WEBVTT

1
00:00:00.020 --> 00:00:00.180
Hello,

2
00:00:00.300 --> 00:00:02.422
welcome back to Papers with Backtest podcast.

3
00:00:02.762 --> 00:00:03.004
Today,

4
00:00:03.043 --> 00:00:05.406
we dive into another Algo Trading research paper.

5
00:00:05.867 --> 00:00:06.125
Today,

6
00:00:06.207 --> 00:00:11.969
we're unpacking a paper titled Advertising Effect Within Stocks by Thomas Cheminor and Ann Yan.

7
00:00:12.836 --> 00:00:16.094
The version we're looking at is from February 2009.

8
00:00:16.172 --> 00:00:16.312
Okay,

9
00:00:16.578 --> 00:00:18.125
Advertising Effect Within Stocks.

10
00:00:18.797 --> 00:00:20.000
So at its heart,

11
00:00:20.406 --> 00:00:22.484
what were these researchers trying to understand?

12
00:00:22.516 --> 00:00:23.062
What was the,

13
00:00:23.203 --> 00:00:23.422
you know,

14
00:00:23.423 --> 00:00:23.906
the puzzle?

15
00:00:24.384 --> 00:00:24.584
Yeah,

16
00:00:24.664 --> 00:00:30.268
the central question was really about how a company's advertising spending actually impacts its stock returns.

17
00:00:30.850 --> 00:00:33.714
They were digging into whether there's a real connection,

18
00:00:33.893 --> 00:00:35.636
not just like immediately,

19
00:00:36.237 --> 00:00:37.917
but also what happens down the line,

20
00:00:38.018 --> 00:00:40.081
the longer term consequences for investors.

21
00:00:40.159 --> 00:00:40.362
Right.

22
00:00:40.425 --> 00:00:41.440
And for you tuning in,

23
00:00:41.800 --> 00:00:45.018
maybe you're looking to get a quick handle on what matters in this kind of research.

24
00:00:45.081 --> 00:00:49.409
Get those aha moments without wading through loads of dense text.

25
00:00:50.096 --> 00:00:54.065
This deep dive is really focused on pulling out the crucial bits from this paper.

26
00:00:54.244 --> 00:00:58.124
Especially anything that looks like a potential trading rule or insight from their back test.

27
00:00:58.226 --> 00:00:58.683
Exactly.

28
00:00:58.726 --> 00:01:00.144
We want those practical takeaways.

29
00:01:00.566 --> 00:01:01.706
And just as a quick preview,

30
00:01:01.824 --> 00:01:08.464
what we found interesting is how the paper suggests this interplay between advertising,

31
00:01:08.746 --> 00:01:09.871
investor attention,

32
00:01:10.503 --> 00:01:11.941
and then the stock's performance.

33
00:01:11.972 --> 00:01:14.128
It sounds like there are some surprising twists.

34
00:01:14.488 --> 00:01:14.863
Definitely.

35
00:01:14.972 --> 00:01:19.269
Could have some real implications for how you might think about trading strategies based on,

36
00:01:19.410 --> 00:01:19.675
say,

37
00:01:20.347 --> 00:01:21.472
advertising spend.

38
00:01:21.597 --> 00:01:21.847
All right.

39
00:01:21.925 --> 00:01:22.691
Let's get into it then.

40
00:01:22.847 --> 00:01:22.972
Yeah.

41
00:01:23.112 --> 00:01:24.073
The core finding,

42
00:01:24.313 --> 00:01:27.838
what was the main discovery connecting advertising and stock returns?

43
00:01:28.119 --> 00:01:28.318
Okay,

44
00:01:28.377 --> 00:01:29.838
so the primary finding is,

45
00:01:30.017 --> 00:01:30.220
well,

46
00:01:30.681 --> 00:01:31.220
it's quite neat,

47
00:01:31.299 --> 00:01:31.603
actually.

