Abstract
This paper analyzes a wide range of flexible drift and diffusion specifications of stochastic-volatility jump–diffusion models for daily s&p 500 index returns. We find that model performance is driven almost exclusively by the specification of the diffusion component whereas the drift specifications is of second-order importance. Further, the variance dynamics of non-affine models resemble popular non-parametric high-frequency estimates of variance, and their outperformance is mainly accumulated during turbulent market regimes. Finally, we show that jump diffusion models yield more reliable estimates for the expected return of variance swap contracts.
Original language | English |
---|---|
Pages (from-to) | 85-103 |
Journal | Journal of Banking & Finance |
Volume | 83 |
DOIs | |
Publication status | Published - Oct 2017 |
Keywords
- Stochastic volatility
- Jump-diffusion models
- Bayesian inference
- Markov chain Monte Carlo
- Particle filter
- Deviance information criteria
- Realized variance
- High-frequency returns
- Variance risk premium