WEBVTT

1
00:00:00.020 --> 00:00:00.200
Hello,

2
00:00:00.400 --> 00:00:02.301
welcome back to Papers with Backtest podcast.

3
00:00:02.461 --> 00:00:02.701
Today,

4
00:00:02.941 --> 00:00:05.359
we dive into another algo trading research paper.

5
00:00:05.703 --> 00:00:05.902
Today,

6
00:00:05.941 --> 00:00:07.680
we're digging into analyst coverage,

7
00:00:07.762 --> 00:00:10.219
information and bubbles by Andrade,

8
00:00:10.445 --> 00:00:16.219
Bian and Birch from the Journal of Financial and Quantitative Analysis published back in 2013.

9
00:00:16.220 --> 00:00:16.406
Right.

10
00:00:16.422 --> 00:00:18.344
And this one tackles a really interesting question.

11
00:00:19.078 --> 00:00:23.547
What role did analysts actually play in that pretty wild

12
00:00:23.891 --> 00:00:26.391
2007 stock market bubble over in China?

13
00:00:26.719 --> 00:00:27.203
Exactly.

14
00:00:27.453 --> 00:00:27.609
And,

15
00:00:27.672 --> 00:00:27.906
you know,

16
00:00:28.312 --> 00:00:29.594
more importantly for our listeners.

17
00:00:29.888 --> 00:00:35.453
What can we learn from that for maybe building trading rules or thinking about backtesting around those kinds of,

18
00:00:35.734 --> 00:00:35.894
well,

19
00:00:36.054 --> 00:00:37.214
extreme market conditions?

20
00:00:37.234 --> 00:00:37.355
Yeah,

21
00:00:37.356 --> 00:00:40.660
because the 2007 Chinese market was quite unique,

22
00:00:40.675 --> 00:00:41.019
wasn't it?

23
00:00:41.136 --> 00:00:41.261
Oh,

24
00:00:41.316 --> 00:00:41.878
absolutely.

25
00:00:41.941 --> 00:00:44.824
You had the government itself warning about a bubble forming.

26
00:00:44.925 --> 00:00:45.206
Really?

27
00:00:45.339 --> 00:00:45.870
The government?

28
00:00:45.871 --> 00:00:46.027
Yeah,

29
00:00:46.370 --> 00:00:49.167
and tons of new retail investors just piling in,

30
00:00:49.292 --> 00:00:51.527
plus major restrictions on short selling.

31
00:00:51.824 --> 00:00:52.027
Okay,

32
00:00:52.105 --> 00:00:54.245
so limited ways to bet against the market.

33
00:00:54.308 --> 00:00:54.855
Precisely.

34
00:00:55.011 --> 00:00:57.620
And it seems a lot of these new investors were actively looking,

35
00:00:57.777 --> 00:00:58.011
you know,

36
00:00:58.214 --> 00:00:59.511
seeking out analyst reports.

37
00:00:59.604 --> 00:01:00.885
probably trying to figure out what to do.

38
00:01:01.266 --> 00:01:01.426
Hmm.

39
00:01:01.666 --> 00:01:01.986
Interesting.

40
00:01:02.127 --> 00:01:03.670
So they weren't just passive observers.

41
00:01:04.389 --> 00:01:07.213
And the sort of headline finding we're going to explore is that

42
00:01:07.951 --> 00:01:09.537
SOX, with more analysts covering them,

43
00:01:09.732 --> 00:01:11.436
actually had smaller bubbles.

44
00:01:11.615 --> 00:01:12.678
That's the core finding,

45
00:01:12.740 --> 00:01:12.998
yes.

46
00:01:13.537 --> 00:01:14.217
Counterintuitive,

47
00:01:14.240 --> 00:01:14.600
perhaps,

48
00:01:14.701 --> 00:01:15.357
but that's what they found.

49
00:01:15.725 --> 00:01:15.959
Okay,

50
00:01:16.006 --> 00:01:16.834
so let's unpack this.

