Table 15: Results of regression analysis on AI affecting salary expenses
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Salary expenses
Salary expenses
Salary expenses
Salary expenses
Salary expenses
Salary expenses
Salary expenses
Salary expenses/
Nb of Employees
Salary expenses/
Nb of Employees
AI 2561.4*** 1234.6*** 320.5*** -92.30*** 21.02 6.066 8.491*** 1.977
(25.53) (11.37) (8.46) (-6.91) (0.94) (0.26) (5.76) (1.16)
Size 474.1***
(25.24)
Output 2.590*** 0.393*** 0.129 0.185 0.00346** -0.000753
Quality (75.87) (24.22) (0.72) (0.98) (2.71) (-1.74)
Liquidity 0.0694*** 0.00119** 0.00129 -0.00549 0.0000912*** -0.0000294
(96.00) (2.80) (0.22) (-1.06) (3.38) (-0.99)
Management 0.125*** 0.0465
Quality (199.14) (1.92)
2012 4.116 5.178**
(0.03) (3.19)
2013 12.11 7.937***
(0.10) (4.84)
2014 16.87 9.984***
(0.14) (6.03)
2015 22.81 12.68***
(0.19) (7.64)
2016 41.33 16.83***
(0.34) (9.99)
2017 60.37 20.01***
(0.49) (11.78)
2018 86.92 24.07***
(0.70) (14.01)
2019 111.8 29.49***
(0.88) (16.85)
Constant 55.72 -3379.9*** 227.4** 11.57 -10.38** 260.5*** 171.1** 66.11*** 80.52***
(1.94) (-24.36) (2.66) (1.12) (-2.87) (16.40) (3.92) 5.178**
Fixed Effect fyear fyear +
gvkey fyear +
gvkey fyear fyear +
gvkey
Clustering fyear +
gvkey fyear +
gvkey fyear +
gvkey
Observations 5540 5540 5540 5463 5463 5426 5426 4546 4483
R2 0.105 0.198 0.000 0.886 0.986 0.996 0.996 0.117 0.847
t statistics in parentheses
* p < 0.05, ** p < 0.01, *** p < 0.001
Notes: The first seven regressions aim to regress the salary expenses of the banks on the use of AI technologies with different control variables. The 8th and 9th specifications regress the ratio of salary expenses to the number of employees on AI. Column (1) is a regression of the salary expenses on AI. Column (2) is a regression of the salary expenses on AI with the control variable size. Column (3) is a regression of the salary expenses on AI with time fixed effect. Column (4) is a regression of the salary expenses on AI with the control variables output quality and liquidity. Column (5) is a
44
regression of the salary expenses on AI with the control variables output quality, liquidity, and management quality.
Column (6) is a regression of the salary expenses on AI with the control variables output quality and liquidity and with time and entity fixed effect and clustering. Column (7) is a regression of the salary expenses on AI with the control variables output quality, liquidity, and management quality and with time and entity fixed effect and clustering. Column (8) is a regression of the ratio of salary expenses to the number of employees on AI with the control variables output quality and liquidity and with time fixed effect. Column (9) is a regression of the ratio of salary expenses to the number of employees on AI with the control variables output quality and liquidity and with time and entity fixed effect and clustering. The industry category “portfolio management” is used as base level. The ownership category “subsidiaries of companies that are not publicly traded” is used as base level.
Table 15 represents regressions of salary expenses on AI. Specification (4) and (6) add the control variables output quality and liquidity. Specification (5) and (7) also include management quality.
Specification (6) and (7) use time and entity fixed effect and clustering. Regression (2) and (3) include respectively size control and time control. Size and time fixed effect do not explain much of the effect of AI on the number of employees as the r-squared it very low. Management Quality is highly correlated with salary expenses (0.993) therefore specification (4) and (6) seem to be the most accurate. Specification (4) shows that banks using AI spend 320.5 million dollars more in salaries that banks that do not use AI. The effect is statistically significant at the 1% significance level and the r-squared is remarkably high (0.954). Specification (4) and (8) show that the salary expenses and salary expenses per employee are increasing in output quality and liquidity and the coefficients are statistically significant at the 1% significance level. Also, banks using AI seem to pay higher salaries per employee than banks that do not use AI. It can be explained by the fact that these banks need additional labour to develop AI technologies and these employees have special skills which make them deserve a higher compensation than the average employee in a bank. This interpretation is also supported by specification (9) which shows that banks using AI pay 1,960 dollars a year more salary per employee. Thus, banks using AI employ more people and at a higher rate.
