PREDICTABILITY OF TOTAL REVENUE IN GLOBAL LARGE COMPANIES FROM FINANCIAL AND ECONOMIC PARAMETERS

Abstract

The interplay between corporate earnings and market returns has been a pivotal area of study in management research for over half a century. Central to this discourse is the exploration of various financial performance metrics, among which revenue growth, stock price, and market capitalization have recently gained prominence in terms of predictability. This paper delves into the predictability of financial performance of large global corporations, emphasizing the increasing importance of revenue as one of the primary financial parameters. The study proposes that while revenue is fundamentally influenced by the company's product or service portfolio and market reception, other factors such as past revenue performance, financial statements line items and broader economic indicators (like GDP and inflation) also play a crucial role in forecasting revenue trends.

Our research encompasses a comprehensive analysis of approximately 2000 global companies over a 12-year period (2010-2022). The findings reveal that revenues can be predicted with a Mean Absolute Percentage Error (MAPE) accuracy of 15%. To achieve this, we employed a set of regression models incorporating Ordinary Least Squares (OLS), Artificial Neural Networks (ANN), and Bayesian Neural Networks (BNN). These methodologies were specifically chosen to elucidate both non-linear relationships and to provide 99% confidence intervals for each revenue prediction.  The models when applied for 10 sector-capital market groupings, showed that, the predictability of revenue for U.S. companies was markedly higher, with their accounting information proving especially pertinent in predictive models. The study also covers the impact of covid pandemic and its recovery on the revenue predictability.

Keywords: Revenue, ANN, BNN, GDP, MAPE, Earning, Ordinary Least Squares