Explanatory Style as a Predictor of Productivity and Quitting Among Life Insurance Sales Agents

Abstract

Tested the prediction of the reformulated learned helplessness model, which claims that the tendency to explain bad events by internal, stable, and global causes potentiates quitting when bad events are encountered. Two studies were conducted, using a total of 197 life insurance agents as Ss. Explanatory style, as measured by the Attributional Style Questionnaire (ASQ), correlated with and predicted the performance of the Ss. In a cross-sectional study, Ss scoring in the top half of the ASQ sold 37% more insurance in their 1st 2 yrs of service than those scoring in the bottom half. In a prospective 1-yr study of newly hired agents, Ss who scored in the top half of the ASQ when hired remained in their job at twice the rate and sold more insurance than those scoring in the bottom half of the ASQ. These 2 studies support the claim that a pessimistic explanatory style leads to poor productivity and quitting when bad events are experienced, and they extend the usefulness of the ASQ to the workplace. (9 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)

Full text

Journal
of
Personality
and
Social
Psychology
1986,
Vol.
50, No. 4,
832-838
Copyright
1986
by the
American
Psychological
Association,
Inc.
0022-3514/86/$00,75
Explanatory
Style
as a
Predictor
of
Productivity
and
Quitting
Among
Life
Insurance
Sales
Agents
Martin
E. P.
Seligman
and
Peter
Schulman
University
of
Pennsylvania
The
reformulated
learned
helplessness
model
claims
that
the
tendency
to
explain
bad
events
by
internal,
stable,
and
global
causes
potentiates
quitting
when
bad
events
are
encountered.
We
tested
this
prediction
in
the
work
setting
with
individuals
who
frequently
experience
bad
events.
Explanatory
style,
as
measured
by the
Attributional
Style
Questionnaire
(ASQ),
correlated
with
and
predicted
the
perfor-
mance
of
life
insurance
sales
agents.
In
a
cross-sectional
study
of 94
experienced
agents,
individuals
scoring
in the top
half
of the ASQ
sold
37%
more
insurance
in
their
first 2
years
of
service
than
those
scoring
in the
bottom
half.
In a
prospective
1-year
study
of 103
newly
hired
agents,
individuals
who
scored
in the top
half
of the ASQ
when
hired
remained
in
their
job at
twice
the
rate
and
sold
more
insurance
than
those
scoring
in the
bottom
half
of the
ASQ.
These
two
studies
support
the
claim
that
a
pessimistic
explanatory
style
leads
to
poor
productivity
and
quitting
when
bad
events
are
experienced,
and
extend
the
usefulness
of the ASQ to the
workplace.
According
to the
reformulation
of the
learned
helplessness
model, individuals with
a
"pessimistic"
explanatory style
are
more
likely
to
display helplessness
deficits
when confronted with
a bad
event than individuals
with
an
"optimistic"
explanatory
style
(Abramson,
Seligman,
&
Teasdale,
1978; Seligman,
Abramson,
Semmel,
& von
Baeyer,
1979). Individuals
who ha-
bitually
construe
the
causes
of bad
events
as
internal, stable,
and
global
("it's
my
fault,
it's going
to
last
forever,
and
it's
going
to
undermine
everything
I
do") should, when they experience
bad
events,
be
more susceptible
to
helplessness
deficits
than those
with
the
opposite
style. Peterson
and
Seligman
(1984)
reviewed
12
studies that
confirm
this model
by finding
depressive
deficits
associated
with
a
pessimistic explanatory
style
in
students,
de-
pressed patients, prisoners,
and
children.
Here
we
report
two field
studies
of
this
model, using
a
theo-
retically
relevant
population,
life
insurance
sales
agents,
and in-
vestigate
a
central helplessness
deficit—quitting.
These studies
have
two
purposes: First, they test
the
Abramson
et
al.
(1978)
model,
in
which
the
pessimistic explanatory style
predisposes
giving
up, and the
rejections inherent
in
selling
life
insurance
trigger
giving
up
when this disposition
is
present.
The
interaction
of
the
pessimistic explanatory style
and of the
rejections, though
neither
necessary
nor
sufficient
conditions,
increases
the
likeli-
hood
of
helplessness deficits. This
is a
species
of a
diathesis-
stress
model,
in
which
the
diathesis, though probably
not
con-
stitutional,
is a
pessimistic explanatory style,
and the
stress
is
The
authors
thank
Dan
Oran,
Judy
Saltzberg,
and
Jack
Riley
for
their
help
at
various
stages
of
this
study.
We
also
thank
Al
Oberlander,
Richard
Calogero,
Susan
Keppler,
Angelina
Bhatia,
Charles
Wyckoff,
and
Robert
Weber
of the
Metropolitan
Life
Insurance
Company
for
their
generous
assistance,
and
John
Creedon
and
Pierre
Maurer
of the
Metropolitan
for
getting
it all
started.
Correspondence
concerning
this
article
should
be
addressed
to
Martin
E. P.
Seligman,
Department
of
Psychology,
University
of
Pennsylvania,
3815
Walnut
Street,
Philadelphia,
Pennsylvania
19104.
repeated
failures.
Second,
we
extend
the
test
of
learned help-
lessness
and
explanatory style
to
performance
in the
workplace.
