In this presentation we tested the effect of optimism on stock returns. To draw our conclusions on the research question we used two methodologies: the first one based on sorted portfolios and the second one based on regressions. We concluded that both monthly and annual returns of firms with optimistic expectations are consistently lower than firms with pessimistic expectations. The effect of optimism seems more strong during turbulent periods, varying depending on the magnitude of the forecast error and there are some key variables that allow us to explain part of the investor’s behavior facing the same miss in earnings.
1. Does optimism distort stock prices?
Recent
studies
have
suggested
that
stock
with
op3mis3c
expecta3on
obtain
lower
returns.
Authors:
Farei
Gabriele
Buchard
Francois
Kybourg
Philippe
Dessimoz
Benoît
Robyr
Steven
Yana
Yuan
01
December
2012
"The most common cause of low prices is pessimism, sometimes
pervasive, sometimes specific to a company or industry. We want to do
business in such an environment, not because we like pessimism but
because we like the prices it produces. It’s optimism that is the enemy of
the rational buyer." Warren Buffett
2. Table of contents
-‐
Reminder
of
our
research
topic
-‐
Database
-‐
Model
-‐
Results
-‐
InterpretaIons
-‐
Open
issues
-‐
Conclusions
Introduction
Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
2
3. Market
prices
seem
to
react
differently
to
the
earnings
announcement
depending
on
the
forecasts
provided
by
analysts.
• Analysts
are
professional
market
watchers
followed
closely
by
investors
In
our
model
the
posiIve
difference
between
the
forecasted
earnings
and
the
actual
earnings
is
called
op#mism.
This
is
the
result
of
individual
irraIonality:
• Analysts
are
not
completely
raIonal
individuals:
overly
opImisIc
forecasts
• Investors
are
not
completely
raIonal
individuals:
overly
opImisIc
reacIons
We
were
inspired
by
Stephen
Ciccone
–
«
Does
Analyst
Op3mism
About
Future
Earnings
Distort
Stock
Prices?
»
With
our
research
we
want
to
:
Understand
the
rela#on
between
op#mis#c
expecta#ons
and
stock
returns
and
the
factors
that
may
amplify
this
rela#onship.
• Does
the
opImism
distort
stock
prices?
• Is
the
effect
Ime
varying
or
constant?
Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
3
Research topic
Introduction
4. Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
4
Database construction: Raw data
Database
Source:
Center
for
Research
in
Security
Prices
(CRSP)
COMPUSTAT
The
InsItuIonal
Brokers
EsImate
System
(I/B/E/S)
Data:
• All
traded
companies
in
US
markets
Constraints:
• Traded
in
US
dollars
• SIll
acIve
• 20
years
of
historical
data
• Fiscal
year
end
month
=
December
Size
(~7400)
Issues:
By
selecIng
only
firms
with
December
fiscal
year-‐end
we
consciously
ignore
certain
sectors
where
tradiIonally
the
fiscal
year
differ.
A
representaIve
check
has
been
done
to
ensure
that
this
is
just
a
minority.
5. Representativiness of the database
Database
Whole
Sample
Sample
of
companies
reporIng
in
December
Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
5
• Fairly
good
representaIon
0
500
1000
1500
2000
2500
3000
3500
6. Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
6
Database construction: Raw data
Introduction
Raw
data
imported:
• Earning
Per
Share
[Annual]
Reported
Earning
per
share
at
the
end
of
the
fiscal
year
• Prices
[Monthly]
• Number
of
Shares
[Monthly]
• Book
Value
[Annual]
• EPS
Forecast
Mean
[Monthly]
Each
months
the
consensus
over
the
EPS
of
the
fiscal
year
change,
since
analyst
update
their
views
• EPS
Forecast
Standard
deviaIon
[Monthly]
Time
series
standard
deviaIon
of
the
EPS
Forecast
Mean
• EPS
Forecast
Number
of
EsImaIons
[Monthly]
Number
of
analysts
per
esImate
mean
7. Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
7
Database construction: variables
Model
Forecast
Proper#es:
Transparency
measures
• Dispersion
Measured
every
fiscal
year
for
each
sample
using
annual
earnings
forecasts,
used
as
proxy
of
the
opacity/transparency
of
the
firm.
