Gender differences in risk taking: A meta-analysis

Abstract

The authors conducted a meta-analysis of 150 studies in which the risk-taking tendencies of male and female participants were compared. Studies were coded with respect to type of task (e.g., self-reported behaviors vs. observed behaviors), task content (e.g., smoking vs. sex), and 5 age levels. Results showed that the average effects for 14 out of 16 types of risk taking were significantly larger than 0 (indicating greater risk taking in male participants) and that nearly half of the effects were greater than .20. However, certain topics (e.g., intellectual risk taking and physical skills) produced larger gender differences than others (e.g., smoking). In addition, the authors found that (a) there were significant shifts in the size of the gender gap between successive age levels, and (b) the gender gap seems to be growing smaller over time. The discussion focuses on the meaning of the results for theories of risk taking and the need for additional studies to clarify age trends.

Full text

Psychological
Bulletin
1999,
Vol.
125,
No. 3,
367-383
Copyright
1999
by the
American Psychological Association, Inc.
0033-2909/99/S3.00
Gender Differences
in
Risk Taking:
A
Meta-Analysis
James
P.
Byrnes, David
C.
Miller,
and
William
D.
Schafer
University
of
Maryland
The
authors conducted
a
meta-analysis
of 150
studies
in
which
the
risk-taking tendencies
of
male
and
female participants were compared. Studies were coded with respect
to
type
of
task (e.g., self-reported
behaviors
vs.
observed behaviors), task content (e.g., smoking
vs.
sex),
and 5 age
levels. Results showed
that
the
average
effects
for 14 out of 16
types
of risk
taking were
significantly
larger than
0
(indicating
greater
risk
taking
in
male
participants)
and
that nearly half
of the
effects
were
greater
than .20. However,
certain topics (e.g., intellectual risk taking
and
physical skills) produced larger gender differences than
others
(e.g., smoking).
In
addition,
the
authors
found
that
(a)
there were
significant
shifts
in the
size
of
the
gender
gap
between successive
age
levels,
and (b) the
gender
gap
seems
to be
growing smaller over
time.
The
discussion focuses
on the
meaning
of the
results
for
theories
of risk
taking
and the
need
for
additional
studies
to
clarify
age
trends.
Risk
taking
is an
important
form
of
human behavior that
has
been
the
subject
of
numerous investigations, scholarly analyses,
and
policy debates (Byrnes, 1998; Slovic, Lichtenstein,
&
Fischhoff,
1988).
To a
large extent, most researchers point
to the
association between risky behaviors (e.g., unprotected sex)
and
serious health problems
(e.g.,
sexually transmitted diseases) when
they
provide
reasons
for
studying
the
former
(e.g.,
DiClemente,
Hansen,
&
Ponton, 1995). However, others have also argued that
risk
taking
should
be
studied because
of its
relevance
to
three
important
issues
in the
field
of
psychology:
the
adaptiveness
of
human
behavior (Byrnes, 1998; Payne,
Bettman,
&
Johnson,
1993),
the
rationality
of
human thought (Baron, 1994),
and the
relative importance
of
genes versus
the
environment
in
determin-
ing
the
phenotypic expression
of
traits (Wilson
&
Daly, 1985;
Zuckerman,
1991).
In
essence,
then,
researchers
have examined
risk
taking
for a
variety
of
reasons. Correspondingly,
the
literature
on
risk
taking
is
both vast
and
diverse.
In
the
present article,
we
review that portion
of the
literature that
is
concerned with gender differences. From
a
scientific standpoint,
gender differences
are of
interest because they
can
often
precipi-
tate important theoretical advances
in a
particular area
of
inquiry
(Halpern,
1992).
For
example,
the
existence
of
gender differences
on
the
Scholastic Aptitude
Test
(SAT) prompted many
researchers
to
search
for
possible causes
of
this difference.
One of the
most
important outcomes
of
this line
of
work
was the
finding
that much
of
the
variance
in
SAT
scores
can be
explained
by
course work
and
math
knowledge (e.g., Byrnes
&
Takahira,
1993).
We
hoped
to
precipitate similar advances
in the field of risk
taking.
Our
review
is
organized
as
follows.
In the first
section
of
this
article,
we
discuss issues related
to the
definition
and
assessment
of
risk
taking
to
provide
an
interpretive context
for the
rest
of the
James
P.
Byrnes
and
David
C.
Miller, Department
of
Human Develop-
ment, University
of
Maryland; William
D.
Schafer, Department
of
Educa-
tional Measurement
and
Statistics, University
of
Maryland.
Correspondence concerning this
article
should
be
addressed
to
James
P.
Byrnes,
Department
of
Human Development, University
of
Maryland,
College Park, Maryland
20742.
Electronic mail
may be
sent
to
JB119@umail.umd.edu.
article.
In the
second section,
we
consider
the
relevance
of
gender
differences
for
various theories
of risk
taking.
In the
third
and
fourth
sections,
we
describe
our
meta-analytic
methodology
and
the
results
of
this analysis.
In the final
section,
we
interpret
the
findings
and
draw conclusions.
The
Nature
of
Risk Taking: Definition
and
Assessment
Issues
Researchers clearly
differ
in the
definitions they provide
for risk
taking,
but
most refer
to
constructs such
as
goals, values, options,
and
outcomes (e.g., Byrnes, 1998; Furby
&
Beyth-Marom,
1992;
Lopes, 1987; Slovic, Lichtenstein,
&
Fischhoff, 1988). Goals
and
values
determine
the
kinds
of
outcomes
that
are
pursued
by an
individual
(e.g., good grades
in
school
vs.
being popular
with
friends)
and
also determine
the
kinds
of
options that
are
considered
(e.g.,
studying
vs.
socializing).
The act of
implementing
a
goal-
directed option
qualifies
as an
instance
of risk
taking whenever
two
things
are
true:
(a) the
behavior
in
question could lead
to
more
than
one
outcome
and (b)
some
of
these outcomes
are
undesirable
or
even dangerous (Furby
&
Beyth-Marom, 1992).
In
essence,
then,
risk
taking involves
the
implementation
of
options that could
lead
to
negative consequences.
This somewhat standard definition implies that
a
wide range
of
behaviors would
qualify
as
instances
of risk
taking (e.g., telling
a
joke,
raising
one's
hand
in
class,
smoking, having unprotected sex,
etc.).
To
some researchers (including
the
present authors),
the
breadth
of
this definition
is
desirable because
it is
consistent with
their
belief
in the
pervasiveness
of risk
taking
in
daily life.
In
addition,
it
gives them
the
latitude
to
study
risk
taking
in
younger
age
groups (e.g.,
one
cannot
ask a
minor
to
consume alcohol).
To
others, however,
the
definition
may
seem
to be too
broad because
it
lumps fairly innocuous behaviors
(e.g.,
spinning
a
roulette wheel
to win
candy) together with rather dangerous ones (e.g., drunk
driving). From
an
assessment standpoint, this split among
re-
searchers
is
problematic
because
it
raises
questions
about
the
validity
of
certain measures
of risk
taking.
One way to
reconcile
the two
approaches
is to
suggest that
the
category
of risky
behaviors
is not an
equivalence class (Byrnes,
367

368
BYRNES,
MILLER,
AND
SCHAFER
1998).
In an
equivalence class,
any two
members
of the
class
are
equally
good examples
of the
category
(e.g.,
the
numbers
3 and
217
in the
case
of
numbers). Clearly, actions that
are
likely
to
lead
to
distressing outcomes (e.g., reckless driving, unprotected sex)
are
generally
thought
to be
more representative
of the
category
of
risky behaviors than actions that
are
significantly less
likely
to
produce
such
outcomes
(e.g.,
driving within
the
speed
limit,
pro-
tected
sex)
or
actions
in
which seemingly trivial outcomes
are at
stake (e.g., voicing
one's
opinion,
not
carrying
an
umbrella
on a
cloudy
day). Thus,
one
could
say
that
researchers
who
prefer
a
more restrictive
definition
would admit only
the
prototypical cases
into
the
category
of risky
behaviors.
In
contrast, those
who
prefer
the
less restrictive
definition
would admit both prototypical
and
less prototypical cases into
the
category.
A
second issue
that
would
affect
a
researcher's judgment
of
validity
is the
distinction between
an
individual's subjective per-
ception
of risk and the
perceptions
of the
larger community (Furby
&
Beyth-Marom,
1992). Consider
the
case
of an
unmarried ado-
lescent
girl
who
values pregnancy. Although many observers
would
argue that
she has
taken
a risk
when
she
engages
in
unprotected sex,
the
teenager might argue that
she has
not. Con-
versely, consider
the
case
of an
8-year-old girl
who
does
not
want
to
lose
a
small amount
of
candy that
she has won on
early trials
of
a
risk-taking
game. Losing candy
may
seem trivial
to an
adult,
but
it
may be
very undesirable
to the
child.
As a
result,
the
task would
be
subjectively
risky to the
child
but not
very
risky to
adult
onlookers. Some researchers have argued that subjective percep-
tions
of risk
constitute
the
minimum standard
to be met in any
study
of risk
taking. Others, however, seem
to
require higher
and
more objective standards (Slovic,
1964).
Either way,
the
subjec-
tivity
criterion implies that people take
risks
only when they
are
aware
of the
fact
that they
are
taking
risks.
However, there
are
many
behaviors
that
seem
to
qualify
as
prototypical
instances
of
risk
taking
that
are
performed
out of
naivete
(e.g.,
unprotected
sex
in
uninformed
teens;
a
young child playing
in the
street; etc.).
Thus,
we
agree
with
Furby
and
Beyth-Marom's
(1992)
suggestion
that
behaviors
can be
appropriately
define'd
as risky
even when
the
person performing these actions
is
unaware
of
possible negative
consequences.
A
third factor that
affects
judgments
of
validity
is the
relation
between
a
person's skill level
and risk
taking.
By
definition,
a
highly
skilled individual
fails
less
often
than
an
unskilled peer
on
tasks relevant
to
that skill.
As a
result, skill-related actions might
only
be risky for the
latter. This analysis implies, therefore, that
a
researcher could
not use the
tendency
to
engage
in
skilled behav-
iors
as the
only evidence
of risk
taking
in a
particular study (Miller
&
Byrnes, 1997).
As we
indicate later, however, many researchers
have
done
so.
A
fourth
and final
issue pertains
to the
contextualization
of
behaviors.
Clearly,
there
are
ways
to
perform
an
action
and
situ-
ations
in
which
it is
performed that make
it
more
or
less
risky. For
example, drinking
a
small amount
of
alcohol
is
less
risky
than
drinking
a
large amount
of
alcohol. Similarly, having unprotected
sex
with
a
stranger
is riskier
than having unprotected
sex
with
a
spouse
who has
been tested
for
sexually transmitted diseases.
Hence, responding "yes"
to
decontextualized
questions regarding
alcohol consumption
and
unprotected
sex may or may not
indicate
risk
taking.
Collectively, these
four
issues illustrate
the
problems associated
with
assessing someone's
risk
taking that have
to be
considered
when
the
results
of a
study
are
interpreted. Note that these prob-
lems should
not
preclude
the
possibility
of
conducting
a
meaning-
ful
review because
the
standard
of
unquestioned validity
is
impos-
sible
to be
fully
met. That
is, all
measures
of
competence have
been questioned
at one
time
or
another
(e.g.,
IQ
tests, achievement
tests, etc.). Questions
of
validity
do,
however,
affect
the
degree
of
concern generated
by a
particular
set of findings. For
obvious
reasons, gender differences
on a
valid
and
widely accepted mea-
sure would tend
to
generate more concern than gender differences
on
an
ambiguous
or
controversial measure.
We
return
to the
issues
of
measurement
and
validity when
we
describe
our
inclusion
criteria
and
discuss
the
implications
of our
results.
For
now,
it is
important
to
point
out an
important implication
of
the
foregoing
analysis:
Risk
taking
can
either
be
adaptive
or
maladaptive.
It is
maladaptive whenever
the
benefits
of
some
activity
are far
less likely
to
occur than
the
potential hazards.
It is
adaptive whenever
the
converse
is
true.
In
other words, people
do
not
successfully adapt
to
their surroundings
by
avoiding
all the
risks
they
face.
In
fact,
it
would
be
impossible
to do so.
Instead,
they
successfully adapt
by
systematically pursuing certain
risks
while avoiding others
(Baumrind,
1991; Byrnes,
1998).
The
study
of
gender differences, then,
has the
potential
for
demonstrating
greater
or
lesser
degrees
of
environmental adaptation
in
women
versus men.
Theories
and
Expectations Regarding Gender Differences
in
Risk Taking
Theories
are
often
judged
to be
adequate
to the
extent that they
can
explain similarities
or
differences
in
performance among var-
ious
groups
(e.g.,
male
vs.
female participants, younger
vs.
older
students, experimental
vs.
control groups).
All
things being equal,
a
theory that
can
explain gender
differences
is
more adequate than
a
theory that cannot.
Our
reading
of the risk-taking
literature
reveals that researchers have
not
been particularly interested
in
explaining
or
uncovering gender
differences
using
the
most widely
cited theoretical
models
of risk
taking. Instead, they have tended
to
examine gender differences
in an
ancillary manner (presumably
because
it
would prove interesting
to the
reader). Nevertheless,
it
is
useful
to
briefly consider
the
types
of
results that would
be
more
or
less consistent with particular types
of
theoretical approaches.
At
a
general level, theories
of risk
taking
fall
into
one of
three
categories
(Lopes,
1987).
The first
category consists
of
theories
that
are
equipped
to
explain
the
differences between
people
who
regularly
take
risks and
people
who
regularly avoid
risks. Two
examples
of
such theories
are
Zuckerman's
(1991)
account
of the
sensation-seeking personality
and the
"Risk
as
Value" hypothesis
as
described
by
Kelling, Zirkes,
and
Myerowitz (1976). Here,
a
single
factor such
as (a) a
naturally
lower
level
of
arousal
in men
or
(b) a
socially instilled belief that
risk
taking
is a
highly valued
masculine tendency motivates high levels
of risk
taking across
contexts
in
men.
In
their simplest form, such theories predict that
the
size
and
direction
of
gender differences would
not
vary
by
context (i.e.,
men
would always take more
risks
than
women
and
the gap
would remain relatively
the
same across contexts). More-
over,
one
need
not be
concerned about
the
type
of
task used
because most
risk-taking
tasks would probably
tap
into this ten-

