Journal of Applied Mathematics
Volume 2010 (2010), Article ID 263451, 5 pages
doi:10.1155/2010/263451
Research Article

An Analytic Solution for a Vasicek Interest Rate Convertible Bond Model

1Department of Applied Mathematics, University of Western Ontario, London, ON, Canada N6A 5B7
2Department of Statistical & Actuarial Sciences, University of Western Ontario, London, ON, Canada N6A 5B7

Received 31 May 2009; Revised 5 November 2009; Accepted 6 January 2010

Academic Editor: Peter Spreij

Copyright © 2010 A. S. Deakin and Matt Davison. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

This paper provides the analytic solution to the partial differential equation for the value of a convertible bond. The equation assumes a Vasicek model for the interest rate and a geometric Brownian motion model for the stock price. The solution is obtained using integral transforms.


This work corrects errors in the original paper by Mallier and Deakin [1] on the Green's function for the Vasicek convertible bond equation. One error involves subtle points of the inverse Laplace transform. We show that the solution of

𝜕 𝑉 = 1 𝜕 𝜏 2 𝜎 2 𝑆 2 𝜕 2 𝑉 𝜕 𝑆 2 𝜕 + 𝜌 𝜎 𝑐 𝑆 2 𝑉 + 1 𝜕 𝑆 𝜕 𝑟 2 𝑐 2 𝜕 2 𝑉 𝜕 𝑟 2 + 𝑟 𝑆 𝜕 𝑉 𝜕 𝑆 + ( 𝑎 𝑏 𝑟 ) 𝜕 𝑉 𝜕 𝑟 𝑟 𝑉 ( 1 ) in the log stock variables 𝑥 = l o g 𝑆 and 𝑆 ̃ 𝑥 = l o g is

𝑉 ( 𝑆 , 𝑟 , 𝜏 ) = 𝑉 0 𝑒 ̃ 𝑥 , ̃ 𝑟 𝐺 ( 𝑟 , ̃ 𝑟 , 𝑥 ̃ 𝑥 ) 𝑑 ̃ 𝑟 𝑑 ̃ 𝑥 , ( 2 ) where 𝑉 = 𝑉 0 ( 𝑆 , 𝑟 ) at 𝜏 = 0 and the Green's function (GF) is

𝐺 ( 𝑟 , ̃ 𝑟 , 𝑥 ̃ 𝑥 ) = e x p ( 𝐹 ) 𝑁 ( 𝑤 , Ξ ) 𝑁 ( 𝛼 , Φ ) . ( 3 ) The normal distribution with variance 𝑤 and argument Ξ is here denoted by

𝑁 ( 𝑤 , Ξ ) = ( 2 𝜋 𝑤 ) 1 / 2 Ξ e x p 2 ( 2 𝑤 ) , ( 4 ) and the coefficients are

𝑤 = 1 𝑒 2 𝑏 𝜏 𝑐 2 2 𝑏 , Ξ = ̃ 𝑟 𝑟 𝑒 𝑏 𝜏 𝐵 𝑎 𝐵 𝑐 2 2 , ( 5 ) 𝐹 = 𝐴 𝐵 𝑟 , Φ = ̃ 𝑥 𝑥 𝐷 Ξ ( 2 𝜌 𝜎 / 𝑐 + 𝐵 ) 1 + 𝑒 𝑏 𝜏 , ( 6 ) 𝛼 = 𝜏 𝜎 2 1 𝜌 2 + 𝑐 𝑏 + 𝜌 𝜎 2 2 𝜏 𝑏 t a n h 𝑏 𝜏 2 𝑣 , ( 7 ) 𝐹 + 𝐷 + 2 = 0 , 𝐴 = ( 𝐵 𝜏 ) 2 𝑎 𝑏 𝑐 2 2 𝑏 2 𝑐 2 𝐵 2 4 𝑏 , ( 8 ) 𝐵 = 1 𝑒 𝑏 𝜏 𝑏 , 𝑣 = 𝜏 𝜎 2 + ( 𝜏 𝐵 ) ( 2 𝜌 𝜎 𝑏 + 𝑐 ) 𝑐 𝑏 2 𝑐 2 𝐵 2 . 2 𝑏 ( 9 ) In the case of the convertible bond, the initial condition 𝑉 0 in (2) is independent of ̃ 𝑟 . Integrating (2) in ̃ 𝑟 , we obtain the simpler Green's function

