Disclosure: I may earn affiliate revenue or commissions if you purchase products from links on my website. The prospect of compensation does not influence what I write about or how my posts are structured. The vast majority of articles on my website do not contain any affiliate links.
My college roommate Alek Mosholt was a professional powerlifter and an amateur value investor. When Chipotle stock went into free fall following the Boston College norovirus debacle, his models indicated that the stock was fundamentally undervalued. He purchased it near its low and realized a profit when it corrected slightly in early March 2016. This incident–the most notable in recent memory–drove home an important point. Investors will react adversely to food safety incidents.
Again this week, Chipotle stock dropped after what was just confirmed as another norovirus outbreak. Food contamination and other sanitation issues (e.g. the Domino’s video that necessitated a comprehensive restructuring of the company) cause investors to lose confidence and sell stock. Maybe my buddy Alek bought again this time, or maybe he stayed away. The point is that we know the price will fall, but don’t know by how much or over what period of time. By the time a mainstream news network breaks the story, it might already be too late to make money. Capital markets are incredibly efficient.
When any news–positive or negative–surrounding a publicly traded company is released, it is digested by the markets almost immediately. While you leisurely read Bloomberg during lunch and consider making trading decisions, you might not know that an algorithm analyzed the sentiment of the article and traded on that information within (milli-; micro-) seconds of it becoming available. Trying to assess this information for yourself and then trade on it isn’t necessarily going to bring you to ruin, but it’s a loser’s game. Sometimes you’ll win, but, in the long run, you’ll be beat by other players who are smarter, faster, and more disciplined than you.
So, if it’s not consistently profitable to trade on information once it becomes public, how can it be possible to profit from stories about food contamination? Here’s a hypothetical situation. Let’s say you live in the rural Northern Tier of Pennsylvania (a really beautiful place!) and one day you look out over the rolling hills and see a United Airlines airplane flying low in the sky. It’s also on fire. There are no major airports near you, and you suspect you might be one of the first people to be witnessing what will probably be called a tragedy on the news the next morning. After you call 911–and as terrible as this sounds–you can profit immensely by then leveraging your life savings and selling United stock.
Is it insider trading when an ordinary person happens to witness a pre-news event? I researched this topic while writing this article, and I couldn’t find any examples of it having been prosecuted presumably because 1. it’s so exceptionally rare and because 2. insider trading specifies a “violation of some duty of trust” which finding yourself in such a situation doesn’t explicitly entail.
Many of the hypotheticals now running through your mind are probably plausible, but won’t happen to you with any regularity. To form a trading system around them and be prepared to deploy a significant amount of capital would be useless. But what if we can detect pre-news events using different resources on the internet?
First, I checked Twitter. Might someone take to Twitter to complain about getting sick at a chain restaurant? If enough people do this in a tight geographical area, it could be a signal that a food contamination news story is developing. Enough in this case meaning significantly more complaints than during a regular day or week. The usefulness of Twitter is that it’s instant, easy to filter by geographical area, and sparsely moderated. This search could, and, ideally, should, be automated, but for this article I just decided to do an advanced search by hand. By searching for the words ‘Chipotle’ or ‘sick’ in tweets within 15 miles of Boston from 11/24/15 and 12/08/15 (when the BC scare became publicized) there were several dozen tweets. However, the only ones making any reference to the Chipotle food contamination story were made by Boston-based news websites after everyone knew about it.
Next, I tried Yelp. When people have negative restaurant experiences, they love to write Yelp reviews. However, what makes Yelp difficult as an information source is that it is moderated significantly more than Twitter. This is especially true considering that people who get sick sometimes go on graphic, profanity-laced tirades that get filtered soon after being posted. The other problem with Yelp reviews is that they take more effort to write than tweets. They might not get posted for days or weeks after a negative experience, whereas if someone was going to post a tweet they’d be more likely to do it immediately. Also, once a Yelp business is considered in the news, Yelp will aggressively moderate the existing and new content. I checked out the reviews of the location at the epicenter of the current outbreak, and found only two reviews mentioning food contamination, but, as you might imagine, they were posted well after the news stories.
There is one website that’s been credited for signaling the most recent Chipotle crisis. Iwaspoisoned.com isn’t for writing food reviews, it’s for reporting when and where you got sick. With this express purpose, it’s likely that other investors are already trawling the data published by the site to determine if there may be a story developing. Such a specialized resource presents two problems. The first is that, as it gets more news coverage, use of the website will grow considerably (at odds with the comparatively mature platforms of Twitter and Yelp). Developing a model that alerts based on spikes in reports won’t work so well if the spikes are just due to a larger percentage of customers reporting their experiences. The other issue is that the information is moderated heavily. Though it looks to be published in real-time, I believe this is a carefully constructed illusion. On one hand, reviews are being screened for inappropriate content, but, on the other hand, the owner of the website has privileged access to a lucrative data source. The moderation isn’t perfect, either. I noticed that there was still a heavy stream of Virginia Chipotle reports streaming in days after the news story broke.
While Yelp and Twitter appear to be useless for detecting food contamination issues, Iwaspoisoned is an intriguing resource that has been credited with uncovering recent outbreaks. With a carefully written computer program, the information published on that website can be used to profit from eventual declines in stock price once the information becomes mainstream. In a future post, I will explore how this may be achieved.
Pingback: Building a Food Contamination Data Set | Tim O'Hearn()