RISK
TAKING
369
dency.
In
more elaborated forms
of
these theories, however,
a
greater
degree
of
context specificity might
be
predicted (e.g.,
see
Arnett's
model below).
The
second category consists
of
theories that
are
equipped
to
explain
the
differences
between
situations
that
promote
risk
taking
(in
most people)
and
situations that promote
risk
aversion.
An
example would
be
Kahneman
and
Tversky's (1979) prospect the-
ory. Prospect theory
was
designed,
in
part,
to
account
for the
fact
that
most people seem
to
prefer
a risky
option over
a
sure thing
when
the
choices
are
framed
in a
positive
way
(e.g.,
the
number
of
people
who
would
be
saved
by a
medication),
but
they
shift
their
preferences when
the
same
choices
are
framed
in a
negative
way
(e.g.,
the
number
who
would
not be
saved).
There
are,
of
course,
participants
in
each study
of
framing
who
fail
to
demonstrate this
shift,
but
prospect theory
was not
designed
to
account
for
such
individual
differences (Lopes, 1987).
As
such, situation-based
theories would have
to
predict
a
constant
effect
size
of d = 0
across contexts (i.e.,
no
gender differences).
The
third category consists
of
theories that
are
equipped
to
explain differences among people
and
situations that promote
risk
taking.
In
other words, these models could explain
why
only
certain people take
risks in
certain situations.
To
illustrate, con-
sider multifactor models that include expectations
and
values
in
their
formulation. According
to
these models, people take
risks in
a
particular context because they
(a)
believe they will
be
successful
and
(b)
value success
in
that context (Atkinson, 1983; Byrnes,
1998;
Irwin
&
Millstein,
1991;
Wigfield
&
Eccles,
1992).
As
people move
from
one
context
to
another, however, they generally
hold different expectations
and
values.
As a
result, these models
suggest that gender differences would vary
by
context
and
that
some contexts would promote greater
risk
taking
on the
part
of
women.
For
example,
if
there
was
reason
to
believe that women
might
feel
more confident
in a
particular situation than men,
and it
mattered more
to
women
to be
successful
in
that situation (e.g.,
volunteering
to
coordinate
a
crucial fundraiser
for
their children's
school), expectancy-value models would support
an
expectation
of
greater
risk
taking
on the
part
of
women.
Other multifactor models also support
the
idea
of
context spec-
ificity,
but for
other reasons.
For
example, Arnett's (1992) theory
of
broad
and
narrow socialization suggests that
the
level
of risk
taking
manifested
by an
individual depends
on two
factors:
(a)
endogenous tendencies such
as
sensation seeking
and (b) the
restrictions placed
on risk
taking
by the
individual's culture (e.g.,
laws, norms, parenting practices, etc.). Whereas cultural restric-
tions
dampen
a
sensation
seeker's
tendency
to
take
risks,
these
restrictions
do not
entirely eliminate
the
tendency.
As
such,
Ar-
nett's
model
leads
to the
prediction that
men
would take more
risks
than
women
in
most cultures (because sensation seeking
is
found
more
often
in men
than women).
However,
the
size
of the
gender
gap
would vary
as a
function
of a
culture's
restrictiveness
and the
norms
for
appropriate gender
role
behaviors.
Our
reading
of Ar-
nett's
model
is
that
it
would
not
necessarily predict
negative
effect
sizes
(i.e.,
more
risk
taking
in
women, using
the
conventional
approach
of
subtracting
the
female mean
from
the
male mean),
but
given
the
fact
that
some
sensation seekers
are
women, negative
effect
sizes would
not be out of the
question.
In
the
same way, Wilson
and
Daly's
(1985)
sociobiological
model suggests that gender differences would
not be
found
for all
contexts,
but
that
it
would
be men who
take more
risks
when
gender differences
do
occur.
These
authors have argued that
risk
taking
is an
"attribute
of the
masculine psychology"
(p. 66)
that
evolved
in
response
to the
competitive demands
of
primate soci-
eties.
According
to
this view, competition forces dominant
indi-
viduals
to
engage
in risk
taking
to
gain their
positions
of
power.
The
greater
the
spread
in
rewards between winners
and
losers,
the
greater
the
incentive
to
take
risks.
This account suggests that
men
would
only
be
more likely
to
take
risks
than women when
a
context involves both competition
and a
large spread
in
rewards
between winners
and
losers.
For all
other contexts,
the gap
would
presumably
be
smaller (though Wilson
and
Daly
do not
explicitly
make this claim). Moreover, their account provides little room
for
the
possibility
of
greater
risk
taking
on the
part
of
women.
One
could,
however, construct
a
different
sociobiological model
to
explain
risky
behaviors that
are
clearly more common
in
women
than
in men
(e.g., binge eating).
So, the
existence
of
negative
effect
sizes
is
more
of a
problem
for
Wilson
and
Daly's
model
than
for
sociobiological models generally.
In
sum, then, there
are
three primary patterns
of
gender
differ-
ences that would
be
more
or
less consistent
with
the
claims
of
particular
theories
of risk
taking.
For
Pattern
1,
effect
sizes
would
show
a
constant gender difference across contexts favoring men.
Pattern
2
would consist
of
varying gender differences across
contexts,
but the
distribution
of
effect
sizes would include only
zeros
and
positive values (i.e.,
it
would
be men who
take more
risks
when
gender differences occur). Pattern
3
would
be a
distri-
bution
of
effect
sizes
that
includes
a
full
range
of
values
(i.e.,
negatives, positives,
and
zeros).
In
addition
to
considering
the
issue
of
context specificity,
we
also considered whether
the
size
of the
gender
gap
would change
with
age.
In so
doing,
we
hoped
to
narrow
the
field
of
plausible
theories
even
further
(beyond that suggested
by the
findings
for
context specificity). Note that relatively
few
theories
of risk
taking
could
explain
a
pattern
in
which
the
gender
gap
increases
mono-
tonically
from
d = 0 in
young children
to d = .40 in
young adults.
Similarly, relatively
few
theories could explain
a
pattern
in
which
the
gender
gap
increases between childhood
and
adolescence (e.g.,
d
= 0 to d =
.20),
and
then decreases between adolescence
and
adulthood (e.g.,
d = .20 to d = 0). In the
final
section
of
this
article,
we
describe
the
extent
to
which existing theories
of risk
taking
are
consistent
with
the age
trends that
we
reveal.
For
now,
we
move
on to a
consideration
of our
meta-analytic
techniques.
Method
Overview
A
total
of 150
studies were retrieved
in five
steps.
The first
step involved
an
extensive
computerized
search
of the
PsycLIT
and
PsycINFO
databases
to
find
empirical studies
in
which
the
researchers examined gender
differ-
ences
in risk
taking. Within this search,
we
crossed
the
keywords risk
and
risk
taking with
the
keywords
gender
differences
and sex
differences
using
the
Boolean
operator AND. Such
an
approach retrieved
all
articles pub-
lished
between
1967
and
1994
that contained both
of the
crossed
keywords
somewhere
in the
title,
abstract,
or
subject identifier
list.
However, only
49
of
these
articles
involved
a
direct comparison
of men and
women
(or
boys
and
girls)
on
some
risk-taking
measure.
The
rest were
of
four
types:
(a)
policy-oriented commentaries,
(b)
studies
in
which
men and
women were
analyzed
separately
(no
means were reported),
(c)
studies
of risk
percep-
tions,
and (d)
articles
on the risk
factors
for
various
diseases
(e.g.,
heart

370
BYRNES,
MILLER,
AND
SCHAFER
disease).
Note that
we
included
all
empirical
articles
that were character-
ized
by
their authors
as
involving
risk
taking (using either
the
broad
or
narrow
definitions that
we
discussed
earlier).
We did so for two
reasons.
First,
we
favor
the
broad definition
because
we
believe that
risk
taking
extends
beyond
the
class
of
prototypical
behaviors.
Second, including both
types
of risk
taking allowed
us to
provide
useful
information
to
researchers
in
both camps (e.g., average
effect
sizes
for
both prototypical
and
less
prototypical types
of risk
taking).
During
the
second step,
we
conducted another search
on an
updated
and
greatly expanded version
of
PsycINFO
that became available
to us
shortly
after
the
first
step
had
been completed.
The new
version used
a
relevance-
based approach that
can be
used instead
of a
Boolean approach.
In the
relevance-based approach, articles that contained both keywords (e.g.,
risk
taking
and
gender differences) were given higher ranks than articles that
contained
only
one of
these terms. However, articles
with
only
one
term
(e.g.,
just
risk
taking)
were
not
excluded
as
they would
be in a
Boolean
approach.
Two new
searches were conducted using
the relevance
method:
one in
which
the
terms
risk
taking
and
gender
differences
were
crossed,
and
another
in
which
the
terms
risk
taking
and sex
differences
were
crossed.
Each
of
these
searches generated
a
rank-ordered list
of 250
articles pub-
lished between 1967
and
1997 that contained
the
original
49
articles plus
many
others (note:
the
system limited searches
to 250 of the
highest ranked
articles).
The
titles
and
abstracts
of all 500 of
these articles were scanned
to
limit
the
selection
to
just those that were empirical studies. This
process
yielded
66
articles that mentioned gender comparisons
in
their abstracts
and
another
125
studies
on risk
taking that failed
to
indicate whether
gender comparisons were made.
All 191 of
these articles were retrieved
and
read
in the
event that
the
latter examined gender differences
but
failed
to
mention this
fact
in
their abstracts.
It
turned out, however, that only
the
articles that mentioned gender differences
in the
abstract actually made
gender comparisons.
In
the
third step,
we
conducted
a
search using
the
names
of the
authors
who
were
found
during
the
first
two
steps. This
was
done because
we
found
that
PsycINFO occasionally
did not
always
place
similar articles
in the
same category.
In
addition,
we
searched
the
MEDLINE
system
to
discover
epidemiological
and
other studies
not
indexed
by
PsycINFO.
In
each
case,
the
searches included articles that
had
dates ranging
from
1967
to
1997.
The net
result
of the
second
and
third steps
was
that
78
additional articles
were
added
to the 49
that were identified
in the first
step.
During
the
fourth
step,
we
entered
the
terms associated with specific
types
of risky
choices such
as
smoking, driving,
and
framing
effects.
In the
case
of
framing,
we
located
103
studies conducted
after
1981 (when
Kahneman
and
Tversky originally identified
the
phenomenon). Although
none
of
these studies mentioned gender differences
in
their
abstracts,
we
retrieved them anyway
and
found
that
8 had
conducted analyses
of
gender
differences.
The
effects
from
the
latter
8
studies were added
to the
data-
base.
Our
search
of
other specific types
of risk
taking failed
to
locate
additional
studies that involved gender comparisons.
In
the fifth and final
step,
we
conducted
a
computerized search
of
Dissertation
Abstracts. This system, which only allows
a
Boolean
ap-
proach
of
dissertations conducted
after
1979, generated
a
list
of 23
disser-
tations
that
examined
gender
differences
in risk
taking.
All 23
were
requested
through
the
interlibrary
loan services
of our
university,
but
only
15
were sent
by
their home institutions. Recall that
a
Boolean search
only
retrieves studies
in
which
all of the
requested terms
are
present
in the
titles or
abstracts. Given
our
primary interest
in
issues
of
context specificity
and
age
trends,
we did not
request additional dissertations
to
check
for the
possibility
that some authors analyzed gender differences
but
failed
to
indicate
this
fact
in
their abstracts.
We
leave that task
to the
interested
reader.
For
now,
it is
sufficient
to
note that
the
means that
we
report
for
dissertations could
be
somewhat
biased
(i.e.,
probably
too
high given
our
experience
with
journal articles
in
which
the
nonreported
comparisons
tended
to
yield very small
effects).
In
total, then,
we
ended
up
analyzing data
from
150
(40%)
of the 374
publications that
we retrieved and read. The remaining 224
(60%) publi-
cations investigated
risk
taking
but did not
provide data
on
gender
differ-
ences. There were
135
published studies
and 15
dissertations
in the final
database. Collectively, these studies comprised
a
total
N of
over
100,000
participants.
In
most studies, more than
one
gender comparison
was
made.
As a
result,
we
were able
to
compute
a
total
of 322
effects
(M
=
2.15
effects
per
study).
When
a
large number
of
comparisons were made
in a
given study
and
all
comparisons involved
the
same content
(e.g.,
gender differences
in
smoking rates
for 22
countries),
a
single average
effect
size
was
computed
for
that study.
Effects
were
not
averaged
in a
study when they pertained
to
different
contents (e.g., smoking rates
and
drinking
rates).
For
each gender
comparison,
we
primarily computed Cohen's
d in one of two
ways:
(a)
directly
from
the
means
and
standard deviations
if
they were provided
or
(b)
indirectly,
by
converting parametric statistics
(e.g.,
F or t) or
chi-square
statistics
using standard conversion formulas
(e.g.,
Rosenthal,
1994). Once
computed, each
d was
corrected
for
bias
and
weighted according
to the
inverse
of its
variance using
the
approaches described
by
Hedges
(1994)
and
Hedges
and
OUan
(1985).
In 9% of the
cases
(i.e.,
30
effects),
d had
to
be
estimated
from
a
probability level (e.g.,
p <
.05)
or
assigned
to a
value
of
zero
if the
authors merely reported that
no
significant difference
was
observed.
Coding
Studies
According
to
Task
and
Content
After
the
studies were retrieved, they were coded primarily according
to
the
type
of
task used
by the
researchers. Three types
of
tasks were
identified.
In the first
type
(hypothetical
choice), participants were asked
to
choose between
two
imaginary options (e.g.,
two fictitious
gambles)
or
choose
a
level
of risk
that they would tolerate
in a
hypothetical situation
(e.g., take
a risk if
they were
70%
sure that things would work out).
Participants were
not
asked
in
these studies
if
they ever engaged
in the
behaviors
described.
In
addition, they
did not
have
to
experience
the
consequences
of
their
choices.
Hypothetical choice tasks
(coded
1)
were
used
in 23% of the
published studies
(n =
31)
and 53% of the
dissertations
(n
= 8).
In the
second type
(self-reported
behavior),
participants reported
how
often
they engaged
in
various
risky
behaviors
(e.g.,
used drugs
or had
unprotected sex).
A
sample item would
be, "In the
last
12
months, have
you
had
sex
without
a
condom?" Hence, whereas hypothetical choice tasks
involved
questions
of the
form
"How
would
you
behave
in
this situation?"
self-report tasks involved questions
of the
form
"How have
you
behaved
in
this
situation?" Self-report tasks
(coded
2)
were used
in 45% of the
published
studies
(n = 61) and 33% of the
dissertations
(n = 5).
In the
third type
(observed
behavior),
participants were observed
by
researchers
as
they engaged
in
various activities that were judged
by the
researchers
to
have some
degree
of risk
(e.g., making
a
left
turn
in
front
of
oncoming
traffic).
Observed behavior tasks (coded
3)
were used
in 33% of
the
published studies
(n = 45) and 20% of the
dissertations
(n = 3).
After
studies were
coded
with
respect
to the
type
of
task used, they were
next
coded according
to the
content
of the
task.
We
created
a
content
category within
a
particular task category
if at
least
10%
of the
effect
sizes
pertained
to
that content. Contents that were composed
of
less than
10%
of
the
effect
sizes
for a
given task were placed into
an
other category.
For
hypothetical
choice studies, three content categories emerged: choice
di-
lemma
tasks
(Kogan
&
Wallach,
1964),
framing
tasks (Tversky
&
Kah-
neman,
1981),
and
other. Choice dilemma tasks (coded
1)
consist
of
presenting participants
with
12
scenarios (e.g.,
a man who is
thinking about
a risky
medical procedure;
a
woman thinking about
her
career options;
etc.).
After
being given
a
certain amount
of
information about several
options, participants
are
asked
to
state
the
minimum level
of
uncertainty
that
they would accept before
the
main character should choose
the riskiest
option.
An
example item
(from
Kogan
&
Wallach, 1964,
p.
257)
is the
following:

