Big data is playing an increasingly important role in the financial industry.
Although this may lead to greater returns, this may lead to biased decisions being made.
More will need to be done going forward in order to limit the negative effects which the use of big data has.
Over the years, we have seen innovation of technology change the ways in which we live, work and spend. Nowhere is this effect more evident than when we look at the impact of machine learning and big data on the financial industry. In fact, the financial industry has been ahead of the curve when it comes to the use of big data and quantitative methods, with some firms, such as Bridgewater Associates and D.E and Shaw, using these types of methods since the 1980s.These methods are commonly used in the industry in order to price and trade securities, as well as measure risk.
Big data can be defined as data which is in high volume, high velocity and high variety. One of the drivers of the sheer amount of data available to financial firms was the electrisation of the New York Stock Exchange (NYSE) in the 1970s. With this, came the introduction of algorithmic trading, which allowed trades to occur at a much faster rate than possible with solely humans involved in the process. The higher frequency of trades resulted in more and more historical data available to firms in order to modify their algorithms to decrease risk. Machine learning has also had a growing impact in the area of asset management. According to research done, the majority of hedge fund managers use machine learning in their investment decisions. Although most funds still have humans involved in the decision-making process, machine learning still does have an impact on how these firms decide to structure their investment portfolios, as well as assist in generating ideas and risk management. According to the OECD, there is some data to suggest that some firms which use artificial intelligence and machine learning are able to outperform conventional hedge funds in terms of returns.
Big data can be defined as data which is in high volume, high velocity and high variety.
Although the uses of big data and machine learning in the financial industry have been successful, as with all things, there are still some risks. One major risk is of biases. The use of machine learning can exasperate human biases, if the training model used includes these same biases. Because the financial services industry does not solely revolve around making profits, but also serving customers, this leads to large issues. As mentioned by Karen Hao, an editor for the MIT Technology Review, problems can occur in the preparation of the data, the framing of the problem or collecting of the data. For example, with firms such as hedge funds which have the main focus of maximising profits, this may lead to “predatory behaviour”, such as investing in subprime mortgages or investing in firms which are on paper profitable, but also engage in morally and ethically dubious actions. These types of decisions can have large consequences for the industry as a whole, which not only supports the global economy, and has already had a problem with public trust since the 2008 financial crisis. It is therefore essential to have humans still be involved in the decision-making process in these firms using machine learning.
Other biases as mentioned by Karen Hao and theOECD which may affect the financial services industry is in the area of credit. If the data used in the training set for the machine learning programming contain human biases, these biases will be transferred to the algorithm being used. This can exasperate already existing problems for credit disparity based on race, ethnicity and gender.
Overall, the use of big data and machine learning is and will continue to have an impact on the finance industry, whether positive or negative. Going forward, there will need to be caution, as well as innovative thinking into how these quantitative methods should be used, and also an emphasis on the importance of still having humans involved in deliberation of these decisions, in order to limit the negative impacts, the implicit biases in the data has.
However, hopefully in the future, firms will be able to minimize or completely eliminate thisbias, so that all the benefits of using these methods will outweigh the risks and disadvantages created.