45
Table 16: Results of regression analysis on AI affecting equipment expenses
(1) (2) (3) (4) (5) (6) (7)
Equipment
Expenses Equipment
Expenses Equipment
Expenses Equipment
Expenses Equipment
Expenses Equipment
Expenses Equipment Expenses
AI 252.5*** 83.82*** 252.9*** 48.10*** 8.698 -3.391 -3.745
(19.01) (5.74) (18.98) (8.29) (1.76) (-0.93) (-0.82)
Size 55.80*** 8.018***
(20.70) (8.19)
Output Quality 0.180*** 0.0602*** 0.0184 0.0249*
(40.65) (13.55) (0.81) (2.66)
Liquidity 0.00676*** 0.000975*** -0.00101 -0.00151
(53.87) (5.67) (-0.95) (-0.74)
Management 0.00747*** 0.00282
Quality (38.69) (0.36)
2012 -4.122
(-0.26)
2013 -5.300
(-0.33)
2014 -7.716
(-0.47)
2015 -11.14
(-0.67)
2016 -6.894
(-0.41)
2017 -7.231
(-0.43)
2018 -2.440
(-0.14)
2019 -3.859
(-0.22)
Constant 8.176 -403.0*** 13.56 4.563** -56.08*** 35.53*** 27.93
(1.94) (-19.92) (1.18) (2.63) (-7.66) (10.22) (1.49)
Fixed Effect fyear fyear +
gvkey fyear +
gvkey
Clustering fyear +
gvkey fyear +
gvkey
Observations 2163 2163 2163 2154 2154 2132 2132
R2 0.143 0.285 0.144 0.857 0.920 0.986 0.987
t statistics in parentheses
* p < 0.05, ** p < 0.01, *** p < 0.001
Notes: The seven regressions aim to regress the equipment expenses of the banks on the use of AI technologies with different control variables. Column (1) is a regression of the equipment expenses on AI. Column (2) is a regression of the equipment expenses on AI with the control variable size. Column (3) is a regression of the equipment expenses on AI with time fixed effect. Column (4) is a regression of the equipment expenses on AI with the control variables output quality and liquidity. Column (5) is a regression of the equipment expenses on AI with the control variables output quality, liquidity, and management quality. Column (6) is a regression of the equipment expenses on AI with the control variables output quality and liquidity and with time and entity fixed effect and clustering. Column (7) is a regression of the equipment expenses on AI with the control variables output quality, liquidity, and management quality and with time and entity fixed effect and clustering. The industry category “portfolio management” is used as base level. The ownership category “subsidiaries of companies that are not publicly traded” is used as base level.
46
Table 16 represents regressions of equipment expenses on AI. Specification (4) and (6) add the control variables output quality and liquidity. Specification (5) and (7) also include management quality. Specification (6) and (7) use time and entity fixed effects and clustering. Management Quality is highly correlated with Equipment Expenses (0.946) therefore specification (4) and (6) seem to be the most accurate models and they have a high explanatory power as the r-squared equals 0.857 and 0.986 respectively. Specification (4) shows that banks using AI spend 48.10 million dollars more per year in equipment expenses than banks not using AI, this result is significant at the 1% significance level. Specification (4) also shows that equipment expenses are increasing in output quality and liquidity and the coefficients are statistically significant at the 1% significance level. We cannot conclude with certitude that AI has a positive effect on equipment expenses as specification (6) shows a negative coefficient. However, this coefficient is not statistically significant as we do not have enough observations of firms using AI. Also, it seems reasonable to think that firms using AI need to spend more funds on software and hardware which would increase equipment expenses. However, equipment expenses of firms using AI seem to decrease over time (figure 14, Appendix) while it is increasing for banks not using AI. It can be explained by the fact that the main part of equipment costs is incurred at the implementation of an AI technology, which could result in even higher profits in the future.