RISK
TAKING
371
Mr.
B., a
45-year-old
accountant,
has
recently been informed
by his
physician
that
he has
developed
a
severe heart ailment.
The
disease
would
be
sufficiently
serious
to
force
Mr. B. to
change many
of his
strongest
life
habits—reducing
his
work load, drastically changing
his
diet, giving
up his
favorite leisure-time pursuits.
The
physician sug-
gests that
a
delicate medical operation could
be
attempted which,
if
successful,
would completely relieve
the
heart condition.
But its
success could
not be
assured,
and in
fact,
the
operation might prove
fatal.
Imagine that
you are
advising
Mr. B.
Listed below
are
several
probabilities
or
odds that
the
operation will prove successful. Please
check
the
lowest probability that
you
would
consider
acceptable
for
the
operation
to be
performed:
Place
a
check here
if you
think
Mr. B
should
not
have
the
operation
no
matter what
the
probability.
The
chances
are 9 in 10
that
the
operation will
be a
success.
The
chances
are 7 in 10
that
the
operation will
be a
success.
The
chances
are 5 in 10
that
the
operation will
be a
success.
The
chances
are 3 in 10
that
the
operation will
be a
success.
The
chances
are 1 in 10
that
the
operation will
be a
success.
On
this task,
any
respondent
who
consistently selects
low
probabilities
across
the 12
scenarios
(e.g.,
3 in 10) is
said
to
have
a
preference
for risk
taking.
In
contrast,
the
tendency
to
choose
the
highest odds
of
success
(e.g.,
9 in 10)
across
the
scenarios
is
taken
to
indicate
a
high level
of
conservatism.
In a
sense, then,
the
12-item
questionnaire that contains these
scenarios (i.e.,
the
Choice Dilemma Questionnaire
or
CDQ)
is
comparable
to
personality questionnaires
in its
assumption that
risk
taking
is a
trait-like
disposition that would
be
expressed across many
contexts.
The CDQ was
used
in 18
(46%)
of the
studies that used
a
hypothetical choice task.
On
framing tasks (coded
2),
participants
are
also presented with
a
hypothetical scenario such
as the
following
from
Tversky
and
Kahneman
(1981):
Imagine that
the
U.S.
is
preparing
for the
outbreak
of an
unusual
Asian disease, which
is
expected
to
kill
600
people.
Two
alternative
programs
to
combat
the
disease
have been proposed. Assume that
the
exact scientific estimates
of the
consequences
of the
program
are as
follows:
If
Program
A is
adopted,
200
people will
be
saved.
If
Program
B is
adopted, there
is a 1/3
probability that
600
will
be
saved
and
a 2/3
probability that
no
people
will
be
saved. Which
of the two
programs
do you
favor?
(p.
453)
Half
of the
participants
received
the
wording above,
and
half
received
a
version
in
which
the
first three lines were
the
same
but the two
options
were stated
as
follows:
"If
Program
C is
adopted,
400
people
will die.
If
Program
D is
adopted there
is a 1/3
probability that nobody would die,
and
a
2/3
probability that
600
people will die." Note that Option
A is
identical
to
Option
C, and
Option
B is
identical
to
Option
D.
Within each
of the two
frames,
there
is a
"sure
thing"
(e.g., Program
A)
contrasted with
a risky
option (e.g., Program
B). The
primary indication
of risk
taking
on
framing
tasks
is the
selection
of the risky
option.
The
tendency
to
choose
the
sure
thing
in the
first
frame
is so
pervasive that
it has
been taken
to
indicate
a
basic
aversion
to
risk taking
in
adults (Lopes, 1987). Framing tasks were
used
in 28% of the
hypothetical choice studies.
The
other category
(coded
3) for
hypothetical
choice
tasks
contained
the
remaining
26% of
measures that
did not fit
either
the
choice dilemma
category
or
framing category.
In all
cases,
participants were
presented
with
a
scenario
and
shown
two or
more ways
one
could behave
in
that scenario.
Some
of the
scenarios included
(a)
making friends
in a new
neighborhood
(Miller
&
Byrnes, 1997),
(b)
walking home
from
the
woods
(Morrongiello
&
Bradley,
1997),
(c)
solving
math
problems
(Foersterling,
1980),
(d)
donating
an
organ
to a
sick child
(Lampert
&
Yassour,
1992),
and (d)
planting
different
types
of
crops (Wilson, Daly, Gordon,
&
Pratt, 1996).
The riskiest
options always entailed
the
possibility
of
negative conse-
quences
(e.g.,
being rejected
in the
case
of
making
new
friends, etc.)
For
self-reported behavior studies,
five
content categories emerged:
drinking
and
drugs (coded
1),
driving (coded
2),
sexual activities (coded
3),
smoking
(coded
4), and
other (coded
5). As
noted earlier,
risk
taking
was
assessed
by
asking participants
if
they ever
(or
frequently)
engaged
in
such
behaviors. Example items
for the
first
four
categories included
(a)
"Have
you
ever smoked
marijuana?";
(b)
"Have
you
ever driven
20
miles
per
hour
above
the
speed
limit?";
(c)
"Have
you
ever
had sex
without
a
condom?";
and
(d) "Do you
currently smoke
cigarettes?"
The
other category included
items
that referred
to
behaviors such
as
vandalism, boating
in a
storm,
and
hitchhiking.
Too few
items focused
on the
latter topics
to
warrant
the
creation
of
additional categories. Some studies examined
a
single type
of
behavior
(e.g.,
use of
condoms), whereas others examined
a
number
of the
aforementioned
behaviors using multi-item instruments that
have
under-
gone reliability
and
validity assessments (e.g., Eysenck
&
Eysenck's, 1977,
Personality Questionnaire).
In
general, then,
the
self-reported-behavior
category mainly refers
to
behaviors that
are
either physically dangerous
or
illegal.
For
observed behavior studies, eight content categories emerged.
The
category
of
informed guessing (coded
1)
included tasks
in
which partici-
pants
could earn points
or
money
for
correct guesses
but
could also
lose
points
or
money
for
incorrect guesses
(e.g.,
standardized achievement tests
that
have penalties
for
incorrect
guesses).
The
category
of
physical activity
(coded
2)
included behaviors such
as the
following: climbing
a
steep
embankment, playing
in the
street,
trying
out
gymnastics
equipment
(e.g.,
a
balance beam),
and
taking
a ride on an
animal (e.g.,
a
donkey).
In
essence,
the
actions were
risky
because
of the
possibility
of
physical harm.
The
category
of
driving (coded
3)
included actions such
as
making
a
left
turn
in
front
of
oncoming
traffic,
gliding through
a
stop sign rather than
coming
to a
complete stop,
and
engaging
in
simulated driving tasks
on a
computer.
For the
real-life driving
behaviors,
the risks
include
damage
to
one's
vehicle, physical
injury,
and
traffic
tickets.
The
category
of
physical
skills (coded
4)
included such things
as
playing
shuffleboard
and
tossing
rings
onto
pegs.
In
most
cases,
options
differed
in
terms
of
their probability
of
success (i.e., high,
medium,
and
low)
and the
number
of
points that
could
be won or
lost (i.e., winning more points
for
success
but
losing more
points
for
failure
on
high-risk
options).
The
category
of
gambling
tasks
(coded
5) was
similar
to the
category
of
physical skills
in
terms
of the
varied
risk/reward
options. However, gambling tasks
differed
in the
sense
that
a
person's skill level
had no
bearing
on the
likelihood
of
success.
Examples included spinning
a
roulette wheel
and
pulling cards
from
a
deck.
The
category
of risky
experiments (coded
6)
involved
an
individual's
willingness
to
participate
in an
experiment
that
was
described
to
them
as
involving
the
chance
of
physical
or
psychological harm.
The
category
of
intellectual
risk
taking (coded
7)
involved tasks that required mathematical
or
spatial reasoning skills. Participants were presented with items
of
various levels
of
difficulty
and
asked
to
indicate their preferred level
of
choice. Unlike
the
tasks
in the
informed-guessing category, points were
not
subtracted
for
incorrect
answers
on the
intellectual
tasks.
Thus,
participants
were
mainly concerned about getting stuck
on
items
or
exposing their lack
of
skill when they
fail.
The final
category
of
other (coded
8)
included
the
following
behaviors that
did not fit the
other seven categories: entrepre-
neurial
activities
in a
simulated classroom economy, lying about
finding
someone
else's
money, cheating during
a
computerized game, playing
a
game
alone
instead
of
teaming
up, and
administering
an
electric
shock
to
a
confederate
to
increase
his
learning
rate.
Given
our
earlier arguments regarding
the
need
to
take into account
factors
such
as
context
and
skill levels when
risk-taking
competence
is
assessed,
we
refined
our
coding
of
tasks
in one
additional way.
In
some
studies, participants pursued courses
of
action that were clearly
not the
best
way
to
proceed
(i.e.,
they
had
better
options
available).
In
other
studies,
however,
the
measures
of risk
taking were more ambiguous.
In
other
words,
one
could
not
definitely
say
whether engaging
in the risky
action

372
BYRNES,
MILLER,
AND
SCHAFER
was
a
good idea
or
not.
Two
judges rated
the
ambiguity
of the
measures
used
in the 150
studies
and
agreed
93% of the
time. Differences were
resolved
by
discussion.
The
unambiguous measures (coded
1, n = 145
effects)
included such things
as
unprotected
sex
with
a
stranger, regular
consumption
of
alcohol
by a
minor, drug
use by a
minor, smoking,
speeding, drunk driving, criminal activities,
and
laboratory
tasks
that
involved
controls
for
skill level
and
probabilities
of
payoffs.
The
ambig-
uous
measures (coded
2, n =
111
effects) included
choice
dilemma tasks,
framing
tasks, making
a
left
turn
in
front
of
traffic,
trying alcohol once,
guessing
on an
objective examination, intellectual risk taking,
entreprenur-
ial
activities
in a
simulated
classroom
economy, lying about
finding
money,
and
laboratory tasks that
did not
control
for
skill level
or
proba-
bilities
of
payoff.
Additional
Codes
In
addition
to
coding studies
in
terms
of
task
and
content,
we
coded
them
according
to the age of the
participants using
the
following scheme:
(a)
below
age 9, (b)
ages
10-13,
(c)
ages 14-17 (high-school level),
(d)
ages
18-21
(college
level),
and (e)
ages
22 and
older (noncollege adult). Studies
were also coded with respect
to
their year
of
publication (i.e.,
1964-1980
vs.
1980-1997)
and
publication type
(e.g.,
dissertation
vs.
journal article).
Table
1
provides
the
entire corpus
of
150
studies that were used
in the
meta-analysis.
As can be
seen,
we
provide
all
effect
sizes
generated
by a
given
study
as
well
as the
corresponding
codes
for
task, content,
and age
(in
parentheses).
We
used
the
convention that positive
effect
sizes
corre-
sponded
to
greater risk taking
on the
part
of
males.
For
hypothetical choice
studies,
then,
a
positive
effect
size means that male participants were more
likely
than female participants
to
choose
the
risky option presented
in the
scenarios
or
accept
a
lower level
of
odds
across
scenarios.
For
self-report
studies,
it
means that male participants were more likely
to say
that they
had
engaged
in
behaviors such
as
reckless driving
or
unprotected sex.
For
observation studies,
it
means that experimenters
saw
male participants
engaging
in risky
behaviors more
often
than female participants.
Results
In
what follows,
we first
describe
the
distribution
of
effects
that
emerged
from
our
meta-analysis,
and
then
we
consider whether
the
size
of the
gender
gap
varied according
to
age, task, content, year
of
publication,
and
publication type.
Distribution
of
Effect
Sizes
For
descriptive purposes,
it is
useful
to
begin
by
subdividing
the
distribution
of
effect
sizes into successive intervals that capture
20%
of the
scores (i.e., quintiles).
The
interval corresponding
to
the first
quintile
was
found
to be
—1.23
to
—.09
(indicating greater
risk
taking
on the
part
of
female participants).
The
interval
for the
second quintile
was
-.08
to .07
(indicating essentially
no
differ-
ence).
The
intervals
for the
third, fourth,
and fifth
quintiles were
.08 to
.27,
.28 to
.49,
and .50 to
1.45, respectively (all indicating
greater
risk
taking
on the
part
of
male participants). Thus,
the
majority
(i.e., 60%)
of the
effects
support
the
idea
of
greater
risk
taking
on the
part
of
males.
In
fact,
nearly half (48%) were larger
than
.20
(the conventional
cutoff
point
for
small
effects).
However,
a
sizable minority (i.e., 40%) were either negative
or
close
to
zero.
Across
all 322
effects,
the
weighted mean
was
found
to be d =
.13, with
a 95%
confidence interval
of .12 to
.14. This
figure is
significantly
larger than zero,
^(l)
=
659.02,
p <
.00001.
To
check
for the
possibility that certain
effects
had an
undue influence
on
the
overall mean,
we
conducted three additional analyses. First,
we
excluded
the 12
effects
that were estimated
to be
zero
and
found
the
identical mean
of
d =
.13.
Hence,
it
cannot
be
said that
the
estimated
effects
artificially deflated
the
overall mean. Next,
we
generated
a
histogram
for the
original
322
scores
(including
the
estimated
effects)
to
consider
the
extent
to
which
the
distribution
deviated
from
the
characteristic shape
of a
normal distribution.
Because
we
searched
for
articles using general terms
(e.g.,
risk
taking)
and
specific terms
(e.g.,
framing),
we
were concerned that
there
may be too
many
articles
of one
type that yielded
too
many
values
in a
given region (causing
a
spike). Inspection
of the
histogram showed that
the
distribution
had the
characteristic nor-
mal
shape
in all
regions except
for the
region corresponding
to
values near zero.
We
trimmed
20
effects
from
that region
(i.e.,
the
12
effects
estimated
to be
zero
and 8
others
selected
at
random)
to
create
a
more proportionate array.
The
resultant distribution
generated
a
slightly higher mean
of d =
.14,
but
note that this
value
is
still within
the
original
95%
confidence interval
of .12 to
.14.
In the
third analysis,
we
examined
the
possibility that
the
mean
was
unduly influenced
by
extreme
scores.
Here,
we
excluded
all
effects
whose values
fell
two
standard deviations above
or
below
the
mean
(n = 30
effects).
The
resultant distribution
for the
remaining
292
effects
generated
a
mean
of d
=
.13,
so
outliers
did
not
appear
to be
biasing
the
mean
in a
particular direction. Overall,
then,
the
original mean
ofd=
.13
appears
to be a
reliable
estimate
of
the
population value.
Partitioning
Effects
Into Homogeneous Subgroups
Task,
content,
and age
effects.
Homogeneity analyses revealed
that
there
was
significant heterogeneity around
the
grand mean
of
d
=
.13,
^(313)
=
2457.83,
p <
.00001.
As a
result,
we
con-
ducted several
additional
analyses
to
partition
the
studies
into
more
homogeneous subgroups using
the
factors
of
task, content,
and
age.
In
the first
analysis,
we
found
that there
was
significant variation
associated with
the
task factor,
^(l)
=
10.85,
p
<
.01.
The
means
and
confidence intervals
for the
three types
of
tasks were
(a)
hypothetical
choice,
d =
.15
(.12
to
.18);
(b)
self-reported
behav-
ior,
d = .12
(.11
to
.13);
and (c)
observed behavior,
d = .19
(.16
to
.22). Using
the
method
of
nonoverlapping confidence intervals
(Schafer,
in
press),
we
found that these results suggest that
the
only
significant
contrast
is
between
the
means
for
self-reported behav-
ior
and
observed behavior.
All
three
of the
task means differed
significantly
from
zero (smallest
x2
=
65.79,
p <
.001).
Further analyses revealed significant heterogeneity around each
of
the
task means
as
well.
As a
result,
we
attempted
to
create more
homogeneous subgroups with each task using
the
factor
of
content.
The
means
and
confidence intervals
for
various contents
are
shown
in
Table
2. As can be
seen,
different
contents produced different
means
within each task.
For
hypothetical choice tasks,
for
exam-
ple,
two
levels
of
gender differences
are
evident.
The first
level
contains
the
relatively small gender differences
generated
by
choice dilemma
(d =
.07)
and
framing
tasks
(d =
.05).
The
means
for
these
two
tasks
did not
differ
significantly,
and
only
the
mean
for
choice dilemma tasks
was
significantly
greater than zero.
The
second level consists
of
tasks that
did not fit
choice dilemma
or
framing
categories.
The
mean
for
this other category
(d =
.35)
differed
significantly
from
the
means
for
choice dilemma
and

