A Stochastic Correlation Model with Time Change for Pricing Credit Spread Options

In this paper, we introduce the stochastic correlation processes for modeling the credit spread. We first model the components of spread process as correlated Ornstein-Uhlenbeck processes and correlation as Jacobi process. Using the properties of Jacobi process, we are able to obtain the analytical solutions for the credit spread option prices. To further enhance the model’s ability to capture the abrupt changes in the observed correlation time series, we construct a new model where the correlation is modeled by a Jacobi process time change by Lévy subordinators. We employ the eigenfunction expansion methods to obtain the closed-form solutions for the option prices. Our empirical study indicates the time changed Jacobi process fits the correlation series significantly better than the Jacobi process.


Introduction
A credit spread option is an option on a particular borrower's credit spread.The credit spread is the difference between the yield on the borrower's debt and the yield on Treasury debt of the same maturity.The credit spread represents the risk premium that the market demands for holding the borrower's debt.Unlike the credit default swaps, the payoff of credit spread options depends solely on the measurement of the credit spread rather than a specific credit event.Credit spread options protect the holder of the borrower's debt from the loss of the rising yield spread due to default or credit rating downgrade.
Most of the models used in the pricing of credit spread options belong to the multi-factor models where state variables are modeled as correlated stochastic processes.[1] proposes a valuation model for pricing credit spread options by assuming that the risk-free interest rate and logarithm of the credit spread follow correlated mean reverting diffusion processes.[2] argues that the spread should be modeled using its two components instead of the spread itself and he extends [1] by assuming that the risk-free rate and the two components of credit spread follow correlated Ornstein-Uhlenbeck (OU) processes.[3] proposes a very general mean reverting process for the credit spread and two stochastic volatility processes, the square-root process and the OU process.The volatility process is assumed to be correlated with the spread process.There are many other examples that employ correlated processes for modeling the credit spread and pricing credit spread options, see e.g.[4]- [9].
Although most of the models stress the role played by the correlation between state variables for pricing credit spread options, a common assumption is that the correlation is constant.There is a growing literature that documents that the correlation is far from being constant and it should be modeled as a random process, just like interest rate, stock price and volatility.[10] provides analytical properties of the correlation process and provides a new approach of modeling correlation as a stochastic process.[11] and [12] propose a modified Jacobi process to evaluate risk premium of the stochastic correlation and develop a series solution for pricing options under the correlation risk.[13] provides a closed-form approximation for pricing of several two-dimensional derivatives under the assumptions of stochastic correlation.[14] and [15] propose to generate the stochastic correlation process through the hyperbolic transformation of any mean-reverting process with positive and negative values.
In this paper, the main problem we are going to solve is how to model the credit spread and price the credit spread options with the stochastic correlation process.We stress that so far most of stochastic correlation models are developed for the equity and foreign exchange derivatives, where the underlying state variables are usually modeled as correlated geometric Brownian motions.To model the credit spread, we need to take the mean-reversion in the state variables into consideration.In particular, we model the components of the spread process as correlated OU processes and stochastic correlation as Jacobi process which guarantees the correlation between state variables is a bounded process.Using the properties of Jacobi process, we are able to derive the analytical formula for the moments of integrated correlation process required for calculating the credit spread options prices.From the data we employ for the empirical study, we observe that the correlation process changes abruptly and a diffusion model may not fully capture its dynamics.We therefore develop a time-changed model for correlation process where the Jacobi process is time changed by Lévy subordinators to yield state-dependent jumps.To the best of our knowledge, the time-changed Jacobi process (TC-Jacobi) has not been studied in the literature for modeling correlation.We notice that the TC-Jacobi process remains analytically tractable and the closed-form solutions for the credit spread options we de-rive under the Jacobi process require only minor changes.Our empirical results demonstrate that indeed the TC-Jacobi process improves the fit of correlation process significantly compared to the Jacobi process.
The structure of the paper is as follows.In Section 2, we introduce the credit spread model with constant correlation.In Section 3 we extend the constant correlation model to stochastic correlation model.Using the properties of Jacobi process, we derive the closed-form solutions for credit spread option prices.We introduce the time change to Jacobi process in Section 4 and demonstrate that the analytical formulas remain little changed by employing the eigenfunction expansion method.We compare the performance of Jacobi process with the time changed process using the correlation time series in Section 5. We also numerically study the implications on the derivatives for both processes.

