Content uploaded by Traci Freling
Author content
All content in this area was uploaded by Traci Freling on Jul 12, 2022
Content may be subject to copyright.
When not to accentuate the positive: Re-examining valence effects
in attribute framing
Traci H. Freling
a,
⇑
, Leslie H. Vincent
b
, David H. Henard
c
a
University of Texas at Arlington, 217 College of Business Administration, Arlington, TX 76019, USA
b
University of Kentucky, Gatton College of Business & Economics, Lexington, KY 40506, USA
c
North Carolina State University, Poole College of Management, Raleigh, NC 27695, USA
article info
Article history:
Received 26 March 2012
Accepted 7 December 2013
Available online 13 March 2014
Accepted by Harris Sondak
Keywords:
Message framing
Framing effects
Construal level theory
Meta-analysis
abstract
While the expanding body of attribute framing literature provides keen insights into individual
judgments and evaluations, a lack of theoretical perspective inhibits scholars from more fully extending
research foci beyond a relatively straightforward examination of message content. The current research
applies construal level theory to attribute framing research. The authors conduct a meta-analysis of 107
published articles and then conceptually expand this knowledge base by synthesizing attribute framing
research and construal level concepts. Results suggest that attribute framing is most effective when there
is congruence between the construal level evoked in a frame and the evaluator’s psychological distance
from the framed event. A follow-up experiment confirms that the congruence between a frame’s
construal level and psychological distance—not simply its valence—appears to be driving attribute
framing effects. This research proposes to shift the focus in attribute framing research from that of
message composition to a more complex relationship between the message and the recipient.
Ó2014 Elsevier Inc. All rights reserved.
Introduction
‘‘4 out of 5 dentists surveyed would recommend sugarless gum...’’
This now infamous advertising tagline for Trident gum has been
used for nearly 50 years and is a prime example of how marketing
managers successfully use framing in persuasive messages. Parallel
to its use in practice, research on framing effects and their impact
on decision-making continues to proliferate and bears tribute to
the interest level in the subject area. In line with Krishnamurthy,
Carter, and Blair (2001) we define framing, in general, as presenting
individuals with logically equivalent options in semantically differ-
ent ways. Framing scholars traditionally focus their research on one
of three frame types: attribute, risky choice, or goal framing. While
each of these focal areas provides insight into various facets of
choice, Levin, Schneider, and Gaeth (1998) warn that the three dif-
ferent types of framing should be examined independently to avoid
unnecessary complexity and confusion that can result from their
idiosyncratic characteristics. In this current research, we therefore
focus our attention on the effects of attribute framing, wherein the
object of the frame is an attribute of the decision option.
Extant empirical works overwhelmingly indicate that people
are more receptive to positive (e.g., 4 out of 5 dentists recommend
Trident sugarless gum) vs. negative (e.g., Only 1 out of 5 dentists
does not recommend Trident sugarless gum) attribute frames.
Krishnamurthy et al. (2001) explain that positive framing is more
effective because it ‘‘generates more positive associations and thus
seems more attractive than negatively framed options’’ (p. 383).
Levin et al. (1998) support this contention, stating ‘‘even at the
most basic level the valence of a description often has a substantial
influence on the processing of that information’’ (p. 184). Given
this, one could view knowledge of attribute framing as fait accom-
pli, concluding that attribute framing effects are so straightforward
that the results are, statistically speaking, nearly always positive
and that any differences in outcomes are simply a matter of degree
or a result of study artifacts.
While valence effects in attribute framing are ‘‘a reliable
phenomenon’’ (Levin, Gaeth, Schreiber, & Lauriola, 2002, p. 413),
viewing them as straightforward is problematic. Research
regularly reveals that seemingly straightforward relationships are
often more complex when viewed from different levels of analysis.
Toward this end, researchers have identified various moderators of
valence effects including the nature of the product (Khan & Dhar,
2010), personal involvement with the framed issue (Chan &
Mukhopadhyay, 2010), and processing motivation and opportunity
(Shiv, Edell Britton, & Payne, 2004). Together, these studies suggest
http://dx.doi.org/10.1016/j.obhdp.2013.12.007
0749-5978/Ó2014 Elsevier Inc. All rights reserved.
⇑
Corresponding author. Fax: +1 (817) 272 2854.
E-mail addresses: freling@uta.edu(T.H. Freling), lvinc2@email.uk y.edu (L.H. Vincent),
dhhenard@ncsu.edu (D.H. Henard).
Organizational Behavior and Human Decision Processes 124 (2014) 95–109
Contents lists available at ScienceDirect
Organizational Behavior and Human Decision Processes
journal homepage: www.elsevier.com/locate/obhdp
that something more than valence could be driving attribute fram-
ing effects.
Since the publication of seminal works in this area
(Krishnamurthy et al., 2001; Levin et al., 1998), there has been a
continued expansion of published research. Thus, the first goal of
this manuscript is to conduct a meta-analysis of the research
stream to update the empirical base of knowledge on attribute
framing. While much of the early work on attribute framing in-
volved exploring the impact of positive vs. negative message attri-
butes (i.e., valence effects), a growing trend toward investigating
other issues such as differing frames of reference and temporal
contexts has developed. It is our view that this is an impactful
and meaningful research evolution. The second goal of this manu-
script is therefore to determine if components of construal level
theory constitute important structural determinants of framing
effects that could possibly encompass both earlier generalizations
focusing on valence and more recent work. We believe that a
theory-driven, micro-focused examination of the attribute framing
literature will yield insights that build upon and extend existing
research.
Construal level theory (Liberman & Trope, 1998; Trope &
Liberman, 2010; Trope, Liberman, & Wakslak, 2007) is a useful
and relevant conceptual lens through which to view attribute
framing effects. This is due to its treatment of events and issues
as differing in terms of construal level and psychological distance,
which together can impact resulting evaluations. The dimensions
of both construal level and psychological distance map favorably
with key variables manipulated in attribute framing research. Fur-
thermore, by incorporating construal level theory into the extant
attribute framing literature, we are better able to meaningfully
understand the nuances of concomitant effects. While much of
the framing literature has focused on message construction, using
a construal level perspective to guide our investigation allows us to
examine the interaction between the message and the recipient,
thus providing a richer understanding of the phenomena. This ap-
proach makes the subsequent findings relevant for any individual
in an organization who is responsible for crafting persuasive mes-
sages in a host of managerial, negotiation, selling, evaluation, or
promotional situations.
In the following section, we discuss the conceptual foundations
that underpin existing attribute framing research. We then con-
duct a meta-analysis designed to both update the current base of
attribute framing research knowledge and expand that knowledge
with a fine-grained theoretical perspective using construal-level
theory. Building on the meta-analysis, we advance the attribute
framing literature by conducting an experiment that investigates
the outcome effects emanating from the congruency between the
evoked construal level of a message frame and the perceived psy-
chological distance of intended message recipients. The manu-
script concludes with a discussion of the results and implications
for framing scholars and practitioners responsible for developing
persuasive messages.
Conceptual development
Attribute framing
Kahneman and Tversky (1979) were the first researchers to
demonstrate that framing (i.e., different wording of formally iden-
tical problems) makes individuals code decision outcomes as gains
or losses relative to a reference point. Since that groundbreaking
work, empirical research on framing effects has flourished across
multiple research domains including cognition, psycholinguistics,
perception, social psychology, health psychology, clinical
psychology, educational psychology, and marketing (Kühberger,
1998). While the term ‘‘framing’’ includes all of the various ways
decision situations are presented that lead decision-makers to
construct markedly different representations of such situations
(Kühberger, 1995), we focus exclusively on attribute framing,in
which a single attribute within a given context is the subject of
the framing manipulation (e.g., describing ground beef as ‘‘80%
lean’’ or ‘‘20% fat’’).
We distinguish attribute framing from two other types of
framing identified by Levin et al. (1998):risky choice framing,
which describes the outcomes of a potential choice involving
options differing in level of risk (e.g., presenting two programs
differing in risk level for reducing cholesterol described in terms
of either positive or negative outcomes); and goal framing, where
the goal of an action or behavior is framed (e.g., stressing either
the positive consequences of reducing red meat in one’s diet or
the negative consequences of failing to do so).
Levin et al. (1998) cogently assert that research exploring these
different types of frames is qualitatively different because attri-
bute, goal, and risky choice framing involve different mechanisms
and consequences, and vary in terms of the information that is
framed, the presumed outcome of the frame, and the manner in
which effects of the frame are measured (see Table 1, p. 151). Levin
et al. (2002) empirically corroborate these theoretical propositions
using a within-subjects framing manipulation in a study conducted
across two sessions in which each subject saw both framing condi-
tions and all three types of frames. Among other key insights, Levin
et al. (2002) demonstrate significant effects for attribute and risky
choice framing, but not goal framing and conduct direct test of
dependency suggesting the three types of framing are governed
by difference processes that are independent of each other. These
results provide further empirical support for the decision to solely
concentrate on attribute framing in this meta-analysis.
Another contribution of Levin et al. (1998) is their identification
of a ‘‘valence-consistent shift’’ that is found in most attribute fram-
ing studies, wherein a positive description of attributes leads to
more favorable evaluations than a negative frame. A classic dem-
onstration of this valence-consistent shift is provided by Levin
and Gaeth (1988), where ground beef was rated as better tasting
and less greasy among subjects exposed to a ‘‘75% lean’’ frame
compared to those in a ‘‘25% fat’’ frame. In other attribute framing
studies, subjects evaluate issues described in terms of ‘‘success’’ or
‘‘survival’’ rates vs. ‘‘failure’’ or ‘‘mortality’’ rates (Davis & Bobko,
1986; Dunegan, 1993; Levin, Schnittjer, & Thee, 1988; Linville,
Fischer, & Fischhoff, 1993; Marteau, 1989), or assess gambling con-
texts that are portrayed in terms of probability of ‘‘winning’’ or
‘‘losing’’ (Levin, Snyder, & Chapman, 1989; Levin et al., 1986). In
such studies, the alternative framed in a more positive light is rou-
tinely rated more favorably than when described negatively.
While this valence-consistent shift has been amply demon-
strated in the literature, the body of work on attribute framing
has tremendously expanded since Levin et al.’s (1998) article and
warrants a new review and synthesis. Interestingly, much of the
recent attribute framing research foregoes a valence manipulation
and explores the effects of presenting numeric information in dif-
ferent formats (e.g., dollars vs. cents), providing different frames of
reference (e.g., self vs. others), or varying the temporal context
(e.g., now vs. in the future). While earlier attribute framing re-
search primarily assessed effects in terms of evaluations, many re-
cent studies use alternative criterion variables such as behaviors,
behavioral intentions, estimates, and predictions. To explore and
better understand these important qualitative differences, we
draw upon construal level theory (CLT) to integrate previous attri-
bute framing research findings into a theoretical framework that
allows us to make specific predictions about when effects should
be stronger across different types and contexts of attribute
framing.
96 T.H. Freling et al. / Organizational Behavior and Human Decision Processes 124 (2014) 95–109
Construal level theory
According to CLT, an association forms between an individual’s
perceptions of his psychological distance from an event being eval-
uated and the construal level of the information he receives about
that event (Liberman & Trope, 1998). This, in turn, influences that
individual’s outcome evaluations. Applying CLT terminology to
attribute framing research, an ‘‘event’’ translates as the object or
situation being evaluated, ‘‘construal level’’ is the mental represen-
tation evoked by the attribute framing information provided, and
‘‘psychological distance’’ equates to any dimension that affects
how closely the individual perceives himself to be from the framed
event or issue. Although attribute framing studies have not previ-
ously drawn upon CLT to explain observed effects, its central con-
structs—construal level and psychological distance—are readily
applicable to this body of work.
