Abstract
We consider the problem of controlling the diffusion coefficient of a diffusion with constant negative drift rate such that the probability of hitting a given lower barrier up to some finite time horizon is minimized. We assume that the diffusion rate can be chosen in a progressively measurable way with values in the interval [0, 1]. We prove that the value function is regular, concave in the space variable, and that it solves the associated HJB equation. To do so, we show that the heat equation on a right triangle, with a boundary condition that is discontinuous in the corner, possesses a smooth solution.
Similar content being viewed by others
References
Abundo, M.: Some conditional crossing results of Brownian motion over a piecewise-linear boundary. Stat. Probab. Lett. 58(2), 131–145 (2002)
Althöfer, I., Beckmann, M., Salzer, F.: On some random walk games with diffusion control. In: Plaat, A., van den Herik, J., Kosters, W. (eds.) Advances in Computer Games, pp. 65–75. Springer International Publishing, Cham (2015)
Ankirchner, S., Blanchet-Scalliet, C., Jeanblanc, M.: Controlling the occupation time of an exponential martingale. Appl. Math. Optim. 76(2), 415–428 (2017)
Asmussen, S., Albrecher, H.: Ruin Probabilities. World Scientific Publishing Co Pte Ltd, Singapore (2010)
Asmussen, S., Taksar, M.: Controlled diffusion models for optimal dividend pay-out. Insurance 20(1), 1–15 (1997)
Avellaneda, M., Levy, A., Parás, A.: Pricing and hedging derivative securities in markets with uncertain volatilities. Appl. Math. Finance 2(2), 73–88 (1995)
Azcue, P., Muler, N.: Stochastic Optimization in Insurance: A Dynamic Programming Approach. Springer, Berlin (2014)
Bäuerle, N., Bayraktar, E.: A note on applications of stochastic ordering to control problems in insurance and finance. Stoch. Int. J. Probab. Stoch. Process. 86(2), 330–340 (2014)
Bayraktar, E., Young, V.R.: Correspondence between lifetime minimum wealth and utility of consumption. Finance Stoch. 11(2), 213–236 (2007)
Bayraktar, E., Young, V.R.: Optimal investment strategy to minimize occupation time. Ann. Oper. Res. 176, 389–408 (2010)
Bayraktar, E., Young, V.R.: Proving regularity of the minimal probability of ruin via a game of stopping and control. Finance Stoch. 15(4), 785–818 (2011)
Bertoin, J., Chaumont, L., Pitman, J., et al.: Path transformations of first passage bridges. Electron. Commun. Probab. 8, 155–166 (2003)
Biane, P., Yor, M.: Quelques précisions sur le méandre brownien. Bull. Sci. Math. 112(1), 101–109 (1988)
Borodin, A.N., Salminen, P.: Handbook of Brownian Motion–Facts and Formulae. Probability and Its Applications, 2nd edn. Birkhäuser, Basel (2002)
Browne, S.: Optimal investment policies for a firm with a random risk process: exponential utility and minimizing the probability of ruin. Math. Oper. Res. 20(4), 937–958 (1995)
Callander, S.: Searching for good policies. Am. Polit. Sci. Rev. 105(4), 643–662 (2011)
Callander, S., et al.: A theory of policy expertise. Q. J. Polit. Sci. 3(2), 123–140 (2008)
Caraco, T., Martindale, S., Whittam, T.S.: An empirical demonstration of risk-sensitive foraging preferences. Anim. Behav. 28(3), 820–830 (1980)
David Promislow, S., Young, V.R.: Minimizing the probability of ruin when claims follow brownian motion with drift. N. Am. Actuar. J. 9(3), 110–128 (2005)
de l’Assurance, F.F.: Les assurances de bien et de responsabilité, données clés 2015, (2016)
Denis, L., Martini, C.: A theoretical framework for the pricing of contingent claims in the presence of model uncertainty. Ann. Appl. Probab. 16(2), 827–852 (2006)