51
00:01:16.881 --> 00:01:18.162
When we say bubble here,

52
00:01:18.787 --> 00:01:22.303
what did that actually look like in the data for 2007 China?

53
00:01:22.787 --> 00:01:22.912
Well,

54
00:01:22.913 --> 00:01:25.131
the paper really paints a vivid picture.

55
00:01:25.209 --> 00:01:27.490
You saw P-E ratios just...

56
00:01:29.153 --> 00:01:52.823
skyrocketing trading activity turnover went through the roof huge volumes massive and loads of new retail trading accounts being open even things like google searches in china for stock market terms just spiked really textbook stuff sounds like classic exuberance across the board did they focus on a specific period for this bubble analysis they did they primarily looked at November 29th,

57
00:01:52.824 --> 00:01:54.557
2006 through to

58
00:01:54.995 --> 00:01:55.260
May 29th,

59
00:01:55.261 --> 00:01:56.760
2007.

60
00:01:56.776 --> 00:01:56.916
Okay,

61
00:01:56.948 --> 00:01:57.495
about six months.

62
00:01:57.736 --> 00:01:57.936
Right.

63
00:01:58.056 --> 00:02:03.022
And that end date is key because it was just before the government unexpectedly tripled the stamp duty,

64
00:02:03.202 --> 00:02:04.382
the security transaction tax.

65
00:02:04.421 --> 00:02:04.542
Oh,

66
00:02:04.543 --> 00:02:05.425
the tax hike.

67
00:02:05.546 --> 00:02:06.405
That must have been a shock.

68
00:02:06.686 --> 00:02:08.327
It definitely seemed to prick the bubble.

69
00:02:08.366 --> 00:02:08.530
Yeah.

70
00:02:08.968 --> 00:02:10.452
It was a major turning point.

71
00:02:10.585 --> 00:02:10.952
Got it.

72
00:02:11.553 --> 00:02:12.928
So a period of,

73
00:02:13.053 --> 00:02:13.374
let's say,

74
00:02:13.632 --> 00:02:16.928
rapid inflation and then a specific catalyst.

75
00:02:17.632 --> 00:02:17.772
Now,

76
00:02:18.178 --> 00:02:22.194
how did they measure the bubble intensity for individual stocks?

77
00:02:22.303 --> 00:02:22.928
Must have been tricky.

78
00:02:23.356 --> 00:02:23.496
Yeah,

79
00:02:23.516 --> 00:02:24.977
they didn't just look at the whole market.

80
00:02:25.038 --> 00:02:27.419
They used five different measures at the stock level.

81
00:02:27.923 --> 00:02:31.165
First was just the cumulative return over that six-month window.

82
00:02:31.462 --> 00:02:31.665
Okay,

83
00:02:31.743 --> 00:02:32.868
how much did stocks go up?

84
00:02:33.188 --> 00:02:33.688
On average,

85
00:02:33.704 --> 00:02:34.626
around 200%.

86
00:02:34.627 --> 00:02:35.165
They tripled.

87
00:02:35.329 --> 00:02:35.626
Wow,

88
00:02:35.930 --> 00:02:36.868
tripled in six months.

89
00:02:37.008 --> 00:02:37.188
Yep.

90
00:02:37.790 --> 00:02:41.618
So you can imagine if you were backtesting a simple trend-following strategy then,

91
00:02:42.008 --> 00:02:43.743
the returns would look amazing.

92
00:02:43.993 --> 00:02:44.477
Phenomenal,

93
00:02:44.602 --> 00:02:47.055
but maybe misleading given the context.

94
00:02:47.165 --> 00:02:47.305
Yeah.

95
00:02:47.462 --> 00:02:48.399
Screams bubble risk,

96
00:02:48.446 --> 00:02:48.665
right?

97
00:02:48.680 --> 00:02:49.133
Absolutely.

98
00:02:49.305 --> 00:02:49.977
Big asterisk,

99
00:02:50.008 --> 00:02:50.415
as you say.

