Psychology, Public Policy, and Law
© 2019 American Psychological Association
ISSN: 1076-8971
2020, Vol. 26, No. 1, 88 –104
http://dx.doi.org/10.1037/law0000215
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Conflict of Interest Disclosure With High-Quality Advice: The Disclosure
Penalty and the Altruistic Signal
Sunita Sah
Daniel Feiler
Cornell University
Dartmouth College
Advisors often have conflicts of interest: a potential clash between professional responsibilities and
self-interests. Disclosure—informing advisees of the conflict—is a common policy response to manage
such conflicts. However, extant research on disclosure has often confounded disclosure with poor-quality
advice. In this article, we explore whether laws requiring conflict of interest disclosure damage the
advisor–advisee relationship more than is intended. Across 6 experiments (N ⫽ 1,766), we examine
situations in which advisors give high-quality advice but still must disclose a conflict of interest. As
predicted, such disclosures yield negative attributions regarding the advisor’s character, even when
advice is of high quality (and advisees have full information to judge advice quality), and even when
the advisor’s professional responsibility and self-interest are aligned, or the advice runs counter to the
advisor’s self-interest. This disclosure penalty decreases trust in honest advisors but can be mitigated if
the advisees are explicitly told that their advisor’s conflict of interest arose from external factors beyond
the advisor’s control. When advisors’ recommendations run counter to their self-interests, conflict of
interest disclosure creates an independent competing effect—the altruistic signal—which increases trust.
The net effect on trust depends on which effect—the disclosure penalty or altruistic signal—is stronger.
We discuss the implications of these findings for law and policy.
Keywords: conflicts of interest, disclosure, trust, advice, judgment and decision-making
Supplemental materials: http://dx.doi.org/10.1037/law0000215.supp
questions and examine whether laws requiring such disclosures
have unintended consequences.
Conflicts of interest (COIs) present a potential clash between an
advisor’s professional responsibility and a personal self-interest
(Boatright, 2000; Carson, 1994; Sah, 2019). By professional responsibility, we mean the responsibilities of advisors to place their
advisees’ needs above their own. By personal self-interest, we
mean advice that gives extra benefits to the advisor, usually (but
not necessarily) extra financial compensation. Among the many
policy options for managing COIs, disclosure (informing advisees
of the COI) is the most commonly implemented solution (Sah,
2017). Disclosure is attractive because it preserves advisee autonomy (Johns, 2007) and, theoretically, enables advisees to hold a
critical lens to the advice they receive such that they can assess the
risk and extent of COI influence on advice-quality (Thompson,
1993). Recipients say they want the disclosures (Grady, Horstmann, Sussman, & Hull, 2006) and policymakers continue to
institute new and broader disclosure mandates that can be delivered in many ways, for example, directly from the advisor, a third
party, or a public website. (Rosenthal & Mello, 2013). While we
do not contest that there are significant merits to COI disclosure,
we examine one important blind spot that comes to light when
taking a behavioral perspective: the effect on the advisor–advisee
relationship when advice is of high-quality.
There has been recent progress toward a better understanding of
the psychological effects of COI disclosure on both the advisor’s
willingness to give biased advice (Sah, 2019) and on whether
Consider two cases in which an honest, and unbiased, financial
advisor is required to disclose that she receives a referral bonus if
her client invests in a particular mutual fund. In one case, the fund
the advisor receives a bonus for, and recommends, is objectively
superior to other funds and best for her client. In another case, the
fund that the advisor receives a bonus for is inferior, but the
advisor still recommends the other (superior) fund, sacrificing her
self-interest. In both cases, the advisor’s recommendation is of
high quality. How does disclosure—informing the client of the
advisor’s referral bonus—affect the client’s trust in the advisor?
Does awareness about the advisor’s conflict of interest create
problems for the advisor-client relationship even if advice quality
is high and the bonus is sacrificed? In this article, we address these
This article was published Online First November 21, 2019.
X Sunita Sah, Johnson Graduate School of Management, Cornell SC
Johnson College of Business, Cornell University; Daniel Feiler, Tuck
School of Business, Dartmouth College.
This article has benefited from discussions with seminar participants at the
Tuck School of Business, Dartmouth College and Yale School of Management, as well as conference participants at the 77th Annual Meeting of the
Academy of Management. Data, preregistrations and stimuli material are
available on the Open Science Framework: https://osf.io/uj8yp/.
Correspondence concerning this article should be addressed to Sunita
Sah, Johnson Graduate School of Management, Cornell SC Johnson College of Business, Cornell University, 326 Sage Hall, Ithaca, NY 14853.
