Are Markets Truly Efficient? Experiments Using Deep Learning Algorithms for Market Movement Prediction

Sanjiv Das,Robbie Culkin,Karthik Mokashi

doi:10.3390/a11090138

Abstract

We examine the use of deep learning (neural networks) to predict the movement of the S&P 500 Index using past returns of all the stocks in the index. Our analysis finds that the future direction of the S&P 500 index can be weakly predicted by the prior movements of the underlying stocks in the index, but not strongly enough to reject market efficiency. Decomposition of the prediction error indicates that most of the lack of predictability comes from randomness and only a little from nonstationarity. We believe this is the first test of S&P 500 market efficiency that uses a very large information set, and it extends the domain of weak-form market efficiency tests.

Highlights

The benefits of free markets derive from their “efficiency”, i.e., the idea that efficient markets accurately and rapidly impound new information into prices via trading
This paper offers a small experiment to assess the potential of deep learning algorithms to predict markets
Unlike existing tests of weak-form market efficiency, which use a single return series to test for autocorrelation, this new approach expands the information set by a huge order of magnitude, i.e., by using return data from all stocks in the index for the past 30 days

Summary

Introduction

The benefits of free markets derive from their “efficiency”, i.e., the idea that efficient markets accurately and rapidly impound new information into prices via trading. Market efficiency implies no information remains to be impounded in the price, all new information is a surprise and is inherently unpredictable. In a follow up article more than three decades later, Fama [2] revisited the evidence on market efficiency, and found in favor of its broad existence using effective methods, especially event studies. These ideas have existed in the finance literature for a few decades. There has been a machine-learning-based literature [3,4,5,6,7,8,9,10,11,12,13] that has explored market prediction with some success, though the jury is still out on whether markets are inefficient and easy to beat

Objectives

Methods

Conclusion