The Effect of Return Policy Leniency on Consumer Purchase and Return Decisions: A Meta-analytic Review

Abstract

While most retail stores offer return policies, some offer more lenient return policies than others. The inherent belief is that lenient return policies are more likely to lead to purchases than to encourage returns. Examining prior research we find that return policy leniency has been characterized in terms of five different dimensions: time, money, effort, scope, and exchange. We conduct a meta-analysis of 21 papers examining the effect of leniency on purchase and return decisions, and demonstrate that overall, leniency increases purchase more than return. Further, we show the return policy factors that influence purchase (money and effort leniency increase purchase) are different from the return policy factors that influence returns (scope leniency increases returns while time and exchange leniency reduces returns).

Full text

Journal
of
Retailing
92
(2,
2016)
226–235
Research
Note
The
Effect
of
Return
Policy
Leniency
on
Consumer
Purchase
and
Return
Decisions:
A
Meta-analytic
Review
Narayan
Janakiraman a,∗,
Holly
A.
Syrdal b,1,
Ryan
Freling c,2
aCollege
of
Business,
University
of
Texas
at
Arlington,
701
S.
West
Street,
Suite
201,
Arlington,
TX
76019,
USA
bCollege
of
Business,
University
of
Texas
at
Arlington,
701
S.
West
Street,
Suite
223,
Arlington,
TX
76019,
USA
cMarketing
Department,
The
Naveen
Jindal
School
of
Management,
The
University
of
Texas
at
Dallas,
SM
32,
800
West
Campbell
Road,
Richardson,
TX
75080-3021,
USA
Available
online
28
November
2015
Abstract
While
most
retail
stores
offer
return
policies,
some
offer
more
lenient
return
policies
than
others.
The
inherent
belief
is
that
lenient
return
policies
are
more
likely
to
lead
to
purchases
than
to
encourage
returns.
Examining
prior
research
we
find
that
return
policy
leniency
has
been
characterized
in
terms
of
five
different
dimensions:
time,
money,
effort,
scope,
and
exchange.
We
conduct
a
meta-analysis
of
21
papers
examining
the
effect
of
leniency
on
purchase
and
return
decisions,
and
demonstrate
that
overall,
leniency
increases
purchase
more
than
return.
Further,
we
show
the
return
policy
factors
that
influence
purchase
(money
and
effort
leniency
increase
purchase)
are
different
from
the
return
policy
factors
that
influence
returns
(scope
leniency
increases
returns
while
time
and
exchange
leniency
reduces
returns).
©
2015
New
York
University.
Published
by
Elsevier
Inc.
All
rights
reserved.
Keywords:
Product
returns;
Money
back
guarantee;
Meta-analysis;
Return
policy;
Post-purchase
decision
making
Return
policies
are
a
consumer
risk
reliever
often
used
by
retailers
(e.g.,
Greatorex
and
Mitchell
1994)
to
increase
con-
sumer
demand.
However
higher
demand
also
likely
leads
to
higher
product
returns.
Product
returns
in
2014
totaled
about
$280
million
across
all
U.S.
retailers—a
figure
that
exceeds
the
annual
sales
revenue
of
all
but
the
top
ranked
retailer.3Returns
can
be
prohibitively
expensive
for
retailers
due
to
the
high
processing
costs
and
low
salvage
values
associated
with
returned
merchandise.
In
fact,
a
recent
research
anecdote
suggests
that
retailers
with
product
return
rates
in
excess
of
20
percent
often
generate
zero
operating
profit.4
∗Corresponding
author.
Tel.:
+1
817
272
2885;
fax:
+1
817
272
2854.
E-mail
addresses:
janakira@uta.edu
(N.
Janakiraman),
hsyrdal@uta.edu
(H.A.
Syrdal),
ref081000@utdallas.edu
(R.
Freling).
1Tel.:
+1
817
272
2885;
fax:
+1
817
272
2854.
2Tel.:
+1
972
679
0803.
3http://www.theretailequation.com/Retailers/images/public/pdfs/industry
reports/ir
2014
nrf
retail
returns
survey.pdf.
4https://www.theretailequation.com/Retailers/images/public/pdfs/white
papers/wp
TRE4013
WhitePaper
ReturnFraud101
Feb2013.pdf.
Despite
the
cost
and
prevalence
of
returns,
most
retailers
offer
a
return
policy
hoping
that
the
positive
effect
on
demand
will
more
than
offset
the
negative
effect
on
returns.
In
a
survey
of
133
stores
in
California
conducted
by
Davis,
Hagerty,
and
Gerstner
(1998),
almost
all
of
the
stores
had
a
return
policy
irrespective
of
whether
they
were
department
stores,
specialty
stores,
or
single
outlet
stores.
However,
the
policies
differed
in
that
some
stores
offered
more
lenient
return
policies
compared
with
others.
In
this
regard,
the
first
question
we
seek
to
address
in
the
current
research
is:
do
lenient
return
policies
increase
product
purchase
more
than
product
returns?
Extant
scholarly
research
offers
no
clear-cut
answers.
In
fact,
research
exploring
the
effect
of
lenient
return
policies
on
product
purchase
and
subsequent
returns
is
inconclusive.
To
illustrate,
while
Petersen
and
Kumar
(2010)
find
that
lenient
return
policies
lead
to
both
higher
purchase
and
higher
returns,
Wood
(2001)
and
Wang
(2009)
find
that
lenient
return
policies
increase
pur-
chase
without
increasing
returns.
