Step-by-Step Calibration of the Option Pricing & Rates Models

Amit Kumar Jha

Introduction
In this short pdf, I am breaking down the calibration of option pricing models and interest rate models step by step, focusing
on clear and practical methods for understanding and implementing these processes. We’ll cover key models such as the
Black-Scholes model, LV model, Vasicek model, CIR and HW Model.

1. Data Collection
• Market Prices of Options (Cmarket ): These are the observed prices of options traded in the market. We collect these
because they represent the ”true” value of the options under current market conditions.
• Underlying Asset Price (S0 ): This is the current price of the stock or asset on which the option is based.
• Strike Price (K): This is the price at which the option can be exercised.
• Time to Maturity (T): The remaining time until the option expires.
• Risk-Free Interest Rate (r): The continuously compounded risk-free rate, typically derived from government bonds.

2. Parameter to Calibrate
• Volatility (σ): This is the parameter we need to calibrate. It represents the standard deviation of the asset’s returns and
is crucial for pricing options.

3. Initial Guess for Volatility

Start with an initial guess for volatility. Let’s assume σ0 = 0.2 (20%).

4. Black-Scholes Option Pricing Formula

The theoretical price of a European call option under the Black-Scholes model is given by:

C = S0 Φ(d1 ) − Ke−rT Φ(d2 )

where:
ln(S0 /K ) + (r + σ2 /2) T
d1 = √
σ T
√
d2 = d1 − σ T
and Φ is the cumulative distribution function of the standard normal distribution.

5. Objective Function
The objective function measures the error between market prices and theoretical prices. We aim to minimize this error.
n
Error = ∑ (Cmarket,i − Cmodel,i (σ))2
i =1

where Cmodel,i (σ ) is the price calculated using the Black-Scholes formula with volatility σ.

1
Illustrative Example
Let’s say we have the following data for a European call option:

• Market Price of the Option (Cmarket ) = $8

• Underlying Asset Price (S0 ) = $100
• Strike Price (K) = $100

• Time to Maturity (T) = 1 year

• Risk-Free Interest Rate (r) = 5% (0.05)

Initial Guess
Let’s start with an initial guess for volatility σ0 = 0.2.

Calculating Cmodel
Compute d1 and d2 :
ln(100/100) + (0.05 + 0.22 /2) · 1 0 + 0.07
d1 = √ = = 0.35
0.2 1 0.2
√
d2 = d1 − 0.2 · 1 = 0.35 − 0.2 = 0.15
Compute Φ(d1 ) and Φ(d2 ): Using standard normal distribution tables or a calculator:

Φ(0.35) ≈ 0.6368

Φ(0.15) ≈ 0.5596
Compute the theoretical option price:

Cmodel = 100 · 0.6368 − 100 · e−0.05·1 · 0.5596

Cmodel = 63.68 − 53.14 = 10.54

Error Calculation
Error = (8 − 10.54)2 = (−2.54)2 = 6.45

Update Volatility
Since the model price (10.54) is higher than the market price (8), we need to reduce the volatility.
New Guess: Let’s try σ1 = 0.15.
Compute d1 and d2 for new σ:

ln(100/100) + (0.05 + 0.152 /2) · 1 0 + 0.0575

d1 = √ = = 0.3833
0.15 1 0.15
√
d2 = d1 − 0.15 · 1 = 0.3833 − 0.15 = 0.2333
Compute Φ(d1 ) and Φ(d2 ) for new σ:
Φ(0.3833) ≈ 0.6494
Φ(0.2333) ≈ 0.5928
Compute the theoretical option price for new σ:

Cmodel = 100 · 0.6494 − 100 · e−0.05·1 · 0.5928

Cmodel = 64.94 − 56.37 = 8.57

2
Error Calculation
Error = (8 − 8.57)2 = (−0.57)2 = 0.32
The error has reduced significantly. We can continue this process, adjusting σ iteratively until the error is minimized.

Final Iteration (if necessary)

If σ = 0.15 still doesn’t provide the desired accuracy, we can adjust it slightly up or down and repeat the calculations. Typ-
ically, using numerical optimization methods (like the Newton-Raphson method) can automate this process and converge
more quickly to the optimal σ.

