a. For Vasicek process dr = a(b-r)dt + σdz, Δt=1/252, then we have Δr = A + Br+ ε. A=abΔt, B=-aΔt, Δt=1/252. To determine a,b and σ, I firstly calculate Δr based on the real world data. My goal is to find the a and b to minimize ∑(Δr − A − B𝑟! )" . Then I calculate Δr − A − B𝑟! and calculate the sum of them. Then I use SOLVER to find the appropriate a and b to minimize ∑(Δr − A − B𝑟! )" . Here I got a= 0.025016511, b= 0.003085497. As for σ, because E( (Δr − A − B𝑟! )" = σ" Δt and E (Δr − A − B𝑟! ) = 0 (since E(z2)=0), the model has sample 𝑉𝑎𝑟(Δr − A − B𝑟! ) = σ" Δt , so σ = # /$% 𝑉𝑎𝑟(Δr − A − B𝑟! ) = 0.002048213.
b. For CIR process dr = a(b-r)dt + σ√rdz, we have Δr/√r = A/√r + B√r + ε. A=abΔt, B=-aΔt, Δt=1/252
&!"#! '&! ) I firstly use SOLVER to find the a and b that minimize ∑( − − 𝐵 1𝑟! )2, and (&! (&!
&!"#! '&! ) also I add a constraint: ∑( − − 𝐵1𝑟! ) = 0 in SOLVER. And for σ, since (&! (&!
&!"#! '&! ) &!"#! '&! )
𝐸( − − 𝐵1𝑟! ) = 0, 𝐸( − − 𝐵1𝑟! ) 2 = σ" Δt, then the model (&! (&! (&! (&!
Finally I get a= 0.01138756, b= -0.000928032, σ= 0.021312886.
Q2. a. I need to convert Problem1.a Vasicek real world model dr = a(b-r)dt + σdz to Risk Neutral model. dr = [a(b-r)- λ(t,r)]dt + σdz Here we already know a= 0.025016511, b= 0.003085497, σ = 0.002048213, and λ(t,r)= λ, so we only need to solve the value of constant λ.
*+',- *+',- Here 𝑑𝑟 = 𝑎 7 * − 𝑟8 𝑑𝑡 + 𝜎𝑑𝑧, so I set b’= * . Here we use the exact formula to price a zero coupon bond(assume par 100): P(0,10)=100*A(0,10)exp[-r0B(0,10)], r0 = 0.0025688. On the other hand, the market price of a 10-year zero bond (par 100) is calculated as : Pm=100/[1+(2.15%/2)^20]= 80.74674044. Here we need to use SOLVER to find a λ which will make the model price equal to the market price. Finally I get λ = -2.138345615.
b. Here we assume λ(t,r)=λ√r, for a Risk Neutral CIR Model: dr = [a(b-r)- λ(t,r)]dt + σ√r dz we use the a, b , σ from Problem 1 (Real World Model) and λ(t,r) to calculate new a and b in the Risk Neutral Model. *+ a’=a+λr, b’=*./ 1. Here we also need to use the exact formula to price a zero coupon bond(assume par 100): P(0,10)=100*A(0,10)exp[-r0B(0,10)], r0 = 0.0025688, and the market price of the bond is also Pm=100/[1+(2.15%/2)^20]= 80.74674044. Finally I use SOLVER to find the λ that makes the model price equals to the market price of the bond. I get λ = -208.4675543
Q3. Here we use the Vasicek risk neutral model dr = a(b-r)dt + σdz, r0 =0.0013950 Again I use the exact price formula for these 7 zero coupon bonds: P(0,T) = 100*A(0,T) exp[-r0B(0,T)], A(0,T) and B(0,T) are calculated by using the formulus containing a,b, and σ. On the other hand, we need to calculate the market price of 1/3/5/10/20/30/50-year zero coupon bond by using the semi-annual spot rates (%) on high quality corporate bonds. PTj= 100/[1+(r/2)^j]. Then we get the market prices of these zero coupon bonds as:
Then I use SOLVER to find a,b, and σ that can minimize ∑(P(0, Tj) − 𝑃23 )" . Finally I get a=0.0405596, b= 0.1375098, σ= 0.0198401. The errors per 100 par are as follow:
Q4. a. For this interest swap, fixed rate = 0.283%, notional = 100, r0=0.0013950, 3-year, fixed payments are semi-annual, and float are received monthly. Our goal is to price this swap, P(swap)=PV(float)-PV(fixed).Firstly, here we need to use the formula: dr = a(b-r)dt + σdz, a=0.0405596, b= 0.1375098, σ= 0.0198401. Then I use the formula above to generate 500 paths Monte Carlo rates.
When calculating the PV of fixed payment, I use the formula:
3 4."67% '9 $ PV(fixed)= 100 ∗ 0.283%/2 ∗ ∑73;# 𝑃 K0, "L = 100 ∗ " ∗ ∑:3;# 𝐸[𝑒 % ] For the calculation of integrals I use the right end points, for example: : : 1 𝐼# = Δ𝑡 ∗ Q 𝑟 # 3 = ∗ Q𝑟 # 3 " #" 12 #" 3;# 3;#
When calculating the PV of float reveived, I use the following formula:
&$ $ ' ∫(&% &' => PV(float)= 100 ∗ ∑7: 3;# 𝐸[ &% ∗𝑒 ] #" For the calculation of integrals I also use the right end points.