A Credit Spread Model with Constant Correlation
Let ( ) Ω, ,P  denote a probability space with an information filtration ( t  ).
We assume that under physical measure P , the dynamics of the two yields in- cluded in the credit spread, denoted by 1 motions with constant correlation coefficient ρ .
To price the credit spread options, we need to work under the risk-neutral measure Q .We assume that under the measure Q , the stochastic processes of ( ) ( ) 2 X t are still correlated Ornstein-Uhlenbeck processes, but with different coefficients to highlight the risk premium.Therefore, ( ( ) and ( ) ( ) Under the above assumptions, we know that at time T, under the measure Q the vector has a bivariate normal distribution with mean µ and covariance matrix Σ given by , , where ( ) C T K ρ denote the price of a call option on credit spread with strike price K and maturity T .The option price can be calculated from where ρ is included in C to highlight that option price depends on the cor- relation coefficient ρ .r is the risk-free short rate and for simplicity, we as- sume it is a constant.Since , X T X T follows a bivariate normal distribution, we know that ( ) ( ) It becomes straightforward to compute the call option prices.We summarize the result in the following proposition (see also [2] and [12]).Proposition 1. Assume the dynamics of 1 X and 2 X are given by ( 4), (5) and ( 6) then the option price ( ) C T K ρ can be calculated by , , exp exp , 2 2π where

Motivation for Stochastic Correlation Process
In this section, we extend the constant correlation model in Section 2 by modeling the correlation coefficient ρ as a stochastic process.To motivate our model, we first take a look at the data we use in the empirical study.Our data contain daily rates for 10-year constant maturity Treasury bonds and Moody's Aaa and Baa seasoned bond indices.The data cover the period January 2, 1996 to July 25, 2016, for a total of 5146 observations.In Figure 1, we plot the time series of three yields and it is clear that all of them display mean reversion.We also construct the time series of correlation by the following procedure (similar to [10] [14] and [15]): 1) At time t , we regress daily changes in the values of ( ) ( ) on the values of ( ) ( ) X t a day before for a time window T n : and 2) The correlation ( ) is calculated by the correlation coefficient between ( ) ( ) ( ) where ( ) ( ) .
We then roll over to time t t + ∆ and so on to obtain a series of correlations through the time.
In Figure 2 we plot the dynamics of ( ) with 30 T n = .We can see clearly that the correlation is far from being constant and it varies very abruptly.The correlation is supposed to be bounded between −1 and 1, but in our dataset, it is concentrated on [0, 1].In addition, the correlation process displays strong mean-reverting property.In Table 1, we provide the summary statistics for ( ) calculated from the two defaultable bond indices.The average correlations between the yields of Treasury bonds and Aaa and Baa bonds are around  0.87, but the correlations fluctuate between 0.33 and 0.99.The density functions for the correlation series have negative skewness and fat tail.
To satisfy the properties of the correlation series displayed in our data, we model the correlation as a Jacobi process, which takes the value between m ρ and M ρ , i.e.Under the martingale measure Q , we assume the stochastic process of ( ) where ( ) tion under the measure Q and independent of ( ) 4) and (5).
The standard Jacobi process Y takes values between 0 and 1.Following [16], we also impose the following conditions to ensure that the boundaries 0 and 1 are inaccessible to the process ( )

Properties of Jacobi Process
It is well known that there is no closed-form expression for the transition density for the Jacobi process (see e.g.[17]).However, the transition density of Jacobi process can be represented in terms of eigenfunction expansions.From ( 19), we know the infinitesimal generator  of Jacobi process is where f is transformation function.f ′ and f ′′ are first-and second-order derivatives of f , respectively.
Let { } n λ be the eigenvalues of − and { } n ψ be the corresponding eigenfunctions, i.e. .
We know for the Jacobi process Y defined on ( ) , the eigenvalues and eigenfunctions can be written as (see e.g.[16] and [17]) and where ( ) where ( ) is the Jacobi polynomials defined as , ; ; F a b c x is the hypergeometric function defined by The spectral decomposition allows us to find the transition density function of Jacobi process Y as where ( ) π ⋅ is the stationary density function of Jacobi process Y , ( ) ( ) ( ) where ( ) where Γ is the standard Gamma function.
The most important property of Jacobi polynomials is that they satisfy the following orthogonality condition (see e.g.[16]): Proposition 2.
Using this property, we can immediately have the following results: k and 2 k be nonnegative integers.
In this paper, we need to compute the following two integrals: where 1 k and 2 k are two nonnegative integers.
We can prove the following results: Lemma 1. (i) where , , ; , ; F a b c d e x is the hypergeometric function defined by Proof.To prove (i), from Corollary 1, we immediately know that for k k n + < , the integral is clearly zero.For 1 2 k k n + ≥ , we can use the following formula (see e.g.[18]): ) To prove (ii), we first use another representation of the Jacobi polynomials, Then, we have Using Corollary 1, we know that for 2 n m − ≥ , the above integral will be zero.If ( ) We can prove the following result: Theorem 1. ( ( ) where 0 : 0 s = , α , β , k N and k λ are assumed to be calculated using the coefficients under the measure Q .
Proof.First, we know ) Using the eigenfunction expansion for the transition density of Jacobi process Y , we have

