WEBVTT 00:01.110 --> 00:04.850 And I want to talk today about efficient markets, which is a 00:04.850 --> 00:10.080 theory that is a half-truth, I will say. 00:10.080 --> 00:14.500 Before I start, I wanted to just give a few thoughts about 00:14.500 --> 00:18.920 David Swensen's lecture last period. 00:18.920 --> 00:23.460 Let me say, first of all, the Efficient Markets Hypothesis 00:23.460 --> 00:27.750 or the Efficient Markets Theory is a theory that 00:27.750 --> 00:31.110 markets efficiently incorporate all public 00:31.110 --> 00:33.340 information. 00:33.340 --> 00:36.990 And that, therefore, you cannot beat the market, 00:36.990 --> 00:39.940 because the market has all the information in it. 00:39.940 --> 00:41.730 You think you're smarter than the market, 00:41.730 --> 00:43.270 that you know something? 00:43.270 --> 00:45.980 No, the market knows more than you do. 00:45.980 --> 00:49.680 And you'll find out that the market wins every time. 00:49.680 --> 00:51.930 That's the Efficient Markets Hypothesis. 00:51.930 --> 00:55.560 So, it's a very far-reaching hypothesis. 00:55.560 --> 00:59.680 It means that, don't even try to beat the market. 00:59.680 --> 01:00.970 That was a very rudimentary 01:00.970 --> 01:04.320 introduction to today's lecture. 01:04.320 --> 01:11.040 But here I brought in David Swensen, who is claimed to 01:11.040 --> 01:14.870 have beaten the market consistently since 1985, and 01:14.870 --> 01:17.730 dramatically. 01:17.730 --> 01:19.170 And so, what do we make of that? 01:22.260 --> 01:25.710 That's the subject of today's lecture. 01:25.710 --> 01:29.090 By the way, after class one of you came up -- thank you, it 01:29.090 --> 01:31.800 was nice, I don't know where you are -- 01:31.800 --> 01:34.220 one of you came up and thanked Swensen for his 01:34.220 --> 01:36.870 scholarship at Yale. 01:36.870 --> 01:43.510 Yale now has need-blind admissions for the world. 01:43.510 --> 01:47.330 And so people, not just from the United States, people are 01:47.330 --> 01:51.820 helped out, so that people who have managed to meet the high 01:51.819 --> 01:55.339 admission standards here, it makes it possible for them to 01:55.340 --> 01:56.800 actually come here. 01:56.800 --> 02:02.070 And that's substantially David Swensen who did that, because 02:02.070 --> 02:04.550 it's not just the generosity of the university. 02:04.550 --> 02:07.250 They have to have the money to do it. 02:07.250 --> 02:10.070 And so, somehow he seems to have made it. 02:10.070 --> 02:12.620 And I know that there are still many cynics -- the 02:12.620 --> 02:19.130 Efficient Markets Hypothesis has a lot of adherents still. 02:19.130 --> 02:22.230 And as I say, it's a half-truth. 02:22.230 --> 02:25.180 So, some people will say, well, Swensen was just lucky. 02:25.180 --> 02:26.950 And I say, how could he have been lucky for 02:26.950 --> 02:29.410 25 years in a row? 02:29.410 --> 02:32.920 Well, not every single year, but pretty much. 02:32.920 --> 02:37.120 And they say, well, you're picking the one guy out of 02:37.120 --> 02:40.440 millions who is just the luckiest. So, those 02:40.440 --> 02:43.740 arguments are made. 02:49.410 --> 02:53.350 Anyway, one of you asked a question, which I thought it 02:53.350 --> 02:55.090 was very good, at the end. 02:55.090 --> 02:59.090 And that is why, in all of my discussion about Swensen and 02:59.090 --> 03:02.110 all of his talk we never mentioned the Sharpe ratio? 03:04.940 --> 03:07.610 Because as we said, the Sharpe ratio 03:07.610 --> 03:12.090 corrects for risk taking. 03:12.090 --> 03:13.870 That was one of our fundamental lessons. 03:13.870 --> 03:16.360 And we showed you the Efficient Portfolio Frontier, 03:16.360 --> 03:18.200 and the tangency line. 03:18.200 --> 03:22.560 You can get any expected return you want at the expense 03:22.560 --> 03:24.630 of higher uncertainty. 03:24.630 --> 03:28.320 You do a very risky portfolio, and you have high expected 03:28.320 --> 03:30.450 return, because of the risk. 03:30.450 --> 03:32.350 If the risk is measured right. 03:36.470 --> 03:39.130 But I think that's a good very good question. 03:39.130 --> 03:42.980 I caught myself not correcting for it, I just said Yale's 03:42.980 --> 03:45.600 portfolio had a high return. 03:45.600 --> 03:52.540 I didn't correct for standard deviation of return. 03:52.540 --> 03:57.720 So, David Swensen, in his answer, as you recall, 03:57.720 --> 04:02.080 essentially said he doesn't believe in Sharpe ratios, 04:02.079 --> 04:05.849 because we can't measure the standard deviation. 04:05.850 --> 04:10.760 The Sharpe ratio is the excess return of a portfolio over the 04:10.760 --> 04:12.500 market, divided by the standard 04:12.500 --> 04:15.160 deviation of the return. 04:15.160 --> 04:19.840 And that scales it down, so if the excess return is very high 04:19.840 --> 04:23.080 but also has a very high standard deviation, that shows 04:23.080 --> 04:24.990 they were just taking risks. 04:24.990 --> 04:28.770 And so the Sharpe ratio would reveal that. 04:28.770 --> 04:31.370 But Swensen said, I don't think that you can measure the 04:31.370 --> 04:33.200 standard deviation of return. 04:33.200 --> 04:34.040 Isn't that what is answer was? 04:34.040 --> 04:36.120 You were here. 04:36.120 --> 04:38.590 I may not be quoting him exactly right. 04:38.590 --> 04:41.050 So, why wouldn't you be able to measure the standard 04:41.050 --> 04:42.590 deviation of returns? 04:45.360 --> 04:48.940 He gave a reason, which was that, well, when you're 04:48.940 --> 04:53.010 looking at a broad portfolio like Yale's, a lot of the 04:53.010 --> 04:55.570 things in there are private equity -- 04:55.570 --> 04:57.220 that means privately held, so it's not 04:57.220 --> 04:59.180 traded on stock exchanges. 04:59.180 --> 05:02.880 Or it's real estate. 05:02.880 --> 05:06.950 Real estate is only traded every 10 years or 20 years, 05:06.950 --> 05:09.220 and so who knows what it's worth? 05:09.220 --> 05:12.090 All you have is an appraisal, but that's just some 05:12.090 --> 05:16.700 appraiser's estimate, so the standard deviation would be 05:16.700 --> 05:18.280 artificially low. 05:18.280 --> 05:19.920 He's right about that, but I think there's 05:19.920 --> 05:21.760 even more to that. 05:21.760 --> 05:23.340 I know there's more to this. 05:23.340 --> 05:24.590 There's a literature on this. 05:29.775 --> 05:37.355 The point that I wanted to make is that you can do a -- 05:37.360 --> 05:39.970 suppose you're managing money. 05:39.970 --> 05:42.860 And suppose the world out there is evaluating you by 05:42.860 --> 05:45.310 your Sharpe ratio. 05:45.310 --> 05:48.700 And suppose you have no ethics. 05:48.700 --> 05:50.770 You just want money. 05:50.770 --> 05:53.770 This isn't so obviously criminal, this is 05:53.770 --> 05:55.710 not criminal I suppose. 05:55.710 --> 05:58.300 So, you say, I just want to have the best Sharpe ratio for 05:58.300 --> 06:01.340 a number of years running. 06:01.340 --> 06:03.710 I'll get more and more people that will put money in my 06:03.710 --> 06:06.550 investment fund, and eventually I 06:06.550 --> 06:08.620 don't care what happens. 06:08.620 --> 06:11.910 I'll move to Brazil or something. 06:11.910 --> 06:13.100 Some foreign country -- pick out one. 06:13.100 --> 06:15.810 I want to get out of here with the money. 06:15.810 --> 06:18.860 So, all I have to do is fool people into thinking I have a 06:18.860 --> 06:21.030 high Sharpe ratio for a while. 06:21.030 --> 06:22.540 So, what do I do? 06:22.540 --> 06:30.670 Well, there's an interesting paper on this by -- 06:30.670 --> 06:32.030 there's a lot of papers on this, but I'm 06:32.030 --> 06:33.950 going to cite one -- 06:33.945 --> 06:40.715 by Professor Goetzmann and co-authors, here at Yale. 06:43.560 --> 06:53.220 It's actually Goetzmann, Ibbotson, Spiegel, and Welch. 06:53.220 --> 06:55.420 Maybe I'll put all their names on. 06:55.420 --> 06:59.290 Roger Ibbotson is a professor here, Matt 06:59.290 --> 07:03.830 Spiegel, and Ivo Welch. 07:03.830 --> 07:06.120 What they did is they calculated the optimal 07:06.120 --> 07:09.360 strategy for someone who wants to play games 07:09.360 --> 07:11.140 with a Sharpe ratio. 07:11.140 --> 07:15.600 So, you want to fool investors and get a spuriously high 07:15.600 --> 07:17.270 Sharpe ratio. 07:17.270 --> 07:20.510 And they found out what the optimal strategy is. 07:20.510 --> 07:24.840 And that is to sell off the tails of your 07:24.840 --> 07:26.680 distribution of returns. 07:26.680 --> 07:30.270 So, if your return distribution looks like this 07:30.270 --> 07:31.520 -- this is returns. 07:34.460 --> 07:37.620 And you have a probability distribution, say a 07:37.620 --> 07:38.870 bell-shaped curve. 07:42.060 --> 07:44.760 And so the mean and standard deviation of this would be the 07:44.760 --> 07:47.130 inputs to the Sharpe ratio. 07:47.130 --> 07:50.730 But if you're cynical and you want to play tricks, what you 07:50.730 --> 07:53.220 can do is sell the upper tail. 07:56.740 --> 07:59.940 These are very unlikely good events. 07:59.940 --> 08:04.400 Sell them and get money now, and then double up on the 08:04.400 --> 08:05.860 lower tail. 08:05.860 --> 08:09.010 So, you push the lower tail to something like that, and you 08:09.010 --> 08:13.030 wipe out the upper tail, so it goes like that. 08:13.030 --> 08:16.680 So, that means you'll get money, because you sold the 08:16.679 --> 08:17.639 upper tail. 08:17.640 --> 08:20.940 You would do that by selling calls -- we haven't talked 08:20.940 --> 08:22.770 about options yet -- but you can do it by selling 08:22.770 --> 08:27.470 out-of-the-money calls. 