Financial market efficiency and return predictability: integrating econometric and machine learning evidence from three Middle Eastern stock markets
This study examines return dynamics and predictability across the Palestine Exchange (PEX), Amman Stock Exchange (ASE), and Tel Aviv Stock Exchange (TASE) using daily data from 2013 to 2024. It combines unit-root and long-memory tests, structural-break analysis, GARCH models, Granger-causality tests, and walk-forward out-of-sample forecasts. Price indices are non-stationary, whereas returns are stationary across all three markets. Evidence of long memory in raw returns is limited and specification-sensitive, whereas persistence is stronger in absolute and squared returns. GARCH estimates indicate substantial but mean-reverting volatility persistence, robust to Student-t innovations. BIC-selected tests identify asymmetric short-run predictive linkages from ASE to PEX, TASE to PEX, and TASE to ASE. However, Random Forest models do not robustly outperform market-specific majority-class benchmarks. Logistic regression provides stronger directional evidence only for ASE, while neither PEX nor TASE shows robust forecasting gains. The findings reveal heterogeneous return and volatility dynamics but provide no evidence of general or economically exploitable nonlinear return predictability. Institutional differences offer comparative context for these patterns but are not identified as causal determinants of market efficiency.
Authors
- Hasan Z. Nuseibeh (ORCID: https://orcid.org/0009-0002-3172-5218)
- Bashar Abdallah
- Ibrahim M. A. Awad (ORCID: https://orcid.org/0000-0001-7549-5261)
Institutions
- Al-Quds University (PS)
Publication Details
- Journal
- Cogent Economics & Finance
- Published
- 2026-11-20
- DOI
- https://doi.org/10.1080/23322039.2026.2760102
- Primary Topic
- Financial Markets and Investment Strategies
- Type
- article
- Field-Weighted Citation Impact
- 0.00