Content uploaded by Gilad Feldman
Author content
All content in this area was uploaded by Gilad Feldman on Sep 10, 2022
Content may be subject to copyright.
Action-Inaction Asymmetries in Emotions and Counterfactual
Thoughts: Meta-Analysis of the Action Effect
[Registered Report Stage 1]
Siu Kit Yeung
Department of Psychology, Chinese University of Hong Kong, Hong Kong SAR, China
1155062203@link.cuhk.edu.hk / yskjdmmh@gmail.com
^Gilad Feldman
Department of Psychology, University of Hong Kong, Hong Kong SAR, China
gfeldman@hku.hk / giladfel@gmail.com
^Corresponding author
Word: abstract – 189, manuscript – 10857
Action-effect: Meta-Analysis [Registered Report Stage 1] 1
Author bios:
Siu Kit Yeung is a Master of Philosophy (MPhil) graduate from the University of Hong Kong
psychology department. He is currently a PhD student with the Chinese University of Hong Kong
psychology department. His research focuses on Emotions, Judgment, Decision-Making, Mental
Health, as well as Open/Meta-Science.
Gilad Feldman is an Assistant Professor with the University of Hong Kong psychology
department. His research focuses on Social Psychology, Judgment and Decision-Making, as well
as Open/Meta-Science.
Declaration of Conflict of Interest:
Several of the studies planned to be included in this meta-analysis were conducted by the authors.
Apart from that, we do not have any other conflicts of interest to report.
Financial disclosure/funding:
The author(s) received no financial support for the research and/or authorship of this article.
Authorship declaration:
Siu Kit Yeung conducted the meta-analysis. Gilad guided Siu Kit Yeung throughout the project
and edited the outputs for submission.
Corresponding author
Gilad Feldman, Department of Psychology, University of Hong Kong, Hong Kong SAR;
gfeldman@hku.hk
Acknowledgement
We wrote this meta-analysis with reference to Meta-Analysis Registered Report Templates by
Yeung et al. (2021). We appreciate very helpful comments and suggestions from Adrien Fillon
and Alison Lam. We are also grateful for important comments and suggestions from the editor
Prof. Chris Chambers and peer reviewers Dr. Dan Quintana, Dr. Priyali Rajagopal, and Dr. Emiel
Cracco.
Action-effect: Meta-Analysis [Registered Report Stage 1] 2
Contributor Roles Taxonomy
Role
Siu Kit Yeung
Gilad Feldman
Conceptualization
V
V
Pre-testing/Simulation
V
Pre-registration
V
Data curation
V
Formal analysis
V
Investigation
V
Pre-registration peer review/verification
V
Literature search
Datafile study/effect coding
Reproducible code (e.g., RMarkdown)
V
Contacting authors
Data analysis peer review/verification
Methodology
V
Project administration
Resources
Software
V
Supervision
V
Validation
Visualization
V
Writing-original draft
V
Writing-review and editing
V
Note. Based on Allen and O’Connell (2014). Will be updated in Stage 2. We intend to add
another author for Stage 2 for datafile study/effect coding and analyses.
Action-effect: Meta-Analysis [Registered Report Stage 1] 3
Abstract
[Note: This is a Stage 1 Registered Report. We will replace the highlighted parts in the Abstract with actual results
by Stage 2.]
Action-effect refers to the phenomenon in which people experience, associate, or attribute
stronger emotions for action compared to inaction. In this registered report, we conducted a meta-
analysis of the action effect literature (k = [enter number of studies by Stage 2], N = [enter no. of
participants by Stage 2], 1982-2021). We found support/no support/mixed support for action-
effect in [positive emotions, g = X.XX, 95% CI [X.XX, X.XX]], support/no support/mixed
support for action-effect in [negative emotions, g = X.XX, 95% CI [X.XX, X.XX]], and
support/no support/mixed support for action-effect in [counterfactual thought, g = X.XX, 95% CI
[X.XX, X.XX]]. Study heterogeneity was [low / low to medium / medium / medium to high /
high], Q(XX) = XXX.XX, p = .XXX / < .001, I² = XX.XX%. [Summarize results of publication
bias tests; to be completed by Stage 2]. Action-effect was stronger [list of conditions in which the
effects were stronger, if there was/were, to be entered by Stage 2]. We pre-registered our meta-
analysis, with all search protocol, datasets, code, and supplementary made available on the OSF:
https://osf.io/acm24/.
Keywords: Action-effect; Action; Inaction; Meta-Analysis; Registered Report; Regret; Decision
Making; Emotions
Action-effect: Meta-Analysis [Registered Report Stage 1] 4
Action-Inaction Asymmetries in Emotions and Counterfactual Thoughts:
Meta-Analysis of the Action Effect
[Registered Report Stage 1]
Introduction
Action and inaction are fundamental aspects of our behavior. One of the most well-known
effects in the action-inaction literature is the action-effect, which is the phenomenon that people
imagine, associate, or experience stronger emotions for action compared to inaction. It was first
demonstrated by Kahneman and Tversky (1982) with a scenario describing two investors who
both ended up with negative outcomes following the same investment, with the main difference
between the two investors being that one investor switched to that investment from a previous
investment (action), whereas the other had considered switching the investment but had finally
decided to stick with his original choice (inaction). They found that most people perceived the
action investor as experiencing stronger regret compared to the inaction investor, concluding that
regret over negative outcomes is stronger when it involved an action decision rather than an
inaction decision.
