Can diversification degree amplify momentum and contrarian anomalies?
Co-authored by:
Houda BEN MHENNI HAJ YOUSSEF
Economics and Applied Finance Research Unit
The Institute of High Commercial Studies
Carthage, Tunisia
hbenmhenni@ yahoo.fr
(216) 21 32 76 48
Lassad EL MOUBARKI
Paris Dauphine University
elmoubarki_lassad@yahoo.fr
(216) 97 37 91 04
Olfa BENOUDA SIOUD
Associate Professsor
Economics and Applied Finance Research Unit
The Institute of High Commercial Studies
Carthage, Tunisia
olfa.benouda@ihec.rnu.tn
(216) 98 64 80 64
Abstract
This paper analyses the impact of industrial diversification on the profitability of contrarian and momentum
strategies. Our findings show that the momentum strategy seems to be no more profitable in the recent years. In
fact, while momentum strategies earn large positive and significant returns before the 2000 crash, these returns are
negative after the crash. Then, firms are classified into two groups according to the intensity of their industrial
diversification by the Ward (1963) method. We find that, for the non -diversified sample, no effect is significant.
However, for the diversified sample, the significant contrarian effect observed in the medium term disappears in the
long run. Before the crash, we find a more pronounced momentum effect for the non -diversified sample. After the
crash, we identify a contrarian effect more important for this sample. Moreover, the Bootstrap without replacement
results do not support the risk based explanation.
1
Keywords: corporate diversification, momentum, Bootstrap, cross-sectional variance of expected return, risk
1. Introduction
The momentum and contrarian anomalies have been witnessed not only in the American market but also in
many international financial markets (Patro and Wu, 2004; Lewellen, 2002; Rouwenhorst, 1998; Richards, 1997).
Based on the fact that the best (bad) performing stocks will continue their upward (downward) trend, the momentum
strategies, which consist in buying stocks with high returns (winners) and selling stocks with low return (losers), are
profitable in the medium run (Jegadeesh and Titman 1993). In the long run, the trend will reverse and the contrarian
strategy, implying a long position in the past losers and short position in the winners, will be profitable during 3-5
years (DeBondt and Thaler, 1985, 1987).
Two competing approaches are put forward to account for the profitability of these trading strategies: the
first is the rational approach that consists in explaining the momentum anomaly by systematic factors of the risk
ignored by the standard models. The second is the behavioral approach that explains the strategy’s profits by
judgment biases. These judgment biases induce investors’ over-reaction or under-reaction to information and
produce continuation and of stock returns’ reversals. The behavioral models also suggest that momentum and
contrarian effects should be more severe for firms that are hard-to-value (Daniel and Titman 1999) and suffering
from a higher level of informational asymmetry (Jegadeesh and Titman 2001).
Prior studies recorded a significant relation between industrial diversification and information
asymmetry.Nonetheless, they offer contradictory predictions about how diversification affects the level of
information asymmetry. On the one hand, the degree of information asymmetry between managers and outsiders
may differ for the diversified versus the focused firms. The reason is twofold. First, diversified firms are required to
report only limited accounting information for their business segment. Therefore, diversified firms’ managers can
observe divisional cash- flows while outsiders can observe only noisy estimates of divisional cash flows (Rajan et al.
2000; Nanda and Narayanan, 1999). Aware of this advantage, managers may manage earnings and exercise
considerable discretion in allocating revenues, costs, and assets across the business segments (Givoly et al. 1999;
Dunn and Nathan, 1998). Thus, the consolidated nature of earnings for the diversified firms could actually imply less
transparency to outsiders and reported earnings will convey less value-relevant information (Habib et al. 1997).
Second, diversified firms operate in many industries and have more complex structures while individual financial
analysts often specialize within a particular industry (Boudoukh et al. 1994).Hence, analyzing the firm’s earnings
2
reports requires more resources and expertise and dealing with a diversified firm will be a difficult task (Zuckerman,
2000; Dunn and Nathan, 1998). As a result, the degree of informational asymmetry is likely to be more acute in
diversified firms.
On the other hand, corporate diversification may reduce the asymmetric information problems because of
the aggregate nature of firms’ accounting reports. Indeed, firms that operate in different industries are likely to
derive cash flows that are not perfectly correlated. Thus, the errors made by outsiders’ in predicting these cash flows
are imperfectly or negatively correlated across the segments and the absolute value of the percentage error in the
forecast of diversified firm’s consolidated cash flows may be lowered. In other words, if the errors that outsiders
make in forecasting segment cash flows are not perfectly positively correlated then the consolidated forecast may be
more precise for a diversified firm than a forecast for a focused firm. Therefore, the degree to which the predictions
of outsiders differ from the managers’ private information could be reduced (Hadlock et al. 1999; Thomas, 2002;
Clarke et al. 2004).
In brief, prior studies affirm that industrial diversification has an impact on the level of information
asymmetry. The behavioral models also suggest that momentum anomaly is affected by information asymmetry.
Specifically, they argue that the momentum (and contrarian) effect is attributed to inefficient stock price reaction to
the firm’s specific information (Jegadeesh and Titman, 2001). Empirical evidence support that it is related to various
proxies for the quality and type of information that is provided about the firm, the relative amounts of information
disclosed publicly and being generated privately (Hong et Stein, 1999). Therefore, mispricing should be more severe
among securities that are hard to value and momentum and contrarian effects are likely to be greater. Thus, this
paper aims to study the impact of industrial diversification on the profitability of contrarian and momentum
strategies. If industrial diversification decreases the level of asymmetric information, the momentum and contrarian
effects must be less pronounced for the diversified firms. Contrarily, if industrial diversification increases the level
of asymmetric information, the momentum and contrarian effects must be less pronounced for the diversified firms.
Using Monthly returns, the WRSS strategy were applied to 249 American listed stocks from January 1994 to
April 2004. In contrast to previous studies using older data, our results show that the momentum strategy seems to
stop being profitable. To study the impact of the 2000 crash on our results, we implement the WRSS strategy during
the [01/1994-03/2000] and [04/2000-04/2004] sub-periods. Our results indicate that while momentum strategies earn
large positive and significant returns during the first sub- period, these returns are negative during the second subperiod. As Hong and Stein (2003) report, the 2000 market crash is qualified by negative returns.
3
Then, firms are classified according to the intensity of their industrial diversification by Ward‘s (1963)
method of hierarchical cluster analysis. According to this method, our sample should be classified into two subsamples: the diversified and non diversified samples. By re-implementing the WRSS strategy on these sub-samples,
we find that, for the non diversified sample, no effect is significant. However, for the diversified sample, the
significant contrarian effect observed in the medium term disappears in the long run. Before the crash, we find a
momentum effect more pronounced for the non diversified sample. After the crash, we note a contrarian effect which
is more important for this sample. The Bootstrap without replacement is used to explore the risk- based explanation
of the profitability of trading strategies. Our results imply that the cross-sectional difference in expected returns does
not, entirely, explain the trading strategies’ profits.
Our investigation extends the literature both in behavioural finance and corporate diversification. Really, it’s
noteworthy to verify the persistence of the momentum and contrarian anomalies in recent years for two reasons: The
first is to avoid data mining problems and the second is to assess the extent to which investors have learned from
earlier trends. Moreover, many studies revealed that the contrarian or momentum portfolios formed by firm
characteristics such as size, book-to-market, and/or dividend can yield a significant positive return. Nevertheless, in
spite of its importance, industrial diversification remains not studied.
Also, our study contributes to the literature on corporate diversification. According to the latest surveys, the
values of diversified firms are discounted compared with non diversified firms (Jensen, 1986; Stulz, 1990; Denis, et
al., 1997). We provided evidence that diversified firms are more efficiently valued. In fact, before the crash, the non
diversified firms were more overvalued and after the crash they become more undervalued.
The paper is outlined as follows. Section 1 presents a literature review. Section 2 introduces the sample and
the methodology. As for the results, they are presented in section 3, while alternative explanations are examined in
section 4, and we shall conclude our research in the last section.