To
provide
a
more
definitive
answer
to
this
question,
in
the
current
research,
we
conduct
a
meta-analysis
by
synthesizing
data
from
original
studies
that
examine
how
return
policy
leniency
affects
two
important
down-
stream
outcomes:
(1)
purchase
proclivity
(i.e.,
purchase
attitudes
http://dx.doi.org/10.1016/j.jretai.2015.11.002
0022-4359/©
2015
New
York
University.
Published
by
Elsevier
Inc.
All
rights
reserved.

N.
Janakiraman
et
al.
/
Journal
of
Retailing
92
(2,
2016)
226–235
227
or
behaviors)
and
(2)
return
proclivity
(i.e.,
consumer
intentions
or
decisions
to
keep
or
return
products).
We
show,
across
studies
and
on
an
overall
basis,
that
while
return
leniency
increases
both
purchase
and
return,
the
effect
on
purchase
proclivity
is
larger
than
the
effect
on
return
proclivity.
In
addition
to
the
basic
question
of
whether
return
policies
increase
purchase
more
than
returns,
a
second
question
we
exam-
ine
is:
what
type
of
return
policies
have
a
greater
impact
on
purchase,
and
what
type
have
a
greater
impact
on
returns?
Extant
research
offers
little
guidance
on
this
topic.
Based
on
a
research
synthesis
of
prior
literature,
we
propose
that
return
policies
can
be
classified
as
being
lenient
or
restrictive
along
five
dimensions:
time
leniency
(e.g.,
60
day
vs.
30
day
return
policy),
monetary
leniency
(e.g.,
offering
100
percent
money
back
vs.
80
percent
money
back),
effort
leniency
(e.g.,
no
forms
required
vs.
forms
required),
scope
leniency
(e.g.,
accepting
returns
on
sale
items
vs.
not),
and
exchange
leniency
(e.g.,
cash
back
vs.
store
credit).
We
code
the
effects
of
return
policy
leniency
reported
in
the
studies
included
in
our
meta-analysis
as
varying
along
these
five
return
policy
leniency
dimensions.
To
account
for
the
heterogeneity
among
papers
included
in
the
meta-analysis,
we
include
various
substantive
moderators
(durability
of
the
product,
online
focus
of
the
retailer,
and
type
of
assortment
carried)
and
various
methodological
moderators
(type
of
journal,
type
of
study,
study
location,
and
type
of
respon-
dent).
Using
an
HLM
model
that
accounts
for
multiple
effects
included
in
each
study,
we
find
a
distinct
set
of
return
policy
parameters
affect
purchase
compared
with
the
parameters
that
affect
returns.
Specifically,
money
leniency
and
effort
leniency
increase
purchase
to
a
greater
extent
than
do
the
other
return
pol-
icy
factors.
On
the
other
hand,
leniency
on
time
and
exchange
reduces
returns
more
than
other
return
policy
factors,
while
leniency
on
scope
increases
returns.
Because
the
leniency
fac-
tors
have
a
differential
impact
on
purchase
and
returns,
retailers
should
consider
creating
return
policies
that
vary
along
multiple
dimensions.
In
the
sections
that
follow,
we
provide
a
review
of
the
rel-
evant
literature
on
return
policies
and
their
effect
on
purchase
and
return
proclivities.
After
delineating
the
meta-analytic
pro-
cedure
employed
and
presenting
the
results,
we
conclude
with
a
discussion
of
how
our
findings
provide
meaningful
contributions
to
both
academicians
and
retail
managers.
Conceptual
Foundations
Does
Leniency
in
Return
Policies
Increase
Purchase
More
Than
Returns?
Because
most
researchers
have
limited
their
examinations
of
return
policy
leniency
and
its
effect
on
product
purchase
or
product
returns
to
only
one
or
two
return
policy
factors,
prior
research
does
not
sufficiently
answer
this
question.
For
example,
a
study
conducted
by
Wood
(2001)
varied
only
the
amount
of
the
monetary
refund,
whereas
Petersen
and
Kumar
(2010)
varied
only
the
scope
of
products,
Wang
(2009)
varied
only
the
amount
of
time
and
guarantee,
and
Janakiraman
and
Ordó˜
nez
(2012)
varied
only
the
amount
of
time
and
effort.
Thus,
we
know
little
about
the
overall
effect
of
return
policies
on
purchase
and
return.
On
the
one
hand,
there
is
evidence
that
varying
some
return
policy
factors
increases
purchase
without
increasing
returns.
For
example,
Wood
(2001)
offered
participants
the
opportunity
to
order
a
pen
or
choose
a
gift
certificate
under
conditions
of
either
a
lenient
or
a
restrictive
return
policy,
which
was
varied
by
amount
of
the
monetary
refund
(i.e.,
full
refund
amount
for
the
lenient
condition
and
partial
refund
amount
for
the
restric-
tive
condition).
Upon
receipt
of
the
pen
and
after
product
trial,
participants
were
asked
to
decide
whether
to
keep
the
pen
or
exchange
it
for
a
gift
certificate.
The
lenient
return
policy
was
found
to
increase
purchase,
but
did
not
result
in
higher
return
rates
compared
with
the
restrictive
return
policy.
Wang
(2009)
found
similar
effects,
in
that
lenient
return
policies
significantly
increased
initial
purchasing
tendency
but
did
not
increase
return
rate.
Some
previous
research
shows
leniency
on
other
return
policy
factors
likely
increases
purchase
as
well
as
returns.
