Stochastic dynamic utilities and intertemporal preferences

Abstract

We propose an axiomatic approach which economically underpins the representation of dynamic intertemporal decisions in terms of a utility function, which randomly reacts to the information available to the decision maker throughout time. Our construction is iterative and based on time dependent preference connections, whose characterization is inspired by the original intuition given by Debreu’s State Dependent Utilities (1960).

This is a preview of subscription content, access via your institution.

Notes

  1. 1.

    [31], p. 139.

  2. 2.

    We point out that the use of the term ‘ preference’ is slightly improper as the ordering will not be a binary relation as it is usually intended.

  3. 3.

    Indeed this choice is not ‘without loss of generality’. Nevertheless, as explained in the Introduction this research is inspired by potential financial applications, and therefore we prefer to choose a more financial friendly setup.

  4. 4.

    This additional requirement is in fact without loss of generality, and allows a useful simplification in the main body of the proof.

  5. 5.

    Recall the definition in Eq. (2.1)

  6. 6.

    Abuse of notation: the precise formulation should be \(C_{s,v}(f)= C_{s,t}(g)\) where \(g\in \mathscr {L}({\mathcal F}_t)\) is a version of \(C_{t,v}(f)\).

  7. 7.

    This property in general holds only for few classes of Risk Measures.

  8. 8.

    This function is well defined and measurable as \(u_i(\cdot ,\omega )\) is strictly increasing for any \(\omega \in \varOmega \).

  9. 9.

    A set \({\mathcal A}\) is downward directed if for any \(f,g\in {\mathcal A}\) the minimum \(f\wedge g\in {\mathcal A}\). The existence of a minimizing sequence is proved in Appendix A.5 of [12]

  10. 10.

    To show the existence of such g we need to consider for any \(\varepsilon >0\), \(C_{i,i+1}(f)-\varepsilon \) so that \(u_i(C_{i,i+1}(f)-\varepsilon )< u_i(C_{i,i+1}(f))=V_{i+1}(f)\); observing that \(u_i(C_{i,i+1}(f)-\varepsilon )\) increases monotonically to \(u_i(C_{i,i+1}(f))\) (for any \(\omega \in \varOmega \)) we can find an \(\bar{\varepsilon }\) such that \(u_0(a)< E_{{\mathbb P}_i}[u_i(C_{i,i+1}(f)-\bar{\varepsilon })]< E_{{\mathbb P}_i}[u_i(C_{i,i+1}(f))]=E_{{\mathbb P}_i}[V_{i+1}(f)]\).

References

  1. 1.

    Anscombe, F.J., Aumann, R.J.: A definition of subjective probability. Ann. Math. Stat. 34(1), 199–205 (1963)

    MathSciNet  Article  Google Scholar 

  2. 2.

    Bernoulli, D.: Exposition of a new theory on the measurement of risk, trans. L. Sommer, Econometrica, 22, 23–36. (1954). Translated from an article originally published in 1738

  3. 3.

    Castagnoli, E., LiCalzi, M.: Benchmarking real-valued acts. Games Econ. Behav. 57, 236–253 (2006)

    MathSciNet  Article  Google Scholar 

  4. 4.

    Debreu, G.: Topological Methods in Cardinal Utility Theory. Mathematical Methods in Social Sciences, pp. 19–26. Stanford University Press, Stanford (1960)

    Google Scholar 

  5. 5.

    de Finetti, B.: Sul significato soggettivo della probabilità. Fundamenta Mathematicae 298–329 (1931)

  6. 6.

    Drapeau, S., Jamneshan, A.: Conditional preference orders and their numerical representations. J. Math. Econ. 63, 106–118 (2016)

    MathSciNet  Article  Google Scholar 

  7. 7.

    Duffie, D., Epstein, L.: Stochastic differential utility. Econometrica 60(2), 353–394 (1992)

    MathSciNet  Article  Google Scholar 

  8. 8.

    Epstein, L., LeBreton, M.: Dynamically consistent beliefs must be Bayesian. J. Econ. Theory 61(1), 1–22 (1993)

    MathSciNet  Article  Google Scholar 

  9. 9.

    Epstein, L., Schneider, M.: Recursive multiple-priors. J. Econ. Theory 113(1), 1–31 (2003)

    MathSciNet  Article  Google Scholar 

  10. 10.

    Epstein, L., Wang, T.: Interteporal asset pricing under Knightian uncertainty. Econometrica 62(2), 283–322 (1994)

    Article  Google Scholar 

  11. 11.

