A Debate on the Application of Algorithms on the Financial Market

By Christina Wei, Research Analyst at King’s Private Equity Club

The 4th industrial revolution has rearranged data processing and brought advanced analytical and cognitive techniques to the world: big data analytics, artificial intelligence and machine learning. The algorithm as the base of these innovative tools is worth to be mentioned. Defined by Cormen et al. (2009), it is a series of steps for transforming input into output and its purpose is to solve a particular defined computational problem. This core technology is being applied in multiple realms rapidly, especially in the capital market and is beneficial to a variety of stakeholders. Hence, in the following article, I will explore how the characteristics of the algorithm affect the efficiency of the financial market first followed by looking at its downsides, based on the existing research.

‘With the emergence of artificial intelligence technology, deep learning will be carried out repeatedly on the basis of massive data and human-computer interaction takes place, where the complex financial data can be displayed anytime and anywhere in the form of natural language, providing enterprises with better financial services and more intelligent financial decision support.’ — Zhao (2018 in Ren, 2021)

The primary stakeholders of the algorithm system in the financial markets are capital providers and analysts. The algorithm mainly functions as the guidance to support those stakeholders’ investment decision-making. This is because, at first, a vast majority of investors, especially the beginners, would show five biases (local bias, loss aversion, overconfidence, gender bias and racial bias) when they invest in the capital market.

What are the Five Biases?

1. Local bias: Investors lean towards selecting their local companies’ stocks.
2. Loss aversion: Investors prefer to avoid losses over achieving the same gains (The Decision Lab, N.D.).
3. Overconfidence bias: Investors tend to overestimate their abilities in investing (Aguilar, 2021).
4. Gender bias: Investors are more likely to inject more capital into male-led ventures. ‘Female-led ventures catering to male-dominated industries receive significantly less funding at significantly lower valuations than female-led ventures catering to female-dominated industries. In contrast, male-led ventures attain similar funding and valuation outcomes regardless of the gender dominance of the industries to which they cater.’ (Kanze et. al, 2020)
5. Racial bias: Venture places a high premium on pedigree and tenure; thereby there is a phenomenon ‘pattern matching’ that investors seek out the entrepreneurs who are similar to them. This perpetuates homogeneity when the entire top leader is white. (Grant, 2020)

Antretter et al. (2020), Cumming and Dai (2009), proved this in their research and demonstrated that those cognitive biases would affect investors’ investment performance by making irrational investment decisions. In contrast, the investigation carried out by Antretter et.al (2020) indicated that the investment algorithm performed better than the average investors as the algorithm’s portfolio selection was less affected by classical investment biases. Some machine learning algorithms even lead to unbiased, time-efficient, and socially and emotionally neutral results due to their capacity to process large amounts of data (Blohm et al. 2020). These findings indicate that it is the lack of emotion and less apprehension that makes the algorithm system assist the investors to consider all-sided and thereby make wise decisions since human beings are prone to be influenced by their prejudices.

In addition to that, according to Blohm et.al (2020), machine learning algorithms are not as overconfident as investors. The feature allows them to take into account and absorb all the available and useful information, alongside, balancing gains and losses. Consequently, they would reduce the chance of misjudgments and make precise selections in the decision making processes. Furthermore, according to the study conducted by Lee et al. (2018), global stock market networks based on machine learning techniques can provide reliable and accurate market direction estimations and risk predictions in market turbulent times. The results of the study also show that when it came to stock market predictions, the use of network indicators on medium-term investments (8-12 weeks) with short-term volatility outperformed short-term investments (1-4 weeks) (ibid.).

Algorithm trading also affects the stock market as a whole. The algorithms could be used to detect fraud in the financial services industry (Buchanan, 2019). A possible result of this would be a significant reduction of financial fraud and an improvement in financial asset security, thereby ensuring the stability of the financial market. In terms of the liquidity of the financial market, the results from Gsell (2008) indicate that the volume of algorithmic trading orders influences the market prices in a climbing trend. Moreover, lower market volatility is likely to result from lower latency at the same time. The finding from Hendershott, Jones and Menkveld (2011) suggests that algorithmic trading could lower the transactions’ costs and increase the amount of information contained in quotes. Therefore, liquidity providers can expect higher revenues although this tends to be a temporary phenomenon. They (Hendershott, Jones and Menkveld, 2011) also found that implementing algorithmic trading could lead to positive spillover effects in other markets by improving linkages between markets.