As a result, I will make 6 fixed payments and received 12 float rates in total. Here, because I use 500-path Monte Carlo method, I will have 500 PVs for fixed payment and 500 PVs for float received. Then I calculated the Average value of PV(fixed)= 0.838079147, and PV(float)= 2.720372293.
With the same process as in (a), I generate 500-path Monte Carlo rates using the formula dr = a(b-r)dt + σdz, a=0.0405596, b= 0.1375098, σ= 0.0198401. Then I calculate the price of the zero coupon bond at t=3 using the formula: ) * P(t,T) = 100 ∗ 𝐸 7𝑒 ' ∫! &' => 8, which is P(3,8) = 100 ∗ 𝐸 7𝑒 ' ∫+ &' => 8here. For the Integral calculation, we need to use the integral from year 3 to year 8, which is # Δ𝑡 ∗ ∑?: ?: 3;7@ 𝑟 & 3 = #" ∗ ∑3;7: 𝑟 & 3 here. &% &%
Then I calculate 500-path put option payoffs at t=3, Payoff= Max{98-P(bond),0}.
Then I discount the payoffs to t=0 to get the price of this put option and calculate the average of these prices. ) + P(put option) = Payoff(t=3)* 𝐸 7𝑒 ' ∫! &' => 8= Payoff(t=3)* 𝐸 7𝑒 ' ∫( &' => 8 In this formula, the integral we use should be from year 0 to year 3, which is # Δ𝑡 ∗ ∑7: 7: 3;# 𝑟 & 3 = #" ∗ ∑3;# 𝑟 & 3 here. &% &%
Finally, I get the price of put option as 11.49601555.
Q5. a. Ho-Lee (RN) : dr = 𝜃(t)dt + σdz 𝜃(t)= Ft(0,t) + σ2t Here 𝜃(t) = 0.0022t1/10 + 0.0001t, so we get Ft(0,t) = 0.0022t1/10. Therefore, F(0,t) = 0.002 t11/10. The graph for 0≤t≤20 is shown below.
b. First, I use Hoo-Lee model to generate monthly step, binomial lattice. dr = (0.0022t1/10 + 0.0001t)dt + .01dz, r0 = 0.02. & Here, 𝑟!.$% = 𝑟! + K0.0022𝑡 &( + 0.0001tL Δt + .01 ∗ √Δ𝑡 ∗ (1 𝑜𝑟 − 1)
For example, 𝑟AA = 𝑟4 + [𝜃(0) + 𝜃(Δt)]Δt + 2 ∗ 0.01√Δ𝑡.
After generating binomial lattice, I calculate the callable bond value at each node. The bond is 1% semi-annual, callable for 100 on coupon dates starting at t=1 (2nd coupon) +, +, 𝑒 '&BA C∗$! ∗ 100.5 = 𝑃A ⎧ '&BA +- = & C∗$! +- ⎪𝑒 ∗ 100.5 = 𝑃A = So for example, 𝐵7 = … ⎨ ⎪ … ⎩ 𝑒 '&B=+, C∗$! ∗ 100.5 = 𝑃=+, And for non-coupon payment day, for example: '&BA C∗$! +- +, +- ⎧ 𝑒 +. & ∗ (𝑃A + 𝑃A = )/2 ⎪𝑒 '&BA = C∗$! ∗ (𝑃A+- = + 𝑃A+. =% )/2 𝐵7E = … #" ⎨ ⎪ … '&B= +- C∗$! +, +- ⎩ 𝑒 ∗ (𝑃= + 𝑃A= )/2
For coupon-payment day (from t=1 to t=2.5), for example:
'&BA C∗$! +( +& +( ⎧min {𝑒 ∗ (𝑃A + 𝑃A = )/2, 100} + 0.5 +( %/ % ⎪ '&BA%/ = & C∗$! 𝑃A = + 𝑃A = min {𝑒 ∗ , 100} + 0.5 𝐵".E = 2 ⎨ … ⎪ … '&B= +( C∗$! = +& +( ⎩min {𝑒 ∗ (𝑃 + 𝑃A= )/2, 100} + 0.5 But for t=0.5, although it is coupon payment day, the bond is not callable, and its calculation is not influenced by the call. Finally, I get the price of this callable bond as 95.59994187.
b. Put K=99 2-yr Zero coupon bond
t=0 t=1 t=3
Firstly, the value of this 2-year zero coupon bond is 100 at t=3. Then I use the same Binomial lattice above to firstly calculate the price of bond at t=1. I use the formula below to price this zero coupon bond: F 0 (!.$!,2).F 1 (!.$!,2) P(t,T) = 𝑒 '&! $! [ " ] Then after I get zero coupon prices at t=0, I use these prices to calculate payoffs of the put option, Payoff= Max{99-P(bond),0}. After that I discount these payoffs to t=0, use the same function P(t,T) = F 0 (!.$!,2).F 1 (!.$!,2) 𝑒 '&! $! [ " ]. Finally I get the put option price as 3.638041232.