    Epstein, L., Zin, S.: Substitution, risk aversion, and the temporal behavior of consumption and asset returns: a theoretical framework. Econometrica 54(4), 937–969 (1989)

    MathSciNet  Article  Google Scholar 

  12. 12.

    F\(\ddot{o}\)llmer, H., Shied, A.: Stochastic Finance. An introduction in discrete time, 3rd ed., de Gruyter Studies in Mathematics (2011)

  13. 13.

    Frittelli, M., Maggis, M.: Conditional certainty equivalent. Int. J. Theor. Appl. Fin. 14(1), 41–59 (2011)

    MathSciNet  Article  Google Scholar 

  14. 14.

    Frittelli, M., Maggis, M.: Dual representation of quasi-convex conditional maps. SIAM J. Fin. Math. 2(1), 357–382 (2011)

    MathSciNet  Article  Google Scholar 

  15. 15.

    Karni, E.: Risk aversion for state-dependent utility functions: measurement and applications. Int. Econ. Rev. 24(3), 637–647 (1983)

    Article  Google Scholar 

  16. 16.

    Kreps, D.M., Porteus, E.L.: Temporal resolution of uncertainty and dynamic choice theory. Econometrica 46(1), 185–200 (1978)

    MathSciNet  Article  Google Scholar 

  17. 17.

    Maccheroni, F., Marinacci, M., Rustichini, A.: Dynamic variational preferences. J. Econ. Theory 128(1), 4–44 (2006)

    MathSciNet  Article  Google Scholar 

  18. 18.

    Marinacci, M.: Model uncertainty. J. Eur. Econ. Assoc. 13, 998–1076 (2015)

    Article  Google Scholar 

  19. 19.

    Merton, R.: Optimum consumption and portfolio rules in a continuous-time model. J. Econ. Theory 3(4), 373–413 (1971)

    MathSciNet  Article  Google Scholar 

  20. 20.

    Musiela, M., Zariphopoulou, T.: Investment and valuation under backward and forward dynamic exponential utilities in a stochastic factor model, pp. 303–334. Advances in Mathematical Finance, Applied and Numerical Harmonic Series (2006)

  21. 21.

    Musiela, M., Zariphopoulou, T.: Portfolio choice under space-time monotone performance criteria. SIAM J. Fin. Math. 326–365 (2009)

  22. 22.

    Pearl, J.: The sure-thing principle. J. Casual Infer. 4(1), 81–86 (2016)

    Article  Google Scholar 

  23. 23.

    Pham, H.: Continuous-time Stochastic Control and Optimization with Financial Applications, Springer Stochastic Modelling and Applied Probability, Vol. 61 (2009)

  24. 24.

    Pratt, A.: Risk aversion in the small and in the large. Econometrica 32(1/2), 122–136 (1964)

    Article  Google Scholar 

  25. 25.

    Riedel, F., Tallon, J.M., Vergopoulos, V.: Dynamically consistent preferences under imprecise probabilistic information. J. Math. Econ. 79, 117–124 (2018)

    MathSciNet  Article  Google Scholar 

  26. 26.

    Rockafellar, R.T.: Integrals which are convex functionals. Pac. J. Math. 24(3), 525–539 (1968)

    MathSciNet  Article  Google Scholar 

  27. 27.

    Skiadas, C.: Conditioning and aggregation of preferences. Econometrica 65(2), 347–367 (1997)

    MathSciNet  Article  Google Scholar 

  28. 28.

    von Neumann, J., Morgenstern, O.: Theory of Games and Economic Behavior, 2nd edn. Princeton University Press, Princeton (1947)

    Google Scholar 

  29. 29.

    Wakker, P.P., Zank, H.: State dependent expected utility for Savage’s state space. Math. Oper. Res. 24(1), 8–34 (1999)

    MathSciNet  Article  Google Scholar 

  30. 30.

    Wang, T.: Conditional preferences and updating. J. Econ. Theory 108, 286–321 (2003)

    MathSciNet  Article  Google Scholar 

  31. 31.

    Zauberman, G., Urminsky, O.: Consumer intertemporal preferences. Curr. Opin. Psychol. 10, 136–141 (2016)

    Article  Google Scholar 

Download references

Author information

Affiliations

Authors

Corresponding author

Correspondence to Marco Maggis.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

The first author would like to thank Marco Frittelli and Fabio Maccheroni for inspiring discussions on this subject.

