WEBVTT

1
00:00:00.040 --> 00:00:00.240
Hello.

2
00:00:00.420 --> 00:00:02.602
Welcome back to Papers with Fact Test podcast.

3
00:00:02.842 --> 00:00:03.223
Today,

4
00:00:03.383 --> 00:00:06.645
we dive into another algo trading research paper.

5
00:00:07.226 --> 00:00:11.169
We're going to unpack abnormal trading volume and the cross-section of stock returns.

6
00:00:11.929 --> 00:00:12.169
Lee,

7
00:00:12.370 --> 00:00:12.670
Kim,

8
00:00:12.910 --> 00:00:14.511
and Kim from 2016.

9
00:00:14.512 --> 00:00:15.052
That's the one.

10
00:00:15.352 --> 00:00:15.492
Yeah.

11
00:00:15.712 --> 00:00:19.055
And the core question we're really tackling for you today is,

12
00:00:19.355 --> 00:00:19.595
you know,

13
00:00:19.755 --> 00:00:23.538
can unusual trading activity actually help predict future stock returns?

14
00:00:23.719 --> 00:00:25.900
And maybe what does it tell us about the market itself,

15
00:00:25.980 --> 00:00:26.100
right?

16
00:00:26.101 --> 00:00:26.220
Right.

17
00:00:26.884 --> 00:00:27.544
How it behaves.

18
00:00:27.624 --> 00:00:28.144
Exactly.

19
00:00:28.204 --> 00:00:29.324
So let's start with the basics.

20
00:00:29.384 --> 00:00:31.284
What kind of data were they working with here?

21
00:00:31.484 --> 00:00:31.704
Okay.

22
00:00:31.744 --> 00:00:32.924
So they really went broad.

23
00:00:32.984 --> 00:00:35.184
They looked at common stocks from the NYSE,

24
00:00:35.384 --> 00:00:35.744
Amex,

25
00:00:36.044 --> 00:00:36.884
and also Nasdaq.

26
00:00:37.084 --> 00:00:37.344
Right.

27
00:00:37.704 --> 00:00:40.184
And the timeframe was pretty extensive,

28
00:00:40.224 --> 00:00:43.664
January 68 all the way through December 2015.

29
00:00:43.704 --> 00:00:43.924
Wow.

30
00:00:43.964 --> 00:00:44.904
That's a lot of data.

31
00:00:45.064 --> 00:00:45.384
It is.

32
00:00:45.524 --> 00:00:47.644
And they build weekly and monthly return data,

33
00:00:47.804 --> 00:00:49.124
plus turnover data,

34
00:00:49.224 --> 00:00:49.464
you know,

35
00:00:49.544 --> 00:00:50.504
how often stocks traded.

36
00:00:50.964 --> 00:00:53.144
They did make some adjustments for Nasdaq volume,

37
00:00:53.164 --> 00:00:54.204
which is standard practice.

38
00:00:54.344 --> 00:00:54.604
Okay.

39
00:00:54.912 --> 00:00:57.232
And you mentioned abnormal volume in the title.

40
00:00:57.292 --> 00:00:58.252
How did they get at that?

41
00:00:58.332 --> 00:00:59.572
What was the key innovation?

42
00:00:59.992 --> 00:01:00.152
Yeah,

43
00:01:00.192 --> 00:01:01.372
this is really the core of it.

44
00:01:01.492 --> 00:01:04.592
Instead of just looking at like the total trading volume.

45
00:01:04.912 --> 00:01:05.672
The raw number.

46
00:01:05.892 --> 00:01:06.252
Exactly.

47
00:01:06.292 --> 00:01:06.852
They split it.

48
00:01:06.853 --> 00:01:08.112
They decompose it into two parts.

49
00:01:08.212 --> 00:01:08.392
Okay.

50
00:01:08.492 --> 00:01:08.752
First,

51
00:01:08.792 --> 00:01:11.852
there's what they call expected trading turnover or E-turn.

52
00:01:11.972 --> 00:01:12.332
E-turn.

53
00:01:12.532 --> 00:01:12.852
Got it.

54
00:01:12.952 --> 00:01:14.772
Think of this as the normal level,

55
00:01:15.492 --> 00:01:18.592
which you'd predict based on the stock's own trading history.

56
00:01:19.272 --> 00:01:20.192
It's usual behavior.

