pith. sign in

arxiv: 2605.22076 · v1 · pith:INK3JTOTnew · submitted 2026-05-21 · 💰 econ.GN · q-fin.EC

Isomorphic Dynamic Programs

John Stachurski , Junnan Zhang This is my paper

Pith reviewed 2026-05-22 02:45 UTC · model grok-4.3

classification 💰 econ.GN q-fin.EC
keywords dynamic programsorder isomorphismEpstein-Zin preferencesKreps-Porteus preferencesrisk-sensitive preferencesvalue function approximationreal business cycle modelsconjugacy methods
0
0 comments X

The pith

When two dynamic programs are connected by an order isomorphism, optimality properties transmit from one formulation to the other.

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

The paper applies conjugacy methods from dynamical systems to relate different dynamic programs in economics. It establishes that an order isomorphism between two programs allows optimality conditions, solutions, and properties to carry over directly because the mapping preserves ordering and feasibility. This link yields a sharp characterization of optimality for Epstein-Zin preferences that include time preference shocks. The same approach shows that multiplicative Kreps-Porteus preferences are isomorphic to risk-sensitive preferences, so established results for the latter extend to the former. The transformations also raise the precision of numerical value function approximations, delivering two orders of magnitude better accuracy in a multisector real business cycle model.

Core claim

When two dynamic programs are connected by an order isomorphism, optimality properties transmit from one formulation to the other. The isomorphism maps states, actions, rewards, and constraints while preserving the partial order and feasibility structure, so that an optimal policy or value function in one program corresponds exactly to an optimal policy or value function in the other.

What carries the argument

Order isomorphism between dynamic programs that preserves ordering and feasibility structure.

If this is right

  • Epstein-Zin preferences with time preference shocks admit a sharp characterization of when optimality holds.
  • Well-known optimality results for risk-sensitive preferences carry over directly to multiplicative Kreps-Porteus preferences.
  • Isomorphic transformations improve the numerical accuracy of value function approximations by two orders of magnitude in multisector real business cycle models.
  • Conjugacy techniques can be applied to relate and solve families of otherwise distinct dynamic programming problems.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Researchers could systematically search for order isomorphisms to convert intractable dynamic programs into equivalent but computationally simpler forms.
  • The same transmission principle might apply to continuous-time or continuous-state formulations, allowing discrete approximations to inherit exact optimality properties.
  • Numerical gains observed in the RBC example suggest testing isomorphic reformulations in other stochastic growth or asset-pricing models to reduce approximation error.
  • Policy functions derived under one preference specification could be mapped to another via the isomorphism to obtain closed-form or semi-closed-form solutions.

Load-bearing premise

The two dynamic programs must be related by a well-defined order isomorphism that preserves the relevant ordering and feasibility structure.

What would settle it

Constructing or simulating two dynamic programs that satisfy the order-isomorphism conditions yet exhibit an optimal policy in one that fails to be optimal in the other would disprove the transmission result.

Figures

Figures reproduced from arXiv: 2605.22076 by John Stachurski, Junnan Zhang.

Figure 1
Figure 1. Figure 1: An order stable map 𝑆 on [0, 1] Example 2.1. Let (𝑉, ⪯, 𝜌) be a partially ordered space (i.e., a metric space (𝑉, 𝜌) where ⪯ is closed under limits). If (𝑉, 𝑆) is globally stable and 𝑆 is order-preserving, then (𝑉, 𝑆) is order stable. To see this, let 𝑣¯ be the unique fixed point of 𝑆 in 𝑉. Upward stability holds because if 𝑣 ∈ 𝑉 and 𝑣 ⪯ 𝑆 𝑣, then, iterating on this inequality and using the fact that 𝑆 is … view at source ↗
read the original abstract