Appendices

On \(\star \)-continuity

Throughout this section we fix a probability space \((\varOmega ,{\mathcal G},{\mathbb P})\) and a random field \(\phi :{\mathbb R}\times \varOmega \rightarrow {\mathbb R}\) such that for each \(f\in \mathscr {L}^\infty ({\mathcal G})\) the map \(\omega \mapsto \phi (f(\omega ),\omega )\) is \({\mathcal G}\)-measurable and for any \(\omega \), \(x\mapsto \phi (x,\omega )\) is non decreasing. For any \(f\in \mathscr {L}^\infty ({\mathcal G})\) we set

$$ \phi (f(\omega )^+,\omega )=\mathop {\hbox {inf}}\limits _{n\in \mathbb {N}}\phi (f(\omega )+1/n,\omega ) \text { and } \phi (f(\omega )^-,\omega )=\sup _{n\in \mathbb {N}}\phi (f(\omega )-1/n,\omega )$$

and define the following sets:

$$\begin{aligned} RD_f&= \{\omega \in \varOmega : \left( \phi (f(\omega )^+,\omega )-\phi (f(\omega ),\omega )\right)>0\} \\ LD_f&= \{\omega \in \varOmega : \left( \phi (f(\omega ),\omega )-\phi (f(\omega )^-,\omega )\right)>0\} \\ D_f&= \{\omega \in \varOmega : \left( \phi (f(\omega )^+,\omega )-\phi (f(\omega )^-,\omega )\right) >0\} \end{aligned}$$

We now prove a useful lemma which allows to give a well-posed definition of continuity for random fields.

Lemma 4

For each \(f\in \mathscr {L}^\infty ({\mathcal G})\) the sets \(RD_f\), \(LD_f\), \(D_f\), defined above, are \({\mathcal G}\)-measurable.

Proof

Observe that the set \(RD_f\) can be written as:

$$\begin{aligned} RD_f&= \bigcap _{n\in \mathbb {N}}\bigcup _{m\in \mathbb {N}}\left\{ \omega \in \varOmega : \left( \phi \left( f(\omega )+\frac{1}{n},\omega \right) -\phi (f(\omega ),\omega ) \right) > \frac{1}{m}\right\} \\&= \bigcap _{n\in \mathbb {N}}\bigcup _{m\in \mathbb {N}} \left[ \phi \left( f(\cdot )+\frac{1}{n},\cdot \right) -\phi \left( f(\cdot ),\cdot \right) \right] ^{-1} \left( \frac{1}{m},+\infty \right) \end{aligned}$$

which is \({\mathcal G}\)-measurable by measurability of the function

$$\begin{aligned} \omega \rightarrow \phi _n(\omega ) = \phi \left( f(\omega )+\frac{1}{n},\omega \right) -\phi (f(\omega ),\omega ). \end{aligned}$$

Clearly a similar argument shows that \(LD_f\in {\mathcal G}\). Finally, \(D_f = LD_f \cup RD_f \in {\mathcal G}\). \(\square \)

Definition 3

The random fields \(\phi \) is \(\star \)-continuous if \({\mathbb P}(D_f)=0\) for every \(f\in \mathscr {L}^\infty ({\mathcal G})\).

Remark 11

Observe that the set \(D_f\) defined in Lemma 4 can be interpreted as:

$$\begin{aligned} D_f = \left\{ \omega \in \varOmega : f(\omega ) \text { is a point of discontinuity of the function } \phi (\cdot ,\omega )\right\} \end{aligned}$$

In particular for any sequence \(\{f_n\}_{n\in \mathbb {N}}\subset \mathscr {L}^\infty ({\mathcal G})\) such that \(f_{n}(\omega )\rightarrow f(\omega )\) we have \(\phi (f_n(\omega ),\omega )\rightarrow \phi (f(\omega ),\omega )\) for any \(\omega \in (D_f)^c\). Moreover it follows that the definition of \(\star \)-continuity is well posed as the set \(D_f\) is measurable by Lemma 4.