𝐺 ( 𝑟 , 𝜏 , 𝑥 ̃ 𝑥 ) = e x p ( 𝐹 ( 𝑟 , 𝜏 ) ) 𝑁 ( 𝑣 ( 𝜏 ) , ̃ 𝑥 𝑥 𝐷 ( 𝑟 , 𝜏 ) ) . ( 1 0 ) The parameters in the solution have the range of values: 𝜎 > 0 , 𝑐 > 0 , | 𝜌 | < 1 , while 𝑎 and 𝑏 are arbitrary since the solutions are analytic in 𝑎 and 𝑏 .

To prove (3), we assume 𝑉 to be bounded as 𝑆 0 and 𝑆 𝑐 0 𝑉 , where 𝑐 𝑜 is a positive constant, is bounded as 𝑆 so that the Mellin transform of 𝑉 exists. Once the solution is determined, the initial condition may be extended to include the more general case where the integral (2) exists (e.g., 𝑉 0 = m a x ( 𝑆 , 1 ) ). In the derivation of the solution, the condition 𝑏 > 0 is assumed in (1).

To solve for 𝑉 in (1), the Mellin and Laplace transform 𝑉 ( 𝑝 ) = [ 𝑉 ] and 𝑉 ( 𝑧 ) = [ 𝑉 ] (equations ( 2 . 6 ) , ( 2 . 7 ) in [1]) are applied to obtain the ODE

𝑐 2 2 𝑉 𝑟 𝑟 + ( 𝑎 𝜌 𝑐 𝜎 𝑝 𝑏 𝑟 ) 𝑉 𝑟 + 2 1 𝜎 2 𝑝 𝑟 ( 1 + 𝑝 ) 𝑧 𝑉 𝑉 = 0 ( 𝑆 , 𝑟 ) . ( 1 1 ) The generalhomogeneous solution ([2, 3] Section V.I, page 249) of (11) is

𝑉 = e x p ( 1 + 𝑝 ) 𝑟 𝑏 𝜈 2 , 1 2 , 𝑢 2 2 𝑧 , ( 1 2 ) 𝜈 = 𝑏 + 2 𝐸 , 𝑢 ( 𝑟 ) = 2 𝑏 3 𝑐 2 𝑟 𝑏 2 𝑎 𝑏 + 𝑐 2 + 𝑝 𝑐 𝑏 𝜎 𝜌 + 𝑐 2 , ( 1 3 ) 𝐸 = ( 1 + 𝑝 ) 2 𝑎 𝑏 𝑐 2 𝑏 𝑝 Λ 3 4 , Λ = ( 𝑐 + 𝑏 𝜎 𝜌 ) 2 + ( 𝑏 𝜎 ) 2 1 𝜌 2 , ( 1 4 ) and is the general solution of the confluent hypergeometric equation ([2, 3] Section V.I). The general solution (12) in terms of the parabolic cylinder function 𝐷 𝜈 ( 𝑢 ) ([2, 3] Section V.II, page 117), with arbitrary constants 𝐶 1 and 𝐶 2 ( 𝜈 0 , 1 , ), is

𝑉 = e x p ( 1 + 𝑝 ) 𝑟 𝑏 2 𝜈 / 2 𝑒 𝑢 2 / 4 𝐶 1 𝐷 𝜈 ( 𝑢 ) + 𝐶 2 𝐷 𝜈 ( 𝑢 ) . ( 1 5 ) Replacing [ 𝑉 0 ( 𝑆 , 𝑟 ) ] in (11) by the delta function 𝛿 ( 𝑟 ̃ 𝑟 ) (c.f., (20) for details), the GF for (11) has the form

