Skip to main content

A least-squares Monte Carlo approach to the estimation of enterprise risk

Abstract

The estimation of enterprise risk for financial institutions entails a re-evaluation of the company’s economic balance sheet at a future time for a (large) number of stochastic scenarios. The current paper discusses tackling this nested valuation problem based on least-squares Monte Carlo techniques familiar from American option pricing. We formalise the algorithm in an operator setting and discuss the choice of the regressors (“basis functions”). In particular, we show that the left singular functions of the corresponding conditional expectation operator present robust basis functions. Our numerical examples demonstrate that the algorithm can produce accurate results at relatively low computational costs.

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

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7

References

  1. Abbé, E., Zheng, L.: A coordinate system for Gaussian networks. IEEE Trans. Inf. Theory 58, 721–733 (2012)

    MathSciNet  Article  Google Scholar 

  2. Alt, H.W.: Linear Functional Analysis. Springer, Berlin (2012)

    MATH  Google Scholar 

  3. Amemiya, T.: Advanced Econometrics. Harvard University Press, Cambridge (1985)

    Google Scholar 

  4. Bauer, D., Bergmann, D., Reuss, A.: Solvency II and Nested Simulations – a Least-Squares Monte Carlo Approach. Paper presented at the ICA 2010, (2010). http://www.actuaries.org/EVENTS/Congresses/Cape_Town/Papers/Life

  5. Bauer, D., Reuss, A., Singer, D.: On the calculation of the solvency capital requirement based on nested simulations. ASTIN Bull. 42, 453–499 (2012)

    MathSciNet  MATH  Google Scholar 

  6. Belloni, A., Chernozhukov, V., Chetverikov, D., Kato, K.: Some new asymptotic theory for least squares series: pointwise and uniform results. J. Econom. 186, 345–366 (2015)

    MathSciNet  Article  Google Scholar 

  7. Benedetti, G.: On the calculation of risk measures using least-squares Monte Carlo. Int. J. Theor. Appl. Finance 20, 1–14 (2017)

    MathSciNet  Article  Google Scholar 

  8. Birman, M.S., Solomyak, M.Z.: Estimates of singular numbers of integral operators. Russ. Math. Surv. 32, 15–89 (1977)

    MathSciNet  Article  Google Scholar 

  9. Breiman, L., Friedman, J.H.: Estimating optimal transformations for multiple regression and correlation. J. Am. Stat. Assoc. 80, 580–598 (1985)

    MathSciNet  Article  Google Scholar 

  10. Brigo, D., Mercurio, F.: Interest Rate Models – Theory and Practice with Smile, Inflation and Credit, 2nd edn. Springer, Berlin (2006)

    MATH  Google Scholar 

  11. Broadie, M., Du, Y., Moallemi, C.: Risk estimation via regression. Oper. Res. 63, 1077–1097 (2015)

    MathSciNet  Article  Google Scholar 

  12. Cambou, M., Filipović, D.: Replicating portfolio approach to capital calculation. Finance Stoch. 22, 181–203 (2018)

    MathSciNet  Article  Google Scholar 

  13. Carrasco, M., Florens, J.: Spectral method for deconvolving a density. Econom. Theory 27, 546–581 (2011)

    MathSciNet  Article  Google Scholar 

  14. Carriere, J.F.: Valuation of the early-exercise price for options using simulations and nonparametric regression. Insur. Math. Econ. 19, 19–30 (1996)

    MathSciNet  Article  Google Scholar 

  15. Clément, E., Lamberton, D., Protter, P.: An analysis of a least squares regression method for American option pricing. Finance Stoch. 6, 449–471 (2002)

    MathSciNet  Article  Google Scholar 

  16. Cohn, D.L.: Measure Theory, 2nd edn. Birkhäuser, Basel (2013)

    Book  Google Scholar 

  17. Floryszczak, A., Le Courtois, O., Majri, M.: Inside the Solvency 2 black box: Net asset values and solvency capital requirements with a least-squares Monte-Carlo Approach. Insur. Math. Econ. 71, 15–26 (2016)

    MathSciNet  Article  Google Scholar 

  18. Glasserman, P.: Monte Carlo Methods in Financial Engineering. Springer, New York (2004)

    MATH  Google Scholar 

  19. Glasserman, P., Yu, B.: Simulation for American options: regression now or regression later? In: Niederreiter, H. (ed.) Monte Carlo and Quasi-Monte Carlo Methods 2002, pp. 213–226. Springer, Berlin (2004)

    Chapter  Google Scholar 

  20. Gockenbach, M.: Linear Inverse Problems and Tikhonov Regularization. (2016). MAA Press

    Book  Google Scholar 

  21. Golub, G.H., Van Loan, C.F.: Matrix Computations, 4th edn. Johns Hopkins University Press, Baltimore (2013)

    MATH  Google Scholar 

  22. Gordy, M.B., Juneja, S.: Nested simulations in portfolio risk measurement. Manag. Sci. 56, 1833–1848 (2010)

    Article  Google Scholar 

  23. Ha, H.: Essays on Computational Problems in Insurance. PhD Thesis, Georgia State University (2016). Available online at https://scholarworks.gsu.edu/rmi_diss/40/

  24. Hastie, T., Tibshirani, R., Friedman, J.: The Elements of Statistical Learning – Data Mining, Inference, and Prediction. Springer, New York (2009)

    MATH  Google Scholar 

  25. Hong, L.J., Juneja, S., Liu, G.: Kernel smoothing for nested estimation with application to portfolio risk measurement. Oper. Res. 65, 657–673 (2017)

    MathSciNet  Article  Google Scholar 

  26. Huang, Q.: Some Topics Concerning the Singular Value Decomposition and Generalized Singular Value Decomposition. PhD Thesis, ASU (2012). Available online at https://keep.lib.asu.edu/items/151128

  27. Kaina, M., Rüschendorf, L.: On convex risk measures on \(L^{p}\)-spaces. Math. Methods Oper. Res. 69, 475–495 (2009)

    MathSciNet  Article  Google Scholar 

  28. Khare, K., Zhou, H.: Rate of convergence of some multivariate Markov chains with polynomial eigenfunctions. Ann. Appl. Probab. 19, 737–777 (2009)

    MathSciNet  Article  Google Scholar 

  29. Krah, A.S., Nikolić, Z., Korn, R.: A least-squares Monte Carlo framework in proxy modeling of life insurance companies. Risks 6, 62–87 (2018)

    Article  Google Scholar 

  30. Liu, M., Staum, J.: Stochastic kriging for efficient nested simulation of expected shortfall. J. Risk 12, 3–27 (2010)

    Article  Google Scholar 

  31. Longstaff, F.A., Schwartz, E.S.: Valuing American options by simulation: a simple least-squares approach. Rev. Financ. Stud. 14, 113–147 (2001)

    Article  Google Scholar 

  32. Makur, A., Zheng, L.: Polynomial spectral decomposition of conditional expectation operators. In: Proceedings of the 54th Annual Allerton Conference on Communication, Control, and Computing, pp. 633–640. IEEE, Monticello, IL (2016)

    Google Scholar 

  33. Moreno, M., Navas, J.F.: On the robustness of least-squares Monte Carlo (LSM) for pricing American derivatives. Rev. Deriv. Res. 14, 113–147 (2003)

    MATH  Google Scholar 

  34. Newey, W.K.: Convergence rates and asymptotic normality for series estimators. J. Econom. 79, 147–168 (1997)

    MathSciNet  Article  Google Scholar 

  35. Nikolić, Z., Jonen, C., Zhu, C.: Robust regression technique in LSMC proxy modelling. Aktuar 01(2017), 8–16 (2017)

    Google Scholar 

  36. Patarroyo, K.: A digression on Hermite polynomials (2019). Preprint (2019). Available online at https://arxiv.org/pdf/1901.01648.pdf

