Dynamic portfolio selection under generalized disappointment aversion
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Quantitative Finance. The day in quantitative finance: models of markets, risk and prices.
Ruth: I'm Ruth.
Warren: And I'm Warren, here with Iris.
Ruth: We're looking at the paper: "Dynamic portfolio selection under generalized disappointment aversion".
Warren: Here's the gist of the paper. Dynamic portfolio selection under generalized disappointment aversion addresses the continuous-time portfolio selection problem under generalized disappointment aversion (GDA) preferences by providing a sufficient and necessary condition for an equilibrium in a fully nonlinear integral equation The findings indicate that under disappointment aversion preferences, non-participation in the stock market is the unique equilibrium The semi-analytical equilibrium strategies obtained under the constant relative risk aversion utility functions reveal that, under GDA preferences, the investment proportion in the stock market consistently remains smaller than the investment proportion under classical expected utility theory The numerical analysis shows that the equilibrium strategy’s monotonicity concerning the two parameters of GDA preference aligns with the monotonicity of the degree of risk aversion.
Ruth: Problem Formulation and Preference Definition The paper begins by formulating a financial market model with one risk-free asset and several risky assets following specific stochastic differential equations It then defines the generalized disappointment aversion (GDA) preference using an implicit definition of the certainty equivalent which introduces time inconsistency into the problem.
Warren: The GDA value ηt(Y) is determined by a specific equation involving the utility function U, adjustment coefficient β, and adjustment coefficient δ.
Ruth: The relationship between the GDA value and the time-t certainty equivalent Ct(Y) is established through a function ϕ defined by an implicit equation involving U and the parameters δ and β.
Paper summary: Ruth: Building on what we just discussed about non-participation being the unique equilibrium under these GDA preferences, this segment focuses on summarizing what this paper actually claims before we get into the deeper implications. We’re talking about how they set up their model and what their main conclusions are regarding the structure of equilibrium strategies >
Warren: They start by formulating a financial market with one risk-free asset and several risky assets following specific stochastic differential equations, and then they define generalized disappointment aversion using an implicit definition of the certainty equivalent that introduces time inconsistency into the problem >
Iris: So, before we get to the solutions, what's their approach to solving this continuous time problem? Are they using standard dynamic programming or something else >
Ruth: They characterize the equilibrium condition as a fully nonlinear integral equation derived from the trade-off between instantaneous risk and return >
Warren: And they conjecture that the equilibrium strategy takes a specific form: it’s the market price of risk multiplied by a scalar determined entirely by the marginal rate of substitution of cumulative risk for cumulative return >
Iris: That sounds complex. What about when they look at simpler cases, like when delta equals one? Does that simplify things significantly >
Ruth: When delta equals one, equilibrium investments remain nonzero whenever the expected return of stocks is non-zero >
Conclusion: Ruth: So, wrapping up this discussion on "Dynamic portfolio selection under generalized disappointment aversion," we've seen that the core thesis is that for these GDA preferences, non-participation in the stock market is the unique equilibrium >
Warren: And their findings under constant relative risk aversion utility functions show that even with this preference structure, if you're comparing it to classical expected utility theory, you end up investing less in stocks >
Iris: I want to talk about what this means for real-world investors. If the model finds non-participation is the unique equilibrium when delta equals one, does that suggest that people with generalized disappointment aversion would just avoid the market entirely in practice >
Ruth: It suggests that under these specific GDA preferences, avoidance of stocks becomes mathematically inevitable as an equilibrium state >
Warren: But it’s important to remember their caveat. They also mention incorporating horizon-dependent risk aversion, or HDRA, to get a more realistic view of investment behaviors as the terminal time approaches >
Iris: And what does that horizon dependence actually suggest about behavior? Do you think people would behave differently when they know the investment period is ending soon >
Ruth: The paper suggests that under HDRA, investments consistently increase and converge toward the Merton solution as time approaches that terminal time >
Department of Mathematical Sciences, Tsinghua University · RCSDS, NCMIS, Academy of Mathematics and Systems Science, Chinese Academy of Sciences
q-fin.MF, q-fin.PM
Submitted: 2024-01-16
Updated: 2026-11-14
Comments: 45 pages, 4 figures
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 86/100
The gist: Dynamic portfolio selection under generalized disappointment aversion addresses the continuous-time portfolio selection problem under generalized disappointment aversion (GDA) preferences by…
Key concepts
- Generalized Disappointment Aversion (GDA)
- This preference model defines how investors react negatively when actual outcomes fail to meet expectations. It introduces a time inconsistency into the standard portfolio selection problem by defining a generalized certainty equivalent based on utility, an adjustment coefficient $\beta$, and an adjustment coefficient $\delta$. This captures the idea that disappointment aversion changes over time.