Newton-Raphson Method for Calibrating the Black-Scholes Model

The Newton-Raphson method is an iterative technique used for finding the roots of a real-valued function. In the context
of calibrating the Black-Scholes model, we use it to minimize the error between the market price and the model price of an
option. Here, we’ll use the Newton-Raphson method to find the volatility (σ) that minimizes the objective function.

Steps to Apply Newton-Raphson Method

1. Define the Objective Function and Its Derivative
Let f (σ ) be the difference between the market price Cmarket and the Black-Scholes model price Cmodel (σ ):

f (σ ) = Cmodel (σ) − Cmarket

The derivative of f (σ) with respect to σ is needed for the Newton-Raphson method. This is denoted as f ′ (σ ).

2. Black-Scholes Model Price and Its Sensitivity (Vega)

The Black-Scholes price for a European call option is:

Cmodel (σ ) = S0 Φ(d1 ) − Ke−rT Φ(d2 )

The vega (sensitivity of the option price with respect to volatility) is:
√
ν = S0 Tϕ(d1 )
where ϕ is the probability density function of the standard normal distribution:
1 2
ϕ(d1 ) = √ e−d1 /2
2π

3. Newton-Raphson Iteration
The Newton-Raphson update formula is:

f (σn )
σn+1 = σn −
f ′ (σn )
Here, f (σn ) is the difference between the model price and the market price, and f ′ (σn ) is the vega.

Example Calculation
Let’s go through a detailed example with the following data:

• Market Price of the Option (Cmarket ) = $8

• Underlying Asset Price (S0 ) = $100

• Strike Price (K) = $100
• Time to Maturity (T) = 1 year
• Risk-Free Interest Rate (r) = 5% (0.05)

3
Initial Guess
Start with an initial guess σ0 = 0.2.

Iteration 1
Compute d1 and d2 for σ0 :

ln(100/100) + (0.05 + 0.22 /2) · 1 0 + 0.07

d1 = √ = = 0.35
0.2 1 0.2
√
d2 = d1 − 0.2 1 = 0.35 − 0.2 = 0.15
Compute Φ(d1 ) and Φ(d2 ):

Φ(0.35) ≈ 0.6368
Φ(0.15) ≈ 0.5596
Compute the theoretical option price:

Cmodel = 100 · 0.6368 − 100 · e−0.05·1 · 0.5596

Cmodel = 63.68 − 53.14 = 10.54
Compute Vega (ν):
1 2
ϕ(d1 ) = √ e−0.35 /2 ≈ 0.3752
2π
√
ν = 100 · 1 · 0.3752 = 37.52
Compute the difference and update σ:

f (σ0 ) = 10.54 − 8 = 2.54

2.54
σ1 = 0.2 − ≈ 0.2 − 0.0677 = 0.1323
37.52

Iteration 2
Compute d1 and d2 for σ1 :

ln(100/100) + (0.05 + 0.13232 /2) · 1 0 + 0.0283

d1 = √ = ≈ 0.2137
0.1323 1 0.1323
√
d2 = d1 − 0.1323 1 = 0.2137 − 0.1323 = 0.0814
Compute Φ(d1 ) and Φ(d2 ):

Φ(0.2137) ≈ 0.5853
Φ(0.0814) ≈ 0.5324
Compute the theoretical option price:

Cmodel = 100 · 0.5853 − 100 · e−0.05·1 · 0.5324

Cmodel = 58.53 − 50.61 = 7.92
Compute Vega (ν):
1 2
ϕ(d1 ) = √ e−0.2137 /2 ≈ 0.3910
2π
√
ν = 100 · 1 · 0.3910 = 39.10
Compute the difference and update σ:

f (σ1 ) = 7.92 − 8 = −0.08

−0.08
σ2 = 0.1323 − ≈ 0.1323 + 0.0020 = 0.1343
39.10

4
Iteration 3 (Optional)
Continue iterating in the same manner until the difference f (σn ) is close to zero, indicating that the theoretical price matches
the market price.

Summary
1. Define the objective function f (σ ) and its derivative (vega).

2. Make an initial guess for volatility.

3. Iteratively update the volatility using the Newton-Raphson formula.
4. Stop when the error between the market price and the model price is minimized.