𝐺 1 ( 𝑟 , ̃ 𝑟 ) = 2 𝑐 2 1 ( 𝑟 ) 2 ( ̃ 𝑟 ) 𝑊 1 1 ( ̃ 𝑟 ) , 2 ( ̃ 𝑟 ) , 𝑟 > ̃ 𝑟 , ( 1 6 ) where 𝑗 are appropriate homogeneous solutions in (15), 𝑊 is the Wronskian, and 𝐺 1 for 𝑟 < ̃ 𝑟 is defined by interchanging 𝑟 and ̃ 𝑟 in 𝑗 , but not in 𝑊 .

For the existence and the evaluation of the inverse Laplace transform (ILT) of 𝐺 1 , the asymptotic expansion, valid for large ( 𝜈 ) in the sector | a r g ( 𝜈 ) | < 𝜋 ,

Γ ( 𝜈 ) 𝐷 𝜈 ( 𝑣 ( 𝑟 ) ) 𝐷 𝜈 ( 𝑤 ( ̃ 𝑟 ) ) 𝜈 2 𝜋 1 / 2 e x p ( 𝜈 ) 1 / 2 ( 𝑣 ( 𝑟 ) 𝑤 ( ̃ 𝑟 ) ) ( 1 7 ) is required where 𝑣 ( 𝑟 ) = ± 𝑢 ( 𝑟 ) and 𝑤 ( ̃ 𝑟 ) = ± 𝑢 ( ̃ 𝑟 ) . The expansion for the Gamma function is given in ([2, 3] Section V.I, page 47). The expansion with a restricted domain for the parabolic cylinder function appears in [2, 3] (Section V.I, page 249 ( 8 ) ) and the general case is proved by applying the Method of Steepest Descent to the integral representation ([4, 5], page 349). The solutions 𝑖 in (16) must be chosen such that 𝐺 1 has an ILT that exists for all 𝑟 and ̃ 𝑟 . For the general case, we define 𝑖 in (15) by replacing 𝐶 𝑗 by 𝐶 𝑖 𝑗 . There are four terms in (16), only one for which the ILT exists: 𝐶 1 2 = 𝐶 2 1 = 0 , 𝑣 𝑤 = ( 2 𝑏 ) 1 / 2 | 𝑟 ̃ 𝑟 | / 𝑐 in (17). Thus,

𝐺 1 = 𝑔 1 ( 𝑟 ) 𝑔 2 ( ̃ 𝑟 ) 𝑐 1 ( 𝑏 𝜋 ) 1 / 2 Γ ( 𝜈 ) 𝐷 𝜈 ( 𝑢 ( 𝑟 ) ) 𝐷 𝜈 ( 𝑢 ( ̃ 𝑟 ) ) , 𝑟 > ̃ 𝑟 , ( 1 8 ) where 𝑔 𝑗 ( 𝑟 ) = e x p [ ( 1 ) 𝑗 ( ( 1 + 𝑝 ) 𝑟 / 𝑏 𝑢 2 ( 𝑟 ) / 4 ) ] . For 𝑟 < ̃ 𝑟 , 𝐺 1 is defined by interchanging 𝑟 and ̃ 𝑟 in 𝐷 𝜈 . However, to explain the results in [1], we compare ( 2 . 1 6 ) to ( 1 6 , 2 0 ) so that 1 𝒱 2 and 2 𝒱 1 in ( 2 . 1 3 ) (change sign on RHS of ( 2 . 1 4 ) , ( 2 . 1 6 ) ). Consequently, 1 and 2 are defined in (15) by taking ( 𝐶 1 = 0 , 𝐶 2 = 1 ) and ( 𝐶 1 = 1 , 𝐶 2 = 1 ) , respectively. The modified GF is 𝐺 𝑚 1 = 𝐺 1 + 𝐺 𝑠 1 where 𝐺 1 and 𝐺 𝑠 1 are defined from 𝐺 1 by changing 𝑢 to 𝑢 and 𝑢 ( 𝑟 ) to 𝑢 ( 𝑟 ) , respectively.