The terms action and inaction have been used in many ways. In the context of the action-
effect, action has been used to refer to changing, switching, deviation from past behavioral norm,
or doing something, whereas inaction has been used to indicate making no changes, doing
nothing, or not doing something (Feldman et al., 2021). Action-effect (Kahneman & Tversky,
1982) has been influential in theoretical developments in the literature covering the emotion of
regret (Huang & Zeelenberg, 2012), counterfactual thinking, and norm theory (Kahneman &
Action-effect: Meta-Analysis [Registered Report Stage 1] 5
Miller, 1986). Action-effect has also served as the basis for studying other important phenomena,
for example, it has been proposed as a partial explanation for the general preference towards
omission over commission as “omission bias”, to avoid possible regret over harm induced
(Anderson, 2003; DeScioli et al., 2011). Action-inaction, associated emotions, and cognitions
have been shown to be important in understanding human judgment and decision making in
various domains, such as investment (Kahneman & Tversky, 1982), health-medicine (Brewer et
al., 2016), career or education (Gilovich & Medvec, 1994), sports (Bar-Eli et al., 2007), and
morality (Jamison et al., 2020).
Over the years, researchers studying the phenomenon have identified moderators and
boundary conditions that have been shown to weaken or even reverse the action-effect, raising
the need for a systematic review meta-analysis of the effect and the constraints over its
generalizability (e.g., Gilovich & Medvec, 1994; Zeelenberg et al., 2002). Scholars have also
suggested several ways to improve various aspects of empirical studies demonstrating the
phenomenon and pushing the literature forward. For example, the literature would greatly benefit
from better clarity regarding the definitions of the terms action and inaction and linkage with the
concepts of normality and exceptionality (Feldman & Albarracín, 2017; Feldman et al., 2020,
2021), a comparison of effects using different methodology and context (Zeelenberg et al., 2002),
measurements, and study designs (N’gbala & Branscombe, 1997; Zhang et al., 2005).
The present meta-analysis registered report aims to contribute to the literature by
conducting a systematic quantitative review of the literature on the action-effect to further our
theoretical and empirical understanding of the phenomenon with categorizations of effects in the
Action-effect: Meta-Analysis [Registered Report Stage 1] 6
literature, assessment of aggregated effects, and the identification and quantification of
moderators.
We begin with an introduction of the classic pioneering study by Kahneman and Tversky
(1982) and a brief review of the theoretical explanations for the action-effect, followed by a
review of the key findings and debates in the literature, identifying potential moderators.
Action-effect
Action-effect was first demonstrated by Kahneman and Tversky in 1982 and has been
immensely influential ever since. At the time of writing (August 2022), we identified 2626
citations of the article (according to Google Scholar), with many important follow-up theoretical
and empirical articles, with the most well-known being Kahneman and Miller’s (1986) norm
theory.
Kahneman and Tversky (1982) presented 138 participants with scenarios describing two
investors: George, who decided to switch his investment, and Paul, who decided not to change
his initial investment with both being described as having experienced the same negative
outcome. Participants were then asked who they thought would experience stronger regret. They
found that most participants perceived that action-George would experience stronger regret than
inaction-Paul. Many follow-up studies have successfully replicated Kahneman and Tversky’s
(1982) findings (e.g. Connolly et al., 1997; Fillon et al., 2022; Gleicher et al., 1990, Landman,
1987; Sepehrinia et al., 2022). For example, one of the very few action-effect studies measuring
both pleasant and unpleasant emotions by Landman (1987) found support for action-effect for
both elation and regret across different scenarios.
Action-effect: Meta-Analysis [Registered Report Stage 1] 7
In this meta-analysis, we focused on action-inaction asymmetries regarding emotions and
counterfactual thoughts. This is different from other related action-inaction judgments, decision
making, and social psychology effects, such as studies of moral judgments or decisions (for a
meta-analysis on omission bias, see Yeung et al., 2021), status quo bias, and default effect
(Feldman et al., 2020).
What are action and inaction?
