Evaluating Diffusing Diffusivity Models in Quantitative Finance - A non-technical summary

Download the full research paper (PDF)

1. What does it mean?

Imagine that price moves as you driving. Volatility = tyre grip. Most days: dry road; now and then: black ice. Old models say grip never changes. Stochastic vol says the road surface changes texture at random. Our physics trick (“diffusing diffusivity”) lets the road’s conditions change randomly.

2. What I tried

• I ran two road‑texture models on VIX:

  • OU – dry, predictable tarmac that regains grip quickly after small bumps.
  • Lévy – same road but with hidden black‑ice patches that send the car skidding now and then.

• Compared to the usual classics (Heston, Hull‑White) using rolling forecasts.

3. What popped out

• OU: great for everyday bumps.
• Lévy: best for the crazy spikes.
• Heston’s OK; Hull‑White did not perform as well as others.

4. Use cases

• Routine hedging: OU / Heston.
• Panic planning: Lévy.

5. What I plan to do next

Flip the models into option‑pricing (risk‑neutral) land.

• Test on more than VIX such as SPX, variance swaps, maybe crypto.

TL;DR Volatility is slippery. Choose your model based on how much black ice you expect.

---

For further details, see the attached PDF or contact the author for discussion.

Share this post