{"categories":["Fixed Income"],"contentHtml":"

LMM

\n

The LIBOR Market Model (or Brace-Gatarek-Musiela Model)

\n

is a model constructed in terms of the forward rates underlying caplet prices.

\n

HJM was defined as a process for the forwards.

\n

Basically the benefit was that it could model the entire term structure, as long as you input all the volatilities for the term structure.

\n

But the problem is the calibration.

\n

Notation

\n

\\(t_k\\) kth reset date

\n

So t would progress like t_0 = 0, t_1 = 0.25, t_2 = 0.5…

\n

\\(F_k\\) forward rate between k and k + 1

\n

m(t) index for next reset date at time t.

\n

So if t = 0.4, then m(t) = 0.5

\n

δ_k = t_k+1 - t_k

\n

So basically you can imagine that you have your forward curve,

\n

xi: ξ(t) = volatility of F_k(t) at time t

\n

Assume that we have only one factor

\n

This factor is the forward risk neutral process with respect to P(t, t_k+1)

\n

Then, our process

\n

$$\ndF_k(t) = \\xi_k (t) F_k(t) dW\n$$

\n

The change is driven by the volatility, existing forward rate, and Brownian motion.

\n

Applying some change of numeraire, skipping some steps, you get

\n

the rolling forward rate process (in notes)

\n

Price of a ZCB at time t_i which is discounted, the ratio of that to t_i+1

\n

$$\n\\frac{P(t, t_i)}{P(t, t_{i+1})} = 1 + \\delta_i F_i(t)\n$$

\n

If you apply the natural log, you get

\n

$$\n\\ln P(t, t_i) - \\ln P(t, t_{i+1}) = \\ln [1 + \\delta_i F_i(t)]\n$$

\n

This is the relationship between the ZCB and the forward.

\n

Then if you apply Ito’s lemma… and equate coefficient of dW

\n

something long…

\n

Then you can take the substitution of the forward process.

\n

$$\n\\frac{dF_k(t)}{F_k(t)} = \\sum_{i= m(t)}^k \\frac{\\delta_i F_i(t) \\xi_i(t) \\xi_k(t)}{1+\\delta_i F_i(t)} dt + \\xi_k(t) dW\n$$

\n

This is the process followed by the forward rate between t_k and t_k+1, in a risk-neutral world.

\n

In the limiting case, when this interval becomes smaller, this converges to HJM

\n

The idea is being able to calibrate this model.

\n

It can be simplified with:

\n

Simplified Model

\n

Assume that ξ_k(t) function only on the number of whole accrued periods between the next date and t_k.

\n

Define Λ_i as the value of ξ_k(t) (volatility of forward) when there are i such accrued periods.

\n

Then the volatility of the forward can be redefined as

\n

\\(\\xi_k(t) = \\Lambda_{k-m(t)}\\) which is a step function

\n

And such values like Λ_i can be estimated from the volatilities used to value caplets in Black’s model.

\n

Recall that to value a caplet, we have

\n

$$\nL\\delta_k P(0, t_{k+1}) [F_k N(d_1) - R_k N(d_2)]\n$$

\n

blah blah blah d_1 and d_2

\n

If we equate the variances between Black’s model and this, we get

\n

$$\n\\sigma_k^2 t_k = \\sum_{i=1}^k \\Lambda_{k-i}^2 \\delta_{i-1}\n$$

\n

Example in the slides for converting Black volatilities to Lambda, use to check

\n

Implementation

\n

How would I use this model to price a bond. If you observe the volatility, then you can price a bond under this?

\n

This would be done via Monte Carlo simulation

\n

$$\n\\frac{dF_k(t)}{F_k(t)} = \\sum_{i=m(t)}^k \\frac{\\delta_i F_i(t) \\Lambda_{i - m(t)}\\Lambda_{k-m(t)}}{1 + \\delta_i F_i(t)} dt + \\Lambda_{k-m(t)} dW\n$$

\n

Then, by Ito’s lemma,

\n

$$\nd\\ln F_k(t) = \\left[\\sum_{i=m(t)}^k \\frac{\\delta_i F_i(t) \\Lambda_{i-m(t)} \\Lambda_{k-m(t)}}{1 + \\delta_i F_i(t)}\\right]dt + \\text{same thing as above}\n$$

\n

Then we can approximate the drift by approximating \\(F_i(t)\\) and \\(t\\)

\n

Then,

\n

$$\nF_k(t_{j+1}) = F_k(t) \\exp\\left[\\biggl(\\sum_{i=j+1}^k \\frac{\\delta_i F_i(t_j) \\Lambda_{i-j-1}}{1 + \\delta_i F_i (t_j)} - \\frac{\\Lambda^2_{k-j-1}}{2}\\biggr) \\delta_j + \\Lambda_{k-j-1} \\epsilon \\sqrt{d_j}\\right]\n$$

\n

Slide 8 has more calculations that I can use for testing.

