Working Paper: NBER ID: w12963
Authors: Torben G. Andersen; Tim Bollerslev; Dobrislav Dobrev
Abstract: We develop a sequential procedure to test the adequacy of jump-diffusion models for return distributions. We rely on intraday data and nonparametric volatility measures, along with a new jump detection technique and appropriate conditional moment tests, for assessing the import of jumps and leverage effects. A novel robust-to-jumps approach is utilized to alleviate microstructure frictions for realized volatility estimation. Size and power of the procedure are explored through Monte Carlo methods. Our empirical findings support the jump-diffusive representation for S&P500 futures returns but reveal it is critical to account for leverage effects and jumps to maintain the underlying semi-martingale assumption.
Keywords: volatility models; leverage effects; jumps; intraday data; nonparametric volatility measures
JEL Codes: C15; C22; C52; C80; G10
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
Cause | Effect |
---|---|
presence of jumps (C69) | return distribution characteristics (D39) |
leverage effects (G41) | semimartingale assumption (C51) |
jumps and leverage effects (C58) | standardized returns (G12) |
standardized returns (adjusted for jumps) (G12) | IID Gaussian (C26) |
model specification (jump-diffusion vs. pure diffusion) (C22) | observed return dynamics (C69) |
market microstructure noise (G14) | findings (Y40) |