2. Literature review: Momentum and Contrarian effects
The profitability of contrarian and momentum strategies was confirmed by empirical studies 1. Momentum
portfolios, which entail long positions in the past best performing stocks (winners) and short positions in the past
worst performing stocks (losers), yield significant positive profits in the medium term: 3–12 months (Jegadeesh and
Titman, 1993). In contrast, a systematic reversal effect is found when a longer holding period (more than 3 years) is
1
Jegadeesh et Titman (1993), DeBondt et Thaler (1985), Daniel et al (1998), Barberis et al (1998), Hong et Stein
(1999), Rouwenhorst (1999).
4
considered, and reversing the momentum strategy (buying past losers and selling past winners) results in the
production of contrarian profits (DeBondt and Thaler 1985).
While most of the empirical works point to some level of predictability in stock returns, the underlying
explanation for this type of predictability remains challenging. The rational approach asserts that the standard
models like CAPM or the three -factor model fail to take into account the systematic risk (Fama, 1998). To bypass
this shortfall, Conrad and Kaul (1998) adopt the bootstrap and Monte Carlo -simulation techniques. They find that
momentum and contrarian profits are explained by the cross-sectional variance of the stocks’ expected returns: The
momentum strategy tends to buy the stocks with higher than average unconditional mean returns and sell the stocks
with lower than average mean returns, and hence the momentum effect is not related to stock predictability.
Advocates of the behavioral approach propose a number of theoretical models of investors’ behaviour to
explain these serial correlation properties in stock prices. The underpinning of the Daniel et al. (1998) model is
investor overconfidence. They consider that stock prices are determined by the informed investors who are subject to
two biases: overconfidence and self-attribution. Overconfidence in their signals causes overreaction to their private
information, and self-attribution causes under-reaction to public information. Overreaction to private information
leads them to push up the prices of the winners above their fundamental values. This trend will be reversed, in the
long run, when public information will be confirmed.
Baberis et al. (1998) present a pattern consisting of a representative investor who believes that earnings tend
to move between two different “regimes”. In regime A, earnings are mean reverting and in regime B earning are
trending. Specifically, when a positive earnings surprise is followed by another positive (negative) surprise, the
investor raises the likelihood that he is in the trending regime, he tends to become too optimistic (pessimistic) about
the future profitability of the firm. As a result, firms realizing good earnings’ growths tend to become overvalued,
and those realizing bad earnings’ growths tend to become undervalued. The prices of these stocks ultimately undergo
reversals as achieved earnings fail to meet expectations. Whereas when a positive surprise is followed by a negative
surprise, the investor raises the likelihood that he is in the mean-reverting regime and he will under-react to earnings’
announcement. When this expectation is not confirmed by later earnings, stock prices show a delayed response to
earlier earnings.
Hong and Stein (1999) propose a model which focuses on the interactions between heterogeneous agents
rather than the psychology of the representative agent. The informed investors or the “news watchers” make
forecasts based only upon private information. The other investors “momentum traders” condition only on the most
recent price change. The information obtained by the news watchers is broadcast slowly and hence is partially
5
incorporated in the prices. This slow diffusion of information generates momentum. To benefit from the prices’
momentum, the momentum traders continue to trade pushing up the prices above their fundamental value and
generating an overreaction that will be later reversed.
3. Sample and methodology
Momentum strategies will be profitable if stock returns display a positive serial correlation, whereas
contrarian strategies will be profitable in case of negative serial correlation of stock returns. In order to examine the
profits of these trading strategies, we employ the weighted relative strength strategy (WRSS) first proposed by Lo
and MacKinlay (1990) and used by Conrad and Kaul (1998), Jegadeesh and Titman (2002), Lewellen (2002) to
mention but a few. Under this strategy, investors hold assets in proportion to their market-adjusted returns.
Specifically, they take long positions in stocks that have above average past returns and short stocks with below
average past returns. So, the investment weight assigned to stock i at time t is given by
Wi , t
1
   * ri , t 1
N
Where:
N

Rt 1 
is the total number of stocks in the sample
ri,t-1 is the return of stock i during ranking period t-1
Rt-1 is the average return across all stocks in the sample at time t-1
The profit of the WRSS strategy is given by:
 it

n
W
r
it it
1
ri,t is the return of stock i during holding period t
A positive (negative) profit indicates that a momentum (contrarian) strategy is profitable. As Conrad and
Kaul (1998) highlight, this strategy has, at least, three advantages. First, it captures the philosophy of return-based
trading strategies since its success depends on the time-series’ behavior of asset prices. Second, the weighing scheme
reflects investor’s beliefs. Third, weights on all stocks sum up to zero; that is, they imply a zero-cost portfolio.
Our sample consists of 250 American firms listed from January 1, 1994, to April 30, 2004. Monthly returns
are yielded from Datastream data files and the data for segment sales were taken from firm’s 10K annual reports
published by the Securities and Exchange Commission (SEC).
6
We, first, implemented the WRSS strategy, on the entire sample, during different time periods ([1994,
2004]; [1994, 03/2000] and [04/2000, 04/2004]), for different holding periods (3, 6, 9 and 12 months) and for
different ranking periods (3, 6, 9, 12 months). We stop our time period to March 2000 because it corresponds to the
millennium crash and to be able to compare our results with prior studies. To minimize small-sample bias and raise
the power of our tests, like Conrad and Kaul (1998) and Jegadeesh and Titman (2002), we construct trading
strategies for overlapping holding periods on a monthly -frequency. Hence, we obtain 16 trading strategies for each
time period. Then, firms of the sample are classified in two sub-samples according to their berry’s index: The
diversified sample (Div) and the non diversified (NDiv) sample. We then re-implemented these strategies on these
two sub-samples.
The Berry Index is an entropy measure of industrial diversification which takes into account both the
number of segments in which a firm operates and the relative importance of each segment in total firm sales
(Jacquemin and Berry, 1979; Palepu, 1985). This measure increases with the number of segments, holding constant
variance of the segment size, and decreases with variance of the segment size, holding the number of segments
constant. Thus, a two-segment firm with equal segment sales is more diversified than a two-segment firm with
unequal segment sales. The lower bound of the Berry index is 0 (a single segment) and it tends to 1 for a highly
diversified firm (but not equal to 1).
