WEBVTT

1
00:00:00.040 --> 00:00:00.200
Hello.

2
00:00:00.460 --> 00:00:02.602
Welcome back to Papers with Backtest podcast.

3
00:00:03.002 --> 00:00:05.885
Today we dive into another Algo trading research paper.

4
00:00:06.185 --> 00:00:06.605
Indeed.

5
00:00:06.685 --> 00:00:07.606
And this one's quite interesting.

6
00:00:07.607 --> 00:00:09.027
It's called Piety and Profit,

7
00:00:09.107 --> 00:00:11.449
Stock Market Anomaly During the Muslim Holy Month.

8
00:00:11.689 --> 00:00:11.949
Right.

9
00:00:12.089 --> 00:00:12.990
By Bielkowski,

10
00:00:13.330 --> 00:00:13.831
Edabari,

11
00:00:14.171 --> 00:00:14.812
and Wisniewski.

12
00:00:15.572 --> 00:00:18.995
And the headline finding is pretty attention grabbing,

13
00:00:19.035 --> 00:00:19.355
isn't it?

14
00:00:19.595 --> 00:00:20.176
It really is.

15
00:00:20.476 --> 00:00:25.140
They looked at stock returns during Ramadan in 14 predominantly Muslim countries.

16
00:00:25.952 --> 00:00:28.274
over a decent period 1989 to 2007.

17
00:00:28.275 --> 00:00:30.576
And found returns were significantly higher.

18
00:00:30.756 --> 00:00:31.377
Significantly.

19
00:00:31.857 --> 00:00:35.360
Almost nine times higher during Ramadan compared to the rest of the year.

20
00:00:35.660 --> 00:00:36.220
And get this,

21
00:00:36.661 --> 00:00:37.722
volatility was lower.

22
00:00:37.882 --> 00:00:40.624
Higher returns and lower risk during that specific time.

23
00:00:40.804 --> 00:00:41.044
Okay,

24
00:00:41.144 --> 00:00:42.525
so for you listening,

25
00:00:42.605 --> 00:00:45.367
this deep dive is really a shortcut to understanding this,

26
00:00:45.628 --> 00:00:45.788
well,

27
00:00:45.868 --> 00:00:49.351
surprising market pattern and what the researchers actually found.

28
00:00:49.431 --> 00:00:49.851
Exactly.

29
00:00:49.951 --> 00:00:50.572
Is it a quirk?

30
00:00:50.732 --> 00:00:51.212
An edge?

31
00:00:51.692 --> 00:00:52.213
Let's dig in.

32
00:00:52.432 --> 00:00:55.815
Let's get right into the trading rules and the backtest results they presented.

33
00:00:56.095 --> 00:00:57.576
They talk about this Ramadan effect.

34
00:00:57.696 --> 00:00:58.737
What exactly is that?

35
00:00:59.097 --> 00:01:00.699
So it's basically this observation,

36
00:01:01.099 --> 00:01:03.301
which they found to be statistically significant,

37
00:01:04.602 --> 00:01:05.963
that during the month of Ramadan,

38
00:01:06.783 --> 00:01:11.147
stock returns in these markets tended to be much higher and price swings,

39
00:01:11.347 --> 00:01:11.587
you know,

40
00:01:11.647 --> 00:01:13.949
volatility were noticeably lower.

41
00:01:14.429 --> 00:01:15.190
Not just a feeling,

42
00:01:15.191 --> 00:01:16.311
they actually measured it.

43
00:01:16.312 --> 00:01:16.531
Oh yeah,

44
00:01:16.731 --> 00:01:17.552
they put numbers on it,

45
00:01:17.672 --> 00:01:18.312
big numbers.

46
00:01:18.913 --> 00:01:21.335
Using MSCI indices denominated in U.S.

