Paper: arXiv 2401.00188

Abstract

We propose a discrete-time econometric model that combines autoregressive filters with factor regressions to predict stock returns for portfolio optimisation purposes. In particular, we test both robust linear regressions and general additive models on two different investment universes composed of the Dow Jones Industrial Average and the Standard & Poor’s 500 indexes, and we compare the out-of-sample performances of mean-CVaR optimal portfolios over a horizon of six years. The results show a substantial improvement in portfolio performances when the factor model is estimated with general additive models.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced econometric techniques including autoregressive filters, generalized hyperbolic distributions, and GAMs for factor modeling, resulting in high math complexity. It is also empirically rigorous, featuring a six-year out-of-sample backtest on two major indices (Dow Jones and S&P 500) with clear performance metrics, though lacking code or public datasets.

Research Flowchart

  flowchart TD
  A["Research Goal"] --> B["Data Preparation<br/>DJIA & S&P 500 Indices"]
  B --> C["Model Estimation<br/>GAM vs. Robust Linear Regression"]
  C --> D["Forecast Generation<br/>Autoregressive Filters + Factor Regressions"]
  D --> E["Portfolio Optimization<br/>Mean-CVaR Optimal Portfolios"]
  E --> F["Out-of-Sample Evaluation<br/>6-Year Horizon"]
  F --> G["Key Finding<br/>GAM factor models substantially<br/>improve portfolio performance"]