The Berry Index is defined as:
n
Berry 1Pi 2
i 1
Where n is the number of industry segments in which the firm is active and Pi is the share of the ith segment in the
total sales of the firm. Firms are taxonomised as diversified and non diversified by Hierarchical cluster -analysis. The
Hierarchical cluster- analysis is a statistical method for finding relatively homogeneous clusters of cases based on
measured characteristics. It starts with each case in a separate cluster and then combines the clusters sequentially,
reducing the number of clusters at each step until only one cluster is left. When there are N cases, this involves N-1
clustering steps, or fusions. Ward (1963) put forward a clustering procedure that computes the distance between two
clusters. This popular method combines the two clusters at each stage which minimize the squared error function, or
Euclidean sum of squares, E. Objective function E is defined as follows:
MinE = kik (xi- k)2
249
15
247
248
241
245
239
240
234
236
226
230
223
224
221
222
214
216
208
211
204
207
200
201
196
197
194
195
191
192
193
187
190
185
186
178
181
176
177
171
172
167
169
159
165
149
150
147
148
143
146
139
140
134
136
128
131
118
124
116
117
114
115
105
111
103
104
95
98
92
93
87
88
82
86
80
81
77
78
73
74
76
69
72
66
68
64
65
60
62
54
55
49
50
43
44
35
39
32
33
30
31
28
29
26
27
23
25
14
22
10
11
8
9
6
2
243
5
145
164
225
1
203
129
108
228
132
84
173
198
206
85
237
130
99
242
57
235
122
157
168
102
162
119
38
137
53
217
67
229
233
170
17
127
232
215
18
227
7
166
20
202
142
63
3
231
12
161
37
48
47
180
100
144
40
151
101
97
70
121
244
126
125
219
199
61
213
56
220
13
138
71
133
91
109
36
183
45
75
46
123
94
246
83
179
58
158
154
163
141
210
189
218
4
113
34
188
19
96
90
120
155
152
212
182
107
112
89
205
209
174
59
51
42
184
160
153
106
24
238
16
79
52
175
135
156
41
110
21
249
248
245
247
240
241
236
239
230
234
224
226
222
223
216
221
211
214
207
208
201
204
197
200
195
196
193
194
190
191
192
186
187
181
185
177
178
172
176
169
171
165
167
150
159
148
149
146
147
140
143
134
136
128
131
118
124
116
117
114
115
105
111
103
104
95
98
92
93
87
88
82
86
80
81
77
78
74
76
69
72
73
66
68
64
65
60
62
54
55
49
50
43
44
35
39
32
33
29
31
26
27
23
25
14
22
10
11
81
9
6
2
164
139
132
5
145
129
57
228
229
198
242
203
84
237
108
122
85
166
99
246
142
12
119
227
161
233
37
17
18
7
202
20
215
38
217
162
225
102
212
173
127
168
130
157
206
137
170
235
232
53
40
67
184
58
158
154
182
46
160
94
156
21
174
96
112
48
113
59
51
42
120
209
100
47
3
163
231
218
4
155
34
15
63
189
19
123
183
179
90
107
141
210
83
135
152
188
175
220
56
243
110
36
13
138
109
125
133
180
45
126
71
101
89
199
61
91
79
219
28
121
97
153
75
238
213
41
205
151
52
24
30
16
244
70
144
106
249
248
245
247
240
241
236
239
230
234
224
226
222
223
216
221
211
214
207
208
201
204
197
200
195
196
193
194
190
191
192
186
187
181
185
177
178
172
176
169
171
165
167
148
149
146
147
136
140
131
134
117
128
111
116
103
105
93
98
88
92
81
82
78
80
76
77
73
74
69
72
63
68
55
60
50
54
48
49
35
39
44
31
33
27
29
23
26
11
14
2
9
5
8
132
43
164
101
143
32
202
47
215
163
159
16
170
151
137
6
157
141
168
166
102
227
64
18
233
173
217
85
209
38
246
162
212
3
145
53
133
225
142
130
67
232
99
237
228
127
89
118
84
108
235
122
37
242
198
203
120
57
4
229
1
129
100
15
13
107
155
135
83
95
34
218
12
189
152
206
10
20
19
119
17
175
46
161
97
213
114
231
22
182
179
96
174
123
106
40
75
45
153
527
154
94
183
156
24
87
42
184
124
113
86
110
51
158
65
70
62
90
91
210
28
25
79
160
205
112
21
219
61
121
199
126
56
244
150
41
30
238
59
220
180
243
139
58
188
138
144
66
104
71
109
125
115
36
249
15
247
248
243
245
240
241
236
239
230
234
224
226
222
223
216
221
211
214
207
208
201
204
197
200
195
196
192
193
194
190
191
186
187
181
185
178
179
176
177
171
172
167
169
149
165
147
148
140
146
134
136
117
131
111
116
103
105
93
98
88
92
81
82
78
80
76
77
73
74
69
72
63
68
55
60
50
54
39
44
49
33
35
29
31
26
27
14
23
11
9
8
2
101
132
180
110
164
47
213
5
202
85
143
119
157
34
215
3
232
168
163
159
227
53
170
122
133
67
84
6
151
17
38
16
233
65
162
141
173
217
212
104
90
145
129
100
225
130
1
229
37
57
4
203
166
118
108
127
86
237
99
142
46
228
198
235
242
174
120
205
66
244
28
126
160
238
150
199
61
112
56
25
18
154
41
58
30
109
139
219
36
138
125
144
135
115
79
21
220
71
91
43
231
102
137
95
114
32
64
97
161
113
75
19
96
42
189
20
209
13
218
10
155
206
121
175
152
22
182
12
107
123
183
83
156
24
59
52
184
89
188
48
87
7
246
128
62
70
51
210
40
124
45
158
153
106
94
1994
1995
132
101
249
15
247
248
243
245
240
241
236
239
230
234
224
226
222
223
216
221
211
214
207
208
201
204
197
200
194
195
196
192
193
190
191
186
187
181
185
178
179
176
177
171
172
167
169
149
165
147
148
140
146
134
136
117
131
111
116
103
105
93
98
88
92
81
82
78
80
76
77
73
74
69
72
63
68
60
62
50
54
55
44
49
35
39
31
33
27
29
23
26
11
14
2
9
47
8
164
213
160
5
219
114
139
110
180
53
118
242
133
4
122
6
151
65
104
84
162
90
143
85
119
173
113
215
168
212
202
138
233
64
232
217
17
67
38
227
141
157
34
159
225
145
52
229
129
137
1
130
100
228
37
120
203
166
237
170
127
108
57
99
86
142
46
174
235
198
210
124
128
158
183
707
106
94
182
40
12
45
153
83
107
123
20
102
19
13
96
121
209
206
175
218
152
189
75
231
10
95
22
163
3
161
97
32
16
155
41
199
61
43
66
112
24
188
246
156
89
51
42
184
25
220
56
244
59
205
48
115
87
109
58
135
30
126
125
36
144
18
154
21
238
150
79
28
91
71
1996
1997
0
0
0
0
0
10
10
10
10
10
20
20
20
20
20
30
30
30
30
30
40
40
40
40
40
50
50
231
170
127
108
212
138
152
106
209
34
214
182
217
242
141
229
218
4
123
107
143
91
246
97
130
37
184
18
215
183
41
90
52
189
48
179
51
132
100
94
75
155
58
113
79
153
42
154
65
233
227
53
57
202
157
160
19
166
168
158
119
198
67
232
102
17
3
161
16
174
162
122
12
175
20
173
99
203
228
206
84
85
137
235
156
237
129
1
225
145
248
249
245
247
240
241
236
239
230
234
224
226
222
223
216
221
208
211
205
207
201
204
197
200
195
196
193
194
191
192
187
190
185
186
178
181
176
177
171
172
167
169
159
163
165
150
151
148
149
146
147
140
142
134
139
128
131
118
124
116
117
114
115
109
111
103
105
95
98
92
93
87
88
82
83
80
81
77
78
74
76
72
73
68
69
64
66
60
62
55
56
50
54
44
47
49
39
40
35
38
31
33
28
29
26
27
23
25
14
22
10
11
8
9
2
7
63
6
180
36
133
220
199
61
210
45
164
238
244
96
89
21
101
135
70
46
188
121
112
86
136
32
59
5
104
30
243
24
13
110
15
125
43
219
71
120
126
213
144
170
231
229
91
212
108
127
100
130
34
232
182
209
141
106
206
138
218
246
4
152
97
107
242
132
123
156
75
214
155
94
179
90
51
42
153
79
48
41
183
46
215
58
18
65
52
37
184
70
188
145
225
1
129
84
228
203
237
235
99
85
173
137
158
17
157
53
67
57
198
19
202
168
160
233
227
16
119
174
166
102
3
154
113
161
20
122
175
162
12
248
249
245
247
240
241
236
239
230
234
224
226
222
223
216
221
208
211
205
207
201
204
197
200
195
196
193
194
191
192
187
190
185
186
178
181
176
177
171
172
167
169
159
163
165
150
151
148
149
146
147
142
143
139
140
131
134
124
128
117
118
115
116
111
114
105
109
103
104
96
98
93
95
88
92
82
87
80
81
77
78
74
76
72
73
68
69
64
66
60
62
55
56
49
50
54
44
47
39
40
35
38
31
33
28
29
26
27
23
25
14
22
10
11
8
9
2
7
217
6
112
86
121
13
189
32
136
43
59
5
238
133
45
36
101
244
164
135
21
89
126
83
210
219
220
180
125
63
61
199
71
144
120
213
15
110
30
243
24
24
30
120
63
213
71
133
112
220
126
144
110
219
61
210
199
164
45
249
51
247
248
241
245
239
240
234
236
224
226
230
222
223
216
221
208
211
205
207
201
204
197
200
195
196
193
194
191
192
187
190
185
186
178
181
176
177
171
172
167
169
163
165
151
159
149
150
147
148
143
146
140
142
134
139
128
131
118
124
115
116
117
111
114
105
109
103
104
96
98
93
95
88
92
82
87
80
81
77
78
74
76
72
73
68
69
65
66
62
64
56
60
54
55
49
50
44
47
40
43
38
39
33
35
29
31
27
28
25
26
22
23
14
15
18
10
11
8
9
2
7
121
61
86
5
32
13