47
00:01:21.395 --> 00:01:21.775
dollars,

48
00:01:22.092 --> 00:01:24.812
The average annualized return during Ramadan was

49
00:01:25.472 --> 00:01:29.052
38.09%. 38%

50
00:01:29.112 --> 00:01:30.412
and compare that to the rest of the year.

51
00:01:30.532 --> 00:01:32.132
Just 4.32%

52
00:01:32.192 --> 00:01:33.452
annualized for the other months.

53
00:01:33.932 --> 00:01:34.572
So like we said,

54
00:01:34.612 --> 00:01:35.992
nearly nine times the return.

55
00:01:36.152 --> 00:01:37.272
It's a massive gap.

56
00:01:37.632 --> 00:01:38.352
That really is striking.

57
00:01:38.372 --> 00:01:39.332
And the volatility part,

58
00:01:39.333 --> 00:01:40.032
you said that went down.

59
00:01:40.212 --> 00:01:40.352
Yeah,

60
00:01:40.592 --> 00:01:43.512
significantly lower volatility in almost all the countries they studied,

61
00:01:44.232 --> 00:01:44.412
which,

62
00:01:44.572 --> 00:01:44.812
you know,

63
00:01:45.172 --> 00:01:48.032
usually you associate higher potential returns with higher risk,

64
00:01:48.172 --> 00:01:48.912
higher volatility.

65
00:01:48.952 --> 00:01:49.152
Right.

66
00:01:49.340 --> 00:01:50.901
That's the standard finance tradeoff.

67
00:01:50.921 --> 00:01:52.182
So this goes against the grain.

68
00:01:52.362 --> 00:01:52.723
It does.

69
00:01:52.883 --> 00:01:53.583
Higher returns,

70
00:01:53.784 --> 00:01:54.524
calmer markets.

71
00:01:54.744 --> 00:01:56.406
It's quite the anomaly.

72
00:01:56.606 --> 00:01:59.628
Did they see any change in like how much trading was happening?

73
00:01:59.688 --> 00:02:01.309
Did volume spike or dip?

74
00:02:01.670 --> 00:02:02.210
Interestingly,

75
00:02:02.330 --> 00:02:02.450
no.

76
00:02:03.271 --> 00:02:08.355
They looked at trading volume and found no discernible change during Ramadan compared to the rest of the year.

77
00:02:08.775 --> 00:02:12.038
So it doesn't seem to be driven by just more or less people trading.

78
00:02:12.418 --> 00:02:12.538
Hmm.

79
00:02:13.539 --> 00:02:13.659
OK.

80
00:02:14.580 --> 00:02:18.263
The paper also mentions using event study analysis.

81
00:02:18.480 --> 00:02:19.901
Can you maybe unpack that a bit for us?

82
00:02:19.981 --> 00:02:20.662
What did that show?

83
00:02:20.862 --> 00:02:21.062
Sure.

84
00:02:21.522 --> 00:02:25.205
An event study tries to isolate the impact of a specific event here.

85
00:02:25.826 --> 00:02:27.687
It's Ramadan starting on stock prices.

86
00:02:28.047 --> 00:02:35.373
It filters out the normal everyday market movements to see if there's an extra return or abnormal return linked just to the event.

87
00:02:35.874 --> 00:02:38.476
So they're trying to see if Ramadan itself causes a bump?

88
00:02:38.616 --> 00:02:39.136
Precisely.

89
00:02:39.496 --> 00:02:44.120
They used a couple of models like a constant mean return model and a market model as benchmarks.

90
00:02:44.460 --> 00:02:45.241
And what did they find?

91
00:02:45.321 --> 00:02:46.782
Were there these abnormal returns?

92
00:02:46.882 --> 00:02:47.102
Yes,

93
00:02:47.302 --> 00:02:47.683
definitely.

94
00:02:47.960 --> 00:02:51.400
They measured something called cumulative abnormal returns or CARS.

95
00:02:51.880 --> 00:02:55.200
Think of it as the total extra return accumulated during the Ramadan period.

