.org/10.1016/0888-613X(90)90014-S , author =

    doi:10.1016/0888-613x(90)90014-s 1990

  23. [23]

    Modeling the Change of Paradigm: Non-Bayesian Reactions to Unexpected News , urldate =

    Pietro Ortoleva , journal =. Modeling the Change of Paradigm: Non-Bayesian Reactions to Unexpected News , urldate =

  24. [24]

    Revisiting Savage in a Conditional World , urldate =

    Paolo Ghirardato , journal =. Revisiting Savage in a Conditional World , urldate =

  25. [25]

    2003 , issn =

    Recursive multiple-priors , journal =. 2003 , issn =. doi:https://doi.org/10.1016/S0022-0531(03)00097-8 , author =

    doi:10.1016/s0022-0531(03)00097-8 2003

  26. [26]

    Proceedings of the National Academy of Sciences , volume =

    Simone Cerreia-Vioglio and Fabio Maccheroni and Massimo Marinacci and Luigi Montrucchio , title =. Proceedings of the National Academy of Sciences , volume =

  27. [27]

    2023 , issn =

    Strength of preference over complementary pairs axiomatizes alpha-MEU preferences , journal =. 2023 , issn =

    2023

  28. [28]

    Symmetry or Dynamic Consistency? , author =. The B.E. Journal of Theoretical Economics , year =

  29. [29]

    Epstein and Kyoungwon Seo , journal =

    Larry G. Epstein and Kyoungwon Seo , journal =. Symmetry of evidence without evidence of symmetry , volume =

  30. [30]

    2003 , author =

    IID: independently and indistinguishably distributed , journal =. 2003 , author =

    2003

  31. [31]

    2024 , url =

    Alpha-maxmin as an aggregation of two selves , journal =. 2024 , url =

    2024

  32. [32]

    2009 , author =

    Subjective states: A more robust model , journal =. 2009 , author =

    2009

  33. [33]

    2007 , author =

    Coarse contingencies and ambiguity , journal =. 2007 , author =

    2007

  34. [34]

    2014 , issn =

    A theory of subjective learning , journal =. 2014 , issn =

    2014

  35. [35]

    Preference for Flexibility

    David M. Kreps , journal =. A Representation Theorem for "Preference for Flexibility" , volume =

  36. [36]

    and Border, Kim C

    Aliprantis, Charalambos D. and Border, Kim C. , publisher =

  37. [37]

    Random Choice and Private Information , volume =

    Jay Lu , journal =. Random Choice and Private Information , volume =

  38. [38]

    Random Ambiguity , volume =

    Jay Lu , journal =. Random Ambiguity , volume =

  39. [39]

    2006 , author =

    A Behavioral Characterization of Plausible Priors , journal =. 2006 , author =

    2006

  40. [40]

    2018 , author =

    Blackwell's informativeness theorem using diagrams , journal =. 2018 , author =

    2018

  41. [41]

    2025 , author =

    A Theory of Coarse Information Structures , journal =. 2025 , author =

    2025

  42. [42]

    and Schneider, Martin , title =

    Epstein, Larry G. and Schneider, Martin , title =. The Review of Economic Studies , volume =. 2007 , month =

    2007

  43. [43]

    2024 , author =

    On comparisons of information structures with infinite states , journal =. 2024 , author =

    2024

  44. [44]

    2025 , author =

    A simple proof of Blackwell’s theorem on the comparison of experiments for a general state space , journal =. 2025 , author =

    2025

  45. [45]

    Equivalent comparisons of experiments , journal =

    David Blackwell , pages =. Equivalent comparisons of experiments , journal =

  46. [46]

    Comparison of experiments , journal =

    David Blackwell , volume =. Comparison of experiments , journal =

  47. [47]

    2021 , author =

    A Proof of Blackwell's Theorem , journal =. 2021 , author =

    2021

  48. [48]

    2019 , author =

    Ambiguous persuasion , journal =. 2019 , author =

    2019

  49. [49]

    Mechanism Design with Ambiguous Communication Devices , volume =

    Subir Bose and Ludovic Renou , journal =. Mechanism Design with Ambiguous Communication Devices , volume =

  50. [50]

    Bayesian Persuasion , volume =

    Emir Kamenica and Matthew Gentzkow , journal =. Bayesian Persuasion , volume =

  51. [51]

    2016 , author =

    Bayesian persuasion with heterogeneous priors , journal =. 2016 , author =

    2016

  52. [52]