48
00:01:31.900 --> 00:01:35.822
They found that when a company spends more on advertising in a particular year,

49
00:01:36.385 --> 00:01:39.135
its stock tends to perform better in that same year.

50
00:01:39.322 --> 00:01:39.463
Okay,

51
00:01:39.510 --> 00:01:40.369
like a spotlight effect,

52
00:01:40.431 --> 00:01:40.603
maybe.

53
00:01:40.853 --> 00:01:41.244
Sort of,

54
00:01:41.275 --> 00:01:41.463
yeah.

55
00:01:42.135 --> 00:01:43.306
But here's the twist.

56
00:01:43.838 --> 00:01:46.197
In the year after that big advertising push,

57
00:01:46.447 --> 00:01:49.931
those same stocks tend to deliver lower returns.

58
00:01:50.213 --> 00:01:50.353
Ah,

59
00:01:50.603 --> 00:01:50.822
okay,

60
00:01:50.853 --> 00:01:52.119
so it's short-term gain.

61
00:01:52.760 --> 00:01:54.401
but maybe long term pain.

62
00:01:54.821 --> 00:01:55.643
That's a good way to put it,

63
00:01:55.663 --> 00:01:58.005
like a sugar rush followed by a bit of a slowdown.

64
00:01:58.006 --> 00:01:59.067
Can we put some numbers on that?

65
00:01:59.184 --> 00:02:00.466
How big is this effect they found?

66
00:02:00.606 --> 00:02:00.825
Sure.

67
00:02:01.052 --> 00:02:02.692
They use Fama-McBeth's regressions.

68
00:02:02.989 --> 00:02:04.013
It's a pretty standard,

69
00:02:04.075 --> 00:02:07.091
robust way to look at this stuff across lots of stocks in time.

70
00:02:07.395 --> 00:02:07.614
Right.

71
00:02:08.153 --> 00:02:14.575
And they found that a one standard deviation increase in advertising spend in a year was linked to about a 3.35%

72
00:02:14.747 --> 00:02:16.341
higher stock return during that year.

73
00:02:17.309 --> 00:02:18.544
3.35%.

74
00:02:18.545 --> 00:02:19.512
That's not insignificant.

75
00:02:19.872 --> 00:02:19.981
No,

76
00:02:20.059 --> 00:02:20.700
it's noticeable.

77
00:02:21.036 --> 00:02:22.698
But here's the interesting part for traders.

78
00:02:23.339 --> 00:02:26.862
That same increase in advertising was associated with a decrease of about

79
00:02:27.362 --> 00:02:29.425
2.92% in the stock's return,

80
00:02:29.862 --> 00:02:32.065
specifically in the second half of the following year.

81
00:02:32.386 --> 00:02:32.589
Wow.

82
00:02:32.690 --> 00:02:32.909
Okay.

83
00:02:32.925 --> 00:02:34.909
So that initial boost really seems to,

84
00:02:35.511 --> 00:02:35.690
well,

85
00:02:35.768 --> 00:02:36.573
reverse later on.

86
00:02:36.675 --> 00:02:37.456
Pretty significantly,

87
00:02:37.487 --> 00:02:37.675
yeah.

88
00:02:37.831 --> 00:02:37.956
Now,

89
00:02:38.050 --> 00:02:39.393
a key question is always,

90
00:02:40.737 --> 00:02:42.300
is this just random noise,

91
00:02:42.753 --> 00:02:44.597
or is it really about the advertising itself?

92
00:02:45.081 --> 00:02:47.222
Did they check if other known factors like

93
00:02:47.360 --> 00:02:48.741
company size or value?

94
00:02:48.862 --> 00:02:49.423
Explain this.

95
00:02:49.522 --> 00:02:50.083
Absolutely.

96
00:02:50.223 --> 00:02:50.864
Critical point.

97
00:02:50.884 --> 00:02:51.124
And yes,

98
00:02:51.143 --> 00:02:51.503
they did.