08:27.470 --> 08:30.920 And you would do this by writing out-of-the-money puts. 08:30.920 --> 08:34.230 OK, but what you do it you make it, so that if there's 08:34.230 --> 08:38.120 really a bad year, it's going to be a doozer bad year for 08:38.120 --> 08:39.800 your investors. 08:39.800 --> 08:41.210 And if there's ever a good year, then, hey, 08:41.210 --> 08:43.390 you won't get it. 08:43.390 --> 08:46.300 But these good or bad years occur only infrequently. 08:46.300 --> 08:49.030 So, in the meantime, you're making profits from these 08:49.030 --> 08:51.720 sales and you have a high Sharpe ratio. 08:51.720 --> 08:55.030 But little do they know -- 08:55.030 --> 08:57.150 you sold off the tails. 08:57.150 --> 09:00.010 And so nothing happens for many years and you just look 09:00.010 --> 09:03.860 like the best guy there. 09:03.860 --> 09:10.790 So, it turns out that this is not just academic. 09:10.790 --> 09:18.640 There was a company called Integral Investment Management 09:18.640 --> 09:20.310 that did something like this strategy. 09:20.305 --> 09:21.715 It was a hedge fund. 09:24.280 --> 09:31.350 So, it was Integral Investment Management. 09:39.370 --> 09:42.700 It did something like this, by trading in options. 09:42.700 --> 09:45.460 And it got lots of investors to put millions in them. 09:45.460 --> 09:50.630 Notably, the Art Institute of Chicago put $43 million into 09:50.630 --> 09:54.470 this fund and its associated funds. 09:54.470 --> 09:59.040 And then, in 2001, when the market dropped a lot, the Art 09:59.040 --> 10:01.630 Institute of Chicago was wiped out. 10:01.630 --> 10:05.420 They lost almost all of their $43 million. 10:05.420 --> 10:10.310 And so, they got really angry, and they sued this company. 10:10.310 --> 10:14.200 Because they said, you didn't tell us. 10:14.200 --> 10:16.150 What have you been doing? 10:16.150 --> 10:19.300 And then the company pointed out, in its defense, that it 10:19.300 --> 10:23.420 actually said somewhere in the fine print that if markets go 10:23.420 --> 10:26.770 down more than 30%, there would be a problem. 10:26.770 --> 10:30.170 And somehow nobody at the Art Institute read that 10:30.170 --> 10:31.860 or figured it out. 10:31.860 --> 10:34.070 They thought the guy was a genius, because this company 10:34.070 --> 10:37.370 had the highest Sharpe ratio in the industry. 10:37.370 --> 10:38.450 You see what they're doing? 10:38.450 --> 10:40.670 They're playing tricks. 10:40.670 --> 10:44.380 They're making it look like there's less risk than 10:44.380 --> 10:45.540 there really is. 10:45.540 --> 10:48.720 And there's a strategy to do that. 10:48.720 --> 10:51.220 But it didn't end well for Integral Investment 10:51.220 --> 10:55.130 Management, because the Art Institute of Chicago managed 10:55.130 --> 10:57.770 to stick them on other things -- they disclosed it. 10:57.770 --> 11:00.120 They told people that they were doing this strategy. 11:03.350 --> 11:07.040 The artists didn't figure it out. 11:07.040 --> 11:11.180 But there were other dishonesties that they nailed 11:11.180 --> 11:12.430 these guys on. 11:19.090 --> 11:22.010 What Goetzmann and his co-authors did is they showed 11:22.010 --> 11:25.140 that you can play tricks with the -- you can play a lot of 11:25.140 --> 11:26.500 tricks in finance. 11:26.500 --> 11:29.370 But one of them is to play a trick with the Sharpe ratio. 11:29.370 --> 11:30.930 But their trick was very explicit. 11:30.925 --> 11:34.125 It involved particular portfolio composition 11:34.130 --> 11:37.690 involving options, and any professional would immediately 11:37.690 --> 11:40.130 know that that's a trick. 11:40.130 --> 11:43.830 And it didn't work for these guys, they were too aggressive 11:43.830 --> 11:45.360 in their manipulation. 11:45.360 --> 11:50.180 But you can do subtler things as a portfolio manager to get 11:50.180 --> 11:51.920 your Sharpe ratio up. 11:51.920 --> 11:57.960 Instead of manipulating with special derivatives positions 11:57.960 --> 12:07.080 you could just buy companies that have large left tails. 12:07.080 --> 12:11.150 They have a small probability of massive losses. 12:11.150 --> 12:14.210 And you haven't done anything but pick a stock. 12:14.210 --> 12:17.370 And nobody knows whether it really has a small probability 12:17.370 --> 12:21.270 of massive losses, and you could 12:21.270 --> 12:24.150 systematically invest in that. 12:24.150 --> 12:27.250 And then you'd have a high Sharpe ratio for a while, and 12:27.250 --> 12:31.510 then you'd just blow up and lose everything eventually. 12:31.510 --> 12:34.220 I was thinking of an example from recent news. 12:34.220 --> 12:38.750 What about the strategy of investing in Egyptian 12:38.750 --> 12:40.620 companies that are tied to Mubarak? 12:45.960 --> 12:48.280 It might have looked very good for a long time, right? 12:48.280 --> 12:51.320 But the companies might have been underpriced, because 12:51.320 --> 12:55.250 people sensed there's some instability in Egypt. 12:55.250 --> 12:58.100 And look how fast it came on. 12:58.100 --> 13:00.790 I don't know what the outcome will be at this point, but it 13:00.790 --> 13:02.680 just happened -- bang. 13:02.680 --> 13:04.360 That's a tail event, right? 13:04.360 --> 13:07.010 You might look at Egyptian securities and think 13:07.010 --> 13:09.560 everything is stable and fine, it's been 30 years, nothing 13:09.560 --> 13:11.440 has happened. 13:11.440 --> 13:13.560 But someone knows or suspects that there's 13:13.560 --> 13:15.980 something maybe unstable. 13:15.980 --> 13:19.120 And you, as an investor, wouldn't know that by looking 13:19.120 --> 13:20.370 at the numbers. 13:24.620 --> 13:28.120 What I'm getting at is really what is the essence of 13:28.120 --> 13:32.440 Swensen's skill or contribution? 13:32.440 --> 13:36.700 It's not his ability to manipulate Sharpe ratios or 13:36.700 --> 13:38.030 numbers like that. 13:38.030 --> 13:41.870 I think what it has, to me, in my mind, is something to do 13:41.870 --> 13:50.370 with character and his real self and his real objectives. 13:53.190 --> 13:57.150 And this gets at what people in finance are really doing. 13:57.150 --> 14:00.670 I think that when you take a finance course in a 14:00.670 --> 14:06.930 university, you may not get a proper appreciation of how 14:06.930 --> 14:09.130 one, in a career in finance, develops a 14:09.130 --> 14:12.130 reputation for integrity. 14:12.130 --> 14:15.940 Nobody can really judge what you're doing as an investor, 14:15.940 --> 14:18.450 because they can't judge it from the statistics. 14:18.450 --> 14:21.150 Even though we've developed this nice theory about Sharpe 14:21.150 --> 14:25.900 ratios and the like, you end up judging the person and what 14:25.900 --> 14:27.590 the person's real objectives are. 14:31.090 --> 14:33.260 I'll probably come back to that theme again. 14:33.260 --> 14:36.900 Let me also say that's about doing the Goetzmann -- you 14:36.900 --> 14:38.150 understand the strategy? 14:41.730 --> 14:44.650 It's like investing in securities with 14:44.650 --> 14:46.710 time bombs in them. 14:46.710 --> 14:48.200 They're going to go off eventually, you don't know 14:48.200 --> 14:49.490 exactly when. 14:49.490 --> 14:53.830 And they look good for a while, but they'll blow up. 14:53.830 --> 14:55.590 What does the law say about this? 14:55.590 --> 15:00.240 Well, the law in the United States and other countries 15:00.240 --> 15:05.310 emphasizes that an investment manager must not fail to 15:05.310 --> 15:09.840 disclose relevant information about a security. 15:09.840 --> 15:12.270 And it has to be more than boilerplate disclosure. 15:16.010 --> 15:18.540 For, say Integral Investment Management, you could write up 15:18.540 --> 15:22.020 a prospectus and say, but of course past returns are not a 15:22.020 --> 15:25.160 guide to the future and something could go wrong. 15:25.160 --> 15:28.300 That's a boilerplate disclosure, because people 15:28.300 --> 15:30.460 think, well, that's what everybody says. 15:30.460 --> 15:34.120 The law says that you have to actually actively disclose. 15:34.120 --> 15:37.970 If there's something that's relevant that would make your 15:37.970 --> 15:42.460 statistics misleading, you have to get their attention 15:42.460 --> 15:44.180 and explain it to them. 15:44.180 --> 15:46.670 That's the law. 15:46.670 --> 15:48.620 So, I think it's laws like that, and it's the 15:48.620 --> 15:50.140 development of -- 15:50.140 --> 15:54.130 it's not something that we specialize in academia. 15:54.130 --> 15:56.770 Well, we are, we're trying to develop character I suppose. 15:56.765 --> 15:59.305 But it's not just Sharpe ratios. 15:59.310 --> 16:03.090 And I think that tendencies to rely on numbers like this has 16:03.090 --> 16:06.780 led to errors in the past. 16:06.780 --> 16:12.440 So, let me go more directly into today's lecture. 16:12.440 --> 16:24.910 It's about the Efficient Markets Hypothesis, which is, 16:24.910 --> 16:30.260 to me, a fascinating theory, which is not completely true, 16:30.260 --> 16:33.090 which makes it all the more interesting. 16:33.090 --> 16:36.850 The first statement that I could find -- 16:36.850 --> 16:38.980 I'm interested in the history of thought -- 16:38.980 --> 16:44.540 so, the first statement of the efficient markets hypothesis 16:44.536 --> 16:52.456 that I could find was in a book by George -- 16:56.630 --> 17:07.110 [SIDE CONVERSATION] 17:07.109 --> 17:10.399 PROFESSOR ROBERT SHILLER: George Gibson, who wrote a 17:10.400 --> 17:18.560 book in 1889 called The Stock Exchanges of London, Paris, 17:18.560 --> 17:19.810 and New York. 17:23.330 --> 17:27.310 I'm quoting George Gibson, 1889: "When shares become 17:27.310 --> 17:31.040 publicly known in an open market, the value, which they 17:31.040 --> 17:35.330 acquire there, may be regarded as the judgment of the best 17:35.330 --> 17:37.830 intelligence concerning them." 17:37.830 --> 17:41.720 He described the stock market as a kind of voting machine 17:41.720 --> 17:42.660 where people vote. 17:42.660 --> 17:47.130 If you think a share is worth more, you vote by buying it. 17:47.130 --> 17:50.580 If you think it's worth less in the market, you sell it. 17:50.580 --> 17:53.560 And everybody in the world can do that. 17:53.560 --> 17:54.630 It's open to the public. 