Abstract
We compute the optimal investment and consumption strategies for an individual who wishes to maximize her expected discounted exponential utility of lifetime consumption, while imposing a constraint on the expected time her wealth spends below a poverty threshold b. First, we compute the optimal strategies for the corresponding (unconstrained) problem with a running penalty for time that wealth spends below b. This penalty acts as a Lagrange multiplier for our original constrained problem, so we recover the optimal strategies for our original problem from the recast problem. We show that (1) if the current wealth is greater than b, then the optimal investment strategy becomes more conservative as the poverty constraint becomes sharper; and (2) if the current wealth is less than b, then the optimal investment strategy is either independent of the poverty constraint or becomes more aggressive as the poverty constraint becomes sharper, depending on the value b. We also show that the optimal rate of consumption (weakly) decreases as the poverty constraint becomes sharper.
We’re sorry, something doesn't seem to be working properly.
Please try refreshing the page. If that doesn't work, please contact support so we can address the problem.
Notes
There is debate over measures of poverty. Total wealth (that is, net assets) is one of the prevailing ones, because it precisely measures the current financial state of an individual.
Think of P as a Lagrange multiplier for the problem in (2.1).
To guarantee the existence of \({\mathcal {H}}_2\)’s zero \(y_b(P) \in (0, 1)\) for some \(P \le 0\), we require \({\overline{P}} < 0\), that is, \(b > \frac{1}{\delta \eta \rho _{-}^2 (\rho _{+} - \rho _{-})} \,\).
References
Atkinson, C., Papakokkinou, M.: Theory of optimal consumption and portfolio selection under a Capital-at-Risk (CaR) and a Value-at-Risk (VaR) constraint. IMA J Manag Math 16(1), 37–70 (2005)
Bayraktar, E., Young, V.R.: Maximizing utility of consumption subject to a constraint on the probability of lifetime ruin. Finance Res Lett 5(4), 204–212 (2008)
Bayraktar, E., Virginia, R.Y.: Optimally investing to reach a bequest goal. Insur Math Econ 70, 1–10 (2016)
Grandits, P.: An optimal consumption problem in finite time with a constraint on the ruin probability. Finance Stochast 19(4), 791–847 (2015)
Karatzas, I., Lehoczky, J.P., Sethi, S.P., Shreve, S.E.: Explicit solution of a general consumption/investment problem. Math Oper Res 11(2), 261–294 (1986)
Lehoczky, J.P., Sethi, S.P., Shreve, S.E.: Optimal consumption and investment policies allowing consumption constraints and bankruptcy. Math Oper Res 8(4), 613–636 (1983)
Merton, R.C.: Lifetime portfolio selection under uncertainty: the continuous-time case. Rev Econ Stat 51(3), 247–257 (1969)
Merton, R.C.: Optimum consumption and portfolio rules in a continuous-time model. J Econ Theory 3(4), 373–413 (1971)
Pirvu, T.A.: Portfolio optimization under the value-at-risk constraint. Quant Finance 7(2), 125–136 (2007)
Samuelson, P.A.: Lifetime portfolio selection by dynamic stochastic programming. Rev Econ Stat 51(3), 239–246 (1969)
Yiu, K.-F.C.: Optimal portfolios under a value-at-risk constraint. J Econ Dyn Control 28(7), 1317–1334 (2004)
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.
Appendix: Proof of Proposition 1
Appendix: Proof of Proposition 1
Note that depending on the value of \(y_b\), (3.1) can be divided into two cases, specifically, \(y_b> 1\) and \(y_b\le 1\).
First, we consider the case for which \(y_b> 1\). In this case, the FBP (3.1) becomes
Define the function \({\mathcal {H}}_1\) by
for \(y > 0\). It follows that
Note that if \(P \le \min \big (\, {{\overline{P}}}, 0 \big )\), in which \({{\overline{P}}}\) is given in (3.7), then \({\mathcal {H}}_1(1; P) \le 0\), which implies that \({\mathcal {H}}_1\) has a unique zero \(y_b= y_b(P) \in [1, \infty )\).
It is tedious, but straightforward, to verify by substitution that \({\mathcal {V}}_1\) given by
classically solves FBP (A.1) and is \({\mathcal {C}}^{1}\) at \(y = y_b\). Here, the coefficients are given in (3.4).
