Efficiency versus Robustness under Tail Misspecification: Importance Sampling and Moment-Based VaR Bracketing

Battle Count: 85

Highly relevant for quantitative risk management within trading desks and asset management firms. The paper directly addresses VaR estimation at regulatory confidence levels (99%, 99.5%), which is critical for position sizing, capital allocation, and risk limits in quantitative trading. The efficiency-robustness trade-off informs practitioners on when to trust model-based VaR vs. when to adopt conservative interval estimates. The stability diagnostics (ESS, weight concentration) are practical tools for monitoring simulation-based risk engines. However, the paper focuses on risk measurement rather than alpha generation or trading strategy development, and is limited to univariate equity returns.

The IS component (exponential tilting, likelihood ratios, bisection root-finding) is moderately complex but well-documented in the rare-event simulation literature. The DMM component requires solving linear programs over moment-feasible sets, which involves constructing discretization grids, enforcing moment constraints, and handling numerical feasibility issues at higher moment orders. The overall simulation framework (multiple replications, parameter sweeps over ν and α, diagnostic tracking) adds engineering complexity. The paper provides sufficient algorithmic detail for reimplementation but does not include code.

The paper provides a detailed simulation design with fixed random seeds, explicit parameter choices (ν ∈ {5,7,10}, α ∈ {0.990, 0.995}), common random numbers for IS bisection stability, and clear algorithmic descriptions for both IS and DMM. The use of publicly available QQQ data and standard distributions (Gaussian, Student-t) aids reproducibility. However, no explicit code repository or data link is provided in the extract. The DMM grid construction and specific implementation details (e.g., grid points, number of replications M) are partially specified.