Abstract. For a squared Bessel process, X, the Laplace transforms of joint laws of (U, Ry 0 X p s ds) are studied where Ry is the first hitting time of y by X and U is a random variable measurable with respect to the history of X until Ry. A subset of these results are then used to solve the associated small ball problems for Ry 0 X p s ds and determine a Chung's law of iterated logarithm.Ry 0 X p s ds is also considered as a purely discontinuous increasing Markov process and its infinitesimal generator is found. The findings are then used to price a class of exotic derivatives on interest rates and determine the asymptotics for the prices of some put options that are only slightly in-the-money. .
IntroductionLet X be a squared Bessel process which is the unique strong solution towhere ν ≥ −1 is a real constant and B is a standard Brownian motion. Letting δ = 2(ν + 1), X is called a δ-dimensional squared Bessel process. We will denote such a process with X 0 = z by BESQ δ (z) and δ and ν will be related by δ = 2(ν + 1) throughout the text. In this paper we are interested in the integral functional where p > −1 and R y := inf{t ≥ 0 : X t = y} for y ∈ [0, ∞) and X is BESQ δ (z). (In the sequel, we will write R δ y only if we need to specify the dimension to avoid ambiguity.) Squared Bessel processes have found wide applications especially in Finance Theory, see Chapter 6 in [8] for a recent account. They can, e.g., be used to model interest rates in a Cox-Ingersoll-Ross framework. In the above setting, if X p models the spot interest rates, then exp Σ δ p,z,y refers to the cumulative interest until the spot rate hits the barrier y p . As such, this random variable is related to certain exotic options on interest rates (see [5] for some formulae regarding barrier options in a similar framework). Bessel processes also appear often in the study of financial bubbles since 1/ √ X is the prime example of a continuous (strict) local martingale when X is a BESQ 3 (see, e.g., [14], [17] and [18] for how strict local martingales, and in particular Bessel processes, appear in mathematical studies of bubbles).In Section 2 we will determine the joint law of (U, Σ δ p,z,y ) by martingale methods, where U is a random variable measurable with respect to the evolution of X until R y . In particular we will obtain the joint distributions of (R y , Σ δ p,z,y ) and (max t≤Ry X t , Σ δ p,z,y ). As a by-product of our findings, if |ν| p+1 = 1 2 , we have a remarkable characterisation of the conditional law of Σ δ p,z,y given that the maximum (resp. minimum) of X at R y is below (resp. above) a fixed level in terms of the first hitting time distributions of a 3-dimensional Bessel process when z ≥ y (resp. z ≤ y).Date: March 9, 2015.