RISK
TAKING
373
framing
tasks. Recall that
the
scenarios
in the
other category asked
participants
how
they would behave
in
situations that involved
such
things
as
making friends
in a new
neighborhood,
walking
home
from
the
woods, solving math problems, donating
an
organ
to
a
sick child,
and
planting different types
of
crops.
For
self-reported behavior, three levels
of
gender differences
can be
observed.
The
first
level
is
composed
of the
mean
for
smoking
(d =
—.02),
which
was
significantly smaller than
all
other means.
The
next level consists
of the
means
for
drinking
and
drug
use (d =
.04)
and
sexual
activities
(d =
.07).
The
third
level
consists
of the
means
for
driving
(d =
.29)
and
other
(d =
.38).
The
latter
two
means were significantly larger than
the
means
for
sexual activities
and
drinking
and
drug use. Earlier,
we
noted that
the
other category included multi-item scales that referred
to
most
of
the
contents listed under self-reported behaviors (e.g., items
for
drinking, smoking,
and
reckless driving,
all on the
same scale).
Individuals
who
engage
in
more than
one of
these behaviors would
get
higher scores than individuals
who
engage
in few of
these
activities.
The
other category also included scores
from
ambiguous
single items with general wording (e.g.,
"I
often
take
risks"), as
well
as
items addressing such things
as
shoplifting, vandalism,
career changes,
and
dangerous activities
for
young children
(e.g.,
running
into
the
street, standing
on
chairs, etc.).
For
observed behaviors, distinct levels
of
gender differences
failed
to
emerge because most confidence intervals overlapped
to
a
certain extent. Nevertheless, there
are
seven contrasts
in
Table
2
that
correspond
to
significant mean differences.
In
brief, these
contrasts amount
to the
differences between
two of the
largest
means (i.e., physical skills
and
intellectual
risk
taking)
and
three
of
the
four
smallest means
(i.e.,
physical activity, driving,
and in-
formed
guessing).
In
addition,
the
difference between physical
skills
and
other
was
also significant.
The
other
category included
the
following behaviors: lying about
finding
a
lost coin, teacher
ratings
of
entrepreneurial activities
in a
simulated economy, com-
peting against
two
individuals
in a
coalition game, cheating
to
earn
points, providing
an
electric shock
to a
confederate
to
produce
faster
learning
of a
rule,
and
choosing
a
person
to
date
from
a set
of
pictures.
As
for age
effects,
heterogeneity
analyses
revealed
considerable
variation
in the
size
of d for the
different
age
groups,
^(4)
=
34.69,
p <
.001.
The
weighted means
and 95%
confidence
intervals
(in
parentheses)
for
these groups were
as
follows:
(a)
ages
3 to 9, d = .19
(.14
to
.24);
(b)
ages
10 to 13, d = .26
(.21
to
.31);
(c)
ages
14 to 17, d = .11
(.09
to
.13);
(d)
ages
18 to 21,
d
= .24
(.22
to
.26);
and (e)
older than
21, d = .05
(.03
to
.07).
Using
the
method
of
nonoverlapping
confidence
intervals,
we
found
that
the
gender
gap for the
oldest
age
group
is
significantly
smaller than
the
gender
gap for
high-school students, which,
in
turn,
is
significantly smaller than
the
gaps
for
children,
preadoles-
cents,
and
college
students.
The
means
for
these
latter
three groups
do not
differ.
Moreover,
as
indicated
by the
fact that none
of the
confidence
intervals include
zero,
all of the
means differed signif-
icantly
from
zero (smallest
x2 =
34.16,
p <
.001).
Considered
as
a
whole, then,
the
results
for age
factor revealed
the
following
wave-like
pattern
in the
relative size
of the
gender gap: Level
1 to
Level
2, no
change; Level
2 to
Level
3,
significant
decrease;
Level
3 to
Level
4,
significant increase; Level
4 to
Level
5,
significant
decrease.
However,
it is
possible
that this wave-like pattern simply
re-
flects
a
confounding
of age and
content. Note that children
un-
der 10
were
not
asked about their driving
or
sexual practices.
Similarly,
participants older than
10
were
not
asked
if
they would
like
to
play
on
exercise
equipment,
and so on.
Overall,
64% of the
possible combinations
of
age, content,
and
task emerged
in our
database (see Table
3). To
consider whether
the age
trends simply
reflect
content
effects,
we
conducted
a
weighted
least
squares
regression analysis
as
suggested
by
Hedges
and
Olkin
(1985,
pp.
173-174).
More specifically,
we
first
separated
the
effects
by
task,
then
entered
the
dummy
codes
for the
various contents
in the
first
step
of a
hierarchical regression.
In
other words,
we
conducted
four
separate regressions,
one for
each type
of
task (because
contents
differed
across tasks).
On the
second
and
third steps
of
each
of
these regressions,
we
entered
the
dummy codes
for the age
groups
and
then
the Age X
Content interaction terms, respectively.
For
both self-reported behaviors
and
observed behaviors,
the
vari-
ation accounted
for by age
remained
after
controlling
for
content,
X*
=
185.35
and
9.83, respectively.
In
addition, however,
the
Age X
Content interaction terms
for all
three tasks explained
significant
variation
in the
effect
sizes even
after
controlling
for
content
and
age. These results show that
the
initial findings
for age
were
not
simply
the
result
of a
confounding
of age and
content.
A
more accurate interpretation
is to say
that
different
contents pro-
duced
different
patterns
of
age-related change (see Table
3).
Whereas some contents produced
an
increase
in the
gender
gap in
college followed
by a
decrease (e.g., drinking
and
drugs, smoking),
others produced monotonic increases
or
decreases
in the
size
of the
gender
gap
(e.g., choice dilemma, driving, sexual activities).
In
any
event, there
was no one
pattern
of age
trends that
was
true
for
all
contents.
As for the
factor
of
ambiguity,
the
means
and
confidence inter-
vals
for the
unambiguous
and
ambiguous measures
of risk
taking
were
d = .12
(.13
to
.15)
and d = .16
(.14
to
.18), respectively.
Both means
differed
significantly
from
zero,
but
they
did not
differ
from
each other.
Year
of
publication
and
publication
type.
To
determine
whether
the
gender
gap in risk
taking
has
grown smaller over time,
we
formed
two
groups
of
studies:
(a)
those conducted between
1964
and
1980
(n
= 83
effects)
and (b)
those conducted between
1981
and
1997
(n = 235
effects).
These intervals were selected
to
divide
the
total span
of 34
years into
two
periods
of 17
years each.
The
means
and
confidence intervals
for
these
two
periods were
d
= .20
(.17
to
.23)
and d = .13
(.12
to
.14),
respectively.
The
difference
between these means
was
significant,
so the
gender
gap
appears
to be
growing smaller over time.
To
determine whether
effect
sizes
covaried
with
the
type
of
publication,
we
created
four
categories.
The
first
category con-
tained
14
studies
from
top-tier journals that have both
a
general
focus
and
high standards
of
admission (e.g.,
an 80%
rejection rate):
Developmental
Psychology, Merrill-Palmer Quarterly,
Child
De-
velopment,
Journal
of
Personality
and
Social Psychology,
and
American Journal
of
Public Health.
The
second category con-
tained
98
studies
from
a
large number
of
second-tier journals
(e.g.,

374
BYRNES,
MILLER,
AND
SCHAFER
Table
1
Studies
Used
in the
Meta-Analysis
Study
Aharoni
(1986)
Anderson
&
Mathieu
(1995)
Anderson
&
Mathieu (1996)
Arenson
(1978)
Arnett
&
Jensen (1994)
Bachman
et
al.
(1991)
Barnes
&
Olson (1977)
Barrett
(1980)
Beutell
&
Brenner (1986)
Bevier (1993)
Block
&
Keller (1995)
Bofinger
(1984)
Bonnelle
(1995)
Booth
(1995)
Boverie
et al.
(1994)
Bradley
et al.
(1972)
Breakwell
et al.
(1991)
Campbell
et al.
(1992)
Carlson
&
Cooper
(1974)
Cassell (1992)
Catania
et al.
(1994)
Catania
et al.
(1995)
Cecil
(1972)
Chapman
et al.
(1980)
Choi
&
Catania
(1996)
Chusmir
&
Koberg
(1986)
Clifford
et al.
(1989)
Clifford
et al.
(1990)
Cochran
et al.
(1991)
Cochran
&
Peplau (1991)
Coet
&
McDermott
(1979)
Cooper
et al.
(1994)
Cvetkovich
(1972)
Dahlback
(1991)
Deldin
&
Levin (1986)
Dolcini
&
Adler
(1994)
Dwyer
et al.
(1994)
Ebbeson
&
Haney
(1973)
ds
(nmaf,
"female,
task, content,
age)
-.13 (54,
50, 1, 3, 1)
.12
(54,
50, 3, 2, 1)
-.57 (72,
48, 2,
3,
4)
.02
(316, 165,
2, 3, 4)
.00
(57,
57, 3, 5, 1)
-.21 (108, 125,
2,
4, 3)
-.05 (108, 125,
2, 1, 3)
.11
(108,
125,2,
3, 3)
.33
(108, 125,
2, 2, 3)
.60
(108, 125,
2, 5, 3)
.02
(8500,
8500,
2, 1, 3)
.23
(150,
150,
2, 5, 3)
.04
(419, 400,
2, 3, 4)
.43
(118,
84,
2,5,4)
-.24(12,
11, 1,
1,2)
.00
(192,
32,
1,
2, 5)
.26
(93,
74, 1, 1, 5)
.43
(93,
74, 1, 3, 5)
.28
(93,
74, 1, 3, 5)
.18
(55, 101,
2, 1, 4)
.65
(55, 101,
2, 3, 4)
.75
(55,
101,
2,
2,
4)
.55
(55, 101,
2, 5, 4)
-.05 (398, 195,
2, 3, 5)
-.24 (398, 195,
2, 3, 5)
.52
(61,
41, 2, 5, 4)
.34
(44,
63, 3, 7, 4)
.24
(240,
331,
2, 5, 4)
.08
(124, 131,
2, 3, 4)
.62
(20,
14, 3, 3, 5)
.92
(20,
14, 3, 3, 5)
.19
(154,
341,
1, 1, 5)
.04
(198, 261,
2, 3, 5)
.03
(171, 195,
2, 3, 5)
.00(119,
115,
1,
1,4)
-.17 (297, 339,
3,
2, 2)
.02
(1022,
680,
3, 2, 1)
.03
(170, 151,
3, 2, 3)
.19
(138, 132,
3, 2, 2)
.33
(145,
65, 3, 2, 3)
.35
(653, 468,
3, 2, 1)
-.13
(1167,
1063,2,
3,5)
-.35 (792, 745,
2, 3, 5)
.33
(96,
59, 2, 5, 5)
.08
(84, 116,
3,7,
1)
.14
(46,
49, 3, 7, 1)
.53
(48,
52, 3, 7, 2)
.83
(54,
52, 3, 7, 2)
-.20 (91,
97, 2, 3, 4)
-.16 (28,
44, 2, 3, 4)
.22
(97,
91, 4, 3, 4)
.25
(91,
97, 4, 3, 4)
.66
(91,
97, 4, 3, 4)
.34
(52,
48, 1, 1, 4)
.51
(52,
48, 1, 1, 4)
.14
(616, 560,
2, 3, 3)
.63
(40,
30, 3, 5, 4)
.08
(77,
86, 2, 5, 5)
.00
(12,
12, 1, 2, 4)
-.19
(88,
95, 2, 4, 2)
-.04 (88,
95, 2, 1, 2)
.60
(88,
95, 2, 3, 2)
-.15 (904, 329,
2, 1, 5)
-.15
(904, 329,
2, 3, 5)
.29
(278, 186,
3, 3, 5)
Study
Eysenck
&
Eysenck (1977)
Fagley
&
Miller
(1990)
Farrington
&
Kidd
(1977)
Finney
(1984)
Flaherty
&
Arenson (1978)
Flisher
&
Chalton
(1995)
Foersterling
(1980)
Freeman
et al.
(1994)
Furnham
&
Saipe
(1993)
Gallois
et al.
(1992)
Gibbons
&
Gerrard
(1995)
Gibbons
et al.
(1995)
Ginsburg
&
Miller
(1982)
Grupp
et al.
(1971)
Hayes (1973)
Heilizer
&
Cutter (1971)
Hudgens
&
Fatkin
(1985)
Ingersoll
et al.
(1993)
Ingersoll
&
Orr
(1989)
Irwin
&
Tolkmitt
(1968)
<k
("male,
"female-
&&<
Content,
age)
.38
(709, 1398,
2, 5, 5)
-.42 (45,
51, 1, 2, 4)
-.22 (45,
51,
1,
2, 4)
-.18 (45,
51, 1, 2, 4)
-.12 (45,
51, 1, 2, 4)
.00
(45,
51, 1, 2, 4)
.00
(40,
54, 1,
2,
4)
.13
(40,
54, 1, 2, 4)
.31
(40,
54, 1, 2, 4)
.49
(40,
54, 1, 2, 4)
.49
(40,
54, 1, 2, 4)
-.21 (49,
35, 3, 8, 5)
.00 (6,
8,
3, 6, 4)
.66
(33,
51, 3, 6, 4)
1.11
(131,82,2,5,4)
.02
(28,
40, 2, 5, 3)
.15
(28,
40, 2, 1, 3)
.03
(28,
40, 2, 2, 3)
.43
(28,
40, 2, 5, 3)
.18
(28,
40, 2,
3,
3)
.15
(28,
40, 3, 1, 3)
.03
(28,
40, 3, 2, 3)
.18
(28,
40, 3, 3, 3)
.02
(28,
40, 3, 5, 3)
.43
(28,
40, 3, 5, 3)
.15
(60,
60, 1, 3, 3)
.21
(60,
60, 1, 3, 3)
.37
(60,
60, 1, 3, 3)
.49
(60,
60, 1, 3, 3)
.67
(60,
60, 1, 3, 3)
-.03 (507, 151,
2, 3, 5)
.38
(239,
74, 2, 3, 5)
-.19 (268,
77, 2, 3, 5)
.85
(41,
29, 2, 2, 5)
.01
(77,
74, 2, 3, 5)
-.17 (303, 376,
2, 3, 4)
.00
(303,
376,
2, 4, 4)
.06
(303, 376,
2, 1, 4)
.48
(303, 376,
2, 2, 4)
.79
(72, 154,
1, 3, 3)
.65
(218, 214,
1, 3, 3)
.54
(152,
89, 3, 2, 1)
.86
(58,
25, 3, 2, 1)
.89
(67,
28, 3, 2, 1)
1.07
(45,
16, 3, 2, 1)
.11
(456,849,2,5,5)
.07
(30,
30, 3, 5, 1)
.43
(84,
60, 3, 1, 4)
.45
(27,
27, 3, 1, 4)
.66
(27,
27, 3, 5, 4)
.77
(84,
60, 3, 5, 4)
-.78 (18,
18, 3, 3, 5)
1.00
(9, 9, 3, 3, 5)
-.22 (704, 672,
2, 5, 3)
-.10 (704, 672,
2, 4, 3)
.00
(704, 672,
2, 1, 3)
.02
(704, 672,
2, 1, 3)
.03
(704, 672,
2, 1, 3)
.06
(704, 672,
2, 2, 3)
.10
(704, 672,
2, 5, 3)
.63
(704, 672,
2, 3, 3)
.23
(754, 754,
2, 3, 3)
.00
(20,
20, 3, 5, 4)

RISK
TAKING
375
Table
1
(continued)
Study
Jackson
&
Gray
(1976)
Jamieson
(1969)
lessor
et
al.
(1995)
Karabenick
&
Addy
(1979)
Kass (1964)
Kelling
et al.
(1976)
Kogan
&
Dorros
(1978)
Kogan
&
Wallach
(1964)
Kohler
(1996)
Kopfstein
(1973)
Kourilsky
&
Campbell (1984)
Kreitler
&
Zigler
(1990)
Krishna
(1981)
Lamm
(1979)
Lampert
&
Yassour
(1992)
Leigh
et al.
(1993)
Lettman
(1981)
Levin
&
Chapman (1993)
Levin
et al.
(1988)
Lupfer
et al.
(1971)
Martinez (1995)
Martuza
(1970)
McCormack
et al.
(1993)
McDonald
(1976)
McGaffney
(1976)
McKelvie
&
Schamer
(1988)
Michaels
&
Getting
(1979)
Miller (1987)
Miller
&
Byrnes (1997)
rfs
("male,
"female,
task.
Content,
age)
-.09
(142,
95, 3, 3, 5)
-.04(142,
95, 3,
3,5)
.22
(142,
95, 3, 3, 5)
.03
(142,
95, 3, 3, 5)
.33
(42,
42, 3, 5, 2)
.14
(639,
847,
2, 5, 3)
.54
(40,
40, 3, 4, 4)
.80
(21,
21, 3, 5, 1)
-.16
(57,
85, 1, 1, 3)
.18
(57,
85, 1,
1,
3)
-.15 (80,
80, 1, 1, 4)
-.10 (80,
80, 1, 1, 4)
-.09 (80,
80, 1, 1, 4)
.00
(80,
80, 1, , 4)
.12
(80,
80, 1, , 4)
.14
(80,
80, 1, , 4)
.20
(80,
80, 1, , 4)
.23
(80,
80, 1, , 4)
.33
(80,
80, 1, , 4)
.33
(80,
80, 1, , 4)
.02(114,
103,
1,
1,4)
-.05
(114,
103,
3,
5,4)
-.59(114,
103,
3,
1,4)
.57
(52,
48, 2, 5, 4)
-.90 (30,
30, 3, 5, 1)
-.14
(30,
30, 3, 5, 1)
.10
(417,
392,
3, 8, 1)
-.05 (30,
30, 3, 4, 1)
.17
(30,
30, 3, 4, 2)
-.30
(100,
100,
1, 1, 3)
-.45 (24,
24, 3, 5, 3)
-.07 (24,
24, 3, 5, 2)
-.24
(181,
189,
1, 3, 5)
.30
(181,
184,
1, 3, 5)
-.09
(646, 767,
2, 3, 5)
.27
(49,
61, 1, 3, 4)
.00
(90, 104,
1, 2, 4)
.10
(50,
60, 1, 2, 4)
.38
(50,
60, 1, 2, 4)
.31
(380,
303,
3, 5, 3)
-.39 (46,
44, 1, 1, 4)
-.35 (46,
44, 1, 1, 4)
.20
(46,
44, 1, 1, 4)
.73
(46,
44, 1, 1, 4)
-.11
(68,92,3,
1,3)
.41
(87,
65, 4, 3, 4)
.49
(76,
57, 2, 3, 4)
.04
(76,
57, 2, 3, 4)
.37
(28,
30, 1, 1, 4)
.27
(28,
44, 3, 8, 4)
.03
(186, 123,
3, 3, 5)
.05
(174, 117,
3, 3, 5)
.29
(75,
75, 3, 6, 4)
.43
(75,
75, 3, 6, 4)
.45
(58,
54, 1, 3, 3)
.49
(33,
32, 3, 4, 2)
.77
(32,
34,
3,
4, 2)
.24
(33,
32, 3, 5, 2)
.26
(33,
32, 3, 5, 2)
.48
(32,
34, 3, 5, 2)
1.17
(32,
34, 3, 5, 2)
.66
(33,
32, 3, 7, 2)
.71
(32,
34, 3, 7, 2)
.10
(55,
60, 1, 3, 2)
.19
(55,
60, 1, 3, 2)
Study
Miller
&
Fagley (1991)
Miller
&
Hoffman
(1995)
Mindock
(1972)
Minton
&
Miller
(1970)
Montgomery
&
Landers
(1974)
Moore
&
Erickson
(1985)
Moore
&
Rosenthal
(1991)
Moore
&
Rosenthal (1992)
Morrongiello
&
Bradley
(1997)
Muldrow
&
Bayton
(1979)
Neale
&
Bazerman
(1985)
Ottomanelli
(1993)
Parra
(1988)
Poppen (1994)
Poppen (1995)
Potts
et al.
(1994)
Potts
et al.
(1995)
Reardon
(1981)
Reddon
et al.
(1996)
Roberts (1975)
Rotheram-Borus
et al.
(1992)
Rothman
et al.
(1993)
Rowe
(1991)
Rudolph
(1996)
Rutte
et al.
(1987)
Sadava
&
Forsyth
(1976)
Scherer (1987)
Schilling
et al.
(1991)
Schwartz
(1983)
Seeborg
et al.
(1980)
Shaw
et al.
(1992)
*
("mate,
"female,
^k,
Content,
age)
.36
(55,
60, 1, 3, 2)
.37
(55,
60, 1, 3, 2)
.55
(55,
60, 1, 3, 2)
.84
(55,
60, 1, 3, 2)
.00
(50,
44, 1, 2, 4)
.58
(1204,
1204,
2, 5, 5)
-.11
(9, 9, 3,
5,4)
.07
(26,
26, 1, 1, 4)
.07
(33,
33, 3, 5, 1)
.07
(33,
33, 3, 5, 1)
.07
(33,
33, 3, 5, 1)
.20
(202,
391,
2, 3, 4)
.30
(263,
674,
2, 3, 4)
.00(71,
118,2,
3,4)
.53
(71, 118,
2,2,4)
.00(71,
118,2,4,4)
.47
(20,
20, 1, 3, 1)
.32
(20,
20, 1, 3, 2)
.95
(100,
100,
1, 1, 5)
.00
(50,
50, 1, 2, 5)
-.01 (97,
69, 2, 1, 5)
-.17
(97,
69, 2, 3, 5)
.18
(30,
30, 3, 8, 4)
.00(105,
110,2,3,4)
.30
(74, 125,
2, 3, 4)
.98
(24,
26, 2, 5, 1)
.74
(39,
44, 2, 5, 1)
.88
(39,
44, 2, 5, 1)
-.49(108,
118,
1, 1, 3)
-.09
(28,
39, 2, 5, 3)
-.05 (50,
16, 2, 5, 3)
.45
(101,
134,
3, 4, 4)
.10
(77,
83, 2, 3, 3)
.59
(104,
89, 1, 2, 4)
.87
(99,
140,
2, 5, 3)
.11
(906,
880,2,4,
3)
.17
(906, 880,
2, 5, 3)
.43
(906, 880,
2, 2, 3)
-.04
(906,
880,
2, 2, 3)
.54
(906,
880,
2, 5, 3)
.00
(33,
33, 1, 2, 4)
.42
(72,
72, 1, 1, 4)
.56
(72,
72, 1, 1, 4)
.94
(195, 142,
2, 5, 4)
.25
(135,
109,
2, 3, 5)
.10
(135,
109,
2, 3, 5)
-.14 (86,
94, 1, 1, 4)
-.55 (51,
42, 1, 1, 4)
-.41
(51,
42, 1, 1, 4)
-38
(51,42,
1,
1,4)
-.36 (51,
42, 1, 1, 4)
-.20
(51,
42, 1, 1, 4)
-.14
(51,
42, 1, 1, 4)
-.02(51,42,
1,
1,4)
.00
(51,
42, 1,
1,
4)
.02
(51,
42, 1, 1, 4)
.16
(51,42,
1,
1,4)
.21
(51,
42, 1, 1, 4)
.36
(51,
42, 1, 1, 4)
.67
(678,
673,
2, 2, 4)
.30
(678,
673,
2, 2, 4)
.64
(678,
673,
2, 5, 4)
.17
(678, 673,
2, 1, 4)
.21
(678, 673,
2, 1, 4)
(table
continues)

