1
00:00:02,720 --> 00:00:07,200
Speaker 1: Bloomberg Audio Studios, podcasts, radio news.

2
00:00:08,160 --> 00:00:09,680
Speaker 2: I was hoping to be like one of those like

3
00:00:09,680 --> 00:00:11,160
Speaker 2: clips on TikTok you see.

4
00:00:11,320 --> 00:00:13,280
Speaker 1: Fake that's the thing you look like you're on a

5
00:00:13,280 --> 00:00:14,760
Speaker 1: podcast pads on home.

6
00:00:14,880 --> 00:00:16,680
Speaker 2: Now it just looks like, you know, instead.

7
00:00:16,400 --> 00:00:20,200
Speaker 3: Of we just randomly got together the chat and microphones

8
00:00:20,680 --> 00:00:23,720
Speaker 3: part of it with Bloomberg Podcasts behind us.

9
00:00:24,079 --> 00:00:26,960
Speaker 1: Yeah. Right, I guess we've conveyed podcasts efficiently with all

10
00:00:27,000 --> 00:00:28,520
Speaker 1: of the Bloomberg podcasts.

11
00:00:28,600 --> 00:00:28,800
Speaker 2: Joe.

12
00:00:29,040 --> 00:00:30,680
Speaker 4: Part of the reason this has to be on video

13
00:00:31,000 --> 00:00:33,839
Speaker 4: is because Matt shaved. Matt has had a beard for

14
00:00:33,960 --> 00:00:35,000
Speaker 4: the past I don't know that.

15
00:00:35,120 --> 00:00:39,960
Speaker 1: I've had a winter beard from like Christmas break through memorild.

16
00:00:39,600 --> 00:00:41,960
Speaker 3: That I shaved over COVID for the first time in

17
00:00:41,960 --> 00:00:44,760
Speaker 3: about thirty years and my kids freaked out.

18
00:00:45,080 --> 00:00:48,400
Speaker 1: Yeah, they were like, you're an alien. Yeah, my kids

19
00:00:48,440 --> 00:00:50,400
Speaker 1: didn't care that much. But one of my sons said,

20
00:00:50,760 --> 00:00:51,800
Speaker 1: it's a different daddy.

21
00:00:52,120 --> 00:00:55,280
Speaker 3: Well, you also have hair on the top of your head.

22
00:00:55,760 --> 00:00:57,880
Speaker 3: When you shave the beard and you don't have any hair,

23
00:00:58,120 --> 00:00:59,320
Speaker 3: Suddenly you're mister clean.

24
00:00:59,480 --> 00:01:01,000
Speaker 1: You just does mister clean?

25
00:01:02,000 --> 00:01:05,639
Speaker 2: No, I don't think so, mister clean. You know clean?

26
00:01:06,400 --> 00:01:09,600
Speaker 3: Come on, I think that's a fair point.

27
00:01:13,280 --> 00:01:16,520
Speaker 1: It's the Money Stuff podcast. We have a guest, Cliff

28
00:01:16,560 --> 00:01:21,520
Speaker 1: has runs AQR. Thanks for coming in, Thanks for having me.

29
00:01:22,200 --> 00:01:23,920
Speaker 1: I always like to ask how to my managers, like

30
00:01:25,200 --> 00:01:26,760
Speaker 1: what do you do for a living? Like what is

31
00:01:26,760 --> 00:01:30,199
Speaker 1: it like economic function of like the business that you run?

32
00:01:30,600 --> 00:01:34,080
Speaker 3: Okay, those are to me slightly different questions. Right.

33
00:01:34,080 --> 00:01:35,680
Speaker 1: One sounds like it's about you. One sounds like it's

34
00:01:35,680 --> 00:01:36,240
Speaker 1: about AQR.

35
00:01:36,280 --> 00:01:39,520
Speaker 3: I'm more interested in even if it's about a q R.

36
00:01:40,160 --> 00:01:43,600
Speaker 3: What you do for a living is And people might

37
00:01:43,600 --> 00:01:47,200
Speaker 3: not like this phraseology, but you're trying to predict what

38
00:01:47,240 --> 00:01:50,000
Speaker 3: happens to securities. You're trying to buy ones that go

39
00:01:50,240 --> 00:01:52,840
Speaker 3: up either in the absolute or more than some benchmark,

40
00:01:52,880 --> 00:01:57,320
Speaker 3: and and sell ones that do the opposite. The broader question,

41
00:01:57,320 --> 00:02:00,800
Speaker 3: which I think is behind what you're saying, is what

42
00:02:00,840 --> 00:02:03,360
Speaker 3: do you do for the world by doing that? And

43
00:02:03,440 --> 00:02:06,520
Speaker 3: they overlap, but they're not exactly the same thing. Something

44
00:02:06,520 --> 00:02:08,640
Speaker 3: can have a net positive effect on the world even

45
00:02:08,639 --> 00:02:11,440
Speaker 3: if you're not waking up. And I'll admit this, I

46
00:02:11,480 --> 00:02:13,840
Speaker 3: don't think most people in their jobs waked up and

47
00:02:13,880 --> 00:02:17,360
Speaker 3: always think that way, and certainly not active managers just

48
00:02:17,440 --> 00:02:19,840
Speaker 3: go I am just making the world a better place.

49
00:02:19,880 --> 00:02:23,720
Speaker 3: They're thinking, is in Nvidia undervalued? Is it overvalued? The

50
00:02:23,800 --> 00:02:26,440
Speaker 3: things that I think you do for the world is

51
00:02:27,000 --> 00:02:32,560
Speaker 3: first take the other side of positions other people disagree

52
00:02:32,600 --> 00:02:35,399
Speaker 3: with you on or don't want to bear. That can

53
00:02:35,440 --> 00:02:39,640
Speaker 3: take two forms. That can mean one of you is

54
00:02:40,080 --> 00:02:43,800
Speaker 3: biased and wrong. You hope it's not you. But in

55
00:02:43,840 --> 00:02:45,520
Speaker 3: that case, what you do for the world is you

56
00:02:45,560 --> 00:02:48,280
Speaker 3: move prices back towards not necessarily all the way too

57
00:02:49,000 --> 00:02:52,400
Speaker 3: in the abstract, the correct price. That's hard to define, actually,

58
00:02:52,480 --> 00:02:54,960
Speaker 3: but something is mispriced and you take the other side

59
00:02:54,960 --> 00:02:57,119
Speaker 3: of that. It could be a risk premium. Other people

60
00:02:57,160 --> 00:03:00,600
Speaker 3: don't want to bear in a lot of strategies necessarily

61
00:03:00,600 --> 00:03:03,400
Speaker 3: that they're making an error. If a merger's announced and

62
00:03:03,639 --> 00:03:05,880
Speaker 3: three quarters of the pop you should get if it

63
00:03:05,919 --> 00:03:08,200
Speaker 3: closes happens, a lot of people might not want to

64
00:03:08,200 --> 00:03:10,919
Speaker 3: stick around for that last quarter. And if you're willing

65
00:03:11,000 --> 00:03:12,400
Speaker 3: to take the other side of that, maybe you get

66
00:03:12,440 --> 00:03:14,639
Speaker 3: paid a little for doing that, and that is a

67
00:03:14,720 --> 00:03:16,680
Speaker 3: service to the market. People want to get out and

68
00:03:16,720 --> 00:03:22,640
Speaker 3: you're helping. That's kind of the positive side. But again

69
00:03:22,760 --> 00:03:24,760
Speaker 3: it's not what you're thinking about when you do the trade.

70
00:03:25,040 --> 00:03:29,600
Speaker 3: There are positions active managers will take that are not

71
00:03:29,720 --> 00:03:31,880
Speaker 3: about that. Let me put it this way. If what

72
00:03:31,919 --> 00:03:35,440
Speaker 3: you're trying to do is predict returns, you can predict

73
00:03:35,440 --> 00:03:40,320
Speaker 3: returns because the price is moving towards truth. But you

74
00:03:40,320 --> 00:03:42,760
Speaker 3: can also make money if you predict the price moves

75
00:03:42,760 --> 00:03:46,440
Speaker 3: further away from truth. You know, if you're a momentum

76
00:03:46,480 --> 00:03:49,440
Speaker 3: memestock investor. And doesn't mean you can't get that right,

77
00:03:49,520 --> 00:03:50,720
Speaker 3: you know. I think of that a little more as

78
00:03:50,760 --> 00:03:54,000
Speaker 3: trading than investing. But they all come together, and it

79
00:03:54,120 --> 00:03:58,160
Speaker 3: even gets complicated within some famous quantitative factors. One famous

80
00:03:58,200 --> 00:03:59,840
Speaker 3: quant factor is the momentum factor.

81
00:04:00,480 --> 00:04:03,520
Speaker 1: Asked the finance professor, I should I have asked? And

82
00:04:03,560 --> 00:04:07,240
Speaker 1: he said, you should ask him if momentum trading makes

83
00:04:07,240 --> 00:04:08,560
Speaker 1: markets more or less efficient.

84
00:04:08,720 --> 00:04:11,000
Speaker 3: We don't fully know, but I can tell you the framework.

85
00:04:11,640 --> 00:04:16,600
Speaker 3: There are two competing explanations in academia and a general

86
00:04:16,760 --> 00:04:19,680
Speaker 3: world that cares about these things. For why momentum on

87
00:04:19,720 --> 00:04:22,760
Speaker 3: average works, there's always a third that it will never

88
00:04:22,800 --> 00:04:24,400
Speaker 3: work again and it was just random and it was

89
00:04:24,480 --> 00:04:26,920
Speaker 3: just luck. But if it's true, why does it work?

90
00:04:27,240 --> 00:04:29,360
Speaker 3: I actually think these can both coexist, so it's not

91
00:04:29,400 --> 00:04:32,960
Speaker 3: truly embarrassing, but it sounds embarrassing that the two major

92
00:04:33,000 --> 00:04:37,960
Speaker 3: explanations one is under reaction and the other is over reaction.

93
00:04:38,360 --> 00:04:40,120
Speaker 3: When you've narrowed it down to two things that at

94
00:04:40,200 --> 00:04:42,479
Speaker 3: least feel like the opposites, you should feel a little shame.

95
00:04:42,520 --> 00:04:45,279
Speaker 1: For a second, there's the inituation, just like there's a

96
00:04:45,279 --> 00:04:48,640
Speaker 1: correct price, and momentum is trading from below to above the.

97
00:04:48,560 --> 00:04:52,000
Speaker 3: Credit friends, Well, I'll give you two scenarios. Information comes

98
00:04:52,040 --> 00:04:56,359
Speaker 3: out and people have a behavioral bias that behavioral psychologists

99
00:04:56,400 --> 00:05:00,320
Speaker 3: would call anchoring an adjustment. They move towards all the

100
00:05:00,360 --> 00:05:02,840
Speaker 3: way to the new information, and I think that fits

101
00:05:02,880 --> 00:05:05,800
Speaker 3: a lot of intuition in the short run that can

102
00:05:05,880 --> 00:05:10,240
Speaker 3: make momentum work. If you're following fundamentals or prices, good

103
00:05:10,240 --> 00:05:13,680
Speaker 3: news comes out or the price moves. If good news

104
00:05:13,680 --> 00:05:17,159
Speaker 3: comes out, on average, the price goes up, but not enough.

105
00:05:17,800 --> 00:05:21,039
Speaker 3: If the price moves, it may on average be responding

106
00:05:21,040 --> 00:05:23,320
Speaker 3: to good news, and simply by observing the price move,

107
00:05:24,120 --> 00:05:27,480
Speaker 3: you can say, Okay, sometimes it's wrong, but on average

108
00:05:27,520 --> 00:05:30,479
Speaker 3: it doesn't quite move enough. If that's the reason, and

109
00:05:30,520 --> 00:05:33,279
Speaker 3: I think the weight of the opinion in academia I

110
00:05:33,320 --> 00:05:37,120
Speaker 3: believe is towards this underreaction explanation. Then, even though you're

111
00:05:37,160 --> 00:05:40,400
Speaker 3: trading on momentum, you're still moving the price towards kind

112
00:05:40,400 --> 00:05:43,640
Speaker 3: of truth or equilibrium or something. But the flip side

113
00:05:43,680 --> 00:05:46,719
Speaker 3: is overreaction. Do you think of that more as just

114
00:05:46,800 --> 00:05:50,840
Speaker 3: your classic positive feedback loop. Someone's buying something just for

115
00:05:51,120 --> 00:05:53,720
Speaker 3: you know, fomo. It's been going up in there, and

116
00:05:54,080 --> 00:05:55,480
Speaker 3: well that could be a negative reason, or they're just

117
00:05:55,560 --> 00:05:59,159
Speaker 3: predicting more people buy it because of fomo. In that sense,

118
00:05:59,360 --> 00:06:02,320
Speaker 3: if you buy some weird meme coin, you could do

119
00:06:02,360 --> 00:06:04,440
Speaker 3: that for a rational reason, not that you are a

120
00:06:04,480 --> 00:06:07,559
Speaker 3: long term holder, but you just believe it's gonna keep going.

121
00:06:07,839 --> 00:06:09,440
Speaker 1: So did you make money on GameStop?

122
00:06:10,680 --> 00:06:12,719
Speaker 3: This the God's honest truth. You won't believe me. I

123
00:06:12,760 --> 00:06:14,400
Speaker 3: have no idea if we were long or short game

124
00:06:14,440 --> 00:06:15,520
Speaker 3: stop during the whole thing.

125
00:06:15,680 --> 00:06:16,920
Speaker 2: You never went back and check.

126
00:06:17,080 --> 00:06:19,520
Speaker 3: No, I never did. I did have an episode where

127
00:06:19,560 --> 00:06:22,520
Speaker 3: I mentioned on perhaps a different TV network that we

128
00:06:22,520 --> 00:06:23,479
Speaker 3: were short, AMC.

129
00:06:23,960 --> 00:06:25,520
Speaker 2: What does your Twitter look like after that?

130
00:06:25,600 --> 00:06:28,080
Speaker 3: It was very ugly. Yeah, I certainly knew of that world,

131
00:06:28,120 --> 00:06:30,880
Speaker 3: even though we're quants. I watch the markets all day,

132
00:06:30,920 --> 00:06:33,279
Speaker 3: even even if I don't do anything about it. It's

133
00:06:33,320 --> 00:06:35,120
Speaker 3: like the old joke about the weather. Everyone talks about it,

134
00:06:35,160 --> 00:06:36,360
Speaker 3: but nobody does anything about it.

135
00:06:36,400 --> 00:06:37,400
Speaker 2: Best thing to talk about.

136
00:06:37,720 --> 00:06:40,640
Speaker 3: So it's not like everything that went on with GameStop

137
00:06:40,720 --> 00:06:44,000
Speaker 3: Melvin Capitol. You know, I'm watching it every day, but

138
00:06:44,279 --> 00:06:47,839
Speaker 3: we take relevant tiny positions in every stock. There was

139
00:06:47,839 --> 00:06:50,160
Speaker 3: nothing weird in our P and L. And yes, I

140
00:06:50,200 --> 00:06:53,160
Speaker 3: was not even curious. It probably wasn't even in our

141
00:06:53,279 --> 00:06:56,000
Speaker 3: universe at that point of things. We trade. But then

142
00:06:56,040 --> 00:07:02,720
Speaker 3: I'm going on this other network. It's allowed, and in

143
00:07:02,960 --> 00:07:04,400
Speaker 3: kind of a pre call of what are we going

144
00:07:04,480 --> 00:07:06,000
Speaker 3: to talk about? You guys know, you don't want to

145
00:07:06,000 --> 00:07:08,159
Speaker 3: get on there and have absolutely nothing to talk about.

146
00:07:08,240 --> 00:07:10,720
Speaker 3: You want to have some not necessarily the answers worked out,

147
00:07:10,720 --> 00:07:13,560
Speaker 3: but agreed upon topics. They're like everyone in this segment

148
00:07:13,640 --> 00:07:16,680
Speaker 3: gives us some lungs and shorts. But I'm saying, as

149
00:07:16,720 --> 00:07:19,880
Speaker 3: a quant that's kind of silly. They're not indicative, and

150
00:07:19,920 --> 00:07:22,880
Speaker 3: we kind of made a deal where they'd let me

151
00:07:23,000 --> 00:07:26,880
Speaker 3: briefly explain that doesn't make a whole lot of sense

152
00:07:26,880 --> 00:07:29,440
Speaker 3: for quantum, but it might be fun. And my way

153
00:07:29,480 --> 00:07:30,880
Speaker 3: of saying it was, if I give you a few

154
00:07:30,960 --> 00:07:33,560
Speaker 3: names long and a few names short, you could look

155
00:07:33,560 --> 00:07:35,640
Speaker 3: in six months later and think we had a fantastic

156
00:07:35,720 --> 00:07:38,200
Speaker 3: year or a terrible year and be terribly wrong in

157
00:07:38,240 --> 00:07:41,440
Speaker 3: either direction because they're tiny. So I went through an

158
00:07:41,480 --> 00:07:45,680
Speaker 3: AMC was I think I'm accidentally doing it again. We'll

159
00:07:45,720 --> 00:07:49,640
Speaker 3: keep going hopefully they don't listen to you, guys. But

160
00:07:51,080 --> 00:07:54,440
Speaker 3: it was bad on every single thing in our model practically,

161
00:07:54,440 --> 00:07:58,080
Speaker 3: which is hard to do. It was expensive, unprofitable, high beta.

