WEBVTT

1
00:00:00.040 --> 00:00:00.180
Hello,

2
00:00:00.320 --> 00:00:02.442
welcome back to Papers with Backtest podcast.

3
00:00:02.882 --> 00:00:05.724
Today we dive into another algo trading research paper.

4
00:00:05.885 --> 00:00:06.105
Yeah,

5
00:00:06.225 --> 00:00:07.206
looking forward to this one.

6
00:00:07.346 --> 00:00:10.288
We're revisiting a really interesting idea in trading,

7
00:00:10.688 --> 00:00:10.808
the

8
00:00:11.429 --> 00:00:12.270
52-week high effect.

9
00:00:13.931 --> 00:00:15.752
You might remember back in 2004,

10
00:00:16.193 --> 00:00:18.695
George and Wang pointed out something fascinating.

11
00:00:18.815 --> 00:00:19.095
Right,

12
00:00:19.215 --> 00:00:22.898
that stocks trading near their highest price in the past year tended to,

13
00:00:23.158 --> 00:00:23.418
well,

14
00:00:23.538 --> 00:00:26.401
do better going forward than those way below their guys.

15
00:00:26.561 --> 00:00:27.061
Exactly.

16
00:00:27.301 --> 00:00:28.422
And And their initial thought,

17
00:00:28.462 --> 00:00:28.923
their theory,

18
00:00:29.244 --> 00:00:30.545
It was about anchoring bias.

19
00:00:30.725 --> 00:00:30.845
Yeah.

20
00:00:30.965 --> 00:00:33.647
Like investors sort of get anchored to that recent high number.

21
00:00:33.848 --> 00:00:34.028
Yeah.

22
00:00:34.168 --> 00:00:36.470
And maybe don't react fully to new good news,

23
00:00:36.510 --> 00:00:36.750
you know.

24
00:00:37.310 --> 00:00:39.892
But what's great is how researchers kept digging into this.

25
00:00:40.112 --> 00:00:41.454
Which brings us to today's paper.

26
00:00:41.734 --> 00:00:43.315
It's from 2011 by Hong,

27
00:00:43.655 --> 00:00:43.975
Jordan,

28
00:00:44.016 --> 00:00:44.296
and Liu.

29
00:00:44.976 --> 00:00:46.077
Industry information and the

30
00:00:46.578 --> 00:00:47.358
52-week high effect.

31
00:00:47.879 --> 00:00:49.720
And this one really tries to get at the why.

32
00:00:49.960 --> 00:00:53.443
Is this effect just a risk thing?

33
00:00:54.123 --> 00:00:56.425
Or is it more about how investors actually behave?

34
00:00:56.564 --> 00:00:57.305
And they had a twist,

35
00:00:57.365 --> 00:00:57.565
right?

36
00:00:57.685 --> 00:00:57.865
Yeah.

37
00:00:58.005 --> 00:01:01.788
Looking at whether it's driven by individual company stuff or maybe broader industry information.

38
00:01:01.948 --> 00:01:02.529
Precisely.

39
00:01:02.969 --> 00:01:08.533
And they found something pretty surprising that maybe an industry-focused strategy could be even more profitable.

40
00:01:08.653 --> 00:01:08.814
Okay,

41
00:01:08.854 --> 00:01:09.574
let's unpack this.

42
00:01:09.934 --> 00:01:12.837
So George and Huang looked at individual stocks hitting highs,

43
00:01:13.757 --> 00:01:14.298
but Hong,

44
00:01:14.498 --> 00:01:14.858
Jordan,

45
00:01:14.958 --> 00:01:16.700
and Liu kind of zoomed out.

46
00:01:17.020 --> 00:01:17.460
They did.

47
00:01:17.660 --> 00:01:18.621
They looked at industries,

48
00:01:18.721 --> 00:01:20.543
and they used a really long time frame,

49
00:01:20.544 --> 00:01:22.985
1963 all the way to 2009.

50
00:01:23.465 --> 00:01:24.626
That's a lot of market history.

51
00:01:24.946 --> 00:01:25.767
Definitely gives it weight.

52
00:01:26.168 --> 00:01:35.976
So our mission today is to really understand the trading rules they tested for both individual stocks and these industry groups and crucially what the back tests showed.

53
00:01:36.196 --> 00:01:36.416
Right.

