Now, Faculty Focus with Scott Weisbenner. In this module, the focus is on Josh Pollet. Josh Pollet is an Associate Professor in Finance at the University of Illinois with research and empirical asset pricing and behavioral finance. Professor Pollet has published in all the leading finance journals as well as the American Economic Review, the flagship journal of the American Economic Association. His work has been cited over 1,500 times by other scholars and has been highlighted in numerous popular press outlets including the Wall Street Journal and New York Times. Fun fact about Josh, his nicknamed the Razor because of his sharp intellect and his ability to slice and dice a poorly thought out argument. Professor Pollet, welcome. Thank you. Sorry about that nickname thing. I made it up on the spot but the sharp intellect part is true. Hopefully there will be no skewering of colleagues today. That'd make for a more interesting Faculty Focus segment. In this module we've talked a bit about inattention on the part of investors in how if there is some as inattention may lead to predictability in stock returns. You're the perfect person to talk to. We haven't highlighted any of your research in the module because they want to talk to you and going to get the insights from you face-to-face. Certainly, it's a pleasure to talk about my work. One of the best parts about behavioral finance is the idea that investor mistakes can actually lead to exploitable trading strategies that sophisticated investors can take advantage of. When we can figure out ways to model those mistakes carefully and then predict the types of mistakes that people will be making, then hedge funds can do the opposite and take advantage and have high abnormal returns. A key thing I like about your work is you are coming up with stories or maybe mistakes are being made that makes sense our priority. It doesn't appear to be an x pulse story that's made up to explain some return anomaly. Exactly. We want to think of a mistake that would be made by many people or a large population and then investigate the predictability of the predictions of that type of mistake to see if we can observe the types of behavior that are consistent with a particular behavioral bias that many people might have. Excellent. So I thought in particular two of your papers would be particularly interesting to our viewers. Your work on Fridays and your work on demographics. Why don't we first talk about your investor inattention and Friday earnings announcement paper. This was published in the Journal of Finance and also was covered very nicely in the Wall Street Journal. What's the idea behind it? So the basic idea is that investors might take a break from their work at different points in time. And in particular that on Fridays and during the weekend, investors are not as focused on work-related tasks as they are other tasks and so they forget what they learned on Friday. And it may not fully incorporate earnings announcements on Friday compared to the way they respond on other days of the week when they're not distracted by the weekend and other non work-related tasks. I see. I understand you brought some figures for us so we can get like a deeper dive into this research. Certainly. The first figure is about abnormal volume. The figure looks at the trading days from just before the earnings announcement to just after the earnings announcement. Day 0 on the figure is the date of the earnings announcement, day negative two is two days before. Day two is two days after. The striking thing that we can see in this figure is that on almost all of the days, the amount of abnormal volume for an earnings announcement on Friday is the same as the amount of abnormal volume on the other days. The dark blue bars and the light blue bars are about the same. Exactly. The dark blue bars and the light blue bars are the same height on all of the days except for day one. Which is Monday following a Friday announcement. Exactly, one day after the announcement on Monday, there seems to be a lot less trading in response to a Friday earnings announcement compared to what happens on an announcement for instance, on Wednesday when you would have one day later would be Thursday and this gap is extremely large. It's about 40 percent less trading following a Friday announcement on Monday than for instance a Thursday announcement on Friday. When we're looking at day one, the blue bar is about 60. The dark blue bar is about 60 percent of the height of the light blue bar. So the natural question is, is there some special types of earnings that are announced on Friday. You're more apt to see negative when you have this less response in terms of the attention paid. I think that's exactly right. If you think that investors are not going to pay as much attention to an earnings announcement on a particular day, what type of earnings announcements would you then. It's not the good stuff. Why would you want to hide good news? You would never want to hide that. You want to trumpet that You want to trumpet that, so you're going to announce it on a different day whereas if you have bad news to announce, you're going to choose to announce it on a Friday. We see that pattern in this figure which is that the light blue bars refer to announcements on other days of the week, the dark blue bars again refer to the announcements on Friday. The light blue bars are set so that all five of the negative quantiles one through five, have the same height based on the distribution of announcements on other days. Similarly, all five of the blue bars to the right of number 6, 7 through 11 have the same height, the light blue bars for the positive announcements on other days. If we just take a step back, I see we have 11 groups of earnings announcements here, one through 11 on the x-axis. Group 6 and the middle is zero