[MUSIC] So in this video, we're going to talk about complicated firms and potential inattention on the part of some segment of the market. And how that can be used to form potentially profitable investment strategies. So the idea is that some group of investors takes a longer time to process relevant, complicated information to value a firm. But at least it's not so complicated that no one can figure it out. Or maybe there just some group of people's not paying immediate attention to all the relevant information for valuing a firm. So that suggests a potential profit opportunity for the group of investors that has the ability to process this complicated information or simply paying attention to the news immediately after it's released. Okay, so the idea is to find things that are complicated for many. So I don't know how many of you remember this Rubik's Cube, but when I was kid it was quite a phenomenon. And you could buy these books, like how can I get this back to like all yellow on one side, all blue on another. So for a lot of people, this may be very complicated. But maybe for you, you can figure things out quite quickly. Now for me, I was actually risk averse. So my mom bought the Rubik's cube. And I actually just kept it like this, because I was very afraid that if I started to monkey with it, I would never get it back to the original condition here. So my Rubik's Cube is somewhere at home exactly like this, kind of very sad level risk aversion here as a kid. So let's pause, think, and answer. What's Le Penseur have in store for us now? What type of firms do you think might be more complicated than others? Right? If the firms aren't that complicated to follow, there probably aren't great profit opportunities, right? Because as soon as any news comes out, everyone instanteounsly understands what it means for the firm and stock prices adjust up or down right away. So you want to look for firms that maybe not everyone can figure out what this news means for firm prospects right away. If we find such firms, what's a portfolio strategy one could construct, assuming you have an edge in processing information about these complicated firms? So let's think about this, and then I'll come back and give you some examples. What type of firms are complicated? What's a portfolio strategy one can construct, assuming you have an edge in processing information about these complicated firms. Well, let's go through a few examples of some research that shows potential returns to some of these inattention or complicated firm investment strategies. So one of these is research known or or titled does attention stop at water's edge. So what's the idea behind this? US firms increasingly have more and more of their sales overseas. So if we want to forecast how a firm's going to be doing in the future, it's not just enough to understand what's happening in the US economy, but you want to know what's happening in economies in Europe and Asia, where these US firms also are doing a lot of business. The question is are investors in these US firms fully keeping track of how these overseas economies are performing and thus affecting the sales, going forward, of these US firms. That's the idea. So perhaps it takes a little time for overseas shocks to the economy to be fully reflected in the stock price of the US firms. Okay, this is research reported by Quak Nguyen in his 2012 paper. And Huang also reports similar results in her research. So what's the portfolio strategy? Let's get down to the nitty gritty here. For each month, from 1999 to 2010, let's rank a firm by a weighted average geographic return. Okay? So it turns out, what do we mean by the weighted average geographic return? Well, the weight will be the share of sales in a particular country. The return is the stock market in that country. Okay? So the idea is let's say Coach purse sells all their products to Portugal. So if we look at the kind of share of sales that Coach purse has, 100% will be in Portugal, and we have data over this period, 1999 to 2010. For each firm, we know the geographic regions where they do sales. Is it in the US? Is it in Portugal? Is it in South Korea? So Coach purse does their sales in Portugal. It turns out last month Portugal's stock return was minus 10%. Let's say Pepsi does their sales, it's 100% in South Korea. Last month South Korea stock return was 10%. So South Korea, 10% return, let's say Portugal was -10%. So we would say, going forward, good news for investors in Pepsi. The economy seems to be doing well in South Korea. Bad news for investors in Coach, the economy seems to be doing not so well in Portugal. So we calculate this weight average geographic return for each firm, and I gave you a simple example with Coach and Pepsi, then, rank firms by this geographic returns. So Pepsi, their sales are in South Korea. South Korea did well last month. So they're going to get a high ranking. Coach, their sales are in Portugal. Portugal did poorly last month. So Coach will get a low ranking in this. Rank firms by this geographic return. Form portfolios over the next month, based on this ranking. So the idea would be U.S. investors don't instantly realize that the good news in South Korea is good news for Pepsi and the bad news for Portugal is bad news for Coach. So let's see if we can earn returns over the next month by investing in Pepsi and avoiding Coach, or shorting Coach. And we're going to repeat this process month after month, after month, okay? And we'll present the returns from this strategy and we'll report monthly returns in percentage points, so we can evaluate is there some inefficiency in the market that this international news doesn't instantaneously get reflected in US stock markets. So let's look at the results here. Does attention stop at the water's edge? Here are the returns that we're looking at for the strategy. So let's just focus on the fifth column here, and here we're looking at those firms that are in the top 20%, based on this geographic return. So on the simple example I had, Pepsi would be