This paper quantifies the economic cost of Samsung Securities’s ghost stock blunder using the synthetic control method. As the financial world becomes more computer based, sudden price fluctuations caused by unintended human input errors appear to be happening more frequently. Samsung Securities, the Samsung conglomerate’s stock trading arm, mistakenly distributed shares worth over US$100 billion to its employees on April 6, 2018, due to a keyboard input error. The difficult process of finding a proper control group plagues comparative case studies. This study overcomes that hurdle by constructing a synthetic version of the event firm. It turns out that the company lost 12.17% (US$428 million) of its pre-event market capitalization, 3000 times the direct loss incurred, due to a fat finger mistake. The results highlight the importance of surveillance in curtailing unintended human errors such as incorrect keyboard inputs or mouse misclicks. Regulatory bodies should monitor financial institutions for their unintended errors and internal control systems.
Albert Einstein once said: “Computers are incredibly fast, accurate, and stupid. Human beings are incredibly slow, inaccurate, and brilliant. Together they are powerful beyond imagination.” However, inaccurate human beings coupled with stupid computers can cause fiascos in the financial market. As the financial world becomes more computer based, sudden price fluctuations caused by unintended human input errors, eg, incorrect keyboard inputs or mouse misclicks, appear to be happening more frequently. In 2001, a Japan-based trader inadvertently placed a sell order of 610 000 shares at six yen rather than six shares at 610 000 yen (see Economist (2009) for further details). This mistakenly placed order cost UBS Warburg approximately £71 million. In 2005, a trader at the Mizuho bank incorrectly sold 610 000 shares for one yen.11 1 See https://bit.ly/3g2U08v. In 2013, a fat finger error at Everbright Securities caused the Shanghai Composite Index to surge approximately 6% in two minutes (Gao et al 2016). In 2014, 617 billion US-dollars worth of Japanese stock was accidentally ordered and subsequently canceled.22 2 See https://bloom.bg/3mIqK8m.
On April 6, 2018, Samsung Securities, the Samsung conglomerate’s stock trading arm, accidentally paid stock dividends worth approximately 100 billion US dollars (30 times the company’s market capitalization) to its employees who joined the stock ownership plan.33 3 Samsung Securities is a financial investment company operating primarily in Korea, China, the United States, the United Kingdom, Japan and Hong Kong. This was an operational mistake, but the incident significantly damaged Samsung Securities’s reputation, as 16 employees sold the shares and made 144 million US dollars instantly before the company became aware of the error.
In this paper, I conduct a comparative case study in order to estimate the market capitalization of Samsung Securities without this fat finger blunder. A synthetic control approach shows that the company lost 12.17% (428 million US dollars) of its pre-event market capitalization, 3000 times the direct loss incurred, due to an incorrect keyboard input. Given the size of the economic cost of the blunder, it is vital for surveillance to curtail unintended human input errors.
This paper is one of the first to quantify the economic cost of a fat finger event. Pervasive in the literature is the belief that one may use an event study to quantitatively capture the impact of corporate practices. An event study is useful when one examines whether or not investors penalize public corporations for their behavior. This method is, however, silent on how much corporations are penalized, as researchers do not know how these companies would have been treated in the absence of such an event. In contrast to the literature, I employ the synthetic control method of Abadie and Gardeazabal (2003) to synthetically create an imaginary company, ie, a synthetic control, that is identical to Samsung Securities but without the fat finger incident. As such, we are able to compute the true treatment effect of this error.
This paper proceeds as follows. Section 2 briefly reviews the chronology of Samsung Securities’s ghost stock blunder. Section 3 distinguishes indirect costs from direct costs. I recap the synthetic control method and summarize my key findings in Section 4. Section 5 summarizes how this study contributes to the literature and provides a conclusion.
2 The blunder
Samsung Securities mistakenly paid the wrong type of dividend. This inadvertent error resulted in issuing ghost stocks amounting to 31.5 times the total outstanding shares of the company. The company intended to pay a KRW 1000 (approximately 0.94 US dollars) cash dividend per share on April 6, 2018. The company’s compliance unit failed to notice that the type of dividend chosen for the company’s employment stock ownership plan (ESOP) accounts was a “share”, ie, not “cash”, and 1000 ghost shares for each ESOP stock were issued. As a result, instead of a KRW 2 812 956 000 cash dividend (2.65 million US dollars), 2 812 956 000 ghost shares (105.52 billion US dollars) were inadvertently distributed to 2018 ESOP accounts at 09:30 on that day.