Action effect has laid the foundation for discussing action and inaction, yet the terms
action and inaction were never clearly defined until recently (Feldman et al., 2021). When
referring to “action” or describing an “action”, some (e.g., Kahneman & Tversky, 1982) may
refer to change or deviation from past behavioral norms, yet this does not always have to be the
case. Sometimes action (e.g., Gilovich & Medvec, 1994) is used to refer to “doing something”,
for example, going to college, whereas inaction refers to “not doing something” or “doing
nothing”, for example, not going to college. “Action” may also mean physically active or making
a proactive choice, whereas “inaction” may mean physically inactive, making a passive choice
(e.g., avoidance), or procrastination (Feldman et al., 2021). However, most studies on the action-
effect seemed to either use the “change vs no change” meaning or the “doing something vs not
doing something” meaning. Apart from the recent work by Sepehrinia et al. (2022) which
manipulated these two types of action-inaction operationalizations in a single study, we are
unaware of any study comparing action-effects with two operationalizations within a study (but
see Yeung & Feldman, 2022 for related discussions regarding discrepancies of findings between
studies). The unclear meanings have been identified as a major gap in the literature leading to
confounded usage by Feldman et al. (2021), and so in our meta-analysis, we aim to try and
carefully disentangle the different types and the possible similarities and/or differences in effects.
Action-effect: Meta-Analysis [Registered Report Stage 1] 8
Norm Theory and Normality
Action-effect is related to the concept of normality and can be explained by norm theory
(Kahneman & Miller, 1986). Normality refers to the extent to which behaviors, circumstances, or
outcomes are perceived to be normal. Normality may depend on cognitive availability of events
or stimuli, and it is likely related to counterfactual thought – retrievability of related stimuli or
events and imagining alternative realities (Feldman et al., 2020). The term exceptionality is often
used to refer to the deviation from a reference point, such as past behavioral norms (Feldman et
al., 2020; Fillon et al., 2020). Regret appears to be stronger when the behavior before the
outcome is more exceptional, and the argument was that is it easier to imagine normal
alternatives (routine behavior) that may have resulted in a more favorable outcome (Feldman,
2020). Regret and upwards counterfactuals are positively correlated, though the causal direction
remains unclear (Coricelli & Rustichini, 2010; Feldman, 2020; Fillon et al., 2020). Norm theory
suggested that in action-effect situations, inaction seems more normal than action, and action is
associated with more counterfactual thoughts. It therefore seems easier to imagine abstaining
from action following the established norm than to imagine action that deviates from that norm.
Thus, people imagine, attribute, or experience stronger regret following action compared to
following inaction. For example, in the context of the Kahneman and Tversky (1982) investor
scenario, action means switching or deviating from a previous investment decision (past
behavioral norm as the reference point), the reference point. In the context of risky financial
investments that may result in negative outcomes, actions to make an investment or to change
investments may be perceived as less normal, or as a deviation from one’s past behavioral norm.
That being said, we note that “action” does not necessarily mean a change or a deviation from
Action-effect: Meta-Analysis [Registered Report Stage 1] 9
one’s past behavioral norm, especially in studies where the experimenters asked participants to
think of “something they did” (e.g., Study 1 and Study 5 of Gilovich & Medvec, 1994). The
norm-theory account may not be applicable to those situations.
As mentioned above, there is likely a positive correlation between regret and
counterfactual thoughts. It had been assumed that higher counterfactual mutability of actions
compared to inactions accounts for the action-effect (Kahneman & Miller, 1986), but very few
studies measured both counterfactual thoughts and regret simultaneously (Fillon et al., 2020;
N’gbala & Branscombe, 1997). In follow-up studies testing the norm theory account of action-
effect, N’gbala and Branscombe (1997) and Davis et al. (1995) failed to find support for the
proposed mediating role of counterfactual thought. Davis et al. (1995) compared counterfactual
mutability of action vs inaction for spouses and parents who experienced accidental traumatic
events and found that people mutated inactions more than actions. N’gbala and Branscombe
(1997) argued that “There are situations where actions and inactions can be perceived as equally
mutable: that is when both can be construed as instrumental or causal for achieving the desired
outcome.” (p. 328). That said, it is premature to conclude that there are no differences in
counterfactual thought between action and inaction based on one or two articles, and so a meta-
analysis on the action-inaction asymmetries of counterfactual thoughts could help assess this
systematically. We also note that there has been evidence, with larger sample sizes and across
several studies, suggesting action is associated with more “what if” thoughts compared to
inaction (Byrne & McEleney, 2000). The discrepancies in findings between studies may be due
to differences in measures, scenarios, or study design. One possible explanation suggested by
Byrne and McEleney (2000) is that there may be a stronger effect with within-subject design.
Action-effect: Meta-Analysis [Registered Report Stage 1] 10
Decision Justification Theory
Another key theory in the domain of action-inaction is the Decision Justification Theory
(DJT; Connolly & Zeelenberg, 2002), which focuses on the justifiability of decisions resulting in
regret. Linking the two theories, exceptional decisions may be more likely to be perceived as less
justified than normal decisions, thereby resulting in stronger regret when things go wrong (Reb &
Connolly, 2010). This is because it is easier to justify routine or normal decisions to oneself and
to others compared to exceptional or abnormal decisions (Reb & Connolly, 2010). In the context
of the two-investor scenario, action (deviating from past personal behavioral norms) may be
perceived as more exceptional and less justified than inaction (sticking with the same option).