99
00:02:51.624 --> 00:02:53.186
They controlled for things like size,

100
00:02:53.225 --> 00:02:54.128
book-to-market ratio,

101
00:02:54.866 --> 00:02:56.210
the value measure and momentum.

102
00:02:56.593 --> 00:02:58.428
Factors we already know predict returns.

103
00:02:58.811 --> 00:03:00.413
And even after accounting for those,

104
00:03:00.452 --> 00:03:01.593
this advertising effect,

105
00:03:01.655 --> 00:03:02.913
both the up and the down,

106
00:03:03.311 --> 00:03:05.077
it remains statistically significant.

107
00:03:05.436 --> 00:03:08.843
So that suggests advertising has its own sort of independent impact here.

108
00:03:09.077 --> 00:03:10.343
That's what the evidence points to.

109
00:03:10.405 --> 00:03:10.561
Yeah.

110
00:03:10.639 --> 00:03:13.452
It's not just a proxy for being a certain size or type of company.

111
00:03:13.593 --> 00:03:13.780
Okay.

112
00:03:13.874 --> 00:03:14.686
So we have this pattern.

113
00:03:15.311 --> 00:03:16.014
Advertising up.

114
00:03:16.484 --> 00:03:17.324
Short-term return up,

115
00:03:17.364 --> 00:03:19.145
longer-term return down.

116
00:03:19.466 --> 00:03:19.845
Why?

117
00:03:20.284 --> 00:03:21.585
What's the theory behind this?

118
00:03:21.725 --> 00:03:26.143
The authors put forward what they called the investor attention hypothesis.

119
00:03:26.206 --> 00:03:26.464
Okay.

120
00:03:26.909 --> 00:03:28.604
The basic idea is pretty straightforward.

121
00:03:29.104 --> 00:03:30.925
When a company advertises more,

122
00:03:31.050 --> 00:03:33.550
it simply grabs more attention from investors.

123
00:03:33.565 --> 00:03:34.815
It becomes more visible.

124
00:03:35.081 --> 00:03:35.628
That makes sense.

125
00:03:35.643 --> 00:03:36.534
You see the ads,

126
00:03:36.535 --> 00:03:37.331
you think about the company,

127
00:03:37.456 --> 00:03:38.253
maybe look up the stock.

128
00:03:38.471 --> 00:03:39.034
Exactly.

129
00:03:39.315 --> 00:03:44.393
And they suggest this increased attention can push prices up short-term in a couple of ways.

130
00:03:44.925 --> 00:03:45.253
First,

131
00:03:45.472 --> 00:03:50.037
just more eyeballs might mean more buying pressure as people notice it.

132
00:03:50.177 --> 00:03:50.876
Simple awareness.

133
00:03:50.978 --> 00:03:51.259
Right.

134
00:03:51.619 --> 00:03:51.978
Second,

135
00:03:52.040 --> 00:03:58.126
more attention might lead to more disagreement or heterogeneity in beliefs about the stock's future.

136
00:03:58.806 --> 00:04:00.384
And if it's hard to short the stock,

137
00:04:00.470 --> 00:04:03.986
the price tends to reflect the more optimistic views pushing it up.

138
00:04:04.329 --> 00:04:04.595
Gotcha.

139
00:04:05.189 --> 00:04:06.111
So the buzz builds,

140
00:04:06.533 --> 00:04:07.814
optimists drive the price.

141
00:04:08.361 --> 00:04:09.595
But why the later reversal?

142
00:04:09.658 --> 00:04:10.595
Why does it come back down?

143
00:04:10.954 --> 00:04:11.142
Well,

144
00:04:11.579 --> 00:04:14.923
the hypothesis suggests that this attention spike is often temporary.

145
00:04:15.396 --> 00:04:16.577
The ad campaigns end,

146
00:04:16.877 --> 00:04:18.137
the news cycle moves on.

