AL-Qadisiyah Journal For Administrative and Economic sciences

Predictive Expectile Regression with SCAD for High-Dimensional Financial Big Data: Evidence from the S&P 500

Document Type : Research Paper

Authors

Saif H. Raheem
Maryam Hussien Kuman

University of Al-Qadisiyah

https://doi.org/10.33916/qjae.2026.02567580
Abstract
This study proposes a Predictive Expectile Regression model with SCAD penalty to improve forecasting accuracy and variable selection in high-dimensional financial data. By integrating the SCAD regularization into expectile regression, the model effectively captures tail risk and manages multicollinearity in non-Gaussian environments.
Simulation experiments under normal, skewed, and heavy-tailed errors show that the proposed ER–SCAD consistently achieves the lowest MSE and MAED, confirming its robustness and sparsity compared with ER, ER–LASSO, and ER–EN. Application to S&P 500 daily data (2015–2024) further demonstrates superior predictive performance and numerical stability.
The findings indicate that combining expectile regression with SCAD penalty provides an efficient and interpretable framework for high-dimensional financial prediction and risk assessment.

Keywords

Expectile Regression
SCAD Penalty
High-Dimensional Data
Financial Forecasting
S&‌P 500
Crossmark
AL-Qadisiyah Journal For Administrative and Economic sciences
Volume 28, Issue 2
Summer 2026
Pages 567-581

Home

Guide for Authors

Submit Manuscript

Reviewers

Contact Us