17:54.625 --> 17:58.665 So, the smartest people get into it, and then they soon 17:58.670 --> 18:01.930 make a lot of money doing it, so they have a lot of votes. 18:01.930 --> 18:04.750 So, the smarter people have more votes. 18:04.750 --> 18:05.850 And you've got everyone there. 18:05.850 --> 18:10.850 If anyone has a special clue, they go right in and they buy 18:10.850 --> 18:13.180 if it's positive, or they sell if it's negative. 18:13.180 --> 18:15.940 So, the smartest people go right in there and 18:15.940 --> 18:18.000 aggressively affect the value. 18:18.000 --> 18:19.770 Until it's right, and then there's no there's no 18:19.770 --> 18:22.190 incentive to buy or sell. 18:22.190 --> 18:23.620 The other thing about Gibson -- 18:23.620 --> 18:26.100 I should have copied the quote, but as I remember from 18:26.100 --> 18:30.870 the book, he says something like ''in our modern electric 18:30.870 --> 18:38.180 age, information flows with the speed of light.'' I say, 18:38.180 --> 18:40.160 what is he talking about in 1889? 18:40.160 --> 18:41.340 Well, you know what he's talking about. 18:41.340 --> 18:42.560 The telegraph. 18:42.560 --> 18:47.900 In fact, they had ticker machines. 18:47.900 --> 18:50.500 They were electronic printers that would 18:50.500 --> 18:53.020 print out stock quotes. 18:53.020 --> 18:57.350 So, they were really in the information age by 1889. 18:57.350 --> 19:00.590 So, this is what happened, Gibson said, you can't beat 19:00.590 --> 19:01.430 the market. 19:01.430 --> 19:05.470 It's just smarter because -- it's like Wikipedia is smarter 19:05.470 --> 19:06.510 than any one of us, right? 19:06.510 --> 19:07.820 Because it puts together all the 19:07.820 --> 19:09.270 thinking of all the people. 19:09.270 --> 19:12.260 Well, they had Wikipedia of a sort, because they had the 19:12.260 --> 19:13.920 stock market. 19:13.920 --> 19:16.460 They had the price. 19:16.460 --> 19:21.800 So, that is the statement that -- 19:21.800 --> 19:23.590 he didn't use the word ''Efficient Markets.'' 19:23.590 --> 19:27.160 Actually, tried to find the origin of "Efficient Markets." 19:27.160 --> 19:31.120 Sometimes in the 19th century, people would say ''Efficient 19:31.120 --> 19:38.840 Markets.'' But it wasn't a cliche yet, it wasn't a phrase 19:38.840 --> 19:41.390 that would be recognizable. 19:41.390 --> 19:42.960 Even in this context, they would use it. 19:42.960 --> 19:44.810 But it was not a name for a theory yet. 19:48.250 --> 19:54.040 The next Efficient Markets theorist -- and I put this on 19:54.040 --> 19:54.990 your reading list -- 19:54.990 --> 20:02.220 Charles Conant, who wrote a 1904 book called Wall Street 20:02.220 --> 20:04.330 and the Country. 20:04.330 --> 20:08.590 I put one chapter on that on the reading list, because it 20:08.590 --> 20:12.970 was a statement of the Efficient Markets Hypothesis, 20:12.970 --> 20:17.410 which was remarkably well-written, I thought. 20:17.410 --> 20:21.330 He starts out the chapter by pointing out that a lot of 20:21.330 --> 20:24.400 people think speculation is a kind of gambling, 20:24.400 --> 20:26.080 or a kind of evil. 20:26.080 --> 20:28.850 Speculating on the stock market, that sounds like some 20:28.850 --> 20:32.870 wild activity that ought to be ruled out. 20:32.870 --> 20:37.260 Then he said, how can that possibly be true? 20:37.260 --> 20:40.970 The stock market is a central institution 20:40.970 --> 20:43.160 of all modern economies. 20:43.160 --> 20:45.580 To think that it's just gambling just 20:45.580 --> 20:47.270 defies common sense. 20:47.270 --> 20:52.440 Then he goes on and describes what it is that it does. 20:52.440 --> 20:57.350 And there's some really nice passages in Conant's book. 20:57.350 --> 21:00.360 In one passage he says, suppose for a moment that 21:00.360 --> 21:01.760 stock markets of the world were 21:01.760 --> 21:04.260 closed, what would happen? 21:04.260 --> 21:07.250 He said no one would know what anything is worth. 21:07.250 --> 21:10.090 No one could make any calculated decisions. 21:10.090 --> 21:13.200 He said that, in fact, capital moves around from one industry 21:13.200 --> 21:17.210 to another in respect to the prices that are quoted in 21:17.205 --> 21:17.875 these markets. 21:17.880 --> 21:20.370 And if you didn't see the prices, you would be blind. 21:23.060 --> 21:26.730 It's often said about the Soviet economy -- 21:26.730 --> 21:29.960 which did not have stock markets or 21:29.960 --> 21:32.190 financial markets -- 21:32.190 --> 21:35.850 that they relied on prices in the rest of the world. 21:35.850 --> 21:38.290 Nobody in the Soviet Union could plan very well, because 21:38.290 --> 21:40.110 they didn't know what anything was worth. 21:40.110 --> 21:42.460 But at least they had the rest of the world, and they 21:42.460 --> 21:44.460 thought, well that's an approximation to what prices 21:44.460 --> 21:46.020 ought to be in the Soviet Union. 21:58.400 --> 22:01.180 I just wanted to reiterate a little bit about the intuition 22:01.180 --> 22:03.940 of Efficient Markets. 22:03.940 --> 22:08.620 The idea is that if you trade securities, the advantage to 22:08.620 --> 22:11.870 being there a little bit ahead of anyone else is enormous. 22:11.870 --> 22:15.450 If you know five minutes before the other investors 22:15.450 --> 22:18.110 about some good news or bad news, either way it doesn't 22:18.110 --> 22:20.940 matter -- you know it five minutes earlier, you jump 22:20.940 --> 22:22.610 right in and trade. 22:22.610 --> 22:25.660 You can trade ahead of them and prices haven't changed 22:25.660 --> 22:26.970 yet, you make money. 22:26.970 --> 22:30.870 So, that has created an industry that speeds 22:30.870 --> 22:33.160 information. 22:33.160 --> 22:37.780 The first such industry that I would tell 22:37.780 --> 22:42.020 you about is Reuters. 22:42.020 --> 22:46.670 Mr. Reuters, I think in the 1840s before the telegraph, 22:46.670 --> 22:51.810 created a financial information service using 22:51.810 --> 22:53.910 carrier pigeons. 22:53.910 --> 22:56.200 You know, these are birds. 22:56.200 --> 22:59.640 So, when he was in London with some information, they had 22:59.640 --> 23:02.490 carrier pigeons that were brought from Paris, and as 23:02.490 --> 23:05.490 soon as new information came out, they would tie it to the 23:05.490 --> 23:07.570 foot of the little bird and they'd let it go, and it would 23:07.570 --> 23:09.170 fly to Paris. 23:09.170 --> 23:11.420 It would go to its roosting place, the message would be 23:11.420 --> 23:14.070 read, and subscribers would be notified. 23:14.070 --> 23:17.700 And that was no joke, that really worked, because you 23:17.700 --> 23:21.340 would have information hours or even days before everyone 23:21.340 --> 23:22.130 else in Paris. 23:22.130 --> 23:23.960 So, you could make a killing. 23:23.960 --> 23:27.300 So, Reuters today, it's now called Thomson Reuters, is 23:27.300 --> 23:29.980 still in that business. 23:29.980 --> 23:33.990 The carrier pigeons were a brilliant idea. 23:33.990 --> 23:36.840 But they had to keep up with the times, because, shortly 23:36.840 --> 23:40.360 thereafter, the telegraph was invented and pigeons no longer 23:40.360 --> 23:42.580 were the leading technology. 23:42.580 --> 23:44.830 But maybe that was the beginning of the information 23:44.830 --> 23:47.030 age, with pigeons. 23:47.030 --> 23:50.410 Now we have beepers, and we have the internet. 23:50.410 --> 23:53.490 A beeper is something you can carry in your pocket that 23:53.490 --> 23:55.250 beeps when there's financial news. 23:55.250 --> 24:01.640 So, suppose a company makes an announcement that that it has, 24:01.640 --> 24:06.930 say, a new drug that's successful in trials. 24:06.930 --> 24:11.210 As soon as they make the announcement, it then goes out 24:11.210 --> 24:15.210 electronically everywhere, and the beepers start beeping. 24:15.210 --> 24:17.460 And all the investment analysts, they drop their 24:17.460 --> 24:20.890 morning coffee, because they know they have to act fast. 24:20.890 --> 24:22.880 So, you get this new announcement. 24:22.880 --> 24:25.010 Now it's t plus 20 seconds. 24:25.010 --> 24:28.200 He's got his drug specialist on the phone. 24:28.200 --> 24:28.980 What does this mean? 24:28.980 --> 24:31.540 Quick, how much is it going to go up? 24:31.540 --> 24:34.020 And so the guy says, I don't know, first thought, maybe 24:34.020 --> 24:36.060 it's going to go up $2 a share. 24:36.060 --> 24:39.480 OK, it's only gone up $1 a share, I'll buy right now. 24:39.475 --> 24:45.395 And now it's two minutes after the announcement, and then the 24:45.400 --> 24:47.690 analyst says, I've thought about it a little bit more. 24:47.690 --> 24:51.260 No, I only think it's $1.50 a share. 24:51.260 --> 24:54.210 This is homing in, and so the price is jiggling around 24:54.210 --> 24:57.310 rapidly as all this is happening for a few minutes, 24:57.310 --> 24:58.430 and then it settles down. 24:58.430 --> 24:59.740 Because after -- 24:59.742 --> 25:03.712 I may be exaggerating -- after 10 minutes, they've kind of 25:03.710 --> 25:05.610 figured it out and it's reached its new level. 