147
00:04:18.377 --> 00:04:19.497
The novelty wears off.

148
00:04:19.657 --> 00:04:20.297
Precisely.

149
00:04:20.918 --> 00:04:23.754
And as that initial surge of attention fades,

150
00:04:23.895 --> 00:04:26.200
the stock price tends to correct downwards,

151
00:04:26.598 --> 00:04:28.958
closer to maybe what the fundamentals justify.

152
00:04:29.379 --> 00:04:31.497
The temporary boost wasn't sustainable.

153
00:04:31.622 --> 00:04:31.793
Okay.

154
00:04:31.840 --> 00:04:33.418
The market sort of recalibrates.

155
00:04:33.434 --> 00:04:37.762
So how do they actually try to measure this investor attention?

156
00:04:38.346 --> 00:04:39.247
Sounds a bit fuzzy.

157
00:04:39.627 --> 00:04:39.868
Yeah,

158
00:04:39.988 --> 00:04:41.229
you can't just ask everyone.

159
00:04:41.389 --> 00:04:43.289
They used a couple of common proxies,

160
00:04:43.332 --> 00:04:45.532
things we can measure that likely correlate with attention.

161
00:04:45.813 --> 00:04:46.954
First was trading volume.

162
00:04:47.172 --> 00:04:48.094
How much the stock trades.

163
00:04:48.157 --> 00:04:48.657
Exactly.

164
00:04:48.774 --> 00:04:49.375
Specifically,

165
00:04:49.797 --> 00:04:51.399
exchange-adjusted turnover.

166
00:04:51.539 --> 00:04:55.047
The idea being more attention equals more trading activity.

167
00:04:55.375 --> 00:05:00.235
The second was the number of financial analysts covering the firm using IBES data.

168
00:05:00.469 --> 00:05:00.672
Right.

169
00:05:00.735 --> 00:05:03.032
More analysts covering it usually means more visibility.

170
00:05:03.204 --> 00:05:03.891
That's the thinking.

171
00:05:03.938 --> 00:05:04.875
And they found a clear link.

172
00:05:05.046 --> 00:05:10.408
Higher ad spending in a year did lead to higher trading volume and more analyst coverage that same year.

173
00:05:10.748 --> 00:05:10.869
OK,

174
00:05:11.029 --> 00:05:14.127
so advertising does seem to attract attention by these measures.

175
00:05:14.228 --> 00:05:14.767
Looks like it.

176
00:05:14.986 --> 00:05:16.392
But the key thing for us,

177
00:05:16.752 --> 00:05:17.588
thinking about trading,

178
00:05:18.025 --> 00:05:19.486
is whether attention itself,

179
00:05:19.713 --> 00:05:20.791
however it's generated,

180
00:05:21.064 --> 00:05:23.767
causes this up-then-down pattern in returns.

181
00:05:24.173 --> 00:05:24.470
Right.

182
00:05:24.611 --> 00:05:26.236
And their analysis showed just that.

183
00:05:26.423 --> 00:05:27.298
Increased attention,

184
00:05:27.517 --> 00:05:28.845
whether from ads or something else,

185
00:05:28.908 --> 00:05:32.252
was linked to higher returns now but lower returns later.

186
00:05:32.782 --> 00:05:36.646
Which strengthens the whole attention-based explanation for the advertising effect.

187
00:05:36.666 --> 00:05:37.166
Exactly.

188
00:05:37.287 --> 00:05:40.748
It suggests the rise and fall of attention is a major driver here.

189
00:05:40.928 --> 00:05:41.092
Okay,

190
00:05:41.146 --> 00:05:42.615
let's get to the really practical stuff.

191
00:05:43.053 --> 00:05:44.310
This is Papers with Backtest,

192
00:05:44.396 --> 00:05:44.756
after all.

193
00:05:45.498 --> 00:05:50.240
What did they actually do in terms of backtesting or simulating trading strategies based on this?