- Integral Equation Equilibrium
- The equilibrium for dynamic portfolio selection is found by solving a fully nonlinear integral equation. This equation balances the instantaneous trade-off between risk and return. The solution represents the optimal investment strategy, which is determined by the marginal rate of substitution of cumulative risk for cumulative return.
- Classical Expected Utility (EU) vs. GDA
- The paper compares investment strategies under classical expected utility theory (EU) with those under GDA. It finds that under GDA, an investor's investment proportion in the stock market is consistently smaller than what they would choose under EU preferences, suggesting that disappointment aversion makes investors less willing to invest.
- Horizon-Dependent Risk Aversion (HDRA)
- HDRA suggests that an investor's risk aversion changes depending on how much time remains until the investment horizon. When considering decisions at time $t$ for outcomes at time $T$, the agent uses a risk aversion coefficient $\rho(t)$. This concept helps model realistic investment behavior where preferences evolve as the terminal time approaches.
Terminology
Summary
Dynamic portfolio selection under generalized disappointment aversion addresses the continuous-time portfolio selection problem under generalized disappointment aversion (GDA) preferences by providing a sufficient and necessary condition for an equilibrium in a fully nonlinear integral equation The findings indicate that under disappointment aversion preferences, non-participation in the stock market is the unique equilibrium The semi-analytical equilibrium strategies obtained under the constant relative risk aversion utility functions reveal that, under GDA preferences, the investment proportion in the stock market consistently remains smaller than the investment proportion under classical expected utility theory The numerical analysis shows that the equilibrium strategy’s monotonicity concerning the two parameters of GDA preference aligns with the monotonicity of the degree of risk aversion. p. 0
Problem Formulation and Preference Definition
The paper begins by formulating a financial market model with one risk-free asset and several risky assets following specific stochastic differential equations It then defines the generalized disappointment aversion (GDA) preference using an implicit definition of the certainty equivalent which introduces time inconsistency into the problem. The GDA value ηt(Y) is determined by a specific equation involving the utility function U, adjustment coefficient β, and adjustment coefficient δ. p. 2 The relationship between the GDA value and the time-t certainty equivalent Ct(Y) is established through a function ϕ defined by an implicit equation involving U and the parameters δ and β. p. 6
Characterization of Equilibrium Strategies
The equilibrium condition for the portfolio selection problem is characterized as a fully nonlinear integral equation derived from the trade-off between instantaneous risk and return. The equilibrium strategy is conjectured to take the form of a market price of risk multiplied by a scalar entirely determined by the marginal rate of substitution of cumulative risk for cumulative return. p. 3 For the case where δ = 6= 1, equilibrium investments remain nonzero whenever the expected return of stocks is non-zero. p. 3
Monotonicity and Risk Aversion Analysis
The paper investigates the monotonicity of the risk aversion of GDA preferences with respect to the two parameters, β and δ. It is shown that under CRRA utility functions, GDA (δ 6= 1) alters the agent’s risk attitude, rendering the agent more risk averse. p. 3 Specifically, a gradual increase in δ leads to a gradual decrease in equilibrium investment when 0 1. p. 3
Conclusion and Extensions
The paper concludes by summarizing that non-participation in the stock market is the unique equilibrium under DA preferences Furthermore, semi-analytical solutions in the case of CRRA utility demonstrate that the equilibrium investment under GDA preferences consistently remains below the optimal investment under EU preferences. The results also suggest incorporating horizon-dependent risk aversion (HDRA) to provide a more realistic investment behaviors as the terminal time approaches. p. 3 The equilibrium strategy under HDRA exhibits a peculiar phenomenon where investments consistently increase and converge towards the Merton solution as time approaches the terminal time. p. 3 The paper establishes that the unique equilibrium in D is found when U is concave and δ = 1. p. 4
A Some technical lemmas
The paper presents several technical lemmas to support its proofs, including Lemma A.1 which extends Stein (1981) regarding the expectation of a function of a standard normal random variable. p. 5 Lemmas B.5 and B.6 establish the continuity properties of the auxiliary functions h and H under certain regularity conditions on U. p. 7 Lemma 3.1 proves that when δ = 6= 1, g is C1 on [0,∞) × R and gy(0,0)gv(0,0) = −2m(0, 0) = 2 U′(1)U′′(1). p. 2 Lemma B.7 proves that when δ = 1 and U is concave, the limit of g-g(0, 0)√v as v→0 is c∗ < 0 where c+βcN(c)+βN'(c)=0. Lemma B.8 proves that the equilibrium strategy π¯ = 0 is an equilibrium when U is concave and δ = 1. p. 0 The paper also includes a proof of Theorem 3. p. 65 establishing the existence and uniqueness of the solution to the integral equation (3. p. 07). The proof of Theorem 4. p. 6 p. 61 shows that for CRRA utility function U = Uρ, the equilibrium in D exists uniquely. p. 7 The numerical analysis confirms that as time t approaches T, the cumulative risk