145
59
225
129
85
84
99
203
173
237
228
100
12
161
3
122
20
158
233
154
119
160
202
227
198
168
102
157
53
229
57
235
179
137
17
166
123
107
132
242
97
90
108
246
170
138
231
212
162
67
232
91
83
34
106
209
182
141
127
4
130
152
175
217
113
218
206
19
214
37
58
52
42
41
183
79
153
189
156
75
155
94
243
174
101
36
180
238
244
135
89
136
125
184
16
215
48
188
70
46
21
249
248
245
247
240
241
236
239
230
234
224
226
222
223
216
221
213
214
208
211
204
207
200
201
196
197
194
195
191
192
193
187
190
185
186
178
181
176
177
171
172
167
169
159
165
149
150
147
148
143
146
139
140
131
134
124
128
117
118
115
116
111
114
104
105
103
98
95
96
92
93
87
88
82
86
80
81
77
78
73
74
76
69
72
66
68
64
65
60
62
55
56
50
54
44
49
40
43
38
39
33
35
31
32
29
30
27
28
25
26
22
23
14
15
10
11
8
9
6
2
164
5
136
59
219
13
37
63
215
113
112
94
155
120
75
156
52
42
18
45
51
180
58
135
184
189
21
188
79
70
16
158
41
126
71
109
133
238
220
125
91
144
106
199
61
183
110
243
121
151
36
101
97
160
174
244
24
67
3
175
12
107
231
162
161
127
108
170
212
90
218
19
182
209
217
34
141
100
166
46
123
83
210
205
163
152
132
4
206
232
235
246
137
17
227
202
168
157
153
122
102
154
233
138
48
71
229
47
119
142
53
20
225
145
129
173
130
179
242
237
228
57
99
85
198
84
203
89
249
248
245
247
240
241
236
239
230
234
224
226
222
223
216
221
212
214
208
211
204
207
200
201
196
197
194
195
191
192
193
187
190
185
186
178
181
176
177
171
172
167
169
159
165
149
150
147
148
143
146
139
140
131
134
124
128
117
118
115
116
111
114
104
105
103
98
93
95
88
92
86
87
81
82
78
80
76
77
72
73
74
68
69
65
66
62
64
55
60
50
54
44
49
40
43
35
39
32
33
30
31
28
29
26
27
23
25
14
22
10
11
8
9
6
2
213
5
108
15
217
164
157
235
119
233
173
154
179
206
137
17
127
53
48
142
166
229
232
202
71
227
20
168
102
225
145
129
203
198
85
84
130
228
237
57
242
99
122
56
220
13
36
91
219
109
133
71
125
180
144
238
199
61
136
151
106
97
126
121
110
101
243
244
45
24
46
189
70
183
52
174
79
21
59
120
135
158
58
123
94
170
160
205
75
42
89
155
51
184
156
246
41
188
16
209
38
12
3
175
18
215
153
162
37
161
67
96
47
138
107
83
90
100
152
231
112
63
141
34
210
19
182
163
113
132
218
4
0
0
0
0
0
10
10
10
10
10
20
20
20
20
20
30
30
30
30
30
40
40
40
40
40
50
50
7
Where for a given classification, firm i belongs to cluster k, xi is the value of the Berry Index for the firm i, and k is
the mean of Berry Index in cluster k. The summations are, firstly, the squared error of firm i in cluster k for the
Berry Index; secondly, for all observations (firms) i within cluster k; and finally, to total the error over all clusters k.
This hierarchical clustering process can be displayed as a tree, or dendrogram, where each step in the clustering
process is illustrated by a join of the tree. The choice of the number of clusters in the classification tree relies on the
Euclidean distances: A great distance between classes indicates a high degree of dissimilarity between classes. Each
year we classify our sample firms according to their Berry Index ( the variable x i ) . The ward methods’ tree,
displayed in figure 1, shows that our sample should be classified in two classes: Diversified and non diversified. In
fact, the Euclidean distance is maximal for a number of classes equal two for all years.
hclust (*, "w ard")
hclust (*, "w ard")
hclust (*, "w ard")
hclust (*, "w ard")
hclust (*, "w ard")
hclust (*, "w ard")
hclust (*, "w ard")
hclust (*, "w ard")
hclust (*, "w ard")
hclust (*, "w ard")
FIGURE 1
The Ward MethodTree
The centre and the number of stocks of each class are presented in Table 1. The number of stocks in the
non -diversified sample is more important than the diversified sample until 1997. After this date, the number of
diversified firms exceeded the number of non- diversified firms indicating that more firms have become diversified.
The centres of the two classes have dropped from 1994 to 2003, while the non diversified sample, plummeted from
0.0535 in 1994 to 0.0077 in 2003.
TABLE 1
CENTRES AND NUMBER OF STOCKS
1998
1999
2000
2001
2002
2003
8
NDiv
Div
centers
0.0535
0.0458
0.02528
0.0028
0.0010
0.0008
0.0007
0.0013
0.0027
0.0077
number
159
159
148
126
121
119
116
98
102
107
centers
0.558
0.5534
0.5232
0.4908
0.4828
0.4894
0.4815
0.5068
0.5116
0.5238
number
90
90
101
123
128
130
133
151
147
142
This implies that firms of the non -diversified samples became more and more specialized. Regarding the
diversified sample, the centre has slightly fallen from 0.558 to 0.5238. These findings are similar to those of Wrigley
(1970) According to this author, American firms chose to be highly- specialized or highly -diversified but not in the
middle state.
We first implemented the strategies on the entire sample and for the [1994-1999] period because it matched
with the time period used in the previous studies (Jegadeesh and Titman, 2001) and we will concentrate on the 6month -ranking period strategies because they were the primary focus of the original Jegadeesh and Titman
(1993)research.
4. Results
4.1 Profitability of the trading strategies: Medium term results
We first implemented the strategies on the entire sample and for the [01/1994-03/2000] period because it
corresponds with the time period used in the early studies (Jegadeesh and Titman, 2001) and we will focus on the six
month -ranking period strategies because they were the primary focus of the original Jegadeesh and Titman (1993)
study.
Table 2 shows the profits of the trading strategies. For the [1994, 2004] time period, fourteen returns out of
sixteen were negative, whereas only two returns were positive. Among the fourteen negative returns, four are not
significant, four are significant at the 10% level and six are significant at the 5% level. The returns of six monthranking period are all negative and are significant for the nine and twelve month- holding periods. The most
profitable strategy is the 12-12 month contrarian strategy with a return of 1.3%. The return of the famous 6-6 month
strategy is negative and not significant.
TABLE 2
AVERAGE RETURNS OF THE MEDIUM TERM TRADING STRATEGIES FOR THE DIFFERENT
PERIODS
9
3
6
9
12
3
-0.0027
(0.18)
-0.0020
(0.31)
-0.0042
(0.04)
-0.0054
(0.03)
6
-0.0019
(0.58)
-0.0038
(0.15)
-0.0066
(0.04)
-0.0089
(0.02)
9
0.0044
(0.339)
-0.0008
(0.070)
-0.0107
(0.040)
-0.0125
(0.038)
12
0.0074
(0.073)
-0.0009
(0.064)
-0.0108
(0.060)
-0.0129
(0.064)
3
0.0005
(0.596)
0.0014
(0.214)
0.0024
(0.058)
0.0039
(0.036)
6
0.0012
(0.292)
0.0030
(0.039)
0.0053
(0.008)
0.0102
(0.000)
9
0.0022
(0.12)
0.0057
(0.01)
0.0092
(0.00)
0.0197
(0.00)
12
0.0042
(0.01)
0.0093
(0.00)
0.0134
(0.00)
0.0290
(0.00)
3
-0.0042
(0.46)
-0.0065
(0.14)
-0.0157
(0.00)
-0.0144
(0.00)
6
-0.0052
(0.68)
-0.0143
(0.02)
-0.0283
(0.01)
-0.0342
(0.00)
9
-0.017
(0.34)
-0.0301
(0.01)
-0.0488
(0.01)
-0.0590
(0.01)
12
-0.0288
(0.07)
-0.0351
(0.00)
-0.0526
(0.01)
-0.0646
(0.01)
Ranking period
04/2000-04/2004
01/94-03/2000
01/94-04/2004
Holding period

This table comprises the average returns of the zero-cost trading strategies implemented during 01/94-04/2004; 01/9403/2000 and 04/2000-04/2004 for 3, 6, 9 and 12 month- holding periods and 3, 6, 9, 12 month- ranking periods. The
trading strategies buy winners and sell losers based on their past performance relative to the performance of an equal-
 it
weighed index of all stocks. The dollar profits are given by

n
W
1
r
it it
where ri,t is the return of stock i
Wi , t
1
   * ri , t 1

Rt 1 
N
during the holding period t and the investment weight assigned to stock i at time t is given by
,
N is the total number of stocks in the sample, ri,t-1 is the return of stock i during the ranking period t-1, Rt-1 is the
average return across all stocks in the sample at time t-1 The numbers in parentheses are significance level of the
strategies ‘real ized profits.