162
00:07:58,120 --> 00:08:01,880
Speaker 3: They were issuing shares, not buying backshit. There are more examples,

163
00:08:02,120 --> 00:08:04,240
Speaker 3: and so I said that, but then I added that

164
00:08:04,600 --> 00:08:09,000
Speaker 3: were only short twelve basis points, so the crazy people

165
00:08:09,000 --> 00:08:13,000
Speaker 3: could be right and it doesn't really matter. I discovered

166
00:08:13,000 --> 00:08:15,160
Speaker 3: two things. They're not going to like a short period,

167
00:08:15,480 --> 00:08:19,000
Speaker 3: and crazy people don't always like being called crazy. Yeah,

168
00:08:19,040 --> 00:08:21,239
Speaker 3: I had to discover that from myself. So my Twitter

169
00:08:21,280 --> 00:08:23,600
Speaker 3: got ugly for a while. You may have noticed. I've

170
00:08:23,600 --> 00:08:26,120
Speaker 3: gotten I think a fair amount better at this, but

171
00:08:26,240 --> 00:08:28,760
Speaker 3: I used to be pretty bad about responding to ugly,

172
00:08:29,400 --> 00:08:30,920
Speaker 3: which you learn your lesson on that.

173
00:08:30,920 --> 00:08:33,240
Speaker 2: You always feed the trolls less so than.

174
00:08:33,160 --> 00:08:33,480
Speaker 1: I used to.

175
00:08:33,840 --> 00:08:36,600
Speaker 3: At least I think I'm better. Maybe I'm wrong, but

176
00:08:37,120 --> 00:08:39,880
Speaker 3: I became public enemy number three to the meme stock

177
00:08:39,960 --> 00:08:44,000
Speaker 3: crowd for a while, I did not really. Yes, Kenny

178
00:08:44,040 --> 00:08:46,880
Speaker 3: Ken Griffin was number one. I don't believe Ken did this,

179
00:08:47,000 --> 00:08:50,920
Speaker 3: but it's the whole poll, the Bible thing. And oddly enough,

180
00:08:50,920 --> 00:08:54,040
Speaker 3: Gary Gensler was clearly number two, Yeah, because they thought

181
00:08:54,080 --> 00:08:58,160
Speaker 3: he was covering for the manipulators and the naked shorts

182
00:08:58,200 --> 00:09:00,400
Speaker 3: and whatnot. I've met both, I know, and a little

183
00:09:00,400 --> 00:09:01,959
Speaker 3: better than Gary. I don't think there are any two

184
00:09:02,000 --> 00:09:05,720
Speaker 3: people on Earth less likely to be in cahoots than

185
00:09:05,960 --> 00:09:07,960
Speaker 3: those two. I think they're on the opposite side of

186
00:09:08,000 --> 00:09:10,960
Speaker 3: most issues. But that was the theory. But both Ken

187
00:09:11,000 --> 00:09:13,760
Speaker 3: and Gary are too smart to respond to them on Twitter,

188
00:09:13,800 --> 00:09:17,120
Speaker 3: so I certainly became the most actively engaged. And then

189
00:09:17,240 --> 00:09:20,640
Speaker 3: I never did that again before today, when I've accidentally

190
00:09:20,640 --> 00:09:21,000
Speaker 3: done it.

191
00:09:21,120 --> 00:09:22,959
Speaker 4: Before we get back to stuff that matters, can I

192
00:09:23,000 --> 00:09:26,520
Speaker 4: just say on AMC, I tweeted, when did June two

193
00:09:26,559 --> 00:09:26,960
Speaker 4: come out?

194
00:09:27,080 --> 00:09:27,920
Speaker 2: It was like last summer.

195
00:09:28,120 --> 00:09:28,720
Speaker 3: Yeah, year ago.

196
00:09:28,800 --> 00:09:31,439
Speaker 4: I tweeted that I fell asleep during Dune two, and

197
00:09:31,600 --> 00:09:34,560
Speaker 4: it reawakened that crowd, at least on my Twitter, because

198
00:09:34,600 --> 00:09:35,839
Speaker 4: I've the same.

199
00:09:35,640 --> 00:09:41,080
Speaker 3: Crowds crowd because he's dissing a movie, don't you understand

200
00:09:41,120 --> 00:09:45,000
Speaker 3: you have to like every movie or else. You're anti America.

201
00:09:45,240 --> 00:09:48,000
Speaker 4: And then there were a lot of conspiracy theories about

202
00:09:48,040 --> 00:09:52,000
Speaker 4: Bloomberg reporter lies about falling asleep in doom too because

203
00:09:52,000 --> 00:09:52,920
Speaker 4: she hates AMC.

204
00:09:53,280 --> 00:09:54,800
Speaker 2: It would be, but it wasn't.

205
00:09:55,760 --> 00:09:59,520
Speaker 3: Oh yeah, you do innocuous that I was obnoxious, so

206
00:09:59,559 --> 00:10:02,040
Speaker 3: I'd kind of deserved it. You didn't deserve Thank you

207
00:10:02,120 --> 00:10:03,600
Speaker 3: for saying that you didn't deserve it.

208
00:10:03,679 --> 00:10:05,520
Speaker 1: I haven't sent. That's not doing what I was good.

209
00:10:05,880 --> 00:10:07,520
Speaker 2: That's all too. It's pretty good.

210
00:10:07,760 --> 00:10:10,560
Speaker 3: No you didn't, Actually I didn't. You were asleep.

211
00:10:10,960 --> 00:10:13,000
Speaker 2: Well no, not the whole time, not the whole.

212
00:10:12,760 --> 00:10:15,679
Speaker 1: Time, but also like you and and they were like,

213
00:10:15,800 --> 00:10:19,000
Speaker 1: give me some shorts, and you gave them some lungs

214
00:10:19,040 --> 00:10:21,720
Speaker 1: and shorts. Would you have known that or did you

215
00:10:21,760 --> 00:10:22,880
Speaker 1: have to be like I got it.

216
00:10:22,960 --> 00:10:24,960
Speaker 3: I would not have known that, right, So you don't know,

217
00:10:25,000 --> 00:10:26,560
Speaker 3: you're ar It would have been better for me if

218
00:10:26,600 --> 00:10:28,559
Speaker 3: we didn't have the call, because then I could have

219
00:10:28,679 --> 00:10:32,720
Speaker 3: just said I don't know. I rarely know individual stocks,

220
00:10:33,120 --> 00:10:35,400
Speaker 3: and if I do know, I'm probably not happy I know.

221
00:10:35,960 --> 00:10:38,120
Speaker 3: And even then, it's like we lost twenty BIPs on

222
00:10:38,160 --> 00:10:40,000
Speaker 3: that today, which is a giant number for us to

223
00:10:40,040 --> 00:10:44,000
Speaker 3: lose on one stock, and even then I probably don't notice.

224
00:10:43,520 --> 00:10:45,240
Speaker 1: Verse twenty bit. If someone comes to you and says

225
00:10:45,880 --> 00:10:46,200
Speaker 1: that's not.

226
00:10:46,280 --> 00:10:49,520
Speaker 3: One, I might be told about it. For us, it's

227
00:10:49,559 --> 00:10:52,280
Speaker 3: whether these seven hundred fifty stocks be these seven hundred

228
00:10:52,280 --> 00:10:56,120
Speaker 3: and fifty stocks. Yeah, I don't memorize fifteen hundred stocks. Now,

229
00:10:56,160 --> 00:10:58,880
Speaker 3: we take a fair amount of risks, some funds more,

230
00:10:58,960 --> 00:11:01,640
Speaker 3: some funds designed to be less. And that's about the

231
00:11:01,679 --> 00:11:05,520
Speaker 3: size of these two positions. So when I say small,

232
00:11:05,559 --> 00:11:07,920
Speaker 3: I'm not saying we're not both taking risk and trying

233
00:11:07,960 --> 00:11:10,920
Speaker 3: to generate pretty decent returns. But if you just think

234
00:11:10,920 --> 00:11:13,439
Speaker 3: about it, a quant is playing the odds. They're saying,

235
00:11:13,440 --> 00:11:17,679
Speaker 3: affirm a company with these characteristics. And this can be

236
00:11:17,960 --> 00:11:20,840
Speaker 3: old school factor quants from the nineteen nineties, these can

237
00:11:20,880 --> 00:11:24,040
Speaker 3: be modern machine learning. But with these characteristics tend to

238
00:11:24,040 --> 00:11:27,200
Speaker 3: beat these characteristics. If that's all you know, and it

239
00:11:27,240 --> 00:11:29,880
Speaker 3: is all we know, why on earth would you take

240
00:11:29,880 --> 00:11:32,360
Speaker 3: a lot of risk in any one company. AMC really

241
00:11:32,360 --> 00:11:33,960
Speaker 3: could have done well. It could have been bad on

242
00:11:34,040 --> 00:11:37,280
Speaker 3: every single thing that on average doesn't work, and it

243
00:11:37,280 --> 00:11:40,400
Speaker 3: could be a special situation that we don't understand. Something

244
00:11:40,400 --> 00:11:43,560
Speaker 3: could be good on everything and the CEO can embezzle

245
00:11:43,600 --> 00:11:46,120
Speaker 3: all the money. We don't want to take a lot

246
00:11:46,160 --> 00:11:48,160
Speaker 3: of risk on any one thing because we have no

247
00:11:48,200 --> 00:11:49,920
Speaker 3: insight in that it's risk for no return.

248
00:11:50,320 --> 00:11:53,000
Speaker 1: What thing you've written is that like over time, the

249
00:11:53,400 --> 00:11:56,439
Speaker 1: quant like factor model has moved closer to being what

250
00:11:56,559 --> 00:11:59,520
Speaker 1: Gramm and dot investors do. Like are you like an

251
00:11:59,559 --> 00:12:02,520
Speaker 1: abstract like Metagram and Dada investors? I like the way

252
00:12:02,559 --> 00:12:03,520
Speaker 1: to think of what you do.

253
00:12:04,320 --> 00:12:07,560
Speaker 3: I think it's moved closer. There are still differences and

254
00:12:07,640 --> 00:12:12,120
Speaker 3: maybe some of the like momentum if it's overreaction, if

255
00:12:12,120 --> 00:12:13,960
Speaker 3: you're riding momentum. I don't think a Graham and DoD

256
00:12:14,000 --> 00:12:16,439
Speaker 3: manager does that. So I don't want to push the analogy.

257
00:12:17,080 --> 00:12:19,760
Speaker 3: But this came out very very early in my career.

258
00:12:19,800 --> 00:12:21,840
Speaker 3: This is like thirty years ago. This is like my

259
00:12:21,880 --> 00:12:25,400
Speaker 3: Goldman Sax days. I started hearing a lot of active

260
00:12:25,400 --> 00:12:28,000
Speaker 3: stock pickers, some I'm still still friends with one guy

261
00:12:28,040 --> 00:12:31,000
Speaker 3: in particular. I was laughing at them, and I was

262
00:12:31,000 --> 00:12:32,720
Speaker 3: telling my friend, you all say the same thing. You

263
00:12:32,760 --> 00:12:35,960
Speaker 3: all say you're looking for valuation plus a catalyst. It's

264
00:12:36,000 --> 00:12:37,840
Speaker 3: like a I don't know if everyone says it, but

265
00:12:37,840 --> 00:12:41,360
Speaker 3: I've heard it many times and I'm making fun of him,

266
00:12:41,480 --> 00:12:42,960
Speaker 3: and at some point he looks at me and goes,

267
00:12:43,320 --> 00:12:47,319
Speaker 3: you do value on momentum, and it was a gotcha.

268
00:12:47,440 --> 00:12:50,360
Speaker 3: He won that round because I'm like, okay, I see

269
00:12:50,400 --> 00:12:53,400
Speaker 3: your point. So even back then, you can think of

270
00:12:53,400 --> 00:12:55,600
Speaker 3: those two together. We literally add them up. But if

271
00:12:55,600 --> 00:12:57,640
Speaker 3: you think of him as this holistic system, we're looking

272
00:12:57,640 --> 00:12:59,719
Speaker 3: for cheap things that are starting to get better in

273
00:12:59,760 --> 00:13:04,160
Speaker 3: price or fundamentals over time. And now I'm I'm not

274
00:13:04,200 --> 00:13:07,840
Speaker 3: talking the really modern stuff, alternative data, machine learning. I'm

275
00:13:07,840 --> 00:13:11,120
Speaker 3: talking just classic quant stuff that's been academia and then

276
00:13:11,200 --> 00:13:15,880
Speaker 3: imports over to applied. Profitability is a factor Robert nov

277
00:13:16,000 --> 00:13:18,880
Speaker 3: Mark's fort a great paper that we all incorporated, where

278
00:13:19,600 --> 00:13:23,240
Speaker 3: also equal give invaluation and momentum. If a company is

279
00:13:23,280 --> 00:13:28,119
Speaker 3: more profitable on some famous scale's gross profitability ROA Roe,

280
00:13:28,480 --> 00:13:31,720
Speaker 3: that's some degree a positive low beta investing. Two of

281
00:13:31,720 --> 00:13:35,839
Speaker 3: my colleagues, Andrea Ferzini and los A. Peterson, resurrecting stuff

282
00:13:35,840 --> 00:13:38,640
Speaker 3: Fisher Black did, and they're very good about saying Fisher

283
00:13:38,640 --> 00:13:42,960
Speaker 3: did it first. That lower beta stocks. If you're famously

284
00:13:43,000 --> 00:13:46,360
Speaker 3: a capital asset pricing model person, they're supposed to sell

285
00:13:46,360 --> 00:13:50,520
Speaker 3: on average underperform higher beta stocks. That's the main output

286
00:13:50,559 --> 00:13:53,400
Speaker 3: of that model. It doesn't work. I mean, it's one

287
00:13:53,400 --> 00:13:56,000
Speaker 3: of the largest empirical failures ever. It doesn't work in

288
00:13:56,000 --> 00:13:59,480
Speaker 3: any kind even outside of stocks people tested in other places. Therefore,

289
00:13:59,480 --> 00:14:02,559
Speaker 3: it kind of makes low beta stocks a little bit

290
00:14:02,559 --> 00:14:05,079
Speaker 3: of a free lunch because they are lower risk and

291
00:14:05,160 --> 00:14:08,800
Speaker 3: they keep up. If you actually go read Graham and DoD,

292
00:14:09,000 --> 00:14:13,000
Speaker 3: they're not just buying low multiples. They're much more holistic

293
00:14:13,040 --> 00:14:16,800
Speaker 3: than that. You know, high quality companies that have a moat,

294
00:14:16,960 --> 00:14:19,240
Speaker 3: that have some kind of margin of safety I think

295
00:14:19,360 --> 00:14:22,240
Speaker 3: was the term they use. Margin of safety and looking

296
00:14:22,240 --> 00:14:26,400
Speaker 3: for low risk doesn't sound so so different. So over

297
00:14:26,480 --> 00:14:29,640
Speaker 3: time I've thought at least a core amount of what

298
00:14:29,880 --> 00:14:32,400
Speaker 3: quants and academics, if you take them as a whole,

299
00:14:32,880 --> 00:14:36,200
Speaker 3: are finding, is the full paneplyy of stuff that a

300
00:14:36,240 --> 00:14:39,080
Speaker 3: Graham and DoD investor. We do it very differently. Again,

301
00:14:39,080 --> 00:14:41,600
Speaker 3: we're betting on the concept working on average. They are

302
00:14:41,680 --> 00:14:44,400
Speaker 3: using it in a soft or a hard sense as

303
00:14:44,440 --> 00:14:46,720
Speaker 3: a screen to look for candidates, and then they're trying

304
00:14:46,760 --> 00:14:50,320
Speaker 3: to learn a lot about that situation. They're upside as

305
00:14:50,360 --> 00:14:52,440
Speaker 3: if they learn a lot about that situation, they could

306
00:14:52,440 --> 00:14:54,840
Speaker 3: be more reliable than my You know, hey, we could

307
00:14:54,880 --> 00:14:57,960
Speaker 3: be wrong about AMC, and their downside is they better

308
00:14:57,960 --> 00:15:00,560
Speaker 3: be right. The concept can work, and they can still

309
00:15:00,560 --> 00:15:03,440
Speaker 3: lose if they're wrong about the specific So over time

310
00:15:03,480 --> 00:15:06,000
Speaker 3: I've gotten a little less hubrious about this. I think

311
00:15:06,120 --> 00:15:08,400
Speaker 3: quants caused the problem, by the way. I think when

312
00:15:08,480 --> 00:15:10,880
Speaker 3: Gene fam and Ken French started looking at like price

313
00:15:10,920 --> 00:15:14,200
Speaker 3: to book in the late eighties and early nineties, and

314
00:15:14,240 --> 00:15:15,560
Speaker 3: there were other people who did it too, I'm just

315
00:15:15,680 --> 00:15:17,800
Speaker 3: Chicago guy, so I'm going to just go with Fama

316
00:15:17,840 --> 00:15:21,640
Speaker 3: and French. They did it best. In my very biased opinion.

317
00:15:22,040 --> 00:15:23,720
Speaker 3: I don't think i'd have to go back and check,

318
00:15:23,760 --> 00:15:25,800
Speaker 3: but I don't think the first few papers use the

319
00:15:25,840 --> 00:15:30,600
Speaker 3: word value investing. Over time, low multiple investing in the

320
00:15:30,680 --> 00:15:34,720
Speaker 3: quant world came to be called value investing, and in

321
00:15:34,760 --> 00:15:36,480
Speaker 3: the Gramm and Dodd world they get kind of mad

322
00:15:36,520 --> 00:15:39,080
Speaker 3: at that, and they'd be like, it's not value investing.