Selling
life
insurance
is a job
particularly
suitable
for the in-
vestigation
of
learned helplessness
and
explanatory style. Sales
agents repeatedly encounter
failure,
rejection,
and
indifference
from
prospective clients. Consequently,
the
turnover rate among
life
insurance agents
is
very
high
(as are the
training
costs).
Studies
by the
Life
Insurance Marketing Research Association
(LIMRA,
1983)
have
found
that
78% of the
life
insurance agents hired
in
the
United States quit within
3
years
of
service.
We
predicted
that individuals with
an
optimistic explanatory style
will
weather
such
a
challenging
job
better.
In
these studies
we
measured explanatory style with
the At-
tributional Style Questionnaire (ASQ; Peterson
et
al.,
1982;
Se-
ligman
et
al.,
1979).
Helplessness deficits were
operationalized
by
two
objective performance
measures:
survival
and
produc-
tivity.
Survived
represents whether
the
agent
is
still working
or
has
quit
after
a
specified
period
of
time.
Productivity
is the
com-
mission earned
by the
agent,
calculated
as a fixed
percentage
of
the
revenues generated
from
the
sale
of a
life
insurance policy.
The
learned helplessness model (Seligman, 1975)
predicts
that
uncontrollable
failure
will
be
followed
by
lowered response ini-
tiation.
In the job of
selling insurance, this
translates
into
fewer
sales
attempts, less
persistence,
and the
ultimate
learned help-
lessness measure, quitting.
The
reformulated learned helplessness
model (Abramson,
et
al.,
1978)
specifies
which individuals
are
more
vulnerable
and
which
are
more
resistant
to
these deficits
when
failure
is
encountered. Individuals
with
a
vulnerable
ex-
planatory style
will
tend
to
explain
the
cause
of
their
failure
as
more internal, stable,
and
global. They
will
therefore blame
themselves
and
expect
failure
to
recur over
a
longer period
of
time
and in
more
situations.
Consequently, they
will
suffer
more
self-esteem
deficits,
and
response
initiation deficits will
be
more
sustained
in
time
and
across
situations than
for
individuals with
the
opposite explanatory style.
So, we
predicted
that
individuals
who
habitually explain
failure
with internal,
stable,
and
global
832

EXPLANATORY
STYLE
833
causes
would
initiate
fewer
sales attempts,
be
less persistent, pro-
duce less,
and
quit more
frequently
than those
with
a
more
op-
timistic
explanatory
style.
Study
1:
Cross-Sectional
Method
Subjects.
Eleven hundred
Attributional
Style Questionnaires,
along
with
postpaid
return envelopes,
were
distributed
to the
entire
sales
force
of the
Pennsylvania region
of the
Metropolitan
Life
Insurance Company.
A
letter
from
the
regional manager encouraging voluntary
participation
but
assuring
sales
agents that taking
it or not
would
in no way
aifect
their
job
status, accompanied
the
questionnaire.
One
hundred sixty-nine
questionnaires were returned
completed,
and
accurate quarterly
pro-
duction
data
(in
dollar
figures)
was
available
for 94 of
these agents
up
until
that time
in
their career.
The
company keeps accurate computerized
production
records
for the
purpose
of
compensating agents.
We
analyzed
the
synchronous
correlation
of
explanatory style
with
production
for
these
94
agents.
Is
this sample
of 94
representative
of the
1,100
agents
in the
Pennsyl-
vania
region? Because
the
return rate
was so
meager,
our
main concern
was
that
there might
be
systematic production
differences
between
the
respondents
and the
nonrespondents.
The
mean quarterly production
figures
were
slightly higher
for
the
respondents (2,620),
but not
significantly
so
from
the
mean
for the
entire region (2,500;
(test
p <
.45).
Questionnaires.
The
sales agents took
the
Attributional
Style
Ques-
tionnaire (ASQ; Peterson
et
al.,
1982;
Seligman
et
al.,
1979). This
self-
report instrument yields scores
for
explanatory style
for bad
events
and
good events
using
three causal
dimensions—internal
versus external, stable
versus
unstable,
and
global versus
specific
causes.
The
format
reflects
the
fact
that
we
wanted
to
assess
how
respondents
view
themselves along
a
continuum
for
each
of the
three dimensions.
We ask
subjects
to
generate
their
own
cause
for a
series
of
hypothetical events,
and
then
to
rate
that
cause
along 7-point
scales
corresponding
to the
intemality,
stability,
and
globality
dimensions.
The ASQ
does
not
create
or
constrain
the
causal
explanations provided
by the
subject,
but at the
same time
it
allows simple
and
objective quantification
of
responses
by
asking
the
subject
to
rate
the
internality,
stability,
and
globality
of the
causes.
The
questionnaire
is
group
or
individually administered,
and the
fol-
lowing
directions
appear
on the first
page
of the
booklet:
Please
try to
vividly
imagine yourself
in the
situations that
follow.
If
such
a
situation happened
to
you,
what would
you
feel
would
have
caused
it?
While events
may
have
many causes,
we
want
you
to
pick
only
one—the
major cause
if
this happened
to
you. Please
write
this cause
in the
blank provided
after
each event. Next,
we
want
you to
answer some questions about
the
cause.
To
summarize,
we
want
you to:
1.
Read each situation
and
vividly
imagine
it
happening
to
you.
