Defining Monte Carlo Greeks for Local Volatility Models
Summary
The document discusses estimating delta, vega, and rho for a basket option priced by simulating several domestic equity indices under local volatility, forward interest rates, and dividend yields. It recommends keeping the same random samples when regenerating paths so that finite-difference comparisons are less affected by simulation noise. It also explains that Greeks become ambiguous when model inputs vary across the market: delta, for example, can hold the volatility skew fixed or move it with spot, producing different measures.
The choice of shocks should follow the intended use. For hedge equivalents, the option and hedging instruments should use consistent definitions; a straddle’s vega, for instance, should be calculated using the same volatility perturbation. For risk reporting aligned with Black-Scholes conventions, the answer suggests ordinary delta, a parallel shift to the implied volatility surface for vega, and a parallel shift to the spot rate curve for rho. These are conventions rather than universal definitions, and the document does not prescribe a single choice for every hedging or reporting purpose.
Key ideas
- Using common random samples across perturbed simulations reduces noise in finite-difference Greek estimates.
- Greek definitions can differ when spot, volatility skew, or other model inputs move together.
- Hedging use calls for consistent Greek definitions across the option and its hedge instruments.
- A Black-Scholes-style convention uses implied-volatility surface shifts for vega and spot-curve shifts for rho.
Tags
Full text
# How to estimate the greeks with a Monte Carlo simulation?
# How to estimate the greeks with a Monte Carlo simulation?
I am simulating the path of three indices to price a 1 year basket option.
All the indices are domestic, so there is no currency component.
At each time step I am using the local volatility determined using Dupire model, the forward interest rate and the forward dividend yield.
For now, I am OK resimulating paths to calculate the greeks.
- Delta $\Delta$: I am shocking $S$ at $t=0$ up and down and generating two new sets of paths. I am not sure how to calculate Vega and Rho.
- Vega $\nu$: Should I apply a parallel shock to the implied volatility surface and regnerate the local volatility surface or apply a parallel shock directly to the local volatility surface?
- Rho $\rho$: Should I apply a parallel shock to the spot rate curve and determine the forward rates or apply a parallel shock directly to the forward curve?
## Answer by Brian B (score 3)
https://quant.stackexchange.com/a/15624
First, please make sure that when you resimulate sample paths, you are keeping your underlying random samples constant, as in this answer.
For your delta, vega and rho there is some ambiguity in the definition of the greeks. Consider the simple case of delta in the presence of a skew $\sigma(K/S)$, and say that the underlying price right now is $S_0$. We might use
$$ \Delta = \lim_{dS \rightarrow 0} \frac{V(S_0+dS, \sigma(K/S_0))-V(S_0, \sigma(K/S_0))}{dS} $$ or we might use what is sometimes called the "total delta" $$ \Delta^{\textrm{tot}} = \lim_{dS \rightarrow 0} \frac{V\left(S_0+dS, \sigma\left(\frac{K}{S_0+dS}\right)\right)-V\left(S_0, \sigma\left(\frac{K}{S_0}\right)\right)}{dS} $$
In the case of rho and vega, as you note you have similar ambiguity about what model inputs should be changing infinitesimally as part of their definition.
Basically, in the presence of variations from the constant-parameter assumptions of the BSM model, the definition of the greeks is up to you.
What you want to think about is why you are calculating these greeks. The main use of greeks is to help us in hedging. Knowing what the delta is, for example, tells us how much underlying to trade (at a given moment) to lay off some of the hedged position risk.
The role of the greeks in this case is to provide some numbers that can quickly translate to equivalent quantities of liquid securities: underlying for delta, straddles for vega and bonds for rho. The main thing you need to be sure of is that your calculation of equivalents is consistent for the hedging instruments. For example, you would want to calculate your straddle vega using the same definition you used for your basket option vega.
The other role of greeks is to give a sense of portfolio risk to managers who are familiar with thinking of options portfolio risk in these terms. In that case, you want to keep it close to the BSM model, which generally entails
- Not using total delta
- Define vega with a parallel shock to the input implied vol surface
- Define rho with a parallel spot curve shock
## Answer by Ana María (score 1)
https://quant.stackexchange.com/a/15703
I suggest that you can use the Pathways - Euler Method to estimate vega, for sure with this method you can approximate the vega value.Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.