100
00:02:50.880 --> 00:02:52.661
Second measure was the average P.E.

101
00:02:52.760 --> 00:02:54.100
ratio during that period.

102
00:02:54.321 --> 00:02:55.120
Price to earnings.

103
00:02:55.202 --> 00:02:55.460
Right.

104
00:02:55.702 --> 00:02:58.303
And they found very high average and median P.E.

105
00:02:58.401 --> 00:02:58.960
ratios.

106
00:02:59.382 --> 00:03:01.585
Clear signal of potential overvaluation.

107
00:03:01.725 --> 00:03:01.905
Okay.

108
00:03:01.944 --> 00:03:05.139
So value strategies would likely be flagging these stocks as expensive.

109
00:03:05.280 --> 00:03:05.663
Definitely.

110
00:03:05.664 --> 00:03:07.546
A flashing red light for value investors.

111
00:03:08.077 --> 00:03:09.671
Third was the announcement return.

112
00:03:09.702 --> 00:03:10.046
What's that?

113
00:03:10.249 --> 00:03:13.483
It's the return after that transaction tax hike was announced.

114
00:03:13.780 --> 00:03:14.421
Specifically,

115
00:03:14.593 --> 00:03:15.780
the five-day return.

116
00:03:16.046 --> 00:03:16.280
Ah.

117
00:03:16.483 --> 00:03:18.608
So how stocks reacted to the bad news.

118
00:03:18.733 --> 00:03:19.452
Exactly.

119
00:03:20.036 --> 00:03:26.301
The thinking is stocks inflated by speculation would get hit harder when trading suddenly became more expensive.

120
00:03:26.402 --> 00:03:26.883
Makes sense.

121
00:03:27.324 --> 00:03:28.644
Less reason to flip them quickly.

122
00:03:28.785 --> 00:03:29.746
And that's what they saw.

123
00:03:30.090 --> 00:03:34.168
Significant negative returns on average for stocks after the announcement.

124
00:03:34.449 --> 00:03:37.996
Which also tells you something important for backtesting,

125
00:03:38.059 --> 00:03:38.246
right?

126
00:03:38.699 --> 00:03:40.059
Transaction costs matter,

127
00:03:40.449 --> 00:03:41.652
especially sudden changes.

128
00:03:42.074 --> 00:03:42.574
Absolutely.

129
00:03:43.012 --> 00:03:45.121
Your models need to account for that possibility,

130
00:03:45.199 --> 00:03:46.668
especially around policy shifts.

131
00:03:46.777 --> 00:03:46.902
OK,

132
00:03:46.996 --> 00:03:47.387
what else?

133
00:03:47.388 --> 00:03:48.137
That's three measures.

134
00:03:48.388 --> 00:03:50.489
Fourth was a composite bubble measure.

135
00:03:50.729 --> 00:03:51.128
Basically,

136
00:03:51.169 --> 00:03:53.229
they combine the first three metrics return,

137
00:03:53.370 --> 00:03:53.749
PE,

138
00:03:54.011 --> 00:03:54.991
announcement reaction,

139
00:03:55.108 --> 00:03:57.788
into a single score for each stock's bubbliness.

140
00:03:58.233 --> 00:03:59.429
Kind of overall indicator.

141
00:03:59.507 --> 00:03:59.694
Yeah,

142
00:03:59.788 --> 00:04:00.671
just a holistic view.

143
00:04:01.030 --> 00:04:01.530
And fifth,

144
00:04:01.733 --> 00:04:03.647
they looked at the China HK premium.

145
00:04:04.054 --> 00:04:06.366
The price difference for stocks listed in both places.

146
00:04:06.585 --> 00:04:06.851
Right.

147
00:04:07.054 --> 00:04:09.179
For companies listed on both the mainland,

148
00:04:09.366 --> 00:04:11.694
Shanghai Shenzhen and Hong Kong exchanges.

149
00:04:12.116 --> 00:04:13.366
Since Hong Kong was more open,

150
00:04:13.538 --> 00:04:14.272
had short selling.