We study relationships between dynamic programs by applying conjugacy methods from dynamical systems theory. When two dynamic programs are connected by an order isomorphism, we show that optimality properties transmit from one formulation to the other. We apply these results to Epstein--Zin preferences with time preference shocks, obtaining a sharp characterization of when optimality holds. We also show that multiplicative Kreps--Porteus preferences and risk-sensitive preferences are isomorphic, so that well-known results for the latter carry over to the former. Finally, we demonstrate how isomorphic transformations can improve the numerical accuracy of value function approximations, with gains of two orders of magnitude in a multisector real business cycle model.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The paper applies conjugacy methods from dynamical systems theory to relate dynamic programs in economics. It claims that when two dynamic programs are connected by an order isomorphism, optimality properties transmit from one to the other. This framework is used to obtain a sharp characterization of optimality for Epstein-Zin preferences with time-preference shocks, to establish that multiplicative Kreps-Porteus preferences are isomorphic to risk-sensitive preferences (allowing results to carry over), and to demonstrate that isomorphic transformations can improve numerical accuracy of value function approximations by two orders of magnitude in a multisector real business cycle model.

Significance. If the transmission result is established rigorously, including under stochastic expectations, the work could unify analyses across recursive preference specifications and provide practical tools for computation in dynamic models. The Epstein-Zin application and the reported numerical gains in the RBC example represent concrete strengths that would strengthen the contribution if the underlying conjugacy arguments are fully verified.

major comments (2)
  1. The central transmission claim for stochastic dynamic programs (applied to Epstein-Zin with time-preference shocks) requires explicit conditions ensuring the order isomorphism commutes with the conditional expectation operator and preserves measurability with respect to the filtration. The conjugacy construction from dynamical systems is typically stated for deterministic settings; without verifying that suprema over plans remain interchangeable with integration, the sharp optimality characterization may hold only pathwise rather than in expectation.
  2. In the section establishing the isomorphism between multiplicative Kreps-Porteus and risk-sensitive preferences, the paper should specify how the order isomorphism maps feasible plans while preserving the relevant ordering and feasibility structure, as this is the weakest assumption needed for optimality transmission to hold.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and insightful comments on our manuscript. These have prompted us to strengthen the presentation of the stochastic transmission results and the explicit construction of the isomorphism in the Kreps-Porteus application. We address each major comment below and indicate the corresponding revisions.

read point-by-point responses

  1. Referee: The central transmission claim for stochastic dynamic programs (applied to Epstein-Zin with time-preference shocks) requires explicit conditions ensuring the order isomorphism commutes with the conditional expectation operator and preserves measurability with respect to the filtration. The conjugacy construction from dynamical systems is typically stated for deterministic settings; without verifying that suprema over plans remain interchangeable with integration, the sharp optimality characterization may hold only pathwise rather than in expectation.

    Authors: We agree that the stochastic extension merits explicit verification. The manuscript already assumes that the order isomorphism is adapted to the filtration and that the relevant suprema and conditional expectations can be interchanged under standard integrability conditions maintained throughout the Epstein-Zin application. To make this fully rigorous, we will add a new lemma (and accompanying discussion) that states the precise measurability and commutation conditions required for the isomorphism to preserve optimality in expectation rather than merely pathwise. The proof will explicitly interchange the supremum over plans with integration by appealing to the monotone convergence theorem under the maintained boundedness assumptions on the aggregator. revision: yes

  2. Referee: In the section establishing the isomorphism between multiplicative Kreps-Porteus and risk-sensitive preferences, the paper should specify how the order isomorphism maps feasible plans while preserving the relevant ordering and feasibility structure, as this is the weakest assumption needed for optimality transmission to hold.