\n

ipynb file is provided

\n

Each approximation of the drift the forward rate remains constant.

\n

If we define the rolling forward in the risk neutral world, it allows us to discount bond from one date to the next one.

\n

In terms of simulation, what’s happening in the code is that if you assume that we want to simulate a zero curve with N accrued periods.

\n

On each trial, we start with the forward rate at time 0, which is calculated from the initial zero curve.

\n

F_0(0), F_1(0), … F_N-1(0)

\n

Then we can use an approximation formula (described above) to calculate F_1(t_1), F_2(t_1)...

\n

Limitations of the theoretical model

\n

Credit Risk

\n

LMM is a little complicated, this is just an introduction to the topic.

\n

Credit risk is the probability that a debtor will default on their debts. We have always assumption that the cash flow is 100% likely i.e. with treasury bonds.

\n

For single-name derivatives, we will look at CDS and CDOs. Then we will look at portfolios. We are interested in expected loss, with a discrete number of names in the portfolio, to estimate the losses in order to price the portfolio correctly.

\n

Rating agencies will rate the credit risk of bonds.

\n

S&P says AAA, AA, A, BBB, BB, B, CCC, CC, C

\n

Moody’s has Aaa, Aa, A, Baa, Ba, B, Caa, Ca and C

\n

Bonds with ratings of BBB and above are investment grade

\n

These are measured as bands of probabilities.

\n

Question about Diversification

\n

By diversifying your portfolio into multiple assets, you decrease the credit risk. Then you can securitize the portfolio by splitting it into tranches, and then get the CDO.

\n

And then payment in the bonds

\n

You can have nonlinear dependence between variables, and tail dependence between variables which is conditional. Then the probability of default, what is the likelihood of this house being on fire if this other house is on fire? If the answer is not 0, then you have tail dependence.

\n

We will look at different models that measure this.

\n

Hazard Rates

\n

Also known as default intensity, this is the probability of default for a certain time period conditional on no earlier default.

\n

Unconditional default probability is from 0.

\n

We are given a table that has cumulative default rates.

\n

In order to calculate the probability that A defaults in the first year, you take the actual probability. Then if it’s the second year, we’re concerned with the cdf(2) - cdf(1).

\n

Then the probability of survival is 1 - this

\n

This is all unconditional default probability.

\n

Then the conditional probabilities are known as default intensities or hazard rates

\n

The unconditinal probability of default within a certain default vs the survival probability.

\n

The survival probability is V(t)

\n

$$\\lambda(t) \\Delta t = \\frac{V(t) - V(t + \\Delta t)}{V(t)}$$

\n

Then you get the ODE

\n

$$\\frac{dV(t)}{dt} = -\\lambda(t) V(t)$$

\n

$$V(t) = e^{-\\int_0^t \\lambda(t) dt}$$

\n

Then the survival probability can be found by integrating over the hazard rate. So if you have a specification of the hazard rate, you can find the cumulative survival probability.

\n

Then we can also discuss the cumulative default probability.

\n

We can use the CDS to determine the piecewise constant hazard rates and then construct the survival probabilities.

\n

So this is a calibration in which we can price the risk of default through CDS.

\n

First we will look at bonds.