25:09.130 --> 25:11.890 They'll be thinking about it the next morning when they're 25:11.890 --> 25:14.770 taking their shower, and they'll get a better and 25:14.770 --> 25:18.690 better, more refined idea of what the price is. 25:18.690 --> 25:20.300 But here's the Efficient Markets theme. 25:20.300 --> 25:22.870 You, the next day, read about it in the Wall Street Journal 25:22.870 --> 25:23.710 in the morning. 25:23.710 --> 25:26.840 You're now 24 hours late. 25:26.840 --> 25:29.950 So, you call your broker and say, maybe I should buy this 25:29.950 --> 25:32.550 stock, they've got this new breakthrough. 25:32.550 --> 25:34.830 Your broker might laugh at you, right? 25:34.830 --> 25:37.580 Because you're 24 hours late. 25:37.580 --> 25:42.850 And what do you know, anyway, about pharmaceuticals? 25:42.850 --> 25:45.970 That's where the Efficient Markets Hypothesis is true. 25:45.970 --> 25:50.870 You can't expect to routinely profit from information that's 25:50.870 --> 25:52.470 already out there. 25:52.470 --> 25:54.550 If you're going to profit, you've got to come up with 25:54.550 --> 26:00.660 something faster, something that you can get faster. 26:00.660 --> 26:04.090 I think this is somewhat what David Swensen was referring to 26:04.090 --> 26:04.550 yesterday [correction: last lecture], 26:04.550 --> 26:08.840 when he talked about different asset classes. 26:08.840 --> 26:11.310 Remember how he talked about comparing the top quartile and 26:11.310 --> 26:17.520 bottom quartile of investment managers, in terms of returns, 26:17.520 --> 26:19.730 or at different asset classes? 26:19.730 --> 26:22.530 Well, the managers weren't able to beat the 26:22.530 --> 26:25.360 bond market very much. 26:25.360 --> 26:27.450 The top quartile wasn't able to beat the 26:27.450 --> 26:28.990 stock market very much. 26:28.990 --> 26:32.930 But when you get to unusual assets, private equity, which 26:32.930 --> 26:36.850 is not traded on stock exchanges, or absolute return 26:36.850 --> 26:38.710 investments that he talked about -- they're unusual, 26:38.710 --> 26:41.990 smaller, rare investments that the public doesn't have a lot 26:41.990 --> 26:43.020 of information about. 26:43.020 --> 26:46.590 And these guys can get ahead on those things. 26:46.590 --> 26:51.750 So, part of what makes Swensen a success is picking his game. 26:51.750 --> 26:54.420 Even so the stock market is not completely efficient, but 26:54.420 --> 26:58.180 it's so much more efficient, because it's so many people 26:58.180 --> 26:59.430 involved in it and watching it. 27:03.200 --> 27:05.110 So, maybe I should write here, because this is what we're 27:05.110 --> 27:06.360 talking about here. 27:06.360 --> 27:09.100 The Efficient Markets Hypothesis. 27:12.474 --> 27:15.574 I'll write it down. 27:15.570 --> 27:23.730 This is the name for this idea that was coined -- 27:23.730 --> 27:29.730 or sometimes people say Efficient Markets theory. 27:29.730 --> 27:32.600 They're referring to what Conant and Gibson and other 27:32.600 --> 27:35.310 people had been talking about for a long time, and it was 27:35.310 --> 27:37.470 common knowledge. 27:37.470 --> 27:48.790 But the first person to use this term apparently was Harry 27:48.790 --> 27:55.240 Roberts, a professor at the University of Chicago. 27:55.240 --> 28:01.610 But he was made famous by Eugene Fama, who referred to 28:01.610 --> 28:04.050 it as Harry Roberts's idea. 28:04.050 --> 28:09.290 Eugene Fama is maybe the best-known finance professor 28:09.290 --> 28:10.140 in the country, I think. 28:10.140 --> 28:11.910 He is also at the University of Chicago. 28:14.806 --> 28:18.206 He's been talked about as a Nobel Prize candidate for a 28:18.210 --> 28:22.280 long time, and he should have won probably, even though his 28:22.280 --> 28:24.270 theory is not entirely right. 28:24.270 --> 28:27.580 I think his chances of getting it have dimmed a little bit, 28:27.580 --> 28:31.670 because the theory is not looked upon as quite such an 28:31.670 --> 28:35.260 absolute truth as it used to be. 28:35.260 --> 28:36.550 As I said, it's a half-truth. 28:39.770 --> 28:44.400 What Conant said is all well-taken and right, but 28:44.400 --> 28:47.950 there's other nuances and it's not exactly -- 28:47.950 --> 28:49.930 when you first read Conant, you 28:49.930 --> 28:50.980 think, the guy is brilliant. 28:50.980 --> 28:53.470 That's what I thought, this is right on. 28:53.470 --> 28:56.180 Then you read it again, you think, well, you know maybe 28:56.180 --> 28:58.750 there is a little gambling in the financial market. 28:58.750 --> 29:00.880 Things don't always work right. 29:00.880 --> 29:04.380 So, he was maybe a little bit too positive. 29:10.700 --> 29:13.410 I can give a little history of this. 29:13.410 --> 29:18.020 It was in 1960. 29:18.020 --> 29:20.190 The University of Chicago is kind of the 29:20.190 --> 29:22.170 forerunner in this. 29:22.170 --> 29:27.170 In 1960, Ford Foundation gave a grant to the University of 29:27.170 --> 29:33.210 Chicago to assemble all stock price data back to 1926. 29:33.210 --> 29:35.550 And to get it right. 29:35.550 --> 29:41.230 So, they set up the Center for Research in Securities Prices 29:41.230 --> 29:42.480 at Chicago. 29:47.080 --> 29:51.540 The Center for Research in Securities Prices, or CRSP, as 29:51.540 --> 29:55.970 it's called, had a Ford Foundation grant to go to the 29:55.970 --> 30:00.330 stock exchanges in the United States and get all the data, 30:00.330 --> 30:01.790 and get it right. 30:01.790 --> 30:04.770 Remember, I told you they have splits in stocks. 30:04.770 --> 30:08.080 So, you see the price of a share, and then suddenly the 30:08.080 --> 30:10.770 price will fall in half or maybe it will double, because 30:10.770 --> 30:14.250 they changed the units of measurement. 30:14.250 --> 30:17.300 If you wanted to know, what is the price history of stocks 30:17.300 --> 30:20.110 over the long haul, no one had ever organized that and 30:20.110 --> 30:23.170 figured out things like that, and got it right. 30:23.170 --> 30:24.430 And when were the dividends paid? 30:24.430 --> 30:27.050 When did you actually get the dividends? 30:27.050 --> 30:29.710 So they said, let's get it right. 30:29.710 --> 30:33.200 Let's put it on -- hey, this is really super modern -- a 30:33.200 --> 30:35.290 UNIVAC tape. 30:35.290 --> 30:37.870 That's a computer tape. 30:37.870 --> 30:40.130 We'll put it on a tape and we'll sell it at cost to 30:40.130 --> 30:41.900 anybody in the world. 30:41.900 --> 30:44.650 And so the Ford Foundation, which is a non-profit, said: 30:44.650 --> 30:47.350 Good idea, let's do it. 30:47.350 --> 30:50.230 So, that CRSP tape has launched a revolution in 30:50.230 --> 30:52.160 finance, because nobody had the data. 30:52.160 --> 30:56.020 They were throwing it away, it was not used. 30:56.020 --> 30:58.050 How do you know what Sharpe ratios are if you 30:58.050 --> 31:00.320 don't have the data? 31:00.320 --> 31:02.330 I mean, you could find it in newspapers, but it wasn't 31:02.330 --> 31:05.280 organized right, it wasn't set up right. 31:05.280 --> 31:10.190 So, with the invention of the CRSP tape in 1960, it really 31:10.185 --> 31:14.065 gave impetus to the Efficient Markets revolution. 31:14.070 --> 31:17.410 And by the end of that decade, there were thousands of 31:17.410 --> 31:19.390 articles testing market efficiency 31:19.390 --> 31:21.910 using the CRSP tape. 31:21.910 --> 31:27.830 And, in particular in 1969, Eugene Fama wrote one of the 31:27.830 --> 31:30.200 most cited articles in the history of finance. 31:30.200 --> 31:34.240 It was called "Efficient Capital Markets: A Review". 31:34.240 --> 31:38.930 So, he reviews all of these studies of the CRSP tape. 31:38.930 --> 31:41.110 And it looked authoritative for the first time, because we 31:41.110 --> 31:43.750 were using the whole universe of stocks, all 31:43.750 --> 31:46.730 the way back to 1926. 31:46.730 --> 31:49.380 That sounded like a long time, and a lot of data. 31:49.380 --> 31:56.580 And Fama said it's not uniform, there are some 31:56.580 --> 31:59.900 negative results, but the evidence is that markets are 31:59.900 --> 32:01.710 remarkably efficient. 32:01.710 --> 32:05.080 And this is a truth that we've discovered. 32:05.080 --> 32:09.020 And that was really a bombshell, because he was 32:09.020 --> 32:13.110 discrediting, or seeming to discredit, practically all the 32:13.110 --> 32:14.770 investment managers in the country. 32:14.770 --> 32:16.260 It was a huge industry. 32:16.260 --> 32:19.220 And he's claiming the successful ones must have just 32:19.220 --> 32:23.060 been lucky, because the market is so efficient that we can't 32:23.060 --> 32:26.560 see any way that you could make money in the market. 32:32.770 --> 32:36.760 The high point of the Efficient Markets Hypothesis 32:36.760 --> 32:40.470 was probably in the 1970s. 32:40.470 --> 32:42.610 I'll call that the high point. 32:42.610 --> 32:46.500 It seemed, at that time, that all of the scholars were 32:46.500 --> 32:50.120 finding that there was no way to beat the market. 32:50.120 --> 32:55.230 But it started to deteriorate, the support for the Efficient 32:55.230 --> 32:56.480 Markets Hypothesis. 33:04.620 --> 33:07.380 There's been a change in thinking. 33:07.380 --> 33:11.550 Efficient Markets is still regarded with respect, but not 33:11.550 --> 33:17.820 the same respect that it had in 1969 or in 1979. 33:17.820 --> 33:22.220 I have here some indication of how thinking has changed about 33:22.220 --> 33:25.240 Efficient Markets. 