Table 17 (Appendix) represents regressions of operating expenses on AI. Specification (4) and (6) add the control variables output quality and liquidity. Specification (5) and (7) also include management quality. Specification (6) and (7) use time and entity fixed effect and clustering.
Management Quality is highly correlated with Operating Expenses (0.97) therefore specification (4) and (6) seem to be the most accurate models and they have a high explanatory power as the r-squared equals 0.904 and 0.987 respectively. Specification (4) shows that banks using AI spend 816.5 million dollars more per year in operating expenses than banks not using AI, this result is significant at the 1% significance level. Regression (6) also shows a positive coefficient for AI.
Specification (4) exhibits that operating expenses are increasing in output quality and liquidity and the coefficients are statistically significant at the 1% significance level. This can be explained by the two previous tables as banks using AI appear to have higher salary and equipment expenses, which are two types of expenses included in operating expenses.
The evidence supports the hypothesis stating that AI increases salary and equipment expenses.
However, output quality and liquidity have a positive effect on salary, equipment, and operating
47
expenses. Banks using AI have higher profits after costs (net income and EBIT) that banks not using AI but they pay higher costs. Therefore, the return drivers of banks using AI does not come from cost reduction.
Conclusion
A difference-in-differences model is built in order to compare the performance of banks using AI and banks not using AI. The results support the main hypothesis that the use of AI improves performance as the regression of EBIT on AI reveals a positive coefficient of 153.7 million dollars which is statistically significant at the 1% significance level. The r-squared of this model reaches 0.955 due to the use of the control variables size, industry, ownership, leverage, age and operating expenses. However, this master’s thesis does not allow to conclude with certitude that AI influences positively bank performance as the results are not statistically significant when using entity fixed effect. Even though the AI coefficient when applying entity fixed effect is not statistically significant, the differences-in-differences model used is strong as the banks observed are both the treatment and the effect. Additionally, a diff-in-diff model allows to obtain a causal estimate as it removes from the model the effect of variables that are constant over time, the effect of macroeconomic cycles and the influence of the sector and individual companies. Moreover, all the regressions of this paper without entity fixed effect are statistically significant at least at the 10%
significance level and they all have a high r-squared of above 0.8 as the control variables added seem relevant. The magnitude of the effect is high and probably overestimated, however it can be expected that the sign of the AI coefficient, when using entity fixed effect on a larger dataset, will remain positive for the independent variables: net income, EBIT, output, number of employees, salary expenses, furniture expenses and operating expenses. This master’s thesis gives an indication that the use of AI increases the performance of banks through a higher production, over the period 2011-2019 in the US. Banks using AI seem successful therefore the banks that do not use AI yet should consider it.
Discussion
The biases inherent to a diff-in-diff model still need to be considered; especially reversed causality omitted variable bias and the influence of extraneous factors on individual companies.
Furthermore, the AI variable is not a perfect proxy for the use of AI thus several banks are included in the control group when they should be in the treatment group. Even though the patent application date gives a clear date for the treatment, the estimate can be contaminated by an
48
anticipated or delayed use of the patent. These limitations of the model should not bias the estimator as the average treatment effect over the period is considered but studying a wider period would allow to obtain a more consistent estimator. Although the number of observations regarding all the banks in the dataset is high (5572), the number of observations regarding the banks using AI is relatively low (452) as we observe only 66 banks using AI and there is only one observation per fiscal year. Thus, increasing the number of years of observation would increase the significance of the results. It would be relevant to update this master’s thesis when the official version of the dataset WRDS US Patents will be released as it will cover the period before 2011. It would also be relevant to study the different subfields of artificial intelligence, different types of institutional investors and different geographical areas such as Europe, for which the European Patent Office (EPO) database can be used. Further studies should also look for another proxy or instrumental variable to account for AI and think about other possible omitted variables. This master’s thesis gives an insight on the use of AI in finance and further research are needed to confirm the results observed. This research paper is hopefully the first of a series of article dealing with Artificial Intelligence and Institutional Investors.
49
Appendix
Source code 1: web scrapping program
Notes: This document is a screenshot of a pdf file extracted from the web application Jupyter Notebook. It shows the first part of the webscrapping algorithm coded in python.