    2024 , author =

    Dynamic Consistency and Ambiguous Communication , journal =. 2024 , author =

    2024

  53. [53]

    2024 , journal=

    Persuasion with Ambiguous Communication , author=. 2024 , journal=

    2024

  54. [54]

    2025 , journal=

    Ambiguous Persuasion: An Ex-Ante Formulation , author=. 2025 , journal=

    2025

  55. [55]

    2021 , journal=

    A Concavification Approach to Ambiguous Persuasion , author=. 2021 , journal=

    2021

  56. [56]

    2017 , journal=

    Public Persuasion , author=. 2017 , journal=

    2017

  57. [57]

    2024 , author =

    Informativeness orders over ambiguous experiments , journal =. 2024 , author =

    2024

  58. [58]

    2012 , author =

    Informativeness of experiments for meu , journal =. 2012 , author =

    2012

  59. [59]

    2016 , author =

    Blackwell's informativeness ranking with uncertainty-averse preferences , journal =. 2016 , author =

    2016

  60. [60]

    2015 , author =

    Comparisons of Ambiguous Experiments , journal =. 2015 , author =

    2015

  61. [61]

    2022 , author =

    Biased learning under ambiguous information , journal =. 2022 , author =

    2022

  62. [62]

    2025 , author =

    Sequential learning under informational ambiguity , journal =. 2025 , author =

    2025

  63. [63]

    American Economic Review , Volume =

    Auster, Sarah and Che, Yeon-Koo and Mierendorff, Konrad , Title =. American Economic Review , Volume =

  64. [64]

    Econometrica , volume =

    Dütting, Paul and Feldman, Michal and Peretz, Daniel and Samuelson, Larry , title =. Econometrica , volume =

  65. [65]

    Theoretical Economics , volume =

    Wolitzky, Alexander , title =. Theoretical Economics , volume =

  66. [66]

    2025 , author =

    Contagious Ambiguity , journal =. 2025 , author =

    2025

  67. [67]

    Theoretical Economics , volume =

    Kosterina, Svetlana , title =. Theoretical Economics , volume =

  68. [68]

    Model Uncertainty , volume =

    Massimo Marinacci , journal =. Model Uncertainty , volume =

  69. [69]

    R. H. Strotz , journal =. Myopia and Inconsistency in Dynamic Utility Maximization , volume =

  70. [70]

    2023 , issn =

    Ambiguous information and dilation: An experiment , journal =. 2023 , issn =

    2023

  71. [71]

    2021 , author =

    Evaluating ambiguous random variables from Choquet to maxmin expected utility , journal =. 2021 , author =

    2021

  72. [72]

    Choice deferral and ambiguity aversion , volume =

    Kopylov, Igor , year =. Choice deferral and ambiguity aversion , volume =

  73. [73]

    SSRN Electronic Journal , year =

    Experimental Persuasion , author =. SSRN Electronic Journal , year =

  74. [74]

    and Maschler, Michael B

    Aumann, Robert J. and Maschler, Michael B. , title =. 1995 , publisher =

    1995

  75. [75]

    A Rule For Updating Ambiguous Beliefs , volume =

    Pires, Cesaltina , year =. A Rule For Updating Ambiguous Beliefs , volume =. Theory and Decision , doi =

  76. [76]

    SSRN Electronic Journal , year =

    Raphael Boleslavsky and Kyungmin Kim , title =. SSRN Electronic Journal , year =. doi:10.2139/ssrn.2913669 , url =

    work page doi:10.2139/ssrn.2913669

  77. [77]

    Econometrica , volume =

    Giacomini, Raffaella and Kitagawa, Toru , title =. Econometrica , volume =

  78. [78]

    Robust Bayesian Analysis for Econometrics , booktitle=

    Giacomini, Raffaella and Kitagawa, Toru and Read, Matthew , editor=. Robust Bayesian Analysis for Econometrics , booktitle=. 2025 , pages=

    2025

  79. [79]

    Model and Predictive Uncertainty: A Foundation for Smooth Ambiguity Preferences , volume =

    Denti, Tommaso and Pomatto, Luciano , year =. Model and Predictive Uncertainty: A Foundation for Smooth Ambiguity Preferences , volume =. Econometrica , doi =

  80. [80]

    Model and Predictive Uncertainty: A Foundation for Smooth Ambiguity Preferences , journal =

    Denti, Tommaso and Pomatto, Luciano , year =. Model and Predictive Uncertainty: A Foundation for Smooth Ambiguity Preferences , journal =

Showing first 80 references.