Next, we show that \({\mathcal {V}}_1(y; P)\) is decreasing and convex with respect to y on \(\mathbb {R}_{+}\), for \(P \in \big [ {\underline{P}}, \, \min \big (\, {{\overline{P}}}, 0 \big ) \big ]\), in which \({\underline{P}}\) is given in (3.3). Rearranging \({\underline{P}}\) yields the useful identity
Define \({\mathcal {P}}_1:\mathbb {R}_{+}/\{0\} \rightarrow \mathbb {R}\) by
Note that \({\mathcal {P}}_1(y_b(P)) = P\) when \(P \le \min \big (\, {{\overline{P}}}, 0 \big )\). Because \({\mathcal {P}}_1\) is decreasing, it has an inverse \({\mathcal {Y}}_1 = {\mathcal {P}}_1^{-1}: \mathbb {R} \rightarrow \mathbb {R}_{+}/\{0\}\), which is also decreasing, with \({\mathcal {Y}}_1(P) = y_b(P)\) when \(P \le \min \big (\, {{\overline{P}}}, 0 \big )\). Next, define \({\mathcal {G}}_1:\mathbb {R} \rightarrow \mathbb {R}\) by
which increases with P because \(({\mathcal {G}}_1)_P = 1 + \frac{\delta \kappa _{-}}{r \eta } \, {\mathcal {Y}}_1^{\rho _{-}} \cdot ({\mathcal {Y}}_1)_P > 0\). From (A.2) and (A.3), we deduce that \({\mathcal {Y}}_1^{\kappa _{-}}({\underline{P}}) = -r \eta \, {\underline{P}}/\delta \); thus, if \(P \ge {\underline{P}}\), then
The second derivative of \({\mathcal {V}}_1\) with respect to y equals
Given the inequality in (A.4), for \( y \in (0, 1)\) and \(P \in \big [ {\underline{P}}, \, \min \big (\, {{\overline{P}}}, 0 \big ) \big ]\),
in which the second inequality follows from \(y_b= y_b(P) > 1\). For \(y \in [1, y_b)\) and \(P \in \big [ {\underline{P}}, \, \min \big (\, {{\overline{P}}}, 0 \big ) \big ]\),
in which the first inequality follows from the fact that the expression in square brackets decreases with respect to y; recall that \(P \le 0\). Clearly, for \(y \in (y_b, \infty )\), \(({\mathcal {V}}_1)_{yy}(y; P) > 0\) because \(C_4 > 0\). Because \(\lim _{y \rightarrow \infty } ({\mathcal {V}}_1)_y(y; P) = \lim _{y \rightarrow \infty } C_4 \kappa _{-} y^{\rho _{-}} = 0\), \({\mathcal {V}}_1\) is decreasing and convex with respect to y on \(\mathbb {R}_{+}\), for \(P \in \big [ {\underline{P}}, \, \min \big (\, {{\overline{P}}}, 0 \big ) \big ]\).
To summarize, if \(b \le \frac{1}{\delta \eta \rho _{-}^2 (\rho _{+} - \rho _{-})} \,\), then \({\overline{P}} \ge 0\); thus, \({\mathcal {V}}_1\) is decreasing and convex with y for all \(P \in \left[ {\underline{P}}, 0 \right] \). If \(\frac{1}{\delta \eta \rho _{-}^2 (\rho _{+} - \rho _{-})} < b \le - \, \frac{1}{r \eta \rho _{-}} \,\), then \({\underline{P}} \le {{\overline{P}}} < 0\); thus \({\mathcal {V}}_1\) is decreasing and convex with y for all \(P \in \left[ {\underline{P}},{\overline{P}} \right] \). Finally, if \(b > - \, \frac{1}{r \eta \rho _{-}} \,\), then there does not exist \(P \le 0\) such that \({\mathcal {V}}_1\) is decreasing and convex because \({\underline{P}} > {\overline{P}}\) in that case.
Second, we consider the case for which \(y_b\le 1\). In this case, the FBP (3.1) becomes
We introduce an identity to use in this proof.
Define the function \({\mathcal {H}}_2\) by
for \(y > 0\). By using (A.6), we compute the following quantities.
If \(P > {\overline{P}}\), in which \({{\overline{P}}}\) is given in (3.7), then \({\mathcal {H}}_2(1; P) > 0\); thus, \({\mathcal {H}}_2\) has an odd number, \(2k + 1\), of zeros in (0, 1), for some \(k = 0, 1, \ldots \). It follows that \(({\mathcal {H}}_2)_y\) has \(2k + 1\) zeros in (0, 1), and \(({\mathcal {H}}_2)_{yy}\) has 2k zeros in (0, 1). From the above expression for \(({\mathcal {H}}_2)_{yy}\), it is clear that the second derivative changes sign at most once in (0, 1); thus, \(k = 0\). Thus, for \(P \in ( {\overline{P}}, 0)\), \({\mathcal {H}}_2\) has a unique zero \(y_b= y_b(P)\) in (0, 1).Footnote 3
As in the case for which \(y_b> 1\), it is tedious, but straightforward, to verify that
classically solves FBP (A.5) and is \({\mathcal {C}}^{1}\) at \(y = y_b\). Here, the coefficients are given in (3.8).