Measured
at
December
31,
t-‐1.
• Forecast
error
Measured
as
the
absolute
normalized
difference
between
the
mean
of
the
forecasts
and
the
actual
EPS.
Used
as
a
proxy
of
the
opacity/transparency
of
the
firm.
Measured
at
December
31,
t-‐1.
Firms
with
high
dispersion/error
are
called
opaque.
Firms
with
low
dispersion/error
are
called
transparent.
Analyst
forecasts
Op#mism
component
Investors
Investors
percepIons
permanent
transitory
Transparent
Opaque
StarIng
from
the
previous
data
several
variables
has
been
constructed:
Op#mism
dummies:
OpImism
is
measured
as
being
the
posiIve
difference
between
forecast
mean
and
the
actual
EPS
,
if
it
is
negaIve
it
is
called
pessimism.
3
dummies
has
been
created
based
on
the
magnitude
of
opImism,
computed
as
being
the
raIo
between
the
forecast
error
and
the
actual
EPS.
RaIo
FE/EPS
0
0,15
0,55
1,00
Dummy
1
Dummy
2
Dummy
3
8. Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
8
Database construction: variables
Model
To
control
for
other
known
relaIonships:
• Book-‐to-‐Market
raIo
For
each
firm
in
the
database
we
computed
the
B2M
raIo
in
December
31,t-‐1.
We
couldn’t
measure
it
just
before
the
return
period
because
of
the
annual
nature
of
the
stockholders’
equity
book
value.
• Size
For
each
firm
in
the
database
we
computed
the
size
in
May
30,
t
:
Just
before
the
return
period.
The
previous
factors
are
added
to
control
the
relaIon
among
size
B2M
and
stock
returns
founded
by
Fama
and
French
in
1992.
• Prior
Loss
dummy
Some
recent
literature
indicate
that
prior
period
loss
affect
future
stock
returns
in
mulIple
ways
(increased
vola3lity
and
mean
return)
9. Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
9
Timeline
Model
June
1
Y
=
t
December
Y
=
t-‐1
Measurement
of
Op/mism/Pessimism
Prior
losses
B2M
Transparency
Measures
Size
Return
period
December
Y
=
t
30
May
Y
=
t
+1
Under
efficient
markets,
we
suppose
that
for
June
1,
the
prior
year
earnings
informaIon
should
be
fully
incorporated
into
the
stock
price
“RevelaIon”
window
“OpImism
rise”
window
• Our
tesIng
includes
only
companies
with
December
Fiscal
year-‐end.
• Two
disInct
return
period
called
:
“OpImism
rise”
window
and
“RevelaIon”
window
That
is
in
the
first
window
we
should
observe
opImism
entering
in
the
stock
price
and
in
the
second
window
we
should
observe
it
reveal
itself
by
a
decreasing
in
the
return.
OpImism
Return
decrease
Pricet
Pricet+1
10. Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
10
Database Management
Model
In
order
to
work
with
a
complete
database,
we
need
to
previously
perform
these
following
checkings:
1. ActualEPS&ForecastMean
are
valid
observaIons
at
the
same
Ime.
2. ForecastMean
implies
others
forecast
properIes.
3. Prices
have
to
be
valid
observaIons
6
months
before
and
aper
fiscal
year
end
to
capture
ingoing
and
outgoing
effects.
4. Previous
year
EPS
must
be
a
valid
observaIon
to
capture
previous
losses
effect.
Past
year
ForecastMean
and
STD
must
be
valid
observaIons
in
order
to
construct
a
transparency
measure.
5. Previous
year
Book
value
has
to
be
a
valid
observaIon.
We
assign
a
binary
to
each
of
the
previous
condiIons,
which
allows
us
to
take
the
firms
that
are
candidates
for
use
in
a
specific
context
(
specific
constraints
at
the
same
Ime)
.
Each
year
we
remove
the
companies
that
do
not
saIsfy
the
criterias
at
that
specific
Ime.