Downs, G.W., Rocke, D.M.: Conflict, agency, and gambling for resurrection: the principal-agent problem goes to war. Am. J. Polit. Sci. 38, 362–380 (1994)
Fokas, A., Kalimeris, K.: The heat equation in the interior of an equilateral triangle. Stud. Appl. Math 124, 283–305 (2010)
Glick, D., Myers, C.D.: Learning from others: an experimental test of brownian motion uncertainty models. J. Theor. Polit. 27(4), 588–612 (2015)
Hipp, C., Vogt, M.: Optimal dynamic xl reinsurance. ASTIN Bull. 33(2), 193–207 (2003)
Iglehart, L.D.: Diffusion approximations in collective risk theory. J. Appl. Probab. 6(2), 285–292 (1969)
Liang, Z., Guo, J.: Optimal proportional reinsurance and ruin probability. Stoch. Models 23(2), 333–350 (2007)
Matoussi, A., Possamaï, D., Zhou, C.: Robust utility maximization in non-dominated models with 2bsde: the uncertain volatility model. Math. Finance 25(2), 258–287 (2015)
McNamara, J., Houston, A., Collins, E.J.: Optimality models in behavioral biology. Siam Rev. 43(3), 413–466 (2001)
McNamara, J.M.: Optimal control of the diffusion coefficient of a simple diffusion process. Math. Oper. Res. 8(3), 373–380 (1983)
A. of British Insurers. U.k. insurance and long-term savings, the state of the market 2019, (2020)
Peng, S.: G-expectation, g-brownian motion and related stochastic calculus of itô type. In: Benth, F.E., et al. (eds.) Stochastic Analysis and Applications, pp. 541–567. Springer, Berlin (2007)
Salminen, P.: On last exit decompositions of linear diffusions. Studia Sci. Math. Hungar. 33(1–3), 251–262 (1997)
Salminen, P., Yor, M.: On hitting times of affine boundaries by reflecting brownian motion and bessel processes. Period. Math. Hung. 62(1), 75–101 (2011)
Schmidli, H.: Optimal proportional reinsurance policies in a dynamic setting. Scand. Actuar. J. 2001(1), 55–68 (2001)
Schmidli, H.: Stochastic Control in Insurance. Springer Science & Business Media, Berlin (2007)
Schmidli, H., et al.: On minimizing the ruin probability by investment and reinsurance. Ann. Appl. Probab. 12(3), 890–907 (2002)
Smith, W., Watson, G.: Diffusion out of a triangle. J. Appl. Probab. 4(3), 479–488 (1967)
Stephens, D.W.: The logic of risk-sensitive foraging preferences. Anim. Behav. 29, 628–629 (1981)
Stephens, D.W., Krebs, J.R.: Foraging Theory, vol. 1. Princeton University Press, Princeton (1986)
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
C. Zhou: Research supported by NSFC Grant No. 11871364 as well as Singapore MOE (Ministry of Educations) AcRF Grants R-146-000-271-112 and R-146-000-284-114.
Appendix. A
Appendix. A
In this appendix we prove the continuity and smoothness properties of the function H claimed in Theorem 3.3. To shorten notation, we take \(b=1\) and we show the respective properties for
Lemma A.1
For every t, \(x\in (t,T)\mapsto \overline{H}(t,x)\) is continuous.
Proof
Let \(t\in [0,T)\) and \(\delta >0\), we prove that for all \(x\in [t,T-\delta ]\), the function is continuous in x. For \(t<x<T-\delta \), using Proposition 3.1, and a simple change of variables, we get
with
Note that \(\sum _{k=-\infty }^{+\infty } h_k(u,x) \leqslant 1\), since it is a probability.
-
1.
For all \(x\in [t,T-\delta ]\),
$$\begin{aligned} \frac{1}{\sqrt{2\pi }u^{3/2}}\exp \left( -\frac{(T-x)^2}{2u}\right) \le \frac{1}{\sqrt{2\pi }u^{3/2}}\exp \left( -\frac{(T-(T-\delta ))^2}{2u}\right) , \end{aligned}$$and this function \(u\mapsto \frac{1}{\sqrt{2\pi }u^{3/2}}\exp \left( -\frac{(T-(T-\delta ))^2}{2u}\right) \) is integrable on \((0,T-t)\).