In line with Trope and Liberman (2010), we conceptualize
construal level as the degree of perceived abstractness that an event
holds for an individual. Low-level construals are relatively unstruc-
tured, contextualized representations that include subordinate and
incidental features of the event. In contrast, high-level construals
are structured, de-contextualized representations that include only
a few superordinate core features of the events (Trope & Liberman,
2003). CLT posits that individuals construct different representa-
tions of the same information at varying levels of abstraction
(Liberman & Trope, 1998). For example, one might have a career
ambition of ‘‘being successful’’ with a personal ambition of
‘‘spending quality time with my family’’ (both abstract, high-level
construals) or a career ambition of ‘‘being a productive researcher
and a respected teacher’’ with a personal ambition of ‘‘visiting
Disneyworld next week with my family’’ (both concrete, low-level
construals).
How an individual construes an event is affected by that person’s
psychological distance from the event—his perceptions of temporal
distance (when an event occurs), spatial distance (where it is likely
to occur), social distance (to whom it occurs), or hypothetical dis-
tance (whether it occurs). The closer an event is to the individual,
the more concrete it is perceived; the greater the distance from
the event, the more abstract the perception. CLT research demon-
strates events that take place farther into the future (Liberman,
Sagristano, & Trope, 2002; Liberman & Trope, 1998; Wakslak,
Nussbaum, Liberman, & Trope, 2008), that occur in more remote
locations (Fujita, Henderson, Eng, Trope, & Liberman, 2006a;
Henderson, Fujita, Trope, & Liberman, 2006), that are less likely to
occur (Todorov, Goren, & Trope, 2007; Wakslak, Trope, Liberman,
& Alony, 2006), and that happen to people less similar to the
evaluator (Liviatan, Trope, & Liberman, 2008; Smith & Trope,
2006) are associated with greater psychological distance and thus
perceived as more abstract. Beyond these four primary psycholog-
ical distance dimensions, recent research identifies additional
distance variables that could also affect construal level, such as
informational distance (i.e., the amount of knowledge or relevant
data the consumers possesses about decision options), experiential
distance (i.e., the degree to which information is based on direct
experience), affective distance (i.e., the emotional intensity of the
decision context), and perspective distance (i.e., one’s commitment
level in the decision process) (Fieldler, 2007).
Studies exploring the effects of attribute framing vary consider-
ably in terms of how the framing information is presented to sub-
jects to evaluate. For example, Shiv et al. (2004) broadly position
the stimulus product in their experiment as simply ‘‘better than’’
or ‘‘worse than’’ the competitor (high-level construal). In contrast,
Buda and Zhang (2000) frame the stimulus product in their exper-
iment using relatively more concrete details such as ‘‘85% of cus-
tomers were satisfied with the product’’ vs. ‘‘15% of customers
were dissatisfied with the product’’ (low-level construal). There
is also substantial variation in studies exploring the effects of attri-
bute framing, in terms of how psychologically distant the decision
event is from the person exposed to the frame. To illustrate, Dun-
egan (1996) requires student subjects to make business resource
allocation decisions (a scenario that is high in hypothetical dis-
tance for this sample), while student subjects in Agrawal and Duh-
achek’s (2010) experiments evaluate ad appeals for anti-drinking
messages (a scenario that is low in hypothetical distance for this
sample). Thus, both construal level and psychological distance ap-
pear to be appropriate conceptual constructs to enhance our
understanding of attribute framing effects.
Hypotheses
Viewing attribute framing research through a CLT lens allows
scholars to investigate certain nuances and micro-level dynamics
at play in persuasive messaging. That is, the application of a CLT
perspective to attribute framing data provides an opportunity to
test theoretical factors that could have an impact on relationships
of interest. In the current research, meta-analysis becomes an
important tool for both testing and expanding theory. Applying
CLT to attribute framing, we assert that an individual’s construal
level is largely dictated by the relative abstractness of the informa-
tion presented in the frame—a research design factor that is wholly
under the control of the message developer and one that can affect
perceptions of the attribute frame. Based on CLT research, we fur-
ther anticipate that the psychological distance inherent in the deci-
sion context or task makes that information more or less
persuasive depending on its congruence with how the information
in the frame is construed. Specifically, attribute framing effects
should be stronger when the construal level and psychological dis-
tance in the framing manipulation are congruent.
We expect that when the construal level of the frame is aligned
with subjects’ perceptions of how distant or proximal the event is,
the framing effect should be stronger and vice versa. This
prediction is consistent with research by Zhao and Xie (2011)
who explore the effectiveness of temporal attribute frames when
social distance is also manipulated. Their findings suggest that con-
gruency of construal levels between distal others and distant time
results in others’ recommendations having a stronger impact on
subjects’ preferences and choice shifts for the distant future than
for the near future (when the construal levels between distal
Table 1
Main effect results for attribute framing effects.
Number of
samples (k)
Number of
observations
(N)
Mean
correlation
(r)
a
Weighted
correlation
(r
W
)
b
Weighted
variance
(var
t
)
95% Confidence
Interval (CI
BS
)
80% Credibility
Interval (CI)
Unaccounted
variance (
v
2
)
Fail-safe
sample size
(N
fsR
)
Evaluations 359 51,665 .25 .25 .041 .22–.28 .02 to .52 3274.56
*
413,085
Estimates 70 9785 .30 .24 .018 .20–.29 .05–.43 491.55
*
16,467
Behaviors 175 26,876 .23 .21 .015 .19–.23 .05–.37 618.21
*
91,196
a
The ‘‘mean correlation’’ is a simple average among all of the coded effect sizes reported for each relationship and is un-weighted.
b
The ‘‘weighted correlation’’ is the reliability-corrected or sample-size weighted mean correlation to account for sampling error.
*
p6.05.
T.H. Freling et al. / Organizational Behavior and Human Decision Processes 124 (2014) 95–109 97
others and proximal time did not match). Other researchers have
employed CLT to explore goal framing in recent empirical work.
Lee, Keller, and Sternthal (2010) explore the effectiveness of pre-
vention vs. promotion frames at high vs. low levels of construal le-
vel, demonstrating that attitudes and performance on subsequent
tasks (i.e., solving anagrams) are enhanced when subjects process
information construed at a level that fits with their regulatory fo-
cus. White, MacDonnell, and Dahl (2011) test a similar ‘‘matching
hypothesis,’’ providing evidence that loss (gain) goal frames more
effectively influence consumer behavior when paired with low-le-
vel, concrete (high-level, abstract) marketing appeals.
These newly published studies inform our theorizing about how
CLT applies to attribute framing. Based in part on findings by Zhao
and Xie (2011), Lee et al. (2010), and White et al. (2011), we predict
stronger attribute framing effects for congruent pairings as com-
pared to incongruent pairings (i.e., psychologically close scenarios
with high-level construal or psychologically distant scenarios with
low-level construal). Consistent with these framing studies and re-
cent CLT research showing that a ‘‘match in messaging’’ results in
increased fluency, ease of understanding, greater processing mean-
ing, and stronger consumer responses (Ulkumen & Cheema, 2011),
we expect stronger framing effects to occur when individuals in a
psychologically close decision situation (e.g., evaluating a new
product) receive information framed at a lower-level construal
(e.g., 85% satisfied vs. 15% dissatisfied) (Buda & Zhang, 2000) and
when individuals in a psychologically distant decision situation
(e.g., making real estate investment decisions) are exposed to a
frame evoking a higher-level construal (e.g., higher profits vs. low-
er expenses) (Brockner, Wiesenfeld, & Martin, 1995). We hypothe-
size this relationship across four facets of psychological distance
(i.e., temporal, hypothetical, affective, informational).
1
Hypothesis 1. Stronger attribute framing effects will occur when
the construal level of the information presented is congruent with
subjects’ psychological distance from the framed event.
Because attribute framing research is replete with demonstra-
tions of the valence-consistent shift (Levin et al., 1998), we also
hypothesize—from a construal level perspective—how information
valence impacts attribute framing effects. Interestingly, CLT
scholars are increasingly treating information valence as a mani-
festation of construal level and suggest that positive information
is construed at a higher level than negative information and that
it is more valued in decision-making (Eyal, Liberman, Trope, &
Walther, 2004). This is theorized to be true because negative infor-
mation is only important when positive information is present,
whereas the importance of positive information does not depend
on the existence of negative information. For example, in deciding
whether to accept a new job opportunity, an individual might con-
sider the positive factors (e.g., assuming greater responsibility and
earning more money) as well as any associated downsides such as
learning a new position, having a longer commute, or increased
overnight travel. If the job opportunity had no apparent benefits,
individuals would likely not inquire about potential drawbacks
and simply decline the offer. In contrast, the same individual
would likely inquire about the job’s benefits even in the absence
of perceived negatives.
A handful of recent CLT studies examine the differential effects
of positive vs. negative information in individual choice models
and demonstrate that negative information is construed at a lower
level than positive information in regards to decision-making. This
research also suggests that positive information—which evokes a
more abstract mindset—should be more persuasive when the
message concerns a more psychologically distant issue, whereas
negative information—which leads to a lower level of construal—
should be more effective for psychologically closer events. For
example, Eyal et al. (2004) demonstrated that, as temporal distance
from an event increased, subjects generated more arguments in fa-
vor of (and fewer arguments against) a particular action. Similarly,
in research conducted by Herzog, Hansen, and Wanke (2007), dis-
tant-future actions were construed in terms of their pro aspects
while near-future actions were construed in terms of their con as-
pects. As such, subjects found it easier to generate arguments in
support of (against) actions pertaining to the distant (near) future.
Likewise, Labroo and Patrick (2009) found that presenting an event
as benign increased subjects’ abstract construal and the adoption
of abstract, future goals while presenting an event as dangerous
caused subjects to focus attention on concrete, proximal concerns
and reduce the adoption of abstract, distant goals. In sum, framing
an event by highlighting its positive aspects and inducing a higher
level of construal is likely to be more persuasive when the individ-
ual evaluating the message perceives the event to be psychologi-
cally distant. Conversely, negative framing is likely to induce a
lower level of construal, which will be more effective when the
event is more psychologically close to the individual evaluating
the message.
This relatively recent development in CLT research is important
because if this subordination of negative information does apply to
attribute framing, it could provide a compelling explanation for the
valence-consistent shift historically documented in attribute fram-
ing research, which would be a meaningful development for fram-
ing researchers. Based on construal level theory and these recent
empirical findings, we offer the following prediction about the im-
pact of valence information on construal level in attribute framing:
Hypothesis 2. Stronger attribute framing effects will occur when
the frame’s valence evokes a construal level that is congruent with
the individual’s psychological distance from the framed event.
Methodology
Dataset development
We identified extant empirical research studies focusing on
attribute framing in multiple ways. First, we collected the articles
referenced in seminal review pieces by Krishnamurthy et al.
(2001) and Levin et al. (1998). Second, we manually searched rel-
evant academic journals including European Journal of Social Psy-
chology,International Journal of Advertising,Journal of Advertising,
Journal of Advertising Research,Journal of Applied Psychology,Journal
of Applied Social Psychology,Journal of Behavioral Decision Making,
Journal of Business Research,Journal of Consumer Psychology,Journal
of Consumer Research,Journal of Experimental Social Psychology,
Journal of Marketing,Journal of Marketing Research,Journal of Prod-
uct & Brand Management, and Organizational Behavior & Human
Decision Processes for the years 1980–2012. Reviewing the refer-
ence sections in the papers identified in these two initial search ef-
forts uncovered additional studies from articles, books, and
dissertations. We then conducted a forward citation search of the
seminal papers in this topic area.