    Authors: We appreciate this suggestion for added clarity. The order isomorphism is defined directly on the product space of consumption and continuation-value sequences and maps each feasible plan in the multiplicative Kreps-Porteus program to a corresponding feasible plan in the risk-sensitive program while preserving the partial order induced by the recursive aggregator. Feasibility is preserved by construction because the budget constraints are identical up to the monotone transformation that defines the isomorphism. In the revision we will insert a short paragraph that explicitly describes this mapping on the set of feasible plans and verifies that both the ordering and the constraint sets are preserved, thereby confirming that the weakest sufficient condition for optimality transmission is satisfied. revision: yes

Circularity Check

0 steps flagged

No circularity: transmission results rest on external conjugacy and order-isomorphism assumptions

full rationale

The derivation applies conjugacy methods from dynamical systems theory to show that order isomorphisms transmit optimality between dynamic programs. The Epstein-Zin application with time-preference shocks and the isomorphism between multiplicative Kreps-Porteus and risk-sensitive preferences are obtained by verifying that the isomorphism preserves feasibility, ordering, and the relevant operators, without defining the target optimality properties in terms of the isomorphism itself or fitting parameters to the outcomes. No load-bearing step reduces by construction to a self-citation or to a renamed input; the arguments remain self-contained once the structural preservation conditions are granted.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the existence of an order isomorphism between the programs under study; no free parameters or new entities are introduced in the abstract.

axioms (1)
  • domain assumption Dynamic programs under consideration admit an order isomorphism that preserves feasibility and ordering of states and actions.
    This structural premise is required for optimality to transmit between formulations.

pith-pipeline@v0.9.0 · 5624 in / 1136 out tokens · 39610 ms · 2026-05-22T02:45:47.268479+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

300 extracted references · 300 canonical work pages · 2 internal anchors

  1. [1]

    Error Propagation for Approximate Policy and Value Iteration , booktitle =

    Farahmand, Amir-Massoud and Munos, R. Error Propagation for Approximate Policy and Value Iteration , booktitle =. 2010 , url =

    work page 2010

  2. [2]

    Sargent and John Stachurski , year=

    Thomas J. Sargent and John Stachurski , year=. Dynamic Programming on Partially Ordered Sets , journal=

    work page

  3. [3]

    Optimal Control on Positive Cones , author=

    work page

  4. [4]

    Advances in Neural Information Processing Systems , volume=

    Distributionally robust linear quadratic control , author=. Advances in Neural Information Processing Systems , volume=

    work page

  5. [5]

    Mathematics of Operations Research , volume=

    Empirical dynamic programming , author=. Mathematics of Operations Research , volume=. 2016 , publisher=

    work page 2016

  6. [6]

    Stochastic Systems , volume=

    Empirical Q-value iteration , author=. Stochastic Systems , volume=. 2021 , publisher=

    work page 2021

  7. [7]

    Computational Economics , volume=

    Continuous state dynamic programming via nonexpansive approximation , author=. Computational Economics , volume=. 2008 , publisher=

    work page 2008

  8. [8]

    Journal of Economic Dynamics and Control , volume=

    Fitted value function iteration with probability one contractions , author=. Journal of Economic Dynamics and Control , volume=. 2013 , publisher=

    work page 2013

  9. [9]

    Risk and Decision Analysis , volume=

    Estimate and approximate policy iteration algorithm for discounted Markov decision models with bounded costs and Borel spaces , author=. Risk and Decision Analysis , volume=. 2017 , publisher=

    work page 2017

  10. [10]

    2007 , publisher=

    Approximate Dynamic Programming: Solving the curses of dimensionality , author=. 2007 , publisher=

    work page 2007

  11. [11]

    Annals of Operations Research , volume=

    Perspectives of approximate dynamic programming , author=. Annals of Operations Research , volume=. 2016 , publisher=

    work page 2016

  12. [12]

    and Chen, Z

    Discounted reinforcement learning is not an optimization problem , author=. arXiv preprint arXiv:1910.02140 , year=

    work page arXiv 1910

  13. [13]

    , author=

    Finite-Time Bounds for Fitted Value Iteration. , author=. Journal of Machine Learning Research , volume=

    work page

  14. [14]