-
2.
\(x \mapsto \frac{1}{\sqrt{2\pi }u^{3/2}}\exp \left( -\frac{(T-x)^2}{2u}\right) \) is continuous on \([t,T-\delta ]\).
-
3.
Let us prove that the series \(\sum _{k=-\infty }^{+\infty } h_k(u,x)\) is a continuous function of x. Since each \(h_k\) is a continuous function of x, it is sufficient to prove that the series converges uniformly, which is implied by the fact that
$$\begin{aligned}&\sum _{k=-\infty }^{+\infty } \sup _{t\le x\le T} \left| h_k(u,x)\right| <+\infty , \text { since}\\&\quad \left| h_k(u,x)\right| \le (T-t)|2k-1|\exp \left( -2\frac{(T-u-t)(T-t)}{u} k(k-1)\right) . \end{aligned}$$
The point 1. above and the fact that \(\sum _{k=-\infty }^{+\infty } h_k(u,x) \leqslant 1\) show that the integrand in (A.1) is bounded from above by an integrable function, which is independent of x. The points 2. and 3. allow to deduce that the integrand in (A.1) is a continuous function of x. Hence \(x\mapsto \overline{H}(t,x)\) is continuous, for \(t<x<T\).
Let us now prove that \(\lim _{x\rightarrow T^-}\overline{H}(t,x)=1\). Recall that
with
Isolating the term \(k=0\) in the sum, we get,
When \(x\rightarrow T\), the second term in the previous integral simplifies to
and
Finally, for \(x=t\), we have
since
goes to 0 as N goes to \(+\infty \), for any \(C>0\). \(\square \)
The following observations will be useful: we can write \(\overline{H}(t,x)\) in the following way:
with \(C_{t,u,T}:= \frac{(T-t)(T-u)}{u-t}\). This can be written
with
Lemma A.2
Both series \(\sum f^1_k\) and \(\sum f^2_k\) are \(\mathcal C^2\) functions of x and
Proof
We will prove the claim for \(\sum f^1_k\), the proof for \(\sum f^2_k\) being very similar. For any \(k\ge 1\), the map \(x\mapsto f^1_k(x,u)\) is \(C^2\) and
The function \(\sinh \) being increasing,
which implies that the series \(\sum \frac{\partial ^2}{\partial x^2} f^1_k(x,u)\) converges uniformy, and from which we deduce the claim. \(\square \)
Lemma A.3
For every t, \(x\mapsto \overline{H}(t,x)\) is \(\mathcal C^2\).
Proof
As previously, let \(x\in [t,T-\delta ]\) where \(\delta >0\). We have
Using Lemma A.2, we get
where
and
We have
For any \(x\in [t, T-\delta ]\), we have
and
The two functions on the RHS are integrable on \((0,T-t)\).
We can show that all the terms appearing in the quantities \(\left| \frac{1}{\sqrt{2\pi }(u-t)^{3/2}}\exp \left( -\frac{(T-x)^2}{2(u-t)}\right) I_2\right| \) and \(\left| \frac{1}{\sqrt{2\pi }(u-t)^{3/2}}\exp \left( -\frac{(T-x)^2}{2(u-t)}\right) I_3\right| \) are bounded from above by \(\frac{K}{(u-t)^{1/2}}\), for \(u\in (t,t+\varepsilon )\), where \(\varepsilon >0\) and where K is a positive constant (independent of u and x). Let us for example study in detail the term
the arguments for the other terms are very similar.