As final measures to locate as many relevant manuscripts as
possible, we conducted keyword searches of appropriate electronic
databases and also requested known research on attribute framing
via academic list serves (e.g., ELMAR) to identify any additional
studies potentially missed in our earlier data collection efforts.
After identifying these studies, the appropriateness of each one
to our research focus was evaluated. Studies were deemed eligible
if the following two conditions were met: (1) the study focused on
1
While CLT suggests that social, spatial, experiential, and perspective distances are
also potential moderating factors, the articles comprising our dataset provided either
insufficient information or sample size to include these factors.
98 T.H. Freling et al. / Organizational Behavior and Human Decision Processes 124 (2014) 95–109
the relationship between attribute framing and respondent out-
comes; and (2) either the correlation between the attribute fram-
ing effect and the outcome effect or a statistical equivalent was
reported (see Glass, McGaw, & Smith, 1981; Janiszewski, Noel, &
Sawyer, 2003). We acknowledge that while unpublished manu-
scripts could provide additional sources of data, the mature nature
of the attribute framing literature coupled with the often less rig-
orous nature of some working and conference papers led us to err
on the side of rigor and include only peer reviewed published stud-
ies in this research. A less developed literature stream might elicit
a different approach.
2
Following existing taxonomies in the CLT literature (e.g., Trope
et al., 2007), we categorized outcome effects into three distinct cat-
egories: evaluations (including attitudes and perceptions regarding
the stimuli), estimates (predictions based on the stimuli), and
behaviors (including choice and behavioral intentions related to
the stimuli). To correctly categorize the frame under investigation
as an attribute frame, and to build an appropriate dataset, we used
extant literature as a guide (Krishnamurthy et al., 2001; Levin et al.,
1998, 2002). We followed accepted meta-analytic procedures (see
Eysenck, 1978; Sharpe, 1997) in establishing decision rules for
determining which articles would comprise our dataset, paying
special attention meta-analytic experts’ warning against ‘‘compar-
ing apples to oranges.’’ This resulted in our exclusion of studies
examining the effects of other types of frames (i.e., goal framing
and risky choice framing), including investigations involving
games of chance and monetary lotteries that differed in terms of
associated outcome risk.
We also consulted a risky choice framing meta-analysis (Küh-
berger, 1998) for more specific guidance on inclusion criteria relat-
ing to experimental task characteristics (e.g., response mode, unit
of analysis, manipulation quality). First, with respect to response
modes elicited in attribute framing studies, only studies in which
an attribute frame addresses features of one object and elicit a
reaction from one subject were included. This decision was based
on research suggesting that presenting two options and forcing
subjects to make a choice artificially enlarges differences between
conditions (Pany & Reckers, 1987; Perner, Gschaider, Kühberger, &
Schrofner, 1999). Second, in regard to manipulation quality, we ap-
plied a strict definition of framing as a ‘‘semantic manipulation of
prospects whereby the exact same situation is simply redescribed’’
(Kühberger, 1998; p. 24). As such, studies featuring ‘‘loose’’ framing
manipulations, where other individual factors and contextual fea-
tures of a situation might result in frames that were not logically
equivalent or directly comparable were excluded. Finally, Kühber-
ger (1998) notes that experiments conducted at the group level are
heterogeneous and introduce additional social and procedural fea-
tures that may lead to a frame being construed differently. Given
this, only studies requiring individuals to evaluate the stimuli
and eliminated any group decision-making scenarios in which sub-
jects interacted with other research participants to arrive at a con-
sensual decision were included. Two of the authors who were blind
to the hypotheses independently coded the data. Inter-coder
agreement was 95.8% with discrepancies rectified through discus-
sion and reference to the coding scheme and confirmation from a
third, independent referee. In all, 107 studies containing 604 fram-
ing effects and 88,326 observations were retained for analysis (see
Appendices A–C for dataset development specifics).
Overview of meta-analytic procedures
We employed analytic techniques prescribed by Hunter and
Schmidt (2004) in data collection and analysis. The effect size coded
for the analyses is the point biserial correlation coefficient, which is
an appropriate metric for our research in that it provides a means
for easy interpretation and meaningful comparison across the effect
sizes reported in the attribute framing literature (Hunter &
Schmidt, 2004). More importantly, it is the dominant metric found
across the population of attribute framing studies, given that the
manipulation of the frame is dichotomous in nature (e.g., positive
vs. negative, self vs. others, now vs. future) while outcome variables
are inherently continuous (i.e., evaluations, estimates, and behav-
iors). We identified potential outliers within our dataset using the
sample-adjusted meta-analytic deviancy (SAMD) statistic (Huffcutt
& Arthur, 1995). In identifying outliers, this procedure uses a boot-
strapping technique where the overall sample-size weighted or
reliability-corrected between attribute framing
3
and its respective
outcomes is calculated k1 times to understand if one sample is
biasing the analysis. This analysis detected a single observation as
an outlier, which was subsequently removed from the dataset.
We additionally calculated the estimated correlation (r
W
) be-
tween the attribute frame and outcomes associated with attribute
framing. To calculate this overall correlation, each study was
weighted by its corresponding sample size. When reported, each
was further corrected for systematic variance. The results were
then averaged across all studies to ensure that sampling error is ac-
counted for in the estimate of the overall effect of framing. From
this, we calculated the average study variance (var
t
) and an esti-
mate of the heterogeneity (i.e., chi-square statistic) across ob-
served effect sizes within our dataset to ascertain the amount of
variance within our observed effects that is explained by sampling
error and study artifacts (see Hunter & Schmidt, 2004).
To help in the interpretation of the significance of the correla-
tion between attribute framing and behaviors, estimates and eval-
uations, we computed the 95% bootstrapped confidence interval
(CI
BS
) and the 80% credibility interval (CI) for each framing relation-
ship. Since collective data often violate the distributional assump-
tions of parametric tests, the use of bootstrapped confidence
intervals that are based on a non-parametric distribution is appro-
priate and provides a more powerful estimate than traditional con-
fidence intervals (Rosenberg, Adams, & Gurevitch, 2000). Finally,
the fail-safe sample size (N
FS
) was calculated to assess the possibil-
ity of publication bias or the ‘‘file-drawer’’ problem (Rosenthal,
1979). This information estimates the number of unpublished
studies with an effect size of zero that would have to exist to ren-
der the observed effects non-significant at the alpha = .05 level
(Janiszewski et al., 2003). A larger N
FS
value conveys greater confi-
dence in the robustness of obtained results.
Coding procedures
Prior studies employing meta-analysis suggest four potential
sources of variation among observed effect sizes within a research
domain: research context; measurement method; estimation pro-
cedure; and model specification (Assmus, Farley, & Lehmann,
1984; Sultan, Farley, & Lehmann, 1990). Given that our analysis
uses correlation as the metric of the observed effects (which are
2
Following Geyskens, Steenkamp, and Kumar (1999), we elected to only include
published, peer-reviewed research in this analysis to maximize the empirical rigor of
our data population. The rationale for this decision was that unpublished studies have
not undergone the same rigorous review process as published studies, and that efforts
by other meta-analysts to uncover unpublished work have not yielded much success
(cf. Krashnikov & Jayachandran, 2008; Palmatier, Dant, Grewal, & Evans, 2006).
3
When reported in original studies, we use the reliability-corrected mean (sample-
size weighted mean corrected for systematic variance due to variability in the
reported reliability of the measure) under the assumption that correlations from
larger samples (central limit theorem) and estimated from more reliable data produce
a mean correlation closer to the true population mean. When the reliability-corrected
mean cannot be estimated due to the absence of reliability data in the original studies,
we use the next most rigorous estimate of the population mean, the sample-size
weighted mean.
T.H. Freling et al. / Organizational Behavior and Human Decision Processes 124 (2014) 95–109 99
unaffected by model specification) and that the model’s estimation
procedure is invariant, we restricted the examination of modera-
tors to those relating to research context and measurement meth-
od as possible explanations for differences in attribute framing
effects. Based on this, we identified variables that are theoretically
justifiable as potential moderating factors and that could be coded
from the extant studies.
With respect to the research context of attribute framing studies,
we used the construal level theory (CLT) literature as a guide to de-
velop a conceptually-focused coding scheme. We first coded which
type of frame each study utilized. Valence frames entail presenting
otherwise identical information in a positive vs. negative format, as
when Jain, Lindsey, Agrawal, and Maheswaran (2007) describe the
advertised toothpaste brand as ‘‘better than’’ the comparison brand
or the comparison brand of toothpaste as ‘‘worse than’’ the adver-
tised brand. Other studies, such as Shu and Gneezy (2010), employ
temporal frames by varying the time context of the information
presented (e.g., measuring redemption intentions for gift certifi-
cates expiring soon vs. sometime in the future). We also coded nu-
meric frames like those used by DelVecchio, Krishnan, and Smith
(2007), which gauged consumer reactions to ‘‘cents off’’ vs. ‘‘per-
cent off’’ promotional offers for shampoo. Finally, our coding
scheme included referent frames which presented identical infor-
mation from different perspectives. To illustrate, White and Peloza
(2009) measure consumer attitudes toward ads for charity that
benefit the self vs. others.
We coded the construal level of the information presented in
the attribute frame for each study (i.e., whether the frame con-
tained predominantly abstract information vs. concrete informa-
tion) by adapting coding schemes recently employed by other
CLT researchers (cf. Bornemann & Homburg, 2011; Magee, Millik-
en, & Lurie, 2010). Specifically, we carefully reviewed and rated
each frame on a 5-point scale, where 1 = very concrete,2=some-
what concrete,3=both concrete and abstract,4=somewhat abstract,
and 5 = very abstract. (Intercoder reliabilities for construal level
coding were
a
= .97.) We then had two independent coders evalu-
ate and rate the few frames that initially received a ‘‘3’’ rating as
either more abstract than concrete,more concrete than abstract,or
equally concrete and abstract. This allowed us to develop a high con-
strual level (i.e., more abstract) and low construal level (i.e., more
concrete) for analysis. (Results for this phase of coding was also
highly reliable:
a
= .90.)
Additionally, using the information available for each study, we
coded four psychological distance variables: temporal distance (i.e.,
whether the decision scenario required an immediate vs. future
response), hypothetical distance (i.e., whether the decision scenario
was likely vs. unlikely for subjects), affective distance (i.e., whether
or not the decision scenario was emotionally intense), and
informational distance (i.e., whether the decision scenario involved
an event that was familiar vs. novel to subjects). We treated these
research context moderators as theoretical moderators and test
specific predictions regarding how their interplay impacts the
effectiveness of attribute framing.
We included several measurement factors, recording informa-
tion about each study’s sample composition (i.e., whether the sam-
ple contained students or nonstudents, US or international
subjects, and women only or males and females). Additionally,
we coded details about the experimental task (i.e., whether sub-
jects performed a product-related task or some other decision
task). Given that measurement factors are less theoretically inter-
esting and practically important, we treated these characteristics
as control variables in the meta-analysis in that they were included
in our GLS regression (see Lynch, 1982; Peterson, 2001) and focus
discussion around the substantive research context moderators
featured in our hypotheses.