50 Source code 2: web scrapping program
Notes: This document is a screenshot of a pdf file extracted from the web application Jupyter Notebook. It shows the second part of the webscrapping algorithm coded in python. The resulting list is composed of only “True” values which means that all of the 66 banks have applied for a patent related to AI.
51
Table 12: Results of regression analysis on AI affecting investment gains
(1) (2) (3) (4) (5) (6)
Investment
Gains Investment
Gains Investment
Gains Investment
Gains Investment
Gains Investment
Gains
AI 53.58*** 18.57*** -8.050* 54.14*** -28.25 -26.42
(12.93) (3.73) (-2.03) (13.05) (-1.04) (-0.85)
Size 12.05*** -2.055* 3.946
(10.86) (-2.25) (0.63)
Leverage 0.000160 0.00681**
(1.12) (4.29)
Age 8.432*** 3.079**
(6.32) (2.86)
Book-to-Market ratio 0.0111 0.00614 0.00642
(0.14) (0.10) (0.80)
Operating Expenses 0.00961*** 0.0157
(11.46) (1.38)
Industry
522110.naics 86.73** 54.04*
Commercial Banking (3.04) (2.38)
522120.naics 94.81*** 52.81*
Savings Institutions (3.30) (2.31)
522292.naics 99.37* 51.12
Real Estate Credit (2.10) (1.37)
522310.naics 94.13 52.66
Mortgage and Other
Loan Brokers (1.42) (1.01)
522390.naics 67.26 41.41
Other Activities Related to Credit Intermediation
(1.63) (1.27)
Ownership
0.stko 18.43 -26.98
Publicly traded company
(0.65) (-1.20)
1.stko 2.178 -19.21
Subsidiary of a publicly
traded company (0.07) (-0.75)
3.stko 36.04 -29.67
Company that is publicly traded but not on a major exchange
(1.26) (-1.31)
Constant 1.393 -207.9*** -9.973 1.348 8.142** -54.95
(1.17) (-4.93) (-0.30) (1.14) (4.18) (-1.23)
Fixed Effect fyear gvkey+fyear gvkey+fyear
Clustering gvkey+fyear gvkey+fyear
Observations 5517 5310 5307 5517 5482 5269
R2 0.029 0.073 0.425 0.032 0.416 0.623
t statistics in parentheses
52
* p < 0.05, ** p < 0.01, *** p < 0.001
Notes: The six regressions aim to regress the investment gains of the banks on the use of AI technologies with different control variables. Column (1) is a regression of investment gains on AI. Column (2) is a regression of investment gains on AI with the control variables size, industry, ownership, leverage, age, book-to-market ratio.
Column (3) is a regression of investment gains on AI with the control variables size, industry, ownership, leverage, age, operating expenses, book-to-market ratio. Column (4) is a regression of investment gains on AI with year fixed effect. Column (5) is a regression of investment gains on AI with time and entity fixed effect and clustering. Column (6) is a regression of investment gains on AI with the control variables size, leverage, operating expenses, book-to-market ratio and with time and entity fixed effect and clustering. The industry category “portfolio management” is used as base level. The ownership category “subsidiaries of companies that are not publicly traded” is used as base level.