For
exam-
ple,
in
a
controlled
field
experiment,
Petersen
and
Kumar
(2010)
varied
a
company’s
product
return
policy
by
either
allowing
for
non-defective
products
to
be
returned
(lenient
return
policy)
or
only
allowing
defective
products
to
be
returned
(restrictive
return
policy).
The
average
dollar
amount
of
returned
products
was
higher
under
the
lenient
return
policy
($67.90)
than
under
the
strict
return
policy
($20.50),
suggesting
a
positive
effect
of
lenient
return
policy
on
return
behavior,
with
higher
levels
of
return
leniency
leading
to
higher
return
proclivity.
There
is
sup-
port
for
such
a
notion
at
the
aggregate
level
as
well.
Bonifield,
Cole,
and
Schultz
(2010)
find
a
positive
relationship
between
number
of
purchases
and
number
of
product
returns,
suggest-
ing
that
increasing
product
purchases
leads
to
greater
product
returns.
To
provide
guidance
on
this
topic,
we
conduct
a
meta-analysis
to
examine
the
overall
impacts
of
return
policy
leniency,
as
well
as
the
individual
influences
of
leniency
factors
that
comprise
return
policies.
We
classify
all
measures
of
purchase
intentions
and
purchase
behavior
as
purchase
proclivity.
Similarly,
we
group
measures
of
both
intentions
to
return
and
actual
returns
as
return
proclivity.
A
summary
of
the
theories
that
are
rele-
vant
to
the
examination
of
the
effect
of
return
policy
leniency
on
purchase
and
returns
follows.
Mechanisms
Driving
Effects
of
Leniency
on
Purchase
In
examining
the
effect
of
return
policy
factors
on
product
purchase,
three
theoretical
mechanisms
have
been
suggested
in
extant
research.
First,
signaling
theory
has
been
put
forth
to
explain
how
return
policies
act
as
positive
quality
signals
by
the
retailer
(e.g.,
Wood
2001).
Second,
consumer
risk
theory
has
been
applied
to
suggest
that
return
policies
should
reduce
the
financial
and
product
risk
that
consumer’s
feel
prior
to
product
purchase
(e.g.,
Van
den
Poel
and
Leunis
1999).
Third,
construal
level
theory
has
been
used
as
a
basis
for
positing
that
individ-
uals
faced
with
lenient
return
policies
are
likely
to
focus
on
the
benefits
of
purchase
rather
than
the
costs
of
purchase
(e.g.,
Janakiraman
and
Ordó˜
nez
2012).

228
N.
Janakiraman
et
al.
/
Journal
of
Retailing
92
(2,
2016)
226–235
A
common
thread
across
the
three
theories
is
they
all
postu-
late
that
lenient
return
policies
should
positively
affect
product
purchase.
However,
the
theories
vary
in
relevance
depending
on
the
return
policy
factor
that
is
examined.
For
example,
researchers
examining
monetary
leniency
or
money
back
guar-
antees
(e.g.,
Suwelack,
Hogreve,
and
Hoyer
2011;
Wood
2001)
have
employed
signaling
theory
and
consumer
risk
theory,
while
those
examining
time
deadlines
have
relied
on
construal
level
theory
(e.g.,
Janakiraman
and
Ordó˜
nez
2012).
Intuition
suggests
that
a
unit
reduction
in
consumer
risk
is
unlikely
to
have
exactly
the
same
impact
on
purchase
as
a
unit
increase
in
consumer
perceived
quality.
Thus,
we
propose
that
while
all
return
policy
factors
are
likely
to
have
a
positive
effect
on
purchase,
the
degree
of
their
influence
will
depend
on
the
specific
return
policy
factor
that
is
varied.
Mechanisms
Driving
Effects
of
Leniency
on
Returns
Similar
to
the
empirical
evidence
reported
earlier,
the
the-
oretical
mechanisms
for
product
returns
are
mixed
in
their
predictions
of
the
effect
of
the
different
return
policy
factors.
Some
theories
predict
lower
returns
with
leniency.
For
instance,
Wood
(2001)
finds
that
lenient
return
policies
enhance
qual-
ity
perceptions
of
the
product
not
only
during
the
pre-purchase
stage,
but
also
during
the
post-purchase
stage.
This
suggests
lenient
return
policies
that
increase
perceived
quality
should
reduce
returns.
This
higher
value
placed
by
the
consumer
on
the
product
increases
with
longer
duration
of
ownership
(i.e.,
an
endowment
effect),
with
longer
return
deadlines
increasing
perceived
product
endowment
more
than
shorter
return
dead-
lines
(Janakiraman
and
Ordó˜
nez
2012;
Wang
2009;
Wood
2001).
This
suggests
lenient
return
policies
that
increase
endowment
should
reduce
return
rates.
However,
other
theories
suggest
increased
leniency
is
likely
to
increase
returns.
For
example,
if
individuals
are
sensitive
to
transaction
costs
post-purchase
(Davis,
Hagerty,
and
Gerstner
1998),
then
lower
transaction
costs
offered
by
lenient
return
policies
should
lead
to
higher
return
rates
by
making
it
easier
to
return
a
product.
Take n
together,
the
various
theories
suggest
that
leniency
on
return
policy
factors
will
differ
not
only
by
degree,
but
also
in
terms
of
the
valence
of
effects
on
product
returns.