Moderator analysis procedures
To explore the influence of moderators in explaining the effects
of attribute framing on its correlates, a weighted generalized least
squares regression (GLS) approach was used (Geyskens et al., 1999;
Lipsey & Wilson, 2001). We used the following equation to esti-
mate the impact of our proposed moderators on each framing ef-
fect separately:
b
¼ðX
0
R
1
XÞ
1
X
0
R
1
d
where dis the transformed correlation associated with the framing
effect coded from the dataset (Raudenbush, Becker, & Kalaian,
1988), Xis the matrix of moderators hypothesized to influence
these framing effects (and included both research context and our
control moderators together), and
R
is a diagonal vector of the var-
iance assigned to each observation (from the sample size of each
study included in our dataset).
However, given that our study focuses on three separate out-
comes of attribute framing, utilizing this univariate approach has
some limitations in that it ignores the potential correlation, or
within study effects, that could be present among the three fram-
ing outcomes (i.e., evaluations, estimates, and behaviors) (Riley,
2009; Riley, Thompson, & Abrams, 2008). Theory suggests that
there is a relationship between subject evaluations and behavior,
therefore suggesting the need to account for within study correla-
tion. Research suggests that one can explore the potential bias of
within study correlation through a series of sensitivity analyses.
According to Riley (2009), ‘‘the most popular approach [to examin-
ing the impact of within study correlation] is a sensitivity analysis
over an (informed) range of imputed correlations; for simplicity
this usually assumes a common within study correlation across
studies’’ (pp. 807–808).
To help in the selection of the appropriate range of within cor-
relations among evaluations and behaviors to use in our sensitivity
analysis (i.e., to make sure that the presence of within correlations
did not bias the results), we were guided by Kim and Hunter (1993)
which found that the average correlation between attitudes and
behavioral intentions is 0.65 (uncorrected weighted mean) and
that across the various moderator analyses, this correlation ranged
from 0.46 to 0.91. They also found that the average correlation be-
tween attitudes and behaviors is 0.47 (uncorrected weighted
mean) and across the various moderator analyses, the correlation
ranged from 0.26 to 0.86. We use these values to estimate the im-
pact of within study correlation in our dataset.
Taking into account the relationship between evaluations and
behaviors, we utilized the methodology by Raudenbush et al.
(1988) where the intercorrelation that is present among outcomes
effects when they are studied together within the same sample is
specified. To model the interdependencies present within these
data, we first computed the variance–covariance matrix for each
sample (
R
i
) and analyzed these together using a full block diagonal
matrix in our analysis on framing effects (
R
)(Raudenbush et al.,
1988). Following the procedures set forth by Raudenbush et al.
(1988), we transformed the point biserial correlation to the dsta-
tistic and use the following equation to understand the impact of
our proposed moderators on attribute framing relationships using
GLS estimation:
d¼Xbþe;
where bis estimated as:
b
¼ðX
0
R
1
XÞ
1
X
0
R
1
d;
and b
*
is estimated via the variance–covariance matrix with the
following:
100 T.H. Freling et al. / Organizational Behavior and Human Decision Processes 124 (2014) 95–109
V
b
¼ðX
0
R
1
XÞ
1
:
We ran the analysis four times using the following estimations
of the evaluation–behavior relationship in our calculation of the
variance–covariance matrix for the impact of framing on evalua-
tions and behaviors: r= .25, r= .45, r= .65, and r= .85.
Results
Main effects
Table 1 provides an overview of the association between attri-
bute framing and the three outcome effects (i.e., 107 studies, 604
framing effects, 88,326 individual observations). In addition to
reporting estimates of the mean true score correlations, it is also
important in meta-analysis to describe variability in the correla-
tions. Accordingly, we report 80% credibility intervals and 95% con-
fidence intervals around the estimated population correlations.
While some meta-analyses report only confidence intervals (e.g.,
Ernst Kossek & Ozeki, 1998) and others report only credibility inter-
vals (e.g., Vinchur, Schippmann, Switzer, & Roth, 1998), it is impor-
tant to report both because each tells different things about the
nature of the correlations (Judge & Piccolo, 2004). Confidence inter-
vals provide an estimate of the variability around the estimated
mean correlation. A 95% confidence interval excluding zero indi-
cates that one can be 95% confident that the average true correla-
tion is non-zero (5% of average correlations would lie beyond the
upper limit of the distribution). Credibility intervals provide an esti-
mate of the variability of individual correlations across studies. An
80% credibility interval excluding zero indicates that 90% of the
individual correlations in the meta-analysis exclude zero (for posi-
tive correlations, 10% are zero or less and 10% lie at or beyond the
upper bound of the interval). Thus, confidence intervals estimate
variability in the mean correlation, whereas credibility intervals
estimate variability in the individual correlations across the studies.
As shown, the sample size-weighted and reliability-corrected
correlation between attribute framing and evaluations outcomes
is .25, indicating a medium-sized statistical relationship (Rosenthal
& Rosnow, 2008). The 95% bootstrapped confidence interval
around the mean correlation ranges from .22 to .28 and the
fail-safe sample size (N
FS
= 413,085) suggests that there is little
indication of publication bias. Similarly, the adjusted correlation
between attribute framing and estimates outcomes is .24, also
indicative of a medium-sized relationship. The 95% bootstrapped
confidence interval around the mean correlation ranges from .20
to .29 and the N
FS
= 16,467. The adjusted correlation between attri-
bute framing and behaviors outcomes is .21, with a 95% confidence
interval of .19–.23 and a fail-safe sample size (N
FS
= 91,196) that
indicates file drawer effect is not an issue for this relationship.
The statistical evidence presented in Table 1 suggests that attri-
bute framing has a positive and statistically significant impact on
evaluations, estimates, and behaviors outcomes. These results both
expand and confirm previously published research findings. Yet
the data range, variance, and heterogeneity indicate that an inves-
tigation of potential moderating variables is warranted. As such,
we next examine factors that might attenuate or mitigate the rela-
tionships between attribute framing and the three outcome effects.
Moderator effects
Table 2 highlights the influence of both theoretical and control
variables in moderating each of the three outcome effects.
Regression results reveal that the correlations observed in prior
research between attribute framing and key outcomes are signifi-
cantly impacted by several moderators. Furthermore, results from
our sensitivity analyses demonstrate that these effects are robust
against varying levels of within study correlation among evalua-
tions and behaviors. The univariate GLS regression analysis results
are presented next to highlight the impact of the theoretical
moderators.
Hypothesis testing
Results indicate that the construal level of the attribute frame is
a statistically significant moderator of its impact on evaluations
(ß= .10, p6.05), estimates (ß= .19, p6.05), and behaviors
(ß= .04, p6.05). These results also hold true for each of the psy-
chological distance moderators across all three criterion variables,
thus providing initial support for the view that psychological dis-
tance moderates the relationship between framing and outcome
effects. To fully evaluate our hypotheses, which predict stronger
framing effects when there is congruence between the attribute
frame’s construal level and the subject’s perceived psychological
distance from the framed event, we conducted additional analyses.
Specifically, we examined the relationship between construal level
and psychological distance for the high vs. low level of each dimen-
sion across each outcome effect variable. Results appear in Table 3.
Table 2
Moderator results for attribute framing.
Factor Outcome effect – estimated separately Evaluations and Behaviors – estimated together modeling within-study variance at
different levels
a
Evaluations Estimates Behaviors r= 0.25 r= 0.45 r= 0.65 r= 0.85
Construal level
High vs. Low .10
*
.19
*
.04
*
0.03
*
0.03
*
0.04
*
0.21
*
Psychological distance
High vs. Low Temporal distance .04
*
.21
*
.22
*
0.54
*
0.53
*
0.50
*
0.63
*
High vs. Low Probability distance .44
*
.07
*
.30
*
0.30
*
0.30
*
0.33
*
0.89
*
High vs. Low Affective distance .14
*
.24
*
.34
*
0.02
*
0.01
+
0.02
*
0.04
*
High vs. Low Informational distance .24
*
.14
*
.13
*
0.14
*
0.14
*
0.16
*
0.57
*
Control variables
Frame type .29
*
.08
*
.20
*
0.17
*
0.18
*
0.20
*
0.14
*
Subject composition .06
*
.13
*
.15
*
0.12
*
0.12
*
0.12
*
0.33
*
Geographic composition .45
*
.05
*
.03
*
0.29
*
0.31
*
0.35
*
0.68
*
Gender composition .35
*
– .94
*
0.26
*
0.30
*
0.35
*
0.52
*
Experimental task .62
*
.16
*
.23
*
0.64
*
0.67
*
0.72 0.91
*
Number of observations 359 69 175
a
To address the potential for interdependence among our dependent variables, we specified the correlation among Evaluations and Behaviors at different levels (i.e., r= .25,
r= .45, r= .65, and r= .85) and ran a series of robustness checks to ensure that interdependence was not an alternate explanation of our findings.
*
p6.05.
+
p6.10.
T.H. Freling et al. / Organizational Behavior and Human Decision Processes 124 (2014) 95–109 101
In support of hypothesis 1, studies featuring framing manipula-
tions with CLT congruency between the psychological distance of
the framed issue and the construal level of information described
have significantly stronger effects. This is true across all four
dimensions of psychological distance coded and analyzed includ-
ing: temporal distance (evaluations: r= .30 vs. r= .28; estimates:
r= .33 vs. r= .22; and, behaviors: r= .30 vs. r= .19); hypothetical
distance (evaluations: r= .28 vs. r= .25; and, behaviors: r= .23 vs.
r= .19); affective distance (evaluations: r= .27 vs. r= .21; estimates:
r= .37 vs. r= .23; and, behaviors: r= .21 vs. r= .20); and, informa-
tional distance (evaluations: r= .27 vs. r= .23; estimates: r= .28
vs. r= .16; and, behaviors: r= .22 vs. r= .19. In sum, these results
lend support to our hypothesis and suggest the interactive effect
between construal level and psychological distance determines
the relative effectiveness of attribute frames.
To test hypothesis 2, we culled 209 valenced attribute frame
observations from our dataset and conducted additional univariate
analyses on this data.
4
We sought to determine if positive attribute
frames (theorized to evoke a higher construal level) are more effec-
tive when greater psychological (temporal, hypothetical, affective,
and informational) distance characterizes the decision scenario. Re-
sults support both our expectations and recent CLT-based theorizing
on information valence. Specifically, we found statistically signifi-
cant stronger correlations (all p6.05) between attribute framing
and evaluations in studies characterized by higher temporal distance
(r= .37 vs. r= .23), higher hypothetical distance (r= .32 vs. r= .25),
higher affective distance (r= .31 vs. r= .26), and higher informational
distance (r= .30 vs. r= .25). Taken together, these results indicate
that the long-established valence-consistent shift found in attribute
framing research is likely a reliable phenomenon (Levin et al., 2002)
due to the interaction between the message and its recipient instead
of an outcome effect solely emanating from the constructed
message.
Analysis of control moderators
Table 4 details the influence of control moderators on each out-
come effect. For all three outcome variables, studies using a valen-
ced decision frame have positive, statistically significant results
(p6.05). Results for evaluations (ß= .29, p6.05), estimates
(ß=.08, p6.05) and behaviors (ß= .20, p6.05) indicate that
Table 3
Weighted univariate results for theoretical moderators.
