[ Music ] >> All right. It's easy to document the market anomalies. The difficult part is interpreting why these anomalies exist or if they really are anomalies. Maybe they represent risk. So when we find some pattern of returns, it could be due to risk so if there's some risk story behind some assets generating high returns, high risk, high return, problem solved. But the key thing is, we can always throw risk out there. Do you have an economic story that is clear as to why investors should care about this risk, and therefore kind of expect higher returns for exposure to it. So what's the story behind the risk? Don't just throw risk out there unless you have a story to kind of back it up. Irrational behavior could explain patterns in returns if you think there's some part of the population that's just making mistakes that leads to mispricing of assets that the smart investors can then take advantage of. But for the irrational behavior story to really kind of be potent in explaining market anomalies, you need to have enough kind of people out there making mistakes and not enough people there that can correct prices. Okay, and then it could just be really is an anomaly that through a pattern you observe is just due to data mining. So when you're looking at people are doing hundreds or thousands of correlations trying to predict stock returns, there's probably bound to be a few weird ones that pop up just by chance. Okay? So, let's talk about market anomalies. They can be attributable to three factors. Let's talk about risk first. Risk is thrown out there. This would be the answer from advocates of efficient market viewpoint. Looking at the out performance of small and value stocks historically, a risk base argument would say hey, these are stocks that are risky. Small firms are riskier than large firms and value stocks maybe have more financial leverage and makes them riskier than growth firms particularly in a recession. This risk causes investors to not value the small and value stocks as high-- that highly and therefore acquiring a higher return as compensation for the risk. Now, notice if there's a risk base explanation for the returns, it would suggest that this market anomaly-- and I put it in quotation marks, will continue. If there's a risk story, the you should have the high returns going forward. You can kind of think of the example if there's a $100 bill laying on the ground, should you pick it up or not? The risk base story would be, "Hey, don't pick up the $100 bill. It could be tied to a bomb that goes off with probability 5 percent." So that would suggest that $100 bill is going to remain on the ground a long time because no one wants to take the chance of it blowing up. So risk base explanation, return pattern should continue. Another possibility for explanation for the market anomaly are the pattern and returns that we observe for like size or value or any other predictability in returns that we have would be just simply irrational behavior on the part of some investors. So perhaps investors really like large-- and are always optimistic going for the latest kind of growth startup, and they neglect the more kind of boring value cement company type stocks and they neglect small stocks. So that causes the small and value stocks to be under priced and yield higher returns in the future. Now, note for the return anomaly to continue under their rational behavior explanation, you need two things to hold. One, there need to be enough of these behavioral investors to keep making these mistakes so that prices are distorted so there's positive returns to be had from the smarter investors. And you'll also need the smart money to not be so large that it fully corrects the distortion in price. Third explanation for market anomalies, kind of predictable patterns and returns simply could be data mining. By chance, some criterion seems to predict returns or when it actually does predict returns in the past, but it's just due to chance. It's just due to this data mining looking at every thousands of variables. You're bound to find a few weird correlations just like if you have a room of 100 people, there's probably going to be a couple if they flip a coin five times, maybe more than that, get five heads-- five heads in a row. Just kind of do by chance. The next five coins they flip probably aren't going to be an additional five heads in a row. So under the data mining explanation, there was a chance correlation in the past returns we observe, but since it's just due to chance, it's unlikely to continue going forward and the return pattern observed in the past under the data mining story would then disappear going forward. Okay, so I know data analytics is kind of very important, but it's always useful to remember the necessity of distinguishing between correlation and causation. So for example, strong correlation in the past between butter production in Bangladesh and the S&P 500 index returns. So of course, people are trying to see what predicts stock market returns, what turns out to be a great prediction of that? Butter production in Bangladesh. So, let's buy our plane ticket and head off to Bangladesh because look at this effect. Ten percent increase in butter production appears to be associated with a 20 percent increase in the S&P 500. So I don't know about you, but I'm already packed and ready to go. So, time to go to Bangladesh. One problem, though. You're not in Bangladesh. That's Vietnam. Bangladesh is over there on the border with India. Sorry, had to pull a little geography quiz here on you here, kind of copyright John Oliver routine. So, this just goes to show it's very important to understand the difference between correlation and causation. In the past, there's been this strong correlation between butter production in Bangladesh and S&P 500 returns, but I don't think that means if I fly to Bangladesh eat a bunch of butter, more butter is produced, the stock market is going to soar. It's just a free correlation. I doubt very much it'll continue going on in the future. The key thing is-- it's always going to be easy for someone ex post, after the fact, to come up with some tortured story to try and sell or motivate why one factor predicts another, although the butter in Bangladesh might be pretty hard to have the story as to why does that predict stock returns, but the key thing is can you come up with a hypothesis prediction ahead of time then go to the data and see that your prediction is actually born out in return patterns in the data? Can you come up with a story that's plausible exante, ahead of time, as opposed to coming up with a story after a data mining exercise to kind of justify an effect you found. So, it's important to realize there's going to be many people correlating many different things with stock returns because there's a kind of a big reward if you can find some correlation there that actually is predictive of returns in the future. When you have so many analyses being run, there's bound to be a few weird anomalies that happen by chance. Now, if you think there's a pattern return that's happened by chance just because you had this pattern in prior years, you wouldn't expect it to continue.