    Annals of Operations Research , volume=

    Markov perfect equilibria in a dynamic decision model with quasi-hyperbolic discounting , author=. Annals of Operations Research , volume=. 2020 , publisher=

    work page 2020

  15. [15]

    Mathematics of Operations Research , volume=

    Robust dynamic programming , author=. Mathematics of Operations Research , volume=. 2005 , publisher=

    work page 2005

  16. [16]

    IEEE Transactions on Automatic Control , year=

    Exact dynamic programming for positive systems with linear optimal cost , author=. IEEE Transactions on Automatic Control , year=

    work page

  17. [17]

    Yu, Zhuodong and Dai, Ling and Xu, Shaohang and Gao, Siyang and Ho, Chin Pang , journal=. Fast

    work page

  18. [18]

    Variance reduced value iteration and faster algorithms for solving

    Sidford, Aaron and Wang, Mengdi and Wu, Xian and Ye, Yinyu , journal=. Variance reduced value iteration and faster algorithms for solving. 2023 , publisher=

    work page 2023

  19. [19]

    2023 , institution=

    Not a Typical Firm: Capital-Labor Substitution and Firms' Labor Shares , author=. 2023 , institution=

    work page 2023

  20. [20]

    Machine learning , volume=

    Q-learning , author=. Machine learning , volume=. 1992 , publisher=

    work page 1992

  21. [21]

    International conference on machine learning , pages=

    A distributional perspective on reinforcement learning , author=. International conference on machine learning , pages=. 2017 , organization=

    work page 2017

  22. [22]

    2023 , publisher=

    Distributional reinforcement learning , author=. 2023 , publisher=

    work page 2023

  23. [23]

    On solutions of the distributional

    Julian Gerstenberg and Ralph Neininger and Denis Spiegel , year=. On solutions of the distributional. 2202.00081 , archivePrefix=

    work page arXiv

  24. [24]

    2021 , eprint=

    Online Abstract Dynamic Programming with Contractive Models , author=. 2021 , eprint=

    work page 2021

  25. [25]

    SIAM Journal on Optimization , volume=

    Regular policies in abstract dynamic programming , author=. SIAM Journal on Optimization , volume=. 2017 , publisher=

    work page 2017

  26. [26]

    SIAM Journal on Control and Optimization , volume=

    Monotone mappings with application in dynamic programming , author=. SIAM Journal on Control and Optimization , volume=. 1977 , publisher=

    work page 1977

  27. [27]

    2012 , publisher=

    Geometrical methods in the theory of ordinary differential equations , author=. 2012 , publisher=

    work page 2012

  28. [28]

    Bulletin of the American Mathematical Society , volume=

    Differentiable dynamical systems , author=. Bulletin of the American Mathematical Society , volume=

    work page

  29. [29]

    2002 , publisher=

    Introduction to dynamical systems , author=. 2002 , publisher=

    work page 2002

  30. [30]

    2023 , eprint=

    Completely Abstract Dynamic Programming , author=. 2023 , eprint=

    work page 2023

  31. [31]

    Journal of Economic Theory , volume=

    Stochastic optimal growth model with risk sensitive preferences , author=. Journal of Economic Theory , volume=. 2018 , publisher=

    work page 2018

  32. [32]

    Lu, Jay and Luo, Yao and Saito, Kota and Xin, Yi , year=2023, institution=. Does

    work page 2023

  33. [33]

    Journal of Economic Theory , volume=

    Stability of equilibrium asset pricing models: A necessary and sufficient condition , author=. Journal of Economic Theory , volume=. 2021 , publisher=

    work page 2021

  34. [34]

    Thompson aggregators,

    Becker, Robert A and Rinc. Thompson aggregators,. Mathematical Social Sciences , volume=. 2021 , publisher=

    work page 2021

  35. [35]

    Annual Review of Economics , volume=

    Firm dynamics and trade , author=. Annual Review of Economics , volume=. 2021 , publisher=

    work page 2021

  36. [36]