E Y s s y s s p s s y s y s y s y s
s s s y For the first integral in the above equation, For the second integral, Using Lemma 1, we can see that ( ) ( ) ( ) . Otherwise, we have ( ( ) Finally, using the definition of k ψ , we obtain the final result.

□
We are also interested in calculating the moments of the following quantity ( ) Using binomial expansion, we immediately have ( ) It is also easy to obtain the centered moments of .
We provide the first two centered moments for ( ) where

Credit Spread Option Prices under Jacobi Process
If ( ) t ρ is not a constant, but deterministic, the credit spread X T X T − defined in Section 2 will still be a Gaussian process with same mean where ( ) T ρ is defined in (50).The pricing formula in (14) will be changed to , , exp exp , 2 2π where If the correlation coefficient ( ) is a stochastic process, but independent of underlying processes for the two yields ( )

we can employ
Taylor series expansion method to obtain the analytical formula for the spread credit spread option prices (see e.g.[11] C T K E ρ denotes the k th derivative of ( ) C T K ρ with re- spect to ρ evaluated at the point In practice, we need to approximate the series using the first n terms to obtain the option prices ( To estimate the truncation error, we use the remainder of Taylor's formula.We know there exists * ρ such that ( Then, the bound for the error of ( ) For example, the second-order approximation to the credit spread option price is ( where .

Stochastic Correlation with Time Changed Jacobi Process
From Figure 2, it is clear that the correlation process changes abruptly through the time.We also know that the density function of correlation process has fat tail.These indicate that a diffusion process such as Jacobi process may not fully capture the main features of the correlation process and it is necessary to include the jumps into the correlation process.One way is to add the jump process directly into the correlation process in (19), which is the approach taken by [19].The shortcoming of this approach is that it is difficult to ensure the correlation stays between m ρ and M ρ .
In this section, we model the correlation with Jacobi process time changed by Lévy subordinators to yield state-dependent jumps.The models with stochastic time changes have been studied in [20] and [21] for equity markets, [22] and [23] for credit markets and [24] and [25] for commodity markets.Here we model the stochastic correlation with time changed process.

Lévy Subordinator
A Lévy subordinator ( ) T t is a non-decreasing Lévy process with jumps and non-negative drift (see e.g.[26]).The Laplace transform of a Lévy subordinator can be written as where function ( ) φ λ is known as the Lévy exponent of the subordinator ( ) and is given by the Lévy-Khintchine formula (see e.g.[26])

The TC-Jacobi Process
We model the correlation as the TC-Jacobi process, i.e.
where ( ) T t is a Lévy subordinator and ( ) Under the martingale measure Q , we assume the stochastic process of ( ) Therefore, the original Jacobi process ( ) is time changed by Lévy subordinator φ to generate jumps.With these specifications, the new correlation process ( ) is guaranteed to lie between m ρ and M ρ .
An important advantage of modeling the correlation as Jacobi process time changed by Lévy subordinator is that we can express the transition density of Y φ via eigenfunction expansion (see e.g.[24] and [25]), i.e.where n λ , n ψ and ( )