151
00:04:14.554 --> 00:04:17.007
The difference could reflect the mainland bubble.

152
00:04:17.280 --> 00:04:18.081
That was the idea,

153
00:04:18.481 --> 00:04:21.723
though the short selling limits in China made pure arbitrage tricky.

154
00:04:22.426 --> 00:04:22.785
Still,

155
00:04:23.004 --> 00:04:25.469
it's another data point on relative valuation.

156
00:04:25.664 --> 00:04:25.789
OK,

157
00:04:25.891 --> 00:04:30.274
so five different ways to gauge the bubble intensity per stock.

158
00:04:30.555 --> 00:04:31.758
Do they all point the same way?

159
00:04:32.016 --> 00:04:32.477
Pretty much.

160
00:04:32.555 --> 00:04:32.711
Yeah.

161
00:04:33.118 --> 00:04:35.602
The paper shows they were all significantly correlated,

162
00:04:36.118 --> 00:04:38.321
gave a consistent picture of which stocks were,

163
00:04:38.446 --> 00:04:38.805
let's say,

164
00:04:39.024 --> 00:04:39.664
frothier.

165
00:04:39.946 --> 00:04:40.243
Got it.

166
00:04:40.618 --> 00:04:41.461
So we have the bubble.

167
00:04:41.899 --> 00:04:42.868
We have ways to measure it.

168
00:04:42.869 --> 00:04:43.008
Now,

169
00:04:43.164 --> 00:04:43.789
the main event,

170
00:04:45.102 --> 00:04:45.914
analyst coverage.

171
00:04:46.764 --> 00:04:48.684
How did they measure that and what was the connection?

172
00:04:49.104 --> 00:04:50.844
They used a pretty straightforward measure,

173
00:04:50.965 --> 00:04:54.547
just the number of cell site analysts covering a specific stock.

174
00:04:54.864 --> 00:04:58.426
That was their proxy for how much public information was being produced about it.

175
00:04:58.465 --> 00:04:58.965
Simple count.

176
00:04:59.184 --> 00:05:00.286
And the key finding,

177
00:05:00.287 --> 00:05:01.309
the really interesting part,

178
00:05:01.567 --> 00:05:07.372
was a strong negative correlation between the number of analysts and all five of those bubble intensity measures.

179
00:05:07.653 --> 00:05:07.762
So,

180
00:05:08.075 --> 00:05:10.215
weight more analysts meant a smaller bubble.

181
00:05:10.977 --> 00:05:11.577
Less froth.

182
00:05:11.797 --> 00:05:12.277
Exactly.

183
00:05:12.498 --> 00:05:15.598
Stocks with greater analyst coverage experience smaller price run-ups,

184
00:05:15.676 --> 00:05:19.700
lower PEs during the period and less negative reaction to the tax hike.

185
00:05:20.020 --> 00:05:20.903
That's quite something.

186
00:05:21.700 --> 00:05:23.262
So the implication for,

187
00:05:23.598 --> 00:05:23.762
say,

188
00:05:23.997 --> 00:05:26.239
backtesting trading rules around bubbles,

189
00:05:27.395 --> 00:05:33.239
maybe filtering by analyst coverage could identify less risky assets.

190
00:05:33.661 --> 00:05:35.286
That's definitely a potential insight.

191
00:05:35.661 --> 00:05:40.380
It suggests that stocks under more scrutiny might have had less extreme behavior.

192
00:05:40.909 --> 00:05:43.892
More information flow seemed to dampen the speculative excess.

193
00:05:43.933 --> 00:05:45.695
Did they check if it was just a size effect?

194
00:05:45.996 --> 00:05:46.234
You know,

195
00:05:46.273 --> 00:05:48.800
bigger companies get more coverage and are less bubbly anyway?

196
00:05:49.175 --> 00:05:49.699
They did.