During the [1994-03/2000] period, all the returns of the 16 strategies are positive and only 4 are not
statistically significant (3-3, 3-6, 3-9 and the 6-3 strategies). The 12-12 month- strategy is the most profitable with a
return of 2.91%. The WRSS strategy based on the previous 6-month performance gains a positive but not significant
return in the following 3 months (holding period = 3 months). This return is 0.31%; 0.58% and 0.94% for 6, 9 and
12-month holding period resp. These results are very close to those found in the previous studies. In fact, Jegadeesh
and Titman (2002) and Conrad and kaul (1998) found a 6-6 month WRSS strategy return of 0.37% and 0.36%
respectively. These results are in favor of the profitability of the momentum strategy in the 1990s. But our results are
10
reversed for the [04/2000, 04/2004] time period. All the return becomes negative and 12 are significant. The 66month contrarian strategy earns 1.44% while the 9-12 month strategy is the most profitable with a return of 5.90%.
These results indicate that while momentum strategies earn large positive and significant returns during the
1994-2000 sample period, the latter are negative or insignificant for the entire sample period. It’s clear that, in later
years, only the contrarian strategy is profitable in the medium run. This finding is in contrast to earlier studies based
on older and different datasets which record a significant momentum effect in the medium run. This is to be
accounted for by our time period which is more recent and by ending our time period in March 2000, we obtain
results that are similar to prior studies. Our results corroborate, however, recent studies that assert that the
momentum anomaly has disappeared in recent years. According to Malkiel (2003), momentum strategies occurred to
generate positive relative returns during some periods of the late 1990s, but highly- negative relative returns during
2000. Mengoli (2004) reports that the most recent period shows a peculiar behavior both in terms of low-momentum
and negative-reversal returns. In the same vein, Campbell (2004) affirms that momentum strategies have performed
well on average but have been highly inconsistent, generating large profits in some years and losses in others.
Schwert (2003) points out two possible explanations for such a pattern. The first is that the apparent profitability of
momentum strategies was an artifact statistic. The second is that investors learn about any true predictable pattern
and exploit it until it becomes no longer profitable. Finally, Shiller (2000) reports a fast growth in prices during the
late 1990s followed by a crash after 2000 in all the international financial markets and attributes the decline in the
market to the shifts in investor psychology. Moreover, regardless of the ranking period, average returns increase
with the holding period for all time periods. Yet, contrary to Conrad et al. (1998), these returns do not go up
geometrically with the length of the holding period so that we could not conclude that security prices follow the
random -walk process.
4.2 Industrial diversification and profitability of the trading strategies
To study the potential impact of industrial diversification on the profitability of trading strategies, we
implement the WRSS strategy on two sub-samples: The diversified sample (Div) and the non- diversified sample
(NDiv). Table 3 displays average returns of 162 trading strategies for 01/1994-04/2004 time period and the
numbers in brackets represent the significance level. Regarding the non -diversified sample, the WRSS strategy
based on the previous 6 months performance presents positive returns of 0.15% ; 0.24% ; 0.01 % respectively, in the
following 3, 6, 9 months and a negative return of 0,2304% during the following 12 months. This evidence is in
favor of momentum profitability and contrarian strategies, respectively, in the medium horizon and the long horizon.
TABLE 3
11
AVERAGE RETURNS OF THE TRADING STRATEGIES IMPLEMENTED ON THE DIVERSIFIED
FIRMS’ SAMPLE AND NON DIVERSIFIED FIRMS’ SAMPLE
Holding
period→
3
6
9
12
NDiv
-0.0008
(0.49)
0.0007
(0.64)
0.0007
(0.74)
-0.0013
(0.61)
Div
-0.0045
(0.18)
-0 .0040
(0.49)
-0.0083
(0.28)
-0.0119
(0.06)
NDiv
0.0014
(0.31)
0.0024
(0.26)
0.0001
(0.97)
-0.0023
(0.56)
Div
-0.0048
(0.13)
-0.0085
(0.07)
-0.0143
(0.05)
-0.0138
(0.06)
NDiv
0.0008
(0.73)
-0.0012
(0.69)
-0.0037
(0.41)
-0.0062
(0.25)
Div
0.0079
(0.01)
-0.0103
(0.04)
-0.0157
(0.04)
-0.0141
(0.08)
NDiv
-0.0022
(0.49)
-0.0039
(0.38)
-0.0064
(0.28)
-0.0097
(0.19)
Div
-0.0077
(0.02)
-0.0124
(0.01)
-0.0168
(0.03)
-0.0153
(0.07)
Ranking
period
3
6
9
12
This table exhibits the average profits to medium -term trading strategies implemented on two sub-samples: The
diversified sample (Div) and the non -diversified sample (NDiv). The Berry Index is defined as:
n
Berry 1Pi 2
i 1
where n is the number of industry segments in which the firm is active and Pi is the share of the ith
 it

n
W
it rit
1
segment in the total sales of the firm. The dollar profits are given by
where ri,t is the return
of stock i during the holding period t and the investment weight assigned to stock i at time t is given by
Wi , t
1
   * ri , t 1
N

Rt 1 
, N is the total number of stocks in the sample, ri,t-1 is the return of stock i during the
month ranking period t-1, Rt-1 is the average return across all stocks in the sample at time t-1 The numbers in
parentheses represent the significance level of realized strategise’ profits.
To gauge the success of the momentum and contrarian strategies on the non -diversified sample, we tested
the statistical significance of the sixteen strategy’s returns. Surprisingly, none is reliably distinguishable from zero.
Opposite to the diversified sample, all the returns are negative and twelve are significant indicating that only the
contrarian strategy is profitable. The 12-9 strategy is the most profitable with a return of -1.68%. When we hold a
ranking period constant, the returns of the contrarian strategies rise with the holding period up to 9 months,but
decrease in the 12th month. For example, the return of a 6- month ranking strategy grows from 0.48% in the next
three months to 1.44% during the next nine months and fall to 1.38% in the next twelve months. This trend leads us
to believe that the return will revert to the mean in the long run.
On the whole, these findings support the existence of a non -significant momentum effect in the medium
term followed by a non -significant contrarian effect in the non -diversified sample. However, for the diversified
sample, our results are in favor of the profitability of the contrarian strategy in the medium term. To study the impact
12
of the 2000 crash on our results, we study the profitability of the 6-month ranking period of the WRSS strategy, on
the two sub-samples, during the [01/1994-03/2000] and [04/2000-04/2004] time periods. Table 4 displays the results.
TABLE 4
RETURNS OF THE TRADING STRATEGIES IMPLEMENTED ON THE
DIVERSIFIED AND THE NON -DIVERSIFIED SAMPLES DURING DIFFERENT
TIME PERIOD
3
6
9
12
1/94-03/2000
04/200004/2004
1/94-04/2004
NDiv
0.0039
(0.01)
-0.0018
(0.483)
0.0014
(0.31)
Div
-0.0008
(0.62)
-0.010191
(0.16)
-0.0048
(0.13)
NDiv
0.0060
(0.01)
-0.0042
(0.249)
0 .0024
(0.26)
Div
0.0008
(0.68)
-0.0215
(0.05)
-0.0085
(0.07)
NDiv
0.0079
(0.02)
-0.0141
(0.01)
0.0001
(0.97)
Div
0.0042
(0.25)
-0.0408
(0.01)
-0.0143
(0.05)
NDiv
0.0111
(0.02)
-0.0240
(0.00)
-0.0023
(0.56)
Div
-0.0092
(0.22)
-0.0229
(0.15)
-0.0138
(0.06)
This table comprises the average profits to medium term trading strategies implemented on the
diversified sample and the non diversified sample during 01/94-04/2004; 01/94-03/2000 and
 it
04/2000-04/2004. The dollar profits are given by

n
W
1
r
it it
where ri,t is the return of
stock i during the holding period t and the investment weight assigned to stock i at time t is given
by
Wi , t
1
   * ri , t 1
N

Rt 1 
, N is the total number of stocks in the sample, ri,t-1 is the return of
stock i during the 6 month ranking period t-1, Rt-1 is the average return across all stocks in the
sample at time t-1 The bracketed numbers stand for the significance level of realized strategies’
profits of.