96
00:02:55.260 --> 00:02:55.420
Okay.

97
00:02:55.660 --> 00:02:56.260
During the actual,

98
00:02:56.300 --> 00:02:56.440
say,

99
00:02:56.480 --> 00:02:58.740
20 or 21 trading days of Ramadan,

100
00:02:59.080 --> 00:03:00.940
the CARS ranged from about 2.5%

101
00:03:01.200 --> 00:03:02.440
to 3.1%.

102
00:03:02.441 --> 00:03:03.720
So an extra 2.5%

103
00:03:03.721 --> 00:03:04.340
to 3%

104
00:03:04.480 --> 00:03:05.540
just from being in that period?

105
00:03:05.920 --> 00:03:06.060
Yeah,

106
00:03:06.120 --> 00:03:08.200
on top of whatever the market was doing anyway.

107
00:03:08.860 --> 00:03:10.040
And what's even more interesting,

108
00:03:10.380 --> 00:03:11.980
if they extended the window slightly,

109
00:03:12.320 --> 00:03:14.820
including days just before and after Ramadan.

110
00:03:14.980 --> 00:03:15.660
It got bigger.

111
00:03:15.880 --> 00:03:16.280
It did.

112
00:03:16.508 --> 00:03:18.510
the Qataris could go as high as nearly 5%.

113
00:03:18.930 --> 00:03:22.933
So the effect seems concentrated around the Ramadan period itself.

114
00:03:23.213 --> 00:03:24.114
Almost 5%

115
00:03:24.354 --> 00:03:25.855
extra return in a short window.

116
00:03:26.416 --> 00:03:28.337
Was this statistically solid though,

117
00:03:28.918 --> 00:03:30.219
or just a fluke in the data?

118
00:03:30.759 --> 00:03:31.580
They ran the tests,

119
00:03:32.140 --> 00:03:32.701
t-tests,

120
00:03:32.801 --> 00:03:33.481
sign tests,

121
00:03:34.242 --> 00:03:35.303
and the results were strong.

122
00:03:35.863 --> 00:03:38.585
P-values were generally well below 0.05,

123
00:03:38.645 --> 00:03:40.287
often even below 0.01.

124
00:03:40.444 --> 00:03:43.687
Which means it's very unlikely these results happened by random chance.

125
00:03:43.747 --> 00:03:44.147
Exactly.

126
00:03:44.167 --> 00:03:46.509
It strongly suggests this Ramadan effect is,

127
00:03:46.749 --> 00:03:46.969
you know,

128
00:03:46.970 --> 00:03:49.011
a genuine pattern in the data they looked at.

129
00:03:49.191 --> 00:03:49.311
OK,

130
00:03:49.351 --> 00:03:50.812
but whenever you see something like this,

131
00:03:50.972 --> 00:03:51.693
you have to ask,

132
00:03:52.393 --> 00:03:53.354
could it be something else?

133
00:03:53.634 --> 00:03:53.895
You know,

134
00:03:53.935 --> 00:03:56.136
some other known market pattern just overlapping.

135
00:03:56.517 --> 00:03:57.437
Did they check for that?

136
00:03:57.698 --> 00:03:58.078
Absolutely.

137
00:03:58.118 --> 00:03:58.658
Crucial point.

138
00:03:58.659 --> 00:04:00.360
They did several robustness checks.

139
00:04:00.520 --> 00:04:00.980
For example,

140
00:04:00.981 --> 00:04:07.265
they made sure the results held up even when using the official Makamun sighting to determine the exact dates of Ramadan each year.

141
00:04:07.485 --> 00:04:07.685
Right.

142
00:04:07.725 --> 00:04:09.287
Because the start date can vary slightly.

143
00:04:09.707 --> 00:04:09.827
Yep.

144
00:04:10.176 --> 00:04:11.797
But the effect was consistent regardless.

145
00:04:11.837 --> 00:04:12.038
Okay.