2.
Decide what
you
feel
would
be the
major
cause
of the
situation
if
it
happened
to
you.
3.
Write
one
cause
in the
blank provided.
4.
Answer three questions about
the
cause.
5.
Go on to the
next situation.
Because
we are
interested
in
style—cross-situational
explanations—we
describe
12
different
hypothetical events. Half
are
good events (e.g.,
you
meet
a
friend
who
compliments
you on
your appearance),
and
half
are
bad
events (e.g.,
you go out on a
date
and
it
goes badly).
After
each
event
are
questions about
the
cause. First,
the
subject
is
asked
to
write down
the
one
major
cause
of the
event. Then
the
subject
is
asked
to
rate
the
cause along
the
three explanatory dimensions.
The
agents' scores
on the
Aptitude Index Battery
(AIB;
LIMRA,
1982),
now
called
the
Career
Profile,
were also available.
The AIB is a
self-report
questionnaire that asks
the
applicant undisguised questions
in six
major
areas:
self-assessment
of
job
relevant
skills
and
abilities, career expecta-
tions,
motivating goals, concerns about career, satisfaction with present
job,
and
potential clients. This
selection
instrument
is
widely
used
throughout
the
insurance industry. Scores
on the AIB
match
the
profile
of
the
applicant
to the
profiles
of
successful insurance agents,
and ap-
plicants
are
hired
if
they match such actuarial
profiles
well,
or if
they
match
them marginally
but do
well
in
interviews.
Dependent
measures.
We
used
three
composite
scores
derived
from
the
ASQ:
composite negative
attributional
style (CoNeg), which
is the
composite score
for the six
negative events, summing across
internal,
stable,
and
global dimensions; composite positive attributional style
(CoPos),
the
composite score
for
the six
positive events;
and a
total
score,
composite positive minus
composite
negative
(CPCN),
the
difference
score
between
CoPos
and
CoNeg.
Past research (Peterson
&
Seligman,
1984)
indicates
that
CoNeg
and
CPCN
are the
most valid empirical predictors
of
depressive deficits.
The AIB
yields
a
single composite
score
that
rep-
resents
the
applicant's likelihood
to
succeed
as an
insurance sales agent.
Productivity
is
measured
by the
agent's
quarterly commissions,
in
dol-
lars,
for
the first
eight quarters
(2
years)
of the
agent's employment. Because
we
used
a
cross-section
of
agents,
however,
not all
agents
had 2
years
of
service
for
which
we
could obtain production data. This measure
is
directly
proportional
to and
perfectly
correlated with
the
amount
of
insurance
sold
in
that period. Commissions
for
renewals
of
previously sold policies
are
excluded
from
the
productivity
figures,
because
it is
believed
that
the
first-time
sale
of a
policy requires more motivation than
the
renewal
of
a
currently held
policy.
Procedure.
The
agents took
the AIB
before
they were hired.
We ad-
ministered
the ASQ
after
they
were
hired
and had
accumulated experience
selling
insurance
for
Metropolitan ranging
from
several months
to
several
decades. Local managers distributed
the ASQ to the
agents,
to be
taken
at
their leisure.
The
questionnaire requires about
20 rain to
take.
The
agents
returned
it
directly
to our
research group,
not to
Metropolitan,
in
individual postpaid
preaddressed
envelopes.
Results
Do
agents
with
an
optimistic explanatory
style
sell more
in-
surance
than
agents with
a
pessimistic
style?
The
answer
is
yes.
Distribution
and
reliability.
The
composite
ASQ
scores
had
the
following
means
and
standard deviations: CoNeg
M=
12.00,
SD
=
2.42;
CoPos
M
=
17.43,
SD
=
1.83; CPCN
M
=
5.42,
SD
=
2.92. These statistics resemble those
of
undergraduate stu-
dent populations.
The
reliabilities,
as
estimated
by
Cronbach's
alpha
(1951),
were
modest:
.75 for
CoNeg
and .62 for
CoPos.
There
were
no
significant
differences
in ASQ
scores
for men
versus
women
(CoNeg
(
=
.26,
ns;
men did
slightly
better),
and
no
significant
differences
by
length
of
service
(CoNeg
r =
.02,
ns),
indicating
that
experienced sales agents
did not
have
a
better
explanatory
style than
new
agents.
The
ASQ and
productivity.
CoNeg correlated
significantly
with
the first 2
years
of
production
(r -
-.18,
p <
.07),
the first
year
alone
(r=—.l9,p<
.07)
and the
second year
of
production
(r
=
-.39,
p <
.01). Agents
who
scored
in the top
half
of the
CoNeg,
using
the
median
cutoff,
sold
37%
more insurance
in
their
first 2
years
of
service than agents
who
scored
in the
bottom
half
(;
=
2.19,
p <
.02). (The
/
and p
statistics
refer
to t
test
analyses
on the
difference
in
production means.) More selective
CoNeg
cutoffs
reveal more striking results.
Agents
who
scored
in the top
decile
of
CoNeg sold
88%
more insurance
in
their
first
2
years than those
who
scored
in the
bottom decile
10%
((
=
2.17,
p <
.03). Furthermore, CoNeg discriminated
the
high
and
low
producers
even
better
in
their second
year
of
service than
in

834
MARTIN
E. P.
SELIGMAN
AND
PETER
SCHULMAN
Table
1
Cross-Sectional
Study:
Top
Half
Versus
Bottom
Half
of
the
Sales
Force
on
CoNeg
and
Their
Productivity
CoNeg
score
Production average:
First
and
second year
Good
CoNeg
<.
11.83
Bad
CoNeg
a
12.00
Production:
First year
Good
CoNeg
s
11.83
Bad
CoNeg
£
12.00
Production: Second year
Good
CoNeg
=s
11.83
Bad
CoNeg
:>
12.00
n
40
55
39
54
15
24
Quarterly
production
average
I
3,105 2.19
2,270
2,762 1.37
2,142
6,242
1.96
2,716
P
.02
.01
.03
Superiority
in
production
37%
29%
130%
Note.
CoNeg
=
Composite
negative
attributional
style.
Good
CoNeg
=
Optimistic
attributional
style
for bad
events.
Bad
CoNeg
=
Pessimistic
attributional
style
for bad
events.
All t
test
results
are
one-tailed.
their
first
year. Agents
who
scored
in the top
half
of
CoNeg sold
29%
more insurance
in
their
first
year
(t
=
2.37,
p <
.01)
and
sold
130%
more insurance
in
their second
year
((
=
1.96,
p <
.03) than agents
who
scored
in the
bottom
half.
Tables
1 and 2
present
different
CoNeg
cutoffs
and the
associated production
differences.
CPCN
did not
significantly
discriminate
productivity
at the
median
division
but did
discriminate
by
quartile
and
decile.
Agents
scoring
in the top
half were
9%
more productive
in the
first
2
years than those
in the
bottom
half
(t =
.63,
ns).
The top
quartile
was 36%
more productive
in the first 2
years than
the
bottom quartile
(t =
1.72,
p <
.05),
and the top
decile
was 67%
more
productive than
the
bottom decile
(t -
1.77,
p <
.05).
CoPos
did not
correlate
significantly
with
production.
The
AIB
and
productivity.
The
industry wide test,
the
AIB,
did not
correlate
significantly
with
the first 2
years
of
production
(/•
=
.12, ns).
It is
important
to
note that
the
distribution
of the
agents'
AIB
scores
in our
sample
was
highly
skewed,
because
most applicants
with
low AIB
scores
are not
hired
and
therefore
did not find
their
way
into
our
pool. Some agents
with
marginal
AIB
scores
are
hired because Metropolitan
allows
its
branch
managers
to
hire people with marginal scores,
if
they
look
very
promising
in
interviews.
The
distribution
of ASQ
scores
was not
skewed
(the
population
had not
been
preselected
by
ASQ),
and
ASQ
scores
did not
correlate
significantly with
AIB
scores
(CoNeg
r =
.09,
ns;
CPCN
r =
-.09, ns). Each
questionnaire,
therefore,
appears
to
measure
different
characteristics
and
they
are not
redundant.
We
used
the
cutoff
score that
at the
time represented
the
"passing"
AIB
score
for the
total
Metropolitan applicant popu-
lation
(greater than
or
equal
to
11).
Agents
who
scored above
this
AIB figure
sold
32%
more insurance
in
their
first 2
years
than
those scoring below
it
((=
2.15,
p<
.02).
The
AIB,
however,
did
not
significantly
discriminate production
at the
median
cutoff
of
this sample (greater than
or
equal
to
13).
The two
tests together discriminated
the
more productive
from
the
less productive agents better than either test alone. Agents
who
scored
in
both
the
upper half (above
the
median)
of the
ASQ
and
above
the AIB
passing score sold
43%
more
in
their
first 2
years,
29%
more
in
their
first
year,
and
196% more
in
their
second year than agents
who
scored
in
both
the
lower half
of the
ASQ
and
below
the AIB
passing score (all
ps <
.05).
Discussion
Agents
who
scored
in the
optimistic half
of
explanatory style
sold
37%
more insurance than agents scoring
in the
pessimistic
half.
Agents
in the top
decile
sold
88%
more
insurance
than
Table
2
Top
Decile
Versus
Bottom
Decile
of
the
Sales
Force
on
CoNeg
and
Their
Productivity
CoNeg
score
Production average: First
and
second
year
Good
CoNeg
s£
9.17
Bad
CoNeg
a
15.17
Production:
First year
Good CoNeg
s
9.17
Bad
CoNeg
a
15.17
Production:
Second year
Good
CoNeg
^9.
17
Bad
CoNeg
2:
15.17
n
11
10
10
9
2
5
Quarterly production
average
3,526
1,874
3,087
1,962
15,320
2,076
t
2.17
2.24
1.69
P
.03
.02
ns
Superiority
in
production
88%
57%
638%
Note.
CoNeg
=
Composite
negative attributional style.
Good
CoNeg
=
Optimistic
attributional
style
for bad
events.
Bad
CoNeg
=
Pessimistic
attributional
style
for bad
events.
All t
test
results
are
one-tailed.

EXPLANATORY
STYLE
835
those
in the
bottom decile. Because these results
are
cross-sec-
tional, there
are
several
possible interpretations
of the
relation
between explanatory style
and
productivity.
First,
it
may
be
that
an
optimistic explanatory
style
predicts
and
precedes
successful
job
performance. Second,
and
less
interesting,
is
that success
may
alter explanatory
style
in the
optimistic direction. Third,
some
third variable
may
produce both
an
optimistic explanatory
style
and job
success. Study
2
tests between
the first and
second
interpretation
by
measuring explanatory
style
on
hiring
and
then
looking
at
sales performance over
the first
year
of
work.
Study
2:
Prospective
Method
Subjects.
One
hundred
four
sales agents
were
hired
by
Metropolitan
in
the
Pennsylvania region
in the
spring
of
1983.
At the
outset
of a
training
course
and
after
hiring,
but
before
any
sales experience
with
Metropolitan,
they
took
the
ASQ.
There
was
100% compliance, even though subjects
were
informed
that
the
test
results would
in no way
affect
their
job
status.
Procedure.
Subjects
took
the
ASQ
as
well
as the
AIB.
We
then obtained
the
quarterly commissions (which
are
directly proportional
to the
amount
of
insurance sold)
for the first
four
quarters
(1
year) that
followed
for 103
of
the
104
new
agents.
In the first
study,
we
could
not
collect information
on
the
theoretically most central
variable—quitting—because
only agents
still working took
the
ASQ.
In
Study
2, we
collected
information
on who
still
remained with Metropolitan
at the end of
their
first
year.
Results
Explanatory
style
significantly
predicted
first
year
survival
as
well
as
productivity
for the
second half
of the
year.
It did not
significantly
predict productivity
for the first
half
of the
year.
Distribution
and
reliability.
The
means
and
standard devia-
tions
of the
three
ASQ
composite scores were
as
follows:
CPCN
M
=
6.24,
SD
=
2.56; CoNeg
M
=
11.52,
SD =
2.10; CoPos
M
=
17.77,
SD =
1.66. There
were
no
significant
differences
in
CPCN
score
by sex (t = .
18,
ns;
males
did
slightly better)
or
race
((
=
.48,
ns;
minorities
did
slightly better), although
the
sample
size
of
minorities
is too
small
to be
conclusive. Agents
who had
experience
selling
insurance prior
to
their employment
with
Metropolitan
had
slightly
better CPCN scores than those without
prior experience
(t =
1.62,
p <
.05).
The
reliabilities
of the
com-
posites,
as
computed
by
Cronbach's
(1951)
alpha,
were
again
modest:
.71
for
CoPos
and .66 for
CoNeg.
We
present
the
results only
for the
CPCN measure,
the one
score that takes
all the
test responses into account. Results
for
CoPos
and
CoNeg were each moderately
significant
predictors,
but
highly
significant
when
combined
to
form
the
CPCN score.
ASQ
predictions:
Survival.
Do
agents with
an
optimistic
ex-
planatory
style survive
longer
than those with
a
pessimistic
ex-
planatory
style? Agents
who
scored
in the
optimistic
half
of
CPCN
survived
at
twice
the
rate
as
agents
who
scored
in the
pessimistic
half.
There
were
42
survivors
and 59
dropouts
in the
sample
of
101
sales agents
for
whom
we
have
status
information.
Of the
42
survivors,
67% (n = 28)
scored
in the top
half
of
CPCN,
33%
(«
= 14) in the
bottom
half,
x2U,
N = 42) =
6.63,
p <
.005.
Quartile
comparisons produced
an
even more accurate prog-
nosis.
The
upper
quartile
of
CPCN
survived
at
almost
3
times
the
rate
of the
lower quartile.
Of the 19
survivors
who
scored
in
either
the
upper
or the
lower quartile
of
CPCN,
74% (n = 14)
scored
in the
upper
quartile,
26% (n = 5) in the
lower quartile,
X2(l,
N = 19)
=
8.37,
p <
.002. Table
3
presents
the
survival
rates
as a
function
of the ASQ and AIB
cutoffs.
The AIB
also predicted survival
significantly
at its
median
cutoff,
as
shown
in
Table
3. Of the
41
survivors
for
whom
AIB
scores
were
available,
68%
scored
in the
upper
57% of AIB and
32%
scored
in the
lower 43%,
x2(U
N = 41) =
4.96,
p <
.01.
(The
distribution
of
AIB
scores
did not
yield
an
exact median.)
The
comparison between upper
versus
lower
quartile
on the AIB
revealed
no
significant
difference
in
survival
rate.
The ASQ and AIB
together
predicted
survival
better than either
test alone.
Of the 25
survivors
who
scored
in
either
the top
half
of
both tests
or in the
bottom half
of
both
tests,
80% (n = 20)
of
these
survivors scored
in the top
half
of
both
and 20% (n =
5)
scored
in the
bottom
half
of
both,
X20,
AT
= 25)
=
9.56,
p <
.001.
Thus
the
survival rate
was 4
times
as
high
for
those
who
scored
well
on
both
the ASQ and AIB as for
those
who
scored
poorly
on
both. Regression analyses suggested that
the
prediction
of
survival
was due
mainly
to the
ASQ,
not to the
AIB. CPCN
significantly
predicted
survival
(F
=
5.72,
p <
.02)
controlling
for
AIB,
but AIB did not
significantly
predict survival
(F =
.51,
ns)
controlling
for
ASQ.
ASQ
predictions:
Productivity.
Do
agents
who
scored
well
on
the ASQ
sell
more insurance than those
who
scored poorly? There
are
two
ways
of
measuring production.
One is to
include pro-
duction
for the
agents
who
have
dropped
out by
entering
a
zero
Table
3
Prospective
Study
Survival
Rates
CPCN scorePercent
survivorsPercent
dropouts
x2
p
Distribution
of
survivors
by top
half versus
bottom
half
of
CPCN"
Good
CPCN
2.6.33
67
(28/42)
41
(24/59) 6.63
.005
Bad
CPCN
s.
6.17
33(14/42)
59(35/59)
Total
100%
Distribution
of
survivors
by top
quartile
versus
bottom
quartile
of
CPCN
Good
CPCN
a;
8.17
74(14/19)
31(9/29)
8.37 .002
Bad
CPCN
s
4.00
26
(5/19)
69
(20/29)
Total
100%100%
Distribution
of
survivors
by top 57%
versus bottom
43% of AIB
Good
AIB
;>
13
68(28/41)
46(26/57) 4.96
.01
Bad
AIB
<.
12
32(13/41) 54(31/57)
Total
100%100%
Distribution
of
survivors
by top 22%
versus bottom
22% of AIB
Good
AIB
fc 16
53(10/19)
46(11/24)
.20 ns
Bad
AIB
^10
47(9/19)
54(13/24)
Total 100% 100%
Note.
Numbers
in
parentheses
represent
the
number
of
survivors
(or
dropouts) falling within
the
test
score range
out of the
total
number
of
survivors
(or
dropouts).
Good
CPCN
=
Optimistic
attributional
style
for
good
and bad
events.
Bad
CPCN
=
Pessimistic
attributional
style
for
good
and bad
events.
AIB =
Aptitude Index Battery.
a
All
chi-square
results
are
one-tailed.

836
MARTIN
E. P.
SELIGMAN
AND
PETER
SCHULMAN
Table
4
Prospective
Study:
Top
Half
Versus
Bottom
Half
of
Sales
Force
on
CPCN
and
Their
Productivity
(Excluding
Dropouts'
Production)
CPCN
score
Production:
First
year
Good
CPCN
£:
6.33
Bad
CPCN
<,
6.17
Production:
First
6
months
Good
CPCN
&
6.33
Bad
CPCN
s
6.
17
Production:
Second
6
months
Good
CPCN
&
6.33
Bad
CPCN
S6.17
n
47
45
47
45
35
33
Quarterly
production
average
2,268
1,993
2,295
2,109
2,617
2,096
t P
1.06
ns
.65
ns
1.55
.06
Superiority
in
production
14%
9%
25%
Note.
CPCN
=
Composite
positive
minus
composite
negative
score.
Good
CPCN
=
Optimistic
attributional
style
for
good
and bad
events.
CPCN
=
Pessimistic
attributional
style
for
good
and bad
events.
for
each quarter
of
production
after
the
individual drops out.
This
measures
the
total economic worth
of the
individual
to the
company
in the first
year,
but
confounds
the
effect
of
quitting
with
the
production
of
those still working.
The
second measure
of
production removes
the
dropouts
from
the
analysis
as
they
drop out. Both measures similarly showed
that
an
optimistic
ASQ
score predicted better productivity
in the
second half
of
the
year,
but not for the first
half
of the
year.
CPCN
correlated moderately with
the
second
6
months
of
productivity
when
dropouts
were
included
(r = .
17,
p <
.09)
and
significantly
when
dropouts
were
excluded
(r =
.27,
p <
.03).
CPCN
did not
correlate
significantly
with
the first 6
months
of
production. Agents
who
scored
in the top
half
of
CPCN sold
35%
more
life
insurance
in
their
second
6
months than
those
who
scored
in the
bottom
half
when
zeroes were entered
for
dropouts'
production
(/
=
1.50,
p <
.07); when dropouts were
omitted,
the top
half
of
CPCN sold
25%
more than
the
bottom
half
(t =
1.55,
p <
.06).
In
accordance with
the findings of the
cross-sectional study, more stringent
ASQ
cutoffs
yielded
larger
productivity
differences.
Agents
who
scored
in the top
quartile
of
CPCN sold
73%
more
in the
second
6
months than those
in
the
bottom
quartile
when
zeroes were entered
for
dropouts'
pro-
duction
(t =
1.80,
p <
.04);
when
dropouts
were
excluded,
the
top
quartile
of
CPCN sold
57%
more than
the
bottom quartile
(t
=
2.05,
p <
.02).
The
rationale behind analyzing
the
second
6
months separately
from
the first 6
months
is
that
it is not
until
after
the first
few
months,
when
agents
are
still undergoing train-
ing
to
acquire
the
specialized sales skills
and
knowledge,
that
differences
in
motivation should account
for the
differences
in
productivity.
When
the
entire
first
year
production
was
examined, agents
in
the top
half
of
CPCN sold
20%
more than those
in the
bottom
half,
with
zeroes entered
for
dropouts'
production
(/ =
1.18,
ns);
when
dropouts
were
excluded,
the top
half
sold
14%
more than
the
bottom
half
(t =
1.06, ns).
The top
quartile sold
50%
more
in
the first
year than
did the
bottom quartile
when
zeroes were
included
(t =
1.75,
p <
.04)
and
sold
40%
more when dropouts
were
excluded
(I
=
1.92,
p <
.03).
See
Tables
4 and 5 for
details.
The
Aptitude Index Battery
did not
significantly
correlate
with
first
year
production
when
dropouts were included
(r = .
15,
p <
.14)
but did
when
dropouts were excluded
(r =
.19,
p <
.07).
AIB
predictions
were
significant
at the
median split,
but not at
more selective
cutoffs.
For the first 6
months, agents
in the top
half
of AIB
sold
34%
more than those
in the
bottom half when
zeroes
were
entered
for
dropouts'
production
(t -
2.03,
p <
.02),
and
sold
38%
more
when
dropouts
were
omitted
((
=
2.53,
p <
.007).
For the
second
6
months, agents
who
scored
in the top
half
of the AIB
sold
51%
more than those
in the
bottom
half
with
zeroes included
(t =
1.95,
p <
.03)
and 37%
more
when
dropouts
were
excluded
(l
=
2.14,
p <
.02).
Table
5
Prospective
Study:
Top
Quartile
Versus
Bottom
Quartile
of
Sales
Force
on
CPCN
and
Their
Productivity
(Excluding
Dropouts'
Production)
CPCN
score
Production:
First
year
Good
CPCN
;>
8.
17
Bad
CPCN
<:
4.00
Production:
First
6
months
Good
CPCN
;>
8.17
Bad
CPCN
<;
4.00
Production:
Second
6
months
Good
CPCN
2
8.
17
Bad
CPCN
<;
4.00
n
21
23
21
23
16
17
Quarterly
production
average
2,689
1,915
2,659
2,004
3,024
1,929
t
1.92
1.53
2.05
P
.03
.07
.02
Superiority
in
production
40%
33%
57%
Note.
CPCN
=
Composite
positive
minus
composite
negative
score.
Good
CPCN
=
Optimistic
attributional
style
for
good
and bad
events.
Bad
CPCN
-
Pessimistic
attributional
style
for
good
and bad
events.

EXPLANATORY
STYLE
837
Table
6
Prospective
Study:
Productivity
of
Sales
Force
in Top
Half
of
Both
Tests
Versus
Bottom
Half
of
Both
Tests
(Excluding
Dropouts'
Production)
Scores
Production:
First
year
Good
CPCN
;>
6.33
and
Good
AIB
2
13
Bad
CPCN
=£
6.
17
and
Bad
AIB
s
12
Production: First
6
months
Good
CPCN
a
6.33
and
Good
AIB
s=
13
Bad
CPCN
s
6.
17
and
Bad
AIB
£ 12
Production: Second
6
months
Good
CPCN
a
6.33
and
Good
AIB
a 13
Bad
CPCN
<;
6.17
and
Bad
AIB
s
12
„
30
22
30
22
23
16
Quarterly
production
average
2,500
1,598
2,558
1,754
2,803
1,607
Superiority
in
(
p
production
2.49 .008
56%
1.94
.03 46%
3.04 .002
75%
Note.
CPCN
=
Composite
positive
minus
composite
negative
score;
AIB
=
Aptitude
Index Battery.
Good
CPCN
=
Optimistic
attributional
style
for
good
and bad
events.
Bad
CPCN
=
Pessimistic attributional style
for
good
and bad
events.
Both
tests
together
predicted productivity better
than
either
test alone.
For the first
half
of the
year,
agents
who
scored
in the
top
half
of
both
tests
sold
37%
more insurance than agents
who
scored
in
the
bottom half
of
both tests
with
zeroes
entered
(I
=
1.57,
p <
.06),
and
sold
46%
more when
dropouts'
production
was
omitted
(t =
1.94,
p <
.03).
For the
second
half
of the
year,
agents
who
scored
in the top
half
of
both tests sold
99%
more
than
did the
agents
who
scored
in the
bottom
half
of
both tests
with
zeroes entered
(t
=
2.40,
p <
.01);
75%
more
when
dropouts'
production
was
omitted
(/
=
3.04,
p <
.002).
See
Table
6 for
details.
Regression analyses further
indicated
that
both
tests
together
predicted second
6
months' productivity more
significantly
than
either
test alone.
By
omitting dropouts, CPCN
significantly
pre-
dicted
second
6
months'
productivity
(F =
7.03,
p <
.01)
con-
trolling
for
AIB,
AIB
significantly
predicted productivity
(F =
3.93,
p <
.05) controlling
for
CPCN,
and the
two-variable model
predicted productivity even more
significantly
(F =
6.12,
p <
.004).
Regression
analysts
also
revealed
that
the two
tests
together
were
more significant predictors
of first
year
production than
either
test alone
when
dropouts'
production
was
included
(F =
2.47,
p <
.09)
and
when
it was
excluded
(F =
3.79,
p <
.03).
General Discussion
We
undertook
two field
studies
of an
occupation
in
which
frequent
failure
is
inevitably
met to
test whether explanatory
style
predicts work productivity
and
quilting.
So
challenging
are
these experiences
that
the
dropout rate
for our
sample
was 58%
in
the first
year.
Four
major
findings
emerged
from
the
prospective study. First,
agents
who had an
optimistic explanatory style,
as
measured
by
the
ASQ,
survived
at a
significantly higher
rate
than agents with
a
more pessimistic explanatory style. Second, agents
with
an
optimistic
style
sold more insurance than agents
with
a
pessimistic
style.
Third, explanatory style predicted survival
and
production
as
well,
and
nonredundantly,
as the
traditional industry
test.
Fourth,
both tests together predicted
survival
and
productivity
better than either test alone.
These
results
follow
from
the
reformulated helplessness model
of
depression
(Abramson
et
al.,
1978). Individuals
who
believe
that
bad
events
are
internally, stably,
and
globally
caused
(and
conversely
for
good events)
will
be
less persistent
after
failure
than
those
with
the
opposite explanatory style.
The
theoretical
significance
of
these
findings is
that
they
support
the
reformu-
lation's
claim that
a bad
explanatory
style
predisposes
to
poor
performance,
and
poor performance
is
then triggered
by
failure
in
those individuals
with
the
predisposing style.
The
interaction
of
the two
components
increases
the
likelihood
of
helplessness
deficits,
here
operationalized
by
quitting
and.
poor
productivity.
These results suggest that
a
depressogenic explanatory style pre-
dicts
performance
deficits
in a
work
setting,
beyond
the
clinical
syndrome
of
depression,
wherein
it has
most
often
been
tested.
These
findings do not
rule
out the
likelihood
of
bidirectional
effects
of
bad
explanatory style
and
failure(e.g.,
Bandura,
1978).
On
this
view,
a bad
explanatory style
leads
to
more
failure,
and
failure
may
also cause
a
deterioration
in
explanatory style. Con-
versely,
a
good style results
in
more success,
and
success
may
enhance
explanatory
style.
So, for
example,
an
agent with
a
good
explanatory
style
might persist more
and
make more sales.
Eventually,
such
an
excess
of
sales will make
him or her
more
optimistic
and
feedback
to an
even better explanatory
style.
Such
bidirectional
effects
have been found previously
in
studies
of the
mutual
effects
of
explanatory style
and
depression (see Peterson
&
Seligman,
1984,
for a
review)
and are
likely
to
exist with
job
productivity
and
explanatory
style.
The
present study
is not fine
grained
enough
to
measure such reciprocal
effects,
but a
design
in
which
waves
of
measurement
of
explanatory style
and
pro-
ductivity
are
taken across time should illuminate
the
issue.
One
practical
implication
of
these
findings
lies
in the
possibility
of
identifying,
in
advance, individuals
who are
particularly suit-
able
or
unsuitable
for
work that entails
frequent
failure
or re-
jection. Matching
the
right explanatory
style
"profile"
to the
work
requirements should
be
useful
both
to the
individual
and

838
MARTIN
E. P.
SELIGMAN
AND
PETER
SCHULMAN
the
organization. Steering vulnerable individuals
away
from
po-
sitions that
are
characteristically
fraught
with
adversity
or en-
gaging
them
in
remedial
attributional
training would also
be a
useful
service.
The
fact
that explanatory style
significantly
cor-
relates with
and
predicts
job
performance
suggests
that
the
mea-
surement
of
explanatory
style
could make
the
process
of
per-
sonnel selection more accurate.
Because
we
used
a
natural
job
setting,
we
forfeited
some
of
the
control
we
have
in
laboratory studies
of
helplessness
and
explanatory
style.
Specifically,
in the
laboratory
we can
control
the
timing
and
amount
of
failure.
In the
work
setting,
we
cannot
control
or
easily
measure
day-to-day
failures
and
rejections.
By
measuring
such naturally occurring
failure
along with explan-
atory
style,
future
research
may
gain
a
clearer picture
of the
process
of
giving
up in the
workplace.
References
Abramson,
L. Y.,
Seligman,
M. E. P., &
Teasdale,
i.
(1978).
Learned
helplessness
in
humans:
Critique
and
reformulation. Journal
of
Ab-
normal
Psychology,
87,
32-48.
Bandura,
A.
(1978).
The
self
system
in reciprocal
determinism.
American
Psychologist.
33,
344-358.
Cronbach,
L. J.
(1951). Coefficient
alpha
and the
internal
structure
of
tests.
Psychometrika,
16,
297-334.
Life
Insurance
Marketing
Research
Association
(1982).
Aptitude Index
Battery
(AIB).
Hartford,
CT:
Author.
Life
Insurance
Marketing
Research
Association
(1983).
The
Manpower
and
Production
Survey.
Hartford,
CT:
Author.
Peterson,
C,
&
Seligman,
M. E. P.
(1984).
Causal
explanations
as a
risk
factor
for
depression:
Theory
and
evidence.
Psychological
Review,
91,
347-374.
Peterson,
C.,
Semmel,
A.,
von
Baeyer,
C.,
Abramson,
L.
Y.,
Metalsky,
G. I, &
Seligman,
M. E. P.
(1982).
The
attributional
style
questionnaire.
Cognitive
Therapy
and
Research, 1982,
6,
287-300.
Seligman,
M. E. P.
(1975).
Helplessness:
On
Depression, Development,
and
Death.
San
Francisco:
W. H.
Freeman.
Seligman,
M. E. P.,
Abramson,
L.
Y.,
Semmel,
A.,
& von
Baeyer,
C.
(1979).
Depressive
attributional
style.
Journal
of
Abnormal
Psychology,
88,
242-247.
Received
May 2,
1985
Revision
received
October
7,
1985
•