However, under some situations and some social environments, action (in the form of change)
may be perceived as more normal and more justified than inaction (in the form of no change),
notably but not limited to situations with negative prior outcomes and problems or certain
situational expectations (Bar-Eli et al., 2007; Connolly & Zeelenberg, 2002; Feldman, 2020;
Feldman & Albarracín, 2017; Olsen, 2017; Zeelenberg et al., 2002). Action norms in some
situations may partly explain findings of inaction-effect, of stronger regret for inaction than for
action (e.g., Feldman & Albarracín, 2017; Zeelenberg et al., 2002). Given prior negative
outcomes or situational expectations for taking action, it would be more justifiable to act in order
to tackle the problem, thereby leading to stronger regret if one failed to act.
Additional theoretical accounts
There are other theoretical explanations of the action-effect. For example, action is
associated with stronger responsibility than inaction and is perceived as more agentic (Connolly
et al., 1997; Yeung et al., 2021). Action would be associated with stronger regret than inaction,
Action-effect: Meta-Analysis [Registered Report Stage 1] 11
given that someone would perceive or experience a stronger sense of responsibility for the
negative outcome resulting from an action (Connolly et al., 1997). Another explanation is that
action is generally perceived as more effortful, and the actor is perceived to experience stronger
regret over wasted effort (N’gbala & Branscombe, 1997). N’gbala and Branscombe (1997) found
that the actor is perceived as less wise as the actor wastes too much effort and time. It seems
plausible that wisdom is likely positively associated with justifiability (Garrett, 1996), but we are
not aware of studies testing both in action-effect studies. We note these accounts, but these
explanations are outside the scope of this meta-analysis.
We summarized the theoretical accounts of action-effect in Table 1. These theoretical
explanations remain under debate and are facing multiple challenges. For example, some
elements of norm theory, such as counterfactual mutation (Kahneman & Miller, 1986), have
received no support from some studies (Davis et al., 1995; N’gbala & Branscombe, 1997). In
these theoretical accounts, constructs such as action, inaction, normality, exceptionality, and
justifiability were not clearly defined and explained, and the expected relationships between these
constructs given theory were not clearly elaborated (Feldman, 2020; Feldman & Albarracín,
2017; Feldman et al., 2021). Theories can improve in outlining testable hypotheses that would
favor one account over the other. However, these theories are not necessarily contradictory and
may jointly explain the phenomenon. An action may be regretted more because of several
different reasons - it is perceived as abnormal, it is perceived as more blameworthy or
responsible, it is perceived as wasted effort, or it is perceived as unjustified. It is possible all these
factors contribute to action-effect in some situations, but one of these accounts may be more
relevant in some situations. That said, very limited studies have compared these different
contributing factors of action-effect and there is much need for more research on the intersection
Action-effect: Meta-Analysis [Registered Report Stage 1] 12
of the theoretical paradigms. Nonetheless, even though still lacking in sound theoretical and
construct specificity, action-effect as a phenomenon seems to have empirically strong
foundations.
Table 1
Theoretical explanations of the action-effect
Norm Theory
Decision Justification
Theory
Action-
responsibility link
Action-effort link
Focus and
assumptions
Focuses on normality and
exceptionality of action
and inaction. Exceptional
behavior is more strongly
regretted than routine
behavior. Assumes
inaction to be more normal
than action.
Focuses on
justifiability. Less
justified decisions are
regretted more than
more justified
decisions.
Action is attributed
more blame than
inaction. Stronger
regret due to
stronger assumed
responsibility.
Action is perceived as
more effortful, and as
wasted effort if the
outcome is negative.
Implications
regarding
action-effect
When inaction is more
normal (e.g., past behavior,
social norms, etc.),
inaction may be regretted
more than inaction.
When inaction is more
justified (e.g., prior
positive outcomes,
situational expectations
for inaction), inaction
may be regretted more
than inaction.
Action is regretted
more than inaction
Action is regretted
more than inaction in
a within-subject
design.
Evidence and
empirical
challenges
1) Assumptions regarding
whether inactions or
actions are more normal
are typically untested
2) N’gbala and
Branscombe (1997) failed
to find support for stronger
counterfactual mutation of
action
3) Byrne and McEleney
(2000) found support for
stronger “if only” thoughts
for action compared to
inaction.
1) Assumptions
regarding whether
inactions or actions are
more justifiable are
typically untested.
2) Unclear link
between normality and
justifiability.
1) Supported by a
recent meta-analysis
(Yeung et al.,
2021).
2) Boundary
condition: under
certain social roles
in which the
decision-maker is
responsible for the
target, such effect is
minimized (Haidt &
Baron, 1996).
1) No direct evidence
regarding effort.
Action is perceived to
be a waste of energy
and time (N’gbala &
Branscombe, 1997)
2) Action-effect
seems stronger in
within-subject
designs
(N’gbala &
Branscombe, 1997;
Zhang et al., 2005).