Table 17: Results of regression analysis on AI affecting operating expenses
(1) (2) (3) (4) (5) (6) (7)
Operating
Expenses Operating
Expenses Operating
Expenses Operating
Expenses Operating
Expenses Operating
Expenses Operating Expenses
AI 6526.5*** 3081.8*** 6532.5*** 816.5*** -92.03* 21.53 -11.05
(25.53) (11.16) (25.51) (9.22) (-2.19) (0.29) (-0.11)
Size 1229.8***
(25.77)
Output Quality 7.821*** 2.985*** 2.102* 2.224*
(98.04) (58.37) (2.35) (2.55)
Liquidity 0.155*** 0.00541*** -0.0213 -0.0360
(92.07) (4.04) (-0.88) (-1.28)
Management 0.274*** 0.101
Quality (138.92) (0.85)
2012 -146.2
(-0.51)
2013 -190.9
(-0.66)
2014 -226.4
(-0.78)
2015 -320.6
(-1.10)
2016 -254.4
(-0.87)
2017 -222.5
(-0.75)
2018 -135.0
(-0.45)
2019 -26.54
(-0.09)
Constant 152.5* -8755.9*** 323.5 28.98 -19.32 649.1*** 454.2
(2.09) (-24.84) (1.57) (1.19) (-1.70) (8.23) (2.25)
Fixed Effect fyear fyear +
gvkey fyear +
gvkey
Clustering fyear +
gvkey fyear +
gvkey
Observations 5552 5552 5552 5464 5464 5427 5427
53
t statistics in parentheses
* p < 0.05, ** p < 0.01, *** p < 0.001
Notes: The seven regressions aim to regress the operating expenses of the banks on the use of AI technologies with different control variables. Column (1) is a regression of the operating expenses on AI. Column (2) is a regression of the operating expenses on AI with the control variable size. Column (3) is a regression of the operating expenses on AI with time fixed effect. Column (4) is a regression of the operating expenses on AI with the control variables output quality and liquidity. Column (5) is a regression of the operating expenses on AI with the control variables output quality, liquidity, and management quality. Column (6) is a regression of the operating expenses on AI with the control variables output quality and liquidity and with time and entity fixed effect and clustering. Column (7) is a regression of the operating expenses on AI with the control variables output quality, liquidity, and management quality and with time and entity fixed effect and clustering. The industry category “portfolio management” is used as base level. The ownership category “subsidiaries of companies that are not publicly traded” is used as base level.
Figure 6: EBIT of the banks using AI and the banks not using AI
Notes: The blue line represents the average EBIT per year of the 66 banks using AI from 2011 to 2019. The red line accounts for the average EBIT per year of the banks that do not use AI. EBIT is expressed in million US dollars. Due to the scale, the average EBIT of the banks not using AI seems to be constant and equal to zero over the period.
However, figure 7 shows that it is not the case.
R2 0.105 0.201 0.105 0.904 0.979 0.987 0.987
54 Figure 7: EBIT of the banks not using AI
Notes: The blue line represents the average EBIT per year of the banks not using AI from 2011 to 2019. EBIT is expressed in million US dollars.
Figure 8: Output of the banks using AI and the banks not using AI
Notes: The blue line represents the average output per year of the 66 banks using AI from 2011 to 2019. The red line accounts for the average output per year of the banks that do not use AI. Output is expressed in million US dollars.
Due to the scale, the average output of the banks not using AI seems to be constant and equal to zero over the period.
However, figure 9 shows that it is not the case.
55 Figure 9: Output of the banks not using AI
Notes: The blue line represents the average output per year of the banks not using AI from 2011 to 2019.
Figure 10: Number of employees of the banks using AI and the banks not using AI
Notes: The blue line represents the average number of employees per year of the 66 banks using AI from 2011 to 2019. The red line accounts for the average number of employees per year of the banks that do not use AI. Due to the scale, the average number of employees of the banks not using AI seems to be constant and equal to zero over the period. However, figure 11 shows that it is not the case.
56
Figure 11: Number of employees of the banks not using AI
Notes: The blue line represents the average number of employees per year of the banks that do not use AI.
Figure 12: Salary expenses of the banks using AI and the banks not using AI
Notes: The blue line represents the average salary expenses per year of the 66 banks using AI from 2011 to 2019. The red line accounts for the average salary expenses per year of the banks that do not use AI. Salary expenses is expressed in million US dollars. Due to the scale, the average salary expenses of the banks not using AI seems to be constant and equal to zero over the period. However, figure 13 shows that it is not the case.
57 Figure 13: Salary expenses of the banks not using AI
Notes: The blue line represents the average salary expenses per year of the banks that do not use AI. Salary expenses is expressed in million US dollars.
Figure 14: Equipment expenses of the banks using AI and the banks not using AI
Notes: The blue line represents the average equipment expenses per year of the 66 banks using AI from 2011 to 2019.
The red line accounts for the average equipment expenses per year of the banks that do not use AI. Equipment expenses is expressed in million US dollars. Due to the scale, the average equipment expenses of the banks not using AI seems to be constant and equal to zero over the period. However, figure 15 shows that it is not the case.