Sample size Number of observations Mean effect size Weighted variance
High temporal distance
a
Evaluations
*
High construal level 8184 66 .3024 .074
Low construal level 6504 62 .2817 .061
Estimates
*
High construal level 2694 28 .3293 .037
Low construal level 4352 27 .2184 .007
Behaviors
*
High construal level 11,046 69 .2979 .031
Low construal level 7456 47 .1897 .007
High hypothetical distance
b
Evaluations
*
High construal level 5503 52 .2816 .087
Low construal level 7037 58 .2469 .045
Behaviors
*
High construal level 5300 47 .2290 .017
Low construal level 4334 25 .1917 .008
High affective distance
Evaluations
*
High construal level 7305 64 .2723 .061
Low construal level 10,459 75 .2142 .037
Estimates
*
High construal level 1944 21 .3733 .040
Low construal level 3460 24 .2253 .015
Behaviors
*
High construal level 7305 52 .2064 .015
Low construal level 10,459 50 .1977 .011
High informational distance
Evaluations
*
High construal level 8479 71 .2660 .019
Low construal level 7513 63 .2263 .025
Estimates
*
High construal level 3571 28 .2829 .026
Low construal level 2575 15 .1601 .010
Behaviors
*
High construal level 8006 51 .2247 .020
Low construal level 575 38 .1894 .013
*
p6.05.
a
For reporting simplicity, each psychological distance reported indicates the ‘‘high’’ observations. Analysis of the corresponding ‘‘low’’ observations reveals perfectly
inverse relationships to those demonstrated above.
b
Insufficient sample sizes for high and low construal levels across Estimates outcomes precluded a statistical examination.
4
Only the relations hip between valenced attribute frames and evaluations
(N= 209) were included in these univariate analyses. We also explored the
relationships between valenced attribute frames and estimates (N= 15) and behav-
iors (N=96). The same pattern of relationships (i.e., all more positive, all p6.05)
were found, so these results are not reported in the text.
102 T.H. Freling et al. / Organizational Behavior and Human Decision Processes 124 (2014) 95–109
studies where the attribute frame involves a valence manipulation
are significantly different than studies where valence is not manip-
ulated. Results reveal that the correlation between attribute fram-
ing and evaluations is statistically stronger in studies manipulating
valence (r= .26), as compared to those where there is no valence
manipulation (r= .22). Valence frames have a similar effect on esti-
mates (r= .24 vs. r= .24) and behaviors (r= .21 vs. r= .21).
Inconsistencies across the observed attribute framing correla-
tions can also be explained by differences in the research design,
samples, experimental stimuli, and measures used in studies com-
prising our dataset. For evaluations, attribute framing effects are
significantly different for studies utilizing student vs. non-student
subjects (ß= .06, p6.05), US respondents vs. non-US respondents
(ß= .45, p6.05), samples comprised of mixed vs. single gender
(ß= .35, p6.05), and when subjects performed a product-related
evaluation vs. another experimental task (ß=.62, p6.05). Corre-
lations are substantially stronger between attribute framing and
evaluations in studies with respondents who were students
(r= .25 vs. r= .21), American (r= .26 vs. r= .20), and mixed gender
(r= .26 vs. r= .08), as well as when the experimental task did not
involve a product-related evaluation (r= .22 vs. r= .29).
Significant moderators of attribute framing effects on estimates
include student vs. non-student respondents (ß=.13, p6.05), US
vs. non-US subjects (ß=.05, p6.05), and product evaluation vs.
non-product evaluation (ß=.16, p6.05). Substantially stronger
correlations between attribute framing and estimates were ob-
served for studies that used non-student (r= .29 vs. r= .23) and
non-US respondents (r= .34 vs. r= .23), and that involved a non
product-related experimental task (r= .26 vs. r= .20). For behav-
iors, attribute framing effects were significantly different for stud-
ies utilizing student vs. non-student subjects (ß= .15, p6.05), for
studies with US respondents vs. non-US respondents (ß= .03,
p6.05), for studies utilizing samples comprised of mixed vs. single
gender (ß= .94, p6.05), and for studies in which subjects per-
formed a product-related evaluation vs. another experimental task
(ß=.23, p6.05). Stronger correlations between attribute framing
and behaviors were observed in studies with samples comprised of
students (r= .22 vs. r= .18), US subjects (r= .22 vs. r= .20), and
mixed genders (r= .21 vs. r= .14), and when the experimental task
focused on a non-product-related evaluation (r= .19 vs. r= .22).
Follow-up experiment
While meta-analysis is extremely useful for quantitatively syn-
thesizing an immense literature base, the technique is limited to
including only variables that can be meaningfully and consistently
coded from existing studies. Further, because the relationships
examined are correlational, causal interpretations should be made
with caution. These methodological limitations prohibit us from
making more definitive conclusions about the impact of valence-
based attribute frames on consumer evaluations and the construal
level mechanisms that we propose drive these effects. To rectify
this, we conduct a follow-up experiment to supplement findings
Table 4
Weighted univariate results for control moderators.
Sample size Number of observations Mean effect size Weighted variance
Valence frame
a
Evaluations
*
30,796 209 .2648 .046
20,869 150 .2163 .026
Estimates 1675 15 .2446 .025
8110 55 .2394 .018
Behaviors
*
14,631 96 .2111 .022
12,245 79 .2067 .011
Subject composition
b
Evaluations
*
42,990 308 .2536 .040
8675 51 .2087 .032
Estimates
*
8247 59 .2318 .022
1538 11 .2854 .003
Behaviors 20,809 142 .2174 .006
5971 33 .1806 .014
Geographic composition
c
Evaluations
*
41,968 291 .2566 .042
9697 68 .2006 .023
Estimates
*
8598 59 .2254 .015
1187 11 .3371 .053
Behaviors
*
18,361 120 .2155 .016
8419 55 .1952 .019
Gender composition
d
Evaluations
*
47,265 348 .2620 .040
4400 12 .0772 .003
Behaviors
*
26,225 172 .2107 .017
555 3 .1355 .004
Experimental task
e
Evaluations
*
30,158 176 .2184 .023
21,507 183 .2852 .061
Estimates
*
3630 23 .2031 .009
6155 47 .2623 .025
Behaviors
*
10,110 59 .1929 .010
16,670 116 .2189 .021
*
p6.05.
a
For each outcome effect, the first line represents studies manipulating valence while the second represents studies that did not.
b
For each outcome effect, the first line represents student subjects while the second represents non-students.
c
For each outcome effect, the first line represents US subjects while the second represents non-US.
d
For each outcome effect, the first line represents mixed gender while the second represents single gender; for estimates outcomes, there was no variation in gender
composition to allow for comparison.
e
For each outcome effect, the first line represents product-related while the second represents non product-related.
T.H. Freling et al. / Organizational Behavior and Human Decision Processes 124 (2014) 95–109 103
emanating from the meta-analysis. Specifically, we explore how
frame valence interacts with social distance—one of the original
dimensions of psychological distance that could not be coded in
the preceding meta-analysis.
Social distance is defined as an individual’s perception of whom
an event references (e.g., one’s self or various others) and the de-
gree of similarity (i.e., ‘‘close’’ or ‘‘distant’’) between that individual
and any others relevant to the event. The further away the event is
perceived from the individual subject, the higher the perceived so-
cial distance. According to construal level theory, individuals natu-
rally regard certain people as being closer to them than other
people (Trope et al., 2007). By definition, the ‘‘self’’ is the most so-
cially proximal entity, ‘‘similar others’’ are more socially close than
‘‘dissimilar others’’, and ‘‘in-group’’ members are more socially
close than ‘‘out-group’’ members (Bar-Anan, Liberman, & Trope,
2006). Thus, an event that is framed as directly involving the sub-
ject would be construed as having a low-level of social distance,
one involving friends or family would be at a higher level, and an
event framed as involving strangers at an even higher perceived le-
vel of social distance.
Consistent with hypothesis 2, we expect stronger framing ef-
fects to occur when individuals in a socially close decision situation
receive information at a lower-level construal and when individu-
als in a socially distant decision situation are presented informa-
tion at a higher-level construal. Again, we conceptualize
information valence as a manifestation of construal level where
positive information evokes a higher level of construal and nega-
tive information leads to a lower construal level. In line with our
meta-analysis results and recent CLT research exploring valence
(Lee et al., 2010; Ulkumen & Cheema, 2011; White et al., 2011),
we expect stronger framing effects for congruent pairings (e.g., a
positively valenced frame for a socially distant scenario), as com-
pared to incongruent pairings (e.g., a positively valenced frame
for a socially close scenario).
Method
A 2 (frame valence: positive vs. negative) 2 (social distance:
high vs. low) between-subjects factorial design was used to test
our hypothesis. One hundred undergraduate students from a large
southwestern university (marketing majors and graduating se-
niors) participated in the study for course credit. Participants were
told that the purpose of the study was to understand how individ-
uals similar to them make decisions.
Following Liviatan et al. (2008), a decision scenario in which
subjects were asked to evaluate a target individual for inclusion
in a group project for class was designed. Social distance (high
vs. low) was manipulated by varying the target person’s similarity
to the subject on many key dimensions, including major, classifica-
tion, and the two most recent semesters of coursework completed
at this university. To manipulate frame valence, we also stated that
one student in the group had worked with the target person on an-
other group project. Consistent with Buda and Zhang (2000),we
included information from this one student that ‘‘four out of six
group members rated the target person positively’’ vs. ‘‘two out
of six group members rated the target person negatively.’’
To develop the stimulus materials, we extensively pretested to
understand students’ criteria in choosing group members for pro-
ject work as well as to assess students’ knowledge of information
presented in the scenarios and their perceived relevance of and
distance from the decision context. Importantly, pretest results
suggested that the manipulation of social distance was not associ-
ated with significant differences in other source characteristics
such as perceptions of the target person’s trustworthiness, credibil-
ity, or attractiveness (Fs < 1). Based on pretest results, four versions
of a decision scenario were developed (see Appendix D) that varied
frame valence and social distance. Following exposure to their
respective decision scenarios, subjects rated the likelihood that
they would invite the target person to join the group using four
7-point semantic differential items (anchored by ‘‘very like-
ly’’...‘‘not at all likely’’, ‘‘very probable’’...‘‘not at all probable’’,
‘‘very possible’’...‘‘not at all possible’’, and ‘‘very certain’’...‘‘not
at all certain’’). Consistent with prior research assessing intentions
(e.g., Bennett & Harrell, 1975; Dover & Olson, 1977; MacKenzie,
1986; Marks & Kamins, 1988; Smith & Swinyard, 1983), an average
of the scale items was used to form a composite behavioral inten-
tion measure.
We included manipulation checks for social distance, valence,
and construal style. For the social distance manipulation check, par-
ticipants indicated how similar and close to themselves they per-
ceived the target person to be using a 7-point response scale
ranging from ‘‘not at all’’ to ‘‘very much’’. These two social-distance
items were reverse-scored for analysis so that greater perceived
similarity corresponded to lower social distance. As an additional
social distance manipulation check, we also had participants com-
plete the Inclusion of Others in Self Scale (IOSS;Aron, Aron, & Smol-
lan, 1992), which measures interpersonal closeness. We followed
established procedures (see Block & Keller, 1995; Maheswaran &
Meyers-Levy, 1990) to assess the effectiveness of our frame valence
manipulation. Specifically, we asked subjects to indicate the extent
to which the information they read portrayed the target person pos-
itively and whether the information they read presented the target
person’s past group project performance in a negative light (both
with 7-point scales ranging from ‘‘not at all’’ to ‘‘completely’’). We
also included the three confounding-check measures suggested by
Block and Keller (1995) by asking subjects to rate the information
presented on 7-point scales anchored by ‘‘very credible’’.... ‘‘not at
all credible’’, ‘‘easy to comprehend’’...‘‘difficult to comprehend’’,
‘‘contained a lot of information’’...‘‘contained little information’’.