Action-effect: Meta-Analysis [Registered Report Stage 1] 13
Boundary conditions of Action Effect
Sometimes action-effect can be weakened or even reversed into inaction-effect. Identified
moderators in the literature so far include study design (N’gbala & Branscombe, 1997; Zhang et
al., 2005), prior outcomes (Zeelenberg et al., 2002), temporal distance (Feldman et al., 1999;
Gilovich & Medvec, 1994), generality (Davison & Feeney, 2008), and study realism (Gilovich &
Medvec, 1994). For example, Gilovich and Medvec (1994) conducted five studies on the
temporal pattern of the action effect using different methods including telephone surveys, two-
scenario experiments, and questionnaires. They found that participants tended to both evaluate
others’ regrets and report their own regrets as being stronger for inaction compared to action, the
more distant in the past the experience was. Using the Kahneman and Tversky (1982) investor
vignette, they showed a reversal of the action-effect by manipulating temporal distance.
Recently, Yeung and Feldman (2022) conducted replications of Gilovich and Medvec
(1994) and found support for the temporal reversal of action-effect in the long run with scenario
experiments, yet failed to find support for an action-effect or an inaction-effect in recalled short-
term and long-term real-life experiences. They summarized it was unclear why they observed
discrepancies between these studies since it could be more than just the contrast between
scenarios and recalls. Another possibility raised was that these may be due to the different
meanings of action and inaction used in these studies. Namely, in the vignette replications action
meant change and inaction meant no change, whereas in the recall replications action meant
doing something and inaction meant not doing something. Other differences between the studies
were about actual first-person experiences versus evaluations of third-person emotions (Yeung &
Feldman, 2022), and methodological/scenario constraints on generality (Byrne & McEleney,
Action-effect: Meta-Analysis [Registered Report Stage 1] 14
2000; Simons et al., 2017). Regardless, these findings suggest that action-inaction-associated
regret is a much more complex phenomenon than initially assumed.
Zeelenberg et al. (2002) showed a reversal of action-effect using prior outcomes. They
showed that with prior positive and neutral outcomes, the action-effect findings from Kahneman
and Tversky (1982) held. However, they showed an inaction effect when prior outcomes were
negative, suggesting that these shift expectations for agents to act attempting to resolve the
problem. A similar demonstration by Feldman and Albarracín (2017) showed similar findings on
action-inaction social norms as moderators of action-effect. Past behavior, expectations, and
social norms serve as reference points that people use to assess what is considered normal and
justifiable (Feldman, 2020). A meta-analysis allows for systematic quantification of the effect,
investigating possible moderators.
Aim and scope
In this meta-analysis, we investigated action-inaction asymmetries of emotions and
counterfactual thoughts. We included counterfactual thoughts as those have been shown as being
associated with emotions. We focused on these dependent variables as these were the first initial
demonstrations of action-inaction asymmetries, and based on our experience in conducting
replications in this domain these dependent variables seem to be the most commonly studied in
the action-inaction domain.
The current project will not cover action-inaction asymmetries regarding moral judgments
and decisions. Some of those asymmetries have already been investigated in other meta-analyses,
such as in the meta-analysis on omission bias by Yeung et al. (2021). Omission bias and action-
Action-effect: Meta-Analysis [Registered Report Stage 1] 15
effect are related but distinct, with different dependent variables, moderators of interest, and
theoretical paradigms (see Feldman et al., 2020; Yeung et al., 2021 for more details). Norm
theory is a framework that has been used to try and align omission bias with action-effect
(Feldman et al., 2020) yet this remains to be studied further, and there are other theoretical
explanations for both effects (decision justification theory, action-effort link, action-causality-
link, etc.) that seem more suitable for explaining one effect yet not the other.
We aimed to address the following research questions: 1) What is the overall empirical
support for the action effect? 2) What factors moderate the action-effect? 3) Under what
conditions are action-effect weakened or reversed into inaction-effect?
Action-effect: Main effect
Our first aim was to examine the overall effect of action-effect. We expected the evidence
to be in support of action-effect with a positive effect meaningfully different from null (null not
included in confidence intervals) (H1overall). Specifically (see Table 2 for term elaboration):
H1a: Action is associated with stronger negative emotion than inaction, given
negative outcome.
H1b: Action is associated with stronger positive emotion than inaction, given
positive outcome.
H1c: Action is associated with more counterfactual thought (number of
counterfactuals) than inaction.
A recent study by Fillon et al. (2022), one of the very few studies that investigated both
regret and joy, found stronger effects for regret and weaker effects for joy, possibly because of
Action-effect: Meta-Analysis [Registered Report Stage 1] 16
negativity bias, meaning that bad seem to be perceived to be more agentic (Feldman et al., 2016)
and has stronger emotional impact than good (Baumeister et al., 2001; Kahneman & Tversky,
1979). Their findings were different than the more classic Landman (1987) and Sepehrinia et al.
(2022) who failed to find support for differences between joy and regret. Given such mixed
findings, we set this direction as exploratory. The vast majority of the studies we know from the
literature only investigated action-effect for negative emotions, with very few studies measuring
positive emotions (e.g., Fillon et al., 2022; Landman, 1987). If we find 5 studies for action-
inaction asymmetries in positive emotions, then we will conduct exploratory moderator analyses
comparing positive emotions studies and negative emotions studies.