( ) (
are the eigenvalues, eigenfunctions and steady-state density of Jacobi process, respectively, which can be found in (24), ( 25) and (30).Thus, the eigenfunction expansion of Y φ remains the same form as Y , but with Using the eigenfunction expansion of the transition density of Y φ , we see that the moments of ( ) by replacing j λ with ( ) j φ λ .We immediately obtain the following result: Theorem 2. ( ) ( ) ( ) where 0 : 0 The credit spread options prices under the TC-Jacobi process can then be computed by employing exactly the same technique as Jacobi process.
for Jacobi process and for TC-Jacobi process.
The MLE approach based on eigenfunctions for Jacobi process has been investigated in [17].They also compare the performance of MLE estimator with other estimators proposed in the literature, such as the estimator in [27], which is a method of moments based on an approximation of score function or with a generalized method of moments (GMM) estimator.The MLE estimator is further compared with computer-intensive simulation-based estimator, such as the simulated method of moments (SMM).They show that MLE outperforms the other estimators relative to bias and variance, while being easy to implement.
In the estimation, we set 0 m ρ = and 1 M ρ = so that the correlation is bounded by 0 and 1.The estimation results are presented in Table 2.We notice that all the parameters for both Jacobi and TC-Jacobi processes are highly significant.For Jacobi process, the mean parameters are very close for the correlations calculated from Aaa and Baa bonds.The correlations for Aaa bonds have higher mean reversion and smaller variance compared to Baa bonds.Whence Lévy subordinator is included, we clearly see that mean reversion and variance parameters drop significantly.This is as expected since the time-change process introduces the mean-reverting jumps into the correlation process, which is contrary to Jacobi process where the only source to generate mean reversion is through diffusion process.
We also report the log-likelihood function values, Akaike Information Criterion (AIC) and the Bayes Information Criterion (BIC) in Table 2.It is clear for the correlation calculated from both Aaa and Baa bonds, the TC-Jacobi model produces much larger likelihood values than Jacobi model, which The standard errors are in the parentheses.significant effect on the prices for different volatility parameters.Finally, we find the higher δ , the lower η in the Lévy process, the higher the price is.

Conclusions
In this paper, we introduce the stochastic correlation process for modeling the credit spread.We model the components of spread process as correlated OU processes and correlation as Jacobi process.We study the properties of Jacobi process and show how to obtain the analytical solutions for credit spread options.To enhance the model's ability to capture the main features of correlation process, we extend the Jacobi process to TC-Jacobi process where the correlation is modeled by a Jacobi process time changed by Lévy subordinators.We demonstrate that the TC-Jacobi process remains analytically tractable and the closed-form solutions for the credit spread options derived under the Jacobi process requires only minor changes.From an empirical study using the correlation series constructed from the yields on Treasury bonds and Aaa/Baa bonds, we demonstrate the superiority of TC-Jacobi compared to Jacobi process for modeling the correlation.We also numerically illustrate that modeling the correlation as a stochastic process will have an effect on the option prices compared to the constant correlation model.Furthermore, there are differences in option prices from Jacobi and TC-Jacobi processes.
To keep our model as parsimonious as possible, we assume that both risk-free short rate and volatility are constants.It will be an interesting extension to allow them to be stochastic.As long as the stochastic processes for the short rate and volatility are independent of correlation process, our method of obtaining the closed-form solutions for option prices will continue to work.We would also like to apply the TC-Jacobi process to the modeling of correlation in equity, foreign exchange and commodity markets.

Figure 1 .
Figure 1.Time series of the yields of Treasury bonds and Moody's Aaa and Baa seasoned bond indices.

Figure 2 .
Figure 2. Correlations between the yields for 10-year constant maturity Treasury bonds and Moody's Aaa and Baa bond indices.Left panel: Aaa bond indices; right panel: Baa bond indices.
we can compute the integral using the definition of function n a .□Inorder to obtain the closed-form solution for the credit spread option prices, we are interested in computing the moments of the following integral general equation for converting the n th uncentered moment to the centered moment is is the drift term of ( ) T t .ν is the Lévy measure which must sa- tisfy -class of Lévy subordinators are the tempered stable subordinators.For such subordinators, the Lévy measure ( ) For such subordinators, the Lévy exponent is given by obtain the closed-form solution for the credit spread option prices under the TC-Jacobi process, we need to calculate the moments of the following integral

Figure 5 .
Figure 5. Credit spread option prices with different parameters from TC-Jacobi process.The parameters for correlation process are obtained from Table 2. 1 0.15, κ = 1 0.06, θ =

Table 1 .
Summary statistics of correlation series.

Table 2 .
Estimation results for the correlation processes.