Finally, we assessed each participant’s construal style, which we
expected to correspond with frame valence. Subjects completed
the Behavioral Identification Form (BIF; see Alter, Zemla, & Oppen-
heimer, 2010; Vallacher & Wegner, 1989), which indicates an indi-
vidual’s preferences for either abstract of concrete description of
thirteen everyday behaviors. For example, subjects indicated their
relative preference for the description of eating as ‘‘chewing and
swallowing’’ (concrete, low-level construal) or ‘‘getting nutrition’’
(abstract, high-level construal). To eliminate any presentation or-
der bias, we alternated which label (Description A or Description
B) referred to the concrete and abstract descriptions. Participants
indicated their relative preference for the two descriptions on a
7-point scale (anchored by ‘‘strongly prefer Description A’’ and
‘‘strongly prefer Description B’’). We then averaged the resulting
preference ratings to generate a score reflecting each participant’s
construal style.
Results
Manipulation checks
To ensure that we correctly induced the intended levels of va-
lence and social distance, we first checked the study’s manipula-
tions. Subjects’ responses to the questions relating to their status
as a college student verified that all were enrolled full-time and
had taken the marketing classes in the socially close scenario
(but not the non-marketing classes in the socially distant scenario).
The two measures assessing participants’ perceived similarity and
closeness to the target person both revealed statistically significant
main effects for social distance. Respondents’ ratings were lower
when social distance was high for both similarity (M
socially close
=
4.23 and M
socially distant
= 3.27; F(1, 98) = 25.29, p6.001) and for
closeness (M
socially close
= 4.32 and M
socially distant
= 3.21;
F(1,98) = 18.00, p6.001). As expected, participants who evaluated
104 T.H. Freling et al. / Organizational Behavior and Human Decision Processes 124 (2014) 95–109
the socially distant decision scenario scored significantly lower
(M= 2.98) on the IOSS than those exposed to the socially close
decision scenario (M= 5.12; t(99) = 11.78, p6.001). The two items
assessing the frame valence manipulation both revealed a statisti-
cally significant main effect for frame valence. Subjects rated the
extent to which the message positively portrayed the target person
higher in the positive frame valence condition (M= 5.93) than in
the negative frame valence condition (M= 4.80; F(1, 98) = 36.41,
p6.001). They also rated the extent to which the message pre-
sented the target person’s past group project performance in a neg-
ative light to be lower in the positive frame valence condition
(M= 4.13) than in the negative frame valence condition
(M= 5.67; F(1,98) = 24.52, p6.001). Importantly, there were no
confounds with any of the manipulations.
With respect to construal level, subjects who were exposed to
the positive frame (high-level construal) preferred the abstract
descriptions of the BIF behaviors more strongly than did
participants who saw the negative frame (low-level construal)
(M
positive frame
= 5.04 and M
negative frame
= 3.11; t(99) = 13.29,
p6.001). Consistent with expectations, exposure to positive
information appeared to lead subjects to prefer the descriptions
consistent with an abstract construal mindset, while exposure
to negative information led participants to prefer descriptions
consistent with a concrete construal mindset.
To rule out any additional unintended confounds, subjects were
asked to indicate how credible, easy to comprehend, and informa-
tive the information presented was. Separate ANOVAs on these
variables revealed no statistically significant treatment effects
(Fs < 1), suggesting that the research treatments were not con-
founded with any of these variables. In summation, the manipula-
tion and confound checks suggest that the intended factors were
successfully manipulated and that our constructs accurately cap-
ture the appropriate underlying dynamics.
Results
A22 ANOVA revealed nonsignificant main effects of social
distance and frame valence (Fs < 1). However, consistent with both
our expectations and hypothesis 2, the interaction between social
distance and frame valence is significant (F(1,98) = 4.37, p6.05).
Stronger framing effects were observed when there was congru-
ence between construal level and social distance.
Table 5 displays the pattern of means for subjects’ behavioral
intentions and highlights support for our expectations. We ana-
lyzed the mean intentions for each frame construal level-social dis-
tance pairing in a series of planned contrasts. As expected, the two
congruent message conditions are associated with the strongest
framing effects. When a positive frame is presented with a socially
distant referent (M= 5.76), intentions are significantly more favor-
able than when subjects see a positive frame with a socially close
referent (M= 3.45; t(38) = 11.45, p6.001) or a negative frame for
a socially distant referent (M= 3.64; t(38) = 10.11, p6.001). Simi-
larly, subjects express significantly more favorable intentions when
a negative frame is presented with a socially close referent
(M= 5.68), as compared to when subjects are exposed to a positive
frame with a socially close referent (M= 3.45; t(38) = 9.23, p6.001)
or a negative frame with a socially distant referent (M= 3.64;
t(38) = 8.67, p6.001). Of particular interest, a planned contrast of
the two congruent message conditions (i.e., positive frame-socially
distant referent vs. negative frame-socially close referent) reveals
no statistically significant differences (t(38) = 0.25, p> .10), sug-
gesting that negatively framed messages can be as effective as pos-
itively framed messages if the issue being evaluated is perceived as
being particularly likely to affect the message recipient.
Discussion
We proposed that congruence between a frame’s construal level
(evoked through valence) and the evaluator’s psychological dis-
tance (via social distance) from the framed event would determine
the effectiveness of that frame. Our experimental results substan-
tiate this expectation by showing that outcome effects are en-
hanced when there is congruence between these two variables.
Both positively framed events presented in a socially distant sce-
nario and negatively framed events presented in a socially close
scenario outperformed incongruent construal-social distance pair-
ings. Importantly, the experimental results reveal that there is no
statistical difference between the two congruent attribute framing
conditions. Despite historical support for a valence-consistent shift
whereby positive frames consistently outperform negative frames,
our findings indicate that it is the congruency between the mes-
sage frame and the message recipient that drives results—and
not simply a positive vs. negative message frame. This is a note-
worthy finding because it suggests that, as long as the construal le-
vel and psychological distance are congruent, both positive and
negative messages can be equally effective.
These results suggest that message framers can choose to in-
voke a particular level of construal among audience members with
their use of positive or negative information in persuasive messag-
ing, depending on the psychological distance that is likely to char-
acterize the issue or event for message recipients. In contrast to the
classic Trident sugarless gum example referenced earlier, consider
a recent ad for the Clearblue Easy pregnancy test kit that uses the
relatively negative tagline: ‘‘1 in 4 women misread a traditional
pregnancy test.’’ In this case, the message is negatively framed to
draw attention to the proportion of individuals who might mistak-
enly interpret test results (rather than the 75% of individuals who
understand the results). This negative positioning is consistent
with a construal level theory view and highlights the importance
of congruence between construal level and social distance. Here,
the presentation of negatively framed information is likely to in-
duce a lower level of construal and should thus resonate with po-
tential target consumers who are psychologically close to this issue
(e.g., females who are sexually active where pregnancy is a
possibility).
For scholars and managers interested in maximizing the effec-
tiveness of attribute frames, these experimental results imply that
when delivering information about a negative event, the greatest
persuasive impact should occur when concrete details are pre-
sented and the event is portrayed as close to the message recipient.
Conversely, positive information should be presented in an ab-
stract manner that accentuates the perceived distance between
the event and the message recipient.
General discussion
Overview
With some recent exceptions, past research examining attribute
frames has largely been concerned with the characteristics of the
message and its impact on consumer outcomes. Our work extends
Table 5
Behavioral intentions toward the new group member.
Low social distance High social distance
Positive frame N=25 N=25
Mean 3.45 5.76
Standard deviation (1.13) (1.02)
Negative frame N=25 N=25
Mean 5.68 3.64
Standard deviation (.97) (1.01)
T.H. Freling et al. / Organizational Behavior and Human Decision Processes 124 (2014) 95–109 105
the current knowledge of framing effects by applying a theoretical
lens to the attribute framing literature. We find strong support for
our assertion that the impact of attribute framing may not be as
straightforward as it is often portrayed. By including a CLT perspec-
tive in our investigation, we empirically demonstrate that it is of-
ten the interaction between the message and the recipient that is
driving evaluations, estimates, and behaviors. These findings are
relevant to a variety of organizational communications designed
to influence behavioral decisions (e.g., leadership, negotiation, pro-
motional). The meta-analytic results for CLT-based theoretical
moderators provide interesting revelations for any individual
tasked with constructing persuasive messages relating to frame va-
lence, construal level, and psychological distance.
Importantly, construal level is one aspect of attribute framing
that is largely under the control of the frame designer. This sug-
gests that the specific language in a message can be purposefully
finessed to strengthen its persuasive impact. By simply considering
how target audience members are likely to regard an issue or event
(in terms of their relative psychological distance from it), one can
then construct a message at an appropriate level of construal to
better achieve a desired result. CLT suggests the psychological dis-
tance of the decision problem affects decision-making, such that
people are predisposed to use high-level construals when thinking
of distant events and low-level construals when thinking of prox-
imal events (Bar-Anan et al., 2006; Liberman, Trope, & Wakslak,
2007). Consistent with this premise, our overall results for the psy-
chological distance variables (see Table 2) provide an initial indica-
tion that these factors do play a role in individuals’ perceptions of
attribute frames. Examining univariate contrasts (see Table 3) pro-
vides an even richer assessment of the proposed relationships.
These key findings echo other CLT research demonstrating stronger
responses when the psychological distance of an object is compat-
ible with its level of abstraction (e.g., Amit, Algom, & Trope, 2009;
Lee et al., 2010).
High temporal distance characterizes situations that are ex-
pected to happen in the future. Our results indicate that when an
event occurs farther in the future, a frame construed in a more ab-
stract manner (i.e., high-level construal) leads to consistently
stronger framing effects. This finding provides converging evidence
that distant future events should be represented in an abstract,
structured manner while messages about relatively near future
events are more effective when represented in a more concrete,
contextualized manner (Liberman et al., 2002; Trope & Liberman,
2000; Trope et al., 2007). Research by Chandran and Menon
(2004) substantiates this relationship between temporal distance
and construal level by demonstrating a higher perceived likelihood
of being affected by Epstein-Barr and cell phone radiation among
subjects exposed to frames detailing the specific number of people
affected each day (vs. each year) and presenting concrete informa-
tion about the symptoms of these health hazards.
High hypothetical distance characterizes situations in which the
subject is unlikely to find oneself. When events are portrayed as
unlikely for the subjects, a similar positive relationship between
construal level and distance is observed. That is, consistent with
prior research (Todorov et al., 2007), attribute framing effects are
stronger when subjects see an abstractly construed frame of an
event that is unlikely to affect them. Kastenmüller et al. (2010) cor-
roborate this assertion, documenting more favorable evaluations of
a terrorism safety policy—a political decision task that represents
an unlikely issue for student subjects to ponder—when information
about the policy is presented in broad, abstract terms.
High affective distance characterizes situations that are rela-
tively less emotional and less intense for subjects. In concert with
previous CLT research on emotional distance (Fujita, Trope, Liber-
man, & Levin-Sagi, 2006b; Labroo & Patrick, 2009), our results indi-
cate that, for events characterized by high affective distance,
frames with a high construal level are most effective. Consistent
with this assertion, subjects in Agrawal and Duhachek’s (2010) re-
search estimate higher incidences of undergraduate binge drinking
after viewing emotionally-charged anti-drinking messages (low in
affective distance) that concretely framed the consequences of
binge drinking (low construal level).
High informational distance characterizes situations or issues
that are likely to be perceived as novel or unique (Fieldler, 2007).
