US Stock Market Efficiency: EMH or AMH?

Abstract

This paper adopts a forecasting method to shed light on efficiency of the US stock market using the S&P500 index in the past 30 years. Daily data is grouped into 30 subsamples. The modified Diebold-Mariano (MDM) test is used to compare the forecasting performances of the random walk model and the Autoregressive Integrated Moving Average (ARIMA) model to show which is a better description of the daily returns of the S&P 500. The results show that the ARIMA model persistently outperforms the random walk model, which suggests that the US market is not in a weak form of efficiency during the sample period. On the other hand, it is also shown that whether the US stock market is efficient or not somehow depends on how investors evaluate their losses.