33:25.240 --> 33:28.050 The textbook that I used to use for this course -- 33:28.050 --> 33:30.650 I've been teaching this course for 25 years -- 33:30.650 --> 33:31.930 Fabozzi et al. 33:31.930 --> 33:34.390 wasn't even written when I started, so I was using a 33:34.390 --> 33:41.290 textbook called Brealey & Myers, Principles 33:41.290 --> 33:44.220 of Corporate Finance. 33:44.220 --> 33:46.730 I still have all these old editions. 33:46.730 --> 33:47.980 It's a very successful textbook. 33:52.530 --> 33:55.160 Since I was teaching out of it all those years, I went back 33:55.159 --> 33:55.749 and looked -- 33:55.750 --> 33:57.860 I don't have the first edition of that book, I have the 33:57.860 --> 34:01.950 second edition, 1984 -- and I was teaching out of it in this 34:01.949 --> 34:05.889 same class in 1984. 34:05.889 --> 34:09.969 At the end of that book, there's a chapter on the seven 34:09.969 --> 34:12.929 most important ideas in finance. 34:12.929 --> 34:15.729 And one of the ideas is Efficient Markets. 34:15.730 --> 34:19.370 And quoting the textbook, Brealey and Myers say: 34:19.370 --> 34:23.870 "Security prices accurately reflect available information 34:23.870 --> 34:27.300 and respond rapidly to new information as soon as it 34:27.300 --> 34:32.370 becomes available." They they do qualify it, in 1984. 34:32.370 --> 34:35.270 "Don't misunderstand the Efficient Markets idea. 34:35.270 --> 34:38.470 It doesn't say there are no taxes or costs. 34:38.469 --> 34:41.269 It doesn't say there aren't some clever people and some 34:41.270 --> 34:45.810 stupid ones, it merely implies that competition in actual 34:45.810 --> 34:48.070 capital markets is very tough. 34:48.070 --> 34:52.390 There are no money machines and security prices reflect 34:52.390 --> 34:57.260 the true underlying value of assets." Let me repeat that: 34:57.260 --> 35:00.930 "Security prices reflect the true underlying value of 35:00.930 --> 35:08.150 assets." That's a pretty strong statement, right? 35:08.150 --> 35:09.730 But that's Efficient Markets. 35:09.730 --> 35:13.910 They just said it in their second edition of the book. 35:13.910 --> 35:15.530 Not many people would say that, right? 35:15.530 --> 35:18.720 Trust the stock market, don't trust people you know and love 35:18.720 --> 35:21.410 and trust. Trust the stock market. 35:21.410 --> 35:24.500 Well, they deleted that from later editions of their book. 35:24.500 --> 35:29.460 I think that's a sign of changes. 35:29.460 --> 35:32.200 So, I was looking at the 2008 edition. 35:32.200 --> 35:34.220 They've now taken on a third author. 35:34.220 --> 35:36.180 They're getting tired of coming out with more and more 35:36.180 --> 35:38.660 editions of their book, so they've taken on Franklin 35:38.660 --> 35:42.050 Allen from the Wharton School. 35:42.050 --> 35:45.030 They've deleted what I just read, and now it says -- 35:45.030 --> 35:49.080 I'm quoting from them: "Much more research is needed before 35:49.080 --> 35:53.710 we have a full understanding of why asset prices sometimes 35:53.710 --> 35:57.630 get so out of line with what appears to be their discounted 35:57.630 --> 36:00.130 future payoffs." That's a complete 36:00.130 --> 36:01.920 turnaround in the textbook. 36:01.920 --> 36:05.330 This is one of the most popular textbooks, and they've 36:05.330 --> 36:07.270 changed completely. 36:07.270 --> 36:13.410 So, I think we have an idea that started around the 1960s. 36:13.410 --> 36:16.870 It was somehow associated with computers, and electronic 36:16.870 --> 36:19.480 databases, and modern thinking, and 36:19.480 --> 36:21.560 mathematical finance. 36:21.560 --> 36:24.120 They kind of went too far with it. 36:24.120 --> 36:28.000 They concluded that you just can't beat the market. 36:28.000 --> 36:30.910 Another thing I put on the reading list is a reading from 36:30.910 --> 36:32.670 The New Yorker magazine. 36:32.670 --> 36:34.240 It just came out in December. 36:36.780 --> 36:38.620 I thought it was relevant. 36:38.620 --> 36:42.660 It's by Jonah Lehrer and the title of the article is "The 36:42.660 --> 36:48.450 Truth Wears Off: Is There Something Wrong With the 36:48.450 --> 36:53.160 Scientific Method?" I don't think there's anything wrong 36:53.160 --> 36:55.750 with the scientific method, but I was interested in this 36:55.750 --> 36:59.710 article, because what The New Yorker article points out is 36:59.710 --> 37:03.400 that a lot of scientists -- and this is outside of 37:03.400 --> 37:07.880 finance, I'm making a parallel here -- a lot of scientists 37:07.880 --> 37:12.440 who follow careful scientific procedures seem to generate 37:12.440 --> 37:15.940 results that are later discredited. 37:15.940 --> 37:17.870 And nobody can figure out why. 37:17.870 --> 37:21.280 It's like the universe is changing. 37:21.280 --> 37:25.060 He gives an example in The New Yorker article -- 37:25.060 --> 37:26.880 and this is from drugs -- 37:26.880 --> 37:30.770 there's a class of drugs called second-generation 37:30.770 --> 37:32.500 antipsychotics. 37:32.500 --> 37:38.110 These are used for people who are either schizophrenic or -- 37:38.110 --> 37:40.760 I guess it can be used more generally than that -- some of 37:40.760 --> 37:44.500 the drugs are called Abilify, Seroquel, Zyprexa. 37:44.500 --> 37:49.540 When these drugs were first introduced, careful studies 37:49.540 --> 37:52.800 that passed muster in the best medical journals found that 37:52.800 --> 37:54.580 they were highly effective. 37:54.580 --> 37:59.130 And they were written up as a godsend, a way of dealing with 37:59.130 --> 38:02.380 problems that used to weigh on people. 38:02.380 --> 38:03.630 Wonderful. 38:05.400 --> 38:09.190 The medical procedures involved careful controls on 38:09.190 --> 38:14.130 studies including a double blind procedure. 38:14.130 --> 38:19.600 When you want to test a drug on human subjects, both the 38:19.600 --> 38:22.660 subject doesn't know whether he or she is getting the drug, 38:22.660 --> 38:26.060 and the experimenter, who runs experiment, doesn't know which 38:26.060 --> 38:27.680 one is the drug. 38:27.680 --> 38:31.430 So, you give bottle A and bottle B to the experimenter, 38:31.430 --> 38:36.050 and the experimenter is never told, which one is Zyprexa and 38:36.050 --> 38:37.790 which one is a placebo. 38:37.790 --> 38:40.640 And then the experimenter has to write up a whole report on 38:40.640 --> 38:44.520 drug A and drug B, not even knowing. 38:44.520 --> 38:47.260 This is to eliminate any possible bias. 38:47.260 --> 38:48.730 So, these drugs passed that. 38:48.730 --> 38:49.340 You see what I'm saying? 38:49.340 --> 38:52.680 The controls were right, everything was good. 38:52.680 --> 38:57.590 And as years go by, the tests start coming out -- the new 38:57.590 --> 39:00.630 attempts to replicate those start 39:00.630 --> 39:02.160 coming out more negative. 39:02.160 --> 39:05.290 They didn't disprove the drugs, they just weren't such 39:05.290 --> 39:08.560 wonder drugs as they thought. 39:08.560 --> 39:10.380 So, how can that be? 39:10.380 --> 39:13.480 And what the article says, well, it must be that somehow 39:13.480 --> 39:16.410 scientific bias, when there's an enthusiasm for some new 39:16.410 --> 39:19.760 theory, it creeps in even if you try to make 39:19.760 --> 39:21.480 the strongest controls. 39:21.480 --> 39:24.930 You say, how could that happen with a double blind procedure? 39:24.930 --> 39:27.120 Well, maybe they broke the double blind somehow. 39:27.120 --> 39:31.090 They tried, but the guy, experimenter, figured it out. 39:31.090 --> 39:34.420 And then he started not deliberately fabricating 39:34.420 --> 39:38.730 results, but it's the kind of thing where one subject says, 39:38.730 --> 39:41.740 I didn't take my Abilify regularly. 39:41.740 --> 39:43.970 I took two tablets last week. 39:43.970 --> 39:46.390 I have to decide whether to throw this person out of the 39:46.390 --> 39:50.360 sample, and then I kind of remember that the drug wasn't 39:50.360 --> 39:53.300 working for this person, and it colors my judgment, so I 39:53.300 --> 39:56.290 throw them out. 39:56.290 --> 39:58.900 And another thing that happens is that the studies that 39:58.900 --> 40:01.310 didn't find it might have been suppressed. 40:01.310 --> 40:03.940 Someone might have done an Abilify test 40:03.940 --> 40:05.790 and gotten bad results. 40:05.790 --> 40:07.860 And then showed it to their superior and said, should I 40:07.860 --> 40:08.530 publish this? 40:08.530 --> 40:10.950 And the superior said, wait a minute, there must be 40:10.950 --> 40:12.000 something wrong here. 40:12.000 --> 40:14.550 Abilify is wonderful, so let's look. 40:14.550 --> 40:16.620 And then they find something that might be wrong, and he 40:16.620 --> 40:17.800 says, you shouldn't publish this, 40:17.800 --> 40:21.170 because they find something. 40:21.170 --> 40:26.340 So, the publication process is biased for a while, but 40:26.340 --> 40:28.790 eventually it catches up. 40:28.790 --> 40:30.690 So, I think the same thing happened with the Efficient 40:30.690 --> 40:31.640 Markets Hypothesis. 40:31.640 --> 40:34.810 In the initial enthusiasm, anybody who found that the 40:34.810 --> 40:38.410 Efficient Markets Hypothesis wasn't supported by the 40:38.410 --> 40:41.350 evidence, that person would be told, look again. 40:41.350 --> 40:43.710 Maybe you've done something wrong. 40:43.710 --> 40:48.390 So that's what happened. 40:51.100 --> 40:58.000 I wanted to do a little bit more history and describe the 40:58.000 --> 41:03.880 concept of Random Walk, which is central to the Efficient 41:03.880 --> 41:05.130 Markets Hypothesis. 41:12.700 --> 41:15.620 Let me start, though, with a little bit more history. 41:15.620 --> 41:16.870 Technical Analysis. 41:25.290 --> 41:27.150 This term goes back -- 41:27.150 --> 41:29.270 must be over 100 years. 