We observed stronger framing effects for messages that use ab-
stract language to describe unfamiliar events. Thus, when develop-
ing an attribute frame for an event that target audience members
will perceive as novel, the response to most decision tasks should
be stronger if the information is construed at a higher level. Consis-
tent with this dynamic, Chan and Mukhopadhyay (2010) observe
higher behavioral intentions to attend a theatrical performance
among student subjects when information about this relatively
unfamiliar concept (high informational distance) is broadly framed
(high construal level).
Although a lack of sufficient extant data meant that social dis-
tance could not be incorporated as a theoretical moderator in the
meta-analysis, we directly examined this psychological distance
variable. In the preceding experiment, we formally manipulated
subjects’ perception of similarity to oneself in valence-based
frames. Results supported predictions that congruence among con-
strual level and frame valence would lead to stronger framing ef-
fects. Subjects expressed more positive evaluations under two
conditions: (1) when a positive frame was used in conjunction
with a socially distant referent; and, (2) when a negative frame
was used in combination with a socially close referent. This finding
further demonstrates that both positive and negative frames can
be equally effective in eliciting evaluations and highlights circum-
stances when the frequently mentioned valence-consistent shift
might not hold. Furthermore, the results of this experiment have
important implications for the optimal way to craft persuasive
messages. For psychologically close issues that are likely to reso-
nate with or concern message recipients, negative frames should
be more effective than positive frames in influencing evaluations.
Contributions
The dual objectives of the current research were to synthesize
and analyze the empirical findings on attribute framing in an effort
to take inventory of existing knowledge and also to apply a con-
strual level theory (CLT) perspective to attribute framing in an ef-
fort to better understand the dynamics at play. This manuscript
first contributes to the literature by making current the empirical
body of knowledge on attribute framing and quantitatively sum-
marizing over three decades of research—much of it published in
Organizational Behavior & Human Decision Processes (see Appendix
A). The application of construal level theory provides evidence that
the attribute framing relationships once described as fairly
straightforward are actually quite intricate. Therefore, another
contribution of the current research is the application of CLT’s core
concepts of construal level and psychological distance to extant
attribute framing research. Results indicate that the long recog-
nized valence-consistent shift document in attribute framing re-
search is likely more than simply a function of positive vs.
negative message positioning, but likely acts via the impact of
the event information on an individual’s perceived construal level
and psychological distance from the event. By manipulating the
mental representation in a framed event, one is able to impact sub-
jects’ evaluations.
In fact, while the meta-analysis clearly demonstrates that attri-
bute framing effects are significantly stronger when messages con-
tain positive frames, a closer analysis of studies manipulating
106 T.H. Freling et al. / Organizational Behavior and Human Decision Processes 124 (2014) 95–109
frame valence suggests the valence-consistent shift described in
earlier summary articles is somewhat conditional. Our univariate
results suggest that positively valenced (higher construal level)
attribute frames are more effective when the psychological dis-
tance of the decision scenario is also higher. This result is consis-
tent with recent research in the CLT tradition (Eyal et al., 2004;
Herzog et al., 2007; Labroo & Patrick, 2009), which intimate that
information valence may be a manifestation of construal level
and share a similar relationship with psychological distance. Our
results add to this developing literature stream and suggest that
the impact of positive attribute framing is more pronounced when
subjects are temporally, hypothetically, emotionally, or informa-
tionally distant from the decision scenario. Our experiment dem-
onstrating a similar relationship between social distance and
frame valence further substantiates the appropriateness of re-
interpreting attribute framing effects using a CLT framework.
Perhaps the most noteworthy contribution of this research is
the finding that the congruence between the evoked construal le-
vel of a framing event and subject’ perceived psychological dis-
tance from that event appears to influence attribute framing
effects. This finding guides scholars and managers as to how to best
frame an event regardless of how far in the future, how likely, how
emotionally intense, or how familiar a scenario is for the individual
interpreting the event. In sum, this current research (1) integrates
and analyzes the body of empirical attribute framing knowledge,
(2) extends the domain of construal level theory, and (3) uncovers
a finer-grained explanation for the valence-consistent shift effect
that is so abundantly observed in attribute framing research.
Limitations and future research
While this manuscript expands the attribute framing knowledge
base, some limitations should be noted. Any quantitative synthesis
is constrained by the nature and scope of the original studies on
which it is based and this shortcoming should be borne in mind
when findings presented here are interpreted. First, not all pub-
lished studies on attribute framing reported correlations or suffi-
cient data to calculate a usable effect size; therefore, some
empirical studies exploring the effects of attribute framing could
not be incorporated into this analysis. Second, the cross-sectional
nature of some of the original studies restricts our ability to make
confident causal inferences. Although time-series data would be
most desirable for these purposes, they are largely unavailable in
the original studies and therefore a reliance on cross-sectional data
for making causal inferences naturally exists in the attribute fram-
ing literature. Third, our analyses were constrained to examining
moderating factors that could be coded from the extant literature.
While the moderating factors studied here provide scholars and
practitioners with useful information, the inability of these
codeable moderators to fully account for the variance in the perfor-
mance outcome correlations indicates that additional measure-
ment and/or contextual factors need to be modeled and reported
in future studies on attribute framing.
Several avenues of future research emanate from our work. As
mentioned previously, this research demonstrates that the effect
of framing on outcomes might not be as straightforward as origi-
nally thought. We provide a good starting point for exploring the
interplay between an attribute frame’s construal level and psycho-
logical distance, showing that the effectiveness of attribute framing
is contingent upon the relationship between the message and the
intended recipient. In this paper, we examined construal level
and each psychological distance variable independently of others.
Future research should examine the interactions of attribute
framing with multiple psychological distance mechanisms
simultaneously and explore the differential effects across the key
outcomes examined.
By far, most attribute framing studies examine the impact of the
frame on evaluations and behaviors; however, some interesting ef-
fects occur when the outcome variables of interest involve esti-
mates. While the focus of this analysis was on theoretical
moderators and not control variables, it is interesting that in stud-
ies which require subjects to make a prediction or estimate, attri-
bute framing is more effective among non-US, non-student
subjects. This finding is at odds with attribute framing studies that
feature evaluations or behaviors as the outcome variable. Future
research would benefit from additional inquiries relating to the ef-
fect of attribute framing on individual predictions.
Finally, while this manuscript appropriately focused solely on
the impact of attribute framing on outcomes, future research
should apply CLT and explore the role of congruence among the
message being framed and the recipient for goal framing, which
frames the relationship between behaviors and goal attainment
(Krishnamurthy et al., 2001). While qualitatively different from
attribute framing, goal framing is also widely used in persuasive
messaging and the vast literature on goal framing could also ben-
efit from a quantitative synthesis and unifying theoretical
framework.
Acknowledgments
The authors thank Ryan Freling, Adwait Khare, Ritesh Saini, and
Zhiyong Yang for constructive comments on previous versions of
this article. They also thank Xiao-Ping Chen and four anonymous
referrers for their insightful comments during the review process.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.obhdp.2013.12.
007.
References
Agrawal, N., & Duhachek, A. (2010). Emotional compatibility and the effectiveness
of antidrinking messages: A defensive processing perspective on shame and
guilt. Journal of Marketing Research, 47, 263–273.
Alter, A. L., Zemla, J. C., & Oppenheimer, D. M. (2010). Missing the trees for the
forest: A construal level account of the illusion of explanatory depth. Journal of
Personality and Social Psychology, 99, 436–451.
Amit, E., Algom, D., & Trope, Y. (2009). Distance dependent processing of pictures
and words. Journal of Experimental Psychology: General, 138(3), 400–415.
Aron, A., Aron, E. N., & Smollan, D. (1992). Inclusion of other in the self scale and the
structure of interpersonal closeness. Journal of Personality and Social Psychology,
63, 596–612.
Assmus, G., Farley, J. U., & Lehmann, D. R. (1984). How advertising affects sales:
Meta-analysis of econometric results. Journal of Marketing Research, 21, 65–74.
Bar-Anan, Y., Liberman, N., & Trope, Y. (2006). The association between
psychological distance and construal level: Evidence from an implicit
association test. Journal of Experimental Psychology: General, 135, 609–622.
Bennett, P. D., & Harrell, G. D. (1975). The role of confidence in understanding and
predicting buyers’ attitudes and purchase intentions. Journal of Consumer
Research, 2, 110–117.
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.
Bornemann, T., & Homburg, C. (2011). Psychological distance and the dual role of
price. Journal of Consumer Research, 38, 490–504.
Brockner, J., Wiesenfeld, B. M., & Martin, C. L. (1995). Decision frame, procedural
justice, and survivors’ reactions to job layoffs. Organizational Behavior and
Human Decision Processes, 63, 59–68.
Buda, R., & Zhang, Y. (2000). Consumer product evaluation: The interactive effect of
message framing, presentation order, and source credibility. Journal of Product
and Brand Management, 9, 229–242.
Chan, E., & Mukhopadhyay, A. (2010). When choosing makes a good thing better:
Temporal variations in the valuation of hedonic consumption. Journal of
Marketing Research, 47, 497–507.
Chandran, S., & Menon, G. (2004). When a day means more than a year: Effects of
temporal framing on judgments of health risk. Journal of Consumer Research, 31,
375–389.
T.H. Freling et al. / Organizational Behavior and Human Decision Processes 124 (2014) 95–109 107
Davis, M. A., & Bobko, P. (1986). Contextual effects on escalation processes in public
sector decision making. Organizational Behavior and Human Decision Processes,
37, 121–138.
DelVecchio, D. H., Krishnan, S., & Smith, D. C. (2007). Cents or percent? The effects of
promotion framing on price expectations and choice. Journal of Marketing, 71,
158–170.
Dover, P. A., & Olson, J. C. (1977). Dynamic changes in expectancy value attitude
model as a function of multiple exposures to product information. In B. A.
Greenberg & D. N. Bellenger (Eds.), Contemporary marketing thought
(pp. 455–459). Chicago: American Marketing Association.
Dunegan, K. J. (1993). Framing, cognitive modes, and image theory: Toward an
understanding of a glass half full. Journal of Applied Psychology, 78, 491–503.
Dunegan, K. J. (1996). Fines, frames, and images: Examining formulation effects on
punishment decisions. Organizational Behavior and Human Decision Processes,
68, 58–67.
Ernst Kossek, E., & Ozeki, C. (1998). Work–family conflict, policies, and the job–life
satisfaction relationship: A review and directions for organizational behavior–
human resources research. Journal of Applied Psychology, 83, 139–149.
Eyal, T., Liberman, N., Trope, Y., & Walther, E. (2004). The pros and cons of
temporally near and distance action. Journal of Personality and Social Psychology,
86, 781–795.
Eysenck, H. J. (1978). An exercise in meta-silliness. American Psychologist, 33, 517.
Fieldler, K. (2007). Construal level theory as an integrative framework for
behavioral decision-making research and consumer psychology. Journal of
Consumer Psychology, 17, 101–106.
Fujita, K., Henderson, M. D., Eng, J., Trope, Y., & Liberman, N. (2006a). Spatial
distance and mental construal of social events. Psychological Science, 17,
278–282.
Fujita, K., Trope, Y., Liberman, N., & Levin-Sagi, M. (2006b). Construal levels and self-
control. Journal of Personality and Social Psychology, 90, 351–367.
Geyskens, I., Steenkamp, J. E. M., & Kumar, N. (1999). A meta-analysis of
satisfaction in marketing channel relationships. Journal of Marketing Research,
36, 223–238.
Glass, G. V., McGaw, B., & Smith, M. L. (1981). Meta-analysis in social research.
Beverly Hills, CA: Sage Publications.