57
00:01:20.352 --> 00:01:23.272
So the baseline activity for that specific stock.

58
00:01:23.784 --> 00:01:24.244
Makes sense.

59
00:01:24.285 --> 00:01:24.905
What's the other part?

60
00:01:25.105 --> 00:01:28.248
The other part is the unexpected trading turnover or U-turn.

61
00:01:28.508 --> 00:01:28.868
U-turn.

62
00:01:29.048 --> 00:01:29.789
And that's basically,

63
00:01:30.850 --> 00:01:30.990
well,

64
00:01:31.050 --> 00:01:32.091
it's the leftover bit.

65
00:01:32.171 --> 00:01:32.391
Okay.

66
00:01:32.392 --> 00:01:35.113
The part of the volume that the past behavior doesn't explain.

67
00:01:35.353 --> 00:01:35.593
Ah,

68
00:01:35.773 --> 00:01:36.954
the surprise element,

69
00:01:37.074 --> 00:01:37.955
the abnormal bit.

70
00:01:37.975 --> 00:01:38.376
Precisely.

71
00:01:38.536 --> 00:01:39.316
It's the deviation.

72
00:01:39.817 --> 00:01:45.081
And understanding the split E-turn versus U-turn is fundamental to understanding your results.

73
00:01:45.461 --> 00:01:45.681
Okay,

74
00:01:45.781 --> 00:01:46.402
great setup.

75
00:01:46.602 --> 00:01:49.484
They've separated volume into expected and unexpected.

76
00:01:49.844 --> 00:01:51.326
Let's get right into the findings.

77
00:01:51.906 --> 00:01:52.667
With an E-turn,

78
00:01:52.984 --> 00:01:53.785
The expected part.

79
00:01:54.145 --> 00:01:55.546
Tell them about future returns.

80
00:01:55.926 --> 00:01:56.046
Well,

81
00:01:56.047 --> 00:01:56.927
this is where it gets interesting,

82
00:01:56.987 --> 00:01:58.408
maybe even a bit counterintuitive.

83
00:01:58.889 --> 00:01:59.909
They found that E-turn,

84
00:02:00.250 --> 00:02:01.671
whether you looked weekly or monthly,

85
00:02:02.271 --> 00:02:04.473
actually negatively predicted stock returns.

86
00:02:05.013 --> 00:02:05.574
Negatively.

87
00:02:05.754 --> 00:02:09.237
So higher expected trading meant lower returns later.

88
00:02:09.317 --> 00:02:10.418
That's what the data showed.

89
00:02:10.918 --> 00:02:11.078
Yeah.

90
00:02:11.579 --> 00:02:15.702
So stocks that everyone expected to trade a lot tended to underperform down the line.

91
00:02:15.962 --> 00:02:16.182
Huh.

92
00:02:16.943 --> 00:02:18.244
Any specific numbers on that?

93
00:02:18.304 --> 00:02:19.725
How strong was the effect?

94
00:02:19.985 --> 00:02:20.185
Sure.

95
00:02:20.866 --> 00:02:22.127
Let's take the monthly results.

96
00:02:22.616 --> 00:02:24.177
They sorted stocks into five groups,

97
00:02:24.498 --> 00:02:25.038
quintiles,

98
00:02:25.278 --> 00:02:25.999
based on E-turn.

99
00:02:26.259 --> 00:02:26.479
Okay.

100
00:02:26.719 --> 00:02:30.522
The group with the lowest expected turnover averaged an excess return of

101
00:02:30.943 --> 00:02:32.924
0.95% per month.

102
00:02:33.184 --> 00:02:33.485
Right.

103
00:02:33.705 --> 00:02:35.066
But the group with the highest E-turn,

104
00:02:35.626 --> 00:02:38.028
that dropped to 0.54%.

105
00:02:38.029 --> 00:02:38.208
Wow,

106
00:02:38.368 --> 00:02:39.049
quite a difference,

107
00:02:39.129 --> 00:02:39.629
and weekly.

108
00:02:40.010 --> 00:02:40.610
Similar story,

109
00:02:40.650 --> 00:02:41.411
just smaller numbers.

110
00:02:41.951 --> 00:02:43.132
It went from 0.25%

111
00:02:43.432 --> 00:02:46.014
for the lowest E-turn group down to 0.19%

112
00:02:46.094 --> 00:02:46.675
for the highest.

113
00:02:46.968 --> 00:02:48.089
a clear downward trend.

114
00:02:48.169 --> 00:02:50.050
And was this just a quick thing or did it last?

115
00:02:50.271 --> 00:02:50.391
No,

116
00:02:50.511 --> 00:02:51.572
it seemed pretty persistent.

117
00:02:51.672 --> 00:02:54.954
The paper shows this negative effect lasting for up to 16 periods,

118
00:02:55.034 --> 00:02:55.655
weeks or months,

119
00:02:55.715 --> 00:02:56.716
depending on the analysis.

120
00:02:57.056 --> 00:02:57.216
Yeah.

121
00:02:57.296 --> 00:02:59.498
And they even mentioned in unreported analyses,

122
00:02:59.518 --> 00:03:00.919
it went out as far as 60 months,

123
00:03:01.680 --> 00:03:02.120
five years.

124
00:03:02.340 --> 00:03:05.482
So sustained high expected volume might be a bit of a,

125
00:03:05.883 --> 00:03:06.063
well,

126
00:03:06.283 --> 00:03:07.684
a longer term red flag,

127
00:03:07.744 --> 00:03:07.984
maybe?

128
00:03:08.245 --> 00:03:09.466
That seems to be the implication.

129
00:03:09.686 --> 00:03:11.727
And they check this against standard risk factors too.

130
00:03:12.108 --> 00:03:12.428
Ah,

131
00:03:12.848 --> 00:03:13.189
good.

132
00:03:13.609 --> 00:03:14.369
Like size.

133
00:03:14.724 --> 00:03:15.244
value,

134
00:03:15.445 --> 00:03:15.965
momentum.

135
00:03:16.125 --> 00:03:16.586
Exactly.

136
00:03:16.626 --> 00:03:19.608
They use the Fama French Carhartt four-factor model,

137
00:03:20.148 --> 00:03:20.709
standard stuff.

138
00:03:21.009 --> 00:03:21.229
Right.

139
00:03:21.549 --> 00:03:24.452
And even after adjusting for those known risk premiums,

140
00:03:24.872 --> 00:03:26.453
this negative prediction from E-turn,

141
00:03:27.714 --> 00:03:28.435
it stuck around.

142
00:03:28.635 --> 00:03:29.636
It didn't just disappear.

143
00:03:29.916 --> 00:03:30.136
Okay.

144
00:03:30.196 --> 00:03:30.837
So E-turn,

145
00:03:31.457 --> 00:03:32.178
negative predictor,

146
00:03:32.358 --> 00:03:33.399
persists long-term,

147
00:03:33.799 --> 00:03:35.280
robust to risk factors.

148
00:03:35.761 --> 00:03:35.881
Now,

149
00:03:35.882 --> 00:03:36.781
what about the other side?

150
00:03:36.841 --> 00:03:37.882
The unexpected part,

151
00:03:38.383 --> 00:03:38.743
U-turn.

152
00:03:39.383 --> 00:03:39.644
Ah,

153
00:03:39.964 --> 00:03:40.584
now U-turn.

154
00:03:40.604 --> 00:03:41.805
That tells a completely different story.

155
00:03:41.945 --> 00:03:42.666
A positive one,

156
00:03:42.706 --> 00:03:43.106
actually.

157
00:03:43.386 --> 00:03:43.647
Okay.

158
00:03:43.888 --> 00:03:45.568
At both weekly and monthly horizons,

159
00:03:45.728 --> 00:03:48.248
U-turn positively predicted future stock returns.

160
00:03:48.868 --> 00:03:51.048
So when trading activity was unexpectedly high.

161
00:03:51.108 --> 00:03:52.768
Returns tended to be higher afterward.

162
00:03:52.769 --> 00:03:53.408
In the short term,

163
00:03:53.508 --> 00:03:53.788
yes.

164
00:03:53.988 --> 00:03:54.868
That's the key difference.

165
00:03:54.869 --> 00:04:01.128
It's like this surprise volume carried positive information or maybe just attention that pushed prices up temporarily.

166
00:04:01.468 --> 00:04:02.608
That's a really neat contrast.

167
00:04:02.628 --> 00:04:05.148
Can you give us the numbers for U-turn like you did for E-turn?

168
00:04:05.388 --> 00:04:05.848
Absolutely.

169
00:04:06.808 --> 00:04:07.548
Monthly horizon,

170
00:04:07.668 --> 00:04:07.888
again,

171
00:04:08.548 --> 00:04:12.848
the lowest U-turn quintile had an average excess return of 0.45%.

172
00:04:13.248 --> 00:04:13.448
Okay.

173
00:04:13.696 --> 00:04:14.156
baseline.

174
00:04:14.677 --> 00:04:16.418
But the highest U-turn quintile,

175
00:04:17.199 --> 00:04:19.921
the stocks with the biggest positive volume surprises,

176
00:04:20.201 --> 00:04:20.682
they average

177
00:04:21.562 --> 00:04:22.483
1.31%. Whoa,

178
00:04:22.484 --> 00:04:23.203
that's a big jump.

179
00:04:23.564 --> 00:04:26.246
Nearly a full percentage point difference per month.

180
00:04:26.426 --> 00:04:27.247
It is significant.

181
00:04:27.327 --> 00:04:28.388
And weekly was similar,

182
00:04:29.068 --> 00:04:29.468
going from

183
00:04:29.869 --> 00:04:33.512
0.01% for the lowest U-turn up to 0.53%

184
00:04:33.632 --> 00:04:34.212
for the highest.

185
00:04:34.332 --> 00:04:34.552
Again,

186
00:04:34.553 --> 00:04:36.094
a strong positive trend.

187
00:04:36.254 --> 00:04:38.015
But you emphasized short term.

188
00:04:38.235 --> 00:04:40.397
How long did this positive effect last?

189
00:04:40.577 --> 00:04:40.817
Right.

190
00:04:41.037 --> 00:04:42.739
Not nearly as long as the E-turn effect.

191
00:04:42.868 --> 00:04:47.191
They found this positive predictive power for U-turn was really concentrated in the first few weeks,

192
00:04:47.692 --> 00:04:48.613
up to about five weeks.

193
00:04:48.733 --> 00:04:49.473
Only five weeks.

194
00:04:49.474 --> 00:04:50.294
What happened after that?

195
00:04:50.714 --> 00:04:50.854
Well,

196
00:04:50.874 --> 00:04:52.996
then things started to reverse quite significantly,

197
00:04:53.056 --> 00:04:53.296
actually.

198
00:04:53.456 --> 00:04:54.037
Reversals.

199
00:04:54.297 --> 00:04:55.238
So the gains faded.

200
00:04:55.418 --> 00:04:55.938
And then some.

201
00:04:56.559 --> 00:04:57.720
For the monthly portfolios,

202
00:04:57.740 --> 00:05:01.122
these reversals started kicking in around month four after formation.

203
00:05:01.756 --> 00:05:03.317
for weekly around week 14.

204
00:05:03.858 --> 00:05:07.801
And the reversal effect actually seemed to peak somewhere around the 10 to 12 month mark.

205
00:05:08.101 --> 00:05:12.605
So a short term pop from surprise volume followed by a longer term pullback.

206
00:05:12.885 --> 00:05:14.566
It sounds almost like an overreaction.

207
00:05:15.006 --> 00:05:16.648
That's definitely one way to interpret it.

208
00:05:17.208 --> 00:05:19.110
The market seems to react strongly,

209
00:05:19.490 --> 00:05:22.672
maybe too strongly to the unexpected activity initially,

210
00:05:23.153 --> 00:05:24.734
and then it corrects over time.

211
00:05:24.994 --> 00:05:27.676
Which leads us back to the whole point of splitting the volume,

212
00:05:27.696 --> 00:05:27.816
right?

213
00:05:27.856 --> 00:05:28.337
Exactly.

214
00:05:28.897 --> 00:05:30.759
Because if you just looked at the raw turnover.

215
00:05:30.968 --> 00:05:32.549
you'd get these confusing signals.

216
00:05:32.890 --> 00:05:34.251
Sometimes high volume looks good,

217
00:05:34.331 --> 00:05:34.931
sometimes bad,

218
00:05:35.351 --> 00:05:36.212
depends on the time frame.

219
00:05:36.372 --> 00:05:36.673
Right.

220
00:05:36.993 --> 00:05:38.834
But this decomposition helps explain it.

221
00:05:39.375 --> 00:05:41.817
The short-term predictability of raw volume,

222
00:05:42.557 --> 00:05:45.599
that seems to be mostly driven by the positive U-turn effect.

223
00:05:45.720 --> 00:05:45.940
Okay.

224
00:05:46.320 --> 00:05:53.686
While the longer-term picture for raw volume is probably muddied or even dragged down by that negative persistent E-turn effect.

225
00:05:53.766 --> 00:05:54.687
That makes a lot of sense.

226
00:05:54.787 --> 00:05:57.569
It clarifies why raw volume alone can be inconsistent.

227
00:05:58.052 --> 00:05:59.533
So speaking of raw volume,

228
00:05:59.793 --> 00:06:03.777
how did it actually perform in their tests when they didn't decompose it?

229
00:06:03.797 --> 00:06:04.637
Just total turn.

230
00:06:05.017 --> 00:06:05.158
Yeah,

231
00:06:05.278 --> 00:06:06.018
they looked at that too.

232
00:06:06.519 --> 00:06:07.319
At the weekly level,

233
00:06:07.740 --> 00:06:12.724
raw turnover to your end did show a positive relationship with next week's returns.

234
00:06:12.804 --> 00:06:13.004
Okay,

235
00:06:13.284 --> 00:06:14.345
short-term positive.

236
00:06:14.545 --> 00:06:14.745
Right.

237
00:06:15.205 --> 00:06:17.807
The highest turnover stocks did better than the lowest,

238
00:06:18.108 --> 00:06:20.289
and the difference was statistically significant.

239
00:06:20.610 --> 00:06:22.751
Q5 minus Q1 was positive.

240
00:06:22.771 --> 00:06:23.512
But my 45...

241
00:06:23.513 --> 00:06:25.654
Monthly was much weaker.

242
00:06:26.034 --> 00:06:26.454
Let's clear.

243
00:06:26.864 --> 00:06:28.425
They found sort of an inverted U shape,

244
00:06:28.505 --> 00:06:28.806
actually.

245
00:06:29.386 --> 00:06:30.507
The middle groups did okay,

246
00:06:30.887 --> 00:06:33.769
but the highest turnover group wasn't significantly better than the lowest.

247
00:06:34.050 --> 00:06:36.552
The Q5-Q1 spread wasn't significant.

248
00:06:36.732 --> 00:06:37.292
So again,

249
00:06:37.332 --> 00:06:42.816
the signal gets weaker or changes over slightly longer horizons if you just use the raw number.

250
00:06:43.017 --> 00:06:43.497
Precisely.

251
00:06:43.737 --> 00:06:45.939
It really highlights the value of that decomposition.

252
00:06:46.219 --> 00:06:49.622
Did they look at like a simple trading strategy based on raw turnover?

253
00:06:49.882 --> 00:06:50.222
Buy high,

254
00:06:50.262 --> 00:06:50.583
sell low?

255
00:06:50.803 --> 00:06:51.163
They did.

256
00:06:51.283 --> 00:06:54.946
And it basically confirmed this short-term versus longer-term story.

257
00:06:55.226 --> 00:06:55.566
How so?

258
00:06:55.967 --> 00:06:56.127
well

259
00:06:56.368 --> 00:06:56.688
A long,

260
00:06:56.748 --> 00:07:02.193
short strategy using raw turnover showed positive profits for like the first one or two weeks only.

261
00:07:02.433 --> 00:07:02.553
OK,

262
00:07:02.753 --> 00:07:03.414
very short term.

263
00:07:03.534 --> 00:07:03.654
Yeah.

264
00:07:03.994 --> 00:07:04.374
After that,

265
00:07:04.454 --> 00:07:07.176
the profits turned negative and actually stayed negative for quite a while,

266
00:07:07.197 --> 00:07:08.858
out to 16 months in their tests.

267
00:07:09.278 --> 00:07:09.618
Wow.

268
00:07:09.859 --> 00:07:12.721
So chasing high raw volume might work for a week or two,

269
00:07:13.541 --> 00:07:14.923
but then it seems to backfire.

270
00:07:15.003 --> 00:07:16.404
That's what the backtest suggests.

271
00:07:16.524 --> 00:07:19.906
It really underlines why just seeing high volume isn't enough.

272
00:07:19.966 --> 00:07:20.547
You need to ask,

273
00:07:20.687 --> 00:07:22.649
is this volume expected or unexpected?

274
00:07:22.809 --> 00:07:23.029
Yeah,

275
00:07:23.129 --> 00:07:24.390
that distinction seems crucial.

276
00:07:24.550 --> 00:07:25.911
Did they do other checks like...

277
00:07:26.284 --> 00:07:30.844
Making sure this U-turn effect wasn't just some other known anomaly in disguise?

278
00:07:30.845 --> 00:07:31.184
It did.

279
00:07:31.324 --> 00:07:33.164
They ran Fama-Macbeth regressions,

280
00:07:33.524 --> 00:07:37.864
which is a way to test if a factor predicts returns even when you control for other known predictors.

281
00:07:38.084 --> 00:07:38.784
Like size,

282
00:07:38.864 --> 00:07:39.404
value,

283
00:07:39.624 --> 00:07:40.224
momentum,

284
00:07:40.384 --> 00:07:40.864
liquidity?

285
00:07:41.104 --> 00:07:41.504
All those,

286
00:07:41.524 --> 00:07:41.664
yeah.

287
00:07:42.304 --> 00:07:44.344
Plus things like idiosyncratic volatility,

288
00:07:44.504 --> 00:07:45.304
analyst dispersion,

289
00:07:45.744 --> 00:07:46.584
earning surprises.

290
00:07:46.664 --> 00:07:47.664
And U-turn held up?

291
00:07:47.764 --> 00:07:48.124
It did.

292
00:07:48.304 --> 00:07:52.364
The positive predictive power of U-turn remained robust across different tests.

293
00:07:52.972 --> 00:07:53.412
Interestingly,

294
00:07:53.553 --> 00:07:57.596
U-turn's negative power was a bit less consistent in these shorter-term regression settings,

295
00:07:58.216 --> 00:08:00.638
but U-turn's positive short-term effect was strong.

296
00:08:01.078 --> 00:08:04.741
What about that idea of a high-volume premium from earlier research?

297
00:08:05.302 --> 00:08:05.862
Gervais,

298
00:08:06.022 --> 00:08:06.503
Kenyal,

299
00:08:06.843 --> 00:08:07.143
Mingle,

300
00:08:07.183 --> 00:08:07.423
Grin?

301
00:08:07.804 --> 00:08:08.184
Good point.

302
00:08:08.185 --> 00:08:09.545
They specifically checked that.

303
00:08:10.145 --> 00:08:16.290
They replicated the method from that 2001 paper to identify stocks that generally have high,

304
00:08:16.611 --> 00:08:16.951
normal,

305
00:08:17.051 --> 00:08:19.973
or low volume based on their own past history.

306
00:08:20.093 --> 00:08:20.313
Okay.

307
00:08:20.814 --> 00:08:22.015
And even within those groups.

308
00:08:22.344 --> 00:08:24.946
among stocks that are already known high volume traders,

309
00:08:25.747 --> 00:08:28.249
U-turn still positively predicted short term returns.

310
00:08:29.089 --> 00:08:30.050
So the U-turn effect,

311
00:08:30.410 --> 00:08:34.774
the unexpected part seems to be distinct from just being a generally high volume stock.

312
00:08:34.894 --> 00:08:35.715
That's the conclusion.

313
00:08:35.775 --> 00:08:35.895
Yeah.

314
00:08:36.115 --> 00:08:37.196
It's a separate phenomenon.

315
00:08:37.456 --> 00:08:37.656
Okay.

316
00:08:37.696 --> 00:08:38.897
So let's try to pull this together.

317
00:08:39.317 --> 00:08:43.180
The main takeaway seems to be that decomposing trading volume is really insightful.

318
00:08:43.360 --> 00:08:44.001
Absolutely key.

319
00:08:44.341 --> 00:08:46.843
Unexpected or abnormal volume.

320
00:08:47.304 --> 00:08:50.286
U-turn gives you a positive signal for short-term returns.

321
00:08:50.660 --> 00:08:51.561
But it reverses.

322
00:08:51.601 --> 00:08:52.361
But it reverses.

323
00:08:53.062 --> 00:08:53.562
Meanwhile,

324
00:08:53.782 --> 00:08:56.665
the expected volume E-turn is actually a negative signal,

325
00:08:57.105 --> 00:08:58.306
especially over the longer term.

326
00:08:58.326 --> 00:09:01.428
And this explains why raw volume gives mixed signals.

327
00:09:01.569 --> 00:09:02.049
Exactly.

328
00:09:02.189 --> 00:09:03.590
It resolves that inconsistency.