  37. Pelsser, A., Schweizer, J.: The difference between LSMC and replicating portfolio in insurance liability modelling. Eur. Actuar. J. 6, 441–494 (2016)

    MathSciNet  Article  Google Scholar 

  38. Retherford, J.: Hilbert Space: Compact Operators and the Trace Theorem. Cambridge University Press, Cambridge (1993)

    Book  Google Scholar 

  39. Risk, J., Ludkovski, M.: Sequential design and spatial modelling for portfolio tail risk measurement. SIAM J. Financ. Math. 9, 1137–1174 (2018)

    Article  Google Scholar 

  40. Stentoft, L.: Convergence of the least squares Monte Carlo approach to American option valuation. Manag. Sci. 50, 1193–1203 (2004)

    Article  Google Scholar 

  41. Tsitsiklis, J.N., Van Roy, B.: Regression methods for pricing complex American-style options. IEEE Trans. Neural Netw. 12, 694–763 (2001)

    Article  Google Scholar 

  42. Vinga, S.: Rényi continuous entropy of DNA sequences. J. Theor. Biol. 231, 377–388 (2004). Supplementary material, available at http://web.ist.utl.pt/susanavinga/renyi/, to Vinga, S., Almeida, J.S., 2004

    MathSciNet  Article  Google Scholar 

  43. Weber, S.: Distribution-invariant risk measures, entropy, and large deviations. J. Appl. Probab. 44, 16–40 (2007)

    MathSciNet  Article  Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Daniel Bauer.

Additional information

Publisher’s Note

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

This paper extends the earlier working paper Bauer et al. [4]. We thank Martin Schweizer (the Editor), an anonymous associate editor, two anonymous referees, Giuseppe Benedetti, Enrico Biffis, René Carmona, Patrick Cheridito, Matthias Fahrenwaldt, Jean-Pierre Fouque, Andreas Reuss, Daniela Singer, Ajay Subramanian, Baozhong Yang and seminar participants at the Bachelier Congress 2014, the 2015 WRIEC, ETH Zurich, Georgia State, Michigan State, St. Joseph’s University, TU Munich, UCSB, UIUC and Université de Montréal for helpful comments.

Appendix: Proofs and technical material

Appendix: Proofs and technical material

Proof of Lemma 2.1

1) Let \(A \in \mathcal{F}_{t}\), \(0 \leq t \leq \tau \). Then, by “taking out what is known”,

$$\begin{aligned} \tilde{\mathbb{P}}[A] =& {\mathbb{E} }^{\tilde{\mathbb{P}}}\left [ { \textbf{1}}_{A} \right ] = {\mathbb{E} }^{\mathbb{P}}\bigg[ \frac{d \tilde{\mathbb{P}}}{d {\mathbb{P}}}{\textbf{1}}_{A} \bigg] = { \mathbb{E} }^{\mathbb{P}}\bigg[ {\mathbb{E} }^{\mathbb{P}}\Big[ \frac{\frac{d {\mathbb{Q}}}{d {\mathbb{P}}}}{{\mathbb{E} }^{\mathbb{P}}[ \frac{d {\mathbb{Q}}}{d {\mathbb{P}}} | \mathcal{F}_{\tau} ] }{ \textbf{1}}_{A}\Big| \mathcal{F}_{\tau} \Big] \bigg] = {\mathbb{P}}[A]. \end{aligned}$$

2) Let \(X:\Omega \rightarrow \mathbb{R}\) be a random variable. Then similarly as above,

$$\begin{aligned} {\mathbb{E} }^{\tilde{\mathbb{P}}}[X | \mathcal{F}_{\tau} ] =& \frac{{\mathbb{E} }^{\mathbb{P}}[ \frac{d \tilde{\mathbb{P}}}{d {\mathbb{P}}} X | \mathcal{F}_{\tau} ] }{{\mathbb{E} }^{\mathbb{P}}[ \frac{d \tilde{\mathbb{P}}}{d {\mathbb{P}}} | \mathcal{F}_{\tau} ]} = {\mathbb{E} }^{\mathbb{P}}\bigg[ \frac{X \frac{d {\mathbb{Q}}}{d {\mathbb{P}}}}{{\mathbb{E} }^{\mathbb{P}}[ \frac{d {\mathbb{Q}}}{d {\mathbb{P}}} | \mathcal{F}_{\tau} ]} \bigg| \mathcal{F}_{\tau} \bigg] = {\mathbb{E} }^{{\mathbb{Q}}} [ X | \mathcal{F}_{\tau} ]. \end{aligned}$$

 □

Proof of Lemma 2.3

Linearity is obvious. For continuity, take \(h^{(n)} \rightarrow h\). Then

$$\begin{aligned} & {\mathbb{E} }^{\mathbb{P}}[ (\mathcal{V}\,h^{(n)} - \mathcal{V}\,h )^{2}] \\ &= {\mathbb{E} }^{\mathbb{P}}\bigg[ \sum _{s,t = \tau}^{T} {\mathbb{E} }^{ \tilde{\mathbb{P}}} [ (h^{(n)}_{s} - h_{s} ) (Y_{s}) | Y_{\tau} ] { \mathbb{E} }^{\tilde{\mathbb{P}}} [ (h^{(n)}_{t} - h_{t} ) (Y_{t}) | Y_{ \tau} ] \bigg] \\ &\leq \sum _{s,t = \tau}^{T} \sqrt{ {\mathbb{E} }^{\mathbb{P}}\big[ \big({\mathbb{E} }^{\tilde{\mathbb{P}}} [ (h^{(n)}_{s} - h_{s} ) (Y_{s}) | Y_{\tau} ] \big)^{2}\big]} \, \sqrt{ {\mathbb{E} }^{\mathbb{P}}\big[ \big({\mathbb{E} }^{\tilde{\mathbb{P}}} [ (h^{(n)}_{t} - h_{t} ) (Y_{t}) | Y_{\tau} ] \big)^{2}\big]} \\ &\leq \sum _{s,t = \tau}^{T} \sqrt{ {\mathbb{E} }^{\tilde{\mathbb{P}}} [ (h^{(n)}_{s} - h_{s} )^{2} (Y_{s}) ]} \, \sqrt{ {\mathbb{E} }^{ \tilde{\mathbb{P}}} [ (h^{(n)}_{t} - h_{t} )^{2} (Y_{t}) ]} \longrightarrow 0 \qquad \text{as $n\rightarrow \infty $}. \end{aligned}$$

 □

Proof of Lemma 2.4

Let \(\mathcal{V}^{(t)}\), \(t=\tau ,\dots ,T\), be the component mappings of \(\mathcal{V}\) introduced in (2.3), such that \(\mathcal{V}^{(t)}: L^{2}(\mathbb{R}^{d},\mathcal{B}, \tilde{\mathcal{P}}_{Y_{t}}) \rightarrow L^{2}(\mathbb{R}^{d}, \mathcal{B},\tilde{\mathcal{P}}_{Y_{\tau}})\) is the conditional expectation. Under the assumption that there exists a joint density, \(\mathcal{V}^{(t)}\) can be represented as

$$\begin{aligned} \mathcal{V}^{(t)}\,\mathbf{x}&=\int _{\mathbb{R}^{d}}\mathbf{x}(y)\, \pi _{Y_{t}|Y_{\tau}}(y|x)\,dy=\int _{\mathbb{R}^{d}}\mathbf{x}(y) \frac{\pi _{Y_{\tau},Y_{t}}(x,y)\,}{\pi _{Y_{\tau}}(x)}\,dy \\ &=\int _{\mathbb{R}^{d}}\mathbf{x}(y) \frac{\pi _{Y_{\tau},Y_{t}}(x,y)}{\pi _{Y_{t}}(y)\pi _{Y_{\tau}}(x)} \pi _{Y_{t}}(y)\,dy=\int _{\mathbb{R}^{d}}\mathbf{x}(y)\,k(x,y)\,\pi _{Y_{t}}(y) \,dy, \end{aligned}$$