194
00:05:50.756 --> 00:05:52.506
So they used portfolio sorts,

195
00:05:52.740 --> 00:05:53.584
standard approach.

196
00:05:54.209 --> 00:05:54.662
Each year,

197
00:05:54.818 --> 00:06:00.021
they ranked all the stocks based on the change in their advertising spending compared to the previous year.

198
00:06:00.162 --> 00:06:00.303
Okay,

199
00:06:00.350 --> 00:06:01.193
so add VT,

200
00:06:01.428 --> 00:06:01.928
the change.

201
00:06:02.068 --> 00:06:02.271
Right.

202
00:06:02.674 --> 00:06:04.015
They grouped them into 10 buckets,

203
00:06:04.135 --> 00:06:04.677
deciles.

204
00:06:04.876 --> 00:06:07.157
Decile 10 had the biggest increase in advertising.

205
00:06:07.497 --> 00:06:09.981
Decile 1 had the biggest decrease or smallest increase.

206
00:06:10.282 --> 00:06:10.583
Got it.

207
00:06:10.739 --> 00:06:12.020
Top group ramped up spending,

208
00:06:12.083 --> 00:06:12.606
bottom group.

209
00:06:12.708 --> 00:06:12.887
Hmm.

210
00:06:13.364 --> 00:06:13.481
Did.

211
00:06:13.762 --> 00:06:14.302
Precisely.

212
00:06:14.809 --> 00:06:20.067
And when they looked at the raw returns in the advertising change happened the advertising year,

213
00:06:20.645 --> 00:06:21.692
the difference was stark.

214
00:06:22.239 --> 00:06:23.395
The top decile,

215
00:06:23.411 --> 00:06:24.317
the big increasers,

216
00:06:24.536 --> 00:06:28.598
outperformed the bottom decile by about 16.8%

217
00:06:28.599 --> 00:06:29.145
on average.

218
00:06:29.302 --> 00:06:29.442
Wow.

219
00:06:29.443 --> 00:06:31.645
16.8%.

220
00:06:31.714 --> 00:06:32.575
That's huge.

221
00:06:33.416 --> 00:06:34.457
Definitely catches the eye.

222
00:06:34.557 --> 00:06:35.617
It absolutely does.

223
00:06:35.998 --> 00:06:36.279
But,

224
00:06:36.479 --> 00:06:37.760
and this is crucial for trading rules,

225
00:06:38.139 --> 00:06:38.963
look what happened next.

226
00:06:38.964 --> 00:06:39.643
The reversal.

227
00:06:39.783 --> 00:06:39.901
Yep.

228
00:06:40.385 --> 00:06:43.104
In the six months immediately after that advertising year,

229
00:06:43.322 --> 00:06:44.307
the top decile,

230
00:06:44.308 --> 00:06:45.166
the previous winners,

231
00:06:45.283 --> 00:06:49.174
actually underperformed the bottom decile by around 9.2%.

232
00:06:49.190 --> 00:06:49.424
Ouch.

233
00:06:49.565 --> 00:06:49.815
Okay.

234
00:06:50.096 --> 00:06:50.830
And it continued.

235
00:06:51.205 --> 00:06:52.268
In the next six months,

236
00:06:52.269 --> 00:06:54.737
so months 7 to 12 after the advertising year,

237
00:06:54.830 --> 00:06:56.362
they underperformed again by about

238
00:06:56.705 --> 00:07:00.143
9.4%. So that massive initial outperformance.

239
00:07:00.474 --> 00:07:04.573
just completely reverses and turns into significant underperformance over the next year.

240
00:07:04.796 --> 00:07:06.675
That's what the raw portfolio sorts show,

241
00:07:06.714 --> 00:07:08.452
a really clear reversal pattern.

242
00:07:08.499 --> 00:07:10.077
That's a massive takeaway for anyone thinking,

243
00:07:10.140 --> 00:07:10.257
oh,

244
00:07:10.374 --> 00:07:11.054
big ad spend,

245
00:07:11.077 --> 00:07:11.655
let's buy.

246
00:07:11.757 --> 00:07:12.280
Absolutely.

247
00:07:12.374 --> 00:07:12.499
Now,

248
00:07:12.515 --> 00:07:16.194
they also checked if this was just because the high advertising firms were different in size or value,

249
00:07:16.515 --> 00:07:18.030
controlled for size and book to market.

250
00:07:18.077 --> 00:07:18.999
Did the effect disappear?

251
00:07:19.405 --> 00:07:19.515
No.

252
00:07:20.358 --> 00:07:22.343
The size of the effect got a bit smaller,

253
00:07:22.624 --> 00:07:23.358
as you'd expect,

254
00:07:23.452 --> 00:07:24.421
but the pattern held.

255
00:07:25.124 --> 00:07:26.155
Even after adjusting,

256
00:07:26.233 --> 00:07:29.687
the top ad increase groups still outperformed in the advertising year,

257
00:07:30.042 --> 00:07:31.704
by about 9.6 percent,

258
00:07:32.204 --> 00:07:39.689
but then underperformed significantly in the next two six-month periods by around 8.4 percent and 8.9 percent.

259
00:07:39.728 --> 00:07:41.712
Still substantial underperformance later on.

260
00:07:41.814 --> 00:07:42.290
Definitely.

261
00:07:42.314 --> 00:07:43.657
What about more advanced models,

262
00:07:43.853 --> 00:07:44.735
like Fama French?

263
00:07:45.196 --> 00:07:46.399
Did those confirm the reversal?

264
00:07:46.665 --> 00:07:46.884
Yes,

265
00:07:47.181 --> 00:07:51.899
they ran Fama French three-factor and Carhartt four-factor models on these portfolios.

266
00:07:52.446 --> 00:07:53.649
The results were consistent.

267
00:07:53.946 --> 00:07:59.274
The high advertising increase portfolio showed significant negative abnormal returns alpha in the six months

268
00:07:59.538 --> 00:08:01.519
and the full year after the advertising year.

269
00:08:01.660 --> 00:08:03.822
Meaning even after accounting for market risk,

270
00:08:04.143 --> 00:08:04.521
size,

271
00:08:04.725 --> 00:08:05.103
value,

272
00:08:05.162 --> 00:08:05.725
and momentum.

273
00:08:06.166 --> 00:08:09.928
There was still this extra underperformance linked to the prior advertising surge.

274
00:08:10.209 --> 00:08:10.670
For instance,

275
00:08:10.671 --> 00:08:11.631
the long-short portfolio,

276
00:08:11.912 --> 00:08:12.912
P10 minus P1,

277
00:08:13.592 --> 00:08:15.998
had a three-factor alpha of menacing at 8.4%

278
00:08:16.217 --> 00:08:18.561
over the next six months and amicable 11.3%

279
00:08:18.639 --> 00:08:19.295
over the next year.

280
00:08:19.639 --> 00:08:19.811
Okay,

281
00:08:19.812 --> 00:08:23.779
so the backtests really drive home that this isn't a simple buy-on-add-spend signal.

282
00:08:24.107 --> 00:08:25.998
The reversal seems very real and robust.

283
00:08:26.311 --> 00:08:27.748
The data strongly suggests

284
00:08:27.850 --> 00:08:30.893
any gains are temporary and likely followed by a payback period.

285
00:08:31.454 --> 00:08:31.575
Now,

286
00:08:31.576 --> 00:08:32.875
you mentioned attention earlier.

287
00:08:33.317 --> 00:08:36.317
The paper looked at whether the persistence of that attention mattered,

288
00:08:36.356 --> 00:08:36.622
right?

289
00:08:36.700 --> 00:08:36.997
Yes,

290
00:08:37.098 --> 00:08:38.418
that was another interesting layer.

291
00:08:38.840 --> 00:08:44.348
They found the initial positive kick in the advertising year was stronger for stocks that got more attention,

292
00:08:44.707 --> 00:08:45.286
more volume,

293
00:08:45.348 --> 00:08:46.832
more analysts that year.

294
00:08:47.129 --> 00:08:47.770
Makes sense.

295
00:08:48.317 --> 00:08:49.129
But crucially,

296
00:08:49.254 --> 00:08:53.661
the negative reversal later on was also stronger for these high attention stocks,

297
00:08:54.489 --> 00:08:56.645
unless that high attention persisted.

298
00:08:56.942 --> 00:08:57.317
Ah.

299
00:08:57.774 --> 00:09:01.654
So if the trading volume and analyst coverage stayed high into the next year...

300
00:09:01.955 --> 00:09:03.537
Then the negative reversal was weaker.

301
00:09:03.834 --> 00:09:06.459
The underperformance wasn't as bad if the buzz stuck around.

302
00:09:06.537 --> 00:09:07.716
That adds a lot of nuance.

303
00:09:08.177 --> 00:09:13.638
Maybe a potential trading rule isn't just fade the ad spike but feed the ad spike unless the attention seems sticky.

304
00:09:14.357 --> 00:09:17.373
It certainly suggests a more sophisticated approach might be needed.

305
00:09:17.795 --> 00:09:23.138
You might want to monitor not just the ad change but the subsequent sustained change in volume or analyst coverage.

306
00:09:23.420 --> 00:09:23.966
Fascinating.

307
00:09:24.570 --> 00:09:28.220
What about things that make it harder for smart traders to correct mispricing?

308
00:09:28.854 --> 00:09:30.275
Like arbitrage costs.

309
00:09:31.056 --> 00:09:31.918
Did that play a role?

310
00:09:31.998 --> 00:09:32.498
It did.

311
00:09:32.797 --> 00:09:38.226
They used proxies for arbitrage costs like illiquidity and idiosyncratic risk,

312
00:09:38.281 --> 00:09:38.765
basically,

313
00:09:38.843 --> 00:09:42.343
how hard or risky it is to trade against potential mispricing.

314
00:09:42.429 --> 00:09:42.547
OK.

315
00:09:43.070 --> 00:09:45.078
And they found the whole advertising effect,

316
00:09:45.203 --> 00:09:47.968
both the initial boost and the subsequent reversal,

317
00:09:48.078 --> 00:09:50.890
was stronger for stocks with higher arbitrage costs.

318
00:09:51.000 --> 00:09:54.593
Meaning where it's harder for arbitragers to step in and correct the price quickly.

319
00:09:54.875 --> 00:09:57.625
The advertising driven attention effect has a.

320
00:09:57.706 --> 00:09:57.986
bigger,

321
00:09:58.287 --> 00:09:59.408
longer lasting impact,

322
00:09:59.548 --> 00:10:01.150
both positive and negative.

323
00:10:01.209 --> 00:10:01.729
That makes sense.

324
00:10:02.072 --> 00:10:03.674
Any differences based on the type of company?

325
00:10:04.033 --> 00:10:04.213
Yeah.

326
00:10:04.271 --> 00:10:05.373
A few interesting points there,

327
00:10:05.432 --> 00:10:05.615
too.

328
00:10:06.018 --> 00:10:06.494
The effect,

329
00:10:06.697 --> 00:10:08.814
particularly the negative long run reversal,

330
00:10:08.900 --> 00:10:11.479
was more pronounced for small firms.

331
00:10:11.604 --> 00:10:11.721
OK.

332
00:10:11.861 --> 00:10:13.158
And also for value stocks.

333
00:10:13.502 --> 00:10:14.182
And interestingly,

334
00:10:14.260 --> 00:10:18.822
for firms that had poor prior stock performance or poor prior operating performance.

335
00:10:19.244 --> 00:10:19.400
Hmm.

336
00:10:20.041 --> 00:10:24.729
So maybe companies that aren't doing so well use advertising as a way to get some temporary attention.

337
00:10:25.291 --> 00:10:26.447
But the reversal hits harder.

338
00:10:26.830 --> 00:10:29.393
That's one interpretation the findings seem to support,

339
00:10:29.434 --> 00:10:29.652
yes.

340
00:10:30.174 --> 00:10:35.258
It might be a tactic used more effectively or perhaps more desperately by firms needing a boost.

341
00:10:35.375 --> 00:10:36.883
And did they do robustness checks,

342
00:10:37.039 --> 00:10:37.523
just briefly?

343
00:10:37.680 --> 00:10:38.039
They did.

344
00:10:38.180 --> 00:10:40.242
They looked at percentage changes in advertising,

345
00:10:40.484 --> 00:10:42.023
controlled for things like sales growth,

346
00:10:42.367 --> 00:10:43.547
checked different time periods,

347
00:10:43.609 --> 00:10:43.828
though,

348
00:10:43.891 --> 00:10:45.656
with some data caveats earlier on.

349
00:10:46.125 --> 00:10:46.531
Generally,

350
00:10:46.562 --> 00:10:47.656
the main results held up.

351
00:10:47.984 --> 00:10:48.234
Okay,

352
00:10:48.484 --> 00:10:48.844
good to know.

353
00:10:49.531 --> 00:10:50.438
So wrapping this up,

354
00:10:50.859 --> 00:10:52.797
what's the key takeaway for our listener?

355
00:10:53.166 --> 00:10:57.992
I think the main point is that while a big jump in advertising can give a stock a short-term lift,

356
00:10:58.371 --> 00:11:02.031
you really need to be wary of the potential for underperformance later on.

357
00:11:02.578 --> 00:11:08.258
It seems less about fundamental value changing overnight and more about this temporary wave of investor attention.

358
00:11:08.484 --> 00:11:11.867
And focusing on those trading rules and backtest results we discussed,

359
00:11:12.758 --> 00:11:17.570
simply chasing high advertising spend looks like a losing strategy long-term.

360
00:11:18.346 --> 00:11:18.566
Yeah,

361
00:11:18.646 --> 00:11:20.647
the reversal is pretty clear in their tests.

362
00:11:20.829 --> 00:11:29.175
Any strategy would need to account for that fade and maybe consider things like how sustainable the attention spike looks or the arbitrage costs involved or the type of company.

363
00:11:29.636 --> 00:11:30.862
It definitely adds complexity.

364
00:11:31.581 --> 00:11:33.604
So maybe a final thought for you listening.

365
00:11:35.128 --> 00:11:36.800
How might you use these insights?

366
00:11:37.800 --> 00:11:44.206
Could tracking changes in advertising alongside subsequent shifts in training volume or analyst coverage give you an edge?

367
00:11:44.706 --> 00:11:46.487
Maybe helping anticipate that fade or...

368
00:11:47.122 --> 00:11:48.563
Identify when it might be less severe.

369
00:11:49.145 --> 00:11:49.924
Something to think about.

370
00:11:50.366 --> 00:11:50.965
Food for thought.

371
00:11:51.087 --> 00:11:51.505
Definitely.

372
00:11:52.169 --> 00:11:54.551
Thank you for tuning in to Papers with Backtests podcast.

373
00:11:54.848 --> 00:11:57.075
We hope today's episode gave you useful insights,

374
00:11:57.309 --> 00:12:02.278
particularly regarding the potential pitfalls of chasing short-term gains linked to advertising expenditure.

375
00:12:02.575 --> 00:12:04.598
Join us next time as we break down more research.

376
00:12:04.942 --> 00:12:06.473
And for more papers and backtests,

377
00:12:06.567 --> 00:12:09.286
find us at https.paperswithbacktests.com.

378
00:12:09.551 --> 00:12:10.083
Happy trading.