ArticlePDF Available

Do Women Panic More than Men? An Experimental Study on Financial Decision

Authors:

Figures

Journal
of
Behavioral
and
Experimental
Economics
52
(2014)
40–51
Contents
lists
available
at
ScienceDirect
Journal
of
Behavioral
and
Experimental
Economics
j
o
ur
na
l
ho
me
page:
www.elsevier.com/locate/jbee
Do
women
panic
more
than
men?
An
experimental
study
of
financial
decisions
Hubert
J.
Kissa,b,
Ismael
Rodriguez-Larac,d,
Alfonso
Rosa-Garciae,
aEötvös
Loránd
University
Department
of
Economics,
Pázmány
Péter
sétány
1/a,
1117
Budapest,
Hungary
bResearch
fellow
in
the
Momentum
(LD-004/2010)
Game
Theory
Research
Group
at
the
MTA
KRTK,
Hungary
cMiddlesex
University
London,
Business
School
Hendon
Campus,
The
Burroughs,
London
NW4
4BT,
UK
dResearch
fellow
at
LUISS
Guido
Carli
University,
Rome,
Italy
eFacultad
de
Ciencias
Juridicas
y
de
la
Empresa,
Universidad
Catolica
San
Antonio,
Campus
de
Los
Jeronimos,
s/n,
Guadalupe
30107,
Murcia,
Spain
a
r
t
i
c
l
e
i
n
f
o
Article
history:
Received
3
April
2013
Received
in
revised
form
10
June
2014
Accepted
16
June
2014
Available
online
24
June
2014
JEL
classification:
C91
D03
D8
G02
J16
Keywords:
Bank
run
Gender
difference
Strategic
uncertainty
Experimental
evidence
Coordination
a
b
s
t
r
a
c
t
We
report
experimental
evidence
on
gender
differences
in
financial
decision-making
that
involves
three
depositors
choosing
whether
to
keep
their
money
deposited
or
to
withdraw
it.
We
find
that
one’s
position
in
the
line,
the
fact
that
one
is
being
observed
and
observed
decisions
are
key
determinants
in
explain-
ing
the
subjects’
behavior.
Our
main
result
is
that
men
and
women
do
not
react
differently
to
what
is
observed.
However,
there
are
gender
differences
regarding
the
effect
of
being
observed:
women
value
the
fact
of
being
observed
more,
while
men
value
the
number
of
subsequent
depositors
who
observe
them.
Interestingly,
risk
aversion
has
no
predictive
power
on
depositors’
behavior.
©
2014
Elsevier
Inc.
All
rights
reserved.
1.
Introduction
Starting
with
the
run
on
Northern
Rock
in
the
UK
in
2007,
the
financial
crisis
has
shown
that
bank
runs
are
still
a
topic
of
first-order
importance
worldwide.
Other
examples
include
the
experiences
of
Washington
Mutual,
Bear
Stearns,
IndyMac
Bank,
the
Bank
of
East
Asia
and
the
fourth
largest
lender
in
Spain,
Bankia.
During
the
previous
waves
of
bank
runs
(the
last
occurred
during
the
Great
Depression),
the
proportion
of
male
depositors
was
higher
than
that
of
female
depositors.
However,
due
to
social
progress
and
changes
in
labor
conditions,
the
gap
in
the
proportion
of
men
and
women
among
banks’
depositors
is
closing.
Currently,
roughly
half
of
the
customers
with
an
account
at
a
formal
finan-
cial
institution
in
the
US
are
women;
this
finding
is
in
contrast
Corresponding
author.
Tel.:
+34
968278662.
E-mail
addresses:
hubert.kiss@tatk.elte.hu
(H.J.
Kiss),
I.Rodriguez-Lara@mdx.ac.uk
(I.
Rodriguez-Lara),
arosa@ucam.edu
(A.
Rosa-Garcia).
to
past
data
(e.g.,
according
to
Wright
(1999),
in
1828,
only
11%
of
the
customers
of
a
bank
in
Philadelphia
were
women).1Similar
changes
are
taking
place
in
developing
countries,
so
it
is
interesting
to
determine
whether
there
are
gender
differences
in
depositors’
behavior.
Gender
differences
in
preferences
have
been
identified
in
sev-
eral
dimensions
(see
Croson
and
Gneezy,
2009
for
a
review),
and
these
results
may
have
relevant
implications
concerning
how
bank
runs
unfold.
More
concretely,
episodes
of
bank
runs
involve
depos-
itors
observing
(at
least
partially)
what
other
depositors
have
done
(Iyer
and
Puri,
2012;
Kelly
and
Grada,
2000),
and
men
and
women
may
react
differently
to
their
observations
of
others’
actions
or
the
fact
that
they
are
being
observed.
Moreover,
women
generally
1Data
from
the
World
Bank
indicate
that
in
the
US,
84.1%
of
women
and
92%
of
men
over
age
15
have
an
account
at
a
formal
financial
institution.
