Working Papers:
[NEW] Robust Estimation and Inference for Stock Return Predictability [Link]
with Hao Xu and Abderrahim Taamouti
(in submission)
Abstract:
We propose a sensitivity analysis approach to estimation and inference for stock return predictabil- ity with multiple persistent predictors and non-normal innovations. Unlike most of the existing literature, we allow the predictors’ residuals to be serially correlated, a case in which ordinary least squares is biased and the conventional T- and F-tests reject too often. We construct an artificial instrument following the IVX approach of Phillips and Magdalinos (2009) and embed it in a minimax sensitivity analysis framework that treats the degrees of endogeneity and persistence, the intercepts, and the serial correlation of the predictors’ residuals as locally misspecified. This delivers estimators and confidence intervals that remain valid in a neighbourhood of a reference model rather than at a single point. Monte Carlo simulations compare their coverage and length with those of ordinary least squares, the IVX test of Kostakis et al. (2015), and the Bonferroni interval of Campbell and Yogo (2006). An application to the S&P 500 index with the dividend-price and earnings-price ratios reassesses the robustness of predictability claims across subsamples and frequencies.
Portfolio Optimization with Growth-at-Risk: A Conditional Sharpe Ratio Approach [Link] [Slides]
with Abderrahim Taamouti
(under review, Journal of Financial Econometrics)
Abstract:
This paper studies portfolio choice under macroeconomic tail risk using the concept of Growth-at-Risk (GaR). We propose a conditional allocation framework that maximizes a Sharpe ratio evaluated in adverse macroeconomic states, thereby incorporating downside growth risk directly into portfolio decisions. The approach is implemented using forward-looking return scenarios generated by multivariate GARCH models, enabling a fully out-of-sample evaluation. Empirically, we consider U.S. sector ETFs and use the Chicago Fed National Activity Index (CFNAI) as a proxy for GaR. The results show that the resulting Conditional Sharpe Ratio (CoSR) portfolio delivers economically significant improvements in risk-adjusted performance relative to standard benchmarks. These gains remain robust after accounting for transaction costs, highlighting the value of conditioning portfolio allocation on macroeconomic downside risk..
Systemic Growth-at-Risk and Growth Spread Measures [Link] [Slides]
with Abderrahim Taamouti
(in submission)
Abstract:
This paper develops two forward‑looking macroeconomic metrics – Systemic Growth-at-Risk (GaR) Measures and Growth Spread Measures – to quantify cross‑country growth risks and net gains from regional economic integration. Systemic GaR extends the standard GaR to a multi‑country setting by capturing a country’s growth shortfall during severe downturns elsewhere. Growth Spread Measures compare expected gains in union‑wide expansions with expected losses in union‑wide recessions. To estimate these measures, we model the joint GDP growth distributions for EU countries using GARCH models with copula and DCC dynamics, with results favoring a factor‑based GARCH-Copula specification emphasizing common shocks and tail dependence. Empirically, net growth dividends vary: countries with stronger fiscal positions, greater trade openness, and lower exposure to global uncertainty capture larger gains and exhibit resilience. Core northern economies face stronger downside exposure to peripheral recessions, underscoring the dual role of economic unions as sources of shared growth and systemic risk.
Enhancing Portfolio Resilience to Systemic Risk: A Neural Network Approach [Link] [Slides]
with Abderrahim Taamouti
(second-round revise & resubmit, Journal of Empirical Finance)
Abstract:
This paper aims to enhance the classical mean-variance portfolio selection model by using machine learning techniques and accounting for systemic risk. The optimal portfolio is obtained through a three-step supervised learning model. Firstly, the Smooth Pinball Neural Network is employed to predict return distributions of individual assets and the market. Secondly, we use a copula to model dependence between assets and the market, based on which we simulate return scenarios. Lastly, we maximize an ex-ante Sharpe ratio conditioning on systemic events. We train our models using a comprehensive dataset of nearly 600 U.S. individual stocks over 37 years; however, the portfolio analysis and backtesting are conducted on several representative portfolio sets with sizes up to 50 assets. This design allows the models to learn from a broad cross-section of firms while focusing the evaluation on computationally tractable portfolio dimensions. Our set of predictors includes 94 firm characteristics, 14 macroeconomic variables, and 74 industry dummies. The backtesting results demonstrate the superiority of our proposed approach over popular benchmark strategies including a GARCH-based model. This outperformance is statistically significant and robust to the inclusion of transaction costs.