","contentMarkdown":"## LMM\n\nThe **LIBOR Market Model** (or **Brace-Gatarek-Musiela Model**)\n\nis a model constructed in terms of the forward rates underlying caplet prices.\n\nHJM was defined as a process for the forwards.\n\nBasically the benefit was that it could model the entire term structure, as long as you input all the volatilities for the term structure.\n\nBut the problem is the calibration.\n\n\n\nNotation\n\n\\\\(t_k\\\\) kth reset date\n\nSo t would progress like t\\_0 \\= 0, t\\_1 \\= 0.25, t\\_2 \\= 0.5…\n\n\\\\(F_k\\\\) forward rate between k and k \\+ 1\n\nm(t) index for next reset date at time t.\n\nSo if t \\= 0.4, then m(t) \\= 0.5\n\n\n\nδ\\_k \\= t\\_k+1 \\- t\\_k\n\nSo basically you can imagine that you have your forward curve,\n\n\n\nxi: ξ(t) \\= volatility of F\\_k(t) at time t\n\n\n\nAssume that we have only one factor\n\nThis factor is the forward risk neutral process with respect to P(t, t\\_k+1)\n\nThen, our process\n\n$$\ndF_k(t) = \\\\xi_k (t) F_k(t) dW\n$$\n\nThe change is driven by the volatility, existing forward rate, and Brownian motion.\n\n\n\nApplying some change of numeraire, skipping some steps, you get\n\nthe rolling forward rate process (in notes)\n\n\n\nPrice of a ZCB at time t\\_i which is discounted, the ratio of that to t\\_i+1\n\n$$\n\\\\frac{P(t, t_i)}{P(t, t_{i+1})} = 1 + \\\\delta_i F_i(t)\n$$\n\nIf you apply the natural log, you get\n\n$$\n\\\\ln P(t, t_i) - \\\\ln P(t, t_{i+1}) = \\\\ln [1 + \\\\delta_i F_i(t)]\n$$\n\nThis is the relationship between the ZCB and the forward.\n\nThen if you apply Ito’s lemma… and equate coefficient of dW\n\n\n\nsomething long…\n\n\n\nThen you can take the substitution of the forward process.\n\n$$\n\\\\frac{dF_k(t)}{F_k(t)} = \\\\sum_{i= m(t)}^k \\\\frac{\\\\delta_i F_i(t) \\\\xi_i(t) \\\\xi_k(t)}{1+\\\\delta_i F_i(t)} dt + \\\\xi_k(t) dW\n$$\n\nThis is the process followed by the forward rate between t\\_k and t\\_k+1, in a risk-neutral world.\n\nIn the limiting case, when this interval becomes smaller, this converges to HJM\n\n\n\nThe idea is being able to calibrate this model.\n\nIt can be simplified with:\n\n## Simplified Model\n\nAssume that ξ\\_k(t) function only on the number of whole accrued periods between the next date and t\\_k.\n\n\n\nDefine Λ\\_i as the value of ξ\\_k(t) (volatility of forward) when there are i such accrued periods.\n\nThen the volatility of the forward can be redefined as\n\n\\\\(\\\\xi_k(t) = \\\\Lambda_{k-m(t)}\\\\) which is a step function\n\nAnd such values like Λ\\_i can be estimated from the volatilities used to value caplets in Black’s model.\n\nRecall that to value a caplet, we have\n\n$$\nL\\\\delta_k P(0, t_{k+1}) [F_k N(d_1) - R_k N(d_2)]\n$$\n\nblah blah blah d\\_1 and d\\_2\n\nIf we equate the variances between Black’s model and this, we get\n\n$$\n\\\\sigma_k^2 t_k = \\\\sum_{i=1}^k \\\\Lambda_{k-i}^2 \\\\delta_{i-1}\n$$\n\n\n\nExample in the slides for converting Black volatilities to Lambda, use to check\n\n\n\n## Implementation\n\nHow would I use this model to price a bond. If you observe the volatility, then you can price a bond under this?\n\n\n\nThis would be done via Monte Carlo simulation\n\n\n\n$$\n\\\\frac{dF_k(t)}{F_k(t)} = \\\\sum_{i=m(t)}^k \\\\frac{\\\\delta_i F_i(t) \\\\Lambda_{i - m(t)}\\\\Lambda_{k-m(t)}}{1 + \\\\delta_i F_i(t)} dt + \\\\Lambda_{k-m(t)} dW\n$$\n\nThen, by Ito’s lemma,\n\n$$\nd\\\\ln F_k(t) = \\\\left[\\\\sum_{i=m(t)}^k \\\\frac{\\\\delta_i F_i(t) \\\\Lambda_{i-m(t)} \\\\Lambda_{k-m(t)}}{1 + \\\\delta_i F_i(t)}\\\\right]dt + \\\\text{same thing as above}\n$$\n\nThen we can approximate the drift by approximating \\\\(F_i(t)\\\\) and \\\\(t\\\\)\n\nThen,\n\n$$\nF_k(t_{j+1}) = F_k(t) \\\\exp\\\\left[\\\\biggl(\\\\sum_{i=j+1}^k \\\\frac{\\\\delta_i F_i(t_j) \\\\Lambda_{i-j-1}}{1 + \\\\delta_i F_i (t_j)} - \\\\frac{\\\\Lambda^2_{k-j-1}}{2}\\\\biggr) \\\\delta_j + \\\\Lambda_{k-j-1} \\\\epsilon \\\\sqrt{d_j}\\\\right]\n$$\n\n\n\nSlide 8 has more calculations that I can use for testing.