David Ardia

Systemic risk forecasts are sensitive to the assumed dependence model, yet the space of plausible specifications is vast. We study the role of dependence uncertainty in out-of-sample forecasting through the lens of a real-time Bayesian learner. We propose a sequential approach for pair-copula constructions that updates beliefs over competing specifications and that forecasts from predictive mixtures. We disentangle different sources of uncertainty and evaluate their relative importance under misspecification. In an application to systemically important US banks, averaging over dependence specifications improves density forecasts, while gains in conditional tail-risk forecasts are confined to the deepest tail configuration and are linked to the dispersion across different tail-dependence shapes.

We test whether the intraday gamma-variance channel documented with proprietary data survives when net gamma is rebuilt from public open interest. Over five years of half-hour SPX observations, a one-standard-deviation increase in non-0DTE net gamma predicts next-bar realized variance about 36 percent lower, nearly twice the published coefficient, external validity of sign, not calibration. A dealer-inventory benchmark sharing an open-interest seed correlates at 0.99 for non-0DTE and 0.53 for 0DTE, so the bucket contrast carries the dilution reading. The association survives dynamic-error models, a HAR horse race, and predictive-ordering checks, reproduces on SPY and QQQ options, and machine learning confirms the linear form's adequacy for the gamma block.

We introduce a Human Capital Concern (HCC) index that captures public concern about financially material social risks. Analyzing U.S. mutual fund flows of retail investors, we document that, during periods of elevated human capital concern, flow sensitivity to fund social risk scores increases by 60%. Non-linear effects in the intensity and duration of the HCC index suggest that investors mainly respond to unexpected social news shocks. Funds with a stronger social score earn higher returns following high-HCC periods, with effects persisting for only two months. This pattern is consistent with investors updating their expectations about the future returns of social-risk assets.

Provides data and functions used in the book "Statistical Methods in Quantitative Finance" by David Ardia (2026).

Provides a comprehensive suite of portfolio spanning tests for asset pricing, such as Huberman and Kandel (1987) , Gibbons et al. (1989) , Kempf and Memmel (2006) , Pesaran and Yamagata (2024) , and Gungor and Luger (2016) .

Estimation, forecasting, simulation, and portfolio construction for regime-switching models with exogenous variables as in Pelletier (2006) .