323
00:15:39,120 --> 00:15:41,280
Speaker 3: There are plenty of low multiple companies that deserve to

324
00:15:41,280 --> 00:15:43,880
Speaker 3: be low multiple, and there are plenty of high multiple

325
00:15:43,880 --> 00:15:47,800
Speaker 3: companies that deserve it, and I think over time, the

326
00:15:48,000 --> 00:15:51,920
Speaker 3: quantitative process agreed with them more and more. So I

327
00:15:51,920 --> 00:15:54,440
Speaker 3: still think it's a communication problem because they'll still talk

328
00:15:54,440 --> 00:15:57,320
Speaker 3: about the value factor. In the quant world, that's just

329
00:15:57,400 --> 00:16:00,840
Speaker 3: low multiples. And if a Gramm and Dodd gets mad

330
00:16:00,920 --> 00:16:03,760
Speaker 3: or any old school active stock picker gets mad at that,

331
00:16:04,000 --> 00:16:06,960
Speaker 3: I'll just say, you're right, because value implies a more

332
00:16:06,960 --> 00:16:08,600
Speaker 3: holistic thing.

333
00:16:08,640 --> 00:16:10,640
Speaker 1: But isn't like modern quantit and what you're doing now

334
00:16:10,760 --> 00:16:13,000
Speaker 1: kind of just that more holistic thing, like you're just

335
00:16:13,160 --> 00:16:15,760
Speaker 1: ingesting more data plants and you have a less linear model,

336
00:16:15,800 --> 00:16:17,760
Speaker 1: and it's like moving towards that.

337
00:16:17,760 --> 00:16:19,960
Speaker 3: Anyway, if you look at machine learning, where either to

338
00:16:20,000 --> 00:16:23,680
Speaker 3: construct factors. One of the best uses of mL we

339
00:16:23,840 --> 00:16:26,920
Speaker 3: found is a subset of mL called natural language processing,

340
00:16:27,320 --> 00:16:30,560
Speaker 3: where you take textual data and you try to say,

341
00:16:30,600 --> 00:16:33,200
Speaker 3: is this good news or bad news? Quants have kind

342
00:16:33,200 --> 00:16:36,440
Speaker 3: of done this forever. You get transcript of an earnings call.

343
00:16:36,560 --> 00:16:38,040
Speaker 3: And the old school way to do this for a

344
00:16:38,040 --> 00:16:41,200
Speaker 3: long time was you count up good words and bad words,

345
00:16:41,640 --> 00:16:45,120
Speaker 3: good phrases and bad phrases, so increasing plus one and

346
00:16:45,160 --> 00:16:47,600
Speaker 3: you tech parse the whole thing, and sure, you guys

347
00:16:47,640 --> 00:16:50,400
Speaker 3: immediately see the problem. If the actual sentence was massive

348
00:16:50,400 --> 00:16:53,960
Speaker 3: embezzlement is increasing, then you were off on that plus one.

349
00:16:54,160 --> 00:16:58,000
Speaker 3: Quantitative stuff can survive doing some horrifically stupid things in isolation,

350
00:16:58,520 --> 00:17:00,720
Speaker 3: If fifty three percent of the time increasing is a

351
00:17:00,720 --> 00:17:04,440
Speaker 3: good word, than forty percent of the time you were stupid,

352
00:17:04,840 --> 00:17:07,840
Speaker 3: turns out natural language processing or NLP, if we want

353
00:17:07,880 --> 00:17:10,720
Speaker 3: to sound like the cool kids. That is, taking that

354
00:17:10,800 --> 00:17:16,480
Speaker 3: same data and training a mL model to say what

355
00:17:16,560 --> 00:17:18,840
Speaker 3: predicts and what doesn't predict, and of course never gets

356
00:17:18,840 --> 00:17:21,320
Speaker 3: near perfect at the end of the day, though we

357
00:17:21,440 --> 00:17:24,240
Speaker 3: believe it does a lot better than the word count

358
00:17:24,280 --> 00:17:27,600
Speaker 3: methods and is additive to a model. But importantly, and

359
00:17:27,640 --> 00:17:31,920
Speaker 3: I think this was your point, Matt, it's not qualitatively different.

360
00:17:32,200 --> 00:17:36,760
Speaker 3: We've looked at both price and fundamental momentum forever. Fundamental

361
00:17:36,760 --> 00:17:39,360
Speaker 3: momentum the classic measures of things like our earnings being

362
00:17:39,440 --> 00:17:41,600
Speaker 3: revised and you want the revision. You want the new

363
00:17:41,640 --> 00:17:46,440
Speaker 3: news up faster or slower. Our earning surprise is coming

364
00:17:46,440 --> 00:17:48,639
Speaker 3: in positive or negative. And this is this anchoring and

365
00:17:48,720 --> 00:17:52,320
Speaker 3: adjustment idea that if that's good or bad news, you

366
00:17:52,400 --> 00:17:54,240
Speaker 3: can make money trading on it. Because it's not fully

367
00:17:54,280 --> 00:17:58,440
Speaker 3: incorporated in the stock price parsing and earnings call poorly

368
00:17:58,680 --> 00:18:01,080
Speaker 3: as in the past or better now, we think of

369
00:18:01,119 --> 00:18:03,440
Speaker 3: it as just another form of our good things happening

370
00:18:03,800 --> 00:18:05,560
Speaker 3: in the new year term, and if so, they're probably

371
00:18:05,680 --> 00:18:10,119
Speaker 3: under appreciated. So yes, I don't think it's changed dramatically.

372
00:18:10,119 --> 00:18:13,520
Speaker 3: It's much more of an evolution rather than a spirit change.

373
00:18:13,760 --> 00:18:16,280
Speaker 3: But we got bigger tools in the tool chest now.

374
00:18:16,840 --> 00:18:18,399
Speaker 1: I mean, like, you know, you ask, like what a

375
00:18:18,560 --> 00:18:20,879
Speaker 1: sort of like traditional old school fundamental manager would do.

376
00:18:20,880 --> 00:18:22,480
Speaker 1: One thing they do is listen to their next call

377
00:18:22,720 --> 00:18:25,280
Speaker 1: and like talk to management. So it's like your machines

378
00:18:25,280 --> 00:18:27,240
Speaker 1: are moving in the direction of being an old school analyst.

379
00:18:27,440 --> 00:18:32,080
Speaker 3: Yeah, as someone with four kids in or around college age,

380
00:18:32,600 --> 00:18:34,199
Speaker 3: my wife and I and she's done a lot more

381
00:18:34,200 --> 00:18:35,680
Speaker 3: of this, has spent a lot of time trying to

382
00:18:35,720 --> 00:18:41,600
Speaker 3: figure out what careers will not be utterly destroyed by MLO,

383
00:18:41,800 --> 00:18:47,680
Speaker 3: and it's not an easy question. A podcaster is probably

384
00:18:47,680 --> 00:18:49,240
Speaker 3: a good one for a while. They have those fakes,

385
00:18:52,000 --> 00:18:55,399
Speaker 3: so it will all be podcasters. But yeah, it's another example.

386
00:18:55,440 --> 00:18:58,800
Speaker 3: I always say, these useless statements like them sufficiently long horizon,

387
00:18:58,840 --> 00:19:01,480
Speaker 3: we're all replaced by machine learning. It's a question of

388
00:19:01,640 --> 00:19:05,960
Speaker 3: what we're getting way too philosophical, but the transition to

389
00:19:06,040 --> 00:19:09,280
Speaker 3: that can be very painful and weird. But a world

390
00:19:09,359 --> 00:19:13,359
Speaker 3: of abundance and leisure. Maybe our horrible fate in the

391
00:19:13,359 --> 00:19:14,520
Speaker 3: long term, I'll take it.

392
00:19:14,640 --> 00:19:27,640
Speaker 5: Yeah.

393
00:19:30,119 --> 00:19:32,280
Speaker 1: I want to talk about market timing because I was

394
00:19:32,320 --> 00:19:35,280
Speaker 1: reading the Virtue of Complexity paper from Brian Kelly at All.

395
00:19:35,200 --> 00:19:36,800
Speaker 3: Which is I will say he may be one of

396
00:19:36,840 --> 00:19:39,679
Speaker 3: the only hosts of a podcast to read that paper.

397
00:19:39,720 --> 00:19:42,560
Speaker 3: This is not a simple paper. It's like he writes,

398
00:19:42,640 --> 00:19:43,000
Speaker 3: very clue.

399
00:19:43,040 --> 00:19:44,880
Speaker 1: It's the sort of like it's the notion of taking

400
00:19:44,920 --> 00:19:46,760
Speaker 1: like a sort of simple factor model and blowing it

401
00:19:46,840 --> 00:19:50,760
Speaker 1: up into like a non linear AI model, and one

402
00:19:51,000 --> 00:19:55,639
Speaker 1: almost throwaway sentence in the pieces like this, like simplified

403
00:19:55,680 --> 00:19:59,200
Speaker 1: AI model that he built for like illustrative purposes, lowered

404
00:19:59,240 --> 00:20:03,119
Speaker 1: its risk before fourteen of the last fifteen recessions. And

405
00:20:03,160 --> 00:20:06,359
Speaker 1: I always thought the naive like the best way to

406
00:20:06,400 --> 00:20:09,199
Speaker 1: invest would be just market timing, just like you know,

407
00:20:09,240 --> 00:20:10,879
Speaker 1: have all your money in the stock market before the

408
00:20:10,880 --> 00:20:12,719
Speaker 1: market goes up and not before it goes down, if

409
00:20:12,760 --> 00:20:15,440
Speaker 1: you can do it. My impression is that like respectable

410
00:20:15,480 --> 00:20:19,399
Speaker 1: headgsplot managers, respectable clime managers, respectable academics say the hardest

411
00:20:19,400 --> 00:20:21,760
Speaker 1: thing in the world is market timing, and like no

412
00:20:21,800 --> 00:20:23,560
Speaker 1: one claims to get alpha from at and it's not

413
00:20:23,600 --> 00:20:27,440
Speaker 1: a thing. Is that changed, and is that changed due

414
00:20:27,480 --> 00:20:28,399
Speaker 1: to like machine learning?

415
00:20:28,840 --> 00:20:33,240
Speaker 3: It has changed a big Do we do trend following

416
00:20:33,520 --> 00:20:36,720
Speaker 3: on macro assets old school CTA stuff. We think we've

417
00:20:36,720 --> 00:20:39,400
Speaker 3: made a new school by incorporating fundamental momentum by doing

418
00:20:39,440 --> 00:20:42,720
Speaker 3: a lot of more esoteric market so we think even

419
00:20:42,720 --> 00:20:45,360
Speaker 3: that's had a march of progress to it, But all

420
00:20:45,359 --> 00:20:48,440
Speaker 3: else equal. I wrote my dissertation on momentum in individual

421
00:20:48,480 --> 00:20:51,720
Speaker 3: stocks for some reason that I cannot explain. If you're

422
00:20:51,800 --> 00:20:54,920
Speaker 3: using past returns to predict the future, just what's going

423
00:20:55,000 --> 00:20:57,560
Speaker 3: up will keep going up, and vice versa to pick

424
00:20:57,600 --> 00:21:01,480
Speaker 3: individual stocks. The whole industry calls momentum. And if you

425
00:21:01,560 --> 00:21:04,040
Speaker 3: think markets tend to keep going in the same direction,

426
00:21:04,119 --> 00:21:07,399
Speaker 3: everyone calls it trend falling. Same thing. Starting in I

427
00:21:07,400 --> 00:21:09,800
Speaker 3: think two thousand and eight, we started. We always had

428
00:21:09,840 --> 00:21:13,000
Speaker 3: it in our macro models, but we started formally offering

429
00:21:13,040 --> 00:21:16,800
Speaker 3: separate trend falling products. It doesn't take hero bets on

430
00:21:16,840 --> 00:21:21,240
Speaker 3: anyone market. It's not Gazarelli selling all the stocks a

431
00:21:21,320 --> 00:21:24,400
Speaker 3: minute before October nineteenth of eighty seven. I'm dating myself.

432
00:21:24,880 --> 00:21:27,520
Speaker 3: My wife's birthday is October nineteenth, which always always gets

433
00:21:27,520 --> 00:21:30,480
Speaker 3: a little amused. Man crash eighty seven meant did you

434
00:21:30,560 --> 00:21:33,159
Speaker 3: really crashed them together? Yeah, there have been some jokes

435
00:21:33,160 --> 00:21:38,000
Speaker 3: over over time. So trend following it's essentially market timing,

436
00:21:38,000 --> 00:21:40,919
Speaker 3: but it's highly diverse, many small bets. Whatever's been happening

437
00:21:40,960 --> 00:21:43,840
Speaker 3: tends to keep happening. That's the main way we'll do

438
00:21:43,920 --> 00:21:44,520
Speaker 3: market timing.

439
00:21:44,600 --> 00:21:46,760
Speaker 1: You won't to like your main equity fun is like

440
00:21:46,840 --> 00:21:47,399
Speaker 1: sometimes that.

441
00:21:48,000 --> 00:21:50,480
Speaker 3: That was kind of a toy model to illustrate a point.

442
00:21:51,040 --> 00:21:52,760
Speaker 3: I know we're not taking a lot of risks on

443
00:21:52,800 --> 00:21:55,399
Speaker 3: that model. Market timing I still think is quite hard.

444
00:21:55,760 --> 00:21:59,680
Speaker 3: Might there be advances in it in the future, absolutely,

445
00:22:00,080 --> 00:22:04,879
Speaker 3: but are we taking significant risk in it now aside

446
00:22:04,880 --> 00:22:06,840
Speaker 3: from trend following, not really.

447
00:22:07,160 --> 00:22:09,800
Speaker 1: You probably sit been more papers to financial journals than

448
00:22:09,800 --> 00:22:11,600
Speaker 1: the average like asset.

449
00:22:11,280 --> 00:22:12,640
Speaker 3: Manager, by like one hundred percent.

450
00:22:12,720 --> 00:22:14,520
Speaker 1: I want to ask two questions. One is like, I

451
00:22:14,520 --> 00:22:16,960
Speaker 1: want to learn about your relationship with academia, because I

452
00:22:16,960 --> 00:22:20,119
Speaker 1: think it is fascinating that like you employ half of

453
00:22:20,160 --> 00:22:24,000
Speaker 1: the l faculty the finance PhD pipeline runs mainly to

454
00:22:24,000 --> 00:22:26,399
Speaker 1: AQR and then too because I think of like what

455
00:22:26,480 --> 00:22:30,560
Speaker 1: HEGEMA manager does as sort of like finding anomalies, finding

456
00:22:30,600 --> 00:22:34,480
Speaker 1: like market inefficiencies, finding factors that are predictive every turns,

457
00:22:34,520 --> 00:22:36,439
Speaker 1: and I think of what a finance academic does is

458
00:22:36,480 --> 00:22:38,960
Speaker 1: mainly also that when do you publish and when do

459
00:22:39,000 --> 00:22:39,679
Speaker 1: you just trade on it?

460
00:22:39,720 --> 00:22:42,199
Speaker 3: Oh, there's so much to talk about here. First, there

461
00:22:42,240 --> 00:22:44,840
Speaker 3: are a lot of reasons we do it. One is

462
00:22:44,920 --> 00:22:48,159
Speaker 3: just personal consumption. We grew up in this world. We

463
00:22:48,359 --> 00:22:50,600
Speaker 3: liked being part of it. We were interested in this

464
00:22:50,600 --> 00:22:54,880
Speaker 3: stuff in a purely academic sense before we got seduced

465
00:22:54,920 --> 00:22:55,760
Speaker 3: into making money.

466
00:22:56,160 --> 00:22:58,960
Speaker 1: I mean, I would like pause it that you think

467
00:22:59,000 --> 00:23:02,840
Speaker 1: there is a value to employing the finance PhDs to

468
00:23:03,000 --> 00:23:06,480
Speaker 1: like build your models, and the way to attract fancy

469
00:23:06,480 --> 00:23:09,679
Speaker 1: finance PhDs is to offer them the most academia like

470
00:23:09,720 --> 00:23:12,679
Speaker 1: possible working.

471
00:23:11,680 --> 00:23:14,200
Speaker 3: And to letting them publish. There are people, including our

472
00:23:14,240 --> 00:23:17,159
Speaker 3: two Yale professors. I don't don't know if would have

473
00:23:17,359 --> 00:23:19,120
Speaker 3: come take you are if we said you can never

474
00:23:19,160 --> 00:23:23,520
Speaker 3: write about this. We do a crass calculation, though, if

475
00:23:23,560 --> 00:23:27,360
Speaker 3: we think there is something that we are reletively unique

476
00:23:27,840 --> 00:23:30,639
Speaker 3: on entirely unique, or a very small handful of people

477
00:23:31,200 --> 00:23:36,240
Speaker 3: know this, we won't publish it. The optimal the optimal

478
00:23:36,280 --> 00:23:39,320
Speaker 3: time to publish a paper is after you've made money

479
00:23:39,359 --> 00:23:42,440
Speaker 3: from something for eleven years and an hour and a

480
00:23:42,480 --> 00:23:44,480
Speaker 3: half before someone else is going to publish the paper,

481
00:23:45,200 --> 00:23:47,800
Speaker 3: and we cannot get that right. I can think of

482
00:23:47,840 --> 00:23:51,520
Speaker 3: one example of momentum in factors, the fact that factors themselves,

483
00:23:51,520 --> 00:23:55,760
Speaker 3: like value, profitability, also exhibit momentum as something that we've

484
00:23:55,760 --> 00:23:58,359
Speaker 3: traded on for many years that we've always refrained from

485
00:23:58,359 --> 00:24:00,880
Speaker 3: writing a paper on because no one else has. And again,

486
00:24:00,920 --> 00:24:02,600
Speaker 3: we knew we couldn't be the only people who do this,

487
00:24:02,680 --> 00:24:04,280
Speaker 3: but we didn't think the cat was out of the bag.