With regard to the [01/1994, 03/2000] time period, the non- diversified sample exhibits a significant
momentum effect for all the holding period. The return hikes from 0.4 % for the 3-month holding period to 1.12 %
for the 12-month holding period but less than geometrically. The return of 0.60% of the 6-6 month strategy which is
almost the double of the return of the entire sample and the return found in earlier studies. Nevertheless, the
diversified sample shows a non- significant momentum effect for the 6 and 9-month holding period. The famous 66-month strategy generates a return of 0.08% witch represents 1/4 of the return of the entire sample but this return is
not significant. For 04/2000- 04/2004 time period, the most important evidence is that the signs of returns are
inverted and all returns become negative. The momentum effect observed in the time period of late March 2000, for
13
the non diversified sample becomes a contrarian effect. In fact, the return of 0.60% for the 6 month holding period
drops to a non significant return of -0.42% and become significant for the nine and twelve month- holding period. As
far as the diversified sample is concerned , all the returns are negative too, but higher then those of the non
diversified sample. The most profitable strategy for the diversified sample is the 6-9 month contrarian strategy with a
return of -4.08%. This return is more important than the most profitable 6-12 month contrarian strategy for the non
diversified sample (2.40%). These findings contradict the pervious literature on the profitability of the momentum
strategy in the medium run. Nonetheless, they are obviously driven by the March 2000 crash. Hong and Stein
(2003) and Hong et al. (2001) state that a market crash is qualified by negative returns. Consequently, two important
results are drawn from Table 4. First, independently from the time period studied, the stock price behavior for the
two groups is different. Second, this behavior has changed in the 2000’s: In the [1994, 2000] time period, the trend
of the non -diversified exhibits a momentum effect but the diversified firms were almost, efficiently priced. By
expanding the time period to the year 2004, the results are reversed. The non diversified sample becomes efficiently
valued and the diversified sample exhibits a contrarian effect instead of a momentum effect in the medium run.
4.3 Industrial diversification and the profitability of trading strategies: Long term results
Figure 2 presents the average return of the 6 -month ranking period strategy during the different holding
periods. For the whole sample, all the returns are negative. This evidence is in favor of the profitability of the
contrarian strategy in both the medium and long run. Indeed, it seems to be driven by the presence of diversified
firms in our sample. Actually, the contrarian effect is very pronounced for the diversified sample. The average return
dips until it reaches
-1.44%, which represents almost 150% the return of the entire sample. From the 12th month, it
starts rising, while for the non diversified sample, the trend indicates a slight momentum- effect that begins to
reverse from the 9th month.
14
0,01
0,005
0
3
6
9
12
15
18
NDiv
24
-0,005
Div
Sample
-0,01
-0,015
-0,02
FIGURE 2
Average Return of the 6 -Month Ranking Period Strategy
Further details on the returns of the trading strategies in the first 24 months are displayed in Table 5.
Among the 24 trading strategies, 20 strategies yield negative returns. The four remaining positive returns are related
to the 24 month holding period in the non diversified sample. For this sample, the six -month ranking -period
strategy generates
TABLE 5
RETURNS OF THE LONG TERM TRADING STRATEGIES IMPLEMENTED ON DIVERSIFIED
AND NON DIVERSIFIED FIRMS SAMPLES
Ranking
period
3
6
9
12
NDiv
-0.0019
(0.51)
-0.00274
(0.55)
-0.0069
(0.29)
-0.009
(0.29)
Div
-0.0050
(0.18)
-0.00638
(0.21)
-0.0064
(0.29)
-0.0098
(0.18)
NDiv
-0.0006
(0.86)
-0.000741
(0.89)
-0.0035
(0.66)
-0.0045
(0.69)
Div
-0.0045
(0.25)
-0.00437
(0.40)
-0.0059
(0.35)
-0.0116
(0.11)
NDiv
0.0012
(0.78)
0.0052
(0.51)
0.0083
(0.47)
0.0123
(0.44)
Div
-0.0037
(0.39)
-0.0070
(0.21)
-0.0113
(0.10)
-0.0164
(0.05)
Holding period

15
18
24
This table contains the average profits to the long term trading strategies implemented on the non diversified firms
sample and the diversified firm sample. The trading strategies buy winners and sell losers based on their past
performance relative to the performance of an equal-weighted index of all stocks. The dollar profits are given by
 it

n
W
1
r
it it
where ri,t is the return of stock i during the holding period t and the investment weight assigned to
Wi , t
1
   * ri , t 1

Rt 1 
N
stock i at time t is given by
, N is the total number of stocks in the sample, ri,t-1 is the return of
stock i during the ranking period t-1, Rt-1 is the average return across all stocks in the sample at time t-1 The numbers
in parentheses are significance level of realize profits of strategies.
15
a return of -0.27% in the fifteen following months. This return rises to -0.0741% in the 18 following months and
become positive in the 24 following months. This upward trend is recorded too for the 3, 9 and the 12- month
ranking period. As for the diversified sample, and independently of the ranking period, the return surges in the 18
following months but drops in the 24 following months. However, the figures’ significance, in Table 5, indicates that
all returns are not reliably different from zero for the two samples.
Thus, we find that, for the non -diversified sample, the momentum in the medium run and the contrarian
effect in the long run are both non significant; whereas , for the diversified sample, the significant contrarian effect
observed in the medium term disappears in the long run .
Table 6 presents the returns of the 6 -month ranking period strategy up to 24 months before and after the
crash. Before the crash, all the returns are positive and statistically significant. First, these results prove the
profitability of the momentum strategy in the long run for the two samples. Second, the significant momentum effect
observed in the medium run in the non diversified sample continues to occur in the long run. The non -significant
momentum effect recorded, in the medium run, in the diversified sample becomes significant in the long run. Third
the momentum effect is more obvious for the non diversified sample. Finally, the returns of the trading strategies
increase dramatically with the holding period in the long run.
TABLE 6
RETURNS OF THE LONG TERM TRADING STRATEGIES IMPLEMENTED FOR
THE DIFFERENT TIME PERIODS
15
18
21
24
01/199403/2000
04/200004/2004
1/199404/2004
NDiv
0.0107
(0.04)
-0.0310
(0.000)
-0.0027
(0.559)
Div
0.0093
(0.007)
-0.0261
(0.034)
0.5156
(0.231)
NDiv
0.0183
(0.007)
-0.0330
(0.000)
-0.0007
(0.896)
Div
0.0160
(0.000)
-0.0260
(0.035)
0.3447
(0.460)
NDiv
0.0365
(0.000)
-0.0368
(0.001)
0.0016
(0.811)
Div
0.0204
(0.000)
-0.0251
(0.035)
-0.0003
(0.546)
NDiv
0.0521
(0.000)
-0.0415
(0.002)
0.0052
(0.512)
Div
0.0154
(0.001)
-0.0344
(0.017)
-0.0071
(0.213)
This table contains average returns to long term trading strategies implemented on the non
diversified firms sample and the diversified firm sample and for different time period. The
trading strategies buy winners and sell losers based on their past performance relative to the
16
performance of an equal-weighted index of all stocks. The dollar profits are given by
 it

n
W
r
it it
1
where ri,t is the return of stock i during the holding period t and the
Wi , t
1
   * ri , t 1

Rt 1 
N
investment weight assigned to stock i at time t is given by
, N is the
total number of stocks in the sample, ri,t-1 is the return of stock i during the 6 month ranking
period t-1, Rt-1 is the average return across all stocks in the sample at time t-1 The numbers in
parentheses are significance level of realize profits of strategies.