E-mail: sunita.sah@cornell.edu
88
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
DISCLOSURE PENALTY
advisees’ choices are influenced by the COI disclosure (Licurse,
Barber, Joffe, & Gross, 2010; Sah, Loewenstein, & Cain, 2013,
2019; Sah, Malaviya, & Thompson, 2018). Some of this research
has revealed unintended consequences of COI disclosure, such as
advisors giving more biased advice (Cain, Loewenstein, & Moore,
2005, although also see Church & Kuang, 2009; Koch & Schmidt,
2010; and Sah, 2019, for instances of COI disclosure improving
advice quality), and advisees feeling pressured to comply with
advice that they do not trust (Sah et al., 2013, 2019). However, less
attention has been given to the downstream effects of such disclosures on the advisor–advisee relationship—mainly, whether advisees wish to maintain the relationship with their advisor or switch
advisors after a COI disclosure. And importantly, much of the
prior research has conflated the presence of a COI with the
provision, or assumption, of biased, poor-quality advice. For example, advisees receiving inferior lottery recommendations (Sah et
al., 2013) or bloggers giving highly positive reviews in sponsored
blog posts (Sah et al., 2018). Therefore, the question remains: How
do advisees respond to disclosure when the advice is uncompromised by the presence of a COI? Indeed, many practitioners argue
that a more typical scenario is the presence of a COI with the
provision of high-quality, honest advice.1
In this article, we address the important question of whether
people punish advisors for disclosing potential COIs, even when
the advice they receive is actually in their best interest. Across six
experiments (one preregistered), we examine the inferences advisees make about their advisors’ character. We focus on advisees’
desire to keep or change advisors after they receive high-quality
advice accompanied with, or without, a disclosure of a COI.
We document two effects. The first is the disclosure penalty,
which leads to increased desire to change advisors with COI
disclosure, even when advice is of demonstrably high quality (and
advisees have full information to judge advice quality), and regardless of whether the advisor’s professional responsibility and
self-interest are actually aligned or conflicted, and whether the
advice is self-serving or self-sacrificing. We demonstrate that the
key boundary condition to the disclosure penalty is whether there
is a salient external factor to which the presence of the advisor’s
personal self-interest can be attributed. The second effect is the
altruistic signal, which decreases desire to change advisors with
COI disclosure, and is present when advisors give advice that
sacrifices (i.e., runs counter to) their self-interests. The net
consequence of these two effects on advisees’ desire to change
advisors depends on which effect—the disclosure penalty or
altruistic signal—is stronger.
Conflicts of Interest and Advisor Bias
COIs are common across many professions: Financial advisors,
real estate agents, physicians and other advisors often receive a
(greater) commission if the advisee selects a particular financial
investment, house, medical treatment, or other product or service.
COIs have the potential to induce bias in advisors (Mullainathan,
Noeth, & Schoar, 2012; Sah & Fugh-Berman, 2013; Sah & Loewenstein, 2015) and disclosure is a popular solution as it alerts
advisees to potential bias in the advice. However, the effects of
disclosure are complex and the COI domain is more intricate than
previous studies reflect, with prior work primarily focusing on, or
89
assuming, the presence of poor-quality advice due to a COI (Sah
et al., 2013, 2018, 2019).
The presence of a COI, however, does not necessarily imply that
advice is biased (Lo & Ott, 2013; McCoy & Emanuel, 2017; Pretty
v. Prudential Insurance Company of America, 2010; Rosenbaum,
2015). Although those that possess a COI are more likely to give
biased advice, not all that possess such conflicts succumb to bias.
A COI, by definition, represents a risk that the advisors’ judgment
will be compromised, but not a determination that such a lapse has
actually occurred (Rothman, 1993).2
Further, in some cases, professional responsibility and personal
interest may not actually be in conflict. They could be aligned. For
example, a high-quality financial advisor may recommend an
investment fund that is the best for her client but also gives the
advisor a larger commission. The distinction between the presence
of a self-interest and advice quality is important for understanding
the psychological dynamics of COI disclosure. These concepts are
often confounded in policy arguments, and lead to considerable
contention between advocates of COI disclosures and advisors
who claim that their integrity ensures that they will give highquality advice and that no conflict actually exists between their
professional and self-interests. In this debate, the mere presence of
a self-interest is sometimes referred to as a confluence of interests
(Cappola & FitzGerald, 2015) or a potential (rather than an actual)
COI (Lo & Ott, 2013). Others reject this nomenclature and state
that there are no “potential” COIs, only the presence or absence of
a COI and the potential for bias (McCoy & Emanuel, 2017).
Another important dimension that we explore is that advisors
may sometimes advise an option that sacrifices their self-interest.