    American Economic Review , volume=

    The nature of firm growth , author=. American Economic Review , volume=. 2021 , publisher=

    work page 2021

  37. [37]

    1990 , publisher=

    Topics in metric fixed point theory , author=. 1990 , publisher=

    work page 1990

  38. [38]

    2002 , publisher=

    An invitation to operator theory , author=. 2002 , publisher=

    work page 2002

  39. [39]

    Recursive utility in a

    Hansen, Lars Peter and Scheinkman, Jos. Recursive utility in a. Proceedings of the National Academy of Sciences , volume=. 2012 , publisher=

    work page 2012

  40. [40]

    2019 , eprint=

    Hyperbolic Discounting and Learning over Multiple Horizons , author=. 2019 , eprint=

    work page 2019

  41. [41]

    EURO Journal on Transportation and Logistics , volume=

    Approximate dynamic programming in transportation and logistics: a unified framework , author=. EURO Journal on Transportation and Logistics , volume=. 2012 , publisher=

    work page 2012

  42. [42]

    Continuous-time

    Guo, Xianping and Hern. Continuous-time. 2009 , publisher=

    work page 2009

  43. [43]

    Krasnosel’skii, M. A. and Vainikko, G. M. and Zabreiko, P. P. and Rutitskii, Ya. B. and Stetsenko, V. Ya. , title =. doi:10.1007/978-94-010-2715-1 , url =

    work page doi:10.1007/978-94-010-2715-1

  44. [44]

    Risk-sensitive

    Howard, Ronald A and Matheson, James E , journal=. Risk-sensitive. 1972 , publisher=

    work page 1972

  45. [45]

    2015 , publisher=

    An Introduction to Dynamical Systems and Chaos , author=. 2015 , publisher=

    work page 2015

  46. [46]

    2014 , publisher=

    Dynamical Systems , author=. 2014 , publisher=

    work page 2014

  47. [47]

    Exponential

    Fei, Yingjie and Yang, Zhuoran and Chen, Yudong and Wang, Zhaoran , journal=. Exponential

    work page

  48. [48]

    American Economic Review , volume=

    Risk preferences are not time preferences , author=. American Economic Review , volume=. 2012 , publisher=

    work page 2012

  49. [49]

    Recursive Utility for Thompson Aggregators: Least Fixed Point, Uniqueness, and Approximation Theories , author=

    work page

  50. [50]

    Handbook of Behavioral Economics: Applications and Foundations 1 , volume=

    Intertemporal choice , author=. Handbook of Behavioral Economics: Applications and Foundations 1 , volume=. 2019 , publisher=

    work page 2019

  51. [51]

    Recursive utility and optimal capital accumulation. I. Existence , author=. Journal of Economic Theory , volume=. 1989 , publisher=

    work page 1989

  52. [52]

    1944 , publisher=

    Theory of games and economic behavior , author=. 1944 , publisher=

    work page 1944

  53. [53]

    Journal of the American Statistical Association , volume=

    The theory of statistical decisions , author=. Journal of the American Statistical Association , volume=. 1951 , publisher=

    work page 1951

  54. [54]

    Journal of Political Economy , volume=

    The utility analysis of choices involving risk , author=. Journal of Political Economy , volume=. 1948 , publisher=

    work page 1948

  55. [55]

    Economics letters , volume=

    Some empirical evidence on dynamic inconsistency , author=. Economics letters , volume=. 1981 , publisher=

    work page 1981

  56. [56]

    Management science , volume=

    Discount rates inferred from decisions: An experimental study , author=. Management science , volume=. 1989 , publisher=

    work page 1989

  57. [57]

    The Theory of Interest as Determined by Impatience to Spend Income and Opportunity to Invest it , author=. Bull. Amer. Math. Soc , volume=

    work page

  58. [58]

    Economic Theory , pages=

    Choquet expected discounted utility , author=. Economic Theory , pages=. 2022 , publisher=

    work page 2022

  59. [59]