Henderson, M. D., Fujita, K., Trope, J., & Liberman, N. (2006). The effects of spatial
distance on social judgment. Journal of Personality and Social Psychology, 91,
845–856.
Herzog, S. M., Hansen, J., & Wanke, M. (2007). Temporal distance and ease of
retrieval. Journal of Experimental Social Psychology, 43, 483–488.
Huffcutt, A. I., & Arthur, W. Jr., (1995). Development of a new outlier statistic for
meta-analytic data. Journal of Applied Psychology, 80, 327–334.
Hunter, J. E., & Schmidt, F. L. (2004). Methods of meta-analysis: Correcting error and
bias in research findings. Thousand Oaks, CA: Sage Publications.
Jain, S. P., Lindsey, C., Agrawal, N., & Maheswaran, D. (2007). For better or for worse?
Valenced comparative frames and regulatory focus. Journal of Consumer
Research, 34, 57–65.
Janiszewski, C., Noel, H., & Sawyer, A. G. (2003). A meta-analysis of the spacing
effect in verbal learning: Implications for research on advertising repetition and
consumer memory. Journal of Consumer Research, 30, 138–149.
Judge, T. A., & Piccolo, R. F. (2004). Transformational and transactional leadership: A
meta-analytic test of their relative validity. Journal of Applied Psychology, 89,
755–768.
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under
risk. Econometrica, 47, 263–291.
Kastenmüller, A., Fischer, P., Jones, E., Greitemeyer, T., Frey, D., Köppl, J., et al. (2010).
Selective exposure: The impact of framing information search instructions as
gains and losses. European Journal of Social Psychology, 40, 837–846.
Khan, U., & Dhar, R. (2010). Price-framing effects on the purchase of hedonic and
utilitarian bundles. Journal of Marketing Research, 47, 1090–1099.
Kim, M., & Hunter, J. E. (1993). Relationships among attitudes, behavioral
intentions, and behavior: A meta-analysis of past research, Part 2.
Communication Research, 20, 331–364.
Krashnikov, A., & Jayachandran, S. (2008). The relative impact of marketing,
research-and-development, and operations capabilities on firm performance.
Journal of Marketing, 72, 1–11.
Krishnamurthy, P., Carter, P., & Blair, E. (2001). Attribute framing and goal framing
effects in health decisions. Organizational Behavior and Human Decision
Processes, 85, 382–399.
Kühberger, A. (1995). The framing of decisions: A new look at old problems.
Organizational Behavior and Human Decision Processes, 62, 230–240.
Kühberger, A. (1998). The influence of framing on risky decisions: A meta-analysis.
Organizational Behavior and Human Decision Processes, 75, 23–55.
Labroo, A. A., & Patrick, V. M. (2009). Psychological distance: Why happiness helps
you see the big picture. Journal of Consumer Research, 35, 800–809.
Lee, A. Y., Keller, P. A., & Sternthal, B. (2010). Value from regulatory construal fit: The
persuasive impact of fit between consumer goals and message concreteness.
Journal of Consumer Research, 36, 735–747.
Levin, I. P., & Gaeth, G. J. (1988). Framing of attribute information before and after
consuming the product. Journal of Consumer Research, 15, 374–378.
Levin, I. P., Gaeth, G. J., Schreiber, J., & Lauriola, M. (2002). A new look at framing
effects: Distribution of effect sizes, individual differences, and independence of
types of effects. Organizational Behavior and Human Decision Processes, 88,
411–429.
Levin, I. P., Johnson, R. D., Deldin, P. J., Carstens, L. M., Cressey, L. J., & Davis, C. R.
(1986). Framing effects in decisions with completely and incompletely
described alternatives. Organizational Behavior and Human Decision Processes,
38, 48–64.
Levin, I. P., Schneider, S. L., & Gaeth, G. J. (1998). All frames are not created equal: A
typology and critical analysis of framing effects. Organizational Behavior and
Human Decision Processes, 76, 149–188.
Levin, I. P., Schnittjer, S. K., & Thee, S. L. (1988). Information framing effects in social
and personal decisions. Journal of Experimental Social Psychology, 24, 520–529.
Levin, I. P., Snyder, M. A., & Chapman, D. P. (1989). The interaction of experiential
and situational factors and gender in a simulated risky decision-making task.
The Journal of Psychology, 122, 173–181.
Liberman, N., Sagristano, M. D., & Trope, Y. (2002). The effect of temporal distance
on level of mental construal. Journal of Experimental Social Psychology, 38,
528–534.
Liberman, N., & Trope, Y. (1998). The role of feasibility and desirability
considerations in near and distant future decisions: A test of temporal
construal theory. Journal of Personality and Social Psychology, 75, 5–18.
Liberman, N., Trope, Y., & Wakslak, C. (2007). Construal level theory and consumer
behavior. Journal of Consumer Psychology, 17, 113–117.
Linville, P. W., Fischer, G. W., & Fischhoff, B. (1993). AIDS risk perceptions and
decision biases. The social psychology of HIV infection (pp. 5–38). Hillsdale, NJ:
Lawrence Erlbaum Associates.
Lipsey, M. W., & Wilson, D. B. (2001). Practical meta-analysis. Thousand Oaks, CA:
Sage Publications.
Liviatan, I., Trope, Y., & Liberman, N. (2008). Interpersonal similarity as a social
distance dimension: Implications for perception of others’ actions. Journal of
Experimental Social Psychology, 44, 1256–1269.
Lynch, J. G. Jr., (1982). On the external validity of experiments in consumer research.
Journal of Consumer Research, 9, 225–239.
MacKenzie, S. B. (1986). The role of attention in mediating the effect of advertising
on attribute importance. Journal of Consumer Research, 13, 174–195.
Magee, J. C., Milliken, F. J., & Lurie, A. R. (2010). Power differences in the construal of
a crisis: The immediate aftermath of September 11, 2001. Personality and Social
Psychology Bulletin, 36, 354–370.
Maheswaran, D., & Meyers-Levy, J. (1990). The influence of message framing and
issue involvement. Journal of Marketing Research, 27, 361–367.
Marks, L. J., & Kamins, M. A. (1988). The use of product sampling and advertising:
Effects of sequence of exposure and degree of advertising claim exaggeration on
consumers’ belief Clarity, belief confidence, and attitudes. Journal of Marketing
Research, 25, 266–281.
Marteau, T. M. (1989). Framing of information: Its influence upon decisions of
doctors and patients. British Journal of Social Psychology, 28, 89–94.
Palmatier, R. W., Dant, R. P., Grewal, D., & Evans, K. R. (2006). Factors influencing the
effectiveness of relationship marketing: A meta-analysis. Journal of Marketing,
70, 136–153.
Pany, K., & Reckers, P. M. J. (1987). With- vs. between-subjects experimental
designs: A study of demand effects. Auditing: A Journal of Practice & Theory, 7,
39–53.
Perner, J., Gschaider, A., Kühberger, A., & Schrofner, S. (1999). Predicting others
through simulation or by theory? A method to decide. Mind & Language, 14,
57–79.
Peterson, R. A. (2001). On the use of college students in social science research:
Insights from a second-order meta-analysis. Journal of Consumer Research, 28,
450–461.
Raudenbush, S. W., Becker, B. J., & Kalaian, H. (1988). Modeling multivariate effect
sizes. Psychological Bulletin, 103, 111–120.
Riley, R. D. (2009). Multivariate meta-analysis: The effect of ignoring within-study
correlation. Journal of the Royal Statistical Society, 172, 789–811.
Riley, R. D., Thompson, J. R., & Abrams, K. R. (2008). An alternative model for
bivariate random-effects meta-analysis when the within-study correlations are
unknown. Biostatistics, 9, 172–186.
Rosenberg, M. S., Adams, D. C., & Gurevitch, J. (2000). MetaWin: A statistical software
for meta-analysis. Sunderland, MA: Sinauer Associates.
Rosenthal, R. (1979). The ‘file drawer problem’ and tolerance for null results.
Psychological Bulletin, 86, 638–641.
Rosenthal, R., & Rosnow, R. L. (2008). Essentials of behavioral research: Methods and
data analysis. Boston, MA: McGraw Hill.
Sharpe, D. (1997). Of apples and oranges, file drawers, and garbage: Why validity
issues in meta-analysis will not go away. Clinical Psychology Review, 17,
881–901.
Shiv, B., Edell Britton, J. A., & Payne, J. W. (2004). Does elaboration increase or
decrease the effectiveness of negatively versus positively framed messages?
Journal of Consumer Research, 31, 199–208.
Shu, S. B., & Gneezy, A. (2010). Procrastination of enjoyable experiences. Journal of
Marketing Research, 47, 933–944.
Smith, R. E., & Swinyard, W. R. (1983). Attitude-behavior consistency: The impact of
product trial vs. advertising. Journal of Marketing Research, 20, 257–267.
Smith, P. K., & Trope, Y. (2006). You focus on the forest when you’re in charge of the
trees: Power priming and abstract information processing. Journal of Personality
and Social Psychology, 90, 578–596.
Sultan, F., Farley, J. U., & Lehmann, D. R. (1990). A meta-analysis of applications of
diffusion models. Journal of Marketing Research, 27, 70–77.
Todorov, A., Goren, A., & Trope, Y. (2007). Probability as a psychological distance:
Construal and preferences. Journal of Experimental Social Psychology, 43,
473–482.
Trope, Y., & Liberman, N. (2000). Temporal construal and time-dependent changes
in preference. Journal of Personality and Social Psychology, 79, 876–889.
108 T.H. Freling et al. / Organizational Behavior and Human Decision Processes 124 (2014) 95–109
Trope, Y., & Liberman, N. (2003). Temporal construal. Psychological Review, 110,
403–421.
Trope, Y., & Liberman, N. (2010). Construal-level theory of psychological distance.
Psychological Review, 117, 440–463.
Trope, Y., Liberman, N., & Wakslak, C. (2007). Construal levels and psychological
distance: Effects on representation, prediction, evaluation, and behavior. Journal
of Consumer Psychology, 17, 83–95.
Ulkumen, G., & Cheema, A. (2011). Framing goals to influence personal savings: The
role of specificity and construal level. Journal of Marketing Research, 48,
958–969.
Vallacher, R. R., & Wegner, D. M. (1989). Levels of personal agency: Individual
variation in action identification. Journal of Personality and Social Psychology, 57,
660–671.
Vinchur, A. J., Schippmann, S. S., Switzer, F. S., III, & Roth, P. L. (1998). A meta-
analytic review of predictors of job performance for salespeople. Journal of
Applied Psychology, 83, 586–597.
Wakslak, C. J., Nussbaum, S., Liberman, N., & Trope, Y. (2008). Representations of the
self in the near and distant future. Journal of Personality and Social Psychology,
95, 757–773.
Wakslak, C. J., Trope, Y., Liberman, N., & Alony, R. (2006). Seeing the forest when
entry is unlikely: Probability and the mental representation of events. Journal of
Experimental Psychology, 135, 641–653.
White, K., MacDonnell, R., & Dahl, D. W. (2011). It’s the mindset that matters: The
role of construal level and message framing in influencing efficacy and
conservation behaviors over the long-term. Journal of Marketing Research, 48,
472–485.
White, K., & Peloza, J. (2009). Self-benefit versus other-benefit marketing appeals:
Their effectiveness in generating charitable support. Journal of Marketing, 73,
109–124.
Zhao, M., & Xie, J. (2011). Effects of social and temporal distance on consumers’
responses to peer recommendations. Journal of Marketing Research, 48,
486–496.
T.H. Freling et al. / Organizational Behavior and Human Decision Processes 124 (2014) 95–109 109