41:29.270 --> 41:33.810 Technical analysis is the analysis of stock prices, or 41:33.810 --> 41:38.480 maybe other speculative asset prices, by looking at charts 41:38.480 --> 41:44.890 of the prices and looking for patterns that suggest 41:44.890 --> 41:47.440 movements in prices. 41:47.440 --> 41:51.500 The classic text of technical analysis is Edwards and McGee. 41:54.190 --> 41:56.980 McGee, there's a famous story about him. 41:56.980 --> 42:02.400 He was not a professor, he was a Wall Street analyst. And the 42:02.400 --> 42:05.470 story about him is that he believed that you look at the 42:05.470 --> 42:07.900 prices and you can predict prices. 42:07.900 --> 42:10.610 And in fact, he said, I don't want to look at anything else. 42:10.610 --> 42:12.270 I just want to see prices. 42:12.270 --> 42:14.840 I'll do plots, and I can -- 42:14.840 --> 42:18.750 using my judgment, I can figure out what 42:18.750 --> 42:19.590 it's going to do. 42:19.590 --> 42:22.510 And so, the story about McGee is, everyone on Wall Street 42:22.510 --> 42:26.220 wants the corner office overlooking the World Trade 42:26.220 --> 42:28.670 Center, whatever. 42:28.670 --> 42:31.830 He said, I wanted an interior office with no windows. 42:31.830 --> 42:34.360 I don't want any distractions, I don't want the real world 42:34.360 --> 42:36.110 impinging in my judgment. 42:36.110 --> 42:39.250 And so, that's McGee. 42:39.250 --> 42:40.680 But he said that there are certain things 42:40.680 --> 42:42.280 that you see obviously. 42:42.280 --> 42:44.490 For example, resistance level. 42:50.850 --> 42:54.700 When the Dow Jones Industrial Average approached 1,000 -- 42:54.700 --> 42:57.620 I think it was in the 1960s -- 42:57.620 --> 42:59.850 it's way above that now, as you know. 42:59.850 --> 43:04.100 But when it first approached 1,000, technical analysts 43:04.100 --> 43:06.350 said, you know maybe it's going to have trouble crossing 43:06.350 --> 43:09.960 1,000, because that's a psychological barrier. 43:09.960 --> 43:13.160 That's sounds magical, how can the Dow be worth over 1,000? 43:13.160 --> 43:15.210 Wow, I'm going to sell. 43:15.210 --> 43:17.560 And so, the idea was that people would sell when it 43:17.560 --> 43:18.590 approached 1,000. 43:18.590 --> 43:21.350 And the technical analysts seemed to be right, because 43:21.350 --> 43:25.270 the Dow bounced around just below 1,000 for a long time. 43:25.270 --> 43:28.520 I guess it was months or a year, like it couldn't cross 43:28.520 --> 43:29.770 the resistance level. 43:32.320 --> 43:33.240 That's one example. 43:33.240 --> 43:34.770 I have another example, which is from 43:34.770 --> 43:36.140 Edwards and McGee's book. 43:38.770 --> 44:02.240 [SIDE CONVERSATION] 44:02.240 --> 44:04.340 PROFESSOR ROBERT SHILLER: That's Edwards and McGee, this 44:04.340 --> 44:06.440 is one of the patterns. 44:06.440 --> 44:13.260 McGee was actually a student of psychology, and he thought 44:13.260 --> 44:15.750 certain kinds of patterns seemed to have 44:15.750 --> 44:18.280 really spooked people. 44:18.280 --> 44:20.630 And this is one pattern, which he called ''Head and 44:20.630 --> 44:25.740 Shoulders.'' That's the head, that's one shoulder, that's 44:25.740 --> 44:26.950 the other shoulder. 44:26.950 --> 44:30.480 He said, when you see this pattern, watch out. 44:30.480 --> 44:34.000 It's going to, actually, totally collapse, as it is 44:34.000 --> 44:35.250 shown doing. 44:37.890 --> 44:41.830 These are stock prices plotted against time. 44:41.830 --> 44:45.800 These are days, each of these points is a day. 44:45.800 --> 44:49.120 And this is from their book, so it's hypothetical. 44:49.120 --> 44:51.670 You hardly ever see such perfect Head 44:51.670 --> 44:54.680 and Shoulders patterns. 44:54.680 --> 45:03.060 And so, maybe that's Edwards and McGee's most famous. 45:03.055 --> 45:04.305 Head and Shoulders. 45:06.830 --> 45:08.520 So, the question is, does it work? 45:08.520 --> 45:09.770 Does it really work? 45:14.540 --> 45:18.020 In the early 1970s, when the efficient markets hypothesis 45:18.020 --> 45:22.770 was really strong, Burton Malkiel, who was a professor 45:22.765 --> 45:25.955 at Princeton, and then later he was the Dean of the Yale 45:25.956 --> 45:29.386 School of Management, wrote a book called A Random Walk Down 45:29.390 --> 45:31.920 Wall Street, which claimed that technical 45:31.920 --> 45:33.910 analysis was bunk. 45:33.910 --> 45:37.170 And he said many studies have shown that it doesn't work. 45:37.170 --> 45:39.260 This Head and Shoulders doesn't work. 45:39.260 --> 45:40.830 None of Edwards and McGee's stuff worked. 45:45.230 --> 45:47.330 There were lots of studies, and I actually met him at a 45:47.330 --> 45:49.810 cocktail party after his book came out. 45:49.810 --> 45:52.600 And I said, you didn't footnote all those studies 45:52.600 --> 45:55.540 about technical analysis. 45:55.540 --> 45:56.250 Where are they? 45:56.250 --> 45:57.180 I can't find them. 45:57.180 --> 45:58.360 I did a search. 45:58.360 --> 46:00.840 Not on the internet, I did it on something else, but I was 46:00.840 --> 46:03.360 able to search. 46:03.360 --> 46:05.670 I couldn't find them, where are they? 46:05.670 --> 46:10.630 And I found that he didn't have an immediate answer. 46:10.630 --> 46:13.670 I suspect that he was extrapolating -- there was a 46:13.670 --> 46:17.440 literature on testing market efficiency. 46:17.440 --> 46:19.400 They looked for things like momentum, 46:19.400 --> 46:20.790 whether that continued. 46:20.790 --> 46:24.390 But there was something a little bit wrong with the 46:24.390 --> 46:26.810 literature. 46:26.810 --> 46:30.830 Not many people really confronted technical analysis. 46:30.830 --> 46:34.410 Later, there were people who did look at some of Edwards 46:34.410 --> 46:37.700 and McGee's points, and they found some 46:37.700 --> 46:39.600 element of truth to them. 46:39.600 --> 46:45.160 So, I think the answer is, McGee wasn't a total idiot, as 46:45.160 --> 46:48.760 you might infer from the Efficient Markets Theory. 46:48.760 --> 46:52.090 But it's not going to make you rich, either. 46:52.090 --> 46:56.540 If anything, technical analysis is a subtle art that 46:56.540 --> 46:58.870 can augment trading strategies. 46:58.870 --> 47:01.080 I bet David Swensen doesn't do it at all. 47:01.080 --> 47:04.500 I could have asked, I don't know for sure. 47:04.500 --> 47:09.010 Let me talk about Random Walk, which is a 47:09.010 --> 47:16.310 central idea in finance. 47:16.310 --> 47:28.430 The idea is that, if stock prices are really efficient, 47:28.430 --> 47:31.000 then any change from day to day has to 47:31.000 --> 47:33.450 be due only to news. 47:33.450 --> 47:37.300 And news is essentially unforecastable. 47:37.300 --> 47:40.760 Therefore, stock prices have to do a random 47:40.760 --> 47:42.010 walk through time. 47:44.710 --> 47:47.490 That means that any future movement in them is always 47:47.490 --> 47:48.740 unpredictable. 47:51.600 --> 47:54.830 The changes are totally random. 47:54.830 --> 48:01.700 So, the term Random Walk is an important term. 48:01.700 --> 48:05.500 It was coined not by a finance theorist, but by a 48:05.500 --> 48:10.700 statistician, Karl Pearson, writing in the scientific 48:10.700 --> 48:18.390 journal Nature in 1905. 48:18.390 --> 48:20.410 Now, he didn't link it to finance. 48:20.410 --> 48:25.310 But what he said is, the movements in some -- well, he 48:25.310 --> 48:26.780 was thinking theoretically. 48:26.780 --> 48:30.740 Actually, I believe he used the example of a drunk. 48:30.740 --> 48:35.090 Let's take someone who is so drunk that 48:35.090 --> 48:37.160 each step is random. 48:37.160 --> 48:42.230 This person has no direction at all, staggering randomly. 48:42.230 --> 48:44.830 So, he starts out at a lamp pole. 48:44.830 --> 48:48.120 And what would you predict -- this is what Pearson asked -- 48:48.120 --> 48:51.450 what would you predict is his position in 10 minutes? 48:51.445 --> 48:54.335 He happens to be at a lamp pole right now. 48:57.230 --> 49:02.000 And what Pearson said is, well, your best forecast is 49:02.000 --> 49:04.750 that he's right where he is now. 49:04.750 --> 49:06.030 Because you have no bias. 49:06.030 --> 49:08.070 He could go in any direction, equally likely. 49:08.070 --> 49:09.510 So, what's most likely? 49:09.510 --> 49:12.230 It's that he stays right where he is. 49:12.230 --> 49:15.130 And what is the probability distribution? 49:15.130 --> 49:18.210 Well, it turns out that the standard deviation around that 49:18.210 --> 49:21.480 point goes up with the square root of n steps. 49:21.480 --> 49:23.130 Because each step is independent of the other, so 49:23.125 --> 49:25.575 the square root rule applies. 49:25.580 --> 49:28.850 So, if you're asked to forecast his position after an 49:28.850 --> 49:32.640 hour -- that's a lot of steps -- you would say, I predict 49:32.640 --> 49:35.110 he's right where he is now, but I now have a big standard 49:35.110 --> 49:37.470 deviation around it. 49:37.470 --> 49:41.160 Pearson's article is a very simple idea. 49:41.160 --> 49:45.650 Among the readers, apparently, was Albert Einstein and 49:45.650 --> 49:48.360 Norbert Wiener, the mathematician who had invented 49:48.360 --> 49:50.470 a continuous version of the Random Walk, 49:50.470 --> 49:52.790 called the Wiener Process. 49:52.790 --> 49:55.580 But it got into finance later. 49:55.580 --> 49:59.460 And in the Efficient Markets revolution, they started to 49:59.460 --> 50:10.630 realize that stock prices look more like more like a Random 50:10.630 --> 50:13.310 Walk than a Head and Shoulders. 