197
00:05:49.957 --> 00:05:52.722
They controlled for market capitalization and other factors,

198
00:05:52.839 --> 00:05:54.003
and the relationship held.

199
00:05:54.464 --> 00:05:56.386
It wasn't solely a large-cap phenomenon.

200
00:05:56.785 --> 00:05:58.644
Analyst attention itself seemed to matter.

201
00:05:58.785 --> 00:05:58.941
Huh.

202
00:05:59.425 --> 00:06:00.050
That's important.

203
00:06:00.410 --> 00:06:00.566
Okay,

204
00:06:00.567 --> 00:06:02.738
but what about when analysts don't agree?

205
00:06:03.816 --> 00:06:05.066
That must muddy the waters.

206
00:06:05.207 --> 00:06:06.238
It absolutely does.

207
00:06:06.300 --> 00:06:07.222
They looked into that too,

208
00:06:07.628 --> 00:06:10.300
measuring the dispersion or disagreement among analysts.

209
00:06:10.521 --> 00:06:13.403
based on their earnings forecasts and their buy-sell recommendations.

210
00:06:13.503 --> 00:06:15.444
So how spread out were their opinions?

211
00:06:15.648 --> 00:06:15.905
Right.

212
00:06:16.007 --> 00:06:21.468
And they found that the bubble-mitigating effect of analyst coverage was weaker when there was more disagreement among them.

213
00:06:21.812 --> 00:06:21.976
Ah,

214
00:06:22.452 --> 00:06:22.694
okay.

215
00:06:23.030 --> 00:06:24.773
So just counting analysts isn't enough,

216
00:06:24.952 --> 00:06:26.296
if they're all saying different things.

217
00:06:26.593 --> 00:06:29.624
Then the coverage doesn't seem to have the same stabilizing effect.

218
00:06:30.280 --> 00:06:34.218
The lack of consensus maybe reduces the coordinating power of the information.

219
00:06:34.546 --> 00:06:35.765
That makes intuitive sense.

220
00:06:36.421 --> 00:06:37.312
For trading rules,

221
00:06:37.374 --> 00:06:37.609
then...

222
00:06:38.121 --> 00:06:40.224
High disagreement could be another red flag,

223
00:06:40.544 --> 00:06:42.187
maybe signaling more uncertainty,

224
00:06:42.605 --> 00:06:44.109
even if a stock looks well covered.

225
00:06:44.589 --> 00:06:45.288
Precisely.

226
00:06:45.609 --> 00:06:52.116
It suggests you might need to look beyond just the raw number of analysts and consider the consensus or lack thereof.

227
00:06:52.234 --> 00:06:52.554
Interesting.

228
00:06:52.898 --> 00:06:55.437
Did they also look at trading activity like turnover?

229
00:06:55.546 --> 00:06:56.171
How did that relate?

230
00:06:56.265 --> 00:06:56.484
Yes,

231
00:06:56.546 --> 00:06:56.859
they did.

232
00:06:57.234 --> 00:07:01.905
They found that more analyst coverage was generally associated with lower stock turnover.

233
00:07:01.984 --> 00:07:02.609
Lower trading.

234
00:07:02.702 --> 00:07:03.249
Why would that be?

235
00:07:03.537 --> 00:07:07.863
The interpretation is that maybe more available information leads to more informed,

236
00:07:07.961 --> 00:07:13.043
perhaps buy and hold type decisions and less speculative churning or flipping of shares.

237
00:07:13.105 --> 00:07:13.746
More conviction,

238
00:07:13.848 --> 00:07:14.832
less noise trading.

239
00:07:15.066 --> 00:07:15.590
Possibly.

240
00:07:16.332 --> 00:07:16.809
But again,

241
00:07:16.887 --> 00:07:18.254
mirroring the other findings,

242
00:07:18.332 --> 00:07:22.816
this turnover reducing effective coverage was also weaker when analyst disagreement was high.

243
00:07:23.035 --> 00:07:23.160
OK,

244
00:07:23.254 --> 00:07:25.832
so another potential signal for backtesting.