Typology
of
Return
Policy
Factors
Previous
researchers
have
attempted
to
classify
the
various
elements
of
return
policy
leniency,
but
little
consensus
exists
regarding
these
factors.
Suwelack
and
Krafft
(2012)
note
that
return
policies
vary
significantly
depending
on
their
terms.
For
example,
in
a
survey
of
133
stores
in
California,
Davis,
Hagerty,
and
Gerstner
(1998)
classify
return
policies
in
terms
of
five
restriction
factors:
(1)
whether
a
store
provides
exchanges
or
cash
refunds;
(2)
whether
a
receipt
is
required
or
not;
(3)
whether
original
packaging
is
required
or
not;
(4)
whether
visible
signs
of
use
is
allowed
or
not;
and
(5)
the
time
limit
for
returns.
Alternatively,
Heiman,
McWilliams,
and
Zilberman
(2001)
char-
acterize
restrictions
in
terms
of
four
factors:
(1)
the
length
of
time
allowed
for
the
return
(typically
between
30
and
60
days);
(2)
the
costs
associated
with
the
product
return
(totally,
partially,
or
not
assumed
by
the
retailer);
(3)
the
terms
of
the
policy
(e.g.,
monetary
refund
vs.
a
replacement
or
an
exchange);
and
(4)
additional
restrictions
(e.g.,
the
obligation
to
have
the
product
returned
in
its
original
package).
Taking
both
of
these
catego-
rization
schemes
into
consideration,
we
classify
return
policy
leniency
as
varying
along
five
dimensions:
1.
Time
leniency.
Retailers
commonly
specify
deadlines
in
their
return
policies
(a
30-day
policy,
a
90-day
policy,
etc.).
Return
policies
that
provide
a
longer
length
of
time
in
which
to
return
products
are
regarded
as
more
lenient.
2.
Monetary
leniency.
Lenient
return
policies
allow
for
a
refund
of
the
full
monetary
amount
paid
for
the
product,
while
strict
policies
allow
for
only
a
portion
of
the
purchase
price
to
be
refunded,
usually
by
imposing
a
“restocking
fee”
or
a
non-
refundable
“shipping
and
handling
fee.”
Policies
that
do
not
impose
monetary
restrictions
are
regarded
as
more
lenient.
3.
Effort
leniency.
Consumer
effort
required
to
execute
returns
varies,
with
some
retailers
creating
“hassles”
for
customers
returning
products
(e.g.,
requiring
the
original
receipt,
tags,
or
product
packaging
be
retained).
Return
policies
requiring
less
effort
on
the
part
of
the
consumer
are
considered
more
lenient.
4.
Scope
leniency.
Stores
limit
items
they
consider
“return-
worthy.”
For
example,
products
purchased
on
sale
may
not
be
eligible
for
return.
Return
policies
with
greater
scope
of
“return-worthy”
items
are
considered
more
lenient.
5.
Exchange
leniency.
While
some
retailers
offer
cash
refunds,
others
offer
store
credit
or
product
exchange
for
the
returned
item.
Return
policies
that
allow
cash
refunds
are
considered
more
lenient.
Fig.
1
visually
represents
the
proposed
linkage
between
the
return
policy
leniency
factors
and
the
behavioral
outcomes
of
purchase
and
return.
The
systematic
coding
of
the
return
policy
leniency
factors
varied
in
each
study
included
in
this
meta-analysis
allows
us
to
address
the
questions
of
which
leniency
factors
exert
a
greater
impact
on
purchase
and
which
have
a
higher
impact
on
returns.
Substantive
Moderators
The
extent
to
which
any
return
policy
might
affect
purchase
and
returns
likely
depends
not
only
on
the
combination
of
return
policy
factors
employed,
but
also
on
the
type
of
retail
store
that
offers
it.
For
example,
Van
den
Poel
and
Leunis
(1999)
exam-
ined
the
effect
of
type
of
retailer
(mail
order/internet/specialty
store/supermarkets)
and
type
of
risk
reliever
(money
back
guar-
antee/price
reduction/known
brand)
as
influencers
of
product
likeability.
They
find
a
significant
interaction
effect
for
store
type
and
risk
reliever,
suggesting
the
type
of
retailer
likely
influ-
ences
the
effect
of
various
risk
relievers.
In
the
current
research,
we
investigate
the
following
substantive
moderators
(see
Fig.
1)
relating
to
suppliers
and
retailers
that
are
likely
to
moderate
the
impact
of
return
policy
leniency
on
purchase
and
return
proclivities:

N.
Janakiraman
et
al.
/
Journal
of
Retailing
92
(2,
2016)
226–235
229
Fig.
1.
Meta-analysis
model.
1.
Assortment.
The
breadth
of
the
assortment
carried
by
retailers
varies
considerably.
In
this
research,
we
term
retailers
car-
rying
a
narrow
assortment
of
product
categories
“specialty
stores”
and
those
carrying
a
broad
assortment
of
product
categories
“general
stores.”
2.
Distribution.
Another
potential
source
of
variance
is
the
retailer’s
primary
method
of
distribution.
Retailers
selling
merchandise
predominantly
through
brick-and-mortar
stores
are
classified
as
“offline
focus,”
while
those
distributing
prod-
ucts
through
e-commerce
web
sites
are
classified
as
“online
focus.”
3.
Durability.
Retailers
selling
products
that
last
a
relatively
longer
time
before
the
consumer
repurchases
or
replaces
the
product
are
classified
as
“durable,”
while
those
selling
prod-
ucts
that
are
immediately
consumed
or
last
a
short
period
of
time
are
categorized
as
“consumable.”