488
00:24:04,840 --> 00:24:07,080
Speaker 3: And then someone else wrote the paper. And I'm sure

489
00:24:07,119 --> 00:24:09,160
Speaker 3: we've done this to other people too, because you never

490
00:24:09,240 --> 00:24:12,240
Speaker 3: know what they're doing internally. So life, you know, you

491
00:24:12,280 --> 00:24:14,040
Speaker 3: do it long enough, life works out kind of fair,

492
00:24:14,240 --> 00:24:15,200
Speaker 3: but that is the goal.

493
00:24:15,320 --> 00:24:18,520
Speaker 1: But you were mad there because, like as academics, you

494
00:24:18,640 --> 00:24:19,920
Speaker 1: wanted credit for that paper.

495
00:24:20,040 --> 00:24:23,000
Speaker 3: It's fun. You could call it childish, but it's human.

496
00:24:23,080 --> 00:24:24,680
Speaker 3: It's it's fun to discover something.

497
00:24:24,680 --> 00:24:27,639
Speaker 1: Well, you also mad because publishing that reduces the value

498
00:24:27,640 --> 00:24:28,679
Speaker 1: of the signal or maybe a.

499
00:24:28,680 --> 00:24:30,880
Speaker 3: Little bit, maybe a little bit. So far, it's still

500
00:24:30,880 --> 00:24:33,439
Speaker 3: worked wonderfully. The trend is still your friend when it

501
00:24:33,440 --> 00:24:36,600
Speaker 3: comes to two factors, but it would be a fireable

502
00:24:36,640 --> 00:24:38,720
Speaker 3: offense for make you are to publish something that we

503
00:24:38,760 --> 00:24:41,879
Speaker 3: thought was truly proprietary. Let me give you my favorite

504
00:24:41,920 --> 00:24:44,159
Speaker 3: example of this, because it came up recently. One of

505
00:24:44,200 --> 00:24:46,840
Speaker 3: the things that we've gotten into in a fairly big way,

506
00:24:46,840 --> 00:24:49,680
Speaker 3: as have some other quants, is what's called alternative data.

507
00:24:50,119 --> 00:24:52,920
Speaker 3: Those are new data sets that put people put together

508
00:24:52,920 --> 00:24:56,280
Speaker 3: with sweat equity. The classic example and the only one.

509
00:24:56,320 --> 00:24:58,200
Speaker 3: And this is the point that I'm allowed to talk

510
00:24:58,240 --> 00:25:01,120
Speaker 3: about our credit card receiver data were a credit cards

511
00:25:01,119 --> 00:25:03,520
Speaker 3: because that's been discussed a million times.

512
00:25:04,080 --> 00:25:04,639
Speaker 1: Alternative there.

513
00:25:04,720 --> 00:25:08,400
Speaker 3: Yeah, yeah, Well that's the point is this stuff is arbitrageble.

514
00:25:08,560 --> 00:25:12,200
Speaker 3: I mean, it can go away. Value, be it the

515
00:25:12,320 --> 00:25:14,520
Speaker 3: narrow Quan sense of value or the more broad gram

516
00:25:14,520 --> 00:25:17,639
Speaker 3: and Odd sense of value, is trying to take advantage

517
00:25:17,680 --> 00:25:20,560
Speaker 3: of I think basic human nature. I think spreads being

518
00:25:20,640 --> 00:25:23,760
Speaker 3: cheap and expensive are wider still than the historical average,

519
00:25:23,800 --> 00:25:25,480
Speaker 3: not tighter. I don't think there's a lot of evidence

520
00:25:25,520 --> 00:25:27,720
Speaker 3: that so much capital is in that that it's making

521
00:25:27,720 --> 00:25:31,000
Speaker 3: you go away. If you get a short term information

522
00:25:31,119 --> 00:25:34,440
Speaker 3: advantage because you have put the time in or paid

523
00:25:34,480 --> 00:25:36,680
Speaker 3: someone who's put the time in to create a new

524
00:25:36,760 --> 00:25:40,159
Speaker 3: data set that other people don't have. They're going to

525
00:25:40,240 --> 00:25:43,600
Speaker 3: have it eventually. So I am talking to the Australian

526
00:25:43,640 --> 00:25:46,600
Speaker 3: Financial Review. I don't even know if we had discussed

527
00:25:46,600 --> 00:25:49,639
Speaker 3: alternative data beforehand, but he asked me about it, and

528
00:25:49,720 --> 00:25:53,440
Speaker 3: I said to him that our head of stock selection,

529
00:25:53,680 --> 00:25:58,560
Speaker 3: a fellow named Andrea Ferzini, has asked me, I might

530
00:25:58,560 --> 00:26:01,320
Speaker 3: have said, told me, but has asked me not to

531
00:26:01,359 --> 00:26:03,320
Speaker 3: talk about these things. There are things I'll talk about,

532
00:26:03,320 --> 00:26:06,200
Speaker 3: but there are things we think are you know, overused word,

533
00:26:06,280 --> 00:26:09,359
Speaker 3: but our true alpha that the world doesn't know, and

534
00:26:09,640 --> 00:26:12,200
Speaker 3: I think it's reasonable, So I really can't talk about him.

535
00:26:12,560 --> 00:26:16,879
Speaker 3: He writes the article and what it says is mostly that,

536
00:26:17,040 --> 00:26:19,080
Speaker 3: but instead of saying he's asked me not to talk

537
00:26:19,080 --> 00:26:23,040
Speaker 3: about it, it says Andrea Ferzini won't tell me what

538
00:26:23,160 --> 00:26:27,800
Speaker 3: we're doing in alternative data, which it has a slightly

539
00:26:27,800 --> 00:26:30,720
Speaker 3: different connotation, and that's a connotation of adult old man,

540
00:26:30,760 --> 00:26:33,240
Speaker 3: don't worry about it. We're doing some stuff here. So

541
00:26:33,480 --> 00:26:36,520
Speaker 3: again it's an example where there certainly are things we

542
00:26:36,560 --> 00:26:37,320
Speaker 3: won't talk about.

543
00:26:37,640 --> 00:26:39,960
Speaker 1: Is that ever true? Are you like most cutting edge

544
00:26:40,000 --> 00:26:42,040
Speaker 1: machine learning people like that don't or are that or

545
00:26:42,080 --> 00:26:43,359
Speaker 1: do you read all the papers?

546
00:26:43,480 --> 00:26:45,480
Speaker 3: I read most of the papers. I don't read all

547
00:26:45,520 --> 00:26:48,479
Speaker 3: of the papers. I at least skim all the papers.

548
00:26:48,480 --> 00:26:51,200
Speaker 3: I know the gist. We do a lot of different

549
00:26:51,200 --> 00:26:53,280
Speaker 3: things at this point. Once you're a quant you want

550
00:26:53,280 --> 00:26:55,240
Speaker 3: more factors, you want to trade it in more places.

551
00:26:55,880 --> 00:26:59,560
Speaker 3: And I will admit that the week doesn't go by

552
00:26:59,560 --> 00:27:01,520
Speaker 3: where I don't and ask someone for a review of

553
00:27:01,520 --> 00:27:04,200
Speaker 3: what we're doing. Somewhere at some point I approved it,

554
00:27:05,000 --> 00:27:09,240
Speaker 3: so Pastcliff had some understanding. Yeah, but also just you know,

555
00:27:09,320 --> 00:27:11,640
Speaker 3: some of this is pretty decent math, and the old

556
00:27:11,680 --> 00:27:14,720
Speaker 3: mathematicians do their best work in their twenties. I can

557
00:27:14,760 --> 00:27:17,200
Speaker 3: tell you I'm probably still decent in math, but I'm

558
00:27:17,200 --> 00:27:19,800
Speaker 3: not what I was when I was twenty two. Wisdom

559
00:27:19,840 --> 00:27:23,480
Speaker 3: has hopefully replaced some of mathematical ability.

560
00:27:23,720 --> 00:27:27,280
Speaker 1: Because the personnel at AQR different from a like I

561
00:27:27,280 --> 00:27:29,640
Speaker 1: don't know who you think of as like quantity competitors,

562
00:27:29,800 --> 00:27:33,280
Speaker 1: but like, my impression is that like you have many

563
00:27:33,320 --> 00:27:37,520
Speaker 1: more finance PhDs than like other places that might have

564
00:27:37,600 --> 00:27:41,199
Speaker 1: more like pure math people or math undergrads or something like.

565
00:27:41,440 --> 00:27:43,760
Speaker 3: Yeah, I think there's some truth to that.

566
00:27:44,000 --> 00:27:45,200
Speaker 1: What makes them better or worse?

567
00:27:45,560 --> 00:27:47,520
Speaker 3: First, I think there's going to be a correlation that

568
00:27:47,560 --> 00:27:50,000
Speaker 3: the closer you get to the high frequency world, the

569
00:27:50,000 --> 00:27:52,720
Speaker 3: more you're going to be more in the pure math realm.

570
00:27:53,160 --> 00:27:57,840
Speaker 3: My very tortured analogy is quantum mechanics versus Newtonian physics.

571
00:27:58,240 --> 00:28:00,720
Speaker 3: When you get into the high frequency world, first you

572
00:28:00,760 --> 00:28:03,800
Speaker 3: have a ton of data. By nature, you just have

573
00:28:03,880 --> 00:28:07,240
Speaker 3: a lot more instances and a lot less theory. So

574
00:28:07,440 --> 00:28:10,400
Speaker 3: turning yourself over to the data more so than having

575
00:28:10,440 --> 00:28:14,560
Speaker 3: an economic rationale is far more rational, and that becomes

576
00:28:14,560 --> 00:28:17,640
Speaker 3: a more mathematical exercise. We tend to think in our

577
00:28:18,080 --> 00:28:21,879
Speaker 3: long seven hundred short seven hundred stock portfolios, average holding

578
00:28:21,880 --> 00:28:25,000
Speaker 3: period is maybe nine months, maybe closer a year. Times.

579
00:28:25,680 --> 00:28:27,160
Speaker 3: That's nowhere near high frequency.

580
00:28:27,400 --> 00:28:28,120
Speaker 2: Yeah, high frequent.

581
00:28:28,119 --> 00:28:29,919
Speaker 3: Two. When I wrote my dissertation and I did have

582
00:28:30,000 --> 00:28:33,479
Speaker 3: this in there was a monthly contrarian strategy. Now we're talking,

583
00:28:33,680 --> 00:28:36,280
Speaker 3: you know, sub seconds kind of thing. I think when

584
00:28:36,320 --> 00:28:40,240
Speaker 3: you're gonna have a medium holding period strategy, as I

585
00:28:40,280 --> 00:28:42,440
Speaker 3: think of us, we're not Warren Buffett, but we're not HFT.

586
00:28:43,520 --> 00:28:48,080
Speaker 3: Even the machine learning stuff needs some economics too. You

587
00:28:48,120 --> 00:28:50,360
Speaker 3: simply don't have enough data and there's too much of

588
00:28:50,400 --> 00:28:54,000
Speaker 3: a dimensionality. Even with Brian's virtue of complexity, you need

589
00:28:54,040 --> 00:28:57,000
Speaker 3: to give it some structure or else it's going to

590
00:28:57,080 --> 00:29:00,880
Speaker 3: overfit and go mad. So I think I think that's the reason,

591
00:29:01,120 --> 00:29:03,440
Speaker 3: not saying we'll never do something in a higher frequency world,

592
00:29:03,440 --> 00:29:05,920
Speaker 3: in which case we'd probably have to shift more. But

593
00:29:06,080 --> 00:29:11,320
Speaker 3: in our world, being a mathematician, being an excellent programmer,

594
00:29:11,480 --> 00:29:14,680
Speaker 3: but also having the economics behind it is kind of

595
00:29:14,680 --> 00:29:15,520
Speaker 3: what we're looking for.

596
00:29:16,240 --> 00:29:21,360
Speaker 1: I think of like renaissances famously employing exclusively people who

597
00:29:21,440 --> 00:29:26,200
Speaker 1: have never thought about markets or financer economics, and they

598
00:29:26,240 --> 00:29:29,360
Speaker 1: come to it pure, and I feel that you are

599
00:29:29,480 --> 00:29:31,800
Speaker 1: very much like people who have math shops, but like

600
00:29:32,920 --> 00:29:35,720
Speaker 1: have economic intuition and think about it as an economy.

601
00:29:36,040 --> 00:29:39,239
Speaker 3: I think that's accurate. I do have a couple of

602
00:29:39,400 --> 00:29:44,200
Speaker 3: Renaissance observations. Okay. One of my favorite questions people asked

603
00:29:44,240 --> 00:29:46,400
Speaker 3: me was howd they do it? And I love that

604
00:29:46,480 --> 00:29:49,480
Speaker 3: question because I get to respond. So your hypothesis is,

605
00:29:49,640 --> 00:29:53,040
Speaker 3: I know, how they took a few billion dollars and

606
00:29:53,040 --> 00:29:54,800
Speaker 3: still take a few billion dollars out of the market

607
00:29:54,880 --> 00:29:57,840
Speaker 3: to share among a relevant tiny group of people every

608
00:29:57,920 --> 00:30:00,760
Speaker 3: year would apparently very low risk, and I choose not to.

609
00:30:01,440 --> 00:30:04,560
Speaker 3: I have some inklings of the general whether they've worked on.

610
00:30:04,600 --> 00:30:07,080
Speaker 3: My biggest guess is that they were ten fifteen years

611
00:30:07,120 --> 00:30:09,120
Speaker 3: ahead of everyone else on most of this stuff and

612
00:30:09,160 --> 00:30:12,480
Speaker 3: are just have developed more sophisticated systems over time. I

613
00:30:12,480 --> 00:30:14,760
Speaker 3: think natural language processing they were very early on.

614
00:30:15,160 --> 00:30:17,640
Speaker 1: It came from like from.

615
00:30:18,480 --> 00:30:20,560
Speaker 3: A few other things. John Leu and I, my one

616
00:30:20,600 --> 00:30:23,600
Speaker 3: of my founding partners, wrote when Fama and Schiller shared

617
00:30:23,600 --> 00:30:26,280
Speaker 3: the Nobel Prize, we wrote a whole overview of market

618
00:30:26,280 --> 00:30:29,080
Speaker 3: efficiency and the debate about it, and I brought them

619
00:30:29,120 --> 00:30:32,680
Speaker 3: up as an example. The Medallion Fund has almost nothing

620
00:30:32,680 --> 00:30:35,680
Speaker 3: to say about market efficiency. It says these guys can

621
00:30:35,720 --> 00:30:40,560
Speaker 3: extract a toll on the market with reliable consistency. But

622
00:30:40,680 --> 00:30:43,720
Speaker 3: in terms of market size it's a giant number for

623
00:30:43,760 --> 00:30:46,120
Speaker 3: a few people to make each year. It's a tiny

624
00:30:46,200 --> 00:30:48,280
Speaker 3: number in terms of whether the markets are efficient, and

625
00:30:48,280 --> 00:30:50,680
Speaker 3: apparently the rest of us can't do it quite as

626
00:30:50,720 --> 00:30:53,000
Speaker 3: well as them. They're the goat when it comes to this. Yeah.

627
00:30:53,040 --> 00:30:56,320
Speaker 1: I think of like, you know, like Jane Street can

628
00:30:56,360 --> 00:30:58,360
Speaker 1: rely excite somebody from the market, and I think of

629
00:30:58,360 --> 00:31:00,360
Speaker 1: that as like a few first service ye, like a

630
00:31:00,400 --> 00:31:03,280
Speaker 1: market maker, rather than being like a predictor of asset precesce.

631
00:31:03,320 --> 00:31:05,440
Speaker 3: I'm gonna guess, and again it's a purely guess that

632
00:31:05,560 --> 00:31:07,959
Speaker 3: not all of it, but a fair amount of Renaissance

633
00:31:07,960 --> 00:31:09,840
Speaker 3: had a similar similar flavor.

634
00:31:09,920 --> 00:31:11,800
Speaker 1: I think when you talk to people who run pod shops, like,

635
00:31:11,840 --> 00:31:14,520
Speaker 1: some of what they do they conceive of as it's

636
00:31:14,520 --> 00:31:17,280
Speaker 1: not literally market making, but it has that sort of flavor.