245
00:07:26.754 --> 00:07:27.863
During bubbly times,

246
00:07:28.504 --> 00:07:31.941
high turnover combined with low coverage or high disagreement.

247
00:07:32.673 --> 00:07:35.074
might flag more speculative,

248
00:07:35.374 --> 00:07:36.394
riskier stocks.

249
00:07:36.673 --> 00:07:38.513
That seems like a very reasonable takeaway,

250
00:07:38.576 --> 00:07:38.935
yes.

251
00:07:39.556 --> 00:07:39.673
Now,

252
00:07:39.674 --> 00:07:43.755
the paper also digs into why coverage might have this dampening effect.

253
00:07:44.318 --> 00:07:46.036
The obvious first guess is...

254
00:07:46.318 --> 00:07:47.958
That analysts were warning everyone,

255
00:07:48.622 --> 00:07:49.701
putting out cell ratings,

256
00:07:50.779 --> 00:07:51.701
the voice of reason.

257
00:07:51.982 --> 00:07:52.544
You'd think so.

258
00:07:53.091 --> 00:07:53.810
But surprisingly,

259
00:07:53.857 --> 00:07:56.716
when they analyzed the actual recommendations over time,

260
00:07:56.763 --> 00:07:59.997
they didn't see a big shift towards cell ratings around the bubble's peak.

261
00:08:00.357 --> 00:08:00.482
No.

262
00:08:01.341 --> 00:08:02.321
They weren't getting more bearish.

263
00:08:02.641 --> 00:08:03.201
If anything,

264
00:08:03.202 --> 00:08:05.883
the average recommendation became slightly more optimistic,

265
00:08:05.941 --> 00:08:06.441
not less.

266
00:08:06.801 --> 00:08:06.922
Huh.

267
00:08:07.324 --> 00:08:09.320
So it wasn't analysts acting as a reality check,

268
00:08:09.500 --> 00:08:11.781
explicitly telling people things were overpriced.

269
00:08:12.000 --> 00:08:13.726
Doesn't seem like that was the main driver,

270
00:08:13.765 --> 00:08:13.883
no,

271
00:08:14.265 --> 00:08:17.719
which kind of weakens the idea that they were just reducing over-optimism directly.

272
00:08:17.937 --> 00:08:19.562
So if it wasn't explicit warnings,

273
00:08:19.906 --> 00:08:23.562
what's the alternative explanation for why more coverage meant smaller bubbles?

274
00:08:24.109 --> 00:08:26.828
The authors leaned towards a belief coordination mechanism.

275
00:08:27.293 --> 00:08:31.137
The idea is that having more information out there being discussed and analyzed.

276
00:08:31.456 --> 00:08:33.217
Even if it wasn't uniformly negative.

277
00:08:33.440 --> 00:08:33.780
Right.

278
00:08:34.241 --> 00:08:36.842
It helped investors converge towards more similar,

279
00:08:36.944 --> 00:08:39.842
maybe more anchored views of a stock's fundamental value.

280
00:08:40.225 --> 00:08:45.186
It reduced the scope for wildly divergent expectations that can fuel a speculative frenzy.

281
00:08:45.592 --> 00:08:45.858
Ah,

282
00:08:45.999 --> 00:08:46.327
I see.

283
00:08:46.686 --> 00:08:54.405
So it's less about the content buy-sell and more about the process of information dissemination leading to more aligned beliefs.

284
00:08:54.624 --> 00:08:55.905
That seems to be their main argument.

285
00:08:56.053 --> 00:09:00.774
More shared information reduces the heterogeneity of beliefs that bubbles often thrive on.

286
00:09:00.832 --> 00:09:02.133
That's a more subtle mechanism.

287
00:09:02.914 --> 00:09:03.075
Now,

288
00:09:03.254 --> 00:09:06.895
the big question with studies like this is always cause and effect,

289
00:09:06.934 --> 00:09:07.137
right?

290
00:09:07.215 --> 00:09:07.817
Endogeneity.