The local volatility model extends the Black-Scholes model by allowing the volatility to be a function of both the underlying
asset price S and time t, i.e., σ = σ (S, t). This model captures the observed market phenomenon where implied volatilities
vary with strike prices and maturities.

Steps for Calibrating the Local Volatility Model

Data Collection
• Market Data: Collect option prices for a range of strike prices and maturities.
• Implied Volatilities: Convert market prices to implied volatilities using the Black-Scholes formula.

Dupire’s Equation
Dupire’s local volatility function σLV (S, t) can be derived from the market prices of European call options C (K, T ):
∂C ∂C
2 ∂T + rK ∂K
σLV (K, T ) = 1 2 ∂2 C
2 K ∂K2

Here, C (K, T ) is the market price of a European call option with strike K and maturity T.

Finite Differences Method

To compute σLV , we need the partial derivatives of the option prices:

∂C ∂C ∂2 C
, ,
∂T ∂K ∂K2

Example Calculation
Let’s go through an example step by step using simple finite differences.

Market Data
Option Prices: We have the following market prices for a given maturity T:

C (90, T ) = 12, C (100, T ) = 8, C (110, T ) = 5

Risk-Free Rate: r = 5% = 0.05

5
Finite Differences
First Partial Derivative with Respect to Strike Price K:

∂C C (K + ∆K ) − C (K − ∆K )
≈
∂K 2∆K
Let’s choose ∆K = 10:

∂C C (110, T ) − C (90, T ) 5 − 12
≈ = = −0.35
∂K K =100 2 · 10 20
Second Partial Derivative with Respect to Strike Price K:

∂2 C C (K + ∆K ) − 2C (K ) + C (K − ∆K )
≈
∂K 2 (∆K )2
Using the same ∆K = 10:

∂2 C C (110, T ) − 2C (100, T ) + C (90, T ) 5 − 2 · 8 + 12 1

≈ = = = 0.01
∂K2 K =100 10 2 100 100
First Partial Derivative with Respect to Maturity T:
Assume we have the prices for slightly different maturities T and T + ∆T. For simplicity, assume:

C (100, T ) = 8, C (100, T + ∆T ) = 7.9

Let ∆T = 0.01 (1 day):

∂C C (100, T + ∆T ) − C (100, T ) 7.9 − 8

≈ = = −10
∂T ∆T 0.01

Compute Local Volatility

Using Dupire’s equation:
∂C ∂C
2 ∂T + rK ∂K
σLV (100, T ) = 2
1 2∂ C
2 K ∂K2

Plugging in the values:

2 −10 + 0.05 · 100 · (−0.35)

σLV (100, T ) = 1 2
2 · 100 · 0.01

2 −10 − 1.75 −11.75

σLV (100, T ) = = = −0.235
50 50
Since volatility cannot be negative, an error might be present in the initial assumptions or calculations. Typically, a larger
data set or smoothing techniques are employed to reduce such discrepancies.

Iterative Process and Updates

The process outlined can be iteratively refined:

• Initial Guess: Start with an initial guess for σLV (K, T ).

• Compute Partial Derivatives: Using market data, compute the partial derivatives using finite differences.
• Update Volatility: Apply Dupire’s equation to update the local volatility surface.

• Repeat: Iterate the process to refine the local volatility estimates.

6
Vasicek Model Calibration
The Vasicek model describes the evolution of the short-term interest rate r (t) with the following stochastic differential
equation (SDE):
dr (t) = κ (θ − r (t)) dt + σ dW (t)
where:
• κ: Speed of mean reversion (how quickly the rate reverts to the long-term mean)
• θ: Long-term mean level of the rate
• σ: Volatility of the rate

• W (t): Standard Brownian motion

Calibration Steps
1. Collect Market Data
To calibrate the Vasicek model, you need market data, such as:
• Zero-coupon bond prices: Prices of bonds with different maturities.
• Interest rate swaps: Market prices of interest rate swaps with various tenors.