54
00:01:36.716 --> 00:01:38.298
What actually worked and how well.

55
00:01:38.718 --> 00:01:40.759
So let's start with that original individual strategy.

56
00:01:40.980 --> 00:01:42.561
The one George and Huang outlined,

57
00:01:42.581 --> 00:01:45.323
which this paper also uses as a sort of baseline.

58
00:01:46.024 --> 00:01:46.684
How did that work?

59
00:01:46.764 --> 00:01:47.385
What are the steps?

60
00:01:47.645 --> 00:01:48.626
It's pretty straightforward,

61
00:01:48.686 --> 00:01:49.026
actually.

62
00:01:49.246 --> 00:01:50.107
At the end of every month,

63
00:01:50.108 --> 00:01:51.888
you calculate a ratio for each stock.

64
00:01:51.968 --> 00:01:53.029
They called it pre-log.

65
00:01:53.189 --> 00:01:53.730
Pre-log.

66
00:01:53.950 --> 00:01:54.070
Yeah.

67
00:01:54.240 --> 00:01:57.022
It's just the current price divided by its 52 week high.

68
00:01:57.302 --> 00:01:57.903
Simple as that.

69
00:01:58.003 --> 00:01:59.804
Price relative to the last year's high.

70
00:01:59.924 --> 00:02:00.265
Got it.

71
00:02:00.725 --> 00:02:03.427
Then you rank all the stocks based on that ratio.

72
00:02:03.627 --> 00:02:04.568
The top 30%,

73
00:02:04.569 --> 00:02:05.749
the ones closest to their high.

74
00:02:05.750 --> 00:02:06.450
That's your winner group.

75
00:02:06.490 --> 00:02:07.751
And the bottom 30%,

76
00:02:07.752 --> 00:02:08.791
the ones furthest away.

77
00:02:08.952 --> 00:02:10.132
Those are the losers.

78
00:02:10.333 --> 00:02:10.573
Okay.

79
00:02:10.753 --> 00:02:13.995
Winners and losers based on individual stock highs.

80
00:02:14.476 --> 00:02:15.497
What's the actual trade?

81
00:02:15.737 --> 00:02:18.459
You buy the winners go long and you short the losers.

82
00:02:18.739 --> 00:02:19.380
Bet against them.

83
00:02:19.500 --> 00:02:20.821
And how are the portfolios weighted?

84
00:02:21.021 --> 00:02:22.302
Equal weighted in this setup,

85
00:02:22.382 --> 00:02:22.502
so

86
00:02:22.796 --> 00:02:26.639
every stock gets the same amount of capital initially and you hold these positions for six months.

87
00:02:26.719 --> 00:02:27.199
Six months,

88
00:02:27.260 --> 00:02:28.080
then rebalance.

89
00:02:28.661 --> 00:02:28.961
Okay.

90
00:02:29.081 --> 00:02:29.982
And the big question,

91
00:02:30.602 --> 00:02:36.627
did this individual strategy actually make money over that long 1963 to 2009 period?

92
00:02:36.967 --> 00:02:37.107
Well,

93
00:02:37.267 --> 00:02:38.228
according to their backtest,

94
00:02:38.308 --> 00:02:38.548
yes.

95
00:02:38.849 --> 00:02:42.311
It generated an average monthly return of 0.43%.

96
00:02:42.312 --> 00:02:44.173
0.43%

97
00:02:44.213 --> 00:02:44.533
a month.

98
00:02:44.573 --> 00:02:45.214
Not huge,

99
00:02:45.314 --> 00:02:46.695
but it adds up over decades,

100
00:02:46.715 --> 00:02:47.135
I suppose.

101
00:02:47.255 --> 00:02:47.615
It can,

102
00:02:47.695 --> 00:02:48.116
absolutely.

103
00:02:48.296 --> 00:02:49.957
But there's always the counter argument,

104
00:02:50.017 --> 00:02:50.157
right?

105
00:02:50.177 --> 00:02:51.058
Maybe it's just risk.

106
00:02:51.136 --> 00:02:54.356
You mean maybe stocks near their highs are just inherently riskier,

107
00:02:54.696 --> 00:02:55.676
higher beta or something?

108
00:02:55.677 --> 00:02:56.216
Exactly.

109
00:02:56.476 --> 00:03:00.356
Maybe that return is just compensation for taking on more market risk.

110
00:03:00.436 --> 00:03:01.816
That's one possible explanation.

111
00:03:02.156 --> 00:03:02.376
Right.

112
00:03:02.916 --> 00:03:04.996
Which makes the industry angle even more interesting.

113
00:03:05.596 --> 00:03:07.316
Did they find a stronger signal there?

114
00:03:07.916 --> 00:03:09.636
How did they build that industry strategy?

115
00:03:09.736 --> 00:03:10.636
Sounds a bit more involved.

116
00:03:11.036 --> 00:03:12.296
It is a few more steps.

117
00:03:12.736 --> 00:03:13.136
First,

118
00:03:13.276 --> 00:03:14.376
they needed industries.

119
00:03:14.876 --> 00:03:17.356
They used two-digit SIC codes,

120
00:03:17.496 --> 00:03:19.096
which basically grouped stocks into...

121
00:03:19.384 --> 00:03:20.485
20 different industries.

122
00:03:20.585 --> 00:03:20.705
OK,

123
00:03:20.825 --> 00:03:22.466
20 broad industry groups.

124
00:03:23.007 --> 00:03:23.407
Then what?

125
00:03:23.787 --> 00:03:25.189
Then for each industry,

126
00:03:25.409 --> 00:03:25.889
each month,

127
00:03:26.089 --> 00:03:29.132
they calculated an average pre-lag for all the stocks within it.

128
00:03:29.632 --> 00:03:30.393
But importantly,

129
00:03:30.513 --> 00:03:32.274
it was a value-weighted average.

130
00:03:32.454 --> 00:03:32.594
Ah,

131
00:03:32.995 --> 00:03:38.839
so bigger companies within an industry had more influence on that industry's average pre-lag score.

132
00:03:39.139 --> 00:03:39.920
That's the idea.

133
00:03:40.080 --> 00:03:43.363
So an industry score reflected whether its larger constituents,

134
00:03:43.443 --> 00:03:44.023
on average,

135
00:03:44.143 --> 00:03:45.304
were near their highs.

136
00:03:45.344 --> 00:03:45.464
OK,

137
00:03:45.564 --> 00:03:46.765
so now we have industry scores,

138
00:03:46.785 --> 00:03:48.046
not just individual stock scores.

139
00:03:48.324 --> 00:03:49.885
How do they form the portfolios then?

140
00:03:50.186 --> 00:03:50.926
Similar idea,

141
00:03:51.066 --> 00:03:52.167
but at the industry level.

142
00:03:52.347 --> 00:03:55.049
They rank the 20 industries by their average prelag.

143
00:03:55.229 --> 00:03:57.591
The top six industries were the winners.

144
00:03:57.831 --> 00:03:58.572
And the bottom six.

145
00:03:58.872 --> 00:04:00.373
The bottom six were the losers.

146
00:04:00.574 --> 00:04:00.814
Right.

147
00:04:00.834 --> 00:04:03.376
So the winner portfolio wasn't just the top stocks.

148
00:04:03.676 --> 00:04:06.298
It was all the stocks in those top six winning industries.

149
00:04:06.778 --> 00:04:07.139
Correct.

150
00:04:07.439 --> 00:04:10.862
And the loser portfolio was all the stocks in the bottom six industries.

151
00:04:10.982 --> 00:04:12.063
And the trading mechanics.

152
00:04:12.663 --> 00:04:12.923
Still,

153
00:04:13.183 --> 00:04:13.864
long winners,

154
00:04:14.044 --> 00:04:14.805
short losers,

155
00:04:15.425 --> 00:04:16.726
equal weighted portfolios.

156
00:04:17.276 --> 00:04:18.297
Held for six months.

157
00:04:18.577 --> 00:04:18.837
Yes,

158
00:04:19.118 --> 00:04:22.600
exactly the same structure in terms of the trade execution and holding period.

159
00:04:23.061 --> 00:04:25.162
Equal weighted across all the stocks selected.

160
00:04:25.202 --> 00:04:26.603
And here's the kicker for you listening.

161
00:04:27.544 --> 00:04:30.446
What did the back test show for this industry approach?

162
00:04:30.867 --> 00:04:31.367
Was it better?

163
00:04:31.767 --> 00:04:33.729
It was significantly better,

164
00:04:33.829 --> 00:04:34.189
actually.

165
00:04:34.530 --> 00:04:38.052
The industry strategy yielded an average monthly return of 0.60%

166
00:04:38.913 --> 00:04:39.874
over that same period.

167
00:04:40.594 --> 00:04:41.415
0.60%

168
00:04:41.435 --> 00:04:43.116
compared to 0.43%

169
00:04:43.236 --> 00:04:44.237
for the individual one.

170
00:04:44.238 --> 00:04:44.417
That's

171
00:04:44.800 --> 00:04:44.920
What,

172
00:04:44.960 --> 00:04:45.621
nearly a 50%

173
00:04:45.981 --> 00:04:46.361
increase?

174
00:04:46.481 --> 00:04:46.862
Roughly,

175
00:04:46.982 --> 00:04:47.142
yeah.

176
00:04:47.302 --> 00:04:48.363
A pretty substantial jump.

177
00:04:48.743 --> 00:04:49.244
And crucially,

178
00:04:49.284 --> 00:04:51.625
this higher return was statistically significant.

179
00:04:51.645 --> 00:04:52.846
It wasn't likely just luck.

180
00:04:53.067 --> 00:04:53.367
Wow.

181
00:04:53.507 --> 00:04:53.687
Okay,

182
00:04:53.707 --> 00:04:55.649
so focusing on industry highs seems,

183
00:04:55.789 --> 00:04:56.549
at least initially,

184
00:04:56.609 --> 00:04:57.470
much more powerful.

185
00:04:57.930 --> 00:04:59.011
But back to that risk question.

186
00:04:59.609 --> 00:05:03.412
Did they check if this bigger return was just because the industry strategy was maybe,

187
00:05:03.413 --> 00:05:04.012
I don't know,

188
00:05:04.073 --> 00:05:05.133
loading up on more risk?

189
00:05:05.213 --> 00:05:06.314
Absolutely critical question.

190
00:05:06.354 --> 00:05:06.675
And yes,

191
00:05:06.715 --> 00:05:08.096
they did extensive checks for risk.

192
00:05:08.116 --> 00:05:10.157
They use things like the Carhartt four factor model.

193
00:05:10.498 --> 00:05:10.718
Right.

194
00:05:10.758 --> 00:05:12.119
That looks at market size,

195
00:05:12.159 --> 00:05:13.920
value and momentum factors.

196
00:05:14.020 --> 00:05:14.521
Exactly.

197
00:05:14.901 --> 00:05:18.644
And they also use something called DGTW benchmark adjusted returns,

198
00:05:18.964 --> 00:05:23.248
which is another way to compare returns against similar stocks to account for characteristics.

199
00:05:23.668 --> 00:05:26.470
So they really tried to isolate if there was genuine alpha.

200
00:05:26.809 --> 00:05:27.049
you know,

201
00:05:27.189 --> 00:05:29.851
excess return beyond just risk compensation.

202
00:05:30.572 --> 00:05:32.594
What happened when they applied these risk controls?

203
00:05:33.114 --> 00:05:34.395
This is where it gets really interesting.

204
00:05:34.655 --> 00:05:36.196
For the individual stock strategy,

205
00:05:36.517 --> 00:05:39.719
its profitability pretty much vanished after these risk adjustments.

206
00:05:39.859 --> 00:05:40.159
Ah,

207
00:05:40.500 --> 00:05:41.801
so the 0.43%

208
00:05:42.001 --> 00:05:42.802
monthly return,

209
00:05:43.462 --> 00:05:46.564
mostly explained by risk factors like size or momentum.

210
00:05:46.745 --> 00:05:47.425
It seems that way.

211
00:05:47.885 --> 00:05:50.648
The DGTW adjuster returns weren't significant.

212
00:05:50.873 --> 00:05:55.116
And the four-factor alpha for the long-short portfolio wasn't statistically significant either.

213
00:05:55.236 --> 00:05:56.397
But the industry strategy,

214
00:05:56.457 --> 00:05:57.218
did that hold up?

215
00:05:57.398 --> 00:05:57.718
It did.

216
00:05:58.019 --> 00:06:01.702
Even after controlling for the Carhartt factors and using DGTW adjustments,

217
00:06:01.862 --> 00:06:05.685
the industry strategy still showed significant positive abnormal returns.

218
00:06:05.765 --> 00:06:06.085
How much?

219
00:06:06.785 --> 00:06:07.006
Let's see,

220
00:06:07.007 --> 00:06:08.187
0.38%

221
00:06:08.327 --> 00:06:09.848
monthly using DGTW,

222
00:06:10.228 --> 00:06:11.809
still statistically significant.

223
00:06:12.230 --> 00:06:14.872
And the four-factor alpha was 0.22%

224
00:06:14.873 --> 00:06:15.232
per month,

225
00:06:15.472 --> 00:06:16.333
also significant.

226
00:06:16.593 --> 00:06:19.676
So the industry effect seems much more robust

227
00:06:19.813 --> 00:06:21.314
to risk explanations.

228
00:06:21.614 --> 00:06:22.695
That's a key finding for you.

229
00:06:23.236 --> 00:06:25.578
And they mentioned the profit came mostly from the long side.

230
00:06:25.918 --> 00:06:26.158
Yes,

231
00:06:26.238 --> 00:06:27.619
that's another important detail.

232
00:06:27.839 --> 00:06:33.023
The buy side buying the stocks and the winning industries seem to be driving most of that persistent alpha.

233
00:06:33.504 --> 00:06:33.924
Interesting.

234
00:06:33.925 --> 00:06:38.368
So it's more about picking the right industries to buy than shorting the losers effectively.

235
00:06:38.888 --> 00:06:42.010
Did they look at other risk checks like mean adjusted returns?

236
00:06:42.090 --> 00:06:42.491
They did.

237
00:06:42.611 --> 00:06:46.154
They compared stock returns to their own historical averages.

238
00:06:46.374 --> 00:06:46.854
Same story,

239
00:06:46.914 --> 00:06:47.314
basically.

240
00:06:47.733 --> 00:06:49.595
Individual strategy profit disappeared.

241
00:06:49.955 --> 00:06:52.016
Industry strategy profit remains significant.

242
00:06:52.557 --> 00:06:57.921
It really bolsters the idea that something beyond standard risk factors is at play with the industry effect.

243
00:06:58.221 --> 00:06:58.341
OK,

244
00:06:58.542 --> 00:07:00.443
so if it's not just risk,

245
00:07:00.943 --> 00:07:04.686
maybe it is that anchoring bias idea George and Wang first proposed,

246
00:07:05.187 --> 00:07:07.929
but maybe it operates more strongly at the industry level.

247
00:07:08.269 --> 00:07:09.570
How did the paper investigate that?

248
00:07:10.311 --> 00:07:12.472
They looked at what institutional investors are doing.

249
00:07:12.693 --> 00:07:15.735
The thinking is if there's underreaction or anchoring,

250
00:07:16.117 --> 00:07:18.697
Maybe sophisticated investors are noticing and trading on it.

251
00:07:18.837 --> 00:07:19.297
Makes sense.

252
00:07:19.337 --> 00:07:19.957
What did they find?

253
00:07:20.397 --> 00:07:22.197
They found that institutional investors,

254
00:07:22.237 --> 00:07:24.897
particularly the more active transient ones,

255
00:07:25.537 --> 00:07:29.997
did tend to increase their holdings in stocks and industries that were getting closer to their

256
00:07:30.517 --> 00:07:31.017
52-week highs.

257
00:07:31.217 --> 00:07:33.277
And decrease holdings when they fell away from the highs.

258
00:07:33.417 --> 00:07:33.977
Exactly.

259
00:07:34.557 --> 00:07:37.577
Their behavior seemed to align with capitalizing on this effect,

260
00:07:37.737 --> 00:07:42.777
which lends support to the underreaction or anchoring explanation rather than just risk.

261
00:07:42.797 --> 00:07:45.097
It's like they see the market underreacting and step in.

262
00:07:45.421 --> 00:07:47.883
That suggests it's a broader behavioral phenomenon,

263
00:07:47.943 --> 00:07:50.105
not just a few naive investors getting stuck.

264
00:07:51.246 --> 00:07:51.366
OK,

265
00:07:51.486 --> 00:07:55.389
so we have this robust industry effect potentially linked to behavior.

266
00:07:56.330 --> 00:08:00.373
How does it stack up against other well-known strategies like momentum?

267
00:08:01.113 --> 00:08:01.634
Good question.

268
00:08:02.074 --> 00:08:03.375
Momentum is obviously related,

269
00:08:03.415 --> 00:08:03.615
right?

270
00:08:04.316 --> 00:08:05.277
Buying past winners.

271
00:08:05.777 --> 00:08:12.762
They specifically compare the industry 52 week high strategy to both individual stock momentum and industry momentum strategies.

272
00:08:12.862 --> 00:08:14.444
Did the 52 week high effect just.

273
00:08:14.685 --> 00:08:16.506
disappear once you accounted for momentum?

274
00:08:16.847 --> 00:08:16.967
No,

275
00:08:17.127 --> 00:08:17.447
actually.

276
00:08:17.547 --> 00:08:18.808
The paper found that the industry

277
00:08:19.449 --> 00:08:23.812
52-week high strategy remained profitable even when controlling for these momentum effects.

278
00:08:24.112 --> 00:08:26.014
It suggests it's capturing something distinct.

279
00:08:26.374 --> 00:08:28.476
So it's not just a different flavor of momentum.

280
00:08:28.736 --> 00:08:30.077
It seems to be its own thing.

281
00:08:30.377 --> 00:08:32.659
They even ran simultaneous regressions,

282
00:08:32.699 --> 00:08:34.340
including both momentum and the

283
00:08:34.821 --> 00:08:35.842
52-week high variables,

284
00:08:36.322 --> 00:08:40.125
and the industry high strategy stayed independently profitable.

285
00:08:40.625 --> 00:08:43.147
That's pretty compelling evidence for it being a separate effect.

286
00:08:43.673 --> 00:08:45.134
What about long-term performance?

287
00:08:45.975 --> 00:08:49.237
Sometimes momentum strategies reverse over longer horizons.

288
00:08:49.357 --> 00:08:50.218
Did they see that here?

289
00:08:50.558 --> 00:08:51.159
Interestingly,

290
00:08:51.279 --> 00:08:51.399
no.

291
00:08:51.799 --> 00:08:54.241
They looked at returns over longer periods,

292
00:08:54.261 --> 00:08:55.562
like three to five years out.

293
00:08:56.203 --> 00:08:59.285
While there's some evidence that individual momentum can reverse,

294
00:08:59.625 --> 00:09:02.928
they found no significant long-run reversal for the industry

295
00:09:03.689 --> 00:09:04.569
52-week high strategy.

296
00:09:04.629 --> 00:09:06.491
Which fits better with an underreaction story,

297
00:09:06.591 --> 00:09:06.791
right?

298
00:09:06.931 --> 00:09:07.992
If it were overreaction,

299
00:09:07.993 --> 00:09:09.934
you might expect it to correct itself eventually.

300
00:09:09.954 --> 00:09:10.394
Exactly.

301
00:09:10.557 --> 00:09:15.741
Under reaction implies the market slowly catches up rather than then overshooting and then snapping back.

302
00:09:16.342 --> 00:09:19.524
The paper also tried to pin down what kind of information drives this.

303
00:09:20.104 --> 00:09:23.227
Is it firm specific news or broader industry trends?

304
00:09:23.862 --> 00:09:24.062
Yeah.

305
00:09:24.322 --> 00:09:28.145
They looked at whether the effect was stronger for stocks that tend to move more with their industry,

306
00:09:28.486 --> 00:09:33.370
those with high industry betas and high R-squared values relative to industry returns.

307
00:09:33.870 --> 00:09:34.010
Yeah,

308
00:09:34.030 --> 00:09:34.310
was it?

309
00:09:34.510 --> 00:09:34.731
Yes.

310
00:09:35.311 --> 00:09:35.431
The

311
00:09:36.112 --> 00:09:37.112
52-week high effect,

312
00:09:37.273 --> 00:09:38.654
particularly the industry one,

313
00:09:38.954 --> 00:09:46.880
was more pronounced for these stocks whose prices seemed more driven by industry-level information rather than just company-specific news.

314
00:09:47.040 --> 00:09:51.243
That really reinforces the industry information part of the paper's title.

315
00:09:51.770 --> 00:09:58.135
It seems the signal is strongest when reflecting broader industry sentiment or trends relative to past highs.

316
00:09:58.435 --> 00:09:59.036
Precisely.

317
00:09:59.596 --> 00:09:59.716
Now,

318
00:09:59.736 --> 00:10:02.138
they also looked at where this effect might be strongest,

319
00:10:02.158 --> 00:10:05.761
considering things like how much information is readily available about a company.

320
00:10:06.001 --> 00:10:06.261
Ah,

321
00:10:06.482 --> 00:10:07.943
like price informativeness,

322
00:10:08.483 --> 00:10:12.927
where maybe anchoring bias would be more likely if prices don't immediately reflect all news.

323
00:10:13.147 --> 00:10:13.647
Exactly.

324
00:10:13.967 --> 00:10:17.610
They hypothesized the strategy might work better for firms where prices are,

325
00:10:17.690 --> 00:10:17.991
let's say,

326
00:10:18.251 --> 00:10:19.772
less efficient or informative.

327
00:10:20.250 --> 00:10:21.170
Think smaller firms,

328
00:10:21.230 --> 00:10:21.910
younger firms,

329
00:10:21.970 --> 00:10:25.450
maybe stocks with higher trading friction or lower analyst coverage.

330
00:10:25.451 --> 00:10:26.170
And is that what they found?

331
00:10:26.490 --> 00:10:26.730
Yes.

332
00:10:27.450 --> 00:10:28.110
The industry,

333
00:10:29.190 --> 00:10:30.070
52-week high strategy,

334
00:10:30.550 --> 00:10:34.390
generated significantly higher profits among firms with these characteristics,

335
00:10:34.750 --> 00:10:35.450
small size,

336
00:10:35.730 --> 00:10:35.990
youth,

337
00:10:36.250 --> 00:10:37.130
high price impact,

338
00:10:37.570 --> 00:10:38.690
low analyst following,

339
00:10:39.030 --> 00:10:40.470
low institutional ownership.

340
00:10:41.090 --> 00:10:44.250
Places where information might travel slower or be harder to interpret.

341
00:10:44.410 --> 00:10:47.390
Which again points towards that behavioral underreaction explanation.

342
00:10:48.010 --> 00:10:48.130
If...

343
00:10:48.238 --> 00:10:50.178
If prices were perfectly efficient everywhere,

344
00:10:50.438 --> 00:10:51.278
you wouldn't expect this.

345
00:10:51.738 --> 00:10:51.998
Right.

346
00:10:52.498 --> 00:10:59.338
It suggests the strategy thrives where there is a bit more informational friction or potential for investors to anchor on past prices.

347
00:10:59.918 --> 00:11:01.598
What about practical implementation?

348
00:11:02.198 --> 00:11:04.018
Holding for six months and rebalancing?

349
00:11:04.498 --> 00:11:07.638
Did they check if the results were sensitive to that specific setup?

350
00:11:07.758 --> 00:11:10.778
Like what if you just bought and held or used different weighting?

351
00:11:11.138 --> 00:11:14.198
They did look at robustness to the rebalancing frequency and weighting.

352
00:11:14.490 --> 00:11:16.553
They found profitability persisted even with,

353
00:11:16.613 --> 00:11:16.794
say,

354
00:11:16.994 --> 00:11:19.097
a buy and hold approach after the initial sort,

355
00:11:19.378 --> 00:11:20.219
at least for a while.

356
00:11:20.620 --> 00:11:22.423
And while equal weighting was the main focus,

357
00:11:22.683 --> 00:11:25.007
they showed it also worked using value weighting,

358
00:11:25.107 --> 00:11:26.689
particularly within small stocks.

359
00:11:27.210 --> 00:11:30.956
So it doesn't seem entirely dependent on that exact six-month rebalance.

360
00:11:31.376 --> 00:11:33.059
And they did other robustness checks,

361
00:11:33.119 --> 00:11:33.299
too,

362
00:11:33.760 --> 00:11:33.900
like

363
00:11:33.974 --> 00:11:34.934
different time periods.

364
00:11:35.014 --> 00:11:35.234
Yeah,

365
00:11:35.294 --> 00:11:42.654
they sliced the data into subperiods and found the industry strategy consistently outperformed the individual one across different decades.

366
00:11:42.814 --> 00:11:44.694
Even excluding major market events.

367
00:11:44.754 --> 00:11:44.934
Yeah.

368
00:11:45.074 --> 00:11:47.994
Like the dot-com bubble or the 2008 crisis.

369
00:11:48.154 --> 00:11:48.394
Yes,

370
00:11:48.654 --> 00:11:49.314
they checked that too.

371
00:11:50.034 --> 00:11:53.354
Removing those extreme periods didn't eliminate the core finding.

372
00:11:53.494 --> 00:11:56.754
They also tested different holding periods like 3 months and 12 months,

373
00:11:57.114 --> 00:11:59.974
and the industry strategy still generally showed an an advantage.

374
00:12:00.262 --> 00:12:03.902
It sounds like a pretty robust finding across different tests and timeframes.

375
00:12:04.202 --> 00:12:08.862
It certainly seems that way based on their analysis over this long historical period.

376
00:12:08.942 --> 00:12:09.062
OK,

377
00:12:09.162 --> 00:12:11.222
so let's distill the key takeaways for you,

378
00:12:11.302 --> 00:12:11.782
the listener.

379
00:12:12.302 --> 00:12:13.762
What does this research really tell us?

380
00:12:14.302 --> 00:12:14.422
Well,

381
00:12:14.482 --> 00:12:14.802
first,

382
00:12:15.102 --> 00:12:16.282
this industry-focused

383
00:12:16.962 --> 00:12:18.282
52-week high trading rule,

384
00:12:18.502 --> 00:12:20.582
which uses publicly available price data,

385
00:12:21.102 --> 00:12:24.502
showed surprisingly strong and persistent returns historically,

386
00:12:24.922 --> 00:12:28.982
significantly more so than just looking at individual stocks near their highs.

387
00:12:29.530 --> 00:12:30.111
And second,

388
00:12:30.631 --> 00:12:34.254
the evidence seems to lean away from this just being about risk.

389
00:12:35.215 --> 00:12:37.777
It points more towards investor psychology specifically,

390
00:12:38.297 --> 00:12:40.979
anchoring or underreaction happening at the industry level.

391
00:12:41.199 --> 00:12:41.499
Right.

392
00:12:41.780 --> 00:12:43.221
So for your own analysis,

393
00:12:43.761 --> 00:12:46.463
focusing on industry price action relative to the

394
00:12:47.024 --> 00:12:52.208
52-week high might offer a more robust signal than the individual stock metric alone.

395
00:12:52.448 --> 00:12:53.028
And finally,

396
00:12:53.509 --> 00:12:56.591
it seems the strategy works particularly well where information might be

397
00:12:58.438 --> 00:13:00.578
A bit fuzzier or less efficiently priced,

398
00:13:00.598 --> 00:13:01.378
like in smaller,

399
00:13:01.578 --> 00:13:02.558
less followed companies.

400
00:13:02.778 --> 00:13:03.258
Exactly.

401
00:13:03.558 --> 00:13:08.518
Those seem to be the areas where this potential anchoring bias has the biggest impact on returns.

402
00:13:08.618 --> 00:13:12.838
It's a fascinating alternative lens through which to view market behavior and potential opportunities.

403
00:13:13.018 --> 00:13:14.338
Definitely provides food for thought.

404
00:13:14.658 --> 00:13:17.158
Thank you for tuning in to Papers with Backtest podcast.

405
00:13:17.378 --> 00:13:19.738
We hope today's deep dive gave you useful insights.

406
00:13:20.058 --> 00:13:21.998
Join us next time as we break down more research.

407
00:13:22.258 --> 00:13:23.658
And for more papers and backtests,

408
00:13:23.758 --> 00:13:27.158
find us at https.paperswithbacktest.com.

409
00:13:27.866 --> 00:13:28.507
Happy trading.

410
00:13:28.908 --> 00:13:30.450
And just a final thought to leave you with.

411
00:13:31.011 --> 00:13:34.056
Could you potentially enhance this kind of strategy by layering in,

412
00:13:34.657 --> 00:13:34.897
say,

413
00:13:35.298 --> 00:13:39.404
industry level sentiment analysis or news flow alongside the pure price signal?

414
00:13:40.065 --> 00:13:45.313
And what are the practical hurdles for an individual trader trying to implement such a broad industry based approach?

415
00:13:45.333 --> 00:13:45.874
Something to ponder.