    Journal of Economic Theory , volume=

    Optimal growth with many consumers , author=. Journal of Economic Theory , volume=. 1984 , publisher=

    work page 1984

  60. [60]

    2002 , publisher=

    Introduction to lattices and order , author=. 2002 , publisher=

    work page 2002

  61. [61]

    arXiv preprint arXiv:2304.06830 , year=

    Recursive Preferences and Ambiguity Attitudes , author=. arXiv preprint arXiv:2304.06830 , year=

    work page arXiv

  62. [62]

    Mathematical Control and Related Fields , volume=

    Time-consistent lifetime portfolio selection under smooth ambiguity , author=. Mathematical Control and Related Fields , volume=. 2023 , publisher=

    work page 2023

  63. [63]

    Scandinavian Actuarial Journal , volume=

    Time-consistent reinsurance and investment strategies for an AAI under smooth ambiguity utility , author=. Scandinavian Actuarial Journal , volume=. 2020 , publisher=

    work page 2020

  64. [64]

    Scandinavian Actuarial Journal , volume=

    Life insurance decisions under recursive utility , author=. Scandinavian Actuarial Journal , volume=. 2019 , publisher=

    work page 2019

  65. [65]

    American Journal of Agricultural Economics , volume=

    The value of biodiversity as an insurance device , author=. American Journal of Agricultural Economics , volume=. 2019 , publisher=

    work page 2019

  66. [66]

    Journal of Risk and Insurance , volume=

    Risk aversion and the value of risk to life , author=. Journal of Risk and Insurance , volume=. 2012 , publisher=

    work page 2012

  67. [67]

    , author=

    Deep learning for solving dynamic economic models. , author=. Journal of Monetary Economics , volume=. 2021 , publisher=

    work page 2021

  68. [68]

    International Economic Review , volume=

    Deep equilibrium nets , author=. International Economic Review , volume=. 2022 , publisher=

    work page 2022

  69. [69]

    arXiv preprint arXiv:2112.14377 , year=

    Deepham: A global solution method for heterogeneous agent models with aggregate shocks , author=. arXiv preprint arXiv:2112.14377 , year=

    work page arXiv

  70. [70]

    2022 , institution=

    Estimating Nonlinear Heterogeneous Agents Models with Neural Networks , author=. 2022 , institution=

    work page 2022

  71. [71]

    arXiv preprint arXiv:2103.16977 , year=

    Solving heterogeneous general equilibrium economic models with deep reinforcement learning , author=. arXiv preprint arXiv:2103.16977 , year=

    work page arXiv

  72. [72]

    1996 , publisher=

    Neuro-dynamic programming , author=. 1996 , publisher=

    work page 1996

  73. [73]

    Recursive preferences, correlation aversion, and the remporal resolution of uncertainty , author=

    work page

  74. [74]

    Econometrica , volume=

    Ambiguity, learning, and asset returns , author=. Econometrica , volume=. 2012 , publisher=

    work page 2012

  75. [75]

    2003 , publisher=

    Handbook of means and their inequalities , author=. 2003 , publisher=

    work page 2003

  76. [76]

    2011 , publisher=

    A weak convergence approach to the theory of large deviations , author=. 2011 , publisher=

    work page 2011

  77. [77]

    1953 , publisher=

    Essays in positive economics , author=. 1953 , publisher=

    work page 1953

  78. [78]

    The Review of Financial Studies , month =

    Gomez-Cram, Roberto and Yaron, Amir , doi =. The Review of Financial Studies , month =

    work page

  79. [79]

    Valuation risk revalued , volume =

    de Groot, Oliver and Richter, Alexander W and Throckmorton, Nathaniel A , journal =. Valuation risk revalued , volume =

    work page

  80. [80]

    The quarterly journal of economics , volume=

    Risk, ambiguity, and the Savage axioms , author=. The quarterly journal of economics , volume=. 1961 , publisher=

    work page 1961

Showing first 80 references.