2. Calculate the Model-Implied Prices

The Vasicek model price of a zero-coupon bond P(t, T ) maturing at time T can be expressed as:

P(t, T ) = A(t, T ) exp (− B(t, T ) · r (t))

where A(t, T ) and B(t, T ) are functions that depend on the model parameters. For the Vasicek model:

1 − exp(−κ ( T − t))
B(t, T ) =
κ
 2 
2 exp(−κ ( T − t))

σ
A(t, T ) = exp θ 2 ( T − t) −
2κ 2κ

3. Objective Function for Calibration

Define the objective function to minimize the difference between the model-implied prices and the market prices. For
zero-coupon bonds:
N
F (κ, θ, σ ) = ∑ ( Pmarket,i − Pmodel,i )2
i =1

where:
• Pmarket,i : Market price of the i-th bond.
• Pmodel,i : Model price of the i-th bond, computed using the Vasicek model.

4. Minimize the Objective Function

Use optimization techniques to find the parameters κ, θ, and σ that minimize the objective function. Common methods
include:

• Gradient Descent: Iteratively adjust parameters in the direction that reduces the objective function.
• Newton-Raphson Method: Uses second-order information (the Hessian matrix) to find the minimum.

7
Detailed Example: Calibration of Vasicek Model
Market Data
Suppose you have the following market prices of zero-coupon bonds with maturities Ti :
• Bond 1 (Maturity = 1 year): Pmarket,1 = 0.98
• Bond 2 (Maturity = 2 years): Pmarket,2 = 0.95
• Bond 3 (Maturity = 3 years): Pmarket,3 = 0.92

Initial Parameter Guesses

Let’s start with the following initial guesses:

κ = 0.5, θ = 0.03, σ = 0.015

Compute Model Prices

For each maturity, compute B(t, T ) and A(t, T ) using the formulas provided above. Assume r (t) is the current short-term
interest rate (say 0.02).
Example Calculation for Bond 1 (Maturity = 1 year):

T=1
1 − exp(−0.5 × 1)
B(t, 1) = ≈ 0.632
0.5
 2 
2 exp(−κ ( T − t))

σ
A(t, 1) = exp θ 2 ( T − t) − ≈ 0.982
2κ 2κ
Pmodel,1 = 0.982 × exp(−0.632 × 0.02) ≈ 0.98
Repeat this process for Bonds 2 and 3.

Minimize the Objective Function

Calculate the Objective Function:

F (κ, θ, σ ) = (0.98 − 0.98)2 + (0.95 − Model Price 2)2 + (0.92 − Model Price 3)2

Optimize: Adjust κ, θ, and σ to minimize F using an optimization algorithm.

Newton-Raphson Method
To use the Newton-Raphson method, we need the gradient of the objective function with respect to the parameters. Let’s
focus on ∂F
∂θ .

Compute the Partial Derivatives

The gradient ∇ F with respect to each parameter κ, θ, and σ is needed. For ∂F
∂θ :

N ∂P
∂F
= −2 ∑ ( Pmarket,i − Pmodel,i ) · model,i
∂θ i =1
∂θ
For the model price:
P(t, T ) = A(t, T ) exp(− B(t, T ) · r (t))
 
∂Pmodel,i ∂A(t, T )
= exp(− B(t, T ) · r (t))
∂θ ∂θ

σ2 2 exp(−κ ( T − t))
 
∂A(t, T )
= A(t, T ) · 2 ( T − t) −
∂θ 2κ 2κ

8
Example Calculation
For Bond 1:
T=1
Pmarket,1 = 0.98, Pmodel,1 = 0.97
∂A(t, T )
= A(t, T ) · 0.000225 [1 − exp(−0.5)]
∂θ
∂Pmodel,1
Compute ∂θ :
 
∂Pmodel,1 ∂A(t, T )
= exp(− B(t, T ) · r (t))
∂θ ∂θ
Compute the Gradient:
∂F ∂P
= −2(0.98 − 0.97) · model,1
∂θ ∂θ

Summary
• Collect Market Data: Zero-coupon bond prices or interest rate swaps.
• Compute Model Prices: Use the Vasicek model formulas for bonds or swaps.
• Define Objective Function: Measure the difference between market and model prices.
• Optimize Parameters: Use numerical methods to find the parameters that minimize the objective function.