Notice also that taking \(f \equiv x \in {\mathbb R}\) then \(D_x=\{\omega \in \varOmega : \phi (\cdot ,\omega ) \text { is discontinuous in } x\}\). Therefore, the condition \({\mathbb P}(D_x) = 0\) means that for \({\mathbb P}\)-a.e. \(\omega \in \varOmega \) the map \(\phi (\cdot ,\omega )\) is continuous in x . On the other hand, if \(\phi \) is \({\mathbb P}-a.s.\) continuous and satisfies the measurability condition of Lemma 4 then it is also \(\star \)-continuous. Hence, the \(\star \)-continuity is a notion of continuity which is deeply related to the probability space (in particular, to the \(\sigma \)-algebra) and is weaker than the \({\mathbb P}\)-a.s. continuity of the trajectories but stronger than the \({\mathbb P}\)-a.s. continuity at fixed points.

State dependent utilities

As in the rest of the paper \((\varOmega ,{\mathcal F})\) denotes a measurable space and \(\mathscr {L}^{\infty }(\mathcal {F})\) is the space of all acts, represented by real valued \({\mathcal F}\)-measurable and bounded random variables. We here use the term “act”  in order to match the terminology adopted in [29] on which this Appendix is based. This term must be used with care in order to avoid confusion with the general notion of Anscombe-Aumann acts. Indeed in [1] acts are functions from the state space \((\varOmega ,{\mathcal F})\) to a convex set of lotteries over a consequence set.

In this appendix the preference relation is a binary relation \(\succeq \) on \({\mathscr {L}^{\infty }(\mathcal {F})}\) : for f, g \(\in {\mathscr {L}^{\infty }(\mathcal {F})}\), if f is preferred to g, write \(f \succeq g\). The preference relation satisfies the following axiom:

(A1):

Preference order: if it is reflexive (\(\forall f \in {\mathscr {L}^{\infty }(\mathcal {F})}\), \(f \sim f\)), complete (\(\forall f,g \in {\mathscr {L}^{\infty }(\mathcal {F})}\), \(f \succeq g\) or \(f \preceq g\)) and transitive (\(\forall f,g,h \in {\mathscr {L}^{\infty }(\mathcal {F})}\) such that \(f \succeq g\) and \(g \succeq h\) then \(f \succeq h\))

Definition 4

A representing function of the preference relation is a function \(V : {\mathscr {L}^{\infty }(\mathcal {F})}\rightarrow \mathbb {R}\) which is order-preserving, i.e.,

$$ f\succeq g \iff V(f) \ge V(g). $$

We use the standard conventions: \(f\preceq g\) if \(g\succeq f\); \(f\sim f\) if both \(g\succeq f\) and \(f\succeq g\); \(g\not \sim f\) if either \(g\nsucceq f\) or \(f\nsucceq g\); \(g\succ f\) if \(g\succeq f\) but \(f \nsucceq g\).

Definition 5

An event \(A\in {\mathcal F}\) is null if \(f\mathbf {1}_A + g\mathbf {1}_{A^c} \sim g\) \(\forall f,g \in {\mathscr {L}^{\infty }(\mathcal {F})}\).We shall denote by \(\mathcal {N}(\mathcal {F})\) be the set of null events.

As a consequence a \(\succeq \)-atom is an element \(A\in {\mathcal F}\) such that for every \(B\in {\mathcal F}\) with \(\varnothing \ne B\subset A\) either B or \(A{\setminus } B\) is null.

An event is essential if it belongs to \({\mathcal F}{\setminus } \mathcal {N}(\mathcal {F})\).

We can consider the following additional Axioms:

(A2):

Strictly monotone if \(x\mathbf {1}_A + f\mathbf {1}_{A^c} \succ y\mathbf {1}_A + f\mathbf {1}_{A^c}\), for all nonnull events \(A\in {\mathcal F}\), for all \(f \in {\mathscr {L}^{\infty }(\mathcal {F})}\) and outcomes \(x > y\).

(A3):

Sure-thing principle: consider arbitrary \(f, g, h \in {\mathscr {L}^{\infty }(\mathcal {F})}\) and \(A\in {\mathcal F}\) such that \(f\mathbf {1}_A + h\mathbf {1}_{A^c} \preceq g\mathbf {1}_A + h\mathbf {1}_{A^c}\) then for every \(c \in {\mathscr {L}^{\infty }(\mathcal {F})}\) we have \(f\mathbf {1}_A + c\mathbf {1}_{A^c} \preceq g\mathbf {1}_A + c\mathbf {1}_{A^c}\).

(A3) holds on \({\mathcal S}({\mathcal F})\) if we substitute in the previous statement \({\mathscr {L}^{\infty }(\mathcal {F})}\) with \({\mathcal S}({\mathcal F})\) (as defined in the paragraph Notations).