50:13.310 --> 50:16.220 You look at these Head and Shoulders patterns and they're 50:16.220 --> 50:18.890 hard to find. 50:18.890 --> 50:23.270 So, what is a Random Walk? 50:23.270 --> 50:24.260 Let me just define it. 50:24.260 --> 50:29.700 A Random Walk is where you have a series x sub t equals x 50:29.700 --> 50:34.720 sub t minus 1 plus epsilon sub t, where 50:34.720 --> 50:37.250 epsilon sub t is noise. 50:37.250 --> 50:40.440 Just unforecastable noise: mean 0 and 50:40.440 --> 50:41.690 some standard deviation. 50:44.140 --> 50:47.160 Ideally, it would be normally distributed, so it would have 50:47.160 --> 50:49.680 a bell-shaped curve, and then the math would be very easy 50:49.680 --> 50:50.930 and very simple. 50:54.290 --> 50:58.620 So, I want I want to contrast that with an alternative, 50:58.620 --> 51:01.670 which is called a ''First-Order 51:01.670 --> 51:03.900 Autoregressive.'' Let me get my notation here. 51:08.010 --> 51:12.110 Let's take a process that starts at 100 -- that's like 51:12.110 --> 51:17.590 the lamp post. We'll say x sub t equals -- how am I 51:17.590 --> 51:18.840 putting this -- 51:21.340 --> 51:29.920 100 plus some number rho -- that's a rho -- times x sub t 51:29.916 --> 51:36.656 minus 1 minus 100 plus epsilon sub t. 51:36.660 --> 51:40.630 So, this is an AR-1. 51:40.630 --> 51:44.570 That's First-Order Autoregressive. 51:44.570 --> 51:51.950 It's like a regression model where 100 times 1 minus rho is 51:51.947 --> 51:53.107 the constant term. 51:53.110 --> 51:58.110 And the coefficient of the lagged x is rho. 51:58.110 --> 52:07.770 And we usually require that rho is between minus 1 and 1. 52:07.770 --> 52:09.940 Normally, rho is positive. 52:09.940 --> 52:18.640 So, that means that it's mean-reverting, but slowly. 52:21.580 --> 52:26.160 First of all, in a special case where rho equals 1, I 52:26.160 --> 52:29.760 wouldn't call it an AR-1 anymore, because it reduces to 52:29.755 --> 52:30.745 a Random Walk, right? 52:30.750 --> 52:35.080 If you make rho 1, then the constant term drops out. 52:35.080 --> 52:38.270 I've got 100 minus 100 -- 52:38.270 --> 52:41.810 there's no constant term -- and then I've got x sub t 52:41.810 --> 52:44.020 equals x sub t minus 1 plus epsilon sub t, 52:44.020 --> 52:46.230 that's a Random Walk. 52:46.230 --> 52:50.870 So, in the extreme case where rho gets to one, then a 52:50.870 --> 52:56.560 First-Order Autoregressive process 52:56.560 --> 52:59.350 converges to a Random Walk. 52:59.350 --> 53:05.220 I wanted to show you some simulations of it. 53:07.760 --> 53:25.520 [SIDE CONVERSATION] 53:25.515 --> 53:30.965 PROFESSOR ROBERT SHILLER: OK, this here is a plot I had. 53:30.970 --> 53:34.010 Let's first look at the black line. 53:34.010 --> 53:39.540 The black line is the Standard & Poor's composite stock price 53:39.540 --> 53:50.700 index in real terms. I have that from 1871 until recently. 53:50.700 --> 53:53.790 That's just there for comparison, that is the actual 53:53.790 --> 53:56.980 stock market. 53:56.980 --> 54:03.070 The pink line is a Random Walk that I generated using this 54:03.070 --> 54:03.280 formula [correction: The pink line is a Random Walk with a 54:03.279 --> 54:05.629 time trend, to be explained below.], and a random number 54:05.630 --> 54:12.310 generator that generates random normal variables. 54:16.060 --> 54:23.000 I started them out at the same level, but don't they look 54:23.000 --> 54:24.250 kind of similar? 54:26.280 --> 54:31.580 If you look at the stock market without comparing it 54:31.580 --> 54:35.680 with a Random Walk, it looks like it has patterns in it. 54:35.675 --> 54:39.845 In fact, here's a Head and Shoulders, right here. 54:39.850 --> 54:42.380 Bang, bang, bang. 54:42.380 --> 54:42.790 When is that? 54:42.790 --> 54:47.730 I think this is 1937. 54:47.730 --> 54:48.020 This is -- 54:48.024 --> 54:49.274 I am not sure -- 54:52.050 --> 54:54.910 1930, 1931? 54:54.910 --> 54:57.470 And this is just before the war. 54:57.470 --> 54:59.600 I'm not sure exactly. 54:59.600 --> 55:00.560 It's a nice Head and Shoulders. 55:00.560 --> 55:03.400 Hey, Edwards and McGee are sort of right, right? 55:03.400 --> 55:05.560 It dropped a lot after that. 55:08.940 --> 55:11.700 You know, I can find Head and Shoulders up here, too, right? 55:11.700 --> 55:12.950 Maybe. 55:14.660 --> 55:17.240 The black line is actual U.S. history, 55:17.240 --> 55:18.950 that's the stock market. 55:18.950 --> 55:28.040 The pink line is a fake stock line generated with pure 55:28.040 --> 55:28.980 random noise. 55:28.980 --> 55:31.450 And the fact that it's going up is just chance. 55:31.450 --> 55:33.740 I can actually use this program to 55:33.740 --> 55:34.990 generate other examples. 55:37.830 --> 55:39.560 This should change, let me see, make sure 55:39.560 --> 55:41.060 it's working here. 55:41.060 --> 55:42.760 The black line is the same. 55:42.760 --> 55:44.340 I'm not going to change the black line, the 55:44.340 --> 55:45.600 black line is history. 55:45.600 --> 55:49.880 I just did a brand new Random Walk calculation using my 55:49.880 --> 55:55.960 random number generator, which is there on Excel. 55:55.960 --> 55:58.710 That looks pretty good, doesn't it? 55:58.710 --> 56:01.670 I'm going to do more for you. 56:01.670 --> 56:03.600 Which one is the real stock market? 56:03.600 --> 56:05.230 I find that hard to tell, right? 56:09.060 --> 56:12.660 The insight is that people get deceived when they look at 56:12.660 --> 56:14.110 stock price charts. 56:14.110 --> 56:16.050 They think they see patterns. 56:16.050 --> 56:19.840 That pink line is guaranteed to have no patterns, because I 56:19.840 --> 56:22.750 generated it, so that there are no patterns, 56:22.750 --> 56:24.390 except random patterns. 56:24.390 --> 56:27.140 But when I look at this pink line, which just came up, look 56:27.135 --> 56:28.305 at that up-trend. 56:28.310 --> 56:29.300 Wow. 56:29.300 --> 56:31.860 That's called a bull market, when it goes up. 56:31.860 --> 56:34.520 And I can make all kinds of theories about why that's 56:34.520 --> 56:36.970 happening, but you know those would be fake theories, 56:36.970 --> 56:38.600 because I know what's really happening. 56:38.600 --> 56:40.970 This is pure randomness. 56:40.970 --> 56:44.290 I'll do it again, I can do this forever. 56:44.290 --> 56:45.670 That's another one. 56:45.670 --> 56:48.930 Here the trend wasn't quite so positive. 56:48.930 --> 56:51.550 Oh, actually, I have to say one thing, I forgot. 56:51.550 --> 56:53.410 I did put an up-trend in the Random Walk. 56:53.413 --> 56:54.283 I am sorry. 56:54.280 --> 57:02.100 In this simulation, I did add a constant, so that 57:02.100 --> 57:03.200 it pushed it up. 57:03.200 --> 57:05.130 Otherwise, it was a Random Walk. 57:05.130 --> 57:06.480 I forgot what I did. 57:06.480 --> 57:08.840 It was a Random Walk with trend. 57:08.840 --> 57:10.860 But it doesn't guarantee that it will go up, 57:10.860 --> 57:12.210 because it's random. 57:12.210 --> 57:13.460 I'll do another simulation. 57:16.620 --> 57:17.960 See, how fast I can do these? 57:17.960 --> 57:19.410 It's the wonder of modern computers. 57:22.260 --> 57:25.760 If this were our history, people would say, the amazing 57:25.760 --> 57:28.390 stock market of the first half -- look at that stock market 57:28.390 --> 57:30.320 in the first half of the 20th century! 57:30.320 --> 57:33.850 We would be devising all kinds of theories to explain it, but 57:33.850 --> 57:36.030 in fact it's just nonsense. 57:36.030 --> 57:37.730 It's just randomness. 57:37.730 --> 57:40.760 I'll do a few more. 57:40.760 --> 57:41.730 Look at that. 57:41.730 --> 57:43.280 Boy, that would be a hump shape for 57:43.280 --> 57:45.640 the whole 20th century. 57:45.640 --> 57:48.410 We'd have historians trying to figure that one out. 57:48.410 --> 57:49.660 Oh, this one downturned. 57:49.660 --> 57:52.430 This is a bad outcome. 57:52.430 --> 57:55.420 Jeremy Siegel would not be pleased with this outcome, 57:55.420 --> 57:59.200 because it has the stock market gaining 57:59.200 --> 58:01.880 nothing in 100 years. 58:01.880 --> 58:06.220 These are all equally likely outcomes. 58:06.220 --> 58:08.820 Look at that one. 58:08.820 --> 58:12.330 If that was the world that we inherited, we would really 58:12.330 --> 58:15.740 think there was a linear trend in the market, right? 58:15.740 --> 58:19.240 People would be making models, saying -- 58:19.240 --> 58:23.650 look at how straight that line is. 58:23.650 --> 58:26.880 The point is that you start to see a sort of reality which is 58:26.880 --> 58:29.420 really just randomness. 58:29.420 --> 58:33.660 That's why Nassim Taleb, a friend of mine, wrote a book 58:33.660 --> 58:34.930 called Fooled by Randomness. 58:34.930 --> 58:36.520 I thought that was a great title for a 58:36.520 --> 58:38.110 book and a great book. 58:38.110 --> 58:42.780 People don't understand how things are just purely random, 58:42.780 --> 58:45.390 and your mind tries to make sense out of them. 58:45.390 --> 58:48.000 And you start looking at patterns and the patterns 58:48.000 --> 58:49.840 don't mean anything. 58:49.840 --> 58:54.660 So, I can just keep doing this, but maybe I'll stop. 58:54.660 --> 58:55.910 Look at that one. 58:58.590 --> 59:01.130 Of course, I'm helped along by the trend that I added in. 59:01.130 --> 59:04.090 So, I want to see if I can get a down-trend one. 59:04.090 --> 59:07.420 Because I put a trend in, it's hard to get a real down-trend. 59:07.420 --> 59:09.550 That's sort of a down-trend. 59:09.550 --> 59:11.260 That's where there were a lot of negative shocks. 59:14.280 --> 59:16.890 So, you see the comparison of the Random Walk with the 59:16.890 --> 59:18.450 actual stock market. 