The role of technological changes in foreign-exchange market inefficiency

Abstract

To study market inefficiency which comes from rapidly developing software and technological progress in whole, we introduce technological bias in the exchange-rate market. The idea of technological bias emerges from the fact that recently innovative approaches have been used to solve trading tasks and to find the best trading strategies. If we consider the same pace of technological progress of trading infrastructure and computational tools along with software in the coming years, the traders who are able to adapt to this technological changes will get more profitable trading solutions than those who will require more time to adjust. Described situation displays market inefficiencies that challenge the idea of the efficient market theory, but are in line with adaptive market hypothesis. To support our suggestion about technological bias we compare the performance of deep learning methods, shallow neural network with ARIMA method and random walk model using daily closing between three currency pairs: Euro and US Dollar (EUR/USD), British Pound and US Dollar (GBP/USD), and US Dollar and Japanese Yen (USD/JPY). The results reveal the convincing accuracy of deep neural networks comparing to the other methods demonstrating the capacity of new computational methods based on evolving software. Shallow neural network outperform random walk model that confirms the idea of market inefficiency, but cannot surpass ARIMA accuracy significantly.