Executive Stock Trading Plans

Executives and corporate leaders sometimes use preset trading plans to avoid accusations of insider trading. The algorithmic plans can get triggered by any number of different parameters, like specific dates, stock prices, or announcements from competitors. The only catch is that the plans can’t be based on inside information. When an executive makes a trade, he or she files a form with the SEC. The Wall Street Journal collected millions of these forms in an attempt to use reverse engineering to see if any of the plans were “opportunistic”—if they appeared to be taking advantage of market timing to increase profits.³²

In this case, the output was observable since the prices of all trades were known. What the WSJ was interested in was reverse engineering how timing information was being used by different plans as an input. Timing was observable, placing this scenario in Figure 1(A) where both inputs and out- puts are known. Essentially the WSJ had a sampled input-output relation- ship for each executive’s plan specified by the documents filed with the SEC. However, what it didn’t know was any of the other inputs that could have also been feeding into these plans. Even though trade forms must be filed, the details of the plans themselves are hidden, leaving the reverse engineer to guess what inputs the algorithm was likely using. Perhaps competitor or sector prices are also inputs to some plans, requiring consideration of each variable in turn to assess whether there were correlations suggesting a connection. This case underscores the challenge with trying to understand which inputs an algorithm pays attention to. There is a huge space of potential inputs, some of which are observable and some of which are not. Practically speaking you have to choose which inputs you want to investigate to see if they are relevant to the algorithm.