Using that \(\sinh \) is increasing and that \(C_{t,u,T} = \frac{(T-t)^2}{u-t}- (T-t)\), we get
Define
where \(c=2(T-t)^2\) and \(c'=2(T-t)\) (\(\frac{c}{u}-c'\ge 0\) for \(u\le T-t\)). We are going to prove that \(\lim _{u\rightarrow 0^+} u\log S(u) = 0\). Assuming momentarily that this limit is indeed equal to 0, then there exists \(\bar{u}>t\) such that for any \(u\le \bar{u}\),
This implies for any \(x\in [t,T-\delta ]\) and any \(u\le \bar{u}\),
This implies, by splitting the interval [t, T] into \([t,t+\varepsilon )\) and \([t+\varepsilon , T]\) with \(\varepsilon \) small enough, that \(\left| \frac{\partial ^2h}{\partial x^2}(x,u)\right| \) is bounded from above by an integrable function which is independent of x, on each of these intervals. Using similar arguments as in the proof of Lemma A.2, we can show that \(x \mapsto \frac{\partial ^2h}{\partial x^2}(x,u)\) is a continuous function. From this, we can conclude that \(x\mapsto \overline{H}(t,x)\) is \(\mathcal C^2\).
It remains to prove that \(\lim _{u\rightarrow 0^+} u\log S(u) = 0\), which is the purpose of Lemma A.4. \(\square \)
Lemma A.4
\(\lim _{u\rightarrow 0^+} u\log S(u) = 0\).
Proof
Define,
which is a differentiable function such that
The sign of this derivative is given by the sign of \(v \frac{c}{u} \cosh \!\left( \!v \frac{c}{u}\!\right) \!-\!\left( \!2 v^2[\frac{c}{u}\!-\!c'] \!-\!3\!\right) \sinh \!\left( \!v \frac{c}{u}\!\right) \), which is going to be negative for u small enough. Indeed, this quantity is negative if and only if
Since \(\tanh \left( v \frac{c}{u}\right) \left( 2v - 2v \frac{c'}{c}u - \frac{3}{v} \frac{u}{c}\right) \) converges to \(2v>1\) as u goes to 0, we have that \(v\mapsto \varphi (v,u)\) is decreasing for u small enough. We can then apply the standard technique of comparison between series and integrals: for \(k \ge 1\),
Taking the sum from \(k=2\) to \(k=N\), and then the limit when N goes to \(+\infty \), we have
Let us compute the integral on the right hand side, (which is well defined since for v in the neighbourhood of \(+\infty \), \(\varphi (v,u) = o(\frac{1}{v^p})\) for any \(p>1\)), that we expand as:
Writing \(\bar{\Phi }_j(x):= E[Z^j\mathbf {1}_{\{Z>x\}}]\), \(j \ge 0\) and \(\bar{\Phi }(x):=\bar{\Phi }_0(x)=P(Z>x)\), with Z a r.v. having a standard Gaussian distribution, we have
where \(\kappa _1(u):= \sqrt{2(\frac{c}{u}-c')}\) and \(\kappa _2(u):= \frac{c}{u\kappa _1(u)}\) (for simplicity, we do not indicate the dependence on u of \(\kappa _1\) and \(\kappa _2\) when there is no ambiguity). Similarly we have
By an integration by parts,
Using these explicit expressions, the value of \(\varphi (1,u)\) given through (A.3) and inequality (A.4), it is straightforward to show that \(u\log S(u)\) converge to 0 as u goes to \(0^+\). \(\square \)
Lemma A.5
For every \(x\in (0,T)\), \(t \in [0,x] \mapsto \overline{H}(t,x)\) is \(\mathcal C^1\).
Proof
Recall that (do a change of variable \(v=\frac{u-t}{T-t}\))
where
Let \(x\in [0,T)\), \(t_0\le 0\) and \(\delta >0\) such that \(t_0+\delta \in [0,x)\).