Exploratory hypotheses (if 5 studies or more on positive emotions): Action-effect
is stronger/weaker for negative emotions compared to positive emotions.
Another exploratory direction is regarding possible differences between upward
counterfactuals and downward counterfactuals. To the best of our knowledge, the vast majority of
studies measured upward counterfactuals, but there might be studies on downward
counterfactuals.
Action-effect: Meta-Analysis [Registered Report Stage 1] 17
Therefore, if we find at least 5 studies on action-inaction asymmetries on downward
counterfactual thoughts (and at least 5 studies for upward counterfactual thoughts), then we
would conduct exploratory analyses for moderation of upward counterfactuals versus downward
counterfactuals.
Exploratory hypotheses (if 5 studies or more on downwards counterfactuals):
Action-effect is stronger/weaker for upwards counterfactuals compared to
downwards counterfactuals
Table 2
Clarifications regarding terms used in the context of the action-effect
Terms
Description with examples
Associated with
1) Type 1: Perception (e.g., Kahneman & Tversky, 1982) in which more
participants perceived that the investor (a third person) who switched
would experience stronger regret.
2) Type 2: Actual experience (e.g., Gilovich & Medvec, 1994 Study 1), in
which more participants experienced stronger regret for inaction than
action when retrospectively thinking about their lifetime regrets.
3) “Action is associated with stronger emotions than inaction” means that
most people perceive and/or experience stronger emotions acting
compared to not acting.
Counterfactual thoughts
Thoughts that are “counter to the factual events”, or possible alternatives of
what if the decision-maker chooses another option.
For example, in N’gbala and Branscombe (1997), the experimenter asked
participants to imagine and describe their “what if” thoughts. After that,
researchers coded the responses as mutations to action or mutations to inaction.
Positive outcome
Negative outcome
The consequence for the decision-maker is desirable. For example, receiving a
better grade in Landman (1987).
The consequence for the decision-maker is undesirable. For example, losing
money in Kahneman and Tversky (1982)
Positive emotion
Pleasant emotions including but not limited to satisfaction, happiness, joy,
elation, feeling good, feeling better, relief. For example, Landman (1987) asked
participants to compare who (the actor vs the non-actor) would feel better.
Negative emotion
Unpleasant emotions including but not limited to guilt, regret, disappointment,
sadness, feeling bad. For example, Kahneman and Tversky (1982) asked
participants to compare regret for the actor and the non-actor.
Action-effect: Meta-Analysis [Registered Report Stage 1] 18
Confirmatory Moderators
Temporal Distance
Does temporal distance moderate the action-effect? Gilovich and Medvec (1994)
replicated action-effect in the short-term yet found support for temporal distance as a moderator.
Weaker or reversed action-effect in the long-term may be due to cognitive accessibility
(Rajagopal et al., 2006), or perhaps because the consequences of actions occur more quickly and
tend to be more limited, but the consequences of inactions may occur later (Gilovich & Medvec,
1995). Another possible explanation has to do with the type of affect elicited by temporal
distance (Gilovich et al., 1998).
Evidence regarding this moderator is mixed. Bonnefon and Zhang (2008), Byrne and
McEleney (2000), Feldman et al. (1999), and Towers et al. (2016) failed to find support for the
temporal pattern of action-inaction regret. Recently, Yeung and Feldman (2022) replicated
Gilovich and Medvec (1994) and found support for action-effect in the short-term and inaction-
effect in the long-term with college decision scenario experiments yet failed to find support for
the temporal effect with real-life experience studies. Null findings do not necessarily mean that
the moderation effect of temporal distance does not exist, as these may be caused by various
methodological factors (Feldman et al., 1999), yet it raises the need to systematically examine
this evidence. To the best of our knowledge, the vast majority of studies for temporal pattern of
action-effect measured negative emotions but not positive emotions, and we therefore do not
have any a-priori expectation regarding such temporal pattern for positive emotions. We use the
Action-effect: Meta-Analysis [Registered Report Stage 1] 19
number of years1 as the unit to code the temporal distance. We also conduct such analyses
categorically (recent or current events – less than a year, more distant events – a year ago or
more, lifetime events). Therefore, we hypothesize the following:
H2: Temporal distance moderates the relationship between action and inaction on
emotions. The more distant2 the behavior is in the past, the weaker the action-
effect. Action-effect is stronger for recent or current events, compared to more
distant (1 year ago or more) or major lifetime events.
Study Design: Between versus within versus comparison
Researchers initially studied action-effect using within-subject or one-sample comparison
designs (N’gbala & Branscombe, 1997; Zhang et al., 2005), yet there is evidence that action-
effect is weaker using between-subject designs (N’gbala & Branscombe, 1997).