surprise. Exactly, group 6 is where you exactly meet median analyst expectations with your earnings announcement. Groups 1 through 5 are evenly divided negative announcements, we take all the negative announcements on other days of the week and evenly divide them into five categories. That's why these light blue bars by definition are the exact same height because you make these cutoffs. Group 1 is a bottom 20 percent of the negative surprises. That's right. Then similarly group 11 is the top 20 percent of the most positive surprises on other days. Seven through 11 are all positives, seven are the weak positive surprises, 11 is the 20 percent of the strongest positive surprises. That's right. What do we see here for the Fridays relative to the non-Fridays in terms of the earnings announcement return? For Fridays we see that the dark blue bar in column 1 is almost twice as big as the light blue bar in group 1, indicating there are almost twice as many extremely negative earnings announcements on Friday compared to other days of the week. Wow I'm sorry, I got ahead of myself. We'll talk about returns later. This is just showing the distribution of, are you announcing good news? Are you announcing bad news? The really bad stuff is happening on the Fridays relative to the other days a week. Exactly, it's much more likely to be announced on Friday and you can see that in both column 1 and column 2. Those are basically the most negative surprises and even a little bit in three and four as well. All the weight is more negative stuff coming out on Friday relative to other days of the week. That's right. This indicates some type of strategic behavior on the part of firms to push their bad news to earnings announcements on Friday. The flip side for good news it looks like here that if you look at at least particularly group 7, group 8, group 9, a little bit on group 10, you see less coming out on Friday. That's right. Again, if I had positive news and I want the market participants to pay attention to it, then I want to announce on other days of the week rather than on Friday. I shift my good news earnings announcements to a Thursday or a Wednesday rather than on a Friday. Why don't we talk about returns and I see you have a few figures detailing map. Of course, the same quantile specification applies here. Groups 1 through 5 are the negative earnings surprises, group 6 is no earnings surprise meeting expectations, and group 7 through 10 are the positive surprises. There are two lines, the dark blue line refers to Friday announcements, and the light blue line refers to announcements on other days of the week. What we see here is that the slope of the two lines is a bit different, the slope of the Friday line is flatter than the slope of the line for the other days, indicating what I would call an attenuated response to inflammation. It's just more muted, it's like less response to negative, less responsive to positive. That's right. There's just less response to both types of extreme announcements because fewer investors are paying attention to whatever the information was that was announced. You can see if we want to just look at groups 1 and groups 11, for group 1 the negative news, more negative reaction on the non-Friday days and the Friday, you go to group 11 the most positive earnings are being released bigger reaction on the other days than on Friday. Exactly. Now really key to your test is if there's this underreaction to the news on Friday, the market may not pick it up in day 0 or day 1 but eventually you think the market would figure this out, so you want to look at drift. That's right. We don t think markets are inefficient forever, and so if they're making mistakes in the way they process information in the short term, eventually that mistake should be reversed as investors catch up with the inattention. I think really a great place to focus is on group 1 here to look at this drift, what's happening to the returns day 2 to day 75 among the firms that report really bad earnings. What we see here is in Group 1 that the performance after the announcement and the days following the announcement until roughly the next earnings announcement is much more negative in the bottom group for Friday announcements. More than undoing the attenuation that we saw right around the announcement, the reaction is much more negative to the stock market reaction, the abnormal return is much lower following a Friday announcement in the bottom group than it is for following announcements on other days in this very bottom group. More muted initial reaction to the bad news on Friday, but then the market gradually figures it out over the next 75 days. Exactly. What my favorite figure from your paper that I think really hits home. This differential drift following good versus bad news on Friday versus other days is this Figure 2. Post-earnings announcement drift has been studied since the 1980s. The idea behind post-earnings announcement drift is that following a good earnings announcement, stock prices keep rising over the next three to six months and following a low earnings announcement, the stock price keeps falling over the next three to six months based on the sign of the earning surprise. Positive surprises, higher returns, negative earnings surprises, negative returns. What we see here is that when we divide up earnings announcements by whether or not they occur on Friday versus other days and then constructed this drift strategy, we find that the drift is much larger for the Friday announcements than the announcements on other days indicating that there's more of a delayed reaction to Friday announcements than there is to the announcements on other days. Interesting. We did talk in class about the post earnings announcement drift. This is a tweak to the strategy. Look at the post-earnings announcement drift following the Fridays