in group five here, because they come from a region, South Korea, whose stock market did very well last month. And here, we're looking at the returns to evaluate this strategy using different benchmarks. So in row one, we're simply looking at what's the excess return of this strategy, the return of the stocks that you know, are in this group, that make sales to regions that did well last month. In our example, Pepsi would be in this group. They're beating the Treasury Bill return by 1.35% on a monthly basis over this period, 1999 to 2010. How about if we do a CAPM model analysis? What's the Alpha from that? 1.36%. Standard errors are in parentheses here. So when we see these standard errors around two or more in magnitude, we know that that's a statistically significant result. We go to the three factor model. We're getting an alpha or outperformance of 0.9% per month, okay? Now remember 0.9% per month, that's over you know 10 percentage points on an annual basis. That's a huge, huge return. How about the firms that are in group one? So in my example that would be Coach. In my example coach sales are all in Portugal. Portugal stock market did poorly last month so we would put Coach and other firms that made sales to regions whose stock market did poorly last month, they would be in group one. So we can see their returns are much lower than those firms that are in group five, okay. So he is kind of the difference. What's the return to the firms like Pepsi minus the return of the firms like Coke? Group five, the high, these are the firms that come make sales to regions last month whose stock market's gone up a lot. Minus the return from firms who make sales to regions whose stock return last month did poorly. So this is the key result, group five minus group one. Regardless of the risk model you use, whether you simply just look at differences in returns, you use a CAPM model to control for risk, you use a FF-3 factor model, you also include momentum effects. Or you even do this five-factor model where you control for market risk, size factor, value factor, momentum effects and liquidity of the underlying investment. All these estimates are exactly the same. They're about 1.4 to 1.5 percentage points on a monthly basis or about 18 percentage points on an annual basis in terms of this portfolio strategy where each month, you invest in US firms that makes sale to regions whose stock market did well last month, short US firms that make sales to regions where whose stock market did poorly last month. So there does seem to be a little bit of inefficiency in the market. Some strong predictability of returns. So how does a payoff from this geographic momentum strategy if you will, how does it vary with the state of economy and other investment strategies. So we're going to focus on this again. Invest in the groups who make sales to regions that did very well last month, short the US stocks that make sales to regions that did very poorly last month. The high minus low group. Let's see, we reported these you know alphas or these differences in performance. They're all the same regardless of the risk model we use, they're all 1.4, 1.5 percentage points per month. About 18 percentage points on an annual basis. Like very lucrative strategy here, what's the r squared of these risk return models? It's basically zero. Usually you would say an r squared of zero is very bad, but in this case I think it's actually good. And if you want to look at it a different way, the p-value of the test, do these models explain any of the returns of the strategy. These p-values are all very high, so the answer is no. What quant Nguyen was finding in his research and Wong as well is that they've uncovered a strategy whose returns aren't sensitive to the state of the economy, right, the coefficient on market conditions is zero. Don't have a size or value tilt, aren't related to the famous momentum strategy, and aren't subject to liquidity risk. So none of these known factors that predict returns can account for the returns from the geographic strategy. So that's important to document. How many months in the future should one be able to earn returns from the geographic momentum strategy, if you will? This kind of gets at, the market might not be perfectly efficient and processing this information, but you probably don't want to sit on this information too long. It'd be surprising if people haven't figured out six months from now that good news in South Korea is good news for Pepsi and bad news in Portugal is bad news for Coach. Maybe they don't figure it out right away in the first month, but they probably figure it out after six months. So let's rank stocks again by their weighted average geographic return. We do that, and then let's see how does that predict returns over the next month. We already showed that results, but how does it predict returns in month two, or month three, or month four, or month five. How long does this predictability last? Here's the results looking at the ranked firms by how the region where they make sales did last month. These are the estimates we already presented. Looking at the group that makes sales to regions that did well last month Short those that makes sales to regions that did poorly last month this return over the next month, this differential return, 1.4 to 1.5 percentage point difference in returns next month. How about if you look at the second month, the third month, the fourth month, the fifth month or the sixth month. Let's just focus on the Fama-French 3-factor alpha. All these alphas are zero. So there is no longer any return predictability once you get past the first month, okay? So it makes the markets pretty efficient, processing this information. It just doesn't do it right away. But if you wait more than a month, too late. Investors have figured it out, prices have adjusted. How about if instead of using the stock market return maybe you don't think that's the best measure to get a sense of the customer's buying power in all these different countries like South Korea