Although the company became aware of the blunder at 09:31, it took 54 minutes for the company to remove all of the ghost shares from the ESOP accounts. Meanwhile, 22 ESOP members submitted a sell order of 12 076 836 shares (13.5% of the total outstanding shares), of which 5 011 616 shares were traded. Consequently, Samsung Securities’s stock experienced a flash crash and rebounded back, as illustrated in Figure 1.44 4 The tick history for the day was obtained from Reuters Data Scope.
The regulatory body imposed direct sanctions on Samsung Securities. The Financial Supervisory Service of Korea ordered Samsung Securities to compensate 398 investors who traded during the flash crash in Figure 1 for their loss.
3 Direct versus indirect costs
Direct costs are mainly out-of-pocket expenses, eg, the drop in market capitalization and the sanctions imposed by the regulatory body. Indirect costs could have a substantially larger magnitude than direct costs, given that indirect costs are mainly opportunity costs (Cutler and Summers 1988). For example, a deteriorating reputation makes it more difficult for Samsung Securities to operate its business, and customer demand for the firm’s service product may decrease. Managers should work to resolve the issue; thus, the value of this time is another source of indirect costs. I thus proceed to estimate the true economic cost of the blunder, which considers both direct and indirect costs, using the synthetic control method.
4 Comparative case study
4.1 Analytical methods
The synthetic control method offers a systematic way to construct a comparison unit in comparative case studies. Unlike difference in differences approaches, this method can account for the time variation of the confounder by matching the treated object before the event. This method also allows econometricians to systematically develop control groups. Among others, Abadie and Gardeazabal (2003) used Catalonia and Madrid to construct an imaginary Basque country without terrorism and to quantify the true economic cost of terrorism in Spain. Abadie et al (2015) studied the economic impact of the German reunification in 1990 by creating a synthetic version of West Germany. Kahane and Sannicandro (2019) also adopted a synthetic control approach to examine the impact of gun law changes on suicide.
I conceptualize an imaginary Samsung Securities company without the blunder of April 6, 2018, against which we can compare the actual Samsung Securities company with that blunder. Let be the number of potential control companies, eg, securities companies in Korea other than Samsung Securities. Let be a () vector of the pre-blunder values of return predictors for Samsung Securities. Let be a () matrix that contains the values of the return predictors for the potential control companies. Let be a diagonal matrix whose diagonal elements denote the relative importance of the return predictors. Consider the below optimization problem:
where . In order to avoid overfitting, I require to be nonnegative. The solution to this optimization problem, , is characterized by the diagonal matrix . I choose such that the monthly returns trajectory of Samsung Securities is best matched with the synthetic control company defined by .
Let be a vector that contains Samsung Securities’s monthly returns. Let be a matrix for the return predictors of the potential control companies. Then,
Finally, one can come up with the weights for the synthetic control company by computing . It should be noted that there are infinitely many solutions, because if is a solution, then for any positive scalar , is a solution as well. To avoid the aforementioned issue, the Euclidean norm of is normalized to one.
4.2 Constructing a synthetic version of Samsung Securities
Building on the idea of the synthetic control method, I create an imaginary firm that replicates Samsung Securities without the blunder. Following Abadie and Gardeazabal (2003), I assume that the imaginary company is a linear combination of other securities companies in Korea.55 5 There are 12 listed securities companies in the original pool. Abadie et al (2015) recommended limiting the donor pool to those companies that are similar to the event company. Because the approach is data driven, Abadie et al (2015, p. 495) state that: “Researchers trying to minimize biases caused by interpolating across regions with very different characteristics may restrict the donor pool to regions with similar characteristics to the region exposed to the event or intervention of interest.” In particular, I apply the following filters to minimize potential biases:
investment banking and brokerage companies on the Korea Composite Stock Price Index66 6 I use the MSCI Global Industry Classification Standard (GICS) to choose investment banking and brokerage firms in Korea. In particular, I choose those firms with GICS sub-industry number 40203020.; and
the same credit rating as that of Samsung Securities.