Methodological
Moderators
Several
methodological
variables,
which
we
term
method-
ological
moderators,
also
possess
the
potential
for
explaining
some
variation
in
leniency-purchase
and
leniency-return
cor-
relations.
Studies
employing
student
versus
general
population
samples
and
U.S.
versus
non-U.S.
samples
may
vary
in
pur-
chase
and
return
outcomes.
Further,
the
journal
quality
(elite
vs.
non-elite)
and
the
research
design
(experimental
study
vs.
field
study)
might
also
affect
purchase
and
return
propensities.
Meta-analysis
Database
Development
To
build
our
database,
we
used
multiple
approaches
to
identify
the
population
of
relevant
research
for
inclusion
in
the
meta-analysis.
First,
we
conducted
a
manual
search
of
leading
journals
(Decision
Sciences,
Electronic
Commerce
Research,
European
Journal
of
Social
Psychology,
Interna-
tional
Journal
of
Retail
&
Distribution
Management,
Journal
of
Applied
Psychology,
Journal
of
Applied
Social
Psychology,
Journal
of
Behavioral
Decision
Making,
Journal
of
Business
Research,
Journal
of
Consumer
Marketing,
Journal
of
Con-
sumer
Psychology,
Journal
of
Consumer
Research,
Journal
of
Economic
Psychology,
Journal
of
Experimental
Social
Psychol-
ogy,
Journal
of
Fashion
Marketing
and
Management,
Journal
of
Marketing,
Journal
of
Marketing
Management,
Journal
of
Mar-
keting
Research,
Journal
of
Operations
Management,
Journal
of
Product
&
Brand
Management,
Journal
of
Retailing,
Journal
of
Retailing
and
Consumer
Services,
Journal
of
Service
Research,
Marketing
Letters,
and
Psychology
and
Marketing)
in
which
articles
investigating
the
influence
of
return
policies
were
most
likely
to
appear.
We
then
searched
electronic
databases
(including
EBSCO-
host
Business
Source
Complete,
Emerald,
Google
Scholar
and
Science
Direct)
with
a
keyword
search,
including
terms
such
as
“money
back
guarantees”
(e.g.,
Suwelack,
Hogreve,
and
Hoyer
2011),
“lenient
return
policies”
(e.g.,
Janakiraman
and
Ordó˜
nez
2012;
Wood
2001),
“restrictive
return
policies”
(e.g.,
Bahn
and
Boyd
2014),
“service
failures”
(e.g.,
Mollenkopf
et
al.
2007),
and
“reverse
logistic
chains”
(e.g.,
Rogi´
c,
Bajor,
and
Roˇ
zi´
c
2010).
We
also
extended
our
search
to
unpublished
work
included
in
the
conference
proceedings
of
the
Society
for
Consumer
Psychology
and
the
Association
for
Consumer
Research.
Next,
in
an
effort
to
prevent
a
“publication
bias,”
which
can
reduce
the
measurement
variability
in
meta-analysis
(e.g.,
Andrews
and
Franke
1991),
we
extended
our
search
to
unpub-
lished
works
by
manually
searching
proceedings
of
additional
conferences
and
contacting
authors
who
have
previously
pub-
lished
in
the
return
policy
domain.
Finally,
employing
an
ancestry
approach,
we
examined
the
references
of
key
concep-
tual
articles
and
studies
identified
through
the
above
efforts.
Altogether,
these
efforts
yielded
46
papers
with
empirical
data
related
to
return
policies.
Based
on
initial
screening
of

230
N.
Janakiraman
et
al.
/
Journal
of
Retailing
92
(2,
2016)
226–235
articles,
a
study
was
deemed
eligible
for
inclusion
in
the
meta-
analysis
if
the
following
information
was
reported:
(1)
the
Pearson
correlation
between
return
policy
leniency/strictness
and
purchase
and/or
return
behaviors
or
intentions
(or
enough
statistical
information
was
provided
to
compute
the
correlation);
and,
(2)
the
sample
size
(Janiszewski,
Noel,
and
Sawyer
2003).
We
coded
159
effect
sizes
from
21
papers
that
met
the
inclusion
criteria.
To
detect
outliers
in
the
meta-analytic
dataset,
we
utilized
the
sample-adjusted
meta-analytic
deviancy
(SAMD)
statistic
recommended
by
Huffcutt
and
Arthur
(1995),
which
takes
into
account
both
the
effect
size
and
study
sample
size
when
identi-
fying
outliers.
The
SAMD
outlier
analysis
led
us
to
remove
five
observations
identified
as
problematic
due
to
extremely
large
variance
and/or
calculated
correlations
near
1.0.
As
a
result,
our
final
dataset
consists
of
154
observations,
of
which
75
related
to
purchase
proclivity
and
79
related
to
return
proclivity.
The
total
sample
size
for
the
studies
included
in
our
database
is
11,662;
however,
because
our
dataset
includes
multiple
effects
from
var-
ious
studies,
the
total
sample
size
for
our
meta-analytic
database
is
51,653
(Table
1).
Coding
of
Return
Policy
Factors
and
Moderator
Variables
For
each
effect
included
in
the
meta-analysis,
three
expert
raters
independently
coded
the
potential
moderators—five
return
policy
factors,
three
substantive
moderators,
and
four
methodological
moderators—based
on
information
about
the
study
reported
in
the
corresponding
paper.