versus the other days. Exactly. What we can see from the figure is that over the next 90 days, the drift is almost twice as large if it's constructed using Friday announcements than if it's constructed just using the announcements on other days. Well, Josh, you saw a pretty compelling set of results. I think the bottom line is, if I see a firm announcing, we're going to move our earnings announcement day to Friday, badder the hatches. But bad weather is coming. It's time to sell. It's time to sell. Well, this was good stuff. Why don't we turn now to your research on demographics and stock valuations. This was published in the American Economic Review, which we know is the flagship journal of the American Economic Association and was also covered in the New York Times, which I always like to see like hey, it's not only published integrate academic journal, but the popular press is all over this as well. What was that paper about? The idea in that paper is that perhaps investors don't understand long-term information very well. But it's very hard to test this in practice because we don't have very good forecasts of the long term in many areas. There is one notable exception to this, which is that we understand demographics over the long term quite precisely. We know roughly how many 25-year-olds will be around 10 years from now based on the number of 15 year olds that are around today and a little bit of adjustment for immigration and mortality. Once we have information about the long-term demographic patterns, we can make very good educated guesses about what those 25-year-olds will be consuming based on what 25-year-olds consume today? Well, if you think of all the other inputs whether it be oil prices, interest rates, there are much more sure footing trying to forecast demographics than any of these other variables. I think so. I think most of these variables are incredibly difficult to forecast. But we do have some few variables involving demand based on demographics that can be fairly precisely measured. Understanding of a few examples of how various consumption goods vary over the age of the household, the demand for the sale, maybe you can take a look at that. Certainly. It's just a coincidence you happen to bring out a profile of beer and liquor by age of the household here. That's right. Beer and liquor. What is interesting about this is even though they're both in the alcohol consumption category, they actually have very different features, which is, the beer consumption peaks much earlier than the liquor consumption. The beer here we have red lines and this is just indicating different years of a survey that you could use to measure, hey, when are people consuming beer? What are 20 years old doing, 30-year-old, 50-year-old doing? That's right. The consumer expenditure survey has done every so often by the federal government and of course, there's more recent data on these profiles. But within that survey, you got very precise data on exactly what a given person is consuming at a point in time and we have their demographic information. The red lines here indicate the consumption of beer in 72 and in 83 by the age of the head of the household. Note that the peak is somewhere in the mid 20s? Yes, that makes sense. Only get to liquor here. It seems like we have and it's really dying out thereafter for the beer consumption. But when we're looking at liquor, that's actually picking up more mid 50s, that's a peak. Indicating that the demographic patterns, it's more, the consumption of liquor depends much more or increases with age, whereas the consumption of Europe years to decrease with age. Indicating that if we want to predict how beer stocks are going to behave, what we really want to understand is the number of 20-year-old or 25-year-olds that are around [inaudible] Today, five years from now, 10 years from now 20 years from now all relevant because the stock price is just the discounted stream of future cashflows. That's right. This is interesting example. Here's another one. We're looking now at bicycles and drugs. Here the peaks are very extreme. Now, what I want to talk about drugs first. Just to clarify, we're talking pharmaceuticals, correct? That's right. These are legal products, Pfizer, Merck, these are not old gone bass product. Pre legalization of marijuana, in which case we might see a different peak than 70 years old for the drug lines. It's certainly possible. In these figures, what we notice is that pharmaceutical consumption appears to be increasing with age of the head of the household. It starts very low in your 20s, and then generally it's upward sloping in all four of the consumer expenditure surveys across the various decades. Compare that to bicycle consumption. Now, what's interesting about bicycle consumption is we're using the age of the head of the household and the age of the head of the peak of bicycle consumption is around age 40. But that's not surprising because if you think people are buying bicycles largely for their children, a 40-year-old probably is the most likely group to have a 10-year-old child who would be riding a bike. Therefore, we see bicycle consumption peaking around when people are having children. Okay, so exactly first one, I was looking at these red lines. I'm like, why isn't this peaking it 5-10? It's obviously the mom and dad are buying the bike there, 40, 45. That's where you see this peak. This is the data you use in your paper to get these forecasts for demand for various industries and then to see is if this demand is incorporated in the stock price. That's right. What we seem to see is that ACT demand. The demographic patterns do clearly predict things like revenue and profitability, indicating they are relevant in any discounted cash-flow model. Then we also see that the stock price responds to these shifts about five years in advance. If