58
Figure 15: Equipment expenses of the banks not using AI
Notes: The blue line represents the average equipment expenses per year of the banks that do not use AI. Equipment expenses is expressed in million US dollars.
Figure 16: Operating expenses of the banks using AI and the banks not using AI
Notes: The blue line represents the average operating expenses per year of the 66 banks using AI from 2011 to 2019.
The red line accounts for the average operating expenses per year of the banks that do not use AI. Operating expenses is expressed in million US dollars. Due to the scale, the average operating expenses of the banks not using AI seems to be constant and equal to zero over the period. However, figure 9 shows that it is not the case.
59
References
Abusalma Abdallah, 2021, “The effect of implementing artificial intelligence on job performance in commercial banks of Jordan”, Management science letters p.2061-2070.
Amer Awad Alzaidi, October 2018, “Impact of Artificial Intelligence on Performance of Banking Industry in Middle East”, IJCSNS International Journal of Computer Science and Network Security, VOL.18 No.10
Artz K. W., Norman P. M., Hatfield D. E., and Cardinal L. B., 2010, “A Longitudinal Study of the Impact of R&D, Patents, and Product Innovation on Firm Performance.” Journal of Product Innovation Management 27 (5): 725–740.
Athanasoglou Panayiotis, Georgiou Evangelia and Staikouras Christos, November 2008,
“Assessing output and productivity growth in the banking industry”. Bank of Greece.
Athanasoglou P. P. and Brissimis S. N., 2004, “The effect of mergers and acquisitions on bank efficiency in Greece” Bank of Greece, Economic Bulletin, 22, 7-34.
Αthanasoglou P. P., Brissimis S. N. and Delis M. D., 2008, “Bank-specific, industry-specific and macroeconomic determinants of bank profitability”, Journal of International Financial Markets, Institutions and Money, 18, 2, 121-136.
Berger A.N., DeYoung R., 1997. Problem loans and cost efficiency in commercial banks. Journal of Banking and Finance 21, 849–870.
Berger A.N., Humphrey D.B., 1997. Efficiency of financial institutions: International survey and directions for future research. Working paper, Board of Governors of the Federal Reserve System.
Bzdok, Altman, and Krzywinski, 2018, Statistics versus machine learning. Nature Methods15:
233–234.
60
Cerniglia Joseph A. and Fabozzi Frank J., November 2020, “Selecting Computational Models for Asset Management: Financial Econometrics versus Machine Learning—Is There a Conflict?”, The Journal of Portfolio Management, 47 (1) 107-118; DOI:
https://doi.org/10.3905/jpm.2020.1.184
Choi N., Fedenia M., Skiba H., Sokolyk T., 2019, “Portfolio Concentration and Performance of Institutional Investors Worldwide”, Journal of Financial Economics 2017 Vol. 123 No. 1
Choueifaty, Y. and Coignard, Y., 2008, “Toward Maximum Diversification” The Journal of Portfolio Management, 35, 40-51. https://doi.org/10.3905/JPM.2008.35.1.40
Frankenfield Jake, March 2021, “Artificial Intelligence (AI)”, Investopedia, accessed 24 June 2021 https://www.investopedia.com/terms/a/artificial-intelligence-ai.asp
Grand View Research, June 2021, “Artificial Intelligence Market Size, Share & Trends Analysis Report By Solution, By Technology (Deep Learning, Machine Learning, Natural Language Processing, Machine Vision), By End Use, By Region, And Segment Forecasts, 2021 – 2028”, Report ID: GVR-1-68038-955-5 , accessed 24 June 2021
https://www.grandviewresearch.com/industry-analysis/artificial-intelligence-ai-market
Grinaker Siw, 2019, “AI – what artificial intelligence is and is not”, accessed 24 June 2021 https://digileaders.com/ai-what-artificial-intelligence-is-and-is-not/
Gu Shihao, Kelly Bryan, Xiu Dacheng, 2020, “Empirical Asset Pricing via Machine Learning”, The Review of Financial Studies, vol 33(5), pages 2223-2273.
Hargreaves Carol, Chandana Vallaru and Reddy Vishnu, November 2017, “Machine learning application in the financial markets industry”, Indian journal of scientific research, Singapore
Harrison J.P. and Samaddar S., 2020, “Who Is Better at Investment Decisions: Man or Machine?”