146
00:04:12.218 --> 00:04:12.538
What else?

147
00:04:12.778 --> 00:04:15.901
They also wondered if maybe the length of the daily fast mattered.

148
00:04:16.201 --> 00:04:18.703
Like maybe longer fasting days had a different impact.

149
00:04:18.763 --> 00:04:19.543
Interesting thought.

150
00:04:20.304 --> 00:04:21.725
But nope.

151
00:04:21.726 --> 00:04:23.547
They didn't find any significant link there either.

152
00:04:24.207 --> 00:04:29.051
The cars didn't seem to correlate with how long the fasting day was in a particular year or location.

153
00:04:29.391 --> 00:04:33.494
So not directly tied to the physical act of fasting duration itself.

154
00:04:33.875 --> 00:04:35.116
Did they try different data?

155
00:04:35.356 --> 00:04:35.696
They did.

156
00:04:36.036 --> 00:04:39.679
They cross-referenced using S&P indices for eight countries.

157
00:04:40.040 --> 00:04:41.281
Instead of the MSCI ones.

158
00:04:41.381 --> 00:04:41.882
And guess what?

159
00:04:42.142 --> 00:04:42.782
Same effect.

160
00:04:43.062 --> 00:04:43.503
Pretty much.

161
00:04:44.143 --> 00:04:47.306
Annualized returns during Ramadan were about 40.8%

162
00:04:47.326 --> 00:04:48.527
versus 11.9%

163
00:04:48.528 --> 00:04:49.107
the rest of the year.

164
00:04:49.548 --> 00:04:52.530
Still a huge difference and statistically significant again.

165
00:04:52.970 --> 00:04:54.852
The t-statistic was 3.65,

166
00:04:54.872 --> 00:04:55.912
which is quite high.

167
00:04:56.033 --> 00:04:57.634
That strengthens the case quite a bit.

168
00:04:57.674 --> 00:04:57.794
Now,

169
00:04:57.854 --> 00:05:02.638
what about those classic calendar anomalies like January effect or sell in May?

170
00:05:02.978 --> 00:05:03.518
Good question.

171
00:05:03.538 --> 00:05:04.619
They tackled those directly.

172
00:05:05.020 --> 00:05:05.940
For the January effect,

173
00:05:05.960 --> 00:05:07.301
they re-ran the analysis.

174
00:05:07.768 --> 00:05:10.805
but excluded any Ramadan periods that happened to start in January.

175
00:05:11.106 --> 00:05:12.454
The Ramadan effect was still there.

176
00:05:12.905 --> 00:05:14.326
Annualized returns were still high,

177
00:05:14.386 --> 00:05:15.587
around 30.4%

178
00:05:16.208 --> 00:05:18.930
during those non-January Ramadans versus the rest of the year.

179
00:05:19.410 --> 00:05:21.672
So it's not just January wearing a disguise.

180
00:05:21.832 --> 00:05:23.473
What about sell in May,

181
00:05:23.854 --> 00:05:24.794
the Halloween effect?

182
00:05:25.215 --> 00:05:25.435
Right.

183
00:05:25.475 --> 00:05:28.137
The idea that returns are better from November to April,

184
00:05:28.617 --> 00:05:29.538
they tested that too.

185
00:05:29.898 --> 00:05:32.841
They excluded Ramadans falling in that no-vapory period.

186
00:05:32.941 --> 00:05:34.242
And did the effect disappear then?

187
00:05:34.402 --> 00:05:35.183
Quite the opposite.

188
00:05:35.184 --> 00:05:36.283
It actually got stronger.

189
00:05:36.624 --> 00:05:41.868
The annualized return during Ramadans outside that no-vapory window jumped to over 51%.

190
00:05:42.593 --> 00:05:44.635
with a t-stat of 4.44.

191
00:05:44.795 --> 00:05:45.155
Wow.

192
00:05:45.555 --> 00:05:48.358
So it seems very distinct from those other calendar patterns.

193
00:05:48.398 --> 00:05:49.078
It really does.

194
00:05:49.458 --> 00:05:51.880
They even went further and used multivariate regressions.

195
00:05:52.201 --> 00:05:53.882
That lets you control for several things at once.

196
00:05:54.062 --> 00:05:54.883
Like Monday effect,

197
00:05:54.943 --> 00:05:55.783
January Halloween,

198
00:05:55.843 --> 00:05:56.324
all together.

199
00:05:56.524 --> 00:05:56.964
Exactly.

200
00:05:57.505 --> 00:06:00.907
And even after accounting for all those potential influences simultaneously,

201
00:06:01.328 --> 00:06:03.870
the Ramadan variable was still statistically significant.

202
00:06:04.030 --> 00:06:05.011
It seemed to be its own thing.

203
00:06:05.291 --> 00:06:05.431
Okay,

204
00:06:05.451 --> 00:06:06.131
that's pretty thorough.

205
00:06:06.472 --> 00:06:09.894
They also compared countries based on Muslim population percentage,

206
00:06:09.954 --> 00:06:10.595
Yes.

207
00:06:11.115 --> 00:06:12.156
Using something called

208
00:06:12.817 --> 00:06:14.599
Seemingly unrelated regression,

209
00:06:14.799 --> 00:06:14.959
or

210
00:06:15.579 --> 00:06:18.422
SUR. It's a technique for comparing groups.

211
00:06:18.782 --> 00:06:19.462
What did that show?

212
00:06:19.683 --> 00:06:21.664
It showed the Ramadan effect was strongest,

213
00:06:21.944 --> 00:06:22.705
most pronounced,

214
00:06:23.005 --> 00:06:25.547
in the countries where the population was majority Muslim.

215
00:06:26.348 --> 00:06:32.613
Which makes intuitive sense if the effect is linked somehow to the shared cultural or religious experience of Ramadan.

216
00:06:32.913 --> 00:06:33.453
Precisely.

217
00:06:33.753 --> 00:06:35.955
It seems tied to that specific context.

218
00:06:36.035 --> 00:06:37.697
And did this effect fade away over time?

219
00:06:37.717 --> 00:06:41.239
Like maybe markets got more efficient towards the end of their study period?

220
00:06:41.379 --> 00:06:42.360
That's another key check.

221
00:06:42.637 --> 00:06:44.178
They looked at the later subperiod,

222
00:06:44.579 --> 00:06:46.140
2001 to 2007,

223
00:06:46.760 --> 00:06:48.181
and the anomaly was still present.

224
00:06:48.662 --> 00:06:52.164
No clear sign that it was disappearing due to increased market efficiency,

225
00:06:52.465 --> 00:06:53.786
at least up to 2007.

226
00:06:54.366 --> 00:06:54.927
One last check,

227
00:06:55.007 --> 00:06:56.728
could it have been currency movements?

228
00:06:56.928 --> 00:07:00.671
Maybe the local currencies were just strengthening against the dollar during Ramadan?

229
00:07:00.791 --> 00:07:01.672
They checked that too,

230
00:07:01.832 --> 00:07:04.294
by looking at returns denominated in local currency.

231
00:07:04.334 --> 00:07:05.395
And the effect was still there?

232
00:07:05.735 --> 00:07:05.875
Yep.

233
00:07:06.476 --> 00:07:07.356
The results held up,

234
00:07:07.637 --> 00:07:10.599
confirming it was mainly driven by the actual stock price changes.

235
00:07:10.813 --> 00:07:12.353
not just exchange rate fluctuations.

236
00:07:12.673 --> 00:07:12.893
Okay.

237
00:07:12.953 --> 00:07:14.593
So pulling all this research together,

238
00:07:15.973 --> 00:07:18.113
what did the paper suggest in terms of,

239
00:07:18.373 --> 00:07:18.573
well,

240
00:07:18.753 --> 00:07:20.173
potential trading strategies?

