CHAPTER 2. LITERATURE REVIEW
2.6. Factors & Events affecting the weak form efficiency
2.6.4. Technological Advances and Algorithm Trading
In recent years, computerised trading systems have replaced mostly human involvement on trading floors in stock exchanges across the world, which is recognised as a significant reform of market trading, and in a study of stock exchanges in 120 countries information about electronic trading was analysed (Jain, 2005). These findings report that over the previous 25 years, 101 stock exchanges in these countries had introduced transparent and fully automated systems of electronic trading. The introduction of computerised trading systems has also initiated more research studies of whether this affects the efficiency of stock exchanges. Findings from a study of Singapore Stock Exchange show that after changing to an electronic system of trading, autocorrelations of returns reduced (Naidu and Rozeff, 1994) (Naidu and Rozeff, 1994), and another study reports that prices are more efficient when more investors use new technology for trading (Kondor, 2009; Oehmke, 2009; Chaboud et al., 2014). In addition, arbitrage is avoided, when quotes reflect new information quickly, and prices reflect information efficiency when traders who use algorithms are better informed and provide liquidity (Hoffman, 2014). Further, Weller (2017) suggested that undermining of financial markets and lower price knowledge were linked to algorithmic trading, while Boehmer et al. (2018) noted greater associated volatility.
The formation of prices in stock markets is positively affected by the advantage of speed of algorithmic traders compared to human traders, as they can respond to public information more rapidly (Biais et al., 2011; Martinez and Rosu, 2011). In addition, Dugast and Foucault (2018) proposed that market actors’ developed data may be transplanted by cutting-edge information technologies’ associated reduced expenditure during collation of data, as well as stronger data dissemination. So that stock prices are informationally efficient when many algorithmic traders operate. These findings are challenged by other studies that have analysed the introduction of new technology into stock exchanges, such as the Toronto Stock Exchange, as no material effect could be identified from the introduction of its computer- assisted trading system on its weak-form efficiency, with the use of rescaled range analysis (Freund et al., 1997). Another study reports that levels of market efficiency in the New York Stock Exchange showed no significant changes following the introduction of different forms of technological automation, which also applied rescaled range analysis (Freund and Pagano, 2000).
In a study conducted using the Dow Jones Industrial Average between 1896 and 1998, market efficiency evolution in the USA stock market was analysed, but recognising that during the past 50 years, information technology had been introduced and that investors had more accurate and up-to-date information to determine their investments, the market moved towards efficiency and levels of autocorrelation were reduced during this more recent period (Gu and Finnerty, 2002). This hypothesis was tested by using runs, serial correlation and VR tests for each of these years to compute first-order autocorrelation between daily returns. These findings report that from 1941 to the 1970s there were high levels of autocorrelation and the market was not weak-form efficient for most years in this period. Within this overall sample time period, the first 35 years reflect a very low level of autocorrelation with moderate fluctuation, and from 1896 to the 1970s there was greater autocorrelation, which is not consistent with the hypothesis of this study, so that for this 103 time period, the first three quarters there was no dominant effect of technological advances on levels of market efficiency.
These findings suggest that improvements in investors’ experience informed by enhanced information technology since the 1970s was due to a sharp decline in autocorrelation, and that since that time many stock markets have evolved and gained efficiency. However, this hypothesis was adopted in another study that focused on the NASDAQ Composite Index between the years 1971 and 2001 to investigate how levels of autocorrelation of daily returns had evolved by using the VR test, but this study reports that from 1991 to 2001 the stock market only demonstrates weak-form efficiency (Gu, 2004).
Findings from previous studies that compared traditional human systems of trading with electronic exchanges show that information efficiency of prices produce mixed theoretical predictions. Some studies report that front running of customers’ orders, insider trading and other abusive practices are reduced and information asymmetry experienced by some investors is reduced, because of better transparency, more publically available information, and lower trading costs with computerised trading that enhances liquidity(Pagano and Röell, 1996; Jain, 2005; Stoll, 2006). In addition, investors can compete with brokers that have exchange seats, barriers to market-making activity are reduced, stock prices are kept closer to equilibrium values by arbitrageurs, because of improved liquidity from higher volumes and lower trading costs. Będowska-Sójka’s (2018) new research evidenced that spreads are beneficially influenced by greater liquidity, itself stemming from high-tech updates to trading
infrastructure. Moreover, algorithmic trading’s extent and the presence of various trading locations’ rivalry may influence liquidity (Gresse, 2017). However, algorithmic trading can have various implications, with data collation being significantly and detrimentally impacted upon, whereas price efficiency is improved through algorithmic trading (Weller, 2018). Overall, stock price changes, the standard of order implementation and the trading conduct of particular stock holders may be influenced by algorithmic processes.
In contrast, other studies suggest that repeated face-to-face interactions enhance the reputations of brokers through lowering bid-ask spread and information asymmetry being mitigated with floor trading, and that compared with floor exchanges, spreads are wider with electronic exchanges (Benveniste et al., 1992; Venkataraman, 2001; Theissen, 2002). In addition, information asymmetry increases and informed traders continue to trade in the anonymous electronic stock market after trading traditionally ends, but evidence indicates that there is a rich exchange of information between humans on trading floors (Pirrong, 1996). Other findings argue that stock prices are driven from their basic values due to excessive uninformed trading, as a result of high turnover and low costs of trading (Shleifer and Summers, 1990). In addition, with or without a trading floor operating, information processing efficiency remains the same, and this current study focuses on whether new technology and the closure of human trading floors influence information efficiency of prices from an empirical perspective.