291
00:09:08.715 --> 00:09:11.075
Could less bubbly stocks just attract more analysts?

292
00:09:11.278 --> 00:09:12.770
A valid concern always.

293
00:09:13.020 --> 00:09:14.410
They tackled this quite carefully.

294
00:09:14.942 --> 00:09:15.223
First,

295
00:09:15.270 --> 00:09:18.098
they used lag analyst coverage from 2005,

296
00:09:18.099 --> 00:09:19.879
well before the main bubble period kicked off.

297
00:09:20.332 --> 00:09:25.504
And they still found that 2005 coverage negatively predicted the bubble intensity in 2007.

298
00:09:25.957 --> 00:09:28.419
That helps mitigate reverse causality concerns.

299
00:09:28.640 --> 00:09:28.820
Okay,

300
00:09:28.900 --> 00:09:31.244
using past coverage predicts future bubble size.

301
00:09:31.544 --> 00:09:31.744
Yeah.

302
00:09:31.745 --> 00:09:31.982
What else?

303
00:09:32.263 --> 00:09:34.709
They also used instrumental variables,

304
00:09:34.966 --> 00:09:37.091
things that likely affect analyst coverage,

305
00:09:37.169 --> 00:09:40.552
but probably not bubble intensity directly except through coverage.

306
00:09:40.568 --> 00:09:41.373
What kind of variables?

307
00:09:41.513 --> 00:09:43.794
They used trading volume and mutual fund ownership,

308
00:09:44.091 --> 00:09:45.263
again from 2005.

309
00:09:45.935 --> 00:09:47.091
Using these instruments,

310
00:09:47.154 --> 00:09:51.779
they still found evidence supporting a causal link from coverage to smaller bubbles.

311
00:09:51.841 --> 00:09:52.060
Right,

312
00:09:52.138 --> 00:09:54.029
using IV techniques adds more weight.

313
00:09:54.405 --> 00:09:55.827
Did they consider other stories?

314
00:09:56.048 --> 00:09:59.395
Like maybe analysts are just lazy and cover safer stocks,

315
00:09:59.953 --> 00:10:03.157
or perhaps it's about pump and dump schemes in the low coverage stocks.

316
00:10:03.529 --> 00:10:06.052
They did address those alternative explanations,

317
00:10:06.213 --> 00:10:07.072
lazy analysts,

318
00:10:07.152 --> 00:10:08.615
institutional selling patterns,

319
00:10:08.732 --> 00:10:09.295
pump and dump,

320
00:10:09.795 --> 00:10:14.076
and argued that the beta didn't really line up strongly with those being the primary drivers.

321
00:10:14.396 --> 00:10:14.521
OK,

322
00:10:14.662 --> 00:10:16.404
so the core findings seems pretty robust.

323
00:10:16.802 --> 00:10:19.107
Let's try to synthesize the key insights for someone,

324
00:10:19.248 --> 00:10:19.482
you know,

325
00:10:19.841 --> 00:10:22.466
building trading rules or running back tests today,

326
00:10:22.935 --> 00:10:25.451
especially if they're looking at potentially frothy markets.

327
00:10:25.748 --> 00:10:25.951
Sure.

328
00:10:25.998 --> 00:10:27.607
I think there are several practical points.

329
00:10:27.795 --> 00:10:28.185
First,

330
00:10:28.607 --> 00:10:30.638
if you see conditions resembling a bubble,

331
00:10:30.779 --> 00:10:31.873
rapid price run ups.

332
00:10:32.321 --> 00:10:36.204
High volume may be considered analyst coverage as a potential risk filter.

333
00:10:36.585 --> 00:10:39.269
So lower coverage might signal higher bubble susceptibility.

334
00:10:39.827 --> 00:10:42.050
Could be a reason to underweight or avoid.

335
00:10:42.355 --> 00:10:42.855
Potentially,

336
00:10:42.909 --> 00:10:43.175
yes.