These criteria leave me with six companies for controls (each firm’s ticker appears in parentheses):
NH Investment & Securities (005940.KS),
Shinyoung Securities (001720.KS),
Daishin Securities (003540.KS),
Meritz Securities (008560.KS),
Korea Investment Holdings (071050.KS) and
Kiwoom Securities (039490.KS).
As per return predictors, I consider the index return, size, long-term momentum and short-term reversal. I rule out the book-to-market ratio, as this ratio is distorted in financial industries. Since securities companies in Korea tend to focus on brokerage, mergers and acquisitions (M&A) and wealth management, I also take into account income from brokerage, income from M&A advisory services and income from wealth management. Given that the synthetic control method is, by nature, data driven, I run a univariate regression using seven variables in order to establish an economically significant association. Table 1 shows the results. Given the statistical significance together with the business environment that securities companies face, I surmise that the index return and income-related variables are economically meaningful and have predictive power to explain the time series patterns in the data.77 7 The data on the financial statement information was collected from the Financial Supervisory Service of Korea.
Now, let be a vector that represents the chosen predictive variables. Let be a matrix of the values of the four predictors of the six control firms. In addition, for all is a weight vector on a subset of controls. Since the synthetic control is determined by the weight vector, the choice of a synthetic version of Samsung Securities can be pinned down to the choice of . The weights are chosen such that the synthetic Samsung Securities best matches the real company before the blunder. Let be a diagonal matrix with nonnegative elements that represent the relative importance of the predictors. One can use his or her prior knowledge with regard to the importance of the predictive variables to choose . I follow Abadie and Gardeazabal (2003) and choose such that the real monthly return path before the blunder is best matched with that of the synthetic Samsung Securities. The synthetic unit is the one that solves for the optimization problem in (4.3):88 8 The MATLAB code is available on Jens Hainmueller’s website: https://stanford.io/3a6PxOh.
4.3 Inference about the economic cost of the blunder
The synthetic version of Samsung Securities is replicated by a linear combination of control firms. Table 2 shows the corresponding weight vector of the synthetic Samsung Securities. The synthetic Samsung Securities is based on the following five control firms: (1) NH Investment & Securities (0.2955), (2) Shinyoung Securities (0.1786), (3) Daishin Securities (0.2695), (4) Korea Investment Holdings (0.1718) and (5) Kiwoom Securities (0.0846). Each firm’s weight appears in parentheses.
|Synthetic Samsung Securities||0.2955||0.1786||0.2695|
|Synthetic Samsung Securities||0.0000||0.1718||0.0846|
Figure 2 visualizes the monthly returns for both the real Samsung Securities and the synthetic one. The two loci closely mimic each other before the blunder. The two graphs nonetheless show a dichotomous pattern in April 2018. The real Samsung Securities has a monthly return of 5.71% in April 2018, whereas the synthetic one has a monthly return of 6.45%. In May 2018, the real Samsung Securities’s monthly return converges to that of the synthetic one. The difference in the monthly holding period returns between the real Samsung Securities and the synthetic one in April 2018 is 12.17% of the company’s market capitalization, implying that the fat finger error caused the company to lose 12.17% of its pre-blunder market capitalization.99 9 This estimate includes both the direct and indirect costs of the corporate event, enabling us to overcome the traditional event study approach taken in Cloninger (1985), Reichert et al (1996), Gunthorpe (1997) and Zeidan (2013).
Samsung Securities faced a six-month suspension of its services to new investors as well as a fine of KRW 140 000 000 (127,000 US dollars). The result from the synthetic control method shows that the real economic cost of the fat finger problem was 12.87% of Samsung Securities’s pre-event market capitalization (428 million US dollars), 3000 times the direct loss incurred.
4.4 Placebo test
To assess the robustness of the results, I conduct a series of falsification tests by applying the synthetic control method to each of the six firms in the control group. Since the blunder was caused by an unprecedented operational mistake, it is unlikely to estimate the same treatment effect in the control firms. For every other firm in the original control group, I construct a synthetic version of the chosen control firm using the rest of the control pool. For each placebo firm, I compute the difference between the real firm and the synthetic one. This iteration procedure furnishes a gap distribution for the firms where no blunder took place. The weights on five placebo tests are presented in Table 3.