637
00:31:17,000 --> 00:31:20,880
Speaker 3: And the constraint on those things is typically capacity. Now

638
00:31:20,880 --> 00:31:23,200
Speaker 3: it's a ton of money for those firms. They're amazing firms,

639
00:31:23,960 --> 00:31:26,800
Speaker 3: but give you an example of Renaissance. I've complimented them

640
00:31:26,840 --> 00:31:28,560
Speaker 3: like crazy, so hopefully they won't be mad at me

641
00:31:28,640 --> 00:31:32,520
Speaker 3: for this. One last comment. Everyone talks about the Medallion Fund,

642
00:31:32,760 --> 00:31:34,640
Speaker 3: but no one's allowed to invest in that. They keep

643
00:31:34,640 --> 00:31:38,200
Speaker 3: that for themselves. Another favorite question I get is this

644
00:31:38,240 --> 00:31:40,160
Speaker 3: doesn't happen much anymore, but every once in a while

645
00:31:40,320 --> 00:31:42,640
Speaker 3: I would get an investor saying, so, Medallion Fund better

646
00:31:42,680 --> 00:31:45,719
Speaker 3: than you guys. I'd be like, oh, hell yes, but

647
00:31:45,800 --> 00:31:48,200
Speaker 3: they won't take your money or my money, and I

648
00:31:48,200 --> 00:31:51,000
Speaker 3: think we're pretty darn good and we will. So why

649
00:31:51,040 --> 00:31:54,200
Speaker 3: is that relevant? But Renaissance also runs a fair amount

650
00:31:54,240 --> 00:31:58,480
Speaker 3: of money in more open institutional products where they look

651
00:31:58,560 --> 00:32:00,880
Speaker 3: very good. I'm not gonna to ride them at all,

652
00:32:01,200 --> 00:32:03,480
Speaker 3: but they don't look better than us. They look like

653
00:32:04,040 --> 00:32:09,360
Speaker 3: really good, solid, regular quants. So what they cannot do,

654
00:32:09,600 --> 00:32:11,960
Speaker 3: I think it's kind of obvious, is take the medallion

655
00:32:12,040 --> 00:32:15,480
Speaker 3: process and scale it up many, many times bigger. They've

656
00:32:15,480 --> 00:32:18,360
Speaker 3: discovered a way, as you put it, to just take

657
00:32:18,400 --> 00:32:20,480
Speaker 3: a certain amount out to provide a service. I called

658
00:32:20,480 --> 00:32:23,880
Speaker 3: it taking a toll, and that's amazing. But they've not

659
00:32:23,920 --> 00:32:27,360
Speaker 3: discovered a way to do that at institutional scale, which

660
00:32:27,400 --> 00:32:29,240
Speaker 3: thank god, because that leaves something for the rest of

661
00:32:29,320 --> 00:32:29,719
Speaker 3: us to do.

662
00:32:30,040 --> 00:32:31,720
Speaker 1: I'll get back to something you talked about the very beginning,

663
00:32:31,720 --> 00:32:33,440
Speaker 1: because you mentioned that paper you wrote with John Leu,

664
00:32:33,520 --> 00:32:38,040
Speaker 1: which you talking about like the two possible explanations for

665
00:32:38,400 --> 00:32:41,560
Speaker 1: making money for anomolies for whatever, which are sort of

666
00:32:41,560 --> 00:32:46,360
Speaker 1: behavioral irrationality or whatever, and you were bearing a risk

667
00:32:46,440 --> 00:32:49,120
Speaker 1: for someone. Everything you write, and like what you said

668
00:32:49,160 --> 00:32:51,880
Speaker 1: here you have, you seem sort of like agnostic about

669
00:32:52,200 --> 00:32:55,840
Speaker 1: which one or what combination there's like do you prefer one,

670
00:32:56,280 --> 00:32:58,280
Speaker 1: is it more reliable to get paid for taking your

671
00:32:58,360 --> 00:32:58,920
Speaker 1: risk or.

672
00:32:59,520 --> 00:33:02,000
Speaker 3: Well forgetting which one you think is really going on?

673
00:33:02,120 --> 00:33:05,240
Speaker 3: They do have different characteristics. The positive behind a risk

674
00:33:05,240 --> 00:33:08,400
Speaker 3: premium is you may be able to get it forever

675
00:33:08,560 --> 00:33:11,880
Speaker 3: because it's rational that you get it. The negative behind

676
00:33:11,920 --> 00:33:14,280
Speaker 3: it is it's a risk premium. You have to figure

677
00:33:14,320 --> 00:33:17,600
Speaker 3: out why. If the risk premium is most of the

678
00:33:17,640 --> 00:33:19,880
Speaker 3: time this is fine, but it has depression risk. It's

679
00:33:19,880 --> 00:33:22,880
Speaker 3: going to do particularly bad in a depression. Well, that's

680
00:33:22,920 --> 00:33:25,320
Speaker 3: not a very pleasant risk to have. You may get

681
00:33:25,320 --> 00:33:30,520
Speaker 3: paid for doing that. The positive of behavioral is it's

682
00:33:30,640 --> 00:33:33,600
Speaker 3: essentially over the long term. Over the short term it

683
00:33:33,600 --> 00:33:36,280
Speaker 3: can be very very painful lunch because the noise work.

684
00:33:36,680 --> 00:33:38,400
Speaker 3: But if on average it works long term, it's a

685
00:33:38,440 --> 00:33:42,040
Speaker 3: free lunch and that you're not taking additional systematic risk.

686
00:33:42,600 --> 00:33:44,600
Speaker 3: The negative is it can go away. It can be

687
00:33:44,680 --> 00:33:47,000
Speaker 3: arbitraged away if that ra stops being made, if too

688
00:33:47,080 --> 00:33:50,000
Speaker 3: much capital chases it, it can be arbitraged away. It's

689
00:33:50,040 --> 00:33:52,800
Speaker 3: hard to arbitrage some of them away, like basic valuation.

690
00:33:52,960 --> 00:33:56,200
Speaker 3: It's very easy to arbitrage some others, like alternative data,

691
00:33:56,240 --> 00:33:58,520
Speaker 3: where the point is to be quicker to getting a

692
00:33:58,600 --> 00:34:01,160
Speaker 3: data set and to trading the day to set. I

693
00:34:01,200 --> 00:34:03,800
Speaker 3: will say this, and I hope teen Fama isn't listening.

694
00:34:04,480 --> 00:34:07,440
Speaker 3: I was probably seventy five twenty five. For risk guy

695
00:34:07,720 --> 00:34:10,600
Speaker 3: thirty years ago, I'm probably seventy five twenty five.

696
00:34:10,640 --> 00:34:13,479
Speaker 1: A behavioral is that GameStop or something not.

697
00:34:13,840 --> 00:34:18,719
Speaker 3: Well, the game stop maybe the unfairly extreme example. It's

698
00:34:18,760 --> 00:34:21,239
Speaker 3: not fair to pick the most extreme craziness is to

699
00:34:21,280 --> 00:34:25,200
Speaker 3: make your point, But yeah, it's real life experience. It's

700
00:34:25,239 --> 00:34:29,120
Speaker 3: watching the spread between cheap and expensive stocks in ninety

701
00:34:29,200 --> 00:34:31,920
Speaker 3: nine two thousand. We started our firm in late ninety eight,

702
00:34:32,040 --> 00:34:35,000
Speaker 3: so right before the real crazy part of what's called

703
00:34:35,040 --> 00:34:38,200
Speaker 3: now the dot com bubble, the spread between how we

704
00:34:38,239 --> 00:34:42,279
Speaker 3: define cheap and expensive, either just using quant multiples or

705
00:34:42,320 --> 00:34:46,520
Speaker 3: adjusting for forms of growth. Set records. We had fifty

706
00:34:46,560 --> 00:34:49,040
Speaker 3: seventy five years of data and we lived through it,

707
00:34:49,120 --> 00:34:53,279
Speaker 3: blowing past those. Then after that, I'm like, all right,

708
00:34:53,320 --> 00:34:55,480
Speaker 3: that was a once in fifty year event. We survived it.

709
00:34:55,560 --> 00:34:57,399
Speaker 3: We ended up being right. We ended up making money

710
00:34:57,480 --> 00:35:01,719
Speaker 3: round trip that was excruciating on the first leg. If

711
00:35:01,760 --> 00:35:03,400
Speaker 3: you ask me back then, am I ever going to

712
00:35:03,440 --> 00:35:06,320
Speaker 3: see that in my career again? I hope I'd be

713
00:35:06,360 --> 00:35:09,160
Speaker 3: smart enough not to say never. None of us in

714
00:35:09,200 --> 00:35:12,560
Speaker 3: our field, yours or mine, should say never. Markets are

715
00:35:12,600 --> 00:35:15,080
Speaker 3: pretty hard things to say never about, but I think

716
00:35:15,120 --> 00:35:17,400
Speaker 3: I would have said it's highly unlikely. For one, it

717
00:35:17,480 --> 00:35:21,320
Speaker 3: was literally the most extreme event in at least fifty years. Second,

718
00:35:21,960 --> 00:35:24,879
Speaker 3: the question presupposes I and people like me will still

719
00:35:24,880 --> 00:35:28,279
Speaker 3: be around. Unpresumably if we're still around, we're in more

720
00:35:28,440 --> 00:35:32,320
Speaker 3: supervisory and authoritative positions. So how's it going to happen again?

721
00:35:33,200 --> 00:35:36,759
Speaker 3: And then it happened again almost exactly twenty years later,

722
00:35:36,840 --> 00:35:40,360
Speaker 3: from twenty eighteen through twenty twenty, you can point to COVID.

723
00:35:40,960 --> 00:35:43,600
Speaker 3: It went absolutely mad during COVID. You guys remember for

724
00:35:43,640 --> 00:35:46,480
Speaker 3: the six month period after COVID when the only two

725
00:35:46,520 --> 00:35:49,239
Speaker 3: stocks you're supposed to own were Peloton and Tesla. One

726
00:35:49,239 --> 00:35:51,080
Speaker 3: of them has still worked out mostly for you. One

727
00:35:51,120 --> 00:35:53,840
Speaker 3: of them did not. But even before COVID, though you

728
00:35:53,880 --> 00:35:56,880
Speaker 3: cannot pin it on COVID, by late nineteen early twenty,

729
00:35:57,080 --> 00:35:59,440
Speaker 3: we were approaching tech bubble levels, which again had not

730
00:35:59,480 --> 00:36:03,200
Speaker 3: been seeing the prior fifty years before. The tech bubble again,

731
00:36:03,400 --> 00:36:07,040
Speaker 3: painful period for us, again made money round trip. I'm

732
00:36:07,040 --> 00:36:09,440
Speaker 3: going to brag about that we survived it. I think

733
00:36:09,440 --> 00:36:12,760
Speaker 3: we're stronger than we were beforehand. But if the first

734
00:36:12,760 --> 00:36:16,280
Speaker 3: one didn't make you start to go this behavioral stuff

735
00:36:16,440 --> 00:36:20,120
Speaker 3: may be real and maybe bigger than it used to be. Yeah,

736
00:36:20,120 --> 00:36:22,080
Speaker 3: and I wrote a piece called the less efficient market

737
00:36:22,160 --> 00:36:26,120
Speaker 3: hypothesis markets getting less efficient. That may be true, but

738
00:36:26,160 --> 00:36:28,680
Speaker 3: I think it's more accurate to say there what my

739
00:36:29,000 --> 00:36:33,080
Speaker 3: evidence is there prone to some extreme bouts of craziness

740
00:36:33,480 --> 00:36:36,200
Speaker 3: on occasion. It's not quite the same as in steady

741
00:36:36,200 --> 00:36:38,960
Speaker 3: state always being less efficient, but probably some of both.

742
00:36:39,239 --> 00:36:41,319
Speaker 3: But I do think that has happened. I probably was

743
00:36:41,360 --> 00:36:44,560
Speaker 3: not giving behavioral enough credit early on, and I think

744
00:36:44,600 --> 00:36:53,280
Speaker 3: it's gotten to be a bigger part.

745
00:37:02,160 --> 00:37:04,319
Speaker 1: I want to read something from the Johnlely paper from

746
00:37:04,360 --> 00:37:07,839
Speaker 1: twenty fourteen. Suppose you imagine some investors get joy from

747
00:37:07,880 --> 00:37:10,359
Speaker 1: owning particular stocks, for example, being able to brag out

748
00:37:10,360 --> 00:37:12,200
Speaker 1: a cocktail party about the growth stocks that they own

749
00:37:12,239 --> 00:37:14,480
Speaker 1: that have done well. One way to describe this some

750
00:37:14,560 --> 00:37:17,239
Speaker 1: investors have a taste for growth stocks beyond simply their

751
00:37:17,280 --> 00:37:19,879
Speaker 1: effect on their portfolios. It can be rational for them

752
00:37:19,880 --> 00:37:22,960
Speaker 1: to accept somewhat lower returns for this pleasure. But even

753
00:37:23,000 --> 00:37:25,560
Speaker 1: if rational to the individuals who have this taste, if

754
00:37:25,560 --> 00:37:27,600
Speaker 1: some investors are willing to give up returns to others

755
00:37:27,600 --> 00:37:30,279
Speaker 1: because they care about cocktail party bragging, can we really

756
00:37:30,280 --> 00:37:32,880
Speaker 1: call that a rational market and feel this statement is useful.

757
00:37:33,080 --> 00:37:35,400
Speaker 1: I feel like that was like a little prescient about

758
00:37:35,600 --> 00:37:37,880
Speaker 1: how I have experienced some of the last five years,

759
00:37:38,080 --> 00:37:41,120
Speaker 1: which is that, like it seems to me that it

760
00:37:41,239 --> 00:37:43,239
Speaker 1: is hard to explain some of the stuff I write about,

761
00:37:43,280 --> 00:37:44,880
Speaker 1: which is maybe not like the most important thing in

762
00:37:44,880 --> 00:37:47,799
Speaker 1: the world. It's often like I think the most important too,

763
00:37:47,800 --> 00:37:49,960
Speaker 1: but it's like, you know, it's maybe not like the biggest,

764
00:37:50,120 --> 00:37:53,000
Speaker 1: you know, dollar, the test LIS's pretty big. It does

765
00:37:53,040 --> 00:37:56,600
Speaker 1: seem like a thing that is regularly happening is that

766
00:37:56,640 --> 00:38:01,719
Speaker 1: people have a taste for stocks beyond any rational or

767
00:38:01,760 --> 00:38:04,160
Speaker 1: irrational calculation about how we'll fight their portfolio.

768
00:38:04,280 --> 00:38:07,840
Speaker 3: Particularly the whole meme world. Meme coins are clearly a

769
00:38:08,040 --> 00:38:09,680
Speaker 3: political statement slash.

770
00:38:11,000 --> 00:38:13,719
Speaker 1: Clearly there is a non economic motivation for some of that,

771
00:38:13,760 --> 00:38:14,600
Speaker 1: and you can callude rational.

772
00:38:14,640 --> 00:38:17,839
Speaker 3: But it's odd, well a few things that paragraph. Even

773
00:38:17,880 --> 00:38:19,759
Speaker 3: writing it, I think we went a little too far

774
00:38:20,400 --> 00:38:22,959
Speaker 3: because I don't think these people, we kind of write

775
00:38:22,960 --> 00:38:25,120
Speaker 3: it like they're consciously doing it, like I just love

776
00:38:25,160 --> 00:38:27,040
Speaker 3: owning these I know I'm gonna lose money. I think

777
00:38:27,040 --> 00:38:30,680
Speaker 3: they kind of end up they resolve cognitive dissonance by

778
00:38:30,680 --> 00:38:33,600
Speaker 3: saying I'm gonna make money even if that's what's really

779
00:38:33,640 --> 00:38:34,000
Speaker 3: going on.

780
00:38:34,360 --> 00:38:36,000
Speaker 1: Right, I think this is maybe a little harsh on

781
00:38:36,040 --> 00:38:38,840
Speaker 1: growth investors in twenty fourteen, but on MEME investors.

782
00:38:38,440 --> 00:38:42,600
Speaker 3: And now instead on meme investors. We were also given

783
00:38:42,640 --> 00:38:46,279
Speaker 3: a little mild shot to some academics who kind of

784
00:38:46,280 --> 00:38:48,759
Speaker 3: try to save market efficiency by saying it's more meta

785
00:38:48,800 --> 00:38:51,279
Speaker 3: efficient if you count people just have tastes for this.

786
00:38:51,360 --> 00:38:53,040
Speaker 3: That argument's been made. I think that's kind of a

787
00:38:53,040 --> 00:38:55,600
Speaker 3: cop out. You know, if we've done market efficiency down

788
00:38:55,600 --> 00:38:58,360
Speaker 3: to that, what do we actually even mean by market efficiency?

789
00:38:58,400 --> 00:39:00,359
Speaker 3: If you go, I know, I'm gonna do terrible on this,

790
00:39:00,480 --> 00:39:04,880
Speaker 3: but it's fun to me, that's functionally a fairly inefficient market.

791
00:39:05,640 --> 00:39:09,160
Speaker 3: So we proposed in that piece which did not catch

792
00:39:09,520 --> 00:39:13,840
Speaker 3: on that classically market efficiency and the testing of it

793
00:39:13,880 --> 00:39:18,200
Speaker 3: has a joint hypothesis problem. You're testing whether markets are efficient,

794
00:39:18,560 --> 00:39:20,600
Speaker 3: but you also need a model for how prices are set,

795
00:39:20,800 --> 00:39:23,759
Speaker 3: and if that model for how prices are set includes

796
00:39:24,080 --> 00:39:26,759
Speaker 3: this is fun to own, then you've pushed it too far.

797
00:39:27,080 --> 00:39:29,520
Speaker 3: It has to be a reasonable hypothesiz.

798
00:39:29,200 --> 00:39:30,719
Speaker 1: This is what I read about this is I want

799
00:39:30,800 --> 00:39:33,040
Speaker 1: someone and it's probably not me. To write a finance

800
00:39:33,080 --> 00:39:36,920
Speaker 1: textbook that incorporates the factor of this is fun to

801
00:39:36,920 --> 00:39:39,680
Speaker 1: own and gives some guidance how to price that.