376
BYRNES,
MILLER,
AND
SCHAFER
Table
1
(continued)
Study
Sheer
&
Cline
(1995)
Sitkin
&
Weingart
(1995)
Sivak
et
al.
(1989)
Slakter
(1967)
Slakter
(1969)
Slakter
et al.
(1971)
Slovic
(1966)
Snyder
(1984)
Sorrentino
et al.
(1992)
Spears
et al.
(1992)
Speltz
et al.
(1990)
Stanford
et al.
(1996)
Stapp
(1986)
Starrett
(1983)
Steptoe
et al.
(1995)
Struckman-Johnson
&
Struckman-
Johnson
(1996)
rfs
(nmale,
nfemal(!,
task,
content,
age)
.20
(678, 673,
2, 1, 4)
.24
(678, 673,
2, 3, 4)
.43
(678, 673,
2, 3, 4)
.29
(678, 673,
2, 5, 4)
.37
(678, 673,
2, 2, 4)
.49
(124, 141,
2, 3, 4)
.46
(122, 138,
2, 5, 4)
.00
(38,
25, 1, 2, 4)
.42
(90,
90, 3, 3, 5)
.57
(90,
90, 3, 3, 5)
.52
(21,31,
3,
1,4)
.59
(21,
31, 3, 1, 4)
-.17 (380, 297,
3, 1, 2)
.08
(522, 548,
3, 1, 3)
.11
(600, 691,
3,
1,4)
-.09 (89,
50, 3, 5, 1)
.11(110,40,3,5,2)
.13
(117,42,
3, 5, 2)
.21
(94,
86, 3, 5, 1)
.33
(108,
46, 3, 5, 2)
.27
(173,
49, 3, 5, 3)
.24
(44,
49, 3, 5, 3)
.32
(44,
49, 3, 5, 3)
.49
(48,
92, 3, 4, 4)
.32
(124, 190,
1, 3, 4)
.31
(129, 124,
2, 5, 1)
.08
(156, 147,
2, 1, 3)
.49
(88, 178,
2, 1, 4)
.23
(156, 147,
2, 2, 3)
.61
(88, 178,
2, 2, 4)
.38
(156, 147,
2, 2, 3)
.54
(88, 178,
2, 2, 4)
-.08 (124,
211,2,5,5)
.19(19,46,
2,
5,4)
.33
(26,
27, 2, 5, 2)
.60
(17,
28, 2, 5, 3)
.01
(7109,
9015,
2, 4, 5)
.43
(53,
54, 1, 3, 4)
Study
Tinsley
et al.
(1995)
Touhey
(1971)
Traub
&
Hambleton
(1972)
Trocki
(1992)
Walesa
(1975)
Ward
et al.
(1988)
Warren
&
Simpson
(1980)
Wayment
et al.
(1993)
West
et al.
(1996)
White
&
Johnson
(1988)
Williams
(1973)
Wilson
et al.
(1996)
Wyatt
(1988)
Yesmont
(1992)
Yinon
et al.
(1975)
Zeff
et al.
(1994)
Zuckerman
et al.
(1990)
ds
(nmale,
nfema!e,
task,
content,
age)
-.19(139,
147,
2,4,
3)
-.10 (139, 147,
2, 1, 3)
-.08 (139, 147,
2, 2, 3)
.06
(40,
40, 1, 1, 4)
.37
(505, 493,
3, 1, 2)
.58
(495, 487,
3, 1, 2)
-.03 (292, 326,
2, 3, 5)
.00
(18,
18, 3, 5, 2)
.00
(18,
18, 3, 5, 2)
.58
(18,
18, 3, 5, 1)
-.25 (30,
30, 3, 8, 4)
.35
(30,
30, 3, 8, 4)
.18
(7673,
1320,
3, 3, 5)
-.43 (50,
50, 2, 1, 5)
.22
(559, 373,
2, 2, 4)
.33
(559, 373,
2, 2, 4)
.02
(559, 373,
2, 2, 4)
.30
(559, 373,
2, 2, 4)
-1.23 (34,
31, 2, 3, 3)
-.96 (151, 155,
2, 3, 3)
-.77
(182,
174,
2, 3, 4)
-.36 (201, 185,
2, 4, 3)
.44
(70, 127,
1, 3, 4)
.53
(36,
68, 1, 3, 4)
-.29 (73, 129,
1, 2, 4)
-.15 (73, 129,
1, 2, 4)
-.11 (73, 129,
1,2,4)
-.08 (73, 129,
1, 2, 4)
-.06 (73, 129,
1, 2, 4)
-.03 (73, 129,
1, 2, 4)
.44
(73, 129,
1, 2, 4)
.41
(94, 159,
2, 3, 4)
.79
(94, 159,
2, 3, 4)
.87
(94, 159,
2, 3, 4)
1.45
(40,
29, 3, 8, 4)
-.15 (26,
19, 2, 5, 5)
-.27 (422, 649,
2, 4, 4)
Note.
The
codes
for
task,
content,
and age are
explained
in the
text.
Journal
of
Drug Issues, Adolescence) that have
a
more specific
focus
(e.g., just articles
on
drug abuse, just articles
on
adolescents,
etc.)
and
sometimes
a
higher acceptance rate than
the
top-tier
journals.
The
third category included
26
studies that were pub-
lished
in
books
or
journals that have lower standards
for
admission
than
the
journals
in the first two
tiers (i.e., Psychological Reports,
Journal
of
Psychology,
Journal
of
General Psychology, Percep-
tual
and
Motor Skills, Research
Quarterly).
The
fourth
category
contained
the 15
dissertation studies.
The
means
and
confidence intervals
for
these
four
categories
were
as
follows:
(a)
top-tier outlets,
d = .03
(.01
to
.05);
(b)
second-tier outlets,
d = .16
(.15
to
.17);
(c)
third-tier outlets,
d =
.12
(.09
to
.15);
and (d)
dissertations,
d = .25
(.21
to
.29).
The
effect
sizes
from
top-tier
outlets
were significantly
smaller
than
the
effect
sizes
from
all
other outlets.
In
addition,
the
effect
sizes
from
second-
and
third-tier outlets were significantly smaller than
the
effect
sizes
from
dissertations. Thus,
effect
sizes seem
to get
smaller
as
higher standards
for
admission
are
imposed
(i.e.,
meth-
odological rigor
and the
need
to
appeal
to a
broad audience).
However,
it is
important
to
qualify
this interpretation
in two
ways.
First,
it
should
be
recalled that
the
mean
for
dissertations might
be
somewhat
inflated because
of the
Boolean search process used
by
the
computerized system
for
dissertations. Second, journals
and
contents were
often
confounded.
In
fact,
data points were missing
for
28% of the
possible combinations
of
journal, task,
and
topic.
As
a
result,
we ran
four
additional analyses (one
for
each task)
in
which
we
examined
the
role
of
journal quality
after
controlling
for
content.
These
analyses showed that journal quality still accounted
for
significant variation
in the
effect
sizes
after
content
was
con-
trolled. However,
the
Quality
X
Content interaction terms were
often
significant
as
well. Hence,
it is
more accurate
to say
that
certain
journals
(e.g.,
Journal
of
Drug
Education)
seem
to be
more
interested
in
publishing gender
differences
on
certain topics (e.g.,
alcohol use) than other journals (e.g., Journal
of
Personality
and
Social
Psychology).
These contents,
in
turn, produced relatively
larger
or
smaller gender
differences.

RISK TAKING
377
Table
2
Mean
Effect
Sizes
by
Task
and
Content
Within
Task
Task
Hypothetical
choice
Choice dilemma
Framing
Other
Self-reported behavior
Smoking
Drinking/drug
use
Sexual activities
Driving
Other
Observed behavior
Physical activity
Driving
Informed
guessing
Gambling
Risky
experiment
Intellectual
risk
taking
Physical skills
Other
M
.07*
.05
.35*
-.02
.04*
.07*
.29*
.38*
.16*
.17*
.18*
.21*
.41*
.40*
.43*
.15*
95%
confidence interval
.05 to .12
-.02
to .12
.29 to .41
-.05
to .01
.02 to .06
.05
to .10
.26 to .32
.35 to .41
.10 to .22
.12
to .22
.13 to .23
.14
to .28
.21 to .61
.25 to .55
.28 to .58
.04 to .26
N
effects
44
27
25
10
19
47
21
35
11
14
11
33
4
7
7
7
Note.
The
asterisk indicates that
the
mean
differs
significantly
from
zero.
The
content labels
are
explained
in the
text.
Discussion
In
the
present
article,
we had two
primary
aims:
(a) to
determine
the
average
effect
size
for
gender differences
in risk
taking
and (b)
to
reveal
the
extent
to
which gender differences vary according
to
context
and age
level.
We
pursued
the
first
aim to
bring order
to a
diverse
and
widely scattered literature.
We
pursued
the
latter
to
test
the
explanatory adequacy
of
several contemporary
theories
of risk
taking.
In
what follows,
we
summarize
and
interpret
our
findings
in
light
of
these
issues.
Overview
of the
Main
Findings
At
a
general level,
our
results clearly support
the
idea that male
participants
are
more likely
to
take
risks
than female participants.
In
nearly every case,
the
mean
effect
size
for a
given type
of risk
taking
was
significantly greater
than
zero (see Table
2), and
almost
half
of the
effects
(i.e., 48%) were larger than
.20
(i.e.,
the
conventional
cutoff
for
small
effects).
However,
a
more qualified
interpretation
of our
results
is to say
that gender differences varied
according
to
context
and age
level.
As
shown
in
Table
3,
certain
topics were associated with
nontrivial
gender differences that
seemed
to
increase with
age
(e.g., driving), whereas others were
associated with considerably smaller gender
differences
at
most
ages (e.g.,
smoking),
or
associated with
shifts from
positive
to
negative
effects
as
children grow older (e.g., sexual activities).
Thus,
if risk
taking
is, in
fact,
"an
attribute
of the
masculine
psychology"
as
Wilson
and
Daly (1985,
p. 61)
suggest,
it
does
not
seem
to
manifest
itself
in a
simple
or
constant
way
across ages
or
contexts.
Of
course,
one
could argue that
the
variability evident
in Ta-
bles
2 and 3 is
largely
due to the
fact
that
we
adopted
an
extremely
broad
definition
of risk
taking.
In
other words,
by
including both
prototypical
and
nonprototypical types
of risk
taking
in our
meta-
analysis,
we
artificially
increased
our
chances
of finding
signifi-
cant
fluctuations
in the
size
of the
gender gap. Note, however, that
there
is
considerable variation
in the
size
of the
gender
gap
across
age
levels
and
topics even within
the
category
of
prototypical
risk
taking (see Table
3).
Thus, whereas
we
agree that
our
results
and
conclusions largely reflect
our
preferred definition
of risk
taking,
Table
3
Age X
Content Interaction Means
Task
Hypothetical choice
Choice dilemma
Framing
Other
Self-reported behavior
Drinking/drugs
Driving
Sexual activities
Smoking
Other
Observed
behavior
Informed
guessing
Physical activity
Driving
Physical skills
Gambling
Risky experiments
Intellectual
risk
taking
Other
1
n/d
n/d
.03
n/d
n/d
n/d
n/d
.55*
n/d
.22*
n/d
-.05
.03
n/d
.10
.10
2
-.24
n/d
.39*
-.04
n/d
.60*
-.19
.33
.31*
-.06
n/d
.48*
.27*
n/d
.68*
n/d
Age
level
3
-.24*
n/d
.53*
.02
.16*
.22*
-.04
.27*
.06
.14
n/d
n/d
.27*
n/d
n/d
n/d
4
.07*
.06
.42*
.17*
.37*
.18*
-.15*
.52*
.09
n/d
n/d
.48*
.31*
.41*
.34
.38*
5
.38*
.00
.13*
-.15*
.85*
-.11*
.01
.29*
n/d
n/d
.17*
n/d
n/d
n/d
n/d
-.21
Sig. contrasts
5>4>3
3>5
4
>3 >5
4,5
> 3
2>3 = 4>5
4>3,
5
1
>3;4>5
2>3,
4
1
>2
2> 1
Note.
Sig.
contrasts means significant
contrasts
(i.e.,
the
confidence intervals
for the
compared age-level means
do
not
overlap).
The
asterisk
indicates that
the
mean
is
significantly larger than
zero,
n/d
stands
for no
data (i.e.,
gender-difference
studies
for
that topic
and age
range could
not be
located).
Age
levels
are
defined
in the
text.

378
BYRNES,
MILLER,
AND
SCHAFER
other definitions would
not
necessarily undermine
our
basic
find-
ings.
One of the
reasons
for
supplying
raw
data
in a
meta-analysis
(Table
1) is to
give researchers
who
hold
differing
perspectives
the
opportunity
to
reorganize, truncate,
or
elaborate
the
data
in
ways
that
might yield alternative interpretations.
One
further
benefit
of
using
the
broad definition
is
that
it
helped
us
reveal
two
additional
findings
that might have been missed
had
we
limited
our
analysis
to
just prototypical
forms
of risk
taking.
The first is an
apparent lack
of
discernment
on the
part
of men and
boys.
In one of our
analyses,
we
showed that males took more
risks
even
when
it was
clear that
it was a bad
idea
to
take
a risk. The
same
analysis revealed
the
opposite
was
true
for
women
and
girls;
that
is,
they seemed
to be
disinclined
to
take
risks
even
in
fairly
innocuous
situations
or
when
it was a
good idea
to
take
a risk
(e.g.,
intellectual
risk
taking
on
practice SATs). Whereas
the
former
finding
suggests
that
men and
boys
would tend
to
encounter
failure
or
other negative consequences more
often
than women
and
girls,
the
latter
finding
suggests that women
and
girls would tend
to
experience success less
often
than they should.
In our
view, both
of
these trends
are a
matter
of
concern.
However,
as we
noted earlier,
the
issue
of
concern
is
very much
a
function
of the
validity
of the
measures used
(i.e.,
a
gender
difference
on a
valid
and
widely accepted measure would tend
to
generate more concern than
a
difference
on an
ambiguous
or
controversial measure)
as
well
as the
consequences that could arise
from
the risky
behaviors
in
question. Some would argue,
for
example, that policymakers
and
researchers
should
be
more con-
cerned about
a
gender difference
in an
actual, dangerous behavior
(e.g.,
reckless driving) than
a
gender difference
in
hypothetical
choices.
It is
interesting
that
the
means
for
certain observed
be-
haviors
were considerably larger than
the
means
for
certain hypo-
thetical choices. These
findings
suggest that gender differences
may
be
more likely
to
emerge when people have
to
actually carry
out
a risky
behavior than when they have
to
simply consider
the
pros
and
cons
of two
options.
If so,
then
the
processes involved
in
the
translation
of
cognitions
to
behavior (e.g.,
fear
responses)
may
explain gender
differences
in risk
taking more adequately than
the
cognitive
processes involved
in the
reflective evaluation
of op-
tions.
This possibility requires
further
study.
Another point
to
make
is
that whereas
the
overall mean
of d =
.13
would
be
labeled small
in
some statistical circles (e.g., Cohen,
1992),
it
conveys
a
different
message when
it is
converted into
a
Binomial
Effect
Size Display
(Rosenthal,
1994).
In
particular,
a
mean
of d =
.13
corresponds
to a 6%
difference
in the risk-taking
rates
of men and
women (e.g.,
53% of men
take
risks vs.
only
47%
of
women).
In
potentially dangerous activities that
are
performed
by
a
large number
of
people (e.g., driving, unprotected sex, etc.),
this
6%
difference
would accumulate across behaviors
and
time
to
produce
a
substantial gender
difference
in the
number
of
expected
injuries
or
death (e.g.,
60,000
if we
assume
one
million female
drivers
and one
million male
drivers).
Thus,
we
believe
that many
of
the
gender
differences
in
Tables
2 and 3 are a
matter
of
concern.
Implications
of the
Findings
for
Theories
of
Risk
Taking
As
we
noted earlier,
a
useful
way to
understand
the
theoretical
implications
of our
results
is to
think
of
theories
as
falling into
one
of
three categories (Lopes, 1987).
The first
category consists
of
theories that
are
equipped
to
explain
the
differences between
people
who
take
risks and
people
who
avoid
risks
(e.g., Zucker-
man,
1991).
The
second category consists
of
theories that
are
equipped
to
explain
the
differences between situations that pro-
mote
risk
taking
(in
most people)
and
situations that promote
risk
aversion
(e.g.,
Kahneman
&
Tversky, 1979).
The
third category
consists
of
theories that
are
equipped
to
explain both types
of
differences
(i.e.,
they could explain
why
only certain
people
take
risks
in
certain situations).
After
categorizing theories
in
this way,
the
next step
is to
consider
the
patterns
of
gender differences that could
be
explained
by
the
theories
in
each group. Whereas Category
1
theories could
explain
a
pervasive pattern
of
gender
differences
across contexts,
Category
2
theories could not. Similarly, whereas Category
3
theories could explain
a
pattern
of
variable gender differences
across contexts, theories
in the
other
two
categories could not.
Given that
we
found
the
variable
pattern,
it
seems
reasonable
to
assert that Category
3
theories demonstrate greater explanatory
adequacy
than
the
theories
in
Categories
1 and 2.
However,
we
need
to add
that some
of the
theories
in
Category
3
(e.g., Wilson
&
Daly's
[1985]
competition model)
are not
equipped
to
explain
two
of our findings: (a) 20% of the
effects
were negative
and
(b)
some
of the
means
in
Table
3
were negative
and
significantly
larger than zero. Hence,
our
results suggest that these theories need
to be
revised
as
well.
Of
course,
to say
that certain theories
are
equipped
to
handle
our
results
is not to say
that
any one of
these theories receives direct
confirmation
by our
results.
The
authors
of the
studies
in
Table
1
were
generally
not
interested
in
providing support
for
specific
theories,
so
they
did not
provide
the
kind
of
data needed
to
show
that
one
theory provides
a
better explanation than another.
For
example, they
did not use
measures tapping
the
core theoretical
constructs
of
more than
one
theory.
As
such,
all one
could
do is
describe
in a
post
hoc
manner
how
different
theories would explain
the
results.
For the
sake
of
brevity,
we
leave that task
to the
interested reader.
There
is,
however,
one
additional aspect
of our
results that
may
ultimately
determine
the
adequacy
of
specific accounts.
We
found
that
some contents produced similar gender differences
at
different
ages, whereas others produced developmental increases
or de-
creases
in the
size
of the
gender gap.
Not all of the
theories
in
Category
3
could explain cyclical developmental trends such
as
these.
Hence,
we
argue that
the age
results narrow
the field
further
to
the
following viable theories:
Byrnes's
(1998) self-regulation
model,
Irwin
and
Millstein's (1991)
biopsychosocial
account,
and
Wigfield
and
Eccles's
(1992) expectancy-value model.
Byrnes
(1998) suggested that developmental increases
in risk
taking
are a
function
of the
fact
that children
are
more likely
to find
themselves
in
novel,
unmonitored
environments
as
they grow older
(e.g., going away
to
college). Resisting
the
temptation
to
take
a
risk
requires
a
certain amount
of
self-regulation (i.e.,
a
calibrated
sense
of
uncertainty,
self-corrective
strategies
for
dealing
with
distractions
and
troublesome personality traits,
and a
tendency
to
learn
from
mistakes).
The
apparent surges
in the
gender
gap may
reflect
double standards with respect
to
parental monitoring (e.g.,
more restrictions placed
on
women
and
girls),
overconfidence
in
men
and
boys,
and
less knowledge
of
self-correcting strategies
in
men and
boys. Irwin
and
Millstein's
(1991)
model
was
crafted
to
explain
the
increase
in risk
taking that
is
said
to
occur between
childhood
and
adolescence.
The
model
was not
meant
to
explain

RISK
TAKING
379
gender differences
per se, but it
seems
to
suggest that surges
in the
gender
gap
would
be due to
periodic changes
in the
following:
(a)
biological maturation,
(b)
cognitive scope (e.g.,
future
time per-
spective),
(c)
self-perceptions (e.g., self-esteem),
(d)
perceptions
of
the
social
environment
(e.g.,
parental
and
peer
influences),
(e)
personal values (e.g., independence),
(f)
risk perception (e.g.,
optimistic
bias),
and (g)
characteristics
of the
peer group (e.g., peer
age). These factors
may
independently
or
collectively influence
males
and
females
in
different
ways
at
different
times.
Wigfield
and
Eccles's
(1992)
model suggests that gender differences would
arise whenever males
and
females hold
different
expectations
and
values. Expectations
and
values
can
show dramatic
shifts
when-
ever children enter novel environments (e.g., children making
the
transition
to
middle school).
At
present,
it is not
clear which
of
these models provides
the
most accurate explanation.
Suggestions
for
Future
Research
Several
findings
require
further
explication
and
analysis
in
subsequent studies.
The
first
pertains
to
intriguing differences
in
the age
trends
for the
four
types
of
self-reported behaviors. Table
3
shows that
the
shift
between high school
and
college seems
to
promote
a
sharper increase
in
drinking
and
drug
use in men
than
in
women.
At
present,
it is not
clear whether this
finding
reflects
the
fact
that
men are
confronted with
risk-inducing
contexts more
often
than women
(e.g.,
they attend
a
greater number
of
parties)
or
whether
women have
a
greater capacity
to
negotiate themselves
through
these
risk-inducing
situations than men.
The
latter would
not
appear
to be the
case because women seem
to be
significantly
more likely
to
smoke during their college years than
men and
also
seem
to be
more likely
to
drink, take drugs,
or
engage
in risky
sexual activities
in
their post-college years (see Table
3).
Such
findings
could
be
interpreted
in one of two
ways.
The
first
would
be
that contexts make
different
demands
on men and
women
at
different
points
in
time.
The
second interpretation would
be
that
men
are
somewhat "precocious" (i.e.,
men
engage
in
these activ-
ities earlier than women
but
women eventually catch
up and
surpass men). Future research should determine which
of
these
explanations seems
to be
more accurate.
In
addition, more work
is
needed
to
understand
the
reasons
for
the age
trends revealed
for
driving
and
gambling behaviors.
At
most
age
levels,
the
effects
for
these
two
behaviors were twice
as
large
as the
overall mean
of d =
.13.
In the
case
of
driving,
the
findings
are
troubling because
the
gender
gap
appears
to
widen
with
age. Note, however, that there
was a
high degree
of
ambiguity
in
the
measurement
of
driving
risks. For
example,
in
some studies,
participants simply rated their
own riskiness. In
others, experi-
menters used ambiguous
indexes
such
as
wait time before making
a
turn.
Hence,
more studies with tighter controls
are
needed
to
verify
the age
trends that
we
revealed
for
driving.
In the
case
of
gambling,
the
measures were less ambiguous,
but it is
still
not
clear
why
males seemed
to be
more inclined
to
take gambling
risks
than
females
and why the
gender
gap
does
not
appear
to
change
with
age.
One
factor that could possibly explain both sets
of
findings
for
driving
and
gambling would
be a
gender-linked dif-
ference
in
competitiveness (Miller
&
Byrnes, 1997; Wilson
&
Daly, 1985). Future studies should explore this
possibility
and
others.
In
general, then,
we
revealed
a
number
of
important trends that
require
further
explication.
To add
greater clarity
to the
study
of
gender
differences,
researchers should
(a)
construct unambiguous
measures
of
both appropriate
and
inappropriate forms
of risk
taking,
(b)
construct valid measures tapping
the
core constructs
of
several
different
theories
of risk
taking
(to see
which
of the
remaining
viable theories
is the
most adequate),
and (c)
give these
measures
to
multiple
age
groups
(to
further
probe
the
meaning
of
our age
trends).
The
findings
from
such studies
would
make
an
important
and
much-needed contribution
to the
literature
and
would
provide important insight into possible ways
to
improve
the
risk-taking
skills
of
children, adolescents,
and
adults.
References
References
marked
with
an
asterisk
indicate
studies included
in the
meta-analysis.
*Aharoni,
H.
(1986).
Assessment
of
children's
risk-taking behavior
as
reflected
in
motor
activity.
Unpublished
dissertation,
Ohio
State
Univer-
sity.
*Anderson,
P.
B.,
&
Mathieu,
D. A.
(1995).
The
relationship
of
alcohol
consumption
as a
sexual
disinhibitor
to
high
risk
behavior:
Gender
differences. Journal
of
Sex
Education
and
Therapy,
21,
217-222.
*Anderson,
P.
B.,
&
Mathieu,
D. A.
(1996).
College
students'
high-risk
sexual
behavior
following
alcohol
consumption.
Journal
of
Sex &
Mar-
ital Therapy,
22,
259-264.
*Arenson,
S. J.
(1978).
Age and sex
differences
in the
probability
prefer-
ences
of
children.
Psychological Reports,
43,
697-698.
Arnett,
J.
(1992).
Reckless
behavior
in
adolescence:
A
developmental
perspective.
Developmental
Review,
12,
339-373.
*Arnett,
J.
J.,
&
Jensen,
L. A.
(1994).
Socialization
and risk
behavior
in
two
countries:
Denmark
and the
United
States,
youth
and
Society,
26,
3-22.
Atkinson,
J. W.
(1983).
Personality,
motivation,
and
action: Selected
papers.
New
York:
Praeger.
*Bachman,
J.
G.,
Wallace,
J. M.,
O'Malley,
P. O.,
Johnston,
L.
D.,
Kurth,
C.
L.,
&
Neighbors,
H. W.
(1991).
Racial/ethnic
differences
in
smoking,
drinking,
and
illicit
drug
use
among
American
high
school
seniors,
1976-89.
American
Journal
of
Public Health,
81,
372-377.
*Barnes,
C. P., &
Olson,
J. N.
(1977).
Usage
patterns
of
nondrug
alterna-
tives
in
adolescence.
Journal
of
Drug Education,
7,
359-373.
Baron,
J.
(1994).
Thinking
and
deciding (2nd ed.).
Cambridge,
England:
Cambridge
University
Press.
*Barrett,
F. M.
(1980).
Sexual
experience,
birth
control
usage,
and sex
education
of
unmarried
Canadian
university
students:
Changes between
1968
and
1978.
Archives
of
Sexual
Behavior,
9,
367-386.
Baumrind,
D.
(1991).
Adolescent
exploratory
behavior:
Precursors
and
consequences.
In L. P.
Lipsitt
& L. L.
Mitnick
(Eds.),
Self-regulatory
behavior
and
risk-taking: Causes
and
consequences
(pp.
109—141).
Norwood,
NJ:
Ablex.
*Beutell,
N.
J.,
&
Brenner,
O. C.
(1986).
Sex
differences
in
work
values.
Journal
of
Vocational
Behavior,
28,
29-41.
*Bevier,
R. A.
(1993).
Risk taking
and
decision making behaviors
of
gifted
and
general education adolescents.
Unpublished
doctoral
dissertation,
Wayne
State
University.
*Block,
L.
G.,
&
Keller,
P. A.
(1995).
When
to
accentuate
the
negative:
The
effects
of
perceived
efficacy
and
message
framing
on
intentions
to
perform
a
health-related
behavior.
Journal
of
Marketing
Research,
32,
192-203.
*Bofinger,
C. J.
(1984).
Risk-taking
in
areas
of
business.
Unpublished
doctoral
dissertation,
University
of
Pittsburgh.
*Bonnelle,
K. P.
(1995).
A
profile
of
sensation
seekers: Risk taking
and

380
BYRNES, MILLER,
AND
SCHAFER
risk
reduction behaviors. Unpublished doctoral dissertation, University
of
Memphis.
*Booth,
R. E.
(1995).
Gender differences
in
high-risk
sex
behaviors among
heterosexual drug injectors
and
crack smokers. American Journal
of
Drug
and
Alcohol Abuse,
21,
419-432.
*Boverie,
P.
E.,
Scheuffele,
D. J., &
Raymond,
E. L.
(1994).
Multimeth-
odological
approach
to
examine risk-taking. Current Psychology:
De-
velopmental,
Learning, Personality, Social,
13,
289-302.
*Bradley,
L.,
Snyder,
C.
R.,
&
Katahn,
M.
(1972).
The
effects
of
subject
race
and sex and
experimenter
race
upon
classroom-related
risk-taking
behavior.
Psychonomic
Science,
28,
362-364.
*Breakwell,
G.
M.,
Fife-Schaw,
C.,
&
Clayden,
K.
(1991). Risk-taking,
control over partner choice
and
intended
use of
condoms
by
virgins.
Journal
of
Community
and
Applied Social Psychology,
1,
173-187.
Byrnes,
J. P.
(1998).
The
nature
and
development
of
decision-making:
A
self-regulation
model.
Hillsdale,
NJ:
Erlbaum.
Byrnes,
J. P., &
Takahira,
S.
(1993). Explaining gender differences
on
SAT-math items. Developmental Psychology,
29,
805-812.
"•Campbell,
S.
M.,
Peplau,
L.
A.,
&
DeBro,
S. C.
(1992). Women, men,
and
condoms:
Attitudes
and
experiences
of
heterosexual college students.
Psychology
of
Women Quarterly,
16,
273-288.
*Carlson,
K.,
&
Cooper,
R. E.
(1974).
A
preliminary investigation
of risk
behavior
in the
real world. Personality
and
Social Psychology Bulle-
tin,
1,
7-9.
*Cassell,
M.
(1992). Locus
of
control
and
propensity
for
risk
taking
as
related
to
achievement
in
higher education. Unpublished doctoral dis-
sertation, Montana State University.
*Catania,
J.
A.,
Coates,
T.
J.,
&
Kegeles,
S.
(1994).
A
test
of the
AIDS
Risk-Reduction
Model:
Psychosocial
correlates
of
condom
use in the
AMEN
cohort survey. Health Psychology,
13,
548-555.
*Catania,
J.
A.,
Stone,
V.,
Binson,
D.,
&
Dolcini,
M.
(1995). Changes
in
condom
use
among heterosexuals
in
wave
3 of the
AMEN survey.
Journal
of
Sex
Research,
32,
193-200.
*Cecil,
E. A.
(1972).
Factors
affecting individual
risk-taking
attitudes.
Journal
of
Psychology,
82,
223-225.
*Chapman,
A.
J.,
Foot,
H.
C.,
&
Wade,
F. M.
(1980). Children
at
play.
In
D. J.
Osborn
& J. A.
Levis
(Eds.),
Human factors
in
transport research.
Volume
2.
User
factors:
Comfort,
the
environment
and
behaviour (pp.
380—388).
London: Academic Press.
*Choi, K-H.,
&
Catania,
J. A.
(1996). Changes
in
multiple sexual partner-
ships,
HIV
testing,
and
condom
use
among
US
heterosexuals
18
to 49
years
of
age, 1990
and
1992. American Journal
of
Public Health,
86,
554-556.
*Chusmir,
L.
H.,
&
Koberg,
C.
(1986). Creativity differences among
managers. Journal
of
Vocational Behavior,
29,
240—253.
*Clifford,
M.
M.,
Chou,
F.
C.,
Mao, K.-N.,
Lan,
W.
L.,
&
Kuo, S.-Y.
(1990). Academic
risk-taking,
development,
and
external constraint.
Journal
of
Experimental
Education,
59,
45—64.
*Clifford,
M.
M.,
Lan,
W.
L.,
Chou,
F.
C.,
& Qi, Y.
(1989). Academic
risk-taking:
Developmental
and
cross-cultural observations. Journal
of
Experimental
Education,
57,
321—338.
*Cochran,
S.
D.,
Mays,
V.
M.,
&
Leung,
L.
(1991). Sexual practices
of
heterosexual
Asian-American
young
adults:
Implications
for
risk of HIV
infection.
Archives
of
Sexual Behavior,
20,
381-391.
*Cochran,
S.
D.,
&
Peplau,
L. A.
(1991). Sexual
risk
reduction behaviors
among young heterosexual adults. Social Science
and
Medicine,
33,
25-36.
*Coet,
L.
J.,
&
McDermott,
P. J.
(1979).
Sex, instructional set,
and
group
make-up:
Organismic
and
situational
factors
influencing
risk-taking.
Psychological
Reports,
44,
1283-1294.
Cohen,
J.
(1992).
A
power primer. Psychological Bulletin, 112,
155-159.
*Cooper,
M.
L.,
Peirce,
R.
S.,
&
Huselid,
R. F.
(1994).
Substance
use and
sexual
risk
taking among black adolescents
and
white adolescents.
Health
Psychology,
13,
251-262.
*Cvetkovich,
G.
(1972).
Effects
of sex on
decision
policies
used
for
self
and
decision policies used
for
other persons. Psychonomic Science,
26,
319-320.
*Dahlback,
O.
(1991). Saving
and risk
taking. Journal
of
Economic Psy-
chology,
12,
479-500.
*Deldin,
P.
J.,
&
Levin,
I. P.
(1986).
The
effect
of
mood induction
in a
risky
decision-making task. Bulletin
of the
Psychonomic Society,
24,
4-6.
DiClemente,
R.
J.,
Hansen,
W. B., &
Ponton,
L. E.
(1995). Handbook
of
adolescent
health
risk
behavior.
New
York:
Plenum.
*Dolcini,
M.
M.,
&
Adler,
N. E.
(1994).
Perceived competencies, peer
group
affiliation,
and risk
behavior among early adolescents. Health
Psychology,
13,
496-506.
*Dwyer,
R.,
Richardson,
D.,
Ross,
M.
W.,
Wodak,
A.,
Miller,
M.
E.,
&
Gold,
J.
(1994).
A
comparison
of
HTV
risk
between women
and men
who
inject drugs. AIDS Education
and
Prevention,
6,
379-389.
*Ebbeson,
E.
B.,
&
Haney,
M.
(1973). Flirting with death: Variables
affecting
risk
taking
at
intersections. Journal
of
Applied
Social Psychol-
ogy,
3,
303-324.
*Eysenck,
S. B.
G.,
&
Eysenck,
H. J.
(1977).
The
place
of
impulsiveness
in a
dimensional system
of
personality description. British Journal
of
Social
and
Clinical Psychology,
16,
57—68.
*Fagley,
N.
S.,
&
Miller,
P. M.
(1990).
The
effect
of
framing
on
choice:
Interactions with
risk-taking
propensity, cognitive style,
and
sex. Per-
sonality
and
Social Psychology Bulletin,
16,
496-510.
*Farrington,
D. P., &
Kidd,
R. F.
(1977).
Is
financial dishonesty
a
rational
decision? British Journal
of
Social
and
Clinical Psychology,
16,
139-
146.
*Finney,
P. D.
(1984).
Coparticipant
effects
on
subjects'
decisions
to
consent
for risk-involving
research. Journal
of
Social Psychology, 124,
35-41.
*Flaherty,
J.
F.,
&
Arenson,
S. J.
(1978).
A
test
of two
theories
in the
initial
process stage
of
coalition formation. Social Behavior
and
Personality,
6,
141-146.
*Flisher,
A.
J.,
&
Chalton,
D. O.
(1995). High-school
dropouts
in a
working
class South African community: Selected characteristics
and
risk-talcing
behavior.
Journal
of
Adolescence,
18,
105-121.
*Foersterling,
F.
(1980).
Sex
differences
in risk-taking:
Effects
of
subjec-
tive
and
objective probability
of
success. Personality
and
Social Psy-
chology
Bulletin,
6,
149-152.
*Freeman,
R.
C.,
Rodriguez,
G.
M.,
&
French,
J. F.
(1994).
A
comparison
of
male
and
female intravenous drug users'
risk
behaviors
for HIV
infection.
American Journal
of
Drug
and
Alcohol Abuse,
20,
129-157.
Furby,
L.,
&
Beyth-Marom,
R.
(1992). Risk taking
in
adolescence:
A
decision-making perspective. Developmental Review,
12,
1-44.
*Furnham,
A.,
&
Saipe,
J.
(1993). Personality correlates
of
convicted
drivers. Personality
and
Individual
Differences,
14,
329-336.
*Gallois,
C.,
Kashima,
Y.,
Terry,
D.,
McCamish,
M.,
Timmons,
P., &
Chauvin,
A.
(1992). Safe
and
unsafe
sexual intentions
and
behavior:
The
effects
of
norms
and
attitudes. Journal
of
Applied
Social
Psychology,
22,
1521-1545.
*Gibbons,
F.
X.,
&
Gerrard,
M.
(1995). Predicting young adults' health
risk
behavior.
Journal
of
Personality
and
Social Psychology,
69,
SOS-
SI?.
*Gibbons,
F.
X.,
Gerrard,
M.,
&
McCoy,
S. B.
(1995). Prototype percep-
tion predicts (lack
of)
pregnancy prevention. Personality
and
Social
Psychology
Bulletin,
21,
85-93.
*Ginsburg,
H.
J.,
&
Miller,
S. M.
(1982).
Sex
differences
in
children's
risk-taking
behavior.
Child
Development,
53,
426-428.
*Grupp,
S.
E.,
McCain,
M.
J.,
&
Schmitt,
R. L.
(1971). Marijuana
use in
a
small college:
A
midwest example. International Journal
of the
Addictions,
6,
463-485.
Halpern,
D. F.
(1992).
Sex
differences
in
cognitive abilities (2nd ed.).
Hillsdale,
NJ:
Erlbaum.

RISK TAKING
381
*Hayes,
C. S.
(1973). Risk taking
by
retarded
and
nonretarded
children.
Psychological
Reports,
32,
738.
Hedges,
L. V.
(1994).
Fixed
effects
models.
In H.
Cooper
& L. V.
Hedges
(Eds.),
The
handbook
of
research synthesis
(pp. 285-299).
New
York:
Russell Sage Foundation.
Hedges,
L.
V.,
&
Olkin,
I.
(1985).
Statistical methods
for
meta-analysis.
New
York: Academic
Press.
*Heilizer,
F., &
Cutter,
H. S. G.
(1971). Generality
and
correlates
of
risks
taking. Journal
of
General Psychology,
85,
259-283.
*Hudgens,
G.
A.,
&
Fatkin,
L. T.
(1985).
Sex
differences
in
risk-taking:
Repeated sessions
on a
computer simulation task. Journal
of
Psychol-
ogy,
119,
197-206.
*Ingersoll,
G. M.,
Grizzle,
K.,
Beiter,
M., &
Orr,
D. P.
(1993). Frequent
somatic complaints
and
psychosocial
risk in
adolescents. Journal
of
Early
Adolescence,
13,
67-78.
*Ingersoll,
G.
M.,
&
Orr,
D. P.
(1989). Behavioral
and
emotional
risk in
early
adolescents.
Journal
of
Early Adolescence,
9,
396-408.
Irwin,
C. E., &
Millstein,
S. G.
(1991). Correlates
and
predictors
of
risk-taking
behavior.
In L. P.
Lipsitt
& L. L.
Mitnick
(Eds.),
Self-
regulatory
behavior
and
risk-taking: Causes
and
consequences
(pp.
3-21).
Norwood,
NJ:
Ablex.
*Irwin,
F. W., &
Tolkmitt,
F.
(1968). Buying insurance
in an
urn-scheme
experiment. Psychonomic Science,
12,
287—289.
*Jackson,
T.
T.,
&
Gray,
M.
(1976). Field study
of risk-taking
behavior
of
automobile drivers. Perceptual
and
Motor Skills,
43,
471-474.
*Jamieson,
B. D.
(1969).
The
influences
of
birth order,
family
size,
and sex
differences
on risk-taking
behaviour. British Journal
of
Social
and
Clinical
Psychology,
8,
1-8.
*Jessor,
R., Van den
Bos,
I.,
Vanderryn,
J.,
Costa,
F.
M.,
&
Turbin,
M. S.
(1995).
Protective factors
in
adolescent problem behavior: Moderator
effects
and
developmental change. Developmental Psychology,
31,
923-
933.
Kahneman,
D.,
&
Tversky,
A.
(1979). Prospect theory:
An
analysis
of
decision under
risk.
Econometrica,
47,
263-291.
*Karabenick,
S.
A.,
&
Addy,
M. M.
(1979).
Locus
of
control
and sex
differences
in
skill
and
chance
risk-taking
conditions. Journal
of
Gen-
eral
Psychology,
100, 215-228.
*Kass,
N.
(1964). Risk
in
decision making
as a
function
of
age, sex,
and
probability
preference.
Child
Development,
35,
577-582.
*Kelling,
G.
W.,
Zirkes,
R.,
&
Myerowitz,
D.
(1976). Risk
as
value:
A
switch
of set
hypothesis. Psychological Reports,
38,
655-658.
*Kogan,
N.,
&
Dorros,
K.
(1978).
Sex
differences
in risk
taking
and its
attribution.
Sex
Roles,
4,
755-765.
*Kogan,
N.,
&
Wallach,
M. A.
(1964).
Risk-taking:
A
study
in
cognition
and
personality.
New
York:
Holt,
Rinehart,
&
Winston.
*Kohler,
M. P.
(1996). Risk-taking behavior:
A
cognitive approach.
Psy-
chological Reports,
78,
489-490.
*Kopfstein,
D.
(1973). Risk-taking
and
cognitive style.
Child
Develop-
ment,
44,
190-192.
*Kourilsky,
M.,
&
Campbell,
M.
(1984).
Sex
differences
in a
simulated
classroom economy: Children's beliefs about
entrepreneurship.
Sex
Roles,
10,
53-66.
*Kreitler,
S.,
&
Zigler,
E.
(1990).
Motivational
determinants
of
children's
probability learning. Journal
of
Genetic Psychology,
151, 301-316.
"Krishna,
K. P.
(1981).
Risk taking among Indian
adolescents.
Journal
of
Social Psychology,
114,
293-294.
*Lamm,
H.
(1979). Geschlecht
und
alter
als
einfluBfaktoren
der
risikobe-
reitschaft
bei
entscheidungen
fiir
die
eigene
und
eine
andere
person
[Sex
and
age
effects
in risk
proneness when making
decisions
for
oneself
and
another
person].
Zeischriftfiir
experimented
und
angewandte
psycholo-
gic,
26,
496-408.
*Lampert,
A.,
&
Yassour,
J.
(1992). Parental investment
and risk
taking
in
simulated
family
situations. Journal
of
Economic Psychology,
13,
499-
507.
*Leigh,
B.
C.,
Temple,
M.
T.,
&
Trocki,
K. F.
(1993).
The
sexual behavior
of US
adults: Results
from
a
national survey. American Journal
of
Public Health,
83,
1400-1408.
*Lettman,
D. S.
(1981). Relationships between
self-esteem,
risk-taking,
involvement,
certainty,
sex,
and
college
major
in
vocational choice.
Unpublished doctoral dissertation, University
of
Cincinnati.
*Levin,
I. P., &
Chapman,
D. P.
(1993). Risky decision making
and
allocation
of
resources
for
leukemia
and
AIDS programs. Health
Psy-
chology,
12,
110-117.
*Levin,
I. P.,
Snyder,
M.
A.,
&
Chapman,
D. P.
(1988).
The
interaction
of
experiential
and
situational
factors
and
gender
in a
simulated
risky
decision-making task. Journal
of
Psychology,
122, 173-181.
Lopes,
L. L.
(1987). Between
hope
and
fear:
The
psychology
of risk. In L.
Berkowitz
(Ed.),
Advances
in
experimental social psychology
(Vol.
20,
pp.
255-295).
New
York: Academic Press.
*Lupfer,
M.,
Jones,
M.,
&
Spaulding,
L.
(1971). Risk-taking
in
cooperative
and
competitive dyads.
Conflict
Resolution,
25,
387-394.
*Martinez,
J. C.
(1995). Risk taking
on
gender-typed tasks
following
an
assignment based
on
sex.
Journal
of
Social Psychology,
135,
573-579.
*Marruza,
V. R.
(1970).
An
investigation
of the
effects
of
strategy avail-
ability, bankroll,
and sex on risk-taking
behavior measured
in a
psycho-
metric context. Journal
of
Personality,
38,
146-160.
*McCormack,
A.
S.,
Anderton,
J.,
&
Barbieri,
T.
(1993).
Gender,
HIV
awareness,
and
prevention among college students. College Student
Journal,
27,
514-520.
*McDonald,
G. W.
(1976).
Sex,
religion,
and risk-taking
behavior
as
correlates
of
death anxiety. Omega,
7,
35-44.
*McGaffney,
T. N.
(1976). Motivational determinants
of
decision-making
in
a
triadic
coalition game. Journal
of
General Psychology,
94,
167-185.
*McKelvie,
S.
J.,
&
Schamer,
L. A.
(1988).
Effects
of
night, passengers,
and
sex on
driver behavior
at
stop signs. Journal
of
Social Psychology,
128,
585-590.
*Michaels,
T.
F.,
&
Getting,
E. R.
(1979).
The
informed consent dilemma:
An
empirical approach. Journal
of
Social Psychology,
109, 223-230.
*Miller,
A.
S.,
&
Hoffman,
J. P.
(1995).
Risk
and
religion:
An
explanation
of
gender differences
in
religiosity. Journal
for the
Scientific
Study
of
Religion,
34,
63-75.
*Miller,
D.
C.,
&
Byrnes,
J. P.
(1997).
The
role
of
contextual
and
personal
factors
in
children's
risk
taking. Developmental Psychology,
33,
814-
823.
*Miller,
K. E.
(1987). Adolescents' same-sex
and
opposite-sex peer rela-
tions:
Sex
differences
in
popularity, social skills,
and
perceived social
competence.
Unpublished doctoral dissertation, Purdue University.
*Miller,
P.
M.,
&
Fagley,
N. S.
(1991).
The
effects
of
framing, problem
variations,
and
providing
rationale
on
choices.
Personality
and
Social
Psychology
Bulletin,
5,
517-522.
*Mindock,
R.
(1972). Risk taking
as a
function
of an
individual's impres-
sion
of the
power situation. Psychological Reports,
31,
471-474.
*Minton,
H.
L.,
&
Miller,
A. G.
(1970).
Group
risk
taking
and
internal-
external control
of
group members. Psychological Reports,
26,
431-
436.
"Montgomery,
G.
T.,
&
Landers,
W. F.
(1974). Transmission
of risk-taking
through
modeling
at two age
levels.
Psychological Reports,
34,
1187-
1196.
*Moore,
D.
S.,
&
Erickson,
P. I.
(1985).
Age,
gender,
and
ethnic
differ-
ences
in
sexual
and
contraceptive knowledge, attitudes,
and
behaviors.
Family
and
Community Health,
8,
38—51.
*Moore,
S.
M.,
&
Rosenthal,
D. A.
(1991).
Condoms
and
coitus: Adoles-
cents'
attitudes
to
AIDS
and
safe
sex
behavior. Journal
of
Adoles-
cence,
14,
211-227.
*Moore,
S.
M.,
&
Rosenthal,
D. A.
(1992).
Australian
adolescents'
per-
ceptions
of
health-related
risks.
Journal
of
Adolescent Research,
7,
177-191.
*Morrongiello,
B.
A.,
&
Bradley,
M. D. M.
(1997).
Sibling
power:
influ-

382
BYRNES, MILLER,
AND
SCHAFER
ence
of
older siblings' persuasive appeals
on
younger siblings' judg-
ments about
risk
taking behaviours.
Injury
Prevention,
3,
23-28.
*Muldrow,
T. W., &
Bayton,
J. A.
(1979).
Men and
women executives
and
processes related
to
decision accuracy. Journal
of
Applied Psychol-
ogy,
2,
99-106.
*Neale,
M.
A.,
&
Bazerman,
M. H.
(1985).
The
effects
of
framing
and
negotiator
overconfidence
on
bargaining behaviors
and
outcomes. Acad-
emy
of
Management
Journal,
28,
34-49.
*Ottomanelli,
C.
(1993). Risk dimensions
of
HIV-tested
substance abusers.
International
Journal
of
the
Addictions,
28,
1487-1495.
*Parra,
M. J.
(1988). Risk seeking
and
risk
aversion
in the
social dating
situation. Unpublished doctoral dissertation, Oklahoma State University.
Payne,
J.
W.,
Bettman,
J. R., &
Johnson,
E. J.
(1993).
The
adaptive
decision-maker.
Cambridge, England: Cambridge University Press.
*Poppen,
P. J.
(1994). Adolescent contraceptive
use and
communication:
Changes over
a
decade. Adolescence,
29,
503—514.
*Poppen,
P. J.
(1995). Gender
and
patterns
of
sexual
risk
taking
in
college
students.
Sex
Roles,
32,
545-555.
*Potts,
R.,
Doppler,
M.,
&
Hernandez,
M.
(1994).
Effects
of
television
content
on
physical
risk-taking in
children. Journal
of
Experimental
Child
Psychology,
58,
321-331.
*Potts,
R.,
Martinez,
I. G., &
Dedmon,
A.
(1995). Childhood
risk
taking
and
injury:
Self-report
and
informant measures. Journal
of
Pediatric
Psychology,
20,
5-12.
*Reardon,
M. M.
(1981). Alternative curriculum
and risk
taking. Unpub-
lished doctoral dissertation, University
of
Connecticut.
*Reddon,
J.
R.,
Pope,
G.
A.,
Friel,
J. P., &
Sinha,
B. K.
(1996).
Leisure
motivation
in
relation
to
psychosocial
adjustment
and
personality
in
young
offender
and
high school samples. Journal
of
Clinical Psychol-
ogy,
52,
679-685.
*Roberts,
G. C.
(1975).
Sex and
achievement motivation effects
on risk
taking.
Research Quarterly,
46,
58-67.
Rosenthal,
R.
(1994). Parametric measures
of
effect
size.
In H.
Copper
&
L. V.
Hedges
(Eds.),
The
handbook
of
research synthesis (pp.
231-244).
New
York: Russell Sage Foundation.
*Rotheram-Borus,
M.
J.,
Meyer-Bahlburg,
H. F.
L.,
Koopman,
C.,
Rosario,
M.,
Exner,
T.
M.,
Henderson,
R.,
Matthieu,
M.,
&
Gruen,
R. S.
(1992).
Lifetime
sexual behaviors among runaway males
and
females. Journal
of
Sex
Research,
29,
15-29.
*Rothman,
A.
J.,
Salovey,
P.,
Antone,
C.,
Keough,
K.,
&
Martin,
C. D.
(1993).
The
influence
of
message framing
on
intentions
to
perform
health
behaviors. Journal
of
Experimental
Social Psychology,
29,
408-
433.
*Rowe,
K. E.
(1991). Adolescents' perceptions
of
family
cohesion
and
family
adaptability
and
their relationship
to
gender
and risk-taking
behaviors.
Unpublished doctoral dissertation, Oklahoma State Univer-
sity.
*Rudolph,
B. A.
(1996). Gender
differences
in
adolescent depression:
A
consequence
of
exposure
to
adverse
life
circumstances
and the
loss
of
parental support. Unpublished doctoral dissertation, University
of
Wisconsin-Madison.
*Rutte,
C.
G.,
Wilke,
H.
A.,
&
Messick,
D. M.
(1987).
The
effects
of
framing
social
dilemmas
as
give-some
or
take-some
games.
British
Journal
of
Social Psychology,
26,
103-108.
*Sadava,
S.
W.,
&
Forsyth,
R.
(1976). Decisions about drug use:
An
application
of the
choice-shifts paradigm. Psychological Reports,
38,
1119-1133.
Schafer,
W. D. (in
press).
An
overview
of
meta-analysis.
Measurement
and
Evaluation
in
Education
and
Development.
*Scherer,
R. F.
(1987).
A
social learning model explanation
for the
development
of
entrepreneurial characteristics
and
career selection.
Unpublished
doctoral dissertation, University
of
Mississippi.
*Schilling,
R.
F.,
El-Bassel,
N.,
Schinke,
S. P.,
Nichols,
S.,
Botvin,
G.
J.,
&
Orlandi,
M. A.
(1991). Sexual behavior, attitudes toward safer sex,
and
gender among
a
cohort
of 244
recovering
IV
drug users. Interna-
tional
Journal
of
the
Addictions,
26,
859-877.
*Schwartz,
H. S.
(1983). Relationships between
risk-taking
judgment
in the
decision-making
process,
self-esteem
and
sex-role orientation
of
adults.
Unpublished doctoral disseration,
Fordham
University.
*Seeborg,
I.,
Lafollette,
W.,
&
Belohlav,
J.
(1980).
An
exploratory analysis
of
effect
of sex on
shift
in
choices.
Psychological Reports,
46,
499-504.
*Shaw,
D.
S.,
Wagner,
E.
F.,
Arnett,
J.,
&
Aber,
M. S.
(1992).
The
factor
structure
of the
Reckless Behavior Questionnaire. Journal
of
Youth
and
Adolescence,
21,
305-315.
*Sheer,
V.
C.,
&
Cline,
R. J. W.
(1995).
Individual differences
in
sensation
seeking
and
sexual behavior: Implications
for
communication interven-
tion
for
HIV/AIDS prevention among college students. Health Commu-
nication,
7,
205-223.
*Sitkin,
S.
B.,
&
Weingart,
L. R.
(1995). Determinants
of risky
decision-
making
behavior:
A
test
of the
mediating role
of risk
perceptions
and
propensity.
Academy
of
Management
Journal,
38,
1573—1592.
*Sivak,
M.,
Soler,
J.,
&
Trankle,
U.
(1989). Cross-cultural differences
in
driver
risk-taking.
Accident Analysis
and
Prevention,
21,
363—369.
*Slakter,
M. J.
(1967). Risk taking
on
objective examinations. American
Educational
Research Journal,
4,
31-43.
*Slakter,
M. J.
(1969).
Generality
of risk
taking
on
objective examinations.
Educational
and
Psychological Measurement,
29,
115—128.
*Slakter,
M.
J.,
Koehler,
R.
A.,
&
Hampton,
S. H.
(1971). Sex, grade level,
and
risk
taking
on
objective examinations. Journal
of
Experimental
Education,
39,
65-69.
Slovic,
P.
(1964). Assessment
of risk
taking behavior. Psychological
Bulletin,
61,
220-233.
*Slovic,
P.
(1966).
Risk-taking
in
children:
Age and sex
differences. Child
Development,
37,
169-176.
Slovic,
P.,
Lichtenstein,
S.,
&
Fischhoff,
B.
(1988).
Decision-making.
In
R.
C.
Atkinson
et
al.
(Eds.),
Steven's handbook
of
experimental
psychol-
ogy.
Vol.
2:
Learning
and
cognition (pp.
673-738).
New
York: Wiley.
*Snyder,
S.
(1984). Risk taking behavior
of
normal
and
mentally handi-
capped
adolescents
and the
relationship
of
this behavior
to
learning.
Unpublished
doctoral dissertation, Indiana University.
*Sorrentino,
R.
M.,
Hewitt,
E.
C.,
&
Raso-Knott,
P. A.
(1992).
Risk-taking
in
games
of
chance
and
skill: Informational
and
affective influences
on
choice behavior. Journal
of
Personality
and
Social Psychology,
62,
522-533.
*Spears,
R.,
Abraham,
S. C.
S.,
Abrams,
D.,
&
Sheeran,
P.
(1992). Framing
in
terms
of
'high-risk
groups" versus "risky practices"
and
prognoses
of
HIV
infection. European Journal
of
Social Psychology,
22,
195-201.
*Speltz,
M.
L.,
Gonzales,
N.,
Sulzbacher,
S.,
&
Quan,
L.
(1990). Assess-
ment
of
injury
risk in
young
children:
A
preliminary study
of the
injury
behavior
checklist. Journal
of
Pediatric Psychology,
15,
373-383.
*Stanford,
M.
S.,
Greve,
K.
W.,
Boudreaux,
J.
K.,
Mathias,
C.
W.,
&
Brumbelow,
J. L.
(1996).
Impulsiveness
and risk-taking
behavior: Com-
parison
of
high school
and
college students using
the
Barratt
Impulsive-
ness Scale. Personality
and
Individual
Differences,
21,
1073-1075.
*Stapp,
B. C.
(1986).
An
investigation
of the
characteristics
of
midlife
career
change
and
midlife
transition. Unpublished doctoral dissertation,
Texas
Women's
University.
*Starrett,
R. H.
(1983).
The
conceptual commonality between impulsive-
ness
as a
personality trait
and as an ego
development stage. Personality
and
Individual
Differences,
4,
265-274.
*Steptoe,
A.,
Wardle,
J.,
Smith,
H.,
Kopp,
M.,
Arpad,
S.,
Vinck,}.,
&
Zotti,
A.
M.
(1995). Tobacco smoking
in
young adults
from
21
countries:
Associations with attitudes
and risk
awareness. Addiction,
90,
571-582.
*Struckman-Johnson,
D.,
&
Struckman-Johnson,
C.
(1996).
Can you say
condom?
It
makes
a
difference
in
fear-arousing
AIDS
prevention
public
service announcements. Journal
of
Applied Social Psychology,
26,
1068-1083.
*Tinsley,
B.
J.,
Holtgrave,
D.
R.,
Reise,
S. P.,
Erdley,
C.,
&
Cupp,
R. G.

RISK
TAKING
383
(1995).
Developmental status,
gender,
age,
and
self-reported
decision-
making influences
on
students'
risky and
preventative
health behaviors.
Health
Education Quarterly,
22,
244-259.
*Touhey,
J. C.
(1971). Risk-taking preferences
of
college students
in
Canada
and the
United States. Psychological Reports,
29,
512.
*Traub,
R. E., &
Hambleton,
R. K.
(1972).
The
effect
of
scoring instruc-
tions
and
degree
of
speededness
on the
validity
and
reliability
of
multiple-choice tests. Educational
and
Psychological
Measurement,
32,
737-758.
*Trocki,
K. F.
(1992). Patterns
of
sexuality
and risky
sexuality
in the
general population
of one
California county. Journal
of Sex Re-
search,
29,
85-94.
Tversky,
A.,
&
Kahneman,
D.
(1981).
The framing of
decisions
and the
psychology
of
choice. Science, 211,
453-458.
*Walesa,
C.
(1975). Children's approaches
to
chance-
and
skill-dependent
risk.
Polish
Psychological Bulletin,
6,
131-138.
*Ward,
D.
A.,
Seccombe,
K.,
&
Bendel,
R.
(1988).
Influenceability
of sex
differences
under conditions
of risk
taking. Journal
of
General
Psychol-
ogy,
115,
247-255.
*Warren,
R.
A.,
&
Simpson,
H. M.
(1980). Exposure
and
alcohol
as risk
factors
in the
fatal nighttime collisions
of men and
women drivers.
Journal
of
Safety
Research,
12,
151-156.
*Wayment,
H.
A.,
Newcomb,
M.
D.,
&
Hannemanm,
V. L.
(1993). Female
and
male intravenous drug users
not-in-treatment:
Are
they
at
differen-
tial
risk for
AIDS?
Sex
Roles,
28,
111-124.
*West,
G. B.,
Moskal,
P.
D.,
Dziuban,
C.
D.,
&
Rumbough,
L. P.
(1996).
Gender
and
marital differences
for risk
taking among undergraduates.
Psychological
Reports,
78,
315-320.
*White,
H.
R.,
&
Johnson,
V.
(1988). Risk taking
as a
predictor
of
adolescent sexual activity
and use of
contraception. Journal
of
Adoles-
cent
Research,
3,
317-331.
Wigfield,
A.,
&
Eccles,
J. S.
(1992).
The
development
of
achievement task
values:
A
theoretical analysis. Developmental Review,
12,
265-310.
*Williams,
A. F.
(1973). Personality
and
other characteristics associated
with
cigarette
smoking among young
teenagers.
Journal
of
Health
and
Social
Behavior,
14,
374-380.
Wilson,
M.,
&
Daly,
M.
(1985). Competitiveness,
risk-taking, and
vio-
lence:
The
young male syndrome. Ethnology
and
Sociobiology,
6, 59-
73.
*Wilson,
M.,
Daly,
M.,
Gordon,
S.,
&
Pratt,
A.
(1996).
Sex
differences
in
valuations
of the
environment? Population
and
Environment,
18,
143-
157.
*Wyatt,
G.
(1988).
Sex and
income differences
in
avoiding costs. Sociol-
ogy
and
Social Research,
72,
168-172.
*Yesmont,
G. A.
(1992).
The
relationship
of
assertiveness
to
college
students'
safer
sex
behaviors. Adolescence,
27,
253-272.
*Yinon,
Y.,
Jaffe,
Y., &
Feshbach,
S.
(1975).
Risky aggression
in
individ-
uals
and
groups. Journal
of
Personality
and
Social Psychology,
31,
808-815.
*Zeff,
L.
E.,
Fremgen,
B.,
&
Martinez,
J. C.
(1994).
Implications
of
gender
differences
for
managers. Psychological Reports,
74,
755—763.
Zuckerman,
M.
(1991).
Psychobiology
of
personality. Cambridge,
En-
gland: Cambridge University
Press.
*Zuckerman,
M.,
Ball,
S.
A.,
&
Black,
J.
(1990). Influences
of
sensation
seeking, gender,
risk
appraisal,
and
situational
motivation
on
smoking.
Addictive
Behaviors,
15,
209-220.
Received July
7,
1997
Revision received September
9,
1998
Accepted October
14,
1998
ORDER
FORM
Start
my
1999
subscription
to
Psychological
Bulletin!
ISSN: 0033-2909
$76.00,
APA
Member/Affiliate
$153.00,
Individual
Non-Member.
$315.00,
Institution
TOTAL
AMOUNT
ENCLOSED
$.
Subscription orders must
be
prepaid. (Subscriptions
are on
a
calendar basis only.) Allow
4-6
weeks
for
delivery
of the
first
issue. Call
for
international subscription rates.
SEND
THIS
ORDER
FORM
TO:
American
Psychological
Association
Subscriptions
750
First
Street,
NE
Washington,
DC
20002-4242
Or
call
(800)
374-2721,
fax
(202)
336-5568.
Send
me a
Free
Sample
Issue
Q
Q
Check
Enclosed
(make
payable
to
APA)
Charge
my:
Q
VISAQMasterCard
Q
American Express
Cardholder
Name
Card
No.
Exp.
date
Signature (Required
for
Charge)
Credit
Card
Billing
Address
_
State
Zip.
Daytime
Phone
SHIP
TO:
Name.
Address
Cirv
State
7,in
APA
Customer
#
TDD/TTf
(202)336-6123.
Email:
subscriptions@apa.org
GAD99_
PLEASE
DO NOT
REMOVE
- A
PHOTOCOPY
MAY
BE
USED