After the crash, all the returns become negative and statistically significant. First, this evidence is in favour
of the profitability of contrarian strategy in the long run for the two samples. In fact, the contrarian effect identified
in the medium run continues to exist in the long run. Second, the contrarian effect become significant in the 6 th
month for the diversified sample and in the 9 th month for the non diversified sample. Third, the contrarian effect that
was more important for the diversified sample in the medium run becomes more important for the non diversified
sample in the long run. The reason is that, for the diversified sample, the return of the contrarian strategy is stabilized
around 2,5% but for the non diversified sample, it grows more quickly and attains 4.1%. This implies that the
diversified firms are more rapidly influenced by the crash but their prices are less affected.
5. Industrial diversification and the profitability of trading strategies: Alternative
explanations
5.1 Behavioral explanation
The behavioral models have attempted to provide a framework for integrating predictability in stock
returns. In this subsection, following Lee and Swaminathan (2000), we briefly summarize our results (Table 7) and
discuss their relation to the behavioral models.
TABLE 7
RESULTS’ SUMMARY: IMPACT OF THE INDUSTRIAL DIVERSIFICATION ON THE PROFITABILITY TRADING
STRATEGIES
Horizon
01/1994-03/2000
04/2000-04/2004
01/1994-04/2004
Medium run
Momentum Effect
Contrarian Effect
Contrarian Effect
Long run
Momentum Effect
Contrarian Effect
Contrarian Effect
non significant
Medium run
Momentum Effect
Contrarian Effect
( 9th month)
Momentum Effect
non- significant
Sample
Non -Diversified
sub-sample
17
Long run
Momentum Effect
more pronounced
Contrarian Effect more
pronounced
Medium run
Momentum Effect
non significant
Contrarian Effect
( 6th month)
Contrarian Effect
Long run
Momentum Effect
less pronounced
Contrarian Effect
less pronounced
Contrarian Effect
non significant
Diversified subsample
none
First, we have noted that the trading strategies implemented on the non diversified sample do not earn
significant returns and they earn negative returns when implemented on the diversified sample in the medium run.
In the long run, however, all the trading strategies do not generate significant returns for the two samples. These
outcomes seem to be driven by the important variability of the return before and after the March 2000 crash. So, we
split our sample period at this date. We have found that, before the crash, the significant momentum effect observed
in the medium run in the non diversified sample continues to exist in the long run. The non significant momentum
effect recorded, in the medium run, in the diversified sample becomes significant in the long run. Nevertheless, the
momentum effect is more pronounced for the non diversified sample. After the crash, only a contrarian effect is
identified for both samples in both the medium and long run but it is more important for the non diversified sample.
Daniel et al. (1998) argue that stocks that are more difficult to value tend to generate greater
overconfidence among investors. Therefore mispricing should be more severe among securities that are hard to
value. If diversified firms tend to be difficult to value, then Daniel et al. (1998) would predict that return’s
anomalies should be stronger among these firms. This is inconsistent with our finding that both before and after the
crash mispricing is more pronounced among non diversified firms. The Hong and Stein (1999) model predicts that
momentum profits should be larger for stocks with slower information diffusion 2. If the following diversified firms
require more resources and skills, these firms suffer from insufficient spread of information. If this is the case, the
Hong and Stein model predict greater mispricing among these firms. Likewise, owing to less public information
available about these stocks, one might expect relatively more private information to be produced. The Daniel et al.
(1998) model suggests that mispricing is more severe among these stocks. Our results revealed this to be true
among non diversified samples. Thus, our results fit with the hypothesis implying that corporate diversification is
associated with a reduction in asymmetric information. Investor’s previsions about diversified firms seem to be less
biased. A plausible explanation is that the errors that outsiders make in the forecasting segment cash flows are not
18
perfectly positively- correlated so that the consolidated forecast is more accurate for a diversified than the forecast
for a focused firm. Therefore, the price of the diversified firms diverge less from their fundamental value.
Nevertheless, the behavioral models predict a medium- horizon momentum -effect that reverse in the long
horizon. Like Jegadeesh and Titman (2001), we found no evidence of reversals of the momentum effect in the
1990s (prior to the crash). After the crash, we identify only a contrarian effect and we find no evidence of
momentum. This result corroborates recent works implying that the momentum strategies are not profitable. Shiller
(2000), Campbell (2004) report that in the last twenty years simple momentum strategies have performed well on
average, but in the period 2000-2003, these strategies performed relatively poorly. This may be accounted for by the
widespread use of this strategy.
5.2 The Risk hypothesis
The momentum (contrarian) profits may be brought about by a cross-sectional dispersion in the expected
stock returns rather than to their time series patterns. According to Conrad and Kaul(1998), purchasing past winners
may imply a long position on high-risk stocks that earn higher expected returns. Inversely, selling past losers entails a
short position on low-risk stocks that earn lower expected returns. Since the standard models like CAPM or the threefactor model fail to account for the cross-section of expected returns, the bootstrap simulation with replacement was
proposed by Conrad and Kaul (1998). This technique allows to eliminate any time series’ properties in each
security’s returns while maintaining the cross-sectional distribution of its mean returns; However, Jegadeesh and
Titman (2002) show that the bootstrap with replacement introduce a small sample bias and they recommend the use
of the bootstrap simulation without replacement. This bootstrap experiment consists in re-sampling without
replacement, for each stock’s actual monthly returns: We generate 500 random samples by drawing without
replacement the monthly returns. On each random sample, called bootstrapped sample, we implement the WRSS
strategy.
Table 8 contains the average actual and average simulated profits of the 6-month ranking period strategies
implemented on the diversified and the non -diversified samples. The holding periods range from six to 24 months.
Panel A presents results for the 01/94-03/2000 time period and panel B presents results for the 01/94-04/2004. The
first column (1) of each panel displays the average actual profit. The second column of each panel, presents the
average simulated profits The simulated profit is an average of the 500 simulated returns yielded from the bootstrap
without replacement with 500 replications. The third column presents percentage of simulated profit in actual profit.
19
TABLE 8
AVERAGE PROFIT OF THE ACTUAL AND SIMULATED TRADING STRATEGIES
IMPLEMENTED ON DIVERSIFIED AND NON DIVERSIFIED FIRMS SAMPLES
Pannel A
01/94-03/2000
(1)
Actual
profit
6
(2)
Simulated
profit
(3)
Percentage
0.0024
-0.0008
-33.33
62.50
-0.0085*
-0.0009
10.59
0.0059
74.68
0.0001
-0.0011
-100.00
0.0042
0.0015
35.71
-0.0144**
-0.0016
11.11
0.0111**
0.0077
69.37
-0.0023
-0.0020
86.96
-0.0092
0.0019
-20.65
-0.0138*
-0.0021
15.22
NDiv
0.0183***
0.0111
60.66
-0.0007
-0.0020
285.71
Div
0.0160***
0.0043
26.88
0.3447
-0.0032
-0.93
NDiv
0.0521***
0.0135
25.91
0.0052
-0.0029
-55.77
Div
0.0154***
0.0069
44.81
-0.0070
-0.0043
61.43
NDiv
Div
9
NDiv
Div
12
NDiv
Div
18
24
(2)
Simulated
profit
(3)
Percentage
0.0060***
0.0040
66.67
0.0008
0.0005
0.0079**
Pannel B
01/94-04/2004
(1)
Actual
profit
This table contains average actual and average simulated profits to 6-month ranking period strategies
implemented on the diversified sample and the non diversified sample. The holding periods range from 6
to 24 months. Pannel A presents results for the 01/94-03/2000 time period and the pannel B presents
results for the 01/94-04/2004. The column (1) of each panel display average actual profit. The column,
(2) of each panel, presents average simulated profits. The simulated profit is an average of the 500
simulated returns obtained from the bootstrap without replacement with 500 replications. The column (3)
presents percentage of simulated profit in actual profit (percentage)
*Significant at 10% level
**Significant at 5%
***Significant at 1%
Before the crash, the simulated profit of the non -diversified sample increases from 0.4% for the 6-month
holding period strategy to 1.35% for the 24-month holding period strategy. However, the simulated profit of the
diversified sample rises from 0.05% for the 6-month holding strategy to 0.69% for the 24-month holding strategy.
