A New Multivariate Predictive Model for Stock Returns

TL;DR

This paper introduces a new multivariate model incorporating financial ratios and consumption-wealth data to predict stock returns, demonstrating superior forecasting accuracy over benchmark models on extensive historical data.

Contribution

The paper develops and empirically tests a novel multivariate predictive model for stock returns that outperforms existing benchmarks in forecasting accuracy.

Findings

The model predicts future stock returns with significant accuracy.

It outperforms benchmark models in RMSE across multiple datasets.

The model is effective on both quarterly and monthly data from extensive historical periods.

Abstract

One of the most important studies in finance is to find out whether stock returns could be predicted. This research aims to create a new multivariate model, which includes dividend yield, earnings-to-price ratio, book-to-market ratio as well as consumption-wealth ratio as explanatory variables, for future stock returns predictions. The new multivariate model will be assessed for its forecasting performance using empirical analysis. The empirical analysis is performed on S&P500 quarterly data from Quarter 1, 1952 to Quarter 4, 2019 as well as S&P500 monthly data from Month 12, 1920 to Month 12, 2019. Results have shown this new multivariate model has predictability for future stock returns. When compared to other benchmark models, the new multivariate model performs the best in terms of the Root Mean Squared Error (RMSE) most of the time.