For example, consider the following real situation that took place
in a U.S. emergency room (ER). The patient, a teenage girl, had
been in a minor car accident. The patient’s father demanded a
computerized tomography (CT) scan of her neck. Upon completing a full history and examination of the patient, the ER physician
did not recommend the scan; there was little concern of a fracture
and no need for even an x-ray. The patient’s father, a lawyer, was
insistent threatening to lodge a complaint if the physician did not
order the scan. However, it was not in the best interests of the
patient, and CT scans can cause harm to the patient later in life due
to radiation exposure. The physician explained to the father that
the potential harm was not trivial but important. His daughter
could get cancer of her thyroid that in later years might actually
be fatal and so the physician believed it more harmful than
beneficial to do the scan. The father was not convinced. The
physician then explained that everything was pushing him to
perform the scan; the demands and subsequent explanations
were taking up the physician’s time, the avoidance of a potential
lawsuit that the father was threatening, and the fact that no one
would blame him in 20 years from now if his daughter got cancer.
1
This research was spurred by numerous conversations with both practitioners and academics who were unclear about the effect on the advisoradvisee relationship of disclosing COIs with high-quality advice.
2
Although a regulator can prosecute physicians who order treatments
that are obviously unnecessary (United States v. Campbell, 1988), it is
usually difficult to verify the existence and effect of bias from COIs. In
many situations in which expert advice is needed (e.g., financial, legal, or
medical decisions), multiple options exist, and it can be impossible to
determine which advice is best and least biased (Sah, 2015).
SAH AND FEILER
90
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Finally, the physician disclosed that he would receive extra compensation for performing the scan due to the U.S. fee-for-service
compensation model (Sah, 2015).3 The doctor disclosed that he
was sacrificing his self-interest for the patient and putting her
interests first by not recommending the scan (This American Life,
2009). This disclosure should serve as a positive altruistic signal,
increasing, rather than decreasing, trust in the doctor.
In this article, we systematically investigate how high-quality
advice accompanied with COI disclosure affects how advisees’
view their advisors and whether they wish to maintain the relationship with their advisor in the future or switch advisors.
The Disclosure Penalty and the Altruistic Signal
The Disclosure Penalty: Negative Attributions
Regarding the Advisors’ Character
COI disclosure (from any source, e.g., directly from the advisor,
a third party, a public website, etc.) informs advisees of their
advisors’ self-interest and thus, the advisee becomes aware of the
possibility that the advisor’s recommendation may be biased by his
or her self-interests. According to the judgment correction model
of decision-making, uncertainty in advice quality should cause the
advice to be discounted (Campbell & Kirmani, 2000; Friestad &
Wright, 1994; Van Swol, 2009). Consistent with this, prior research reveals that (given sufficient resources to deliberate on and
process the meaning of the COI disclosure) advisees report reduced trust in their advisors with COI disclosure (Hwong, Sah, &
Lehmann, 2017; Sah et al., 2013, 2018).
The judgment correction perspective suggests, however, that this
process will be activated only if there is some uncertainty as to the
advice quality. Such uncertainty stems from advisors usually having
greater expertise (more information) than advisees, making it difficult
for advisees to assess advice quality. This implies that if there is no
advisor–advisee information asymmetry, then advisees can judge
advice quality. High advice quality should therefore mitigate advisees’ concerns about the advice and subsequent distrust in the advisor.
However, drawing from attribution theory, the news of the COI
itself may affect perceptions of the advisor, specifically the trustworthiness of the advisor’s character. If COI disclosure provides not only
a signal regarding the potential (low) quality of the advice but also a
negative signal about the advisor’s character in general, then disclosure will lead to less favorable judgments of the advisor even when
advice quality is high. We propose that COI disclosure could cause
advisees to ask themselves not only “How does the conflict of interest
affect the advice?” but also “What kind of advisor enters a situation in
which he or she possesses a conflict of interest?” One of the foundational results in social psychology is the human tendency toward
attributing observed behavior to the character or disposition of the
individual rather than an external cause— often referred to as the
fundamental attribution error (Gilbert & Malone, 1995; Ross & Nisbett, 1991). Thus, even in the presence of high-quality advice, COI
disclosure will have an adverse consequence on the advisor–advisee
relationship, decreasing trust in the advisor’s character, and increasing desire to change advisors. We refer to this effect as the disclosure
penalty: The mere knowledge of the advisor’s personal self-interest,
even in the presence of high-quality advice, will yield negative attributions about the advisor’s character.
At first glance, the disclosure penalty may appear intuitive.
However, the attribution lens we apply results in three important
(and less intuitive) dynamics that we predict and examine experimentally. First, the disclosure penalty will emerge irrespective of
advice quality; in fact, it will emerge even when advice quality is
high, and advisees have full information to judge this quality.