v(t) diminishes and the MRS closely approximates that of the EU preference. p. 0 The paper also shows that as β gradually increases, signifying a higher aversion to disappointment, the equilibrium investment decreases. p. 3 p. 3 The discussion on HDRA suggests that when making decisions at time t and outcomes occur at time T, the agent adopts ρ(t) as the coefficient of risk aversion. p. 2 The paper notes that an equilibrium strategy is found by solving an ODE derived from the integral equation. p. 4 The final result shows that the equilibrium strategy is determined by a function a satisfying at = m(sZ T t as 2ds, Z T t a⊤ s λ(s)dsλ(t), five or more words copied word for word>. p. 3 The paper also shows that the equilibrium investment under GDA preferences consistently remains below the optimal investment under EU preferences. p. 0 The numerical analysis reveals that in the case δ > 1, a larger δ > 1 yields more investment in the stock. p. 0 p. 0
Where the gains come from
The core contribution lies in redefining the certainty equivalent for GDA preferences with general utility functions and a broader range of parameters when δ > 1. The second major finding is that under GDA preferences (δ 6= 1), the unique equilibrium strategy is always 0, signifying non-participation in the stock market regardless of β > 0. p. 3 The semi-analytical solutions in the CRRA utility framework demonstrate that the equilibrium investment under GDA preferences consistently remains smaller than those of EU preferences. p. 0 The numerical analysis indicates that when 0 < δ < 1, a gradual increase in δ leads to a gradual decrease in equilibrium investment, while the opposite holds true for the case δ > 1. p. 3 The paper also shows that as β gradually increases, signifying a higher aversion to disappointment, the equilibrium investment decreases.
The technical lemmas provide the rigorous foundation for the main results, such as Lemma B.1 which shows that h and H are continuous if U is 0-th-order regular. p. 7 Lemma B.2 provides explicit expressions for Hx and Hy under certain conditions on U. p. 6 Lemma 3.7 shows that m(x, y) can be expressed in terms of H. p. 3 These lemmas are crucial for proving the existence and uniqueness of the solution to the integral equation (3. p. 07). The proof of Theorem 4.1 relies on these results to show that a is given by an ODE with separated variables. The technical machinery ensures the rigorous derivation of the equilibrium strategy. p. 0 The paper also shows that the equilibrium investment under GDA preferences consistently remains below those of EU preferences. p. 0 The numerical analysis reveals that in the case δ > 1, a larger δ > 1 yields more investment in the stock.
Discussion
The discussion section explores conventional investment wisdom and proposes incorporating horizon-dependent risk aversion (HDRA) to address the time-dependent nature of optimal strategies. The paper highlights a peculiar phenomenon where as time approaches the terminal time, equilibrium investments consistently increase and converge towards the Merton solution. p. 3 It is also noted that decisions made by the naive agent and the sophisticated agent are completely unrelated in general time inconsistency problems. p. 2 The paper suggests that an equilibrium strategy can be found by solving an ODE derived from the integral equation. p. 4 The overall finding is that GDA preferences exhibit more risk aversion than EU preference, as shown by the discussion on the MRS of g in Subsection 3. p. 61. The paper also shows that the equilibrium strategy is determined by a function a satisfying at = m(sZ T t as 2ds, Z T t a⊤ s λ(s)dsλ(t), five or more words copied word for word>. p. 3
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Develop AI systems capable of solving continuous-time portfolio selection problems under generalized disappointment aversion by formulating and solving fully nonlinear integral equations to find sufficient and necessary equilibrium conditions.
-
Create AI models that can predict investment strategies for agents with generalized disappointment aversion, specifically demonstrating that
non-participation in the stock market is the unique equilibrium
under disappointment aversion preferences. -
Enhance risk assessment capabilities in financial AI by using the GDA preference framework to show how investment proportions consistently remain smaller than those under classical expected utility theory when considering risk aversion parameters beta and delta.
-
Implement AI systems that can distinguish between different types of risk aversion by analyzing the monotonicity of the degree of risk aversion concerning the two parameters of GDA preferences, beta and delta, against the monotonicity observed in equilibrium strategies.
-
Design AI agents that can adapt their investment behavior based on horizon-dependent risk aversion (HDRA) by incorporating a time-varying coefficient for relative risk aversion rho(t) into the preference functional to model realistic long-term investing.
Sources
- Being serious about non-commitment: subgame perfect equilibrium in continuous time
- Dynamic portfolio selection for nonlinear law-dependent preferences
- Equilibrium stochastic control with implicitly defined objective functions
Related papers
- Valuation Reveals Uncertainty
- GROMM: Generative Risk-neutral Option Market Models
- Partially Active Automated Market Makers
- Sublinear Markov Semigroups Beyond Uniqueness
- Price-impact Martingale optimal transport
- Low-rank and graphon limits for dynamic threshold distress contagion in heterogeneous financial networks