337
00:10:43.769 --> 00:10:44.136
Second,

338
00:10:44.691 --> 00:10:46.933
don't stop at just the number of analysts.

339
00:10:47.378 --> 00:10:50.073
Look at the dispersion in their forecasts or recommendations.

340
00:10:50.245 --> 00:10:53.402
Because high disagreement might negate the stabilizing effect.

341
00:10:53.464 --> 00:10:54.089
Exactly.

342
00:10:54.245 --> 00:10:58.027
High coverage with low disagreement might be the most stabilizing combination.

343
00:10:58.573 --> 00:11:00.495
High disagreement could still signal trouble.

344
00:11:00.855 --> 00:11:01.152
Got it.

345
00:11:01.308 --> 00:11:01.636
What else?

346
00:11:02.357 --> 00:11:02.637
Third,

347
00:11:02.918 --> 00:11:04.159
pay attention to turnover.

348
00:11:04.759 --> 00:11:11.730
High turnover in stocks with low coverage or high disagreement during these periods could be a warning sign of excess speculation.

349
00:11:11.886 --> 00:11:13.090
Another potential risk flag.

350
00:11:13.324 --> 00:11:13.652
Fourth,

351
00:11:14.269 --> 00:11:17.293
that transaction tax hike reaction is a stark reminder.

352
00:11:17.933 --> 00:11:21.980
Bubble prone assets can be very sensitive to trading costs and policy changes.

353
00:11:22.402 --> 00:11:25.902
So robust transaction cost modeling and backtests is crucial,

354
00:11:26.105 --> 00:11:27.777
especially anticipating potential shifts.

355
00:11:28.027 --> 00:11:28.418
Definitely.

356
00:11:28.793 --> 00:11:29.414
And finally,

357
00:11:29.574 --> 00:11:33.416
while the China HK premium was hard to trade directly due to restrictions there.

358
00:11:33.596 --> 00:11:35.299
It hints at potential valuation gaps.

359
00:11:35.760 --> 00:11:36.022
Right.

360
00:11:36.400 --> 00:11:39.322
It highlights that cross-market discrepancies can exist,

361
00:11:39.447 --> 00:11:40.463
especially in bubbles,

362
00:11:40.963 --> 00:11:45.869
and might offer opportunities in less restricted markets or if new instruments become available.

363
00:11:46.291 --> 00:11:46.416
OK,

364
00:11:46.729 --> 00:11:48.072
that's a lot of food for thought.

365
00:11:48.635 --> 00:11:57.197
So the big picture takeaway seems to be that public information proxied here by analysts coverage wasn't just noise during the 2007 China bubble.

366
00:11:57.717 --> 00:11:59.577
It actually seemed to play a moderating role.

367
00:11:59.778 --> 00:11:59.998
Yeah,

368
00:12:00.217 --> 00:12:02.617
likely by helping coordinate investor beliefs,

369
00:12:02.797 --> 00:12:05.817
even if the analysts themselves weren't explicitly bearish.

370
00:12:06.176 --> 00:12:11.660
It suggests information flow can be a subtle but important factor in market stability or instability.

371
00:12:11.918 --> 00:12:12.778
Fascinating stuff.

372
00:12:13.184 --> 00:12:19.840
Definitely relevant for thinking about risk and potential signals when navigating potentially overheated markets in our backtests and strategies.

373
00:12:20.028 --> 00:12:24.824
It really underscores how complex market dynamics can be during these extreme periods.

374
00:12:25.365 --> 00:12:27.627
Thank you for tuning into Papers with Backtest podcast.

375
00:12:27.988 --> 00:12:30.168
We hope today's episode gave you useful insights.

376
00:12:30.590 --> 00:12:32.633
Join us next time as we break down more research.

377
00:12:33.215 --> 00:12:34.472
And for more papers and backtests,

378
00:12:34.496 --> 00:12:37.660
find us at https.paperswithbacktest.com.

379
00:12:38.277 --> 00:12:38.855
Happy trading.