1 CIR Model
The CIR model describes the short-term interest rate r (t) with the following stochastic differential equation (SDE):
q
dr (t) = κ (θ − r (t)) dt + σ r (t) dW (t) (1)
Where:
• κ: Speed of mean reversion
• θ: Long-term mean level of the rate
• σ: Volatility of the rate
• W (t): Standard Brownian motion

2 Calibration Steps
2.1 Collect Market Data
As with the Vasicek model, you need market data such as:
• Zero-coupon bond prices: Prices of bonds with different maturities.
• Interest rate swaps: Prices of swaps with various tenors.

2.2 Compute the Model-Implied Prices

For the CIR model, the price of a zero-coupon bond P(t, T ) maturing at time T is given by:

P(t, T ) = A(t, T ) exp (− B(t, T ) · r (t)) (2)

Where A(t, T ) and B(t, T ) are functions that depend on the model parameters. For the CIR model:

2 (exp(κ ( T − t)) − 1)
B(t, T ) = (3)
2κ + (σ2 /θ ) (exp(κ ( T − t)) − 1)
σ2
  
2κθ
A(t, T ) = exp B ( t, T ) − ( T − t ) − ( B ( t, T ) − ( T − t )) (4)
σ2 2κ

9
2.3 Objective Function for Calibration
Define the objective function to minimize the difference between the model-implied prices and the market prices:
N
F (κ, θ, σ ) = ∑ ( Pmarket,i − Pmodel,i )2 (5)
i =1

Where Pmodel,i is the price computed using the CIR model and Pmarket,i is the market price.

2.4 Minimize the Objective Function

Use optimization techniques to find the parameters κ, θ, and σ that minimize the objective function.

3 Detailed Example: Calibration of CIR Model

3.1 Market Data
Suppose you have the following market prices of zero-coupon bonds with maturities Ti :

• Bond 1 (Maturity = 1 year): Pmarket,1 = 0.98

• Bond 2 (Maturity = 2 years): Pmarket,2 = 0.95
• Bond 3 (Maturity = 3 years): Pmarket,3 = 0.92

3.2 Initial Parameter Guesses

Let’s start with the following initial guesses:

• κ = 0.5
• θ = 0.03

• σ = 0.015

3.3 Compute Model Prices

Compute B(t, T ) and A(t, T ):
For each maturity, compute B(t, T ) and A(t, T ) using the formulas provided above. Assume r (t) is the current short-term
interest rate (say 0.02).
Compute Pmodel,i :
Use the computed A(t, T ) and B(t, T ) to get the model prices.
Example Calculation: For Bond 1:

• T=1
Compute B(t, 1) and A(t, 1):
2 (exp(0.5 × 1) − 1)
B(t, 1) =  2
 (6)
2 × 0.5 + 0.015
0.03 (exp(0.5 × 1) − 1)

2 × 0.5 × 0.03 0.0152

  
A(t, 1) = exp B(t, 1) − (1 − 0) − ( B(t, 1) − (1 − 0)) (7)
0.0152 2 × 0.5 × 0.03
Compute these values to find Pmodel,1 .

3.4 Minimize the Objective Function

Calculate the Objective Function:

F (κ, θ, σ ) = (0.98 − Model Price 1)2 + (0.95 − Model Price 2)2 + (0.92 − Model Price 3)2 (8)
Optimize:
Adjust κ, θ, and σ to minimize F using an optimization algorithm.

10
3.5 Gradient Calculation for Newton-Raphson
∂F
To use the Newton-Raphson method, we need the gradient of the objective function with respect to κ, θ, and σ. For ∂θ , the
gradient is:
N ∂P
∂F
= −2 ∑ ( Pmarket,i − Pmodel,i ) · model,i (9)
∂θ i =1
∂θ
∂Pmodel,i
Where ∂θ can be derived from:

∂Pmodel,i ∂A(t, T )
= · exp (− B(t, T ) · r (t)) (10)
∂θ ∂θ
∂A(t,T )
To compute ∂θ :