802
00:39:39,960 --> 00:39:43,880
Speaker 3: Be and how long it's going to continue.

803
00:39:43,960 --> 00:39:47,719
Speaker 1: And people like someone is making money on this, oh yeah,

804
00:39:47,719 --> 00:39:49,680
Speaker 1: and like not just like fading it. Someone is like,

805
00:39:50,120 --> 00:39:52,759
Speaker 1: I have a model for which meme coins will go up,

806
00:39:53,239 --> 00:39:54,000
Speaker 1: And well.

807
00:39:53,880 --> 00:39:55,040
Speaker 2: People have tried to do that.

808
00:39:55,120 --> 00:39:58,359
Speaker 4: You think about all the different buzzy strategies, something about

809
00:39:58,360 --> 00:40:02,000
Speaker 4: ETFs real cool, you can do it. Yeah, but you

810
00:40:02,000 --> 00:40:04,680
Speaker 4: know there have been attempts if you scraped social media,

811
00:40:04,760 --> 00:40:06,840
Speaker 4: et cetera and find out what people are buzzing about.

812
00:40:06,920 --> 00:40:09,960
Speaker 3: But it's the old value manager's lament. But it's also

813
00:40:10,160 --> 00:40:14,320
Speaker 3: I think true if ultimately it's a bad investment in

814
00:40:14,640 --> 00:40:18,160
Speaker 3: return sense that will happen eventually. I do think that

815
00:40:18,320 --> 00:40:21,600
Speaker 3: time horizon and the extremes have lengthened.

816
00:40:21,320 --> 00:40:22,520
Speaker 1: A lot of the point of this way, Why will

817
00:40:22,560 --> 00:40:23,320
Speaker 1: it happen eventually?

818
00:40:23,400 --> 00:40:26,040
Speaker 3: Why will it happen eventually? Well, if you buy a

819
00:40:26,160 --> 00:40:30,160
Speaker 3: company that continues to perform poorly and you paid a

820
00:40:30,200 --> 00:40:32,680
Speaker 3: ton for it, I think there's only so long that

821
00:40:32,680 --> 00:40:34,960
Speaker 3: can go on. I think it gets more and more obvious.

822
00:40:35,000 --> 00:40:35,520
Speaker 3: For one thing.

823
00:40:35,960 --> 00:40:38,160
Speaker 1: I think a lesson of the meme stack and frankly,

824
00:40:38,200 --> 00:40:42,320
Speaker 1: if crypto is that fundamental setay flora and valuation. But

825
00:40:43,040 --> 00:40:45,160
Speaker 1: you know, you know he's like doing LBI, right. But

826
00:40:45,440 --> 00:40:46,960
Speaker 1: I don't want to just be mean to bitcoin for

827
00:40:47,040 --> 00:40:50,160
Speaker 1: no reason. But like, you can certainly have a model

828
00:40:50,160 --> 00:40:52,280
Speaker 1: in which you say the fundamentals of bitcoin are nothing,

829
00:40:52,520 --> 00:40:55,319
Speaker 1: and you could then say and in fifty years it'll

830
00:40:55,320 --> 00:40:57,480
Speaker 1: go to zero, but like that's a weird thing to say. Now,

831
00:40:57,760 --> 00:40:58,120
Speaker 1: I don't know.

832
00:40:58,200 --> 00:41:02,080
Speaker 3: Well, let me be clear. I one short Bitcoin with

833
00:41:02,160 --> 00:41:05,239
Speaker 3: my worst enemies portfolio, and I one hundred percent sure

834
00:41:05,280 --> 00:41:07,280
Speaker 3: I'll never say money is a little bit of magic.

835
00:41:07,320 --> 00:41:09,319
Speaker 3: What becomes money is a little bit of magic. But

836
00:41:09,400 --> 00:41:11,880
Speaker 3: I think most of the probability is eventually zero, but

837
00:41:11,920 --> 00:41:13,880
Speaker 3: it may be a very long time arizon. I think

838
00:41:13,960 --> 00:41:17,240
Speaker 3: what we've learned watching this stuff is that time horizon

839
00:41:17,320 --> 00:41:17,800
Speaker 3: is lengthened.

840
00:41:17,880 --> 00:41:18,160
Speaker 1: Okay.

841
00:41:19,840 --> 00:41:23,600
Speaker 3: In my piece on this, I try to hypothesize, and

842
00:41:23,760 --> 00:41:28,359
Speaker 3: I admit it's real opinionated hypothesizing. You cannot prove if

843
00:41:28,400 --> 00:41:32,600
Speaker 3: I'm right that markets are somewhat less efficient. Why gets

844
00:41:32,680 --> 00:41:35,719
Speaker 3: even harder. But one of my favorite explanations is an

845
00:41:35,760 --> 00:41:40,440
Speaker 3: old man complaining about social media and twenty four to

846
00:41:40,440 --> 00:41:43,719
Speaker 3: seven gamified trading. I don't think most people need a

847
00:41:43,719 --> 00:41:45,719
Speaker 3: lot of convincing that this stuff has made say, our

848
00:41:45,760 --> 00:41:48,839
Speaker 3: politics worse, made us more in bubbles hate each other

849
00:41:48,880 --> 00:41:50,319
Speaker 3: more when it was supposed to make us love each

850
00:41:50,320 --> 00:41:53,200
Speaker 3: other more. Marcus are just voting mechanisms. I don't think

851
00:41:53,239 --> 00:41:57,040
Speaker 3: it's any different than politics. So this notion that things

852
00:41:57,120 --> 00:41:59,440
Speaker 3: get crazier and can go on for longer, and I

853
00:41:59,480 --> 00:42:02,840
Speaker 3: have a very cynical view of your statement that this

854
00:42:02,840 --> 00:42:08,080
Speaker 3: stuff might take fifty years. I think the more absolutely

855
00:42:08,160 --> 00:42:12,520
Speaker 3: unsubstantiated by anything something is the longer the craziness can

856
00:42:12,560 --> 00:42:12,759
Speaker 3: go on.

857
00:42:12,960 --> 00:42:15,120
Speaker 1: Right. I think that if you compare the longevity of

858
00:42:15,160 --> 00:42:17,239
Speaker 1: bitkind to like the game Stop premium, I think right.

859
00:42:17,440 --> 00:42:20,920
Speaker 3: But go back to the tech bubble. Cisco Systems great company,

860
00:42:20,920 --> 00:42:23,600
Speaker 3: but was selling at a very stupid price, selling it about

861
00:42:23,560 --> 00:42:26,399
Speaker 3: one hundred PE when the e was gigantic. You can

862
00:42:26,440 --> 00:42:29,680
Speaker 3: have a small startup company it's growing super rapidly, sell

863
00:42:29,680 --> 00:42:32,319
Speaker 3: at one hundred pee. One of the largest companies in

864
00:42:32,320 --> 00:42:34,120
Speaker 3: the world with some of the largest earnings in the

865
00:42:34,120 --> 00:42:36,520
Speaker 3: world selling at one hundred pe. You needed some very

866
00:42:36,560 --> 00:42:39,960
Speaker 3: heroic and I would argue near I'll never say impossible,

867
00:42:40,000 --> 00:42:44,720
Speaker 3: but near impossible assumptions even if the tech bubble didn't

868
00:42:44,760 --> 00:42:47,279
Speaker 3: break in March of two thousand when it did, and

869
00:42:47,320 --> 00:42:48,719
Speaker 3: by the way, I still don't know why it broke

870
00:42:48,719 --> 00:42:50,960
Speaker 3: in March of two thousand and not a year earlier

871
00:42:51,040 --> 00:42:53,440
Speaker 3: or a year later. Once you're well passed what I

872
00:42:53,480 --> 00:42:57,520
Speaker 3: would consider rational saying you know exactly where but the

873
00:42:57,600 --> 00:43:00,120
Speaker 3: time horizon, you can imagine that going on for when

874
00:43:00,160 --> 00:43:03,239
Speaker 3: the growth is good, maybe even great, but not nearly

875
00:43:03,320 --> 00:43:06,480
Speaker 3: enough to justify that price. It gets more and more

876
00:43:06,480 --> 00:43:11,360
Speaker 3: obvious if something is based completely on air. One of

877
00:43:11,400 --> 00:43:13,600
Speaker 3: the weird of down Matt's point that there's there's no

878
00:43:13,760 --> 00:43:17,560
Speaker 3: arm to the upside, there's no LBO mechanism, and shorting

879
00:43:17,600 --> 00:43:20,319
Speaker 3: it is just frankly too dangerous. It's a weird way

880
00:43:20,360 --> 00:43:22,560
Speaker 3: to say. Sometimes people say markets are efficient because you

881
00:43:22,600 --> 00:43:25,359
Speaker 3: can't make money from these things, and I'm like, it's

882
00:43:25,360 --> 00:43:28,239
Speaker 3: another weird way to defend efficient markets to go they

883
00:43:28,239 --> 00:43:32,680
Speaker 3: can be so friggin stupid that they're terrifying to make efficient.

884
00:43:33,239 --> 00:43:36,279
Speaker 3: That may be totally true, but it also is a

885
00:43:36,280 --> 00:43:38,120
Speaker 3: weird way to argue that markets are efficient.

886
00:43:38,280 --> 00:43:41,600
Speaker 4: What does it mean for AQR if markets are less efficient.

887
00:43:41,200 --> 00:43:45,560
Speaker 3: Well, any active management is an inherently arrogant act. You

888
00:43:45,600 --> 00:43:48,759
Speaker 3: cannot tell the average person we should all be active managers,

889
00:43:49,239 --> 00:43:50,560
Speaker 3: we should all have podcasts.

890
00:43:51,120 --> 00:43:52,959
Speaker 2: But I agree, So it's.

891
00:43:54,520 --> 00:44:00,000
Speaker 3: It's top to believe you should be an active manager

892
00:44:00,280 --> 00:44:02,080
Speaker 3: and to believe you're doing something good for your clients,

893
00:44:02,080 --> 00:44:04,880
Speaker 3: you have to believe two things. That you have alpha

894
00:44:05,239 --> 00:44:07,400
Speaker 3: and that you're not charging the full extent of that

895
00:44:07,440 --> 00:44:10,440
Speaker 3: alpha through your fees. And we do believe that, and

896
00:44:10,480 --> 00:44:13,480
Speaker 3: a lot of active management managers believe it. But it's

897
00:44:13,520 --> 00:44:17,160
Speaker 3: an inherently arrogant act. It's consistent to say most people

898
00:44:17,200 --> 00:44:20,920
Speaker 3: shouldn't do this, but we should, though the arrogance is obvious.

899
00:44:21,200 --> 00:44:23,600
Speaker 1: What percent of your alpha should in your face?

900
00:44:24,239 --> 00:44:27,200
Speaker 3: That is a super hard question. I've thought about writing

901
00:44:27,239 --> 00:44:30,359
Speaker 3: about this at one point. It kind of depends on

902
00:44:30,400 --> 00:44:31,759
Speaker 3: how unique your alpha.

903
00:44:31,520 --> 00:44:34,800
Speaker 1: Is, Okay, because I would think that if you got

904
00:44:34,920 --> 00:44:40,799
Speaker 1: a podshot manager really drunk, they would say, no.

905
00:44:40,920 --> 00:44:43,040
Speaker 3: That's what they can charge for their fees, not what

906
00:44:43,080 --> 00:44:45,400
Speaker 3: they should charge. I don't think they'd ever admit them

907
00:44:45,440 --> 00:44:49,640
Speaker 3: an algie now, like will versus shall.

908
00:44:49,640 --> 00:44:51,880
Speaker 1: Deep in their soul, they want one hundred and ten percent.

909
00:44:52,239 --> 00:44:54,919
Speaker 3: It's a great question, but it's a hard question if

910
00:44:54,960 --> 00:44:57,680
Speaker 3: your alpha is doing FOM and French price to book.

911
00:44:58,120 --> 00:44:59,600
Speaker 3: By the way, I still think Farm and French are

912
00:44:59,600 --> 00:45:00,960
Speaker 3: going to be right in the next thirty years. They

913
00:45:00,960 --> 00:45:03,280
Speaker 3: haven't been right in a while, but spreads have gotten

914
00:45:03,320 --> 00:45:06,280
Speaker 3: wider and wider, and that's been a wind in their face.

915
00:45:06,320 --> 00:45:08,319
Speaker 3: I wrote a piece on this saying the long run

916
00:45:08,400 --> 00:45:11,400
Speaker 3: is lying to you, saying that x the spread widening

917
00:45:11,480 --> 00:45:15,280
Speaker 3: values to the simple Farm of French value has delivered alpha.

918
00:45:15,600 --> 00:45:18,600
Speaker 3: It's just lost on the repricing. But what you can

919
00:45:18,640 --> 00:45:21,839
Speaker 3: actually charge for doing price to book should be very low,

920
00:45:21,920 --> 00:45:26,479
Speaker 3: a very small fraction of the expected return. That's alpha data, right,

921
00:45:26,560 --> 00:45:28,560
Speaker 3: But that's kind of what we're saying. Everything exists on

922
00:45:28,600 --> 00:45:31,000
Speaker 3: a spectrum from one to the other. If you have

923
00:45:31,160 --> 00:45:35,440
Speaker 3: discovered and built the database yourself, an alternative data source

924
00:45:35,480 --> 00:45:38,719
Speaker 3: that you have built. This is extreme and I can't

925
00:45:38,760 --> 00:45:41,120
Speaker 3: think of an example at AQR that fits this. But

926
00:45:41,400 --> 00:45:43,520
Speaker 3: you can charge nearly one hundred percent of that alpha

927
00:45:43,960 --> 00:45:47,520
Speaker 3: because what they're getting is still absolutely unrelated to everything

928
00:45:47,520 --> 00:45:50,279
Speaker 3: else what they're doing. There's some deminimus notion that if

929
00:45:50,320 --> 00:45:53,360
Speaker 3: you charge ninety nine percent, maybe no one would bother

930
00:45:53,880 --> 00:45:55,920
Speaker 3: to do it. But you can charge a large fraction

931
00:45:56,000 --> 00:45:58,480
Speaker 3: of the alpha because what you're delivering is still ultimately

932
00:45:58,840 --> 00:46:02,160
Speaker 3: net returns that you can not get elsewhere that aren't

933
00:46:02,160 --> 00:46:05,440
Speaker 3: correlated to the rest of the portfolio are worth it.

934
00:46:06,040 --> 00:46:08,440
Speaker 1: We can talk abstractly about what percentage of your alpha

935
00:46:08,440 --> 00:46:10,000
Speaker 1: you should charge, but like, no one could send a

936
00:46:10,040 --> 00:46:11,640
Speaker 1: bill for like ninety percent on the alpha, right, Like

937
00:46:12,320 --> 00:46:17,320
Speaker 1: is pricing sort of like set by just like anchored norms.

938
00:46:17,440 --> 00:46:19,879
Speaker 3: Anchored norms is another way of saying, what are other

939
00:46:19,960 --> 00:46:22,160
Speaker 3: people charge or in the ballpark of what you're doing?

940
00:46:22,280 --> 00:46:25,080
Speaker 3: So of course that matters, right if you are way

941
00:46:25,120 --> 00:46:27,640
Speaker 3: off the anchor, way off on the high side, no

942
00:46:27,640 --> 00:46:29,799
Speaker 3: one should invest with you. If you're way off on

943
00:46:29,800 --> 00:46:30,399
Speaker 3: the low side.

944
00:46:30,480 --> 00:46:34,040
Speaker 1: People who charge really high fees and are pretty good

945
00:46:34,040 --> 00:46:36,480
Speaker 1: at demonstrating they have alpha, and oh.

946
00:46:36,400 --> 00:46:39,360
Speaker 3: Some of the block shops charge insane fees and have

947
00:46:39,440 --> 00:46:42,160
Speaker 3: been very good, And that goes to that are they

948
00:46:42,160 --> 00:46:44,680
Speaker 3: doing something unique? And I think to some extent they are.

949
00:46:45,160 --> 00:46:47,680
Speaker 3: I think their problem becomes kind of in the direction

950
00:46:47,800 --> 00:46:49,719
Speaker 3: of Medallion without going all the way. I don't think

951
00:46:49,719 --> 00:46:53,680
Speaker 3: they've rebuilt medallion. Nothing is that, but they should charge

952
00:46:53,680 --> 00:46:56,200
Speaker 3: a higher percent if they're doing something very unique, and

953
00:46:56,239 --> 00:46:59,239
Speaker 3: they do. I have a very you can you know

954
00:46:59,400 --> 00:47:02,719
Speaker 3: violin playing I think rosy view of how we think

955
00:47:02,760 --> 00:47:06,319
Speaker 3: about fees. We're building a long term business. We think

956
00:47:06,320 --> 00:47:08,279
Speaker 3: we have business value. We do not think we're just

957
00:47:08,360 --> 00:47:11,359
Speaker 3: a hedge fund. We run a lot of traditional assets too,

958
00:47:11,480 --> 00:47:15,359
Speaker 3: even the hedge funds. We were a big pioneer in

959
00:47:15,719 --> 00:47:19,160
Speaker 3: doing the more obvious strategies and considerably lower fees, starting

960
00:47:19,160 --> 00:47:21,839
Speaker 3: in merger, ARB and TREND. Following the way we broke

961
00:47:21,880 --> 00:47:24,560
Speaker 3: into some of those as standalone products, not as part

962
00:47:24,600 --> 00:47:29,160
Speaker 3: of our multistrats was charging less and saying, you know,

963
00:47:29,239 --> 00:47:31,960
Speaker 3: this is real and it's good. But you know you

964
00:47:31,960 --> 00:47:34,439
Speaker 3: can do one of every merger and make a fair

965
00:47:34,440 --> 00:47:36,840
Speaker 3: amount of money. But that's not magic, and we shouldn't

966
00:47:36,880 --> 00:47:40,400
Speaker 3: charge magic fees for it. But in that kind of kumbay,

967
00:47:40,480 --> 00:47:45,120
Speaker 3: a big picture sense, charging fees such that clients are

968
00:47:45,160 --> 00:47:48,160
Speaker 3: happy with the long term results is probably how you

969
00:47:48,200 --> 00:47:50,560
Speaker 3: build business value if you step.