Next, we show that \({\mathcal {V}}_2(y; P)\) is decreasing and convex with respect to y on \(\mathbb {R}_{+}\) for appropriate ranges of P, specified below. The second derivative of \({\mathcal {V}}_2\) with respect to y is given by
If \(y \in [1, \infty )\), it is clear that \(({\mathcal {V}}_2)_{yy}(y; P) > 0\) because \(D_4 > 0\). If \(y \in (y_b, 1)\), then, for all \(P \in \big ( {\overline{P}}, 0 \big ]\),
in which the first inequality follows because the expression in square brackets decreases with respect to y.
It remains to show that \(({\mathcal {V}}_2)_{yy}(y; P) > 0\) for \(y \in (0, y_b)\). To that end, define \({\mathcal {P}}_2:\mathbb {R}_{+}/\{0\} \rightarrow \mathbb {R}\) by
Note that \({\mathcal {P}}_2(y_b(P)) = P\) when \(P \in \big (\, {{\overline{P}}}, 0 \big ]\). From \({\mathcal {P}}_2\)’s expression, we obtain
in which we use the identity in (A.6) to simplify \(({\mathcal {P}}_2)_y(1)\). From the above information, we deduce that \({\mathcal {P}}_2\) has an odd number, \(2k + 1\), of zeros in (0, 1), for some \(k = 0, 1, \ldots \). It follows that \(({\mathcal {P}}_2)_y\) has \(2k + 1\) zeros in (0, 1), and \(({\mathcal {P}}_2)_{yy}\) has 2k zeros in (0, 1). From the above expression for \(({\mathcal {P}}_2)_{yy}\), it is clear that the second derivative changes sign at most once in (0, 1); thus, \(k = 0\). Thus, for \(P \in \big (\, {{\overline{P}}}, 0 \big ]\), \({\mathcal {P}}_2\) has a unique zero \({{\hat{y}}} = {{\hat{y}}}(P)\) in (0, 1), and \({\mathcal {P}}_2\) is decreasing on \(\big [{{\hat{y}}}, 1 \big ]\).
We, then, define
which is also decreasing, with \({\mathcal {Y}}_2\big ({\overline{P}} \big ) = 1\), \({\mathcal {Y}}_2(0) = {\hat{y}}\), and \({\mathcal {Y}}_2(P) = y_b(P)\) more generally. Then, for \(y \in (0, y_b)\),
in which the inequality follows because the expression in square brackets decreases with respect to y. Next, define \({\mathcal {G}}_2: \big [ {\overline{P}}, 0 \big ] \rightarrow \mathbb {R}\) by
It follows that
If \(b \le - \, \frac{1}{r \eta \rho _{-}}\),
Thus, because \({\mathcal {G}}_2\) increases with respect to P, if \(b \le - \, \frac{1}{r \eta \rho _{-}}\) and \(P \in \big [ {\overline{P}}, 0 \big ]\), then \({\mathcal {G}}_2(P) \ge 0\), or equivalently, \(({\mathcal {V}}_2)_{yy}(y; P) \ge 0\) for \(y \in (0, y_b)\). If \(b > - \, \frac{1}{r \eta \rho _{-}}\), then \({\mathcal {G}}_2 \big ( {\overline{P}} \big ) < 0\), and there exists \({\hat{P}} \in ({\overline{P}}, 0]\) such that \({\mathcal {G}}_2\big ( {\hat{P}} \big ) = 0\), and for \(P > {\hat{P}}\), \({\mathcal {G}}_2(P) > 0\).
To summarize, if \(\frac{1}{\delta \eta \rho _{-}^2 (\rho _{+} - \rho _{-})} < b \le - \, \frac{1}{r \eta \rho _{-}} \,\), then there exists a decreasing, convex solution \({\mathcal {V}}_2\) of (A.5) for all \(P \in \big ( {\overline{P}}, 0 \big ]\). If \(b > - \, \frac{1}{r \eta \rho _{-}} \,\), then there exists such a solution for all \(P \in \big [ {{\hat{P}}}, 0 \big ]\). \(\square \)
Rights and permissions
About this article
Cite this article
Li, D., Young, V.R. Maximizing expected exponential utility of consumption with a constraint on expected time in poverty. Ann Finance 16, 63–99 (2020). https://doi.org/10.1007/s10436-019-00354-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10436-019-00354-z