59:18.450 --> 59:20.650 So, the actual stock market looks a lot 59:20.650 --> 59:22.060 like a random walk. 59:22.060 --> 59:24.010 One thing is different, though, you don't -- look at 59:24.007 --> 59:24.837 this pattern, here. 59:24.835 --> 59:29.355 In 1929, and this is the crash, after '29. 59:29.360 --> 59:31.980 You know, I'm not getting that in any of my simulations, and 59:31.980 --> 59:33.840 you know why I'm not? 59:33.840 --> 59:37.540 Because I chose a normally distributed shock. 59:37.540 --> 59:41.590 No fat tails, I didn't put fat tails into my simulation. 59:41.590 --> 59:43.640 And you didn't notice that, right? 59:43.640 --> 59:48.830 But in this simulation, we never see such sudden drops. 59:48.830 --> 59:51.890 So, there's something that, if you spend time searching on 59:51.890 --> 59:53.720 it, you might see something not quite right. 59:53.720 --> 59:57.340 But basically, the Random Walk looks a lot like the actual 59:57.340 --> 59:58.980 stock market. 59:58.980 --> 1:00:03.960 I'm trying to get one that really matches up, but I'm not 1:00:03.960 --> 1:00:05.930 quite succeeding. 1:00:05.930 --> 1:00:09.550 That's pretty good, isn't it? 1:00:09.550 --> 1:00:13.640 In this simulation, 1929 wasn't quite as strong and the 1:00:13.640 --> 1:00:16.030 Depression wasn't as bad. 1:00:16.030 --> 1:00:18.460 Anyway, what I want to do now is compare the Random Walk 1:00:18.460 --> 1:00:19.710 with the AR-1. 1:00:26.030 --> 1:00:27.860 Now I'm going to do the same thing, but I'm going 1:00:27.860 --> 1:00:29.110 to do it with -- 1:00:31.690 --> 1:01:13.520 [SIDE CONVERSATION] 1:01:13.520 --> 1:01:15.900 PROFESSOR ROBERT SHILLER: I know what I did, sorry. 1:01:15.900 --> 1:01:20.220 The pink line is now an AR-1 process, 1:01:20.220 --> 1:01:22.680 which is this process. 1:01:22.680 --> 1:01:26.000 So, it's mean reverting now. 1:01:26.000 --> 1:01:29.770 I'm not comparing the Random Walk with the AR-1, I'm just 1:01:29.770 --> 1:01:33.030 doing the same thing now with an AR-1. 1:01:33.030 --> 1:01:37.210 Now, the thing about AR-1 is, you realize that it wants to 1:01:37.210 --> 1:01:40.080 come back to 100. 1:01:40.080 --> 1:01:43.550 I put a trend in. 1:01:43.550 --> 1:01:47.350 So, it's actually coming back to a linear up-trend. 1:01:50.950 --> 1:01:55.120 What I did is I put in a time trend as well. 1:01:58.420 --> 1:02:02.830 But the point is that it tends to hug the trend somewhat. 1:02:06.930 --> 1:02:08.450 I have it here shown not around a 1:02:08.450 --> 1:02:10.230 trend, but around 100. 1:02:10.230 --> 1:02:12.530 What an AR-1 does is, say rho is 1/2. 1:02:15.210 --> 1:02:20.650 If rho is 1/2, then it means that if x sub t minus 1 was 1:02:20.650 --> 1:02:21.510 above 100 -- 1:02:21.510 --> 1:02:23.380 last period it was above 100 -- 1:02:23.380 --> 1:02:26.790 it will be above 100 this time, but only half as much 1:02:26.790 --> 1:02:30.570 above 100, so it's going back to 100. 1:02:30.570 --> 1:02:34.000 And then the next time, it'll only be half, again, as much 1:02:34.000 --> 1:02:37.200 above 100 as it was the last time. 1:02:37.200 --> 1:02:40.390 So, it tends to go back to a trend. 1:02:40.390 --> 1:02:44.700 But what I've shown here is a simulation with a random 1:02:44.700 --> 1:02:52.800 number generator of an AR-1 around a trend, where the 1:02:52.800 --> 1:02:55.250 trend matches the actual trend in the stock market. 1:02:58.320 --> 1:03:02.070 Now, in this case, this is not Random Walk. 1:03:02.070 --> 1:03:05.660 And there is a profit opportunity, and the profit 1:03:05.660 --> 1:03:09.350 opportunity is when it's below trend, buy, when it's above 1:03:09.350 --> 1:03:10.760 trend, sell. 1:03:10.760 --> 1:03:12.550 Because it will tend to come back to trend. 1:03:15.190 --> 1:03:19.170 I chose a rho which was very small, something like 1/2, so 1:03:19.165 --> 1:03:22.455 it tends to come rapidly back to trend. 1:03:22.460 --> 1:03:24.830 This does look different than the actual stock market, 1:03:24.830 --> 1:03:26.570 doesn't it? 1:03:26.570 --> 1:03:29.890 You see how much it hugs the trend? 1:03:29.890 --> 1:03:33.630 So, it doesn't seem to fit as well. 1:03:33.630 --> 1:03:35.610 I can do simulations of this, too. 1:03:38.940 --> 1:03:41.760 This is different. 1:03:41.760 --> 1:03:43.380 You can see the difference, right? 1:03:43.380 --> 1:03:47.650 This pink line doesn't look as much like the actual stock 1:03:47.650 --> 1:03:52.620 market, because it really wants this trend. 1:03:52.620 --> 1:03:55.220 We saw trendy ones occasionally, by chance with a 1:03:55.220 --> 1:03:59.900 Random Walk, but here we're seeing it's always on a trend. 1:03:59.900 --> 1:04:05.880 And so, in this world, if the stock market were an AR-1, 1:04:05.880 --> 1:04:07.440 there would be a profitable strategy. 1:04:07.440 --> 1:04:10.090 Always buy when it's below trend, and sell when it's 1:04:10.090 --> 1:04:10.990 above trend. 1:04:10.990 --> 1:04:13.260 Because you know it'll come back, you see how reliable 1:04:13.260 --> 1:04:15.360 this comes back to a trend? 1:04:19.310 --> 1:04:21.430 So, you can see that there's a fundamental difference, the 1:04:21.430 --> 1:04:27.310 Random Walk seems to fit the data better than the AR-1. 1:04:27.310 --> 1:04:30.540 The Random Walk theory says that stock prices 1:04:30.540 --> 1:04:33.630 are not mean reverting. 1:04:33.630 --> 1:04:36.590 Where they go from today is all random. 1:04:36.590 --> 1:04:39.070 If they're above the historical trend, it's 1:04:39.070 --> 1:04:39.960 meaningless. 1:04:39.960 --> 1:04:41.970 The historical trend is just nonsense. 1:04:41.970 --> 1:04:43.230 It's just random. 1:04:43.230 --> 1:04:46.450 And forget trends, forget anything. 1:04:46.450 --> 1:04:49.020 It's always the drunk at the lamp post, no matter where you 1:04:49.020 --> 1:04:51.410 are in history. 1:04:51.410 --> 1:04:54.970 This one, if rho is substantially less than 1, it 1:04:54.970 --> 1:04:57.920 looks a lot different doesn't it? 1:04:57.920 --> 1:05:01.550 With rho equals 1/2, this is not the world we live in. 1:05:01.550 --> 1:05:03.690 It would be too easy to make money. 1:05:03.690 --> 1:05:06.420 All these little oscillations around the trend I could 1:05:06.420 --> 1:05:07.410 profit from. 1:05:07.410 --> 1:05:09.810 But in the real world, it's not like that. 1:05:09.810 --> 1:05:19.000 But suppose the real world is AR-1 with rho equal to 0.99. 1:05:19.000 --> 1:05:21.070 What about that? 1:05:21.070 --> 1:05:24.070 Well that's not much different from a Random Walk, is it? 1:05:24.070 --> 1:05:25.720 It's going to be hard to tell the difference. 1:05:29.340 --> 1:05:33.050 In the Efficient Markets Theory period, people were 1:05:33.050 --> 1:05:36.100 really excited about the Random Walk Hypothesis. 1:05:36.100 --> 1:05:40.160 That's why Burton Malkiel's book, A Random Walk Down Wall 1:05:40.156 --> 1:05:41.746 Street, which he came out with -- 1:05:41.750 --> 1:05:44.870 I think it was in 1973, right after Fama. 1:05:44.870 --> 1:05:50.120 It became a huge best seller, it sold over a million copies, 1:05:50.120 --> 1:05:53.810 because at that time people were thinking, this is 1:05:53.810 --> 1:05:56.160 exciting new wisdom. 1:05:56.160 --> 1:05:59.130 We've learned that the stock market is a Random Walk. 1:05:59.130 --> 1:06:01.790 And there's all kinds of implications for that. 1:06:01.790 --> 1:06:02.860 The problem is -- 1:06:02.860 --> 1:06:05.170 I have to wrap up -- 1:06:05.170 --> 1:06:07.790 that the Random Walk Hypothesis 1:06:07.790 --> 1:06:10.340 wasn't exactly right. 1:06:10.340 --> 1:06:13.380 It's sort of right, you've gotten some insights. 1:06:13.380 --> 1:06:16.840 But you know, maybe the real world is AR-1 with a 1:06:16.840 --> 1:06:18.750 rho close to 1. 1:06:18.750 --> 1:06:22.590 And in that world, there are profit opportunities, but they 1:06:22.590 --> 1:06:24.760 take a long time to come. 1:06:24.760 --> 1:06:29.050 So, if the real world is AR-1 with rho equal to 0.99 or 1:06:29.050 --> 1:06:34.570 0.98, that means you can buy stocks when 1:06:34.570 --> 1:06:35.950 they're below trend. 1:06:35.950 --> 1:06:38.680 But then you have to wait 10, 20 years for them 1:06:38.680 --> 1:06:40.140 to get back to trend. 1:06:40.140 --> 1:06:44.350 So, it's like the drunk on the lamp pole. 1:06:44.350 --> 1:06:47.580 The drunk is standing next to a lamp pole, it 1:06:47.580 --> 1:06:48.530 was a random walk. 1:06:48.530 --> 1:06:51.520 But now we put in elastic band around the drunk's ankle and 1:06:51.520 --> 1:06:55.190 tie it to the lamp pole and it pulls him back. 1:06:55.190 --> 1:06:58.440 Now, if we have a very loose elastic, this guy can wander 1:06:58.440 --> 1:07:01.980 for a long time, but will eventually be pulled back. 1:07:01.980 --> 1:07:04.440 You'd never know when. 1:07:04.440 --> 1:07:08.590 But if you have a tight elastic, then it would be 1:07:08.590 --> 1:07:11.170 obvious that the drunk is coming back. 1:07:11.170 --> 1:07:14.430 The problem is that the real world seems, maybe, to be more 1:07:14.430 --> 1:07:16.900 like the drunk with the loose elastic. 1:07:16.900 --> 1:07:18.590 And so, it's kind of unsatisfying. 1:07:18.590 --> 1:07:23.610 You can beat the market, but simple trading rules like 1:07:23.610 --> 1:07:28.810 Edwards and McGee are not powerful, short-run profit 1:07:28.810 --> 1:07:29.610 opportunities. 1:07:29.610 --> 1:07:32.630 In that sense, the Efficient Markets Hypothesis is right. 1:07:32.630 --> 1:07:34.800 So, don't forget the Efficient Markets Hypothesis. 1:07:34.800 --> 1:07:37.130 I'll repeat what I said at the beginning. 1:07:37.130 --> 1:07:40.530 It's a half-truth, it's half true. 1:07:40.530 --> 1:07:43.620 Remember that, but don't put too much faith in it, either.