329
00:09:03.690 --> 00:09:05.832
The short term is U-turn's game.

330
00:09:06.292 --> 00:09:08.894
The long term is influenced more by E-turn's drag.

331
00:09:09.195 --> 00:09:12.137
That feels like the real aha moment from this paper.

332
00:09:12.417 --> 00:09:14.038
It takes something seemingly simple,

333
00:09:14.339 --> 00:09:14.839
volume,

334
00:09:15.139 --> 00:09:18.602
and shows there's important hidden information if you just look a bit closer.

335
00:09:19.082 --> 00:09:19.483
Definitely.

336
00:09:19.723 --> 00:09:20.243
And the paper...

337
00:09:20.720 --> 00:09:24.423
briefly touches on why this might happen behavioral stuff.

338
00:09:24.903 --> 00:09:25.063
Oh,

339
00:09:25.304 --> 00:09:25.644
like what?

340
00:09:26.164 --> 00:09:28.166
Things like investor overconfidence,

341
00:09:28.626 --> 00:09:31.188
maybe biased self-attribution after gains,

342
00:09:31.649 --> 00:09:32.689
the disposition effect,

343
00:09:32.870 --> 00:09:33.990
even just attention shifts.

344
00:09:34.431 --> 00:09:40.215
They suggest the U-turn effect might be more tied to these behavioral biases than just simple attention grabbing.

345
00:09:40.576 --> 00:09:41.016
Interesting.

346
00:09:41.036 --> 00:09:45.720
So maybe the unexpected volume reflects moments when these biases are driving trading?

347
00:09:45.780 --> 00:09:46.260
Plausibly,

348
00:09:46.320 --> 00:09:46.600
yes.

349
00:09:47.161 --> 00:09:49.583
They also found that short sale constraints didn't seem to make

350
00:09:49.712 --> 00:09:51.054
the predictability stronger,

351
00:09:51.515 --> 00:09:53.678
which sometimes happens with mispricing stories.

352
00:09:53.818 --> 00:09:53.939
OK,

353
00:09:54.319 --> 00:09:55.261
so wrapping up,

354
00:09:55.281 --> 00:09:57.264
the big insight is this decomposition.

355
00:09:57.761 --> 00:09:58.141
For sure.

356
00:09:58.362 --> 00:09:59.442
Which leads to a final thought,

357
00:09:59.542 --> 00:10:00.403
maybe for you listening.

358
00:10:00.823 --> 00:10:03.405
Even if you're not calculating E-turn and U-turn formally.

359
00:10:03.726 --> 00:10:03.906
Right.

360
00:10:04.226 --> 00:10:05.507
How can you use this concept?

361
00:10:05.687 --> 00:10:09.850
Maybe it's about paying attention to significant deviations from a stock's normal trading pattern.

362
00:10:10.231 --> 00:10:10.351
Yeah,

363
00:10:10.471 --> 00:10:15.555
perhaps looking for those unusual spikes or lulls as potential short-term signals.

364
00:10:15.915 --> 00:10:20.119
While being maybe a bit more wary of stocks that just trade heavily,

365
00:10:20.239 --> 00:10:20.939
consistently,

366
00:10:21.119 --> 00:10:21.960
month after month.

367
00:10:22.000 --> 00:10:24.202
Could be a practical way to think about applying this idea.

368
00:10:24.702 --> 00:10:25.843
recognizing that not all

369
00:10:25.953 --> 00:10:27.594
all volume is created equal.

370
00:10:27.815 --> 00:10:28.995
Definitely food for thought.

371
00:10:29.336 --> 00:10:30.497
This was a great breakdown.

372
00:10:30.537 --> 00:10:30.657
Yeah,

373
00:10:30.797 --> 00:10:34.159
really interesting how separating those components changes the picture so much.

374
00:10:34.400 --> 00:10:37.022
Thank you for tuning in to Papers with Backtest podcast.

375
00:10:37.402 --> 00:10:39.984
We hope today's episode gave you useful insights.

376
00:10:40.464 --> 00:10:42.606
Join us next time as we break down more research.

377
00:10:43.167 --> 00:10:44.588
And for more papers and backtests,

378
00:10:44.628 --> 00:10:47.710
find us at https.paperswithbacktest.com.

379
00:10:48.030 --> 00:10:48.711
Hackey Trading.