-
For all \(t\in [0, t_0+\delta ]\)
$$\begin{aligned} \left| \frac{\partial f}{\partial t} \right|&=\frac{T-x}{2\sqrt{2\pi }(T-t)^{3/2}(u)^{3/2}}\exp \left( -\frac{(T-x)^2}{2(T-t)u}\right) \\&\quad +\frac{(T-x)^3}{\sqrt{2\pi }(T-t)^{3/2}(u)^{5/2}}\exp \left( -\frac{(T-x)^2}{2(T-t)u}\right) \\&\le \frac{T-x}{2\sqrt{2\pi }(T-t_0-\delta )^{3/2}(u)^{3/2}}\exp \left( -\frac{(T-x)^2}{2Tu}\right) \\&\quad +\frac{(T-x)^3}{\sqrt{2\pi }(T-t_0-\delta )^{3/2}(u)^{5/2}}\exp \left( -\frac{(T-x)^2}{2Tu}\right) . \end{aligned}$$The right hand side is an integrable function on [0, 1] independent of \(t\in [0, t_0+\delta ]\). The function \(t\mapsto \frac{\partial f}{\partial t}\) is continuous on \([0, t_0+\delta ]\)
-
Note that \(g(t,x,u)=f(t,x,u)/(T-x)\). Hence the functions \(t\mapsto g(t,x,u)\) and \(t\mapsto \frac{\partial g}{\partial t}\) are continuous on \([0, t_0+\delta ]\) and can be bounded from above by a function independent of f and integrable w.r.t u.
-
Let us prove that the \(\sum _{k=-\infty , k\ne 0}^{+\infty } h_k(t,x,u) \) is a \(\mathcal C^1\) function. Since
$$\begin{aligned}&\sup _{t\in [0,t_0+\delta ]}\left| \frac{\partial h_k(t,x,u)}{\partial t}\right| \\&\quad =\sup _{t\in [0,t_0+\delta ]}\left| 2k\exp \left( \frac{1-u}{u}\left[ -2k^2(T-t)+2k(T-x)\right] \right) \right. \\&\qquad + \left. 2k^2\frac{1-u}{u}(T-x-2k(T-t))\exp \left( \frac{1-u}{u}\left[ -2k^2(T-t)+2k(T-x)\right] \right) \right| \\&\quad \le 2|k|\exp \left( \frac{1-u}{u}\left[ -2k^2(T-t_0-\delta )+2k(T-x)\right] \right) \\&\qquad +2k^2\frac{1-u}{u}(T-x+2|k|T)\exp \left( \frac{1-u}{u}\left[ -2k^2(T-t_0-\delta )+2k(T-x)\right] \right) \\&\quad \le 2|k|\exp \left( \frac{1-u}{u}\left[ -2k^2(T-x)+2k(T-x)\right] \right) \\&\qquad +2k^2\frac{1-u}{u}(T-x+2|k|T)\exp \left( \frac{1-u}{u}\left[ -2k^2(T-x)+2k(T-x)\right] \right) , \end{aligned}$$which is the term of a convergent series, the result follows.
-
\(|\frac{\partial g(t,x,u)}{\partial t}\sum _{k=-\infty , k\ne 0}^{+\infty } h_k(t,x,u)|\le |\frac{\partial g(t,x,u)}{\partial t}|\), which is bounded from above by an integrable function independent of \(t\in [0, t_0+\delta ]\) (see above).
-
Define
$$\begin{aligned} \widetilde{S}_n(u):= \sum _{k=1}^{+\infty } k^n \exp \left( -k^2c\left[ \frac{1-u}{u}\right] \right) \sinh \left( kc\frac{1-u}{u}\right) , \quad 0<u\le (T-t), \end{aligned}$$where \(c=(T-x)\), \(n=1,2,3\). Similar arguments as the ones used in the proof of Lemma A.4 imply that \(\lim _{u\rightarrow 0^+} u \log \widetilde{S}_n(u) = 0\). Then there exists \(\bar{u}\) such that for any \(u\le \bar{u}\)
$$\begin{aligned} \left| g(t,x,u)\sum _{k=-\infty , k\ne 0}^{+\infty }\frac{\partial h_k(t,x,u)}{\partial t}\right| \le C\frac{1}{\sqrt{u}}, \end{aligned}$$where C is constant independent of u and t.\(\square \)
Rights and permissions
About this article
Cite this article
Ankirchner, S., Blanchet-Scalliet, C., Kazi-Tani, N. et al. Gambling for Resurrection and the Heat Equation on a Triangle. Appl Math Optim 84, 3111–3136 (2021). https://doi.org/10.1007/s00245-020-09741-9
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00245-020-09741-9