On the other side, Yadav (2015) argues that securities markets are experiencing an inefficient allocation of capital due to algorithmic trading. This is because ‘algorithmic markets evidence a systemic degree of model risk that stylized programming and financial modelling fails to capture the messy details of real-world trading (ibid., 2015)’. It means the algorithm system is set up by the traders in advance, however, the traders are not able to forecast and detect the risks in the market comprehensively and accurately since the uncertainties and instabilities mutually exist in the financial market; therefore, those limitations prevent algorithms from handling exceptional circumstances that beyond their programming. Further, the study mentions that the model risks raised by the algorithmic system could cause serious informational deficits. Compared with informed traders, the algorithmic traders would gain other people’s information to save time and money, as well as receive earnings without any effort, which could lead to less incentive for informed traders to invest in long-term research and analysis.

As analysed above, even with some shortcomings, algorithms and the techniques derived from algorithms are clearly gaining a competitive edge in the investment and financial market.

References:

Aguilar, O. (2021) Fundamentals of behavioral finance: overconfidence bias. Available at: https://www.schwabassetmanagement.com/content/overconfidence-bias (Accessed: 26 January 2022).

Antretter, T., Blohm, I., Siren, C., Grichnik, D., Malmstrom, M. and Wincent, J. (2020) ‘Do algorithms make better — and fairer — investments than angel investors?’, Harvard Business Review, 2 November. Available at: https://hbr.org/2020/11/do-algorithms-make-better-and-fairer-investments-than-angel-investor s (Accessed: 26 January 2022).

Blohm, I., Antretter, T., Sirén, C., Grichnik, D. and Wincent, J. (2020) ‘It’s a peoples game, isn’t it?! A comparison between the investment returns of business angels and machine learning algorithms’, Entrepreneurship Theory and Practice. doi: 10.1177/1042258720945206

Cormen, T.H., Leiserson, C.E., Rivest, R.L. and Stein, C. (2009) Introduction to algorithms. 3rd edn. Cambridge: MIT Press.
Cumming, D. and Dai, N. (2010) ‘Local bias in venture capital investments’, Journal of Empirical Finance, 17(3), pp. 362-380. doi: 10.1016/j.jempfin.2009.11.001

Grant, J. (2020) ‘Investing in racial diversity: a call to action to the venture capital community’, Nature Biotechnology, 38(8), pp. 925-927. doi: 10.1038/s41587-020-0624-y Gsell, M. (2008) Assessing the impact of Algorithmic Trading on markets: a simulation approach (No. 2008/49). Frankfurt: Goethe University Frankfurt, Center for Financial Studies.

Hendershott, T., Jones, C.M. and Menkveld, A.J. (2011) ‘Does algorithmic trading improve liquidity?’, Journal of Finance, 66(1), pp. 1-33. doi: 10.1111/j.1540-6261.2010.01624.x Kanze, D., Conley, M.A., Okimoto, T.G., Phillips, D.J. and Merluzzi, J. (2020) ‘Evidence that investors penalize female founders for lack of industry fit’, Science Advances, 6(48), p. eabd7664. doi: 10.1126/sciadv.abd7664

Lee, T.K., Cho, J.H., Kwon, D.S. and Sohn, S.Y. (2019) ‘Global stock market investment strategies based on financial network indicators using machine learning techniques’, Expert Systems with Applications, 117, pp. 228-242. doi: 10.1016/j.eswa.2018.09.005
Ren, J. (2021) ‘Research on financial investment decision based on artificial intelligence algorithm’, IEEE Sensors Journal, 21(22), pp. 25190-25197. doi: 10.1109/JSEN.2021.3104038

The Decision Lab (2020) Why do we buy insurance? Available at: https://thedecisionlab.com/biases/loss-aversion/ (Accessed: 27 January 2022).
Wood, B. (2019) Artificial intelligence in finance: new landscaping report from The Alan Turing Institute. Available at: https://www.turing.ac.uk/news/artificial-intelligence-finance-new-landscaping-report-alan-tur ing-institute (Accessed: 27 January 2022).

Yadav, Y. (2015) How algorithmic trading undermines efficiency in capital markets. Vanderbilt Law Review, 68(6), pp. 1607-1671.