∂A(t, T ) 2σ2
= A(t, T ) · 2 [ B(t, T ) − ( T − t)] (11)
∂θ θ

Hull-White Model
The Hull-White model describes the short-term interest rate r (t) with the following stochastic differential equation (SDE):

dr (t) = θ (t) dt + σ (t) dW (t)

where:
• θ (t) is the deterministic drift term.
• σ (t) is the time-dependent volatility term.
• W (t) is a standard Brownian motion.
For the Hull-White model, the term structure of the interest rate is given by:

P(t, T ) = A(t, T ) exp(− B(t, T ) · r (t))

where A(t, T ) and B(t, T ) are functions of the model parameters. Specifically:

1 − e−α( T −t)
B(t, T ) =
α

σ2 
 
2
A(t, T ) = exp B(t, T ) − ( T − t)
2α2
where:
• α is the mean-reversion speed.
• σ is the volatility of the short-term interest rate.

Calibration Steps
Collect Market Data
To calibrate the Hull-White model, you need:
• Zero-Coupon Bond Prices: Prices of bonds with different maturities.
• Interest Rate Swaps: Prices of swaps with various tenors.

Compute Model-Implied Prices

For zero-coupon bonds, use the Hull-White model formulas to compute the model price:

Pmodel (t, T ) = A(t, T ) exp(− B(t, T ) · r (t))

11
Define the Objective Function
The objective function F to minimize is the sum of squared differences between market prices and model prices:

N
F (α, σ ) = ∑ ( Pmarket,i − Pmodel,i )2
i =1

Optimize Parameters
Use optimization techniques to find the parameters α and σ that minimize the objective function.

Detailed Example: Calibration of Hull-White Model

Market Data
Suppose you have the following market prices of zero-coupon bonds with maturities Ti :
• Bond 1 (Maturity = 1 year): Pmarket,1 = 0.98
• Bond 2 (Maturity = 2 years): Pmarket,2 = 0.95
• Bond 3 (Maturity = 3 years): Pmarket,3 = 0.92

Initial Parameter Guesses

Let’s start with the following initial guesses:
• α = 0.5

• σ = 0.015

Compute Model Prices

Compute B(t, T ) and A(t, T ):
For Bond 1 (Maturity = 1 year):
1 − e−0.5×1
B(t, 1) = ≈ 0.632
0.5

0.0152 
 
2
A(t, 1) = exp 0.632 − 1
2 × 0.52
 
0.000225
A(t, 1) ≈ exp (0.399 − 1) ≈ exp(−0.000045) ≈ 0.999955
0.5
So, the model price for Bond 1:
Pmodel,1 = 0.999955 × exp(−0.632 × r (t))

Compute Prices for Other Bonds

Repeat the process for Bonds 2 and 3 using their maturities.

Define and Minimize Objective Function

Calculate the Objective Function:
Define F (α, σ ) as:
F (α, σ ) = (0.98 − Pmodel,1 )2 + (0.95 − Pmodel,2 )2 + (0.92 − Pmodel,3 )2

Optimize Parameters
Use numerical optimization techniques (e.g., gradient descent, Newton-Raphson) to adjust α and σ to minimize F.

12
Gradient Calculation for Newton-Raphson
∂F
To use the Newton-Raphson method, we need the gradients of F with respect to α and σ. For ∂α , the gradient is:
N ∂P
∂F
= −2 ∑ ( Pmarket,i − Pmodel,i ) · model,i
∂α i =1
∂α
∂Pmodel,i
where ∂α can be derived from:
 
∂Pmodel,i ∂A(t, T )
= · exp(− B(t, T ) · r (t))
∂α ∂α

Similarly, compute the gradient with respect to σ:

N ∂P
∂F
= −2 ∑ ( Pmarket,i − Pmodel,i ) · model,i
∂σ i =1
∂σ
∂Pmodel,i
where ∂σ can be derived from:
 
∂Pmodel,i ∂A(t, T )
= · exp(− B(t, T ) · r (t))
∂σ ∂σ

Summary
1. Collect Market Data: Zero-coupon bond prices or interest rate swaps.

2. Compute Model Prices: Use Hull-White model formulas.

3. Define Objective Function: Measure the difference between market and model prices.
4. Compute Gradients: Derive partial derivatives for Newton-Raphson.

5. Optimize Parameters: Use numerical methods to minimize the objective function.

13