970
00:47:50,320 --> 00:47:52,719
Speaker 4: Back from just AQR though, I'm curious to hear your

971
00:47:52,719 --> 00:47:56,799
Speaker 4: thoughts on your industry overall and whether alternative strategies in

972
00:47:56,840 --> 00:47:58,480
Speaker 4: general are too expensive.

973
00:47:59,040 --> 00:47:59,719
Speaker 3: Yes, they are.

974
00:48:00,640 --> 00:48:03,120
Speaker 4: I have some numbers to back up that statement. There's

975
00:48:03,160 --> 00:48:05,400
Speaker 4: a new study out there. My understanding of it is

976
00:48:05,400 --> 00:48:08,319
Speaker 4: that you basically take a sixty to forty since two

977
00:48:08,400 --> 00:48:11,000
Speaker 4: thousand and eight, you add Alt's exposure to it in

978
00:48:11,080 --> 00:48:16,440
Speaker 4: various proportions, and that blended portfolio basically trails that benchmark,

979
00:48:16,760 --> 00:48:19,560
Speaker 4: basically in close proportions to the fees that are charged

980
00:48:20,040 --> 00:48:21,799
Speaker 4: by some of the ALTS managers.

981
00:48:21,960 --> 00:48:23,000
Speaker 2: And I don't know.

982
00:48:23,080 --> 00:48:25,040
Speaker 4: I read that and I was just kind of wondering.

983
00:48:25,080 --> 00:48:26,960
Speaker 4: I mean, you could make the case that why are

984
00:48:27,040 --> 00:48:27,839
Speaker 4: we charging so much?

985
00:48:27,840 --> 00:48:32,839
Speaker 3: Well, this is almost a mathematical certainty. Yeah, forget about

986
00:48:32,880 --> 00:48:36,160
Speaker 3: ALTS for a second. You cannot tell me the average

987
00:48:36,719 --> 00:48:42,719
Speaker 3: active portfolio beats the market after fees and costs. The

988
00:48:42,800 --> 00:48:45,879
Speaker 3: average active adds up to the market because for every

989
00:48:45,920 --> 00:48:48,759
Speaker 3: deviation one way, there's deviation the other way. When I

990
00:48:48,800 --> 00:48:50,840
Speaker 3: said it to inherently an arrogant act to be an

991
00:48:50,880 --> 00:48:54,000
Speaker 3: active manager, it means you think you got it, even

992
00:48:54,040 --> 00:48:58,440
Speaker 3: though if you buy one of each you can't have it.

993
00:48:58,520 --> 00:49:01,400
Speaker 3: I mean a subset of the market like could. But

994
00:49:01,480 --> 00:49:04,560
Speaker 3: I think a lot of that still applies. It's inherently

995
00:49:04,640 --> 00:49:07,240
Speaker 3: arrogant act. But we've been saying this for at least

996
00:49:07,320 --> 00:49:09,799
Speaker 3: twenty four years. We wrote a paper, and yes, I

997
00:49:09,800 --> 00:49:11,800
Speaker 3: started a lot of sentences with We wrote a paper

998
00:49:12,080 --> 00:49:14,560
Speaker 3: in two thousand and one called do hedge funds Hedge

999
00:49:15,200 --> 00:49:18,160
Speaker 3: where we took the known indices of hedge funds and

1000
00:49:18,760 --> 00:49:20,840
Speaker 3: we tried to take out just the market beta. You

1001
00:49:20,880 --> 00:49:23,319
Speaker 3: could argue for a more sophisticated risk model, and that could,

1002
00:49:23,719 --> 00:49:27,000
Speaker 3: as usual, go down that rabbit hole. But we found

1003
00:49:27,040 --> 00:49:29,280
Speaker 3: that betas were First of all, this is more mundane

1004
00:49:29,360 --> 00:49:33,759
Speaker 3: but statistically underestimated, partly because a lot of hedge funds

1005
00:49:33,800 --> 00:49:37,000
Speaker 3: do some stuff that is not of perfect liquidity. And

1006
00:49:37,160 --> 00:49:41,040
Speaker 3: this was a small early version of what I got

1007
00:49:41,120 --> 00:49:43,479
Speaker 3: into at the private world later on. But if something

1008
00:49:43,520 --> 00:49:45,960
Speaker 3: doesn't trade all the time and you try to estimate

1009
00:49:46,000 --> 00:49:49,880
Speaker 3: its correlation with the market, you will underestimate it because

1010
00:49:50,040 --> 00:49:52,600
Speaker 3: one great way to look uncorrelated is to have a

1011
00:49:52,640 --> 00:49:54,920
Speaker 3: three day leg, and when you trade, your returns will

1012
00:49:54,920 --> 00:49:58,040
Speaker 3: be off. So the betas were underestimated by earlier studies.

1013
00:49:58,120 --> 00:50:00,760
Speaker 3: Given the correct betas, there was pretty much no alpha

1014
00:50:01,280 --> 00:50:03,279
Speaker 3: to the hedge fund world. First of all, I was

1015
00:50:03,280 --> 00:50:05,120
Speaker 3: a lot younger than a lot less well known, so

1016
00:50:05,200 --> 00:50:09,480
Speaker 3: I cared. Nowadays, I quite obviously court controversy. But back

1017
00:50:09,520 --> 00:50:12,000
Speaker 3: then I probably had ten famous managers call me and

1018
00:50:12,080 --> 00:50:15,120
Speaker 3: yell at me about that paper. Yeah. The first time,

1019
00:50:15,160 --> 00:50:17,239
Speaker 3: I was, of course obnoxious. They called and said, why

1020
00:50:17,239 --> 00:50:20,719
Speaker 3: did you write this? And I gave the obnoxious answer

1021
00:50:20,840 --> 00:50:23,400
Speaker 3: because we think it's true. Now it was apparently not

1022
00:50:23,480 --> 00:50:26,840
Speaker 3: an acceptable answer. I will only call out one person

1023
00:50:26,880 --> 00:50:29,960
Speaker 3: on the positive side, Richard Perry, famous hedge fund manager.

1024
00:50:30,320 --> 00:50:32,040
Speaker 3: He called me up. And I already been yelled at

1025
00:50:32,040 --> 00:50:34,040
Speaker 3: by a whole bunch of people. And so when I

1026
00:50:34,080 --> 00:50:36,080
Speaker 3: heard Richard Perry on the phone, people like Richard Perry

1027
00:50:36,080 --> 00:50:38,760
Speaker 3: didn't call people like me. Back then. He was big,

1028
00:50:38,800 --> 00:50:40,440
Speaker 3: I was small. So I'm like, I know what this is.

1029
00:50:40,480 --> 00:50:42,640
Speaker 3: He's just gonna yell at me. He gets on the phone,

1030
00:50:42,640 --> 00:50:46,000
Speaker 3: he goes, that paper you wrote, that's just correct, good job, right,

1031
00:50:46,120 --> 00:50:48,040
Speaker 3: And I'm only telling that. I'm not giving the names.

1032
00:50:48,040 --> 00:50:50,960
Speaker 3: And I do remember him, of course, and in a

1033
00:50:51,000 --> 00:50:54,719
Speaker 3: book somewhere, it's all. I have a list of Vedettas

1034
00:50:54,880 --> 00:50:56,040
Speaker 3: in my brain pursuit.

1035
00:50:56,040 --> 00:50:59,000
Speaker 1: Wait, so you've been you've been making fun of private

1036
00:50:59,000 --> 00:51:01,959
Speaker 1: aquery managers on some lines recently. Do you get calls

1037
00:51:01,960 --> 00:51:02,600
Speaker 1: from them?

1038
00:51:02,880 --> 00:51:05,960
Speaker 3: No, they're so fat and happy. They don't even care

1039
00:51:06,040 --> 00:51:08,120
Speaker 3: about me making fun. No, I get yelled at by

1040
00:51:08,120 --> 00:51:11,759
Speaker 3: some usually friends. I live in Greenwich connection, so you

1041
00:51:11,840 --> 00:51:12,279
Speaker 3: all hang.

1042
00:51:12,239 --> 00:51:14,680
Speaker 1: Out the country club. They're like, I, I'm.

1043
00:51:14,520 --> 00:51:19,320
Speaker 3: Actually a country but the old often used in a

1044
00:51:19,400 --> 00:51:21,440
Speaker 3: very bad way. Some of my best friends are, so

1045
00:51:21,480 --> 00:51:23,880
Speaker 3: I'm off the hook. I have some good friends who

1046
00:51:23,920 --> 00:51:27,480
Speaker 3: are private equity managers. Most of them can accept the

1047
00:51:28,040 --> 00:51:30,640
Speaker 3: you're right about the industry but not our firm, which

1048
00:51:30,719 --> 00:51:33,520
Speaker 3: is essentially what I'm saying. So I can't not being

1049
00:51:34,160 --> 00:51:35,080
Speaker 3: mean about this.

1050
00:51:35,280 --> 00:51:38,480
Speaker 1: Right, You're not like your criticism is volatility, lunder Right,

1051
00:51:38,520 --> 00:51:41,440
Speaker 1: Your criticism is that private equity seems to have a

1052
00:51:41,480 --> 00:51:43,919
Speaker 1: higher sharp because it has a lower of volatility because

1053
00:51:43,960 --> 00:51:45,040
Speaker 1: it doesn't report.

1054
00:51:46,480 --> 00:51:49,440
Speaker 3: Yes, that is my main criticism, which I think is

1055
00:51:49,520 --> 00:51:52,440
Speaker 3: quite obviously true. That's just me. Some people do disagree.

1056
00:51:53,320 --> 00:51:55,880
Speaker 3: The best disagreement I've heard, and I have some sympathy

1057
00:51:55,880 --> 00:51:58,120
Speaker 3: for this because I do not think markets are perfect,

1058
00:51:58,520 --> 00:52:02,839
Speaker 3: is we are right about the valuations. You are right

1059
00:52:03,239 --> 00:52:05,800
Speaker 3: that we move at a highly damped version of the market,

1060
00:52:06,400 --> 00:52:09,920
Speaker 3: but market moves too much. We are right. My response

1061
00:52:09,960 --> 00:52:11,560
Speaker 3: to that is, why don't we get to do that?

1062
00:52:11,920 --> 00:52:13,759
Speaker 3: It's fair, you know, when we've had a tough year

1063
00:52:13,800 --> 00:52:16,080
Speaker 3: because the market's gone crazy and we were on the

1064
00:52:16,080 --> 00:52:18,640
Speaker 3: wrong side of that. I have to tell my clients

1065
00:52:18,719 --> 00:52:21,080
Speaker 3: we're down twelve percent. I don't get to tell my

1066
00:52:21,160 --> 00:52:24,680
Speaker 3: clients we're up based on where I'd mark the portfolio.

1067
00:52:24,760 --> 00:52:26,600
Speaker 3: So why one group? And by the way, they could

1068
00:52:26,640 --> 00:52:29,560
Speaker 3: market just like we market. They're brilliant at valuing companies

1069
00:52:29,600 --> 00:52:30,799
Speaker 3: and they can tell you where they could sell it

1070
00:52:30,840 --> 00:52:34,600
Speaker 3: for today. So it's like an institutional legal quirk that

1071
00:52:34,680 --> 00:52:36,160
Speaker 3: they get to do it one way and we have

1072
00:52:36,200 --> 00:52:39,480
Speaker 3: to do it another. Everyone can mark their portfolio at

1073
00:52:39,480 --> 00:52:42,640
Speaker 3: what they could sell it for today or what they

1074
00:52:42,680 --> 00:52:44,759
Speaker 3: think it's worth, and one side gets to do it

1075
00:52:44,760 --> 00:52:47,520
Speaker 3: one one gets to do the other. So that's a

1076
00:52:47,560 --> 00:52:49,759
Speaker 3: reasonable argument, but I still think it leads to an

1077
00:52:49,840 --> 00:52:54,160
Speaker 3: unreasonable conclusion. I will say ninety percent of my critique

1078
00:52:54,200 --> 00:52:57,040
Speaker 3: is about the volatility or the beta, is about saying

1079
00:52:57,080 --> 00:53:00,000
Speaker 3: these things are low risk. Ten percent is about percent

1080
00:53:00,360 --> 00:53:04,759
Speaker 3: if not trailing future returns, because if I'm right about

1081
00:53:04,760 --> 00:53:08,920
Speaker 3: the volatility laundering, it has implications for returns going forward.

1082
00:53:09,280 --> 00:53:11,680
Speaker 3: If when David Swenson was pioneering, which is what he

1083
00:53:11,719 --> 00:53:14,839
Speaker 3: called his book, Private Equity is part of an institutional

1084
00:53:14,960 --> 00:53:18,640
Speaker 3: endownment portfolio. He's quite clear. It's a lot of it.

1085
00:53:18,640 --> 00:53:21,520
Speaker 3: It's an ill liquidity premium that no one wants. Illiquidity.

1086
00:53:21,600 --> 00:53:24,000
Speaker 3: Everyone's scared of it. So if you're willing to do that,

1087
00:53:24,040 --> 00:53:27,479
Speaker 3: you get paid extra. If I'm right that people love

1088
00:53:28,840 --> 00:53:30,799
Speaker 3: the fact that they don't have to look at the volatility,

1089
00:53:31,280 --> 00:53:34,800
Speaker 3: that means illiquidity is no longer a bug. It is

1090
00:53:34,840 --> 00:53:38,280
Speaker 3: now a feature and very simple model for how expective

1091
00:53:38,320 --> 00:53:41,720
Speaker 3: returns are set. You get paid a higher expective return

1092
00:53:41,760 --> 00:53:43,480
Speaker 3: in something if you have to bear a bug nobody

1093
00:53:43,520 --> 00:53:47,440
Speaker 3: wants and you pay through a lower expector return. If

1094
00:53:47,520 --> 00:53:49,799
Speaker 3: you have a feature everyone wants, then you have to

1095
00:53:49,800 --> 00:53:52,560
Speaker 3: pay up for it. So the chance that that is

1096
00:53:52,560 --> 00:53:54,680
Speaker 3: going on going forward, I think is quite high. Whether

1097
00:53:54,719 --> 00:53:56,759
Speaker 3: it means there's no edge to private equity or a

1098
00:53:56,800 --> 00:54:00,560
Speaker 3: negative edge or a smaller positive edge can tell you that.

1099
00:54:00,760 --> 00:54:02,760
Speaker 3: But I do think if you think of risk ajusy

1100
00:54:02,840 --> 00:54:06,399
Speaker 3: return as numerator of return denominator of risk, I think

1101
00:54:06,520 --> 00:54:10,319
Speaker 3: my statements are mostly about the denominator, but they're ten

1102
00:54:10,400 --> 00:54:27,360
Speaker 3: percent now about the numerator. Going forward, you're.

1103
00:54:27,200 --> 00:54:30,040
Speaker 1: In my conservative private equity private markets by which is

1104
00:54:30,080 --> 00:54:33,080
Speaker 1: that it seems to me that like there is a

1105
00:54:33,160 --> 00:54:37,200
Speaker 1: ton of fee pressure in public markets and everyone has

1106
00:54:37,239 --> 00:54:39,359
Speaker 1: kind of like learned the gospel of like Bilow cost

1107
00:54:39,400 --> 00:54:42,920
Speaker 1: index funds and even charging for farma French factors is

1108
00:54:42,920 --> 00:54:48,120
Speaker 1: not a two and twenty business. And private markets because

1109
00:54:48,120 --> 00:54:51,120
Speaker 1: they're not indexible because it's harder to like extract factors

1110
00:54:51,200 --> 00:54:54,920
Speaker 1: because they're not liquid, don't have those problems, and you

1111
00:54:54,960 --> 00:54:56,800
Speaker 1: can charge two and twenty for a lot of private stuff.

1112
00:54:56,800 --> 00:54:58,840
Speaker 1: And so like it seems to me that there is

1113
00:54:58,880 --> 00:55:01,480
Speaker 1: a move to you put a lot of stuff that

1114
00:55:01,520 --> 00:55:04,440
Speaker 1: would have previously been public into private markets, and to

1115
00:55:04,560 --> 00:55:07,200
Speaker 1: say we can put privates into four oh one ks

1116
00:55:07,920 --> 00:55:10,360
Speaker 1: and there's a good economic grastionals we're putting private assets

1117
00:55:10,360 --> 00:55:12,680
Speaker 1: into four one case because you don't need liquidity. But

1118
00:55:12,719 --> 00:55:14,799
Speaker 1: there's also like this like really overwhelming if.

1119
00:55:14,680 --> 00:55:17,120
Speaker 3: There's a positive illiquidity premium.