(A4’):

Norm continuity if \(\forall f \in {\mathscr {L}^{\infty }(\mathcal {F})}\) the sets \(\{ g \in {\mathscr {L}^{\infty }(\mathcal {F})}: g \succeq f \}\) and \(\{ g \in {\mathscr {L}^{\infty }(\mathcal {F})}: f \succeq g \}\) are \(\Vert \cdot \Vert _{\infty }\)-closed.

Theorem 8

(Debreu 1960, state-dependent expected utility for finite state space) Let \({\mathscr {L}^{\infty }(\mathcal {F})}\) the set of acts and \(\succeq \) a preference relation on it. Let the state space \(\varOmega = \{ \omega _1, \ldots , \omega _n \}\), where at least three states are nonnull. Then the following two statements are equivalent:

  1. (i)

    There exist n continuous functions \(V_j : \mathbb {R} \rightarrow \mathbb {R} \), \(j = 1, \ldots , n\), that are strictly increasing for all nonnull states and constant for all null states, and such that \(\succeq \) is represented by

    $$\begin{aligned} V(f) = \sum _{j=1}^{n} V_j(f(\omega _j)). \end{aligned}$$
    (B.1)
  2. (ii)

    \(\succeq \) is a norm continuous, strictly monotonic preference order that satisfies the sure thing principle.

The following uniqueness holds for (1) : \(W(f) = \sum _{j=1}^{n} W_j(f(\omega _j))\) represent \(\succeq \) if and only if there exist \(\tau _1, \ldots , \tau _n \in \mathbb {R}\) and \(\sigma > 0 \) such that \(W_j = \tau _j + \sigma V_j\) \(\forall j\), implying that \(W = \tau + \sigma V\) for \(\tau = \tau _1 + \cdots + \tau _n\).

In [29] the previous Theorem is generalized to an infinite state spaces \(\varOmega \) when \(\varOmega \) contains no atoms. We here recall the integral reformulation of the Debreu representation given in [3] under pointwise continuity.

Definition 6

A preference order is

(A4):

Pointwise continuous if for any uniformly bounded sequence \(\{f_n\}\subseteq {\mathscr {L}^{\infty }(\mathcal {F})}\), such that \(f_n(\omega )\rightarrow f(\omega )\) for any \(\omega \in \varOmega \) then \(\forall g \in {\mathscr {L}^{\infty }(\mathcal {F})}\) such that \(g \succ f\) (resp. \(g \prec f\)) \(\exists J \in \mathbb {N}\) such that \(g \succ f^j\) (resp. \(g \prec f^j\)) \(\forall j > J\).

Theorem 9

([29], Theorem 12 and [3], Theorem 5) Let \({\mathscr {L}^{\infty }(\mathcal {F})}\) be the set of acts and \(\succeq \) the preference relation on it. Assume that \({\mathcal F}\) contains at least three disjoint essential events. Then the following two statements are equivalent:

  1. (i)

    There exists a countably additive measure \(\mathbb {P}\) on \(\varOmega \) and a function (the state-dependent utility) \(u(\omega , \cdot ) : \mathbb {R} \rightarrow \mathbb {R}\) strictly increasing \(\forall \omega \in \varOmega \), such that \(\succeq \) is represented by the pointwise continuous integral

    $$ f \rightarrow \int _{\varOmega } u(\omega , f(\omega )) d\mathbb {P}. $$
  2. (ii)

    \(\succeq \) satisfies: (A1), (A2), (A3) on \({\mathcal S}({\mathcal F})\), (A4).

The following uniqueness holds: the couple \((\mathbb {P},u)\) can be replaced by \((\mathbb {P}^*, u^*)\) if and only if \(\mathbb {P}\) and \(\mathbb {P}^*\) are equivalent and \({\mathbb P}(u^* = \tau + \sigma \delta u)=1\), where \(\tau : \varOmega \rightarrow \mathbb {R}\) is \({\mathcal F}\)-measurable, \(\sigma > 0\) and \(\delta \) is the Radon-Nikodym density function of \(\mathbb {P}\) with respect to \(\mathbb {P}^*\).

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Maggis, M., Maran, A. Stochastic dynamic utilities and intertemporal preferences. Math Finan Econ (2021). https://doi.org/10.1007/s11579-020-00290-x

Download citation

Keywords

  • Intertemporal decisions
  • Stochastic dynamic utility
  • Conditional preferences
  • Sure thing principle

JEL Classification

  • C02
  • D81