This implies that the average simulated return goes up with the holding period but less than geometrically for the two
samples. Second, the simulated returns of the diversified firms’ sample are lower than those of the non-diversified
sample. The percentage of the simulated profit in actual profit is less than 100% for all the strategies and it doesn’t
exceed 75% for the non diversified firm sample and 63% for the diversified firm sample. Therefore, the crosssectional difference in expected returns does not, utterly, explain trad the ing strategies’ profits. By expanding the
time period to 2004, the results indicate that only for the diversified sample, a contrarian strategy generates
significant profits. The simulated profit of this strategy does not exceed 15.5%3.
6. Conclusion
3
Here we ignore non significant actual profit
20
Prior studies affirm that industrial diversification has an impact on the level of information asymmetry. The
behavioral models also suggest that momentum anomaly is affected by information asymmetry. Thus, the purpose of
this paper is to analyze the impact of industrial diversification on the profitability of contrarian and momentum
strategies. Using Monthly returns, the WRSS strategy is applied to 249 American listed stocks from January1,994 to
April ,2004. Unlike the previous studies, which used older data, our findings reveal that the momentum strategy
seems to stop being profitable. In order to study the impact of the 2000 crash on our results, we implemented the
WRSS strategy during the [01/1994-03/2000] and the [04/2000-04/2004] sub-periods. Our results indicate that while
momentum strategies earn large positive and significant returns during the first sub-period, these returns are negative
during the second sub- period. As Hong and Stein (2003) report, the 2000 market crash is qualified by negative
returns.
The stocks are divided into two groups according to the intensity of their industrial diversification using
Ward (1963) method of hierarchical classification. Our results entail that the trading strategies implemented on the
non -diversified sample do not earn significant returns but earn negative returns when implemented on the
diversified sample in the medium run. In the long run, however, all the trading strategies do not generate significant
returns for the two samples. These results seem to be driven by the great variability of return before and after the of
March 2000 crash. In fact, we find that, before the crash, the significant momentum effect observed in the medium
run in the non diversified sample continues to occur in the long run. The non -significant momentum effect
recorded, in medium run, in the diversified sample becomes significant in the long run. Nevertheless, the
momentum effect is more pronounced for the non diversified sample. After the crash, only a contrarian effect is
identified for both samples in medium and long run but it is more important for the non diversified sample. We then
try to explain these results with the predictions of several behavioral models. Like Lee and Swaminathan (2000), we
conclude that each model has specific features that help explain some aspects of our findings but that no single
model accommodates all our findings. Then, the Bootstrap without replacement is used to explore the risk- based
explanation of the profitability of the trading strategies. We find that the cross-sectional differences in expected
returns do not, entirely, account for the trading strategies’ profits.
BIBLIOGRAPHY
Barberis N, Shleifer A. and Vishny R. A model of investor sentiment. Journal of Financial Economics1998; 49;
307–343.
21
Boudoukh J, Richardson M. and Whitelaw R. “Industry returns and the Fisher effect”, Journal of Finance 1994; 49;
1595-1616.
Campbell, J. Y. “Understanding Momentum,” Arrowstreet Capital; 2004.
Clarke J E , Fee C. E. and Thomas S. Corporate diversification and asymmetric information: evidence from stock
market trading characteristics. Journal of Corporate Finance 2004; 10; 105-129
Conrad J, Kaul G. An Anatomy of Trading Strategies. Review of Financial Studies 1998; 11; 489-519.
Daniel K, Hirshleifer D. and Subrahmanyam A. Investor Psychology and Security Market Under-and Overreactions.
Journal of Finance 1998; 53; 1839-1886
Daniel K and Titman S. Market Efficiency in an Irrational World. Financial Analyst Journal 1999; 55; 28-40.
DeBondt W, Thaler R. H. Does the stock market overreact?”, Journal of Finance, 40 (1985), 793-805.
Denis, D. J.; Denis, D. K.; and Sarin, A. 1997. “Agency problems, equity ownership, and corporate diversification”.
Journal of Finance 52 (March): 135-60.
Dodd, P. Merger proposals, management discretion and stockholder wealth. Journal of Financial Economics 1980; 8
105-137.
Dunn K , Nathan S. The Effect of Industry Diversification on Consensus and Individual Analysts' Earnings
Forecasts. SSRN Working Paper 1998.
Fama, E. Market Effciency, Long-Term Returns, and Behavioral Finance. Journal of Financial Economics 1998; 49
283–306.
Givoly D, Hayn C, D'Souza J. Measurement Errors and the Information Content of Segment Reporting. Review of
Accounting Studies 1999; 29 ; 15-43
Habib M A, Johnsen B and Naik N. Spinoffs and information. Journal of Financial Intermediation 1997; 6 ;153-176
Hadlock, C, Ryngaert M and Thomas S. Corporate structure and equity offerings: are there benefits to
diversification. Journal of Business 1999; 74 ; 613-35
Hong H, Stein J. A Unified Theory of Underreaction, Momentum Trading and Overreaction in Asset Markets.
Journal of Finance 1999; 54 ; 2143-2184.
Hong H, Stein J. Differences of Opinion, Short-Sales Constraints and Market Crashes. Review of Financial Studies
2003; 16; 487-525.
Hong H, Chen J, and Stein J. Forecasting Crashes: Trading Volume, Past Returns and Conditional Skewness in
Stock Prices. Journal of Financial Economics 2001; 61; 345-381.
Jacquemin A P , Berry C H. Entropy measure of diversification and corporate growth. The Journal of Industrial
Economics 1979; 27; 359-369.
Jegadeesh N, and Titman S. Returns to buying winners and selling losers: Implications for stock market efficiency.
Journal of Finance 1993; 48; 65–91.
Jegadeesh N, and Titman S. Profitability of momentum strategies: An evaluation of alternative explanations. The
Journal of Finance 2001a; 56 ; 699–720.
Jegadeesh N, Titman S. Momentum. University Of Illinois. Working Paper 2001b.
Jegadeesh N, and Titman S. Cross-Sectional and Time-Series Determinants of Momentum Returns. Review of
Financial Studies 2002; 15; 143-157
22
Jensen M.C. Agency costs of free cash flow, corporate finance, and takeover. American Economic review 1986; 76;
323-329.
Lee C and Swaminathan B. Price Momentum and Trading Volume. Journal of Finance 2000; 55 ;1217-1269
Lewellen J. Momentum and autocorrelation in stock returns. The Review of Financial Studies 2002; 15; 533–563.
Lo A, and MacKinlay A C. When are Contrarian Profits Due to Stock Market Overreaction?. Review of Financial
Studies1990; 3 ; 175-208.
Malkiel B .The Efficient Market Hypothesis and Its Critics. Journal of Economic Perspectives 2003; 17; 59 – 82
Mengoli S. On the source of contrarian and momentum strategies in the Italian equity market. International Review
of Financial Analysis 2004; 13 ; 301– 331
Nanda V and Narayanan M P. Disentangling Value: Financing Needs, Firm Scope, and Divestitures. Journal of
Financial Intermediation 1999; 8 ; 174-204
Palepu K. Diversification strategy, profit performance and entropy measure. Strategic management journal 1985; 6
239-255.
Patro D K and Wu Y. Predictability of short-horizon returns in international equity markets. Journal of Empirical
Finance 2004; 11 ; 553-584
Rajan, R, Servaes H , Zingales L. The cost of diversity: The diversification discount and inefficient investment.
Journal of Finance 2000; 55; 35-80.
Richards A J. Winner-Loser Reversals in National Stock Market Indices: Can They Be Explained?. Journal of
Finance1997; 52 ; 2129-2144
Rouwenhorst K G. International momentum strategies. Journal of Finance1998; 53 ; 267–284
Schwert, G. W. Anomalies And Market Efficiency, 1st ed , Handbook of the Economics of Finance,
Shiller Robert J. Irrational Exuberance. Princeton University Press; 2000.
Stulz R M. Managerial discretion and optimal financing policies . Journal of Financial Economics 1990; 26; 3-27
Thomas S. Firm Diversification and Asymmetric Information: Evidence from Analysts’ Forecasts and Earnings
Announcements. Journal of Financial Economics 2002 ; 46 ; 373-396.
Ward J H. Hierarchical grouping to optimize an objective function. Journal of the American Statistical Association
1963; 58 ; 236–244.
Zuckerman E. Focusing the corporate product: Securities analysts and de-diversification. Administrative Science
Quarterly 2000; 45 ; 591-619
23
24