there's a bunch of people that will turn, let's say 21 or be in their mid-twenties, three years from now, that's already reflected in the beer stock price. That's right. But if that's happening, let's say eight years down the road, not necessarily. It is not in the stock price. Two years from now, that's when it's likely to enter the stock price. It's a good time to buy years by beer, about five years before the demographic shifts occurs. It's like almost there was a five-year forecasting period, if you will, that the information is more efficiently embedded in the price. Exactly, it takes about five years. Given your research, let me ask you a fundamental question. Finance. What's your take on market efficiency? How efficient do you think the markets are? I generally think that the market is fairly efficient, especially over medium time periods. Of course, minute-to-minute we might see mistakes and they might be exploitable with high-frequency trading. But when we really talking about the functioning of the market and the allocation of capital over longer periods of time, I largely think the market is fairly efficient. It's hard to come up with strategies that can take advantage of the market. Except that you found two. I've found two strategies. However, if you actually try to exploit these strategies, you will quickly find that you start pushing prices back to fundamentals very quickly. It's not clear that a large hedge fund could really operationalize these strategies without moving the market prices so much that the benefits of the strategies quickly. Because for a big investor to get a sizable return, they need to make big investments. That big investment by itself moves the price, eliminating the returns to the strategy. That's leads to another question I had. Do you think these ideas could at least partially be exploited by a mutual fund or hedge fund? I think they can. I think they absolutely can. As long as there's only a small number of mutual funds that are trying to exploit it at any one time. The problem occurs if everyone starts adapting the same strategy very quickly the abnormal performance in these strategies will start to disappear. Well excellent stuff. Really think I appreciate you for taking part in this. Before you wrap it up though we have this favorite segment here called Awkward Moment. Would you be willing to participate? Absolutely. Excellent. This is great news, so let's take a quick break and we'll be right back. I hope you are enjoying this interview with Professor Josh Pollet. Remember that this module will actually feature two Faculty Focus episodes. Please see my upcoming interview with Professor Jake Wing Wong, here you will likely pay more attention to Amazon product ratings after that Faculty Focus episode. Please be sure and stay tuned here, I have an extra special awkward moment with geopolitical implications in store for Josh. As researchers, we all strive for our papers to have impact on the profession. Usually that might be measured by citation counts or popular press mentions. Josh's work has all that but perhaps much more. Josh, I tried to follow world affairs, it's a hobby of mine. I recall back on August 8, 2008, Russia invaded or entered Georgia. I guess it depends on your perspective. We don't have to get into that in this show. We're not going to take a stand. I do remember the date. It was also the opening ceremony for the Beijing Olympics 888. Do you happen to know what day the week this Georgia invasion was? Absolutely no idea. Would it surprise you to learn? It was a Friday. Wow. I did not know that. It's interesting. Yeah. Our staff here and by staff, I mean me, did a little digging. We found it easy to find your published Friday paper. We see a date here of being published April of 2009, it's clearly after this incursion into Georgia. But there's a working paper that we've found that you had produced here. It's actually posted on a publicly available website called ssrrn.com. I was wondering if you could share with us the date that this was posted for public consumption. Apparently it was posted on November 13th, 2005. Exactly. Maybe we could allow our viewers at home to see this as well. Clearly before the Friday in question? Now it's time for the awkward moment. Josh, has your research directly or indirectly played a role in the Friday events in Georgia and more generally do you help time the provide advice on the timing of military excursions to minimize media coverage. I have not ever provided that advice. However, I will admit that the research has been motivated by a well-understood phenomena in politics that bad news is released on Fridays. Indeed there was actually a West Wing episode about this phenomena where politicians choose to release bad news on Fridays to minimize the damage that it does. You always hear the phrase Friday news dump. Friday news dump, exactly. It's bad news. it's almost always. It's invariably bad news. That was what motivated the initial idea of trying to think of whether or not corporations did the same thing. However, I certainly would not like to claim any credit if other governments decide to act badly on Fridays. If other leaders read your papers that's not your responsibility. It's not my responsibility. Friday earnings announcements, yes. Friday invasions or excursions into other countries, no. Exactly, no support for that. I hope you can. This is great, Josh. I hope you don't mind me asking here but we really enjoyed it. I'm sure our viewers got a lot out of your research. I certainly hope so. Remember in Module 4 we actually have two Faculty Focus segments. Please see my interview with Professor Jacon Wong later in the module. As always, thanks for watching. Josh, I forgot to mention, in case you need to cash in these rubles. I know a guy. He's still TV? Yeah, don't worry we might sort this out later. Okay.