The falsification test confirms that the effect of Samsung Securities’s ghost stock blunder is driven by the event itself. Figure 3 plots the gaps in market capitalization for each placebo exercise. The figure illustrates that the gap between the real Samsung Securities and the synthetic one is located far below the implied distribution of the placebos. The locus of Samsung Securities plummets to the extreme negative domain in April 2018. It is thus safe to state that the economic cost of Samsung Securities’s blunder is indeed significant.
5 Concluding remarks
Samsung Securities’s fat finger error of April 6, 2018, offers a unique opportunity to gauge the impact of human input errors on a firm’s value. The difficult process of finding a proper control group plagues comparative case studies. I overcome this hurdle by constructing a synthetic version of the event firm. The difference in market capitalization between the synthetic Samsung Securities and the real company is estimated to be 428 million US dollars. The results highlight the importance of surveillance in curtailing unintended human errors such as incorrect keyboard inputs or mouse misclicks. Regulatory bodies should monitor financial institutions with regard to their unintended errors and internal control systems.
Declaration of interest
The author reports no conflicts of interest. The author alone is responsible for the content and writing of the paper.
This study was supported by the Research Program funded by Seoul National University of Science and Technology (SeoulTech).
- Abadie, A., and Gardeazabal, J. (2003). The economic costs of conflict: a case study of the Basque country. American Economic Review 93(1), 113–132 (https://doi.org/10.1257/000282803321455188).
- Abadie, A., Diamond, A., and Hainmueller, J. (2015). Comparative politics and the synthetic control method. American Journal of Political Science 59(2), 495–510 (https://doi.org/10.1111/ajps.12116).
- Cloninger, D. O. (1985). An analysis of the effect of illegal corporate activity on share value. Journal of Behavioral Economics 14(2), 1–13 (https://doi.org/10.1016/0090-5720(85)90015-4).
- Cutler, D. M., and Summers, L. H. (1988). The costs of conflict resolution and financial distress: evidence from the Texaco–Pennzoil litigation. Rand Journal of Economics 19(2), 157–172 (https://doi.org/10.2307/2555697).
- Economist (2009). The World of Business: From Valuable Brands and Games Directors Play to Bail-Outs and Bad Boys. Economist Books.
- Gao, M., Liu, Y.-J., and Wu, W. (2016). Fat-finger trade and market quality: the first evidence from China. Journal of Futures Markets 36(10), 1014–1025 (https://doi.org/10.1002/fut.21771).
- Gunthorpe, D. L. (1997). Business ethics: a quantitative analysis of the impact of unethical behavior by publicly traded corporations. Journal of Business Ethics 16, 537–543 (https://doi.org/10.1023/A:1017985519237).
- Kahane, L. H., and Sannicandro, P. (2019). The impact of 1998 Massachusetts gun laws on suicide: a synthetic control approach. Economics Letters 174, 104–108 (https://doi.org/10.1016/j.econlet.2018.11.004).
- Reichert, A. K., Lockett, M., and Rao, R. P. (1996). The impact of illegal business practice on shareholder returns. Financial Review 31(1), 67–85 (https://doi.org/10.1111/j.1540-6288.1996.tb00864.x).
- Zeidan, M. J. (2013). Effects of illegal behavior on the financial performance of US banking institutions. Journal of Business Ethics 112, 313–324 (https://doi.org/10.1007/s10551-012-1253-2).
Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.
You are currently unable to print this content. Please contact [email protected] to find out more.
You are currently unable to copy this content. Please contact [email protected] to find out more.
Copyright Infopro Digital Limited. All rights reserved.
You may share this content using our article tools. Printing this content is for the sole use of the Authorised User (named subscriber), as outlined in our terms and conditions - https://www.infopro-insight.com/terms-conditions/insight-subscriptions/
If you would like to purchase additional rights please email [email protected]
Copyright Infopro Digital Limited. All rights reserved.
You may share this content using our article tools. Copying this content is for the sole use of the Authorised User (named subscriber), as outlined in our terms and conditions - https://www.infopro-insight.com/terms-conditions/insight-subscriptions/
If you would like to purchase additional rights please email [email protected]