Replication Dataset and Code for: "Defying the Forecast Combination Puzzle: Bridging Financial Theory and Stock Price Prediction"

This dataset accompanies the study titled "Defying the Forecast Combination Puzzle: Bridging Financial Theory and Stock Price Prediction." It includes three CSV files — combined_data.csv, training_data.csv, and testing_data.csv — used to evaluate and compare 15 univariate, multivariate, and forecast combination methods across three U.S. stock indices. Variables span index closing prices and economic indicators from 1997 to 2024, specifically: • SP500: S&P 500 Index• NASDAQ: NASDAQ 100 Index• DOW JONES: Dow Jones Industrial Average• T10Y2Y: 10-Year Treasury Constant Maturity Minus 2-Year Treasury Constant Maturity (Yield Curve Spread)• DTB3: 3-Month Treasury Bill: Secondary Market Rate• DCOILWTICO: West Texas Intermediate (WTI) Crude Oil Prices• VIXCLS: CBOE Volatility Index (VIX)• DEXJPUS: Japanese Yen to U.S. Dollar Spot Exchange Rate• DEXCHUS: Chinese Yuan to U.S. Dollar Spot Exchange Rate• USEPUINDXD: U.S. Economic Policy Uncertainty Index• BAMLC0A1CAAA: ICE BofA AAA U.S. Corporate Index Option-Adjusted Spread• WLEMUINDXD: U.S. Equity Market-related Economic Uncertainty Index In addition to the data, this repository includes three R scripts that support close to full replication of the study’s methodology and results: • Nasdaq Code.R – Forecasting and evaluation procedures for the NASDAQ Index• SP500 Code.R – Forecasting and evaluation procedures for the S&P 500 Index• Dow Jones Code.R – Forecasting and evaluation procedures for the Dow Jones Index Data sources: Yahoo Finance, Stooq, and the Federal Reserve Economic Data (FRED). Please cite this dataset using the DOI provided. All files are released for replication purposes.