A
binary
coding
scheme
was
applied
to
each
of
the
variables,
with
each
rater
independently
rating
each
effect
for
the
potential
moderators.
For
example,
each
rater
coded
the
leniency
factor
variables
by
inputting
a
“1”
if
the
return
policy
factor
was
manipulated
or
measured
in
the
study
and
“0”
if
it
was
not
manipulated
or
mea-
sured
in
the
study.
The
specific
way
each
return
policy
variable
was
classified
as
a
“0”
or
a
“1”
is
detailed
in
Table
2.
The
expert
raters
also
coded
the
type
of
focal
products
exam-
ined
in
the
study,
the
type
of
focal
retail
store
mentioned
in
the
study,
and
the
type
of
research
design
on
which
the
study
was
based.
Table
2
details
the
binary
coding
scheme
used
to
code
the
various
substantive
and
methodological
moderator
variables.
There
was
almost
100
percent
agreement
among
the
judges
on
return-policy
factors,
and
89
percent
agreement
on
the
moder-
ator
variables.
The
coding
was
compared
across
the
raters
and
discrepancies
were
resolved
by
choosing
the
coding
structure
that
was
chosen
by
a
majority
of
the
coders
(i.e.,
two
out
of
the
three
coders).
Analytic
Procedures
To
analyze
the
impact
of
return
policy
leniency
on
consumers’
purchase
and
return
proclivities
we
conducted
both
univariate
analyses—following
procedures
described
by
Borenstein
et
al.
(2009)—and
a
multivariate
approach
employing
a
hierarchical
linear
model
as
suggested
by
Bijmolt
and
Pieters
(2001).
The
effect
sizes
coded
for
the
analyses
were
zero-order
product-moment
correlation
coefficients,
which
are
easy
to
interpret
and
to
compare
across
studies
(e.g.,
Brown
and
Stayman
1992).
We
calculated
and
analyzed
the
estimated
true
correlation
(rT)
between
the
return
policy
leniency
factors
and
key
outcomes.
In
order
to
calculate
the
mean
rT,
each
correla-
tion
for
a
given
study
was
weighted
by
its
sampling
variance
(which
is
a
function
of
the
sample
size
for
the
observation)
and
then
averaged
across
all
studies.
Next,
the
total
heterogeneity
of
the
sample
(a
weighted
sum
of
squares
statistic
comparable
to
a
total
sum
of
squares
in
an
ANOVA)
is
calculated
and
tested
against
a
chi-square
(χ2),
distribution
to
test
whether
variance
among
the
effect
sizes
is
greater
than
expected
by
sampling
error
(Rosenberg,
Adams,
and
Gurevitch
2000).
A
significant
chi-square
test
indicates
that
inconsistent
find-
ings
across
the
observed
correlations
with
respect
to
product
returns
cannot
be
fully
explained
by
statistical
artifacts
and
suggests
the
presence
of
other
factors
(i.e.,
moderators)
in
the
exploration
of
heterogeneity
(Hunter
and
Schmidt
2004).
The
95
percent
bootstrapped
confidence
interval
(CIBS)
for
each
correlation
is
also
calculated
in
an
effort
to
examine
the
signifi-
cance
of
the
mean-corrected
correlations.
Because
meta-analytic
data
often
violate
the
distributional
assumptions
of
paramet-
ric
tests,
the
use
of
bootstrapped
confidence
intervals,
which
are
based
on
a
non-parametric
distribution,
is
appropriate
and
provides
a
more
powerful
estimate
than
traditional
confidence
intervals
(Rosenberg,
Adams,
and
Gurevitch
1997).
Finally,
the
fail-safe
sample
size
(NFS)
using
Rosenthal’s
(1979)
method
was
calculated
to
assess
the
possibility
of
publication
bias
or
a
“file-drawer”
problem.
This
information
estimates
the
number
of
unpublished
studies
with
an
effect
size
of
zero
that
would
have
to
exist
to
render
the
observed
effects
insignificant
at
the
alpha
equal
to
a
.05
level
(Janiszewski,
Noel,
and
Sawyer
2003),
with
a
larger
NFS value
conveying
greater
confidence
in
the
results.
Multivariate
Analysis
using
a
Hierarchical
Linear
Model
(HLM).
Bijmolt
and
Pieters
(2001)
suggest
the
nested
nature
of
meta-analytic
data
imposes
a
correlation
structure
that
is
best
accommodated
by
specifying
a
hierarchical
linear
model
(HLM).
To
jointly
assess
the
influence
of
the
proscribed
vari-
ables
on
purchase
(return)
proclivity
we
specify
a
model
in
which
the
coded
effect
sizes
are
a
linear
function
of
the
return
policy
factors
and
the
substantive
and
methodological
moderators.
The
software
package
HLM
7
was
used
to
estimate
our
models
of
purchase
(return)
proclivity
on
leniency
factors,
retailer
factors
and
manuscript
control
variables
and
is
formally
described
in
Eqs.
(1)
and
(2).
The
variables
included
in
our
study
of
the
impact
of
return
policy
leniency
on
purchase
and
return
proclivities
vary
at
the
effect-size
level
(level
1)
and
the
independent
sample
level
(level
2).
Due
to
the
limited
number
of
observations
in
our
meta-
analytic
dataset,
we
test
only
for
main
effects
in
the
HLM.
1.
Level
1—Effect
Size
Model
The
level
1
model
explains
the
variation
of
effect
size
i
(where
i
=
1.
.
.nj)
around
the
intercept
for
study
j
(j
=
1.
.
.J).
At
level
1,
the
return
policy
leniency
dimension
under
inves-
tigation
is
captured
by
p
categorical
variables
where
p
=
5
(Time,
Money,
Effort,
Scope,
Exchange).
When
the
variables
are
mean
centered,
π0j represents
the
mean
effect
size
for

N.
Janakiraman
et
al.
/
Journal
of
Retailing
92
(2,
2016)
226–235
231
Table
1
Articles
used
in
meta-analysis
of
return
policy
effects.
Authors
Year
#
of
studies
#
of
observations
Research
design
Dependent
variable
Product
category
Leniency
factors
examined
Bahn
and
Boyd 2014
2
170
experiment intentions
durables
time
Bonifield,
Cole,
and
Schultz 2010
2
541
content
analysis;
experiment
intentions
consumables
time,
money,
effort,
exchange
Bower
and
Maxham
2012
2
1,630
longitudinal
event
field
study
behaviors
consumables
money
d’Astous
and
Guèvremont
2008
1
164
experiment
intentions
durables
time
and
exchange
Derbaix
1983
1
200
cross-sectional
study
intentions
durables
money
Griffis
et
al.
2012
1
445
experiment
behaviors
durables
effort
Huppertz
2007
1
338
experiment
intentions
durables
effort
and
scope
Janakiraman
and
Ordonez
2012
3
430
experiment
behaviors
consumables
time
and
effort
Javadi
et
al.
2012
1
107
cross-sectional
study
intentions
consumables
time,
money,
effort,
scope,
and
exchange
Kang
and
Johnson
2009
1
246
cross-sectional
study
behaviors
durables
time,
money,
effort,
scope,
and
exchange
Kim
and
Wansink
2012
2
367
experiment
intentions
durables
and
consumables
time,
money,
effort,
scope,
and
exchange
Lantz
and
Hjort 2013
1
4,000
field
experiment
behaviors
consumables
money
Maity
and
Arnold 2013
1
193
cross-sectional
study
intentions
durables
time,
money,
effort,
scope,
and
exchange
Pei,
Paswan,
and
Yan
2014
1
300
cross-sectional
study
intentions
consumables
time,
money,
effort,
scope,
and
exchange
Posselt,
Gerstner,
and
Radic
2008
2
952
content
analysis
behaviors
durables
time
and
money
Powers
and
Jack
2013
1
308
cross-sectional
study
behaviors
consumables
time,
money,
effort,
scope,
and
exchange
Shao,
Chang,
and
Zhang
2014
1
182
experiment
intentions
durables
money
Suwelack,
Hogreve,
and
Hoyer
2011
2
600
experiment
intentions
consumables
time,
money,
effort
Van
den
Poel
and
Leunis
1999
1
93
experiment
intentions
durables
money
Wang
2009
1
137
experiment
behaviors
consumables
time
and
money
Wood
2001
3
259
experiment
intentions
and
behaviors
durables
and
consumables
time,
money,
effort,
scope,
and
exchange
Total
(21
papers)
31
11,662
Note:
There
are
75
effect
sizes
for
purchases
drawn
from
18
studies,
and
79
effect
sizes
for
returns
drawn
from
17
studies.
Table
2
Meta-analysis
coding
scheme.
Factor
Variable
Binary
coding
scheme
Return
Policy
Time
If
time
deadline
varied
in
study
=
1
If
time
deadline
not
varied
in
study
=
0
Return
Policy
Money
If
money
back
varied
in
study
=
1
If
money
back
not
varied
in
study
=
0
Return
Policy
Effort
If
effort
for
returns
varied
in
study
=
1
If
effort
for
returns
not
varied
in
study
=
0
Return
Policy
Scope
If
discounted
products
accepted
for
returns
=
1
If
discounted
products
not
accepted
=
0
Return
Policy
Exchange
If
money
is
offered
on
returns
in
study
=
1
If
credit
or
product
exchange
offered
on
returns
in
study
=
0
Substantive
Moderator
Type
of
Product
If
focal
product
was
durable
product
=
1
If
focal
product
was
not
durable
product
=
0
Substantive
Moderator
Assortment
Focus
If
retail
store
described
in
study
specialty
store
=
1
If
retail
store
described
in
study
not
specialty
store
=
0
Substantive
Moderator
Channel
Focus
If
retail
store
described
in
study
was
online
store
=
1
If
retail
store
described
in
study
was
not
online
store
=
0
Methodological
Moderator
Journal
Quality
If
paper
published
in
journal
as
described
by
UT
Dallas
list
as
elite
=
1
If
paper
published
in
journal
as
described
by
UT
Dallas
list
as
not
elite
=
0
Methodological
Moderator
Research
Design
If
study
involved
experimental
manipulation
=
1
If
study
did
not
involve
experimental
manipulation
=
0
Methodological
Moderator Study
Location
If
study
was
conducted
in
the
United
States
=
1
If
study
was
not
conducted
in
the
United
States
=
0
Methodological
Moderator
Participant
Details
If
study
was
among
student
participants
=
1
If
study
was
not
among
student
participants
=
0

232
N.
Janakiraman
et
al.
/
Journal
of
Retailing
92
(2,
2016)
226–235
Table
3
Overall
effect
of
return
policy
leniency
on
purchase
proclivity
and
return
proclivity
(based
on
meta-analysis
of
21
papers).