where \(\mathbf{x} \in {L}^{2}(\mathbb{R}^{d},\mathcal{B},\tilde{\mathbb{P}}_{Y_{t}})\), \(\pi _{Y_{t}}(y)\) and \(\pi _{Y_{\tau}}(x)\) are the marginal density functions for \(Y_{t}\) and \(Y_{\tau}\) in \({L}^{2}(\mathbb{R}^{d},\mathcal{B},\tilde{\mathbb{P}}_{Y_{t}})\) and \({L}^{2}(\mathbb{R}^{d},\mathcal{B}, {\mathbb{P}}_{Y_{\tau}})\), and \(k(x,y)= \frac{\pi _{Y_{\tau},Y_{t}}(x,y)}{\pi _{Y_{t}}(y)\pi _{Y_{\tau}}(x)}\). Thus \(\mathcal{V}^{(t)}\) is an integral operator with kernel \(k(x,y)\), and

$$ \int _{\mathbb{R}^{d}}\int _{\mathbb{R}^{d}}|k(x,y)|^{2}\,\pi _{Y_{t}}(y) \pi _{Y_{\tau}}(x)\,dy\,dx= \int _{\mathbb{R}^{d}}\int _{\mathbb{R}^{d}} \pi _{Y_{t}|Y_{\tau}}(y|x)\, \pi _{Y_{\tau}|Y_{t}}(x|y)\,dy\,dx< \infty . $$

Thus \(\mathcal{V}^{(t)}\) is a Hilbert–Schmidt operator (e.g. Alt [2, Proposition 10.15]) and therefore compact. Finally, \(\mathcal{V}\) is the sum of the \(\mathcal{V}^{(t)}\) and therefore also compact. □

Proof of Proposition 3.1

\({\mathbb{P}}_{Y_{\tau}}\) is a finite Borel measure, and hence \({L}^{2}\left (\mathbb{R}^{d},\mathcal{B},{\mathbb{P}}_{Y_{\tau}} \right )\) is separable (see Cohn [16, Proposition 3.4.5]). Now if \(e_{k},\;k=1,\dots ,M\), are independent, we can find by Gram–Schmidt an orthonormal system \(S = \{f_{k}, k=1,\dots ,M\}\) with \({\mathrm{lin}}\{e_{k}, k=1,\dots ,M\} = {\mathrm{lin}}\, S\). For \(S\), in turn, we find an orthonormal basis \(\{f_{k}, k \in \mathbb{N}\} = S' \supseteq S\). Hence

$$ \widehat{C}_{\tau}^{(M)} = \sum _{k=1}^{M} \alpha _{k}\,e_{k} = \sum _{k=1}^{M} \underbrace{\tilde{\alpha}_{k}}_{\displaystyle{=\langle C_{\tau},f_{k} \rangle}}\,f_{k} \longrightarrow \sum _{k=1}^{\infty} \tilde{\alpha}_{k} \,f_{k} = C_{\tau} \qquad \text{as $M\rightarrow \infty $,} $$

where

$$ \| \widehat{C}_{\tau}^{(M)} - C_{\tau} \|^{2} = \sum _{k=M+1}^{\infty} | \langle C_{\tau},f_{k} \rangle |^{2} \longrightarrow 0 \qquad \text{as $M\rightarrow \infty $} $$

by Parseval’s identity. For the second part, we note that

$$ (\hat{\alpha}_{1}^{(N)},\dots ,\hat{\alpha}_{M}^{(N)})' = \hat{\alpha}^{(N)} = (A^{(M,N)} )^{-1} \frac{1}{N} \sum _{i=1}^{N} e (Y_{ \tau}^{(i)} )\,V_{\tau}^{(i)}, $$

where \(e(\, \cdot \, ) = (e_{1}(\, \cdot \, ),\dots , e_{M}(\, \cdot \, ))'\) and

$$ A^{(M,N)} = \bigg( \frac{1}{N} \sum _{i=1}^{N} e_{k}(Y_{\tau}^{(i)}) \,e_{\ell}(Y_{\tau}^{(i)}) \bigg)_{1\leq k,\ell \leq M} $$

is invertible for large enough \(N\) since we assume linear independence. Hence

$$ \hat{\alpha}^{(N)} \longrightarrow \alpha = (\alpha _{1},\dots , \alpha _{M})' = (A^{(M)} )^{-1} {\mathbb{E} }^{\tilde{\mathbb{P}}} \bigg[e (Y_{\tau} )\,\bigg(\sum _{t=\tau}^{T} \,x_{t} \bigg)\bigg] $$

\(\tilde{\mathbb{P}}\)-a.s. by the law of large numbers, where \(A^{(M)} = ({\mathbb{E} }^{\tilde{\mathbb{P}}} [e_{k} (Y_{\tau} )\,e_{\ell}(Y_{\tau} ) ] )_{1\leq k,\ell \leq M}\). Thus

$$ \widehat{C}_{\tau}^{(M,N)} = e'\,\hat{\alpha}^{(N)} \longrightarrow e' \alpha = \widehat{C}_{\tau}^{(M)} $$

\(\tilde{\mathbb{P}}\)-a.s. Finally, for the third part, write

$$\begin{aligned} V_{\tau}^{(i)}&=\sum _{t=\tau}^{T}x_{t} (Y_{t}^{(i)} )=\sum _{k=1}^{M} \alpha _{k}\,e_{k} (Y_{\tau}^{(i)} )+\epsilon _{j}, \end{aligned}$$

where \(\mathbb{E} [\epsilon _{j}|Y_{\tau} ]=0\), \({\mathrm{Var}}[\epsilon _{j}|Y_{ \tau} ]=\Sigma (Y_{\tau})\) and \({\mathrm{Cov}}[\epsilon _{i},\,\epsilon _{j}|Y_{\tau} ]=0\). Thus (see e.g. Amemiya [3, Chap. 1])

$$ \sqrt{N}(\alpha -\hat{\alpha}^{(N)}) \stackrel{\mathcal {D}}{\longrightarrow }{\mathcal{N}}\Big(0, \underbrace{ (A^{(M)} )^{-1}\big(\mathbb{E}^{\mathbb{P}} [e_{k}(Y_{\tau})e_{\ell}(Y_{\tau})\Sigma (Y_{\tau}) ]\big)_{1\le k, \,\ell \le M} (A^{(M)} )^{-1}}_{ \displaystyle{=\tilde{\xi}}}\Big) $$

so that

$$ \sqrt{N} (\widehat{C}_{\tau}^{(M)}-\widehat{C}_{\tau}^{(M,N)} )=e'( \alpha -\hat{\alpha}^{(N)})\sqrt{N} \stackrel{\mathcal {D}}{\longrightarrow} {\mathcal{N}}(0,\xi ^{(M)}), $$

where \( \xi ^{(M)}=e'\,\tilde{\xi}\,e \). □

Proof of Proposition 3.2

Since \((V_{\tau}^{(i)},Y_{\tau}^{(i)})\) are i.i.d. as Monte Carlo trials, the first part of Newey [34, Assumption 1] is automatically satisfied. The conditions in the proposition are then exactly Assumptions 1 (part 2), 2 and 3 in Newey [34] for \(d=0\). Thus the claim follows by the first part of Newey [34, Theorem 1]. □