Second, the disclosure penalty will emerge even when the advisor’s professional responsibility and self-interest are not actually in
conflict. If we consider COIs only in terms of uncertainty in advice
quality, bias, and compliance with advice (as much of the prior
literature does), we would miss that the mere presence of a COI
creates negative attributions about the advisor that effects trust and
the advisor–advisee relationship despite clear high-quality advice.
Third, the disclosure penalty may be mitigated if advisees are
explicitly informed that their advisor’s COI arose from external factors beyond the advisor’s control. To elaborate on this final dynamic,
consider that advisors’ COIs usually fall into two main categories.
One category contains those conflicts created by external factors
beyond the advisor’s control, sometimes referred to as systemic COIs
(Williams, Mayes, Komesaroff, Kerridge, & Lipworth, 2017). For
example, the physician fee-for-service model, which is the predominant form of physician payment in the U.S., or financial advisors who
work in institutions that compensate them with commissions. The
other category contains those conflicts created by the advisor’s own
volition, sometimes referred to as personal COIs (Williams et al.,
2017), such as entering a relationship with a third party who may have
other interests. For example, physicians accepting gifts or sponsorships from the pharmaceutical industry. Advisors who engage in the
personal category are often viewed as biased and untrustworthy
(Rosenbaum, 2015). According to the fundamental attribution error,
one should expect that advisees will generate (whether or not they are
fully cognizant of it) a personal attribution regarding their advisors’
COIs, regardless of whether the COI is actually derived systemically
or personally.4 This personal attribution will cause the advisee to
distrust the advisor, regardless of advice quality, whether the advisor’s
actual recommendation is actually in conflict with what is best for the
advisee and whether the advice is self-serving or self-sacrificing (see
Figure 1).
Following our theory, if the disclosure penalty arises from advisees
generating a personal attribution regarding their advisors’ COIs, the
penalty could be mitigated if advisees are informed that the presence
of their advisor’s COI is due to a systemic cause (i.e., an external
cause beyond the advisor’s control). Understanding the effects of
personal versus systemic attribution for the existence of the advisor’s
COI has implications for the role that norms and the salience of the
systemic attribution play in determining whether disclosure damages
the advisor–advisee relationship. If all advisors have COIs and this is
salient in the advisees’ mind, then having a COI does not reflect
anything about the advisor’s character (Sabini, Siepmann, & Stein,
2001), and the disclosure penalty may be eliminated (a point we return
to in the General Discussion section).
3
The physician (or hospital) would receive more compensation for more
complex patients and CT scans can serve as an indication to insurance
companies of a more complex patient.
4
One can argue as to whether any COI can be considered systemic if
advisors ultimately have to accept the presence of the COI to operate
within that sphere of influence.
DISCLOSURE PENALTY
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Different COI situaons
1.*
Low advice quality –
Self-serving advice –
2.
High advice quality –
Self-serving advice –
3.
High advice quality –
Self-sacrificing advice –
4.
High advice quality –
Self-sacrificing advice –
Salient external attribution
for the existence of the COI
Disclosure Penalty:
Awareness of a COI
91
Altruisc Signal:
Self-sacrificing advice
Figure 1. The disclosure penalty and altruistic signal effects on trust. The quality of advice (low or high) is not
relevant for the disclosure penalty, which reduces trust and is present regardless of the advice quality. The signal
from self-sacrificing advice increases trust. When advisees learn about their advisors’ COI, these signals are
activated. In Situations 1, 2, and 3, there is no salient external attribution for the existence of the COI. ⴱIndicates
COI situation commonly examined in past research. See the online article for the color version of this figure.
Furthermore, the disclosure (news of the COI) may come from
any source (i.e., directly from the advisor, a third party, or public
website). As has been shown in prior research (and confirmed by
experiments shown in the online supplemental material), the
source of the disclosure does not have a large effect on advisees’
trust—it is the information regarding the COI that reduces trust in
the advisor regardless of the source (Sah et al., 2013, 2019).5
sacrifices her self-interest (the altruistic signal)—are opposing
and the net effect on trust depends on which effect is stronger
in a given circumstance.
The Present Research
Background and Overview
The Altruistic Signal: Positive Attributions Regarding
the Advisors’ Character
We contrast the disclosure penalty effect with a second dynamic
by which COI disclosure can increase trust: When advisors recommend an option that sacrifices their self-interest, COI disclosure
reveals not only the existence of the conflict to the advisee but also
that the advisor has made a self-sacrificing recommendation. This
self-sacrifice would be unknown to the advisee without disclosure
and, once revealed, provides a positive signal regarding the advisor’s character. We refer to this signal as the altruistic signal.
The altruistic signal should increase trustworthiness attributions
regarding the advisor. The advice “proves” that the advisor prioritizes
his or her integrity and benevolence over personal gain. By contrast,
recommendations that are consistent with both the advