WEBVTT

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

2
00:00:00.300 --> 00:00:02.242
welcome back to Papers with Backtest podcast.

3
00:00:02.623 --> 00:00:05.344
Today we dive into another algo trading research paper.

4
00:00:05.566 --> 00:00:11.352
Today we're digging into shared analyst coverage unifying momentum spillover effects.

5
00:00:11.648 --> 00:00:14.117
It's by Ali and Hirschleifer from 2019.

6
00:00:14.391 --> 00:00:14.531
Right.

7
00:00:14.555 --> 00:00:18.375
And this one seems to touch on quite a few areas looking at those GEL codes,

8
00:00:18.578 --> 00:00:19.953
G12 asset pricing,

9
00:00:20.438 --> 00:00:22.078
G14 market efficiency.

10
00:00:22.500 --> 00:00:23.000
Exactly.

11
00:00:23.250 --> 00:00:23.500
And,

12
00:00:23.609 --> 00:00:23.844
you know,

13
00:00:23.859 --> 00:00:26.172
G24 financial institutions.

14
00:00:26.516 --> 00:00:28.718
It really revolves around momentum spillovers,

15
00:00:28.798 --> 00:00:31.400
how one stock's momentum might predict another's.

16
00:00:31.802 --> 00:00:33.181
They call it CF momentum.

17
00:00:33.384 --> 00:00:33.544
Yeah,

18
00:00:33.662 --> 00:00:35.005
CF for connected firm momentum.

19
00:00:35.021 --> 00:00:36.888
The whole idea is linking firms,

20
00:00:36.904 --> 00:00:39.732
but specifically through analyst co-coverage.

21
00:00:39.810 --> 00:00:40.005
Okay,

22
00:00:40.209 --> 00:00:40.951
let's unpack that.

23
00:00:41.349 --> 00:00:42.169
So the core idea,

24
00:00:42.373 --> 00:00:43.255
if I'm getting this right,

25
00:00:43.318 --> 00:00:46.834
is that all these different ways momentum seems to spill over,

26
00:00:47.209 --> 00:00:49.552
like between industries or firms in the same region,

27
00:00:49.959 --> 00:00:50.505
supply chains.

28
00:00:50.506 --> 00:00:51.380
Just here in tech,

29
00:00:51.396 --> 00:00:51.646
yeah,

30
00:00:51.943 --> 00:00:52.380
all of those.

31
00:00:52.755 --> 00:00:55.490
The paper suggests maybe they aren't separate things after all.

32
00:00:55.927 --> 00:00:56.052
but

33
00:00:56.152 --> 00:00:58.434
they all trace back to shared analyst coverage.

34
00:00:58.655 --> 00:00:59.596
That's the hypothesis.

35
00:00:59.735 --> 00:01:04.596
The analyst covering the same set of firms is sort of the underlying connection driving these spillovers we see.

36
00:01:05.003 --> 00:01:05.378
Interesting.

37
00:01:05.401 --> 00:01:11.104
So our mission for this deep dive then is to really get into the weeds of this CF momentum strategy.

38
00:01:11.385 --> 00:01:11.549
Yeah.

39
00:01:11.784 --> 00:01:13.409
We need to understand the actual trading rules.

40
00:01:13.534 --> 00:01:13.753
Right.

41
00:01:13.862 --> 00:01:17.768
How they built it step by step and crucially what the back tests showed.

42
00:01:18.206 --> 00:01:21.393
And how it stacks up against the other momentum effects we already know about.

43
00:01:21.690 --> 00:01:22.643
It sounds like it could be a,

44
00:01:23.299 --> 00:01:23.440
well,

45
00:01:23.643 --> 00:01:24.706
simpler way to think about.

46
00:01:24.828 --> 00:01:25.889
cross-asset momentum,

47
00:01:25.929 --> 00:01:26.149
maybe.

48
00:01:26.350 --> 00:01:26.889
Potentially,

49
00:01:27.010 --> 00:01:27.190
yeah.

50
00:01:27.370 --> 00:01:28.913
A more unified view instead of,

51
00:01:29.034 --> 00:01:29.291
you know,

52
00:01:29.354 --> 00:01:30.893
a whole zoo of different effects.

53
00:01:31.135 --> 00:01:31.315
Okay,

54
00:01:31.316 --> 00:01:32.116
so first things first,

55
00:01:32.354 --> 00:01:33.635
defining connected firms.

56
00:01:33.979 --> 00:01:36.362
The key is this analyst co-coverage,

57
00:01:36.377 --> 00:01:36.596
right?

58
00:01:37.237 --> 00:01:38.838
How exactly do they define that?

59
00:01:39.002 --> 00:01:39.143
Yep,

60
00:01:39.479 --> 00:01:40.081
that's central.

61
00:01:40.581 --> 00:01:44.409
Two stocks are connected if there's at least one analyst who covered both of them.

62
00:01:44.643 --> 00:01:45.752
And covered means?

63
00:01:45.971 --> 00:01:48.737
It means the analyst issued at least one earnings forecast,

64
00:01:48.799 --> 00:01:50.956
like for the current or next fiscal year,

65
00:01:51.440 --> 00:01:52.721
sometime in the past 12 months.

66
00:01:53.184 --> 00:01:57.908
They used IBES data from late 83 to late 2015.

67
00:01:58.428 --> 00:02:01.389
What's the significance of using this specific definition?

68
00:02:02.076 --> 00:02:03.576
Why not just use industry groups?

69
00:02:03.897 --> 00:02:04.053
Well,

70
00:02:04.092 --> 00:02:06.193
the authors argue it's more granular,

71
00:02:06.295 --> 00:02:07.576
more firm specific.

72
00:02:07.959 --> 00:02:10.162
It's based on what analysts actually do,

73
00:02:10.381 --> 00:02:12.225
which firms they choose to follow together.

74
00:02:12.568 --> 00:02:17.084
It might capture subtler fundamental links than just being in the same broad sector.

75
00:02:17.209 --> 00:02:17.428
Okay,

76
00:02:17.490 --> 00:02:18.068
that makes sense.

77
00:02:18.428 --> 00:02:21.318
So once you know which germs are connected to which,

78
00:02:21.852 --> 00:02:23.053
How do you calculate this

79
00:02:23.574 --> 00:02:25.616
CF portfolio return for a stock?

80
00:02:25.755 --> 00:02:25.977
Right,

81
00:02:26.036 --> 00:02:27.137
the CFRET.

82
00:02:27.419 --> 00:02:28.540
For any given stock,

83
00:02:28.637 --> 00:02:29.481
say Stocki,

84
00:02:29.919 --> 00:02:34.661
its CFRET is basically a weighted average return of all the stocks it's connected to,

85
00:02:34.723 --> 00:02:35.583
all the stock Js.

86
00:02:35.762 --> 00:02:36.544
Weighted average?

87
00:02:36.708 --> 00:02:37.755
How are the weights determined?

88
00:02:37.942 --> 00:02:41.770
It's based on the number of analysts covering both Stocki and Stockj.

89
00:02:42.598 --> 00:02:45.989
So if a connected Stockj shares more analysts with Stocki,

90
00:02:46.130 --> 00:02:48.973
it gets a higher weight in that average return calculation.

91
00:02:49.176 --> 00:02:49.317
Ah,

92
00:02:49.583 --> 00:02:49.801
okay.

93
00:02:50.248 --> 00:02:54.052
The idea being that more shared analysts signifies a stronger connection.

94
00:02:54.112 --> 00:02:54.634
Precisely.

95
00:02:54.770 --> 00:02:56.235
And just for context,

96
00:02:56.236 --> 00:02:59.458
they found the average stock was connected to about 86 other firms.

97
00:02:59.661 --> 00:03:00.458
86.

98
00:03:00.520 --> 00:03:00.856
Wow.

99
00:03:01.075 --> 00:03:01.278
Yeah.

100
00:03:01.802 --> 00:03:02.700
And they also noted,

101
00:03:02.778 --> 00:03:03.802
which isn't too surprising,

102
00:03:03.958 --> 00:03:06.044
that analysts tend to cover larger stocks.

103
00:03:06.083 --> 00:03:08.770
So these connected firms often had higher market caps.

104
00:03:08.989 --> 00:03:09.114
OK,

105
00:03:09.224 --> 00:03:09.786
so we have this

106
00:03:10.208 --> 00:03:16.989
CFRAT number for every stock reflecting the recent performance of its analyst linked peers.

107
00:03:17.692 --> 00:03:18.661
How do you trade on that?

108
00:03:18.974 --> 00:03:19.661
What's the strategy?

109
00:03:20.116 --> 00:03:22.498
It's a pretty standard momentum approach in structure.

110
00:03:22.779 --> 00:03:23.680
At the end of every month,

111
00:03:23.820 --> 00:03:26.441
you take all your stocks and you sort them into quintiles,

112
00:03:26.543 --> 00:03:30.402
five groups based purely on their CFRET over the past one month.

113
00:03:30.769 --> 00:03:33.832
So rank them by how well their connected firms did last month.

114
00:03:33.887 --> 00:03:34.410
Exactly.

115
00:03:34.652 --> 00:03:36.309
Then you form a long short portfolio.

116
00:03:36.574 --> 00:03:38.074
You go along the top quintile,

117
00:03:38.137 --> 00:03:39.996
the stocks whose connected firms did best,

118
00:03:40.059 --> 00:03:41.699
and you short the bottom quintile,

119
00:03:41.918 --> 00:03:43.559
the ones whose connected firms did worst.

120
00:03:43.699 --> 00:03:44.918
And rebalance monthly.

121
00:03:44.919 --> 00:03:45.621
Rebalance monthly,

122
00:03:45.668 --> 00:03:45.855
yes.

123
00:03:45.856 --> 00:03:46.137
All right.

124
00:03:46.215 --> 00:03:46.918
The crucial part,

125
00:03:47.730 --> 00:03:48.465
the back test.

126
00:03:49.412 --> 00:03:51.255
What happened when they ran this in the U.S.

127
00:03:51.314 --> 00:03:51.653
market?

128
00:03:51.835 --> 00:03:52.435
Did it make money?

129
00:03:52.614 --> 00:03:52.735
Oh,

130
00:03:52.774 --> 00:03:52.956
yeah.

131
00:03:53.415 --> 00:03:55.978
The results were pretty compelling.

132
00:03:56.017 --> 00:03:57.040
They looked at the alphas,

133
00:03:57.060 --> 00:04:00.079
the excess returns after accounting for the usual suspects,

134
00:04:00.142 --> 00:04:00.821
market risk,

135
00:04:00.899 --> 00:04:01.243
size,

136
00:04:01.282 --> 00:04:01.767
value,

137
00:04:01.907 --> 00:04:02.423
momentum,

138
00:04:03.126 --> 00:04:04.524
the FAMA French factor.

139
00:04:04.689 --> 00:04:05.431
Four-factor model.

140
00:04:05.493 --> 00:04:05.774
Right.

141
00:04:06.149 --> 00:04:08.649
And even adding a fifth factor for short-term reversal,

142
00:04:08.696 --> 00:04:11.415
the strategy still showed significant positive alphas.

143
00:04:11.618 --> 00:04:12.774
How significant are we talking?

144
00:04:12.775 --> 00:04:13.743
What were the numbers like?

145
00:04:14.040 --> 00:04:14.196
Well,

146
00:04:14.415 --> 00:04:15.837
for the valuated portfolio,

147
00:04:15.978 --> 00:04:18.806
the five-factor alpha was 1.19%.

148
00:04:18.932 --> 00:04:19.432
per month.

149
00:04:20.032 --> 00:04:20.713
That's substantial.

150
00:04:20.832 --> 00:04:22.172
The T-stat was way up there,

151
00:04:22.232 --> 00:04:23.033
6.71.

152
00:04:23.252 --> 00:04:24.752
1.19%

153
00:04:24.853 --> 00:04:25.252
per month.

154
00:04:25.275 --> 00:04:26.115
That's huge.

155
00:04:26.392 --> 00:04:26.775
It is.

156
00:04:27.037 --> 00:04:29.131
And the equal weighted version was even stronger,

157
00:04:29.310 --> 00:04:30.396
2.1%

158
00:04:30.678 --> 00:04:31.498
per month alpha,

159
00:04:31.717 --> 00:04:33.092
T-stat of almost 12.

160
00:04:33.318 --> 00:04:33.576
Wow.

161
00:04:34.076 --> 00:04:36.849
And did the profits show up on both the long and short sides?

162
00:04:37.412 --> 00:04:37.693
Yes.

163
00:04:37.834 --> 00:04:40.349
They mentioned it was profitable on both legs of the trade,

164
00:04:40.350 --> 00:04:41.490
which is always a good sign.

165
00:04:42.178 --> 00:04:44.178
And there was a clear monotonic relationship.

166
00:04:44.179 --> 00:04:46.334
The higher the CFRET quintile,

167
00:04:46.335 --> 00:04:47.943
the better the future stock performance.

168
00:04:48.325 --> 00:04:49.301
Did the effect last?

169
00:04:49.647 --> 00:04:50.947
Or was it just a one month blip?

170
00:04:51.227 --> 00:04:51.868
That's another key,

171
00:04:51.928 --> 00:04:52.907
finding it persisted.

172
00:04:52.987 --> 00:04:57.010
The positive returns continued over the next 11 months after forming the portfolio.

173
00:04:57.127 --> 00:04:59.366
So it wasn't just immediate mean reversion or something?

174
00:04:59.452 --> 00:04:59.569
No,

175
00:04:59.709 --> 00:05:00.108
exactly.

176
00:05:00.233 --> 00:05:01.670
Over a full 12 months,

177
00:05:01.913 --> 00:05:06.288
the cumulative return for the long short strategy was 3.21%

178
00:05:06.289 --> 00:05:09.256
value weighted and 6.68%

179
00:05:09.319 --> 00:05:09.788
equal weighted.

180
00:05:10.006 --> 00:05:13.772
This persistence really points towards market under reaction to this information.

181
00:05:13.991 --> 00:05:17.756
The market is slow to price in the implications of how connected firms are doing.

182
00:05:18.095 --> 00:05:19.056
That seems to be the story,

183
00:05:19.156 --> 00:05:19.376
yes.

184
00:05:20.157 --> 00:05:22.058
And they noted that if they used deciles,

185
00:05:22.218 --> 00:05:23.659
10 groups instead of five,

186
00:05:23.702 --> 00:05:25.421
the results were even more pronounced.

187
00:05:25.604 --> 00:05:25.804
Okay,

188
00:05:25.882 --> 00:05:28.386
so CF momentum looks strong on its own.

189
00:05:29.229 --> 00:05:32.667
But the paper's big claim is that it unifies other effects.

190
00:05:33.190 --> 00:05:34.058
How did they test that?

191
00:05:34.073 --> 00:05:34.792
How does it compare?

192
00:05:34.964 --> 00:05:35.151
Right,

193
00:05:35.167 --> 00:05:36.370
this is where it gets really interesting.

194
00:05:36.417 --> 00:05:37.933
They did spanning tests.

195
00:05:38.229 --> 00:05:42.339
They looked at seven other known cross-asset momentum anomalies.

196
00:05:42.386 --> 00:05:43.604
Like industry momentum.

197
00:05:43.667 --> 00:05:43.839
Yep,

198
00:05:43.995 --> 00:05:44.417
industry,

199
00:05:44.636 --> 00:05:45.761
geographic momentum,

200
00:05:46.058 --> 00:05:47.245
customer momentum.

201
00:05:47.587 --> 00:05:49.209
supplier industry momentum,

202
00:05:49.589 --> 00:05:52.171
even links between single and multi-segment firms,

203
00:05:52.510 --> 00:05:53.874
and technology momentum.

204
00:05:53.995 --> 00:05:54.694
Quite a list.

205
00:05:54.733 --> 00:05:54.854
OK,

206
00:05:55.053 --> 00:05:56.495
so they built factors for all of these.

207
00:05:56.698 --> 00:05:57.120
They did,

208
00:05:57.214 --> 00:05:59.401
using similar methods ranking based on linked firms'

209
00:05:59.417 --> 00:06:00.221
past returns,

210
00:06:00.542 --> 00:06:02.120
forming long-short portfolios.

211
00:06:02.339 --> 00:06:02.659
And then,

212
00:06:03.159 --> 00:06:05.768
did CF momentum get explained away by these others,

213
00:06:05.799 --> 00:06:06.971
or did it do the explaining?

214
00:06:07.065 --> 00:06:08.331
It was definitely the latter.

215
00:06:08.424 --> 00:06:08.846
First off,

216
00:06:08.862 --> 00:06:11.393
the CF momentum factor itself was highly profitable,

217
00:06:11.518 --> 00:06:13.112
even against the five-factor model,

218
00:06:13.174 --> 00:06:14.237
1.68%

219
00:06:14.299 --> 00:06:14.956
alpha per month.

220
00:06:15.440 --> 00:06:16.503
But the crucial part

221
00:06:17.039 --> 00:06:19.902
When they added the CF momentum factor to the regressions,

222
00:06:19.943 --> 00:06:22.783
trying to explain the returns of those other seven momentum factors,

223
00:06:23.267 --> 00:06:24.326
their alphas disappeared.

224
00:06:24.408 --> 00:06:24.888
Pretty much.

225
00:06:25.166 --> 00:06:26.048
They became small,

226
00:06:26.470 --> 00:06:28.072
statistically insignificant,

227
00:06:28.126 --> 00:06:29.173
sometimes even negative.

228
00:06:29.869 --> 00:06:34.080
It strongly suggested that CF momentum was capturing the same underlying phenomenon.

229
00:06:34.595 --> 00:06:35.501
It subsumed them.

230
00:06:36.111 --> 00:06:37.517
While the reverse wasn't true,

231
00:06:38.173 --> 00:06:40.595
the other factors couldn't explain CF momentum.

232
00:06:40.705 --> 00:06:41.126
Correct.

233
00:06:41.361 --> 00:06:46.564
CF momentum remained highly profitable even when controlling for those other factors.

234
00:06:46.695 --> 00:06:50.479
It seems to be capturing something more fundamental that drives these other effects.

235
00:06:50.999 --> 00:06:52.561
It's like finding the common denominator.

236
00:06:52.682 --> 00:06:53.581
That's a good way to put it.

237
00:06:53.682 --> 00:06:55.546
They also did cross-sectional regressions,

238
00:06:55.585 --> 00:06:55.780
right?

239
00:06:56.007 --> 00:06:58.710
Looking at predicting individual stock returns.

240
00:06:58.803 --> 00:06:59.046
Yes,

241
00:06:59.249 --> 00:07:00.592
Fama-McBeth regressions.

242
00:07:01.030 --> 00:07:02.272
And the story was consistent.

243
00:07:02.811 --> 00:07:07.171
Past one month CF return was a strong predictor of next month's stock return.

244
00:07:07.530 --> 00:07:10.655
And adding it to the model weakened the other momentum variables.

245
00:07:10.999 --> 00:07:11.624
Significantly.

246
00:07:12.139 --> 00:07:13.467
Especially for larger stocks,

247
00:07:13.483 --> 00:07:13.858
things like

248
00:07:14.019 --> 00:07:15.000
industry momentum,

249
00:07:15.240 --> 00:07:16.442
geographic momentum,

250
00:07:16.741 --> 00:07:17.724
customer momentum,

251
00:07:18.185 --> 00:07:22.185
lost their predictive power once you accounted for the stock's CF return.

252
00:07:22.388 --> 00:07:22.567
Okay,

253
00:07:22.606 --> 00:07:24.106
so strong results in the U.S.

254
00:07:24.528 --> 00:07:25.208
Did this travel?

255
00:07:25.474 --> 00:07:26.810
What about international markets?

256
00:07:27.231 --> 00:07:27.513
Yes,

257
00:07:27.630 --> 00:07:29.052
they tested it internationally too.

258
00:07:29.356 --> 00:07:33.481
Found strong CF momentum pretty much across the board in developed markets.

259
00:07:33.638 --> 00:07:33.903
Really?

260
00:07:34.060 --> 00:07:34.560
Same pattern?

261
00:07:34.685 --> 00:07:34.856
Yeah,

262
00:07:35.013 --> 00:07:39.044
significant alphas in 10 out of the 11 developed countries they looked at,

263
00:07:39.403 --> 00:07:41.825
even after controlling for local industry momentum.

264
00:07:42.466 --> 00:07:42.669
And

265
00:07:42.783 --> 00:07:43.424
And interestingly,

266
00:07:43.644 --> 00:07:44.525
just like in the U.S.,

267
00:07:45.064 --> 00:07:49.869
adding CF momentum often made the industry momentum factor insignificant in those countries.

268
00:07:49.990 --> 00:07:54.072
So CF momentum seems to dominate industry momentum internationally as well.

269
00:07:54.431 --> 00:07:55.072
It appears so.

270
00:07:55.455 --> 00:07:58.619
They even showed a profitable global ex-U.S.

271
00:07:58.939 --> 00:08:00.486
CF momentum strategy.

272
00:08:00.861 --> 00:08:05.299
Were the results sensitive to how they define things like industries or the time period?

273
00:08:05.752 --> 00:08:06.533
They checked that,

274
00:08:06.814 --> 00:08:08.346
used different industry definitions,

275
00:08:08.361 --> 00:08:11.908
different geographic definitions like state level and the results held up.

276
00:08:12.167 --> 00:08:17.053
They also split their sample period in half and CF momentum was strong in both halves.

277
00:08:17.311 --> 00:08:18.494
Seems quite robust.

278
00:08:18.592 --> 00:08:21.955
Did they compare it to any other similar ideas like maybe peer momentum?

279
00:08:22.275 --> 00:08:22.533
Yes,

280
00:08:22.619 --> 00:08:27.783
they did compare it to a measure from Israelson's 2016 paper on peer momentum.

281
00:08:28.104 --> 00:08:34.174
The CF momentum strategy still generated significant alpha even when benchmarked against that Israelson measure.

282
00:08:34.752 --> 00:08:35.252
In fact,

283
00:08:35.346 --> 00:08:36.283
in their analysis,

284
00:08:36.518 --> 00:08:39.721
the Israelson measure itself wasn't showing significant returns.

285
00:08:40.039 --> 00:08:40.480
So again,

286
00:08:40.621 --> 00:08:42.344
see if momentum seemed to subsume it.

287
00:08:42.526 --> 00:08:42.725
Okay,

288
00:08:42.766 --> 00:08:43.870
let's talk practicalities.

289
00:08:44.311 --> 00:08:45.170
Trading costs.

290
00:08:46.006 --> 00:08:49.090
Monthly rebalancing sounds like it could generate high turnover.

291
00:08:49.170 --> 00:08:49.769
It does,

292
00:08:49.830 --> 00:08:51.451
especially for the one-month strategy.

293
00:08:51.552 --> 00:08:52.853
They acknowledged the high turnover,

294
00:08:52.892 --> 00:08:55.517
but they calculated the break-even transaction costs.

295
00:08:55.892 --> 00:08:59.197
Meaning how high costs could be before wiping out the profit.

296
00:08:59.298 --> 00:08:59.759
Exactly.

297
00:09:00.041 --> 00:09:01.564
And the levels suggested that,

298
00:09:01.705 --> 00:09:01.884
well,

299
00:09:01.939 --> 00:09:03.439
for large institutions,

300
00:09:03.939 --> 00:09:06.533
large arbitragers with low trading costs,

301
00:09:06.580 --> 00:09:08.705
it could still be profitable net of fees.

302
00:09:09.002 --> 00:09:10.377
What about the longer-term effect?