for Pepsi, Portugal for Coach, using my example. What if instead, you look at GDP growth? So gross domestic product, that's reported on a quarterly basis. So instead of looking at the stock market return each month, every quarter we could look at the GDP growth of the different countries where a firm has sales. And then rank firms by their weighted average of what's the GDP growth of the regions where they make sales. And it's simply weighted by how many sales do you have in that region? So if you do a lot of sales in one region, we're going to give a high weight to that region's GDP growth. So then, we do this four times a year, as opposed to 12 times a year. Ranking firms by their GDP growth over the past quarter, the regions where they do sales. Does that predict future returns? Okay, The answer is yes, it still does. So these coefficient estimates here are again focusing on the difference between the top quintile and bottom quintile. Top quintile firms, those are US firms that make sales to regions that had the highest GDP growth over the last quarter. The bottom group are those US firms that make sales to regions that had the smallest GDP growth over the last quarter. What do we see going forward on a monthly basis? We're getting excess, we're getting risk adjusted returns, 0.6 to 0.9% on this measure here. So these returns are lower than what we had when we do the monthly adjustment with stock returns. When we form portfolios on a monthly basis with stock returns. Here we just formed portfolios four times per year once we had the different GDP growth. These returns still are almost 10 percentage points on an annual basis. They're lower, but we're making less frequent adjustments to the portfolio, so lower transactions cost. Okay? So kind of the bottom line, whether you rank Firms by the stock market performance in the regions where they do sales or the economic growth, the GDP growth in the regions you have sales. You still have predictability, the market doesn't seem to fully incorporate immediately good news in overseas markets and what it means for US firms. So another example of the valuation of complicated firms, kind of very similar in spirit. Can one use a performance of easy-to-analyze firms to predict the performance of their more complicated peers? This is research by Lauren Cohen and Dong Lou. They're basically documenting that you can predict the return of conglomerate firms. Firms that have multiple divisions, by paying attention to the performance of stand-alone firms. Okay? So let's look at their results here. They're kind of using a similar methodology. Let's look at a firm, and they have four divisions. So, last month, let's see how the stand-alone firms did in all four of those divisions. So if your division and let's make it simpler since I don't want to do four examples. Let's just say they have two examples. I have a railroad and I sell soft drinks okay. So let last month, let's look at the return of the railroad industry, let's look at the return of soft drink firms. And let's see if these returns of your stand-alone competitors predict your return next month. Okay? So under efficient markets, if good news happens to soft drinks, that should be good news for me as well because part of my business is soft drinks. There shouldn't be any predictability. The fact soft drink industry did well last month shouldn't predict me doing well this month. I should've done well last month as well. But, if it takes a little while for people to process the news, hey, good news for soft drinks isn't only good news for Coke and Pepsi, it's good news for me as well. It's maybe 40% of my business, but that's still good news for me, if it takes a little while for people to figure that out, maybe there can be some predictability, right. Good news for Coke and Pepsi last month predicts kind of good things happening to my stock price this month. And this is what Cohen and Lou look at. So let's look at the kind of results here. So first, let's weight firms each month, let's categorized firms, each month, by what was the performance of their stand-alone competitors. So group one, these are firms that their stand-alone competitors did very well. That predicts me doing very well next month. If I look at a CapM model, I outperform my benchmark by 0.7%. Next month in a 3-factor model, I outperformed by 0.5% next month, with T statistics here in parentheses. So statistically significant results. So the fact my stand-alone competitors did well last month predicts that I do well, in terms of higher stock returns, next month. How about those groups, group 1, where my stand-alone competitors did poorly last month? Well that predicts me under-performing my benchmark next month. In the CapM model I under-perform by 0.5%. In the 3-factor model I outperform by 0.7%. So the fact Coke and Pepsi did poorly last month, that predicts my stock return doesn't go down last month. Or doesn't fully go down last month, it goes down the next month because it takes a little while for the infer, for people to process hey, bad news about soft drink. Also affects Scott's firm because he's 40% in the soft drink industry. It takes a little while for the market to process that. And then the key is a final row, which is long short. Where we're long, we're investing in the stocks whose stand-alone competitors did well last month. We're shorting the stocks where the stand-alone competitors did poor last month. We see here, these risk adjusted returns, kind of similar to the does the water stop, does the, kind of news stop at the water's edge, paper the geographic momentum. Paper having 1.2% monthly returns to the strategy. About 15% on an annual basis. So quite impressive here. So conglomerates, firms with you know sales overseas both seem to be complicated firms where the information isn't you know kind of fully efficiently processed in real time. And then here is just looking at these returns to the strategy whether you look at the blue line where we're equal weight firms or you look at the red line where we're