Number
of
samples
in
each
analysis
(k)
Number
of
observations
in
the
k
samples
(N)
Mean
uncorrected
correlation
(r)
Mean
weighted
corrected
correlation
(rt)
Estimated
population
variance
(σ2
t)
95
percent
bootstrapped
confidence
intervals
(CIBS)
Effect
size
Variance
unaccounted
for
across
samples
(χ2)
Fail-safe
N
for
each
variable
(NfsR)
Purchase
proclivity
75
21,052
0.212
0.217** 0.139
0.175
0.261
Small
469.25
22,297
Return
proclivity
79
20,831
0.053
0.080** 0.203
0.021
0.137
Small
1,859.82
569.8
** p
<
.01.
study
jk.
eij represents
the
sampling
error
of
effect
size
Yij,
which
is
assumed
to
be
normally
distributed
with
mean
of
0
and
known
variance
Vi—which
is
calculated
based
on
statis-
tics
collected
from
the
included
study.
Yij =
π0j+
P

p=1
πpj apij +
eij (1)
2.
Level
2—Independent
Sample
Level
Model
At
level
2,
we
model
the
intercept
of
study
j
as
a
function
of
the
retailer
characteristics
and
study
control
variables
that
vary
across
studies
X(q=1–8) (Durable
Good,
Specialty
Store,
Online,
Elite
Journal
Publication,
U.S.
Respondents,
Student
Respondents,
Experimental
Setting).
Again,
mean
centering
the
variables
at
level
2
means
the
intercept,
β00 is
interpreted
as
the
average
effect
size.
The
error
term
r0jis
assumed
to
be
normally
distributed
with
mean
of
0
and
variance
τπ.
π0j=
β00 +
Q

q=1
β0qXqj +
r0j(2)
Results
Overall
Results.
The
strength
of
meta-analysis
lies
in
its
pooling
of
estimates
of
association
across
studies
and
analyzing
differences
in
relationship
strength
according
to
factors
that
can
distinguish
these
effects.
Table
3
provides
an
overview
of
the
main
effects
associated
with
return
policy
leniency
on
purchase
indicators
and
return
indicators.
The
overall
correlation
between
return
leniency
and
purchase
indicators
is
0.218
and
that
between
return
leniency
and
return
proclivity
is
0.080,
suggesting
that
return
leniency
is
associated
with
both
increased
purchase
proclivity
and
increased
return
pro-
clivity.
Further,
the
effect
on
purchase
proclivity
is
greater
than
the
effect
on
return
proclivity.
Consistent
with
Rosenthal
and
Rosnow
(1991),
these
effects
are
categorized
as
small
and
are
significant.
The
high
NFS for
both
effects
suggests
that
publica-
tion
bias
is
not
a
problem.
Given
the
heterogeneity
present
within
the
dataset
(χ2
pur =
469.25,
df
=
74,
p
<
.01)
(χ2
ret =
1859.82,
df
=
78,
p
<
.01),
an
examination
of
the
moderators
of
the
rela-
tionship
between
return
leniency
and
purchase
and
return
is
warranted.
Multivariate
Analysis
using
HLM.
The
multi-level
model
as
specified
in
Eqs.
(1)
and
(2)
that
accounts
for
multiple
effects
per
study
was
used
to
estimate
the
impact
of
the
Table
4
Estimation
results—HLM
model—fixed
effects
for
each
DV.
Purchase
proclivity
Return
proclivity
Level
1
Coefficients
Return
Policy
Factors
Time
0.028
−0.381*
Money 0.089*0.021
Effort 0.122*−0.156
Scope
0.163
0.049+
Exchange
0.040
−0.582*
Level
2
Coefficients
Intercept
0.243*0.111*
Substantive
Moderators
Type
of
Product
0.155+0.102
Assortment
Focus
0.140
−0.264
Channel
Focus
−0.013
−0.457
Methodological
Moderators
Journal
Quality
0.213+−0.019
Research
Design
0.085
−0.726+
Study
Location
−0.016
0.198
Participant
Details
−0.068
0.195+
Variance
Component
u0
0.015*0.048*
Variance
Explained
35.9
percent
41.7
percent
AIC
434.70
752.83
BIC
467.10
783.63
*p
<
.05.
+p
<
.1.
influencing
variables
on
purchase
proclivity
and
return
pro-
clivity.
The
meta-regression
models
of
purchase
and
return
proclivites
on
the
influencing
variables
are
provided
in
Table
4.
The
reported
results
are
estimated
using
Restricted
Maximum
Likelihood
techniques,
which
are
more
appropriate
for
estimat-
ing
the
variance
parameters
due
to
the
small
number
of
level
two
units
(independent
samples)
in
our
dataset
(Raudenbush
and
Bryk
2002),
and
provide
the
results
of
a
fully
conditional
model
for
both
purchase
and
returns.
The
variables
were
cen-
tered
around
the
grand
mean
which
allows
interpretation
of
the
intercept
as
the
adjusted
mean
effect
size
for
study
j.5
5The
simple
full
conditional
model
was
found
to
be
significantly
better
(based
on
AIC/BIC)
than
a
non-conditional
model
that
did
not
include
any
of
the
vari-
ables,
or
a
partial
conditional
model
that
included
only
the
return
policy
variables
and
ignored
the
moderator
variables.
Further,
within
the
full
conditional
model
we
ran
different
models
by
dropping
variables
such
as
Specialty
and
Online.</