303
00:09:10.378 --> 00:09:10.877
You mentioned

304
00:09:10.950 --> 00:09:12.431
And it persisted for 12 months.

305
00:09:12.512 --> 00:09:12.712
Right.

306
00:09:12.713 --> 00:09:15.794
And they pointed out that if you implemented a strategy based on,

307
00:09:16.016 --> 00:09:16.255
say,

308
00:09:16.895 --> 00:09:18.938
12-month CF momentum instead of one month,

309
00:09:19.098 --> 00:09:20.520
the turnover would be much lower.

310
00:09:20.762 --> 00:09:23.544
And potentially similar net returns after costs.

311
00:09:23.559 --> 00:09:24.059
Potentially,

312
00:09:24.223 --> 00:09:24.481
yes.

313
00:09:25.005 --> 00:09:26.106
Lower gross alpha maybe,

314
00:09:26.286 --> 00:09:28.684
but much lower transaction cost drag.

315
00:09:28.887 --> 00:09:29.762
So wrapping this up,

316
00:09:30.309 --> 00:09:35.372
the big picture seems to be that this shared analyst coverage idea isn't just another factor,

317
00:09:35.887 --> 00:09:38.356
but potentially a unifying explanation.

318
00:09:38.654 --> 00:09:41.116
for a lot of momentum spillover effects we see.

319
00:09:41.497 --> 00:09:42.880
That really is the core message.

320
00:09:43.176 --> 00:09:49.044
Shared analyst coverage identifies this potent CF momentum effect that seems to drive,

321
00:09:49.208 --> 00:09:50.208
or at least absorb,

322
00:09:50.528 --> 00:09:54.012
many previously documented cross-asset momentum anomalies.

323
00:09:54.512 --> 00:09:58.536
And the back tests show it's generated significant persistent returns.

324
00:09:58.926 --> 00:09:59.348
And for you,

325
00:09:59.395 --> 00:09:59.911
the listener,

326
00:10:00.411 --> 00:10:02.505
trying to get a handle on all these different market effects,

327
00:10:03.036 --> 00:10:06.520
this research suggests a potentially more parsimonious view.

328
00:10:07.194 --> 00:10:11.734
Focusing on analyst connections might give you a simpler way to think about these complex relationships.

329
00:10:11.773 --> 00:10:11.996
Right.

330
00:10:12.094 --> 00:10:14.957
Instead of tracking a dozen different types of momentum spillover,

331
00:10:15.176 --> 00:10:18.660
maybe focusing on this analyst linkage gives you a more fundamental signal.

332
00:10:18.777 --> 00:10:24.136
The real aha moment here is just how powerful that seemingly simple metric,

333
00:10:24.324 --> 00:10:26.371
which analysts cover which stocks turns out to be,

334
00:10:26.511 --> 00:10:28.996
that it could explain this whole zoo of momentum effects is,

335
00:10:29.152 --> 00:10:29.277
well,

336
00:10:29.293 --> 00:10:30.074
it's quite surprising.

337
00:10:30.496 --> 00:10:31.496
It definitely is.

338
00:10:31.574 --> 00:10:33.324
And it leaves you with something to think about.

339
00:10:33.822 --> 00:10:37.766
How could you use this idea of shared analyst coverage in your own analysis?

340
00:10:38.407 --> 00:10:42.371
Could looking at these connections offer a more holistic view of momentum?

341
00:10:42.848 --> 00:10:46.957
And does that apparent underreaction the paper finds signal real opportunities?

342
00:10:47.153 --> 00:10:48.239
Definitely food for thought.

343
00:10:48.637 --> 00:10:51.106
Thank you for tuning in to Papers with Backtest podcast.

344
00:10:51.262 --> 00:10:53.387
We hope today's episode gave you useful insights.

345
00:10:53.824 --> 00:10:55.903
Join us next time as we break down more research.

346
00:10:56.559 --> 00:10:57.824
And for more papers and backtests,

347
00:10:57.825 --> 00:11:00.309
find us at https.paperswithbacktest.com.

348
00:11:00.621 --> 00:11:01.246
Happy trading.