20 October 2022 in Energy Reports
10 September 2025 in Journal of Business & Economic Statistics
1 January 2023 in SSRN Electronic Journal
We propose a time series forecasting method designed to effectively handle large sets of predictive signals, many of which may be irrelevant or short-lived over time. The method transforms predictive signals into candidate density forecasts via time-varying coefficient models, and subsequently combines them into an aggregate density forecast via time-varying subset combination. Our approach is computationally efficient because it uses online prediction and updating. Through extensive simulation analysis, we find that our approach outperforms competitive benchmark methods in terms of forecast accuracy and computing time. We further demonstrate the capabilities of our method in applications to forecasting aggregate daily stock returns and quarterly inflation.
23 September 2026 in SSRN Electronic Journal
Equal weighting is hard to beat. Yet this need not mean predictive information has little portfolio value. Comparing 1/N with an optimized portfolio bundles two decisions: which assets to hold and how much to allocate to each. We study the intermediate case in which predictive information selects a fixed-size equal-weight subset and, holding membership fixed, receives progressively greater authority over relative position sizes. In a stylized model, the predictive gain from selection is first order in the scale of predictive information, whereas attainable weighting gains are at most second order. In a Controlled-Information Experiment on realized U.S. stock returns, selection becomes valuable at weaker predictive alignment than weighting and remains more important when forecast magnitudes are attenuated accordingly. Feasible return forecasts reproduce the asymmetry pseudo-out-of-sample. Equal weighting can thus let predictive information determine membership while restricting its authority over position sizing; shrinkage regulates how much authority it receives.