As outlined in [1], the ILT 𝐺 2 = 1 ( 𝐺 1 ) ( ( 2 . 1 7 ) , [1]) is equal to the contributions from the simple poles of Γ ( 𝜈 ) at 𝜈 = 𝑛 ( 𝑛 = 0 , 1 ). 𝐺 2 is equal to a sum involving Hermite polynomials ([2, 3] Section V.II, page 194 ( 2 2 ) ) so that

𝐺 2 = 𝑁 ( 𝜂 , ̃ 𝑟 𝑟 ) e x p 2 𝑏 4 𝑐 ( 𝑟 ̃ 𝑟 ) 𝑠 1 𝑠 2 𝜆 𝑏 𝜏 8 + 𝑏 𝜏 2 2 𝑏 𝐸 𝜏 ( 1 + 𝑝 ) 𝑏 ( 𝑟 ̃ 𝑟 ) , ( 1 9 ) where 𝑠 𝑚 = 𝑢 𝑚 ( 𝑟 ) + 𝑢 𝑚 ( ̃ 𝑟 ) , 𝜂 = 𝜏 𝑐 2 s i n h ( 𝑏 𝜏 ) / ( 𝑏 𝜏 ) , 𝜆 = ( 2 / ( 𝑏 𝜏 ) ) t a n h ( 𝑏 𝜏 / 2 ) . The semicircle's contribution to 𝐺 2 goes to zero as the radius goes to infinity follows from the approximation of 𝐺 1 in (18) via (17). For the modified GF, 𝐺 𝑚 2 = 𝐺 2 + 𝐺 𝑠 2 where 𝐺 2 and 𝐺 𝑠 2 are formally defined by the contributions from the poles: 𝐺 2 = 𝐺 2 , 𝐺 𝑠 2 = 𝐺 2 e x p ( 𝑢 ( 𝑟 ) 𝑢 ( ̃ 𝑟 ) / s i n h ( 𝑏 𝜏 ) ) .

The last step is to evaluate the inverse Mellin transforms (IMT; ( 2 . 1 8 ) , [1]) 𝐺 3 = 1 𝐺 2 and, for the modified GF, 𝐺 𝑚 3 = 𝐺 3 + 𝐺 𝑠 3 , where 𝐺 𝑠 3 = 1 𝐺 𝑠 2 . To do this, the argument of the exponential in 𝐺 2 and 𝐺 𝑠 2 is expressed in the form 𝛼 𝑝 2 / 2 + 𝛽 𝑝 + 𝛾 , and formula ( 2 . 2 9 ) in [1] is applied. For 𝐺 2 , 𝛼 is given by (7). For 𝐺 𝑠 2 , 𝛼 = 𝛼 𝑠 is given by (7) where t a n h is replaced by c o t h . Correcting the error in [1] (page 228, L.-4, (+) to ( ) ), then 2 𝛼 + = 𝛼 and 2 𝛼 = 𝛼 𝑠 , where 𝛼 ± appear in ( 2 . 2 7 ) and ( 2 . 3 3 ) . Assuming that ( 𝑐 / 𝑏 + 𝜌 𝜎 ) 0 , then there is a positive number 𝜏 𝑜 such that 𝛼 < 0 for 0 < 𝜏 < 𝜏 𝑜 . Thus the IMT of 𝐺 𝑠 2 does not exist for 0 < 𝜏 < 𝜏 𝑜 , and 𝐺 1 in (18) is the correct Green's function. For 𝐺 3 , we have 𝐺 3 = e x p 𝛾 𝑁 ( 𝜂 , ̃ 𝑟 𝑟 ) 𝑁 ( 𝛼 , 𝛽 l o g 𝑆 ) . The variables ( 𝑉 , 𝑉 0 , 𝐺 1 ) and ( 𝑉 , 𝑉 0 , 𝐺 3 ) are connected by