Comparison studies (where the dependent variable is a choice between action or inaction)
and within-subject studies (where the dependent variables are two separate ratings of action and
inaction) allow for an evaluation of action against inaction and therefore may provide better
control for contextual factors and confounding variables. Yet, some have argued that effects in
within and comparison designs may be the result of stronger demand effects. The argument is
that participants may be able to guess the hypothesis of the study and answer in a way they think
1 For short-term regrets that are less than a year ago, we estimated based on provided information. For example, for
Gilovich and Medvec (1994) Study 5 past-week regret of participants, the estimated no. of “year” is 0.0096 (3.5 days
/ 365 days, as it is possible for the decision to be 7 days ago or less than a day ago, we picked the middle number).
2 We note that for lifetime regrets, we coded the studies with reference to the age of participants, or estimated age
based on relevant information (e.g., college students are around 20 years old). In any case, we justified our
estimation in the coding sheet.
Action-effect: Meta-Analysis [Registered Report Stage 1] 20
would best fit with the experimenter’s expectations (Charness et al., 2012; Kahneman & Tversky,
1982b), though recent studies found little support for that argument (Lambdin & Shaffer, 2009;
Mummolo & Peterson, 2019).
Another argument in this debate is that between subject designs are more realistic. In real-
life, people more often face situations in which they are only presented with either action or
inaction resulting in a negative outcome, rather than both together. However, between-subject
designs often lack a meaningful reference point to be used for comparison (Zhang et al., 2005).
This may result in arbitrary choices of some other contextual cues to compare against.
We expected to find support for an action-effect, regardless of design, yet weaker effects
in between designs compared to comparison studies and within-subject studies. We also expected
a stronger effect using comparison designs compared to within-subject designs and between-
subject designs. Fillon et al. (2020) argued that comparison designs create an artificial contrast
leading to an overestimated effect size for exceptionality bias, a related yet distinct judgment and
decision-making bias. They found effect sizes in comparison studies were over three times larger
than those in experimental studies.
Therefore:
H3a: H1 will be supported across between, within, and one-sample comparison
designs (confidence intervals do not overlap with the null).
H3b: Action-effect is stronger using within-subject designs compared to between-
subject designs.
H3c: Action-effect is stronger using comparison designs compared to within-
subject designs.
Action-effect: Meta-Analysis [Registered Report Stage 1] 21
[H3b+H3c: Action-effect is stronger using comparison designs compared to
between-subject designs.]
Normality: Prior outcomes and social norms
Zeelenberg et al. (2002) found that given negative prior outcomes, participants associated
more regret with inaction compared to action, possibly due to an expectation or formed norm for
an action to change so that future losses would be avoided (Feldman & Albarracín, 2017). This is
in line with decision justification theory (Connolly & Zeelenberg, 2002), arguing that decisions
(action given negative prior outcomes) that are better socially justified are regretted less. Most
studies demonstrating the action-effect (e.g., Kahneman & Tversky, 1982) adopted isolated
decision scenarios, in which there was no background information regarding the outcomes of
prior decisions.
We are interested in investigating if there would be a substantial difference in effect size
and a possible reversal of the sign with prior negative outcomes compared to having no prior
outcomes, neutral prior outcomes, or positive prior outcomes.
H4a: H1 will be supported given no prior outcomes, or with neutral or positive
prior outcomes.
H4b: Action-effect is moderated given negative prior outcomes (weaker or
reversed into an inaction effect).
Based on norm theory (Kahneman & Tversky, 1982), people experience stronger regret
for abnormal decisions and weaker regret for normal decisions. Feldman and Albarracín (2017)
and Feldman (2020) proposed that there are three types of normality: past behavioral norm, social
Action-effect: Meta-Analysis [Registered Report Stage 1] 22
norm, and situational/role norm, and that given no information about normality and under
inaction norm, action-effect is more likely to occur. In contrast, under action norm, action-effect
may be weakened or even reversed into inaction-effect. Therefore:
H4c: Action-effect is stronger under inaction norms compared to action norms.
H4d: Action-effect is stronger given no norm information compared to action
norms.
H4e: Action-effect is stronger under inaction past behavior compared to action
past behavior.
H4f: Action-effect is stronger given no mention of past behavior compared to
action past behavior.
Specific vs General
Gilovich and Medvec (1995) proposed that action regrets are more likely to be based on
decisions at a specific time, and inaction regrets are more likely to be about accumulated or
generalized events. To test this idea, Davison and Feeney (2008) examined regret from the
perspective of autobiographical memory, and found that people experienced more inaction regret
for general events, with elderly participants and participants at 40s, but failed to find support for a
meaningful difference between general memories and specific memories for action-regret. Their
findings were consistent with studies that demonstrated inaction effects for general life events
(Gilovich & Medvec, 1994; Yeung & Feldman, 2022). However, there appear to be very limited
follow-up studies or replications with other samples. Therefore:
H5: Action-effect is stronger for specific events compared to general events.