This
methodology
allowed
us
to
perform
dynamic
regressions
by
dynamicallay
changing
the
number
of
companies
as
we
progress
in
Ime,
increasing
the
number
of
firm
per
year
drasIcally.
11. Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
11
1st Approach: Portfolios
Model
• We
divided
in
two
different
porqolios,
the
opImist
companies
and
the
pessimist
companies,
and
we
observe
them
during
all
the
return
periods
of
the
past
20
years.
Then
we
take
the
difference:
Pessimist
returns
–
OpImisIc
returns.
• Firms
taken
in
consideraIons:
Binary1*Binary2
from
our
iniIal
database
[Valid
forecast
Mean,
EPS,
and
the
12
months
of
returns
used
in
the
return
period]
Here’s
our
dynamic
database
over
the
years,
the
total
firms-‐years
are
29’101.
t-1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
559
588
652
811
872
999
1160
1314
1354
1440
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
1437
1444
1517
1690
1854
2055
2236
2309
2304
2506
12. 1st Approach: Mean monthly return
Results
• PessimisIc
stocks
consistently
outperform
opImisIc
ones
• Between
June
to
December,
before
earnings
are
released
and
before
opImism
is
revealed,
the
return
is
greater.
• During
the
next
period,
where
the
opImism
is
gradually
revealed,
we
note
that
the
spread
between
both
is
reduced,
or
even
negaIve
in
May
Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
12
Month&
Mean&difference&(%):&
Pessimistic&Less&Optimistic&
June&t& 1.51%%
July&t& 1.87%%
August&t& 1.05%%
September&t& 1.62%%
October&t& 2.03%%
November&t& 0.96%%
December&t& 1.02%%
January&t+1& 1.20%%
February&t+1& 1.74%%
March&t+1& 0.58%%
April&t+1& 0.16%%
May&t+1& .0.11%%
%
EPS
announcement
13. 1st Approach: Annual Buy-and-Hold Returns
Results
Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
13
Annual
Buy-‐and-‐Hold
Returns,
July
Year
t
Through
June,
Year
t+1
Year
t
Pessimis#c
Op#mis#c
Difference
1992
0.28
0.101
0.179
1993
0.299
0.153
0.146
1994
0.146
0.021
0.125
1995
0.203
0.077
0.126
1996
0.473
0.248
0.225
1997
0.243
0.019
0.224
1998
0.385
0.243
0.142
1999
0.069
-‐0.089
0.158
2000
0.321
0.1
0.221
2001
0.323
0.029
0.294
2002
0.2
0.111
0.089
2003
0.177
0.015
0.161
2004
0.433
0.368
0.065
2005
0.17
0.068
0.102
2006
0.325
0.123
0.202
2007
0.227
0.111
0.116
2008
-‐0.043
-‐0.244
0.201
2009
-‐0.058
-‐0.264
0.206
2010
0.372
0.29
0.082
2011
0.313
0.229
0.084
Mean
0.243
0.085
0.157
• As
we
expected
the
difference
is
consistently
posiIve
in
every
year
• InteresIng
to
noIce
that
the
spread
seems
to
increase
during
turbulent
periods
(dot-‐com
bubble
and
financial
crisis)
14. 1st Approach: Optimise rise vs Revelation window
Results
Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
14
• As
we
intuiIvely
saw
in
the
monthly
average,
there
is
a
difference
between
the
first
window
and
the
revelaIon
window.
• As
opImism
reveal
itself,
the
return
of
opImist
firms
is
slightly
reduced
-0.400
-0.200
0.000
0.200
0.400
0.600
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
June-December
Pessimistic Optimistic
-0.600
-0.400
-0.200
0.000
0.200
0.400
0.600
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
January-May
Pessimistic Optimistic
15. 2nd approach: Enhanced Fama&MacBeth model
Model
Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
15
We
run
a
regression
using
the
previously
defined
variables
as
independent
variables
and
the
monthly
return
per
firm
as
dependent
variable.
We
did
it
for
each
company
candidate
for
a
regression
at
that
Ime,
for
each
month
of
each
return
period,
for
the
past
19
years.
!!
!"#$!!"
= ! + !. !!"#! + !. !2!! + !. !"#$%$&%!!"##$1! + !. !"#$%$&%!!"##$2! + !. !"#$%$&%!!"##$3!
+ !. !"#$%&#"'$()*'#%+"'! + !. !"#$"%$&&!!"##$! + !!!"!!!!"#!!!
!"#"$"%&!!"#$ =! !"#$%"&'!
!
!!!
!
0
200
400
600
800
1000
1200
1400
1600
1800
2000
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Dynamic database
• Tot
firms-‐year
=
21’989
16. 2nd approach: Enhanced Fama&MacBeth model
Model
Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
16
T-‐1
T
Return
period:
12
regressions
12
*
K
coefficients
=
K
AVG
coefficients
1991
2011
19
years
19*K
AVG
Coefficients
=
K
Final
Coefficients
K=
number
of
dependent
variables
+
constant
Intui#on
of
the
methodology:
Aper
we
have
run
the
regressions
for
each
month
and
each
year,
we
averaged
the
coefficients
in
order
to
interpret
on
average
how
our
model
behaves.
17. 2nd approach: Enhanced Fama&MacBeth model
Results
Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
17
Regression
1:
Op#mism
only
Constant
Dummy
Small
Error
Dummy
Medium
Error
Dummy
Large
Error
Coeff
1.8367
-‐0.9881
-‐0.8144
-‐1.2708
Tstat
2.7416
-‐0.9716
-‐0.9321
-‐1.2925
Pvalue
0.0763
0.3294
0.3328
0.2423
Regression
2
:
Using
Previous
Standard
Devi#ons
of
forecasts
as
transparency
measure
Constant
Dummy
Small
Error
Dummy
Medium
Error
Dummy
Large
Error
Book
to
Market
Size
Dummy
Prior
Loss
Previous
Forecast
STD
Coeff
1.6503
-‐0.9446
-‐0.8275
-‐1.3208
0.1619
0.0000
0.1949
0.0074
Tstat
2.2791
-‐0.9379
-‐0.8609
-‐0.9737
0.7602
-‐0.2741
-‐0.4647
0.1087
Pvalue
0.0880
0.3336
0.3718
0.2840
0.3493
0.3998
0.1650
0.4002
Regression
2
:
Using
Forecast
error
as
Transparency
measure
Constant
Dummy
Small
Error
Dummy
Medium
Error
Dummy
Large
Error
Book
to
Market
Size
Dummy
Prior
Loss
Previous
Forecast
Error
Coeff
1.2545
-‐0.6504
-‐0.8326
-‐0.9967
0.3433
0.0000
0.2732
0.0420
Tstat
1.5169
-‐0.6405
-‐0.6239
-‐0.6516
0.6621
-‐0.2822
0.0804
0.2287
Pvalue
0.1292
0.4199
0.3901
0.3423
0.3577
0.4220
0.2770
0.4294
18. Conclusions
Results
Behavioral
Finance
2012
–
Does
opImism
distort
stock
prices?
18
PorZolios
conclusion:
• Investor
opImism
is
reflected
in
stock
prices.
• The
disappointment
caused
by
the
missing
in
earnings
seem
to
reduce
the
stock
return
with
respect
to
firms
without
such
expectaIons
• We
can
idenIfy
pauern
of
opImisIc
firms
in
the
revelaIon
window
• Market
efficiency
seem
to
hold
since
opImisIc
return
is
fully
revealed
by
June
à
Investor
behavior
play
an
important
role
in
the
stock
market.
Regressions
conclusions:
• Lack
of
staIsIcal
significance,
model
issue
or
data
issue?
• Even
so,
the
signs
suggest
the
same
interpretaIon
given
in
the
porqolios
approach.
• The
raIonale
behind
this
negaIve
coefficients
could
be
that
investors
overesImate
growth
prospects
which
arIficially
increase
stock
prices.
As
the
opImisIc
expectaIons
are
not
fulfilled,
the
returns
of
these
stocks
are
low.
• The
relaIon
between
the
opImism
and
stock
return
seem
to
be
slightly
affected
by
the
change
of
measures
of
transparency:
the
opImism
component
seems
to
reduce
the
return
(negaIve
relaIon)