Proof of Proposition 3.3

Analogously to the proof of Proposition 3.2, the first part of Newey [34, Assumption 1] is automatically satisfied. The conditions in the proposition are taken from the second part of Assumption 1, Assumption 8, the discussion following Assumption 8, and Assumption 9 in Newey [34]. Thus the claim follows by the first part of Newey [34, Theorem 4]. □

Proof of Proposition 4.2

We approximate \(\mathcal{V}\) by an arbitrary rank-\(M\) operator \(\mathcal{V}_{F}^{(M,e)}\), which can be written as \(\mathcal{V}_{F}^{(M,e)}=\sum _{k=1}^{M}\alpha _{k}\,\langle \,\cdot \,, u_{k}\rangle \,e_{k}\), where \((\alpha _{k})_{k=1}^{M} \subseteq \mathbb{R}_{+}\), \((u_{k})^{M}_{k=1}\) are orthonormal in ℋ and \((e_{k})_{k=1}^{M}\) are orthonormal in \({L}^{2} (\mathbb{R}^{d},\mathcal{B},{\mathbb{P}}_{Y_{\tau}} )\). Denote by \(\mathcal{V}_{F}^{(M,e^{*})}\) the operator we obtain when choosing \((\alpha _{k}, u_{k}, e_{k})=(\omega _{k}, s_{k}, \varphi _{k})\). Then

$$\begin{aligned} \inf _{\mathcal{V}_{F}^{(M,e)}}\|\mathcal{V}-\mathcal{V}_{F}^{(M,e)} \|^{2}&\le \sup _{\|\mathbf{x}\|=1}\|\mathcal{V}\mathbf{x}- \mathcal{V}_{F}^{(M,e^{*})}\mathbf{x}\|^{2} \\ &=\sup _{\|\mathbf{x}\|=1}\bigg\| \sum _{k=M+1}^{\infty}\omega _{k} \langle \mathbf{x}, s_{k}\rangle \varphi _{j}\bigg\| ^{2} \\ &=\sup _{\|\mathbf{x}\|=1}\sum _{k=M+1}^{\infty}\omega _{k}^{2}\, \langle \mathbf{x}, s_{k}\rangle ^{2} \; =\; \omega _{M+1}^{2}. \end{aligned}$$

On the other hand, consider any alternative system \((\alpha _{k}, u_{k}, e_{k})\) for an arbitrary finite-rank operator \(\mathcal{V}_{F}^{(M,e)}\). Then choose a non-zero \(\mathbf{x}_{0}\) such that

$$ \mathbf{x}_{0}\in {\mathrm{lin}}\{s_{1},\dots ,s_{M+1}\}\, \cap \, {\mathrm{lin}} \{u_{1},\dots ,u_{M}\}^{\bot} \neq \{0\}. $$

Note that \(\mathcal{V}-\mathcal{V}_{F}^{(M,e)}\) is compact and bounded. Therefore

$$\begin{aligned} \|\mathcal{V}-\mathcal{V}_{F}^{(M,e)}\|^{2}&\ge \frac{\|\mathcal{V}\mathbf{x}_{0}-\mathcal{V}_{F}^{(M,e)}\,\mathbf{x}_{0}\|^{2}}{\|\mathbf{x}_{0}\|^{2}} \\ &=\frac{\|\mathcal{V}\mathbf{x}_{0}\|^{2}}{\|\mathbf{x}_{0}\|^{2}}= \frac{\sum _{k=1}^{M+1}\omega _{k}^{2}|\langle \mathbf{x}_{0}, s_{k}\rangle |^{2}}{\sum _{k=1}^{M+1}|\langle \mathbf{x}_{0},s_{k}\rangle |^{2}} \\ & \ge \omega _{M+1}^{2}. \end{aligned}$$

Hence

$$ \inf _{\mathcal{V}_{F}^{(M,e)}}\|\mathcal{V}-\mathcal{V}_{F}^{(M,e)} \|^{2}=\omega _{M+1}^{2}=\|\mathcal{V}-\mathcal{V}_{F}^{(M,e^{*})}\|. $$

Now since \(\inf _{\mathcal{V}_{F}^{(M,e)}}\|\mathcal{V}-\mathcal{V}_{F}^{(M,e)} \|^{2}=\inf _{\{e_{1},\dots ,e_{M}\}}\|\mathcal{V}-P^{(M,e)} \mathcal{V}\|^{2}\), the claim follows by (4.1). □

Proof of Proposition 4.3

Proceeding as in (4.3) and with (4.1), we obtain

$$\begin{aligned} & \inf _{\alpha _{M}} \sup _{y\in \mathcal{Y}} \bigg|C_{\tau}(y)- \sum _{k=1}^{M}\alpha _{M,k}\,e_{k}(y)\bigg| \\ &\le \sup _{y\in \mathcal{Y}} \bigg|C_{\tau}(y)-\widehat{C}_{\tau}^{(M)}(y) \bigg| =\sup _{y\in \mathcal{Y}}\bigg|\sum _{k=M+1}^{\infty}\omega _{k} \,{\langle \mathbf{x}, s_{k}\rangle}\,\varphi _{k}(y)\bigg| \\ &\le \sum _{k=M+1}^{\infty}\omega _{k} \, \|\mathbf{x}\|\,\|s_{k}\|\, \sup _{y\in \mathcal{Y}}|\varphi _{k}(y)| =\sum _{k=M+1}^{\infty} \omega _{k}\,\|\mathbf{x}\|\, \sup _{y\in \mathcal{Y}}|\varphi _{k}(y)|=O \left (\omega _{M}\right ) \end{aligned}$$

for a fixed \(\mathbf{x}\) since the \(\varphi _{k}\) are uniformly bounded, where the second inequality follows from the triangle and Cauchy–Schwarz inequalities.

Then, going through the assumptions of Proposition 3.2 with the choices \(B=I\) and \(e^{(M)}=(e_{1},\dots ,e_{M})'\), we obtain

$$ \mathbb{E}^{\tilde{\mathbb{P}}} [\tilde{e}^{(M)}(Y_{\tau})\tilde{e}^{(M)}(Y_{\tau})' ]=I $$

due to the orthonormality of the singular functions. Therefore, the smallest eigenvalue is bounded away from zero uniformly for every \(M\). Moreover, for fixed \(y\in \mathcal{Y}\), \(|\tilde{e}^{(M)}(y)|=\sqrt{\varphi _{1}(y)^{2}+\cdots + \varphi _{M}(y)^{2}}\) so that

$$\begin{aligned} \sup _{y\in \mathcal{Y}}|\tilde{e}^{(M)}(y)| \le &\sqrt{\sum _{k=1}^{M} \,\sup _{y\in \mathcal{Y}}\varphi _{k}(y)^{2}}\le \sqrt{M \max _{1 \le k\le M} \sup _{y\in \mathcal{Y}}\varphi _{k}(y)^{2}}=C\sqrt{M}= \zeta _{0}(M) \end{aligned}$$

as the \(\varphi _{k}\) are uniformly bounded. Thus the claim follows by Proposition 3.2. □

