Let me approach this one from two different angles. Point number 1: People who predict the market for a living are bad at it. Point number 2: There are a host of reasons to think that they will still be better than you.
Many research papers have sought to rigorously analyze the accuracy of target prices published by equity analysts. The idea being that those whose job depends upon their ability to predict equity prices are those most likely to get that job done. Common sense suggests that if an analyst, devoting as much time as humanly possible to predicting prices for a small set of equities cannot do it, then it is unreasonable to assume that an amateur working a few hours a week will do any better. We are not that smart.
A recent paper (Bonini, Stefano, et al. “Target price accuracy in equity research.” Journal of Business Finance & Accounting 37.9‐10 (2010): 1177-1217.) developed a systematic way to assess the accuracy of price target forecasts from equity analysts. The results are neither surprising nor anywhere close to what I would characterize as “good”. More specifically, the authors find that: “the frequency of accurate prediction is extremely low, and the size of the prediction error is impressively large”
This is an egghead’s way of saying “these forecasts suck!” But let’s be a bit more precise. We note that this study was conducted in Italy. This may seem odd, but it is for a very good reason. Italy is the only G7 country that requires mandatory publication of any research issued by authorized financial intermediaries. In other words, the analysts are required to share these forecasts in a centralized database for future study and analysis. This is key because it helps to avoid a sort of survivorship bias. The firm issuing the forecast cannot simply “take it down” after it is proven to be inaccurate. In addition, no firm can say “Well, we know that would happen. We just didn’t publish it.” if the forecast appears to be correct in hindsight.
At the end of the day the authors find that “errors are consistent, auto-correlated, non-mean reverting and large.” In addition, they find that, “the size of the forecast errors increases with the predicted growth in the stock price, the size of the company and for loss making firms.” Notice that the forecasts with the largest predicted increase in stock price are among the worst. In addition, forecasts of some sort of turn-around in performance are also among the forecasts most likely to be wrong. If one is looking to use forecasts to help make purchase decisions, aren’t these the forecasts you are looking for? In some sense, this is not entirely surprising. It is rare to see a great jump in price and it is difficult to turn a business around in a short span of time, but if the most exciting forecasts are likely to be the worst, what’s the point?
In addition, the authors write that: “prediction errors are consistently different from zero, auto correlated, non mean-reverting, and positive in sign.” Let’s translate this egghead gobbledygook into a bit of simple English. To say that prediction errors are consistently different from zero suggests that if we use analyst forecasts as inputs into a model to predict stock prices after some lag, the error terms do not center around 0. Any student of regression models will immediately recognize that if this is the case, there is something systematically wrong with the model. It’s not just an absence of explanatory power due to prices being completely random. Having residual values not centered around 0 suggests that there is something important to consider that is not being accounted for. (I am going to argue that overconfidence is that factor, but that is just conjecture on my part.)
When we state that errors are auto-correlated, this means that if you are wrong today (either high or low) you are quite likely to wrong tomorrow, and in the same direction. The fact that the errors are “non mean-reverting” is a way of saying that a big error today does not suggest a smaller error tomorrow. If you have a big error on the high side today, it may be comforting to think that you have an elevated chance of seeing a smaller error, possibly on the low side tomorrow, but no such luck here.
The errors are positive in sign simply means that most of the forecasts are simply too high. This may be related to several incentives in place. When I predict that stock X will rise over the next quarter, customers are more likely to increase their holding in Stock X. Furthermore, company X is likely to feel all warm and fuzzy about me as an analyst. In no way am I accusing analysts of lying. I am simply noting the fact that when one lives in a sea of incentives, they tend to work eventually.
What may be even more surprising is that the researchers found “analyst accuracy to be negatively correlated with research intensity” This is almost shocking. The greater the number of analysts paying attention to a stock, the more likely they are, as a group, to be wrong. In addition, they find that: “analysts are slow in recognizing when a stock is approaching exceedingly high levels and show a consistent overoptimism.” Again, putting this together it says that when the analysts tell you a price is going to rise, they are slow to alter that prediction, and it is typically wrong. To make matters worse, the more analysts screaming that the price will rise, the worse the prediction is likely to be.
However, there is also a more subtle issue at play here. The authors find that the errors do have a bit of a pattern, especially if the analyst predicts a strong price rise. Apparently, issuing a recommendation to buy, can serve to increase the stock price in the short term, but the market eventually digests this information, and prices revert closer to the original level. This suggests that for a long-term investor, these forecasts are particularly irrelevant.
Adding additional support to point 2, another recent paper (Merkle, Christoph. “Financial overconfidence over time: Foresight, hindsight, and insight of investors.” Journal of Banking & Finance 84 (2017): 68-87.) focused on overconfidence as a behavioral trait among individual investors, and where this leads. This study is of particular relevance to us because it focused on survey data and actual performance from fairly affluent (but not super-rich) brokerage clients. Most of these folks fit our description of do-it-yourself investors. The surveys generated data about stock market and portfolio expectations. These fairly affluent, self-directed investors were surveyed every 3 months over a 2-year period to measure overconfidence and the evolution of that trait over time. Using this data the authors derive metrics of “overconfidence” from the responses. The authors were then able to match the survey data with actual portfolio holdings and performance. They find that the overconfident investor tends to trade too much, overestimates their level of diversification, overestimates their ability to estimate market characteristics, and takes more risk than they think they are taking.
Many investors over-reported their recent returns to a large degree. This suggests that experience leads to a distortion of memory. In cases in which investors overestimated their past performance this fed their level of confidence going forward. This leads to an overplacement of relative expectations. In other words, these folks expect to outperform other investors over time. Such people are not likely to be impressed with returns that match the broader market, because they “remember” that their portfolio outperformed the market in the recent past, and also expect it to do so in the future. To make matters worse, when an investor does outperform in one period (which will happen about ½ the time with randomly selected stocks), they raise their expectations for the next period. Funny thing about people – when we feel as though we ”beat the house” we keep betting and we tend to go bigger. In addition, about 75% of the trades that these investors made were in individual stocks, as opposed to mutual funds, index funds, or ETFs.
The median investor in this data set trades about 5 times per quarter and holds 12 different stocks. Thus, they trade too much and diversify too little. Investors who expect to beat the market trade more frequently, trade at higher volumes, and churn their portfolio more often.
Let’s bring all of this back to the original point. The people who study the company much more than you have time to do; the people who are paid very well to get this right; the people who have every advantage possible in this task cannot do it – – – and surveys of individuals who do this on their own show that we typically do much worse. If the best trained, best educated, best equipped, best supported, and best paid cannot do this very well, perhaps your 8-10 hours per week of research (or whatever it is) is not likely to do any better. In this sense, the market is smarter than you are. When you believe that the market is wrong, and that you know better, you are setting yourself up for failure. Don’t Do It. Keep it simple to avoid letting your ego cost you a fortune.