value weighting firms. Firms. Okay, and we're simply reporting the returns to this hedge fund strategy of buy the stocks whose stand-alone competitors did well last month, short the stocks where the stand-alone competitors do poorly. You see there's this big 1% return to that strategy in the first month, and then, if anything, the return to the strategy goes up a little bit, but it doesn't reverse. It doesn't reverse over time here. Final example of this drift strategy here, using information that's publicly available to predict future returns. A classic drift strategy here is based on the under-reaction to accounting news. Awkward acronym here: PEAD. P-E-A-D, Post-Earnings Announcement Drift. So the idea is invest in companies that had a positive reaction to the most recent earnings report for the next three months. They continue to do well, okay? Short or avoid companies that had a negative reaction to the most recent earnings report for the next three months. They continue to do poorly. So it's not surprising, if firms announce good earnings, their stock price goes up. What's interesting is that historically you look, the prices continue to drift up after that. It's not surprising if a firm announces bad earnings that the market doesn't expect, the stock price falls on that day. But what is surprising, is the price continues to go down. So let's think a little more about this. What would be the returns from this Post-Earnings Announcement Drift strategy in a fully efficient market? What should the returns from this Post-Earnings Announcement Drift strategy be if there's some behavioral factor so there's under-reaction? To this accounting news. So I think I kind of set this up pretty well. Hopefully these questions aren't too challenging. Think of your responses, then I'll give you my take. What should the returns from this Post-Earnings Announcement Drift strategy be in a fully efficient market? Zero, right? Publicly available information shouldn't predict future returns unless that's associated with some type of risk so the fact a firm had good earnings, that should cause the price to go up when the earnings are announced, but there should be this positive drift afterwards, okay? How about if there's some under-reaction? Well then if there's under-reaction, kind of by definition if the price isn't going up as much as it should to reflect the good news. Then, eventually, there is this drift as the prices get to fundamentals. And historically, that's what we observe. So a lot of people have documented this. I thought one recent study that kind of showed these results fairly well was Andrea Frazzini (2006). So we're looking at the monthly returns to Post-Earnings Announcement Drift strategy. At the beginning of each calendar month we simply rank stocks by their stock return on their most recent earnings announcement date. So the idea is, let's have the market tell us which firms had the most surprising good earnings, and the most surprising bad earnings. So we're going to rank firms by what was the stock market reaction? When they most recently announced earnings, and we're going to put them in groups of five, where group five are the 20% of firms that had the highest return on the announced earnings, the biggest earnings surprise. And group one, are those that had the worst earnings announcement terms or the most negative surprise, most negative announcement. We're going to kind of control for risk in a Fama French three factor model, we've kind of done that quite a bit at this point. So we're going to report the Alphas, or the risk adjusted returns controlling for the market risk of the investments, controlling for the size and the value characteristics of the investment. And we're going to look our hedge fund strategy that's long short, is going to be long. Those stocks that reported the most positive earnings had the biggest market surprise, and short those that had the most negative surprise, and the key thing is, we'll look at this strategy going forward. So it's not surprising, good earnings lead to a good stock return the day it's announced, bad earnings lead to a bad stock return the day it's announced. What's surprising is, is there any return to this drift strategy? Okay, and we're looking here at t-statistics here in parenthesis to the strategy, what do we find? What we find over the next three months, those firms that had the great earnings announcement, they underperformed their benchmark. They continue by 0.6%, on a monthly basis, or 1.8% over the next quarter. There's a symmetry here, those bad earnings announcers, they continue to underperform their benchmark by -0.6% on a monthly basis, or it'd be about 1.8% on a quarterly basis. Then you look at the difference here, invest in the stocks that had the good earnings announcement, short those that had the bad, the difference between the two is 1.2%, on a monthly basis or 3.6 percentage points difference over the next quarter. So the good earnings announcers, it's hard to predict which firms will have the good earnings announcements or the bad earning announcement. It seems like it should be easy to read the Wall Street Journal and just read which had big surprises, which had negative surprises. These results here suggest that simple strategy of just investing in the past good announcers shorting the past bad announcers, still yields non-trivial returns going forward. Okay, so stay tuned, Faculty Focus Episode, talking more about inattention and predictability in returns. Faculty Focused Episode that I have in store, you'd be quite a treat here. When I interview Josh Pollet, we'll talk about two of his areas of research. One is, are people fully paying attention to firm earnings announcements that happen on Friday? And the second is, does the market fully price future changes and demographics that may lead to more demand or less demand for certain industries, okay? And one thing I really want you to focus on in this Faculty Focus Episode. Is pay attention to the awkward moment, it's going to be quite a shocker. [SOUND]