\n\nipynb file is provided\n\n\n\nEach approximation of the drift the forward rate remains constant.\n\nIf we define the rolling forward in the risk neutral world, it allows us to discount bond from one date to the next one.\n\nIn terms of simulation, what’s happening in the code is that if you assume that we want to simulate a zero curve with N accrued periods.\n\nOn each trial, we start with the forward rate at time 0, which is calculated from the initial zero curve.\n\nF\\_0(0), F\\_1(0), … F\\_N-1(0)\n\nThen we can use an approximation formula (described above) to calculate F\\_1(t\\_1), F\\_2(t\\_1)...\n\n\n\nLimitations of the theoretical model\n\n## Credit Risk\n\nLMM is a little complicated, this is just an introduction to the topic.\n\nCredit risk is the probability that a debtor will default on their debts. We have always assumption that the cash flow is 100% likely i.e. with treasury bonds.\n\n\n\nFor single-name derivatives, we will look at CDS and CDOs. Then we will look at portfolios. We are interested in expected loss, with a discrete number of names in the portfolio, to estimate the losses in order to price the portfolio correctly.\n\n\n\nRating agencies will rate the credit risk of bonds.\n\nS\\&P says AAA, AA, A, BBB, BB, B, CCC, CC, C\n\nMoody’s has Aaa, Aa, A, Baa, Ba, B, Caa, Ca and C\n\nBonds with ratings of BBB and above are investment grade\n\nThese are measured as bands of probabilities.\n\n\n\n### Question about Diversification\n\nBy diversifying your portfolio into multiple assets, you decrease the credit risk. Then you can securitize the portfolio by splitting it into tranches, and then get the CDO.\n\nAnd then payment in the bonds\n\nYou can have nonlinear dependence between variables, and tail dependence between variables which is conditional. Then the probability of default, what is the likelihood of this house being on fire if this other house is on fire? If the answer is not 0, then you have tail dependence.\n\nWe will look at different models that measure this.\n\n\n\n### Hazard Rates\n\nAlso known as default intensity, this is the probability of default for a certain time period conditional on no earlier default.\n\nUnconditional default probability is from 0\\.\n\nWe are given a table that has cumulative default rates.\n\nIn order to calculate the probability that A defaults in the first year, you take the actual probability. Then if it’s the second year, we’re concerned with the cdf(2) \\- cdf(1).\n\nThen the probability of survival is 1 \\- this\n\nThis is all unconditional default probability.\n\n\n\nThen the conditional probabilities are known as default intensities or hazard rates\n\n\n\nThe unconditinal probability of default within a certain default vs the survival probability.\n\nThe survival probability is V(t)\n\n$$\\\\lambda(t) \\\\Delta t \\= \\\\frac{V(t) \\- V(t \\+ \\\\Delta t)}{V(t)}$$\n\nThen you get the ODE\n\n$$\\\\frac{dV(t)}{dt} \\= \\-\\\\lambda(t) V(t)$$\n\n$$V(t) \\= e^{-\\\\int\\_0^t \\\\lambda(t) dt}$$\n\nThen the survival probability can be found by integrating over the hazard rate. So if you have a specification of the hazard rate, you can find the cumulative survival probability.\n\n\n\nThen we can also discuss the cumulative default probability.\n\n\n\nWe can use the CDS to determine the piecewise constant hazard rates and then construct the survival probabilities.\n\nSo this is a calibration in which we can price the risk of default through CDS.\n\nFirst we will look at bonds.","dataUrl":"https://sharifhsn.dev/api/posts/advanced-derivatives-week-08.json","date":"2025-03-20","datePublished":"2025-03-20","description":"The LIBOR Market Model (or Brace-Gatarek-Musiela Model)","site":"https://sharifhsn.dev","slug":"advanced-derivatives-week-08","source":"Advanced Derivatives","sourceUrl":null,"tags":["Fixed Income","LIBOR Market Model","Interest Rate Models","Hazard Rates"],"title":"LIBOR Market Model and Hazard Rates","url":"https://sharifhsn.dev/blog/advanced-derivatives-week-08/","version":"1","wordCount":1135}