1120
00:55:17,160 --> 00:55:20,560
Speaker 1: Yeah too right, sorry, go no. I just like you know,

1121
00:55:20,719 --> 00:55:24,480
Speaker 1: you've written for years ago like criticizing people for charging

1122
00:55:24,520 --> 00:55:26,560
Speaker 1: alpha fees for beta, and it seems to me that

1123
00:55:26,640 --> 00:55:30,279
Speaker 1: like the reprivatization on the market is a way to

1124
00:55:30,320 --> 00:55:31,720
Speaker 1: sneak some alpha fies onto beta.

1125
00:55:31,840 --> 00:55:34,520
Speaker 3: I think a tremendous amount of the private world is

1126
00:55:34,600 --> 00:55:37,800
Speaker 3: charging massive alpha fees for beta. I won't mince words

1127
00:55:38,080 --> 00:55:42,160
Speaker 3: about that. If they outperform or underperform on net after

1128
00:55:42,200 --> 00:55:44,680
Speaker 3: all these massive fees, and if performance going forward, it

1129
00:55:44,920 --> 00:55:47,000
Speaker 3: is tougher. And by the way, your point about the

1130
00:55:47,040 --> 00:55:50,480
Speaker 3: fee compression in the public world only makes my hypothesis

1131
00:55:50,520 --> 00:55:52,640
Speaker 3: that they won't beat the public world by the same

1132
00:55:52,680 --> 00:55:55,799
Speaker 3: amount going forward stronger. I think privates have a big

1133
00:55:55,840 --> 00:55:57,440
Speaker 3: function in the world. I don't think they're going away

1134
00:55:57,840 --> 00:55:59,319
Speaker 3: what you started out with me, what do you guys

1135
00:55:59,400 --> 00:56:01,000
Speaker 3: do for the world. There are things that are in

1136
00:56:01,040 --> 00:56:03,280
Speaker 3: between what should be public and what's mom and pop.

1137
00:56:03,800 --> 00:56:06,319
Speaker 3: But I think where we are now, I think a

1138
00:56:06,320 --> 00:56:11,680
Speaker 3: lot of institutions are giving up some amount of expected

1139
00:56:11,719 --> 00:56:15,920
Speaker 3: return for the ease of limit, of reducing their agency

1140
00:56:15,960 --> 00:56:18,320
Speaker 3: problem of sticking with something. Now, if they're going to

1141
00:56:18,400 --> 00:56:21,120
Speaker 3: be terrible and not stick with things, it might be

1142
00:56:21,239 --> 00:56:24,280
Speaker 3: rational to give up some return. But you can't double

1143
00:56:24,360 --> 00:56:27,719
Speaker 3: count and say we're making ourselves better investors giving up

1144
00:56:27,760 --> 00:56:30,359
Speaker 3: some expected return and oh yeah, our expected returns are going.

1145
00:56:30,280 --> 00:56:30,760
Speaker 1: To be higher.

1146
00:56:31,080 --> 00:56:34,680
Speaker 3: If that's the rationale, then you've accepted my argument, and

1147
00:56:34,719 --> 00:56:36,560
Speaker 3: I think you have to say, this is what we

1148
00:56:36,600 --> 00:56:38,480
Speaker 3: get paid for by making your life easier.

1149
00:56:39,120 --> 00:56:41,239
Speaker 4: I am curious what you make of the push to

1150
00:56:41,280 --> 00:56:44,719
Speaker 4: put privates into more retail accessible writers. I'm talking about

1151
00:56:44,719 --> 00:56:47,840
Speaker 4: et avs, but I guess I'm also talking about interval

1152
00:56:47,840 --> 00:56:49,000
Speaker 4: funds a little bit.

1153
00:56:49,239 --> 00:56:51,359
Speaker 2: It just seems like that's where the world is headed.

1154
00:56:51,440 --> 00:56:53,680
Speaker 3: I'm going to hedge this and say it very carefully.

1155
00:56:53,840 --> 00:56:54,920
Speaker 3: I think it's a terrible idea.

1156
00:56:55,120 --> 00:56:55,719
Speaker 5: Okay, go on.

1157
00:56:56,000 --> 00:56:58,600
Speaker 3: There are things that end up in retail in a

1158
00:56:58,680 --> 00:57:00,920
Speaker 3: very good way eventually, but we've done some of this

1159
00:57:00,920 --> 00:57:03,399
Speaker 3: when we introduce mutual funds. I can't say always going

1160
00:57:03,440 --> 00:57:05,719
Speaker 3: to retail is a bad thing. You can price it reasonably,

1161
00:57:05,760 --> 00:57:08,360
Speaker 3: you can say these are strategies you've never had before.

1162
00:57:09,040 --> 00:57:13,120
Speaker 3: This feels a little bit more like we've exhausted the institutions.

1163
00:57:13,280 --> 00:57:15,440
Speaker 3: I think I saw a number saying endown. It's like

1164
00:57:15,480 --> 00:57:18,560
Speaker 3: forty three percent for a number I can't verify. That's

1165
00:57:18,640 --> 00:57:21,840
Speaker 3: wildly specific, but that's the number I remember. So it

1166
00:57:21,880 --> 00:57:23,880
Speaker 3: has a feel of who else we're going to get

1167
00:57:23,880 --> 00:57:25,240
Speaker 3: to own this stuff.

1168
00:57:25,240 --> 00:57:27,440
Speaker 1: It seems so explicitly, and it's really wild.

1169
00:57:27,880 --> 00:57:31,560
Speaker 3: I think it's explicitly that it's one of these things

1170
00:57:31,560 --> 00:57:33,960
Speaker 3: that for the cynics, I don't think they'll ever be

1171
00:57:34,000 --> 00:57:36,280
Speaker 3: a satisfying moment where we get to say we were right.

1172
00:57:36,400 --> 00:57:40,040
Speaker 3: It'll just be somewhat worse over the next decade. It's

1173
00:57:40,040 --> 00:57:42,880
Speaker 3: not one of these things like a tech bubble in

1174
00:57:43,000 --> 00:57:45,920
Speaker 3: ninety nine two thousand that at least, I don't think

1175
00:57:46,120 --> 00:57:49,560
Speaker 3: there are scenarios where things get worse rapidly and the

1176
00:57:49,720 --> 00:57:52,280
Speaker 3: secondary sales. We got close to some of that in

1177
00:57:52,320 --> 00:57:55,680
Speaker 3: the GFC. I was on some investment committees where we

1178
00:57:55,680 --> 00:57:57,160
Speaker 3: were talking we didn't do it, but we were talking

1179
00:57:57,200 --> 00:57:59,280
Speaker 3: about whether we would have to do that. So I'm

1180
00:57:59,280 --> 00:58:01,000
Speaker 3: not saying it's a chance of ugliness, but I think

1181
00:58:01,000 --> 00:58:04,560
Speaker 3: the meat of the probability is it's just somewhat worse.

1182
00:58:05,160 --> 00:58:09,240
Speaker 3: But it does cause me some worry sadness to say, yeah,

1183
00:58:09,280 --> 00:58:11,160
Speaker 3: what we really need is to stick a liquids in

1184
00:58:11,240 --> 00:58:12,040
Speaker 3: four to one case.

1185
00:58:12,200 --> 00:58:12,600
Speaker 2: Yeah.

1186
00:58:12,640 --> 00:58:15,480
Speaker 3: So yeah, bluntly, I think it's a bad idea.

1187
00:58:15,520 --> 00:58:17,439
Speaker 4: Well, I just wanted to talk about and I don't

1188
00:58:17,440 --> 00:58:18,920
Speaker 4: know if this goes too far, but it feels like

1189
00:58:18,960 --> 00:58:20,640
Speaker 4: you had a change of heart when it comes to

1190
00:58:20,720 --> 00:58:23,600
Speaker 4: machine learning and AI in general. This is something we've

1191
00:58:23,600 --> 00:58:25,080
Speaker 4: spoken about before.

1192
00:58:25,120 --> 00:58:25,720
Speaker 3: I did well.

1193
00:58:25,880 --> 00:58:28,040
Speaker 4: Was that like a light bulb moment or was that

1194
00:58:28,160 --> 00:58:30,520
Speaker 4: just you know, maybe people on your team wearing you

1195
00:58:30,600 --> 00:58:31,360
Speaker 4: down over time?

1196
00:58:31,720 --> 00:58:34,200
Speaker 3: It was more the ladder. Yeah, the ladder is the

1197
00:58:34,200 --> 00:58:37,040
Speaker 3: second one, right, I gotta think that through every time

1198
00:58:37,480 --> 00:58:39,200
Speaker 3: I tell that people this, I've said in a lot

1199
00:58:39,240 --> 00:58:41,720
Speaker 3: of public arenas I sell it to clients. I think

1200
00:58:41,720 --> 00:58:43,400
Speaker 3: I probably slowed us down by a couple of years

1201
00:58:43,880 --> 00:58:45,880
Speaker 3: in machine learning. I think it probably costs us some

1202
00:58:45,920 --> 00:58:49,000
Speaker 3: money because the stuff has worked pretty well. We've always

1203
00:58:49,000 --> 00:58:51,360
Speaker 3: described ourselves and we were getting into this a little

1204
00:58:51,360 --> 00:58:55,640
Speaker 3: bit earlier. As a blend of data, back tests are nice,

1205
00:58:56,040 --> 00:59:01,040
Speaker 3: out of sample long periods are even nicer. But also theory.

1206
00:59:01,640 --> 00:59:03,720
Speaker 3: Theory can be a formal economic theory, but it can

1207
00:59:03,760 --> 00:59:05,920
Speaker 3: also just mean a common sense story where you think

1208
00:59:05,960 --> 00:59:09,520
Speaker 3: you understand why you're making money. And there's no way

1209
00:59:09,560 --> 00:59:12,680
Speaker 3: to say exactly what percentage of both. But I've often

1210
00:59:12,680 --> 00:59:14,720
Speaker 3: described it as an attempt to be fifty to fifty.

1211
00:59:14,880 --> 00:59:17,280
Speaker 3: We want something to make sense to us as economists

1212
00:59:17,720 --> 00:59:20,880
Speaker 3: and have strong data when you move into the machine

1213
00:59:20,960 --> 00:59:23,200
Speaker 3: learning world. I don't think you have to abandon theory,

1214
00:59:23,200 --> 00:59:24,600
Speaker 3: and this is something I think we do a little

1215
00:59:24,640 --> 00:59:26,840
Speaker 3: different than most. I think there are some in the

1216
00:59:26,840 --> 00:59:28,960
Speaker 3: machine learning world to kind of throw theory out the window.

1217
00:59:28,960 --> 00:59:32,560
Speaker 3: We're still using common sense and theory to kind of

1218
00:59:32,560 --> 00:59:35,080
Speaker 3: limit the scope, but you are leaning more on the data.

1219
00:59:36,000 --> 00:59:39,000
Speaker 3: And again I'm making up these numbers. But if regular

1220
00:59:39,000 --> 00:59:41,760
Speaker 3: stuff is fifty to fifty, machine learning seventy five twenty

1221
00:59:41,800 --> 00:59:45,440
Speaker 3: five data even for us, that was uncomfortable for me.

1222
00:59:45,640 --> 00:59:48,120
Speaker 3: When you've been telling a story for twenty five years

1223
00:59:48,400 --> 00:59:50,880
Speaker 3: and it's worked for you and your clients, it's not

1224
00:59:51,000 --> 00:59:54,640
Speaker 3: easy to move. I also think it's not improper for

1225
00:59:54,680 --> 00:59:57,160
Speaker 3: the role for the old man of the firm to go,

1226
00:59:57,360 --> 01:00:00,400
Speaker 3: let's just slow down. You guys come in here doing

1227
01:00:00,440 --> 01:00:05,280
Speaker 3: this with a Southern droll and your new fangled machine learning. No,

1228
01:00:05,360 --> 01:00:06,439
Speaker 3: that's the villainous rule.

1229
01:00:06,480 --> 01:00:07,680
Speaker 2: Oh no, I think it's kind of cool.

1230
01:00:07,720 --> 01:00:11,920
Speaker 3: Yeah, okay, so I actually can defend it as entirely appropriate.

1231
01:00:11,960 --> 01:00:14,000
Speaker 3: But yeah, I had to be convinced. It wasn't a

1232
01:00:14,080 --> 01:00:16,960
Speaker 3: light bulb moment. It was people like Brian Kelly, Andrea Ferzini,

1233
01:00:17,000 --> 01:00:20,480
Speaker 3: Laura Serb and all partners of mine presenting great results

1234
01:00:20,480 --> 01:00:23,080
Speaker 3: that made increase it and I did a lot more reading.

1235
01:00:23,120 --> 01:00:25,919
Speaker 3: I was probably I programmed in a programming language called

1236
01:00:26,040 --> 01:00:29,000
Speaker 3: LISP in the nineteen eighties. That was an early machine

1237
01:00:29,040 --> 01:00:32,480
Speaker 3: learning or AI language. I'm not even machine learning, and

1238
01:00:32,520 --> 01:00:34,560
Speaker 3: then I didn't do anything about that for the next

1239
01:00:34,560 --> 01:00:37,240
Speaker 3: thirty years, so I think part of it was just

1240
01:00:37,320 --> 01:00:39,760
Speaker 3: me getting up to speed. Frankly, the thing I.

1241
01:00:39,800 --> 01:00:42,880
Speaker 1: Find compelling in the Kelly paper is like you feed

1242
01:00:42,920 --> 01:00:44,800
Speaker 1: like a sort of toy set of factors into like

1243
01:00:44,800 --> 01:00:48,040
Speaker 1: a ten thousand neuron model, and what he argues is

1244
01:00:48,040 --> 01:00:51,240
Speaker 1: basically you might get a better view of the actual

1245
01:00:51,360 --> 01:00:54,360
Speaker 1: like function that generates the results than if you just

1246
01:00:54,400 --> 01:00:58,360
Speaker 1: try to like use your economic intuition in a lineary aggression. Right.

1247
01:00:58,400 --> 01:01:00,280
Speaker 1: There's like theory behind it. In the theory is like

1248
01:01:00,320 --> 01:01:03,840
Speaker 1: things are not as linear as you know traditional methods.

1249
01:01:04,040 --> 01:01:07,560
Speaker 3: And it's basically saying it's still a sin to only

1250
01:01:07,600 --> 01:01:11,120
Speaker 3: look for patterns. You still need some economics, but machine

1251
01:01:11,160 --> 01:01:13,520
Speaker 3: learning is better at balancing those trade offs. So the

1252
01:01:13,600 --> 01:01:16,080
Speaker 3: idea of throwing more at it when you have a

1253
01:01:16,080 --> 01:01:20,000
Speaker 3: better technique for that, you lean in that direction. I

1254
01:01:20,040 --> 01:01:22,560
Speaker 3: did have a better title, I think than him. I

1255
01:01:22,880 --> 01:01:25,960
Speaker 3: wanted him that's not better. His titles great pretty well, no,

1256
01:01:26,160 --> 01:01:31,680
Speaker 3: but I wanted to call it simply Ackham was wrong. Okay,

1257
01:01:31,760 --> 01:01:34,480
Speaker 3: all right and maybe not bad, but I still want

1258
01:01:34,560 --> 01:01:36,560
Speaker 3: him to use that for maybe a follow up paper.

1259
01:01:37,360 --> 01:01:38,560
Speaker 1: All right, Cliff, thanks for coming.

1260
01:01:38,600 --> 01:01:39,080
Speaker 3: Oh this was fun.

1261
01:01:39,160 --> 01:01:40,680
Speaker 2: Thanks for video with us.

1262
01:01:47,080 --> 01:01:48,520
Speaker 1: And that was the Money Stuff Podcast.

1263
01:01:48,680 --> 01:01:50,640
Speaker 6: I'm Matt Levi and I'm Katie Grifeld.

1264
01:01:51,040 --> 01:01:53,120
Speaker 1: You can find my work by subscribing to The Money

1265
01:01:53,120 --> 01:01:56,280
Speaker 1: Stuff newsletter on Bloomberg dot com.

1266
01:01:55,840 --> 01:01:58,360
Speaker 6: And you can find me on Bloomberg TV every day

1267
01:01:58,400 --> 01:02:01,480
Speaker 6: on Open Interest between nine to eleven am Eastern.

1268
01:02:01,840 --> 01:02:03,560
Speaker 1: We'd love to hear from you. You can send an

1269
01:02:03,560 --> 01:02:06,800
Speaker 1: email to Moneypod at Bloomberg dot net, ask us a

1270
01:02:06,840 --> 01:02:08,240
Speaker 1: question and we might answer it on air.

1271
01:02:08,520 --> 01:02:10,680
Speaker 6: You can also subscribe to our show wherever you're listening

1272
01:02:10,760 --> 01:02:12,760
Speaker 6: right now and leave us a review. It helps more

1273
01:02:12,800 --> 01:02:13,680
Speaker 6: people find the show.

1274
01:02:14,400 --> 01:02:17,640
Speaker 1: The Money Stuff Podcast is produced by Anamasarakus and Moses

1275
01:02:17,640 --> 01:02:19,919
Speaker 1: ONAM Special. Thanks this week, Tota see Wan.

1276
01:02:20,240 --> 01:02:22,320
Speaker 6: Our theme music was composed by Blake Maples.

1277
01:02:22,680 --> 01:02:25,360
Speaker 1: Brandon Francis Newdam is our executive producer.

1278
01:02:25,040 --> 01:02:27,200
Speaker 6: And Stage Bauman is Bloomberg's head of podcasts.

1279
01:02:27,400 --> 01:02:29,760
Speaker 1: Thanks for listening to The Money Stuff Podcasts. We'll be

1280
01:02:29,800 --> 01:02:31,320
Speaker 1: back next week with more stuff