Action-effect: Meta-Analysis [Registered Report Stage 1] 23
Hypothetical vs real-life experience
Some studies presented participants with hypothetical vignette scenarios (e.g., Kahneman
& Tversky, 1982; Landman, 1987) whereas other studies (e.g., Feldman et al., 1999; Gilovich &
Medvec, 1994) tested autobiographical real-life experiences of action-inaction emotions. Real-
life events are different from well-controlled hypothetical scenarios with the same outcome, as
negative outcomes from action and inaction may not be equal in strength, resulting in systematic
variations between action and inaction experiences (Feldman et al., 1999). There appear to be
more successful demonstrations of action-effect with hypothetical scenario studies (Gleicher et
al., 1990; Kahneman & Tversky, 1982; Landman, 1987), but more mixed findings, null findings,
or even reversed findings with real-life experience studies (Feldman et al., 1999; Gilovich &
Medvec, 1994; Yeung & Feldman, 2022). The mechanism is unclear and may be because
inaction regrets in real-life (e.g., not going to college) tend to be longer-lasting and result in more
consequential outcomes (Gilovich & Medvec, 1995). The differences between hypothetical
scenarios and real-life events may be confounded by the fact that most hypothetical scenarios
seem to focus on short-term scenarios, whereas most real-life events studies seem to focus on
lifetime or long-term experiences. Temporal distance may be a potential moderator. Another
explanation proposed by Feldman et al. (1999) is that “if it is easier to anticipate the potential for
harm due to action than due to inaction, people might be more
conservative in their choice of which actions to take. Consequently, greater harm and more
intense regrets would result from failures to act” (p. 254). Therefore:
H6: Action-effect is stronger for studies using hypothetical scenarios than for real-
life events.
Action-effect: Meta-Analysis [Registered Report Stage 1] 24
Exploratory Directions
Target: Self versus others
Some studies asked participants to imagine, attribute and/or compare the level of emotion
of others (e.g., Kahneman & Tversky, 1982 Action George vs Inaction Paul), whereas some
studies asked participants to rate or compare regret for action vs inaction for self (e.g., Gilovich
& Medvec, 1994). Self-other distinction is often a key moderator in judgment and decision-
making research. A recent meta-analysis on omission bias (Yeung et al., 2021) investigated self-
other differences and found mixed results with different models yet seemed to generally suggest
omission bias is stronger if the target is self, compared to when the target is others. Based on the
findings from the omission/action bias literature (Zikmund-Fisher et al., 2006), it seems plausible
to expect a stronger action-effect for self compared to for others.
Meanings of action and inaction
In the classic Kahneman and Tversky (1982) demonstration of action-effect, “action”
meant changing from a prior decision (deviating from past behavioral norm) whereas “inaction”
meant persisting with a prior decision. However, over the years there have been many other uses
of the terms action and inaction. See our discussion in the introduction. We planned to compare
action-effect given meanings of “change vs no change” (example: Kahneman & Tversky, 1982,
which seem to be more related to norm theory and the concept of normality, see Yeung &
Feldman, 2022), “doing something vs not doing something/failure to do something” (example:
Gilovich & Medvec, 1994 Study 5), and other meanings (depending on what meanings we find in
the literature). If there are too few studies (under five) using any of the meanings, we would not
analyze those meanings. It seems plausible that action-effect would be stronger if action involves
Action-effect: Meta-Analysis [Registered Report Stage 1] 25
changing from one investment target/school/company to another one, compared to when action
does not involve switching (Sepehrinia et al., 2022). However, since we are only aware of one
study (Sepehrinia et al., 2022) directly comparing these two operationalizations and they found
support for differences for two domains but failed to find evidence for two other domains, we do
not have specific hypotheses and aim to examine these as an exploratory moderator of action-
effect.
Methods
[Note: Written in past tense to demonstrate methods section after completion, but has yet to be conducted. Will be filled and
updated after pre-registration and data collection.]
Open Science Disclosures
We shared all procedures, materials, datasets, and code on the Open Science Framework:
https://osf.io/acm24/.
We conducted an initial unstructured pre-search to construct, test, and refine our search
syntax. We then posted notices on listservs asking authors to alert us of possible related articles
and unpublished studies: Society for Judgment and Decision Making (SJDM), European
Association for Decision Making (EADM), European Association of Social Psychology (EASP),
Society for Personality and Social Psychology (SPSP), and Social Psychology Network (SPN) on
[Dates to be inserted in Stage 2]. [Note: The list will be updated in Stage 2 if changed]
We provided a notice template in the Template for Contacting Authors for Published and
Unpublished Data on Listservs section of the Supplementary. Systematic data collection has not
been conducted for this project. There are no other unreported pre-registrations for this meta-
analysis project. See Open Science Disclosures in Supplementary for details.
Action-effect: Meta-Analysis [Registered Report Stage 1] 26
Literature search
To find articles relevant to our topic, we used Google Scholar (Walters, 2007),
PsychINFO, Scopus, ProQuest Dissertations and Theses Global, Web of Science (Moreau &
Gamble, 2020), and Opengrey.eu. We used Boolean Logic operators such as “OR”, and “AND”
in the search pattern to connect action-inaction/commission-omission and positive and negative
emotions. Table 3 lists all search patterns. We validated and pre-tested the search pattern with 10
randomly selected articles out of all articles related to action-effect mentioned in this manuscript.
At Stage 2, we reran searches at least two times to ensure w