The Journal of Ialth Management Winter 2020, 23 (3) 70-84; DOI:
https://doi.org/10.3905/jwm.2020.1.119
61
Harvey, Rattray, Sinclair and Van Hemert, 2017, “Man vs. Machine: Comparing Discretionary and Systematic Hedge Fund Performance”, The Journal of Portfolio Management Summer 2017, 43 (4) 55-69; DOI: https://doi.org/10.3905/jpm.2017.43.4.055
Kayid Amr, 2020, “The role of Artificial Intelligence in future technology”, DOI:10.13140/RG.2.2.12799.23201
Kissell Robert and Bae Jungsun, 2018, “Machine Learning for Algorithmic Trading and Trade Schedule Optimization”, The Journal of Trading Fall 2018, 13 (4) 138-147; DOI:
https://doi.org/10.3905/jot.2018.13.4.138
Kwan H., 2003, “Operating performance of banks among Asian economies: An international and time series comparison”, Journal of Banking & Finance Volume 27, Issue 3, March 2003, Pages 471-489 https://doi.org/10.1016/S0378-4266(02)00384-9
Lommers K., El Harzli O., Kim J., 2021, “Confronting Machine Learning With Financial Research”, arXiv:2103.00366 [pdf]
López de Prado M., 2018, “The 10 Reasons Most Machine Learning Funds Fail”, The Journal of Portfolio Management, Special Issue Dedicated to Stephen A. Ross 2018, 44 (6) 120-133; DOI:
https://doi.org/10.3905/jpm.2018.44.6.120
Maresch, Fink, Harms, 2016, “When Patents Matter: The Impact of Competition and Patent age on the Performance Contribution of Intellectual Property Rights Protection.” Technovation, 57–
58: 14–20.
Mester, L.J., 1996, “A study of bank efficiency taking into account risk-preferences”, Journal of Banking and Finance 20, 1024–1045.
Pandio, 2021, “How To Use Machine Learning in Finance”, accessed 24 June 2021, https://pandio.com/blog/machine-learning-in-finance/
62
Pensions & Investments, 2017, “80% of equity market cap held by institutions”, accessed 24 June 2021, https://www.pionline.com/article/20170425/INTERACTIVE/170429926/80-of-equity-market-cap-held-by-institutions
Prabhakar Mali Gaurav, 2020, “PROS AND CONS OF ARTIFICIAL INTELLIGENCE (AI) IN BANKING”, accessed 24 June 2021, https://www.smallbusinessbonfire.com/artificial-intelligence-banking/
Rapach, Strauss, Tu, Zhou, 2019, “Industry Return Predictability: A Machine Learning Approach”, The Journal of Financial Data Science Summer 2019, 1 (3) 9-28; DOI:
https://doi.org/10.3905/jfds.2019.1.3.009
Rasoulian, Gregoire, Legoux, and Senecal, 2017, “Service Crisis Recovery and Firm Performance:
Insights from Information Breach Announcements.” Journal of the Academy of Marketing Science 45 (6): 789–806.
Rotman David, September, 2015, “Who Will Own the Robots”, MIT Technology Review
Rudin C., 2015, “Can Machine Learning Be Useful for Social Science”, The Cities: An essay collection from the Decent City initiative, vol. 9: 86-90.
Santhanam R. and Hartono E., 2003, “Issues in Linking Information Technology Capability to Firm Performance.” Mis Quarterly 27 (1): 125–153.
Scanlon Luke, 2020, “Risk issues before considering AI for financial services”, accessed 24 June 2021, https://www.pinsentmasons.com/out-law/analysis/risk-ai-financial-services
Snow Derek, 2020, “Machine Learning in Asset Management: Part 2: Portfolio Construction—
Weight Optimization”, The Journal of Financial Data Science Spring 2020, jfds.2020.1.029; DOI:
https://doi.org/10.3905/jfds.2020.1.029
The informative, 2020, “7 ARTIFICIAL INTELLIGENCE SUBFIELDS EVERYONE SHOULD KNOW IN 2021”, accessed 24 June 2021, https://theinformativo.com/artificial-intelligence-subfields/#1_Neural_Networks