𝑉 = 𝑉 0 𝐺 ( 𝑆 , ̃ 𝑟 ) 1 𝑑 ̃ 𝑟 , 𝑉 = 1 𝑉 0 𝐺 3 𝑑 ̃ 𝑟 . ( 2 0 ) Using the convolution theorem ( ( 2 . 3 0 ) , [1]), the solution is (2), where

𝜎 𝐺 ( 𝑟 , ̃ 𝑟 , 𝑥 ̃ 𝑥 ) = e x p ( 𝛾 ) 𝑁 ( 𝜂 , ̃ 𝑟 𝑟 ) 𝑁 ( 𝛼 , ̃ 𝑥 𝑥 + 𝛽 ) , ( 2 1 ) 𝛼 = 𝜏 2 1 𝜌 2 + 𝑐 𝑏 2 ( 1 𝜆 ) 𝜙 2 , 𝜙 = 1 + 𝜌 𝜎 𝑏 𝑐 , ( 2 2 ) 2 𝛽 = 2 ( 𝑟 ̃ 𝑟 ) 𝜌 𝜎 𝑐 𝑐 + 𝜏 𝑏 2 ( 1 𝜆 ) 𝑑 1 𝜙 + 𝜎 2 ( 𝑟 + ̃ 𝑟 ) 𝜙 + 2 𝜎 𝑎 𝜌 𝑐 ( ( 2 3 ) 2 𝛾 = 𝑟 ̃ 𝑟 ) ( ( 𝑟 + ̃ 𝑟 ) 𝑏 2 𝑎 ) 𝑐 2 𝑎 + 𝜏 𝑏 𝑐 2 + ( 𝑐 / 𝑏 ) 2 ( 1 𝜆 ) 𝑑 2 2 ( 𝑟 + ̃ 𝑟 ) 1 𝑎 𝑏 𝑐 2 𝑟 2 + ̃ 𝑟 2 ( 𝑏 / 𝑐 ) 2 2 , ( 2 4 ) 𝑑 𝑚 = 𝑞 𝑚 ( 𝑟 ) + 𝑞 𝑚 ( ̃ 𝑟 ) , and 𝑞 ( 𝑟 ) = ( 𝑟 𝑏 2 𝑎 𝑏 + 𝑐 2 ) / 𝑐 2 . Extensive algebraic manipulations are required to express 𝐺 in (21) in the final form (3). The Green's function in (3) has the expected property: 𝐺 𝛿 ( 𝑟 ̃ 𝑟 ) 𝛿 ( 𝑥 ̃ 𝑥 ) and 𝑉 ( 𝑆 , 𝑟 , 𝜏 ) 𝑉 0 ( 𝑆 , 𝑟 ) as 𝜏 0 in (2).

References

  1. R. Mallier and A. S. Deakin, “A Green's function for a convertible bond using the Vasicek model,” Journal of Applied Mathematics, vol. 2, no. 5, pp. 219–232, 2002.
  2. A. Erdélyi, W. Magnus, F. Oberhettinger, et al., Higher Transcendental Functions. Vol. I, Robert E. Krieger, Melbourne, Fla, USA, 1981.
  3. A. Erdélyi, W. Magnus, F. Oberhettinger, F. G. Tricomi, et al., Higher Transcendental Functions. Vol. II, Robert E. Krieger, Melbourne, Fla, USA, 1981.
  4. E. T. Whittaker and G. N. Watson, A Course of Modern Analysis, Cambridge University Press, New York, NY, USA, 2nd edition, 1927.
  5. E. T. Whittaker and G. N. Watson, A Course of Modern Analysis, Cambridge University Press, New York, NY, USA, 4th edition, 1962.