These
amounts
(female/male)
are
fairly
close
to
other
developed
countries
(e.g.,
88.8%/92.5%
in
the
Euro
area
and
96.8%/96%
in
Japan),
but
there
are
sizable
differences
in
less
developed
countries
(e.g.,
2.6%/15.4%
in
Afghanistan
and
17%/32.7%
in
North
Africa).
2214-8043/$
see
front
matter
©
2014
Elsevier
Inc.
All
rights
reserved.
http://dx.doi.org/10.1016/j.socec.2014.06.003
H.J.
Kiss
et
al.
/
Journal
of
Behavioral
and
Experimental
Economics
52
(2014)
40–51
41
exhibit
a
higher
degree
of
risk
aversion
(e.g.,
Charness
and
Gneezy,
2012;
Croson
and
Gneezy,
2009),
making
them
possibly
more
likely
to
withdraw
funds
early
to
guarantee
a
sure
payoff.
We
report
experimental
evidence
on
gender
differences
in
depositor
decision-making.
Our
experimental
design
is
based
on
the
coordination
problem
formulated
by
Diamond
and
Dybvig
(1983),
which
we
modify
to
allow
for
different
levels
of
observ-
ability.
We
consider
the
simplest
model,
where
one
impatient
depositor
needs
to
withdraw
funds
immediately
and
two
patient
depositors
with
no
urgent
liquidity
needs
decide
between
keeping
their
funds
deposited
(which
we
also
call
“waiting”)
and
withdraw-
ing
their
funds,
with
the
former
action
yielding
the
highest
payoff
if
they
both
choose
it.
We
define
a
bank
run
as
a
situation
in
which
at
least
one
of
the
patient
depositors
withdraws
funds.
In
line
with
Diamond
and
Dybvig
(1983),
liquidity
needs
are
private
informa-
tion,
and
there
is
no
aggregate
uncertainty
about
the
number
of
patient
and
impatient
depositors.
One
noteworthy
aspect
of
our
design,
however,
is
that
depositors
choose
sequentially
between
waiting
or
withdrawing
their
money,
implying
that
(i)
depositors
may
observe
what
other
depositors
have
done
before
making
their
decision,
and
(ii)
depositors
know
whether
other
depositors
will
observe
their
decisions.2
Based
on
previous
results
in
the
literature
(Garratt
and
Keister,
2009;
Kiss
et
al.,
2014),
we
hypothesize
that
in
addition
to
gender,
three
forces
may
affect
the
decisions
of
the
patient
depositors.
The
first
concerns
their
observations
of
other
depositors’
decisions.
For
example,
knowing
that
another
depositor
has
already
withdrawn
funds
may
foster
panicking
behavior
and
favor
further
withdrawals
because
a
patient
depositor
observing
that
someone
else
with-
draws
funds
does
not
know
if
(s)he
is
observing
an
impatient
or
a
patient
depositor.
On
the
other
hand,
depositors
at
the
begin-
ning
of
the
line
may
behave
differently
if
subsequent
depositors
are
observing
their
actions.
More
precisely,
if
a
patient
depositor
is
observed
by
the
other
depositors,
then
(s)he
may
decide
to
wait
to
induce
the
other
patient
depositor
to
wait
as
well,
guarantee-
ing
the
highest
possible
payoff.
Finally,
we
aim
to
analyze
whether
attitudes
toward
risk
have
some
predictive
power
in
depositors’
decisions
because
risk
aversion
has
been
frequently
considered
a
key
determinant
in
financial
decisions.
In
addition
to
the
previously
mentioned
factors,
this
paper
also
examines
whether
there
are
gender
differences
in
withdrawal
decisions
after
controlling
for
risk
preferences.
Gender
differences
in
other
financial
settings
(apart
from
depositor
behavior)
have
been
studied
extensively.
Many
studies
analyze
gender
differ-
ences
in
different
investment
decisions
and
in
portfolio
selection
(e.g.,
Bernasek
and
Shwiff,
2001;
Dwyer,
Gilkeson,
and
List,
2002;
Felton,
Gibson,
and
Sanbonmatsu,
2003;
Martenson,
2008;
Sunden
and
Surette,
1998;
Watson
and
McNaughton,
2007),
finding
that
women
are
more
risk
averse
and
choose
more
conservative
invest-
ment
strategies.
Other
papers
find
gender
differences
in
the
way
people
reacted
to
the
recent
financial
crisis
(see,
for
instance,
Söderberg
and
Wester,
2012),
finding
that
women
were
less
likely
to
take
action
in
response
to
the
distress.
Although
there
is
a
growing
experimental
and
empirical
literature
on
bank
runs
(see
Schotter
and
Yorulmazer,
2009;
Garratt
and
Keister,
2009;
Starr
and
Yilmaz,
2007;
Iyer
and
Puri,
2012;
Brown,
Trautmann,
and
Vlahu,
2012,
for
some
recent
examples,
and
Dufwenberg,
2012,
for
a
sur-
vey
on
experimental
banking,
including
a
section
on
bank
runs),
to
the
best
of
our
knowledge,
this
is
the
first
experimental
study
that
specifically
investigates
gender
differences
in
this
context.
We
2Sequential
decisions
have
recently
been
considered
in
bank
run
experiments
(see
Garratt
and
Keister,
2009;
Kiss
et
al.,
2012,
2014;
Schotter
and
Yorulmazer,
2009).
are
only
aware
of
two
empirical
studies
that
refer
to
gender
differ-
ences
in
bank
run
situations
(Kelly
and
Grada,
2000;
O’Grada
and
White,
2003).
Both
study
two
bank
runs
in
New
York
in
1854
and
1857,
and
gender
was
not
clearly
found
to
play
a
role
in
explaining
panicking
behavior.
The
remainder
of
the
paper
is
organized
as
follows.
In
Section
2,
we
present
the
bank
run
game
that
is
played
in
our
experiment,
which
is
detailed
in
Section
3.
We
summarize
our
research
ques-
tions
in
Section
4.
Section
5
contains
the
experimental
results,
and
Section
6
concludes.
2.
The
bank
run
game
with
observability
of
actions
In
this
section,
we
describe
the
coordination
problem
that
is
played
in
each
round
of
the
experiment.
We
extend
the
model
of
Diamond
and
Dybvig
(1983)
to
allow
for
observability
of
actions,
following
Kiss
et
al.
(2014).
The
game
has
three
different
stages,
as
detailed
below.
2.1.
Time
t
=
0.
Deposits
At
t
=
0,
a
bank
with
three
depositors
is
formed.
Each
depositor
deposits
her/his
initial
endowment
(in
our
experiment,
80
ECUs)
in
this
bank,
which
therefore
initially
has
240
ECUs
to
be
invested
in
a
project.
The
project
yields
a
guaranteed
high
return
in
period
t
=
2,
but
the
investment
can
be
liquidated
at
no
cost
at
t
=
1.
2.2.
Time
t
=
1.
Types,
network
structure
and
depositors’
decisions
At
t
=
1,
the
depositors
must
choose
whether
they
want
to
with-
draw
their
money
from
the
bank
or
keep
it
deposited.
We
assume
that
one
of
the
depositors
is
hit
by
a
liquidity
shock
at
the
begin-
ning
of
t
=
1
and
is
forced
to
withdraw
money.
We
follow
Diamond
and
Dybvig
(1983)
and
further
assume
that
there
is
no
aggregate
uncertainty
about
the
liquidity
demand;
i.e.,
it
is
common
knowl-
edge
that
one
of
the
three
depositors
will
need
the
money
and
will
withdraw
with
certainty.
We
refer
to
this
depositor
as
the
impa-
tient
depositor,
whereas
the
depositors
who
can
wait
to
withdraw
their
money
are
called
patient
depositors.
Both
the
patient
and
impatient
depositors
choose
their
actions
in
an
exogenously
determined
sequence.
Depositor
i
chooses
in
position
i,
where
i
=
1,2,3.
Before
choosing
between
withdrawing
or
waiting,
depositor
i
learns
whether
a
subsequent
depositor
j
>
i
will
observe
his/her
choice.
If
depositor
j
does
observe
the
choice
of
depositor
i,
we
say
that
the
link
ij
exists,
for
i,j
{1,2,3},
and
i
<
j.3
We
model
the
information
flow
among
the
depositors
through
a
network.
A
network
is
the
set
of
existing
links
among
the
depos-
itors.
In
our
setup,
there
are
8
possible
networks:
(12,
13,
23),
(12,
13),
(12,
23),
(13,
23),
(12),
(13),
(23),
(),
where
()
stands
for
the
empty
network
that
has
no
links
at
all,
whereas
the
structure
(12,
13,
23)
contains
all
of
the
possible
links;
i.e.,
the
link
12,
the
link
13
and
the
link
23.
This
later
network
therefore
represents
a
fully
sequential
setup,
meaning
that
the
depositors
observe
all
of
their
predecessors’
actions.
In
particular,
(i)
depositor
1
knows
that
depositors
2
and
3
will
observe
his/her
decision,
(ii)
depositor
2
chooses
after
learning
what
depositor
1
has
done
and
is
aware
that
depositor
3
will
observe
his/her
decision,
and
(iii)
depositor
3
makes
his/her
decision
after
learning
what
depositors
1
and
2
have
done.
The
empty
network
()
represents
the
opposite
situation.
This
network
resembles
the
simultaneous
move
in
Diamond
and
Dybvig
(1983),
where
depositors
decide
with
no
information
about
3As
a
natural
consequence
of
the
existence
of
the
link
ij,
depositor
j
chooses
after
learning
what
depositor
i
has
done,
for
i
<
j
and
i,j
{1,2,3}.
42
H.J.
Kiss
et
al.
/
Journal
of
Behavioral
and
Experimental
Economics
52
(2014)
40–51
Table
1
The
payoff
table.
Number
of
previous
withdrawals
If
you
withdraw
the
first
year
In
you
decide
to
wait
in
the
first
year
and
withdraw
in
the
second,
the.
.
.
If
you
both
wait
and
only
the
computer
withdraws
If,
in
addition
to
the
computer,
the
other
depositor
withdraws
0
100
140
60
1
100
140
60
2
40
Not
applicable
60
the
other
depositors’
actions.4In
our
model,
depositors
have
only
local
knowledge
of
the
information
structure;
that
is,
they
know
their
own
links
but
do
not
know
whether
the
other
two
depositors
are
linked.5
The
depositors’
payoffs
are
independent
of
their
type
and
of
the
network
structure,
but
their
payoffs
do
depend
on
their
position
in
the
line
and
their
decisions.
If
a
depositor
decides
to
withdraw,
(s)he
immediately
receives
100
ECUs
as
long
as
there
is
enough
money
in
the
bank
to
pay
this
amount
(out
of
this
amount,
80
ECUs
cor-
respond
to
the
initial
endowment
and
20
ECUs
are
obtained
in
the
form
of
interest).
In
our
experiment,
if
depositors
1
or
2
withdraw,
they
definitely
receive
100
ECUs.
However,
if
depositor
3
decides
to
withdraw
after
two
withdrawals,
(s)he
only
receives
40
ECUs
(because
the
first
two
depositors
who
withdrew
received
100
ECUs
each,
and
the
bank
has
only
40
ECUs
to
pay
depositor
3).
However,
if
depositor
3
withdraws
after
less
than
two
withdrawals,
the
bank
pays
her/him
100
ECUs.
2.3.
Time
t
=
2.
Depositors
who
waited
receive
their
payoffs
Depositors
who
decide
to
wait
at
t
=
1
receive
their
payoff
in
period
t
=
2.
The
amount
that
the
depositors
receive
in
t
=
2
depends
on
the
total
number
of
waitings.
If
only
one
depositor
keeps
her/his
money
deposited,
(s)he
receives
60
ECUs.
If
two
depositors
wait,
then
their
payoff
is
140
ECUs.6
2.4.
Coordination
problem
and
strategic
uncertainty
Our
game
is
such
that
for
patient
depositors,
the
payoff
conse-
quences
of
each
decision
can
be
summarized
in
Table
1.7
Given
these
payoffs,
depositor
3
should
always
wait
if
(s)he
is
patient.8For
a
patient
depositor
in
position
1
or
2,
it
pays
off
to
wait
4However,
position
in
the
line
is
known
in
our
setup.
5For
instance,
in
the
complete
network
(12,13,23)
depositor
3
chooses
after
observing
what
depositors
1
and
2
have
done.
Because
depositor
3
does
only
know
his/her
own
links,
(s)he
does
not
know
whether
depositor
2
has
observed
depositor
1’s
decision
or
not.
In
that
regard,
when
depositor
3
chooses,
(s)he
does
not
know
whether
(s)he
is
in
network(12,13,23)
or
(13,23).
Both
possibilities
are
equally
likely
in
our
framework.
6Remember
that
one
of
the
three
depositors
is
impatient
and
always
withdraws
at
t
=
1.
7The
subjects
received
a
copy
of
this
table
in
the
instructions
for
the
experiment,
where
the
computer
that
acted
as
the
impatient
depositor
was
programmed
to
always
withdraw
(see
Section
3).
8In
our
setup,
waiting
in
position
3
yields
140
ECUs
(60
ECUs)
if
only
the
impa-
tient
depositor
has
withdrawn
(if
in
addition
to
the
impatient
depositor,
the
other
patient
depositor
has
withdrawn),
whereas
withdrawing
yields
100
ECUs
(40
ECUs).
We
note
that
the
amount
that
a
patient
depositor
receives
if
(s)he
waits
alone
(60
ECUs)
is
smaller
than
her
initial
endowment
(80
ECUs)
but
larger
than
the
amount
that
remains
in
the
bank
after
two
withdrawals
(40
ECUs).
The
rationale
for
these
payoffs
can
be
related
to
the
idea
of
partial
deposit
insurance.
If
a
bank
run
is
already
underway,
depositors
who
decided
to
wait
may
receive
more
than
what
remains
in
the
bank
after
paying
all
of
the
depositors
in
t
=
1,
although
these
depositors
may
not
receive
their
initial
endowment.
Kiss
et
al.
(2012)
investigate
how
the
level
of
if
(s)he
knows
or
believes
that
the
other
patient
depositor
does
so
as
well.
We
define
a
bank
run
as
a
situation
in
which
at
least
one
of
the
patient
depositors
withdraws.
Although
this
situation
resembles
the
coordination
problem
in
Diamond
and
Dybvig
(1983),
one
distinctive
feature
of
our
model
is
that
other
depositors’
decisions
can
be
observed,
depending
on
the
network
structure,
which
substantially
affects
the
degree
of
strate-
gic
uncertainty.
As
is
standard
in
the
literature,
we
define
strategic
uncertainty
as
uncertainty
regarding
the
purposeful
decision
of
players
in
an
interactive
situation.
More
precisely,
in
a
setup
in
which
a
patient
depositor
observes
a
withdrawal,
(s)he
will
not
know
if
the
withdrawal
was
due
to
the
impatient
depositor
or
to
the
patient
depositor
who
decided
to
run
the
bank.
If
a
patient
depos-
itor
observes
a
waiting,
there
is
no
doubt
about
the
action
of
the
other
patient
depositor
(i.e.,
strategic
uncertainty
disappears).
This
feature
is
absent
in
the
study
by
Diamond
and
Dybvig
(1983),
where
decisions
are
taken
simultaneously
(i.e.,
patient
depositors
always
choose
without
knowing
what
other
depositors
do).
3.
Experimental
design
and
procedures
We
recruited
a
total
of
60
subjects
(30
men
and
30
women)
with
no
previous
experience
in
coordination
problems
or
experiments
on
financial
decisions.
We
ran
two
sessions
at
the
Laboratory
for
Research
in
Experimental
Economics
(LINEEX)
of
Universidad
de
Valencia
in
June
2013,
with
an
even
distribution
of
genders
within
each
session.
All
of
the
participants
in
the
experiment
were
stu-
dents
from
the
Economic
and
Business
School
at
the
university.
The
men
and
women
who
participated
in
the
experiment
did
not
differ
by
background,
minimizing
the
possibility
of
a
selection
bias.9
The
experiment
was
programmed
using
the
z-Tree
software
(Fischbacher,
2007).
Instructions
were
read
aloud
and
the
bank
run
game
described
in
Section
2
was
played
for
15
rounds.
At
the
beginning
of
each
round,
each
subject
was
informed
that
(s)he
had
been
matched
randomly
with
another
subject
and
assigned
a
third
depositor
(simulated
by
the
computer)
to
form
a
three-depositor
bank.
Likewise,
each
subject
was
told
that
(s)he
had
deposited
her/his
initial
endowment
(80
ECUs)
in
the
common
bank.
All
of
this
information
was
known
publicly,
as
was
the
fact
that
the
com-
puter
was
programmed
to
always
withdraw
and
would
act
as
the
impatient
depositor.
Appendix
A
contains
the
instructions.
In
each
round,
each
depositor
was
privately
informed
about
her/his
position
in
the
sequence
of
decisions
(i
=
1,2,3).
It
was
com-
mon
knowledge
that
this
position
was
randomly
and
exogenously
determined,
so
subjects
were
equally
likely
to
be
at
any
position.
The
depositors
then
decided
in
sequence
(according
to
their
posi-
tion
in
the
line)
whether
to
withdraw
their
money
from
the
bank
or
keep
it
deposited.
If
the
action
of
a
depositor
would
be
observed
by
some
subsequent
depositor,
(s)he
was
also
informed.
Similarly,
if
a
depositor
was
observing
a
predecessor
in
the
line,
(s)he
was
informed
about
that
depositor’s
action
before
making
her/his
own
decision.
In
Fig.
1,
we
present
some
screenshots
from
our
experi-
ment.
As
shown
in
Fig.
1,
the
network
structure
was
presented
graph-
ically
to
the
subjects
on
the
right-hand
side
of
the
screen.
On
the
left-hand
side,
a
block
of
text
informed
the
subjects
about
their
deposit
insurance
affects
depositors’
decisions,
depending
on
whether
the
decisions
are
sequential
or
simultaneous.
Another
possible
interpretation
of
the
payoffs
(that
we
use
in
our
experiment)
is
that
the
amount
not
withdrawn
from
the
bank
earns
money
on
the
project
that
it
was
invested
in.
The
project
yields
a
sure
return,
but
benefits
will
depend
upon
the
amount
that
is
invested
in
the
project
till
the
end.
9The
Kruskal–Wallis
test
indicates
that
there
are
no
gender
differences
with
regard
to
age
or
major
(p-values
=
0.5535
and
0.8416,
respectively).
See
Appendix
B
for
further
details.
H.J.
Kiss
et
al.
/
Journal
of
Behavioral
and
Experimental
Economics
52
(2014)
40–51
43
Fig.
1.
Screenshots
for
depositors
1
and
2.
44
H.J.
Kiss
et
al.
/
Journal
of
Behavioral
and
Experimental
Economics
52
(2014)
40–51
position
in
that
round
and
verbally
summarized
the
(local)
infor-
mation
flow
among
the
depositors.
In
Fig.
1A,
we
illustrate
the
case
of
depositor
1,
who
knows
that
depositor
2
will
observe
her/his
decision.
Because
there
is
no
link
between
depositor
1
and
depos-
itor
3,
depositor
1
knows
that
depositor
3
will
not
observe
her/his
action.
The
fact
that
the
information
is
local
implies
that
depositor
1
is
unaware
whether
depositor
3
observes
depositor
2
(note
the
“?”
symbol
on
the
line
that
connects
depositors
2
and
3).
Fig.
1B
presents
the
case
of
a
depositor
2
who
has
observed
a
waiting
and
must
then
decide
what
to
do.
Depositor
2
knows
that
depositor
3
will
observe
her/his
decision.
At
the
bottom
of
the
screen,
the
sub-
jects
were
reminded
of
the
payoff
consequences
of
each
possible
action
and
the
fact
that
the
computer
was
programmed
to
always
withdraw
(acting
as
the
impatient
depositor).
In
each
round,
the
subjects
were
asked
to
choose
between
waiting
or
withdrawing,
with
their
position
in
the
line
and
the
information
structure
changing
across
rounds
(i.e.,
in
each
round,
the
subjects
were
placed
in
a
different
position
and/or
they
were
placed
in
a
different
network
so
that
their
links
were
different).10
In
both
sessions,
the
subjects
were
divided
into
three
match-
ing
groups
of
10.
Subjects
from
different
matching
groups
never
interacted
with
each
other
during
the
session.
Subjects
within
the
same
matching
group
were
randomly
and
anonymously
matched
in
pairs
at
the
end
of
each
round.
At
the
end
of
the
experiment,
the
subjects
filled
out
a
question-
naire
that
was
used
to
collect
additional
information
about
gender
and
degree
of
risk
aversion.
We
elicited
risk
attitudes
using
the
investment
decision
in
the
study
by
Gneezy
and
Potters
(1997).
Each
subject
hypothetically
received
10
Euros
and
was
asked
to
choose
how
much
of
it,
$x,
(s)he
wanted
to
invest
in
a
risky
option
and
how
much
(s)he
wished
to
keep.
The
amount
invested
yielded
a
dividend
$2.5x
with
½
probability;
it
was
otherwise
lost.
The
money
not
invested
in
the
risky
option
$(10
x)
was
kept
by
the
subject.
In
this
situation,
the
expected
value
of
investing
is
higher
than
the
expected
value
of
not
investing;
therefore,
a
risk-neutral
(or
risk-
loving)
subject
should
invest
the
10
Euros,
whereas
a
risk-averse
subject
will
invest
less.
The
amount
not
invested
in
the
risky
asset
is
a
natural
measure
of
risk
aversion.11
Each
session
lasted
approximately
90
min,
and
the
subjects
received
on
average
16
Euros.
For
the
payment,
we
used
a
random
lottery
incentive
procedure
by
which
one
choice
(i.e.,
one
of
the
rounds)
was
paid
out,
with
ECUs
transformed
into
Euros
using
the
exchange
rate
10
ECUs
=
1
Euro.
4.
Research
questions
We
want
to
explore
gender
differences
in
behavior
in
a
bank
run
situation,
where
depositors
may
observe
other
depositors’
actions.
Kiss
et
al.
(2014)
show
that
if
they
are
being
observed,
depositors
tend
to
wait
to
induce
other
depositors
to
follow
suit.12 Our
first
research
question
is
therefore
to
investigate
gender
differences
in
behavior
when
depositors
know
that
subsequent
depositor(s)
will
observe
their
decisions.
10 Observations
are
balanced
across
networks
so
that
men
and
women
are
equally
likely
to
decide
in
any
particular
network
(see
Appendix
B).
11 Our
questionnaire
also
contained
four
different
lotteries
in
the
spirit
of
Holt
and
Laury
(2002).
Although
henceforth
we
use
the
risk
aversion
elicited
á
la
Gneezy
and
Potters
(1997),
we
can
also
proxy
risk
aversion
by
the
number
of
times
that
a
subject
made
the
risky
choice
in
the
Holt
and
Laury
task.
The
results
presented
in
the
next
section
are
invariant
to
the
measure
of
risk
aversion
that
we
use
because
both
measures
are
correlated
(Correlation
coefficient
=
0.26,
p-value
<
0.0000).
Appendix
B
presents
further
details
about
our
measures
of
risk
aversion.
12 This
finding
is
in
line
with
experimental
evidence
in
public
good
games
showing
that
if
they
are
set
in
a
network
structure,
subjects
may
decide
to
contribute
to
show
their
strategic
commitment
(Choi
et
al.,
2011).
Q1.
How
do
men
and
women
behave
when
nobody
observes
their
choices?
How
do
they
choose
when
they
know
that
subsequent
depositor(s)
will
observe
their
decisions?
In
our
game,
patient
depositors
receive
the
highest
possible
payoff
if
they
coordinate
on
waiting,
but
they
can
observe
what
another
depositor
has
done
before
only
if
the
network
structure
allows
them
to
do
so.
In
turn,
depositors’
decisions
can
take
place
in
the
presence
(absence)
of
strategic
uncertainty.
As
noted
in
Sec-
tion
2,
strategic
uncertainty
is
absent
if
a
subject
can
infer
the
decision
of
another
subject
in
the
room
without
a
problem
before
deciding
what
to
do.
Depositor
1
has
no
information
about
what
other
depositors
have
done
when
making
his/her
choice,
so
(s)he
always
decides
in
an
environment
of
strategic
uncertainty.
Depos-
itors
2
and
3,
however,
make
their
decisions
in
the
absence
of
strategic
uncertainty
if
they
observe
a
waiting,
or
if
depositor
3
observes
the
two
previous
actions.
When
depositor
2
or
depositor
3
observes
either
nothing
or
a
withdrawal,
their
decisions
are
made
in
a
context
of
strategic
uncertainty.
Our
second
research
question
is
aimed
at
assessing
the
importance
of
strategic
uncertainty
as
well
as
investigating
how
men
and
women
react
to
what
they
observe.13
Q2.
Does
strategic
uncertainty
(e.g.,
the
observation
of
a
with-
drawal)
favor
withdrawals?
Do
men
and
women
react
differently
to
what
they
observe?
One
of
the
areas
in
which
gender
differences
have
been
observed
is
risk
attitudes.
In
that
vein,
an
interesting
question
to
be
addressed
concerns
the
predictive
power
of
risk
aversion.
Consider
a
patient
depositor
at
position
1.
Upon
withdrawal,
(s)he
definitely
receives
100
ECUs,
while
waiting
may
yield
140
ECUs
or
60
ECUs,
depending
on
the
decision
of
the
other
patient
depositor.
Let
p
denote
the
probability
that
the
patient
depositor
in
position
1
assigns
to
the
event
that
the
other
patient
depositor
will
wait.
Depositor
1
will
be
indifferent
between
waiting
and
withdrawal
if
u(100)
=
pu(140)
+
(1
p)u(60)
(1)
If
we
assume
that
utilities
are
described
by
the
constant
relative
risk
aversion
utility
function14 u(c)
=
c1/(1
),
where
>
1
rep-
resents
the
degree
of
risk
aversion,
then
we
can
substitute
it
into
the
above
expression
and
obtain:
p
=(100)1
(60)1
(140)1
(60)1
The
derivative
of
p
with
respect
to
is
positive,
indicating
that
the
probability
assigned
to
the
event
that
the
other
depositor
waits
should
be
higher
for
more
risk-averse
depositors,
so
that
the
indif-
ference
relation
holds.15 Therefore,
more
risk-averse
depositors
are
willing
to
wait
only
if
they
are
more
optimistic
about
the
other
depositor’s
decision.
More
precisely,
if
two
depositors
have
the
same
belief
about
the
likelihood
of
successful
coordination,
and
for
the
less
risk-averse
depositor
relation
(1)
holds,
then
the
more
13 To
the
best
of
our
knowledge,
there
are
no
other
studies
that
investigate
how
men
and
women
react
when
varying
the
degree
of
strategic
uncertainty,
although
there
are
studies
that
investigate
gender
differences
in
games
that
are
character-
ized
by
strategic
uncertainty
(e.g.,
in
the
ultimatum
game,
strategic
uncertainty
arises
because
the
proposer
does
not
know
the
minimum
acceptable
offer
of
the
responder).
In
these
games,
no
clear,
general
difference
in
choices
between
men
and
women
has
been
identified.
14 This
assumption
is
common
in
the
theoretical
literature
on
bank
runs.
For
instance,
Green
and
Lin
(2003)
and
Ennis
and
Keister
(2009)
use
this
utility
function
that
cleanly
captures
risk
aversion
through
.
15 This
result
does
not
depend
on
the
specific
payoffs,
and
it
holds
for
any
payoff
scheme
that
has
the
same
relationship
between
payoffs:
the
highest
payoff
goes
to
the
patient
depositors
if
both
wait,
followed
by
the
payoff
related
to
withdrawal,
and
the
lowest
payoff
corresponds
to
waiting
alone.
H.J.
Kiss
et
al.
/
Journal
of
Behavioral
and
Experimental
Economics
52
(2014)
40–51
45
0
0.1
0.2
0.3
0.4
0.5
Depositor 1
Depositor 2
Depositor
3
Men Women
Withdrawal rate
Fig.
2.
Decisions
in
each
position
by
gender.
risk-averse
depositor
prefers
withdrawal
to
waiting. 16 Heinemann
et
al.
(2009)
find
that
risk
aversion
is
negatively
correlated
with
the
likelihood
of
choosing
the
risky
choice
in
a
coordination
game
(for
further
evidence,
see
Goeree
et
al.,
2003),
although
gender
differ-
ences
are
not
analyzed.
Q3.
Does
risk
aversion
predict
withdrawal
decisions?
Do
women
(who
are
generally
more
risk
averse
than
men)
withdraw
more
often
than
men
do,
ceteris
paribus?
5.
Experimental
results
5.1.
Aggregate
data
This
section
presents
our
main
results.
In
Fig.
2,
we
report
the
likelihood
of
withdrawal
in
each
position
by
gender.
Fig.
2
shows
that
the
position
in
the
line
affects
the
depositors’
behavior,
with
depositors
more
likely
to
withdraw
in
position
1
or
2
than
in
posi-
tion
3.
Remember
that
depositor
3
has
a
dominant
strategy
and
should
always
wait
if
patient.
Gender
does
not
seem
to
system-
atically
affect
depositors’
behavior.
We
observe
that
women
(men)
withdraw
more
frequently
in
position
2
and
position
3
(position
1),
but
gender
differences
are
never
significant
(p-values
>
0.146).17
As
noted
above,
we
are
interested
in
looking
at
the
depositors’
behavior
and
gender
differences
in
behavior
with
regard
to
strate-
gic
uncertainty.
In
Fig.
3,
we
group
the
data
according
to
the
cases
in
which
depositors
2
and
3
choose
in
the
presence
(absence)
of
strate-
gic
uncertainty.
We
do
not
consider
this
distinction
for
depositor
1
because
(s)he
always
decides
in
a
context
of
strategic
uncertainty.
In
line
with
Fig.
2,
we
observe
that
men
and
women
behave
similarly
once
we
condition
on
the
presence
or
absence
of
strate-
gic
uncertainty
and
on
a
particular
position
in
the
line
(p-values
>0.443).
This
finding
is
in
line
with
previous
evidence
in
the
literature
on
coordination
problems
that
reports
no
gender
differences
(e.g.,
Dufwenberg
and
Gneezy,
2005;
Heinemann
et
al.,
2009),
although
our
game
differs
from
the
previous
studies
in
that
we
allow
for
the
observability
of
actions
that
are
conditional
on
the
information
structure.
Along
these
lines,
one
interesting
insight
from
Fig.
3
is
that
strategic
uncertainty
increases
the
withdrawal
rate
for
both
men
and
women
in
positions
2
and
3
(p-values
<
0.0080).
This
16 Heinemann
et
al.
(2009)
find
evidence
that
supports
the
idea
that
participants
decide
as
if
they
have
probabilistic
beliefs
about
the
outcome
of
coordination
games.
There
is
scarce
empirical
or
experimental
evidence
about
the
relationship
between
risk
aversion
and
belief
about
another
individual’s
decision.
Fehr-Duda
et
al.
(2006)
elicit
probability
weights
using
gambles
and
find
that
on
average,
women
are
more
pessimistic
than
men
in
the
gain
domain,
and
they
are
also
more
risk
averse.
Translated
to
our
environment,
this
result
would
imply
that
women
believe
less
in
obtaining
the
largest
payoff
(140
ECUs)
through
successful
coordination.
17 Unless
otherwise
noted,
the
tests
refer
to
the
t-test
and
the
Mann–Whitney
U.
finding
highlights
the
importance
of
strategic
uncertainty,
although
gender
differences
are
not
crucial
in
explaining
the
depositors’
behavior.
5.2.
Depositors’
behavior
and
observability
of
actions
In
our
context,
the
position
in
the
line
and
the
information
struc-
ture
determine
what
depositors
observe
and
whether
subsequent
depositors
can
observe
their
actions.
To
disentangle
whether
there
are
gender
differences
in
the
effects
of
these
variables
on
the
with-
drawal
decision,
we
analyze
the
depositors’
behavior
in
more
detail
by
performing
an
econometric
analysis.
We
estimate
a
logit
model
on
the
probability
of
withdrawal
in
each
position.
The
set
of
inde-
pendent
variables
includes
the
subject’s
gender,
the
observation
possibilities
and
the
interaction
effects.
Our
regressions
also
control
for
risk
aversion,
which
is
measured
using
the
investment
decision
in
Gneezy
and
Potters
(1997).18 Because
the
subjects
are
asked
to
make
decisions
during
15
rounds,
we
also
control
for
the
history
of
decisions,
as
in
the
study
by
Garratt
and
Keister
(2009).
In
par-
ticular,
the
variable
History
measures
the
proportion
of
previous
rounds
in
which
the
subject
witnessed
a
bank
run.
Next,
we
ana-
lyze
the
behavior
of
each
depositor
separately
by
focusing
on
the
observability
of
actions
in
Sections
5.2.1,
5.2.2
and
5.2.3.
We
discuss
the
predictive
power
of
the
history
of
decisions
and
the
degree
of
risk
aversion
in
Section
5.2.4.
5.2.1.
Depositor
1’s
behavior
When
depositor
1
chooses
whether
to
wait
or
withdraw,
(s)he
has
no
information
about
what
will
occur
in
the
bank.
However,
depositor
1
does
know
whether
subsequent
depositors
will
observe
her/his
decision,
which
can
affect
depositor
1’s
decision
because
a
patient
depositor
would
like
to
wait
to
induce
the
other
patient
depositor
to
follow
suit.
Our
data
are
consistent
with
such
a
behav-
ioral
pattern.
When
they
are
not
being
observed,
men
withdraw
57%
of
the
time,
and
women
withdraw
64%
of
the
time.
These
withdrawal
rates
decrease
to
32%
and
24%,
respectively,
when
one
other
depositor
will
observe
the
action
of
the
first
depositor,
and
to
11%
and
16%,
respectively,
when
depositor
1
knows
that
the
two
subsequent
depositors
will
observe
her/his
action.
In
the
logit
specification,
the
independent
variables
are
the
sub-
ject’s
gender
(W
is
equal
to
0
for
men
and
1
for
women),
the
number
of
subsequent
followers
observing
the
depositors
(F1
equals
1
if
depositor
1
is
observed
by
only
one
depositor,
and
F2
equals
1
if
depositor
1
is
observed
by
two
depositors)
and
the
interac-
tion
effects,
controlling
for
risk
aversion