241
00:07:20.293 --> 00:07:21.953
If someone listening is intrigued,

242
00:07:22.093 --> 00:07:23.093
what are the implications?

243
00:07:23.813 --> 00:07:23.933
Well,

244
00:07:23.934 --> 00:07:25.173
the paper floats the idea,

245
00:07:25.473 --> 00:07:26.453
based on these findings,

246
00:07:26.953 --> 00:07:34.873
that an investor looking for short-term profits could potentially buy stocks in these markets just before Ramadan starts.

247
00:07:34.913 --> 00:07:35.493
Go long.

248
00:07:36.073 --> 00:07:36.233
Right.

249
00:07:36.473 --> 00:07:37.793
Establish long positions.

250
00:07:38.161 --> 00:07:42.985
and then sell them near the end of Ramadan or maybe shortly after the Eid al-Fitr festival that marks the end.

251
00:07:43.125 --> 00:07:44.186
Seems straightforward enough.

252
00:07:44.486 --> 00:07:45.207
Any catches?

253
00:07:45.507 --> 00:07:48.049
The main caveat they mention is transaction costs,

254
00:07:48.129 --> 00:07:48.549
obviously.

255
00:07:48.929 --> 00:07:49.610
Broker fees,

256
00:07:49.710 --> 00:07:50.531
potential slippage.

257
00:07:50.991 --> 00:07:52.292
That always eats into profits.

258
00:07:52.672 --> 00:07:55.635
But they do suggest that the size of the abnormal returns they found,

259
00:07:55.695 --> 00:07:56.475
that 2.5%

260
00:07:56.476 --> 00:07:56.816
to 5%

261
00:07:57.016 --> 00:07:57.516
extra kick,

262
00:07:58.117 --> 00:08:00.238
might be large enough to potentially overcome those costs.

263
00:08:00.258 --> 00:08:00.779
Potentially.

264
00:08:00.879 --> 00:08:02.940
And was there any other approach mentioned?

265
00:08:03.361 --> 00:08:06.683
They also noted a kind of more passive strategy.

266
00:08:07.201 --> 00:08:07.521
Basically,

267
00:08:07.541 --> 00:08:09.823
if you were planning to buy or sell in those markets anyway,

268
00:08:10.243 --> 00:08:14.247
you might simply adjust your timing slightly to align better with this Ramadan pattern,

269
00:08:14.267 --> 00:08:17.609
maybe delay sales until after Ramadan or bring purchases forward.

270
00:08:17.990 --> 00:08:18.370
Interesting.

271
00:08:18.371 --> 00:08:19.531
So less active trading,

272
00:08:19.611 --> 00:08:20.692
more timing adjustment.

273
00:08:20.732 --> 00:08:27.837
It really makes you think about how factors beyond pure economics like culture and religion might ripple through financial markets.

274
00:08:27.977 --> 00:08:33.281
Are these exploitable anomalies or just fascinating quirks of human behavior reflected in prices?

275
00:08:33.602 --> 00:08:35.023
Something definitely to mull over.

276
00:08:35.273 --> 00:08:35.733
Absolutely.

277
00:08:35.734 --> 00:08:37.735
It challenges the idea of purely rational markets,

278
00:08:37.736 --> 00:08:38.195
that's for sure.

279
00:08:38.956 --> 00:08:41.438
Thank you for tuning in to Papers with Backtests podcast.

280
00:08:41.578 --> 00:08:43.880
We hope today's episode gave you useful insights.

281
00:08:44.320 --> 00:08:46.262
Join us next time as we break down more research.

282
00:08:46.722 --> 00:08:48.163
And for more papers and backtests,

283
00:08:48.303 --> 00:08:51.526
find us at https.paperswithbacktests.com.

284
00:08:51.806 --> 00:08:52.527
Happy trading.