Proof of Lemma 4.4

The assertions on the conditional distributions are standard. In order to show that \(\mathcal{V}\) is compact, we check that the transition and the reverse transition density functions satisfy the condition in Lemma 2.4. Note that the transition density function can be written as

$$\begin{aligned} \pi _{Y_{T}|Y_{\tau}}(y|x) &= g\big(y; \mu _{T}+\Gamma '\Sigma _{\tau}^{-1}(x- \mu _{\tau}), \Sigma _{T|\tau}\big) \\ &= \frac{1}{(2\pi )^{d/2}|\Sigma _{T|\tau}|^{1/2}} \frac{ |\Sigma _{\tau}(\Gamma ')^{-1}\Sigma _{T|\tau}\Gamma ^{-1}\Sigma _{\tau} |^{1/2}}{ |\Sigma _{\tau}(\Gamma ')^{-1}\Sigma _{T|\tau}\Gamma ^{-1}\Sigma _{\tau} |^{1/2}} \\ & \phantom{=:} \times \exp \bigg(-\frac{1}{2}\big(x-\mu _{\tau}-\Sigma _{\tau}( \Gamma ')^{-1}(y-\mu _{T})\big)'\Sigma _{\tau}^{-1}\Gamma \Sigma _{T| \tau}^{-1} \\ & \phantom{=\times \exp \Bigg(-\frac{1}{2}\big(} \times \Gamma '\Sigma _{\tau}^{-1}\big(x- \mu _{\tau}-\Sigma _{\tau}(\Gamma ')^{-1}(y-\mu _{T})\big)\bigg) \\ &= \frac{|\Sigma _{\tau}|}{|\Gamma |} g\big(x;\mu _{\tau}+\Sigma _{ \tau}(\Gamma ')^{-1}(y-\mu _{T}), \Sigma _{\tau}(\Gamma ')^{-1} \Sigma _{T|\tau}\Gamma ^{-1}\Sigma _{\tau}\big). \end{aligned}$$

We evaluate the integral

$$\begin{aligned} &\int _{\mathbb{R}^{d}}\pi _{Y_{T}|Y_{\tau}}(y|x)\pi _{Y_{\tau}|Y_{T}}(x|y)dx \\ &=\frac{|\Sigma _{\tau}|}{|\Gamma |} \int _{\mathbb{R}^{d}}g\big(x; \mu _{\tau}+\Sigma _{\tau}(\Gamma ')^{-1}(y-\mu _{T}), \Sigma _{\tau}( \Gamma ')^{-1}\Sigma _{T|\tau}\Gamma ^{-1}\Sigma _{\tau}\big)\, \\ & \phantom{=\frac{|\Sigma _{\tau}|}{|\Gamma |} \int _{\mathbb{R}^{d}}} \times g\big(x;\mu _{\tau}+\Gamma \Sigma _{\tau}^{-1}(y- \mu _{T}), \Sigma _{\tau |T}\big)\,dx \\ &=\frac{|\Sigma _{\tau}|}{|\Gamma |(2\pi )^{d/2}} \frac{1}{ |\Sigma _{\tau}(\Gamma ')^{-1}\Sigma _{T|\tau}\Gamma ^{-1}\Sigma _{\tau}+\Sigma _{\tau |T} |^{1/2}} \\ & \phantom{=:} \times \exp \bigg(-\frac{1}{2}(y-\mu _{T})'V^{-1}(y-\mu _{T})\bigg) \\ &=C_{1}\, g(y;\mu _{T}, V), \end{aligned}$$

where

$$\begin{aligned} V^{-1} =& (\Gamma ^{-1}\Sigma _{\tau}-\Sigma _{T}^{-1}\Gamma ' ) \big(\Sigma _{\tau}(\Gamma ')^{-1}\Sigma _{T|\tau}\Gamma ^{-1}\Sigma _{ \tau}+\Sigma _{\tau |T}\big)^{-1} \big(\Sigma _{\tau}(\Gamma ')^{-1}- \Gamma \Sigma _{T}^{-1}\big) \end{aligned}$$

and \(C_{1}\) is an appropriate constant to obtain \(g(y; \mu _{T}, V)\) (see the results from Vinga [42] on the product of Gaussian densities). Therefore,

$$ \int _{\mathbb{R}^{d}}\int _{\mathbb{R}^{d}}\pi _{Y_{T}|Y_{\tau}}(y|x) \pi _{Y_{\tau}|Y_{T}}(x|y)\,dx\,dy=\int _{\mathbb{R}^{d}}C_{1}g(y;\mu _{T}, V)\,dy=C_{1} < \infty . $$

 □

Proof of Lemma 4.5

The operator \(\mathcal{V}^{*}\) can be found via

$$\begin{aligned} {\langle \mathcal{V}h,m\rangle}_{\pi _{Y_{\tau}}}&=\int _{\mathbb{R}^{d}} (\mathcal{V}h)(x)\,m(x)\,\pi _{Y_{\tau}}(x)\,dx \\ &=\int _{\mathbb{R}^{d}} \bigg(\int _{\mathbb{R}^{d}} h(y)\pi _{Y_{T}|Y_{ \tau}}(y|x)\,dy\bigg)\,m(x)\,\pi _{Y_{\tau}}(x)\,dx \\ &=\int _{\mathbb{R}^{d}} h(y)\bigg(\int _{\mathbb{R}^{d}} m(x)\pi _{Y_{ \tau}|Y_{T}}(x|y)\,dx\bigg)\pi _{Y_{T}}(y)\,dy= {\langle h, \mathcal{V}^{*}m\rangle}_{\pi _{Y_{T}}}, \end{aligned}$$

where \((\mathcal{V}^{*}m)(y)=\int _{\mathbb{R}^{d}} m(x)\pi _{Y_{\tau}|Y_{T}}(x|y) \,dx\). We obtain for \(\mathcal{V}\mathcal{V}^{*}\) that

$$\begin{aligned} (\mathcal{VV^{*}}\varphi )(x) &=\int _{\mathbb{R}^{d}} (\mathcal{V}^{*} \varphi )(s)\pi _{Y_{T}|Y_{\tau}}(s|x)\,ds \\ & =\int _{\mathbb{R}^{d}}\bigg[\int _{\mathbb{R}^{d}} \varphi (y) \pi _{Y_{\tau}|Y_{T}}(y|s)\,dy \bigg]\pi _{Y_{T}|Y_{\tau}}(s|x)\,ds \\ &=\int _{\mathbb{R}^{d}} \varphi (y) \underbrace{\int _{\mathbb{R}^{d}} \pi _{Y_{\tau}|Y_{T}}(y|s)\pi _{Y_{T}|Y_{\tau}}(s|x)\,ds}_{ \displaystyle{=K_{A}(x,y)}}\,dy. \end{aligned}$$

It is useful to express the reverse density as in the proof of Lemma 4.4 as

$$\begin{aligned} &g(y; \mu _{Y_{\tau}|s}, \Sigma _{{\tau}|T})= \frac{|\Sigma _{T}|}{|\Gamma |}g\big(s; \mu _{T}+\Sigma _{T}\Gamma ^{-1}(y- \mu _{\tau}),\,\, \Sigma _{T}\Gamma ^{-1}\Sigma _{{\tau}|T}(\Gamma ')^{-1} \Sigma _{T}\big). \end{aligned}$$

Hence we get

$$\begin{aligned} K_{A}(x,y)&=\int _{\mathbb{R}^{d}} \pi _{Y_{\tau}|Y_{T}}(y|s) \pi _{Y_{T}|Y_{ \tau}}(s|x)\,ds \\ &= \frac{1}{(2\pi )^{d/2} |\Gamma \Sigma _{T}^{-1} (\Sigma _{T}\Gamma ^{-1}\Sigma _{{\tau}|T}(\Gamma ')^{-1}\Sigma _{T}+\Sigma _{T|{\tau}} )\Sigma _{T}^{-1}\Gamma ' |^{1/2}} \\ & \phantom{=:} \times \exp \bigg(-\frac{1}{2}\big(y-\mu _{\tau}-\Gamma \Sigma _{T}^{-1} \Gamma '\Sigma _{{\tau}}^{-1}(x-\mu _{{\tau}})\big)'(\Gamma ^{-1})' \Sigma _{T}\, \\ & \phantom{= \times \exp \bigg(-\frac{1}{2}\big(} \times \big(\Sigma _{T}\Gamma ^{-1}\Sigma _{{ \tau}|T}(\Gamma ')^{-1}\Sigma _{T}+\Sigma _{T|{\tau}}\big)^{-1} \Sigma _{T}\Gamma ^{-1} \\ & \phantom{= \times \exp \bigg(-\frac{1}{2}\big(} \times \big(y-\mu _{\tau}-\Gamma \Sigma _{T}^{-1} \Gamma '\Sigma _{{\tau}}^{-1}(x-\mu _{{\tau}})\big)\bigg) \\ &=g \big(y; \mu _{\tau}+ \underbrace{\Gamma \Sigma _{T}^{-1}\Gamma '\Sigma _{{\tau}}^{-1}}_{ \displaystyle{=A}}(x-\mu _{{\tau}}),\, \Sigma _{\tau}-\Gamma \Sigma _{T}^{-1} \Gamma '\Sigma _{\tau}^{-1}\Gamma \Sigma _{T}^{-1}\Gamma '\big) \\ &=g\big(y; \underbrace{\mu _{\tau}+A(x-\mu _{{\tau}})}_{ \displaystyle{=\mu _{A}(x)}},\, \underbrace{\Sigma _{\tau}-A\Sigma _{\tau} A'}_{\displaystyle{= \Sigma _{A}}}\big) =g\big(y; \mu _{A}(x), \Sigma _{A}\big), \end{aligned}$$

where we again rely on results from Vinga [42]. The formula for \(\mathcal{V}^{*}\mathcal{V}\) can be derived analogously. □

Proof of Lemma 4.6

We start by recalling the considerations from Khare and Zhou [28]. Let \((X_{t})\) with values in \(\mathbb{R}^{d}\) be an MAR(1) process satisfying the equation

$$ X_{t}=\Phi X_{t-1}+\eta _{t},\qquad t\ge 1, $$
(7.1)

where \(\Phi \in \mathbb{R}^{d\times d}\) and \((\eta _{t})_{t\ge 1}\) are i.i.d. \(\sim N(0, H)\). The process \((X_{t})\) has a unique stationary distribution \(N(0, \Sigma )\) if and only if \(H=\Sigma -\Phi \Sigma \Phi '\), and it is reversible if and only if \(\Phi \Sigma =\Sigma \Phi '\). [28] show that if these assumptions are satisfied, the transformed Markov operator for (7.1) has eigenvalues which are products of eigenvalues of \(\Phi \) and the corresponding eigenfunctions are products of Hermite polynomials.

Now note that for \(Y \sim K_{A}(x,\,\cdot \,) \), we can write

$$ Y-\mu _{\tau}=A(x-\mu _{\tau})+\zeta _{A}, $$

where \(\zeta _{A}\sim N(0, \Sigma _{A})\). Since from Lemma 4.5, we have \(\Sigma _{A}=\Sigma _{\tau}-A\Sigma _{\tau}A'\) and

$$ A\Sigma _{\tau}= \Gamma \Sigma _{T}^{-1}\Gamma ' = \Sigma _{\tau}\, A', $$

the operator \(\mathcal{V}\,\mathcal{V}^{*}\) has for \(\Sigma = \Sigma _{\tau}\) the same structure as the Markov operator for (7.1) which is reversible and stationary.

Following [28], denote by \(\Sigma _{\tau}^{1/2}\) the square root matrix of \(\Sigma _{\tau}\). Then

$$ \Sigma _{\tau}^{-1/2}A\Sigma _{\tau}^{1/2} = \Sigma _{\tau}^{-1/2}\, \Gamma \Sigma _{T}^{-1}\Gamma '\Sigma _{{\tau}}^{-1/2} $$

is symmetric and thus orthogonally diagonalisable as

$$ \Sigma _{\tau}^{-1/2} A \Sigma _{\tau}^{1/2} = P \Lambda P' $$

for an orthogonal matrix \(P\). Then

$$ A=(\Sigma _{\tau}^{1/2} P)\Lambda (P'\Sigma _{\tau}^{-1/2}), $$

and the entries of the diagonal matrix \(\Lambda \) are the eigenvalues of \(A\).

Now for the transformation \(z_{P}(Y)\) defined in (4.4), we obtain

$$\begin{aligned} \mathbb{E}_{K_{A}} [z^{P}(Y)|x ] =&P'\Sigma _{\tau}^{-1/2}A(x-\mu _{ \tau}) \\ =&P'\Sigma _{\tau}^{-1/2}\Sigma _{\tau}^{1/2}P\Lambda P'\Sigma _{ \tau}^{-1/2}(x-\mu _{\tau}) =\Lambda z^{P}(x) \end{aligned}$$

and

$$\begin{aligned} {\mathrm{Var}}_{K_{A}} [z^{P}(Y)|x ] =& P'\Sigma _{\tau}^{-1/2}\Sigma _{A} \Sigma _{\tau}^{-1/2}P \\ =& P'\Sigma _{\tau}^{-1/2}(\Sigma _{\tau}-A\Sigma _{\tau}A') \Sigma _{ \tau}^{-1/2}P =I-\Lambda ^{2}. \end{aligned}$$

Moreover,

$$\begin{aligned} \mathbb{E}_{\pi _{Y_{\tau}}} [z^{P}({Y_{\tau}}) ] =&P'\Sigma _{\tau}^{-1/2} \mathbb{E}_{\pi _{Y_{\tau}}}\left [Y_{\tau}-\mu _{\tau}\right ]=0, \\ {\mathrm{Var}}_{\pi _{Y_{\tau}}} [z^{P}({Y_{\tau}}) ] =& P'\Sigma _{\tau}^{-1/2} \Sigma _{\tau}\Sigma _{\tau}^{-1/2}P = I. \end{aligned}$$

The second part follows analogously. □

Proof of Proposition 4.7

For fixed \(z_{i}^{P}(Y)\), from Carrasco and Florens [13], the univariate orthonormal Hermite polynomial of order \(n_{i}\) is an eigenfunction under \(K_{A}\), so that

$$ \mathbb{E}_{K_{A}}\big[h_{n_{i}}\big(z_{i}^{P}(Y)\big)\big|x\big]= \lambda _{i}^{n_{i}} h_{n_{i}}\big(z_{i}^{P}(x)\big). $$

Moreover, the products of these polynomials are also eigenfunctions since

$$ \mathbb{E}_{K_{A}}\bigg[\prod _{i=1}^{d}h_{n_{i}}\big(z_{i}^{P}(Y) \big)\bigg|x\bigg]=\prod _{i=1}^{d}\mathbb{E}_{K_{A}}\big[h_{n_{i}} \big(z_{i}^{P}(Y)\big)\big|x\big]. $$

The orthogonality of the eigenfunctions is proved in Khare and Zhou [28]. Note that the product of normalised Hermite polynomials is already normalised since

$$\begin{aligned} \mathbb{E}_{\pi _{Y_{\tau}}}\bigg[\bigg(\prod _{i=1}^{d}h_{n_{i}}\big(z_{i}^{P}(Y) \big)\bigg)^{2}\bigg] =\prod _{i=1}^{d}\mathbb{E}_{\pi _{Y_{\tau}}} \big[h_{n_{i}}\big(z_{i}^{P}(Y)\big)^{2}\big]=1. \end{aligned}$$

The right singular functions are obtained similarly from \(z_{i}^{Q}(X)\). □

Proof of (5.5)

We have \(B_{r}(t,T)=\frac{1-e^{-\alpha (T-t)}}{\alpha}\), \(B_{\mu}(t,T)=\frac{e^{\kappa (T-t)}-1}{\kappa}\) and

$$\begin{aligned} A&\!(t,T) \\ =&\exp \bigg(\bar{\gamma}\big(B_{r}(t,T)-T+t\big) \\ &\hphantom{\exp \bigg(}{} +\frac{1}{2}\bigg( \frac{\sigma _{r}^{2}}{\alpha ^{2}}\Big(T-t-2B_{r}(t,T) + \frac{1-e^{-2\alpha (T-t)}}{2\alpha}\Big) \\ & \hphantom{\exp \bigg( +\frac{1}{2}\bigg(} +\frac{\delta ^{2}}{\kappa ^{2}}\Big(T-t-2B_{ \mu}(t,T)+\frac{e^{2\kappa (T-t)}-1}{2\kappa}\Big) \\ &\hphantom{\exp \bigg( +\frac{1}{2}\bigg(} + \frac{2\rho _{23}\sigma _{r}\delta}{\alpha \kappa}\Big(B_{\mu}(t,T)-T+t+B_{r}(t,T)- \frac{1-e^{-(\alpha -\kappa )(T-t)}}{\alpha -\kappa}\Big)\bigg)\bigg). \end{aligned}$$

 □

Proof of Lemma 5.1

Under ℙ, the solutions of (5.1)–(5.3) are

$$\begin{aligned} q_{\tau}&=q_{0}+\bigg(m-\frac{1}{2}\sigma _{S}^{2}\bigg)\tau +\sigma _{S} \int _{0}^{\tau}\,dW_{s}^{S}, \\ r_{\tau}&=r_{0}e^{-\alpha \tau}+\gamma (1-e^{-\alpha \tau} )+\sigma _{r} \int _{0}^{\tau}e^{-\alpha (\tau -t)}dW_{t}^{r}, \\ \mu _{x+\tau}&=\mu _{x} e^{\kappa \tau}+\delta \int _{0}^{\tau}e^{ \kappa (\tau -u)}dW_{u}^{\mu}. \end{aligned}$$

Thus \(Y_{\tau}= (q_{\tau},r_{\tau},\mu _{x+\tau})'\) is distributed according to \({\mathcal{N}}(\mu _{\tau},\Sigma _{\tau})\) with

μ τ = ( q 0 + ( m 1 2 σ S 2 ) τ r 0 e α τ + γ ( 1 e α τ ) μ x e κ τ ) , Σ τ = ( σ S 2 τ ρ 12 σ S σ r B r ( 0 , τ ) ρ 13 σ S δ B μ ( 0 , τ ) ρ 12 σ S σ r B r ( 0 , τ ) σ r 2 1 e 2 α τ 2 α ρ 23 σ r δ 1 e ( α κ ) τ α κ ρ 13 σ S δ B μ ( 0 , τ ) ρ 23 σ r δ 1 e ( α κ ) τ α κ δ 2 e 2 κ τ 1 2 κ ) .

To derive the distribution under \(\mathbb{Q}_{E}\), first note that for \(\tau \le s< T\),

$$\begin{aligned} r_{s}&=e^{-\alpha (s-\tau )}r_{\tau}+\bigg(\bar{\gamma} - \frac{\sigma _{r}^{2}}{\alpha ^{2}}\bigg) (1-e^{-\alpha (s-\tau )} ) \\ & \phantom{=:} +\frac{\sigma _{r}^{2}}{2\alpha ^{2}} (e^{-\alpha (T-s)}-e^{-\alpha (T+s-2 \tau )} ) \\ & \phantom{=:} -\frac{\rho _{23}\sigma _{r}\delta}{\kappa} \bigg( \frac{e^{\kappa (T-s)}-e^{-\alpha (s-\tau )+\kappa (T-\tau )}}{\alpha -\kappa}- \frac{1-e^{-\alpha (s-\tau )}}{\alpha}\bigg) \\ & \phantom{=:} +\sigma _{r}\int _{\tau}^{s}e^{-\alpha (s-y)}dZ_{y}^{r}, \end{aligned}$$

so that \(\int _{\tau}^{T} r_{s}\,ds\) can be evaluated, using the stochastic Fubini theorem, as

$$\begin{aligned} \int _{\tau}^{T} r_{s} ds &=\frac{1-e^{-\alpha (T-\tau )}}{\alpha} r_{\tau}+\bigg(\bar{\gamma}-\frac{\sigma _{r}^{2}}{\alpha ^{2}}\bigg) \bigg(T-\tau -\frac{1-e^{-\alpha (T-\tau )}}{\alpha}\bigg) \\ & \phantom{=:} +\frac{\sigma _{r}^{2}}{2\alpha ^{2}}\bigg( \frac{1-e^{-\alpha (T-\tau )}}{\alpha}- \frac{e^{-\alpha (T-\tau )}-e^{-2\alpha (T-\tau )}}{\alpha}\bigg) \\ & \phantom{=:} -\frac{\rho _{23}\sigma _{r}\delta}{\kappa}\bigg( \frac{e^{\kappa (T-\tau )}-1}{\kappa (\alpha -\kappa )}- \frac{e^{\kappa (T-\tau )}-e^{-(\alpha -\kappa )(T-\tau )}}{\alpha (\alpha -\kappa )} \\ & \phantom{=:} \quad \qquad \qquad -\frac{1}{\alpha}\Big(T-\tau - \frac{1-e^{-\alpha (T-\tau )}}{\alpha}\Big)\bigg) \\ & \phantom{=:} +\sigma _{r}\int _{\tau}^{T}\frac{1-e^{-\alpha (T-y)}}{\alpha}dZ_{y}^{r}. \end{aligned}$$

Thus under \(\mathbb{Q}_{E}\) with known \(Y_{\tau}\), the solutions of (5.6)–(5.8) are

$$\begin{aligned} q_{T}&= \mu _{q_{T}|q_{\tau}} +\sigma _{S}\int _{\tau}^{T}dZ_{s}^{S}+ \sigma _{r}\int _{\tau}^{T}\frac{1-e^{-\alpha (T-y)}}{\alpha}dZ_{y}^{r}, \\ r_{T}&=\mu _{r_{T}|r_{\tau}} +\sigma _{r}\int _{\tau}^{T}e^{-\alpha (T-y)}dZ_{y}^{r}, \\ \mu _{x+T}&=\mu _{\mu _{x+T}|\mu _{x+\tau}}+\delta \int _{\tau}^{T}e^{ \kappa (T-t)}dZ_{t}^{\mu}, \end{aligned}$$

so that the conditional distribution of \(Y_{T}|Y_{\tau}\) is Gaussian with parameters

$$\begin{aligned} \mu _{r_{T}|r_{\tau}}& =e^{-\alpha (T-\tau )}r_{\tau}+\bigg(\bar{\gamma}- \frac{\sigma _{r}^{2}}{\alpha ^{2}}\bigg) (1-e^{-\alpha (T-\tau )} ) \\ & \phantom{=:} +\frac{\sigma _{r}^{2}}{2\alpha ^{2}} (1-e^{-2\alpha (T-\tau )} ), \\ \sigma _{r_{T}|r_{\tau}}^{2}&= \sigma _{r}^{2} \frac{1-e^{-2\alpha (T-\tau )}}{2\alpha} \\ & \phantom{=:} -\frac{\rho _{23}\sigma _{r}\delta}{\kappa}\bigg( \frac{1-e^{-(\alpha -\kappa )(T-\tau )}}{\alpha -\kappa}- \frac{1-e^{-\alpha (T-\tau )}}{\alpha}\bigg), \\ \mu _{\mu _{x+T}|\mu _{x+\tau}}&= \mu _{x+\tau}e^{\kappa (T-\tau )}- \frac{\rho _{23}\sigma _{r}\delta}{\alpha}\bigg( \frac{e^{\kappa (T-\tau )}-1}{\kappa}- \frac{1-e^{-(\alpha -\kappa )(T-\tau )}}{\alpha -\kappa}\bigg), \\ \sigma _{\mu _{x+T}|\mu _{x+\tau}}^{2}&= \delta ^{2} \frac{e^{2\kappa (T-\tau )}-1}{2\kappa}, \\ \sigma _{r_{T}, \mu _{x+T}|r_{\tau}, \mu _{x+\tau}}& = \rho _{23} \sigma _{r}\delta \frac{1-e^{-(\alpha -\kappa )(T-\tau )}}{\alpha -\kappa}, \\ \mu _{q_{T}|q_{\tau}} &= q_{\tau}+B_{r}(\tau ,T)r_{\tau}+\bigg( \bar{\gamma}-\frac{\sigma _{r}^{2}}{\alpha ^{2}}\bigg)\bigg(T-\tau - \frac{1-e^{-\alpha (T-\tau )}}{\alpha}\bigg) \\ & \phantom{=:} +\frac{\sigma _{r}^{2}}{2\alpha ^{2}}\bigg( \frac{1-e^{-\alpha (T-\tau )}}{\alpha}- \frac{e^{-\alpha (T-\tau )}-e^{-2\alpha (T-\tau )}}{\alpha}\bigg) \\ & \phantom{=:} -\frac{\rho _{23}\sigma _{r}\delta}{\kappa}\bigg( \frac{e^{\kappa (T-\tau )}-1}{\kappa (\alpha -\kappa )}- \frac{e^{\kappa (T-\tau )}-e^{-(\alpha -\kappa )(T-\tau )}}{\alpha (\alpha -\kappa )} \\ & \phantom{=} \quad \qquad \qquad -\frac{1}{\alpha}\Big(T-\tau - \frac{1-e^{-\alpha (T-\tau )}}{\alpha}\Big)\bigg)-\frac{1}{2}\sigma _{S}^{2}(T- \tau ) \\ & \phantom{=:} -\frac{\rho _{12}\sigma _{S}\sigma _{r}}{\alpha}\bigg(T-\tau - \frac{1-e^{-\alpha (T-\tau )}}{\alpha}\bigg) \\ & \phantom{=:} -\frac{\rho _{13}\sigma _{S}\delta}{\kappa}\bigg( \frac{e^{\kappa (T-\tau )}-1}{\kappa}-T+\tau \bigg) \\ & \phantom{=:} -\frac{\delta ^{2}}{\kappa}\bigg( \frac{e^{2\kappa (T-\tau )}-1}{2\kappa}- \frac{e^{\kappa (T-\tau )}-1}{\kappa}\bigg), \\ \sigma _{q_{T}|q_{\tau}}^{2}&=\sigma _{S}^{2}(T-\tau )+ \frac{\sigma _{r}^{2}}{\alpha ^{2}}\bigg(T-\tau -2 \frac{1-e^{-\alpha (T-\tau )}}{\alpha}+ \frac{1-e^{-2\alpha (T-\tau )}}{2\alpha}\bigg) \\ & \phantom{=:} +\frac{2\rho _{12}\sigma _{S}\sigma _{r}}{\alpha}\bigg(T-\tau - \frac{1-e^{-\alpha (T-\tau )}}{\alpha}\bigg), \\ \sigma _{q_{T}, r_{T}|q_{\tau}, r_{\tau}} &=\rho _{12}\sigma _{S}\sigma _{r} \frac{1-e^{-\alpha (T-\tau )}}{\alpha}+\frac{\sigma _{r}^{2}}{\alpha} \frac{1-2e^{-\alpha (T-\tau )}+e^{-2\alpha (T-\tau )}}{2\alpha}, \\ \sigma _{q_{T}, \mu _{x+T}|q_{\tau}, \mu _{x+\tau}}& = \rho _{13} \sigma _{S}\delta \frac{e^{\kappa (T-\tau )}-1}{\kappa} \\ & \phantom{=:} +\frac{\rho _{23}\sigma _{r}\delta}{\alpha}\bigg( \frac{e^{\kappa (T-\tau )}-1}{\kappa}- \frac{1-e^{-(\alpha -\kappa )(T-\tau )}}{\alpha -\kappa}\bigg). \end{aligned}$$

We can write the conditional mean of \(Y_{T}\) given \(Y_{\tau}\) in the affine form

( μ q T | q τ μ r T | r τ μ μ x + T | μ x + τ ) = ( 1 1 e α ( T τ ) α 0 0 e α ( T τ ) 0 0 0 e κ ( T τ ) ) = H ( q τ r τ μ x + τ ) + C τ =H Y τ + C τ ,

where \(C_{\tau}\) is a constant matrix defined by the remaining terms of the mean vector of \(Y_{T}|Y_{\tau}\) after defining \(HY_{\tau}\). The unconditional distribution of \(Y_{T}\) under \(\tilde{\mathbb{P}}\) is also Gaussian since \(Y_{\tau}\) and \(Y_{T}|Y_{\tau}\) follow Gaussian distributions. Thus it suffices to specify the mean vector and the covariance matrix of \(Y_{T}\) under \(\tilde{\mathbb{P}}\) to specify its distribution, and we have

$$\begin{aligned} \mu _{T}=\mathbb{E}^{\tilde{\mathbb{P}}}[Y_{T}]&=\mathbb{E}^{ \mathbb{P}}\big[\mathbb{E}^{\mathbb{Q}_{E}}[Y_{T}|Y_{\tau}]\big]= \mathbb{E}^{\mathbb{P}} [HY_{\tau}+C_{\tau} ]=H\mu _{\tau}+C_{\tau}, \\ \Sigma _{T} = {\mathrm{Cov}}^{\tilde{\mathbb{P}}}[Y_{T}]&={\mathrm{Cov}}^{ \mathbb{P}}\big[\mathbb{E}^{\mathbb{Q}_{E}}[Y_{T}|Y_{\tau}]\big]+ \mathbb{E}^{\mathbb{P}}\big[{\mathrm{Cov}}^{\mathbb{Q}_{E}}[Y_{T}|Y_{\tau}] \big] \\ &={\mathrm{Cov}}^{\mathbb{P}}\left [HY_{\tau}+C_{\tau}\right ]+\mathbb{E}^{ \mathbb{P}} [\Sigma _{T|\tau} ] =H\Sigma _{{\tau}}H'+\Sigma _{T|\tau}. \end{aligned}$$

The final step is to specify the joint distribution of \(Y_{\tau}\) and \(Y_{T}\) by finding \({\mathrm{Cov}}(Y_{\tau}, Y_{T})\). Note that

$$\begin{aligned} \Gamma = {\mathrm{Cov}}(Y_{\tau}, Y_{T})&=\mathbb{E}^{\tilde{\mathbb{P}}}[Y_{\tau}Y_{T}']-\mathbb{E}^{\tilde{\mathbb{P}}}[Y_{\tau}] \,\mathbb{E}^{ \tilde{\mathbb{P}}}[Y_{T}']=\mathbb{E}^{\mathbb{P}}\big[\mathbb{E}^{ \mathbb{Q}_{E}}[Y_{\tau}Y_{T}'|Y_{\tau}]\big]-\mu _{\tau}\mu _{T}' \\ &=\mathbb{E}^{\mathbb{P}} [Y_{\tau}(Y_{\tau}'H'+C_{\tau}') ]-\mu _{\tau} \mu _{T}'=\Sigma _{\tau}H'. \end{aligned}$$

 □

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Ha, H., Bauer, D. A least-squares Monte Carlo approach to the estimation of enterprise risk. Finance Stoch 26, 417–459 (2022). https://doi.org/10.1007/s00780-022-00478-7

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00780-022-00478-7

Keywords

  • Risk management
  • Least-squares Monte Carlo
  • Basis functions

Mathematics Subject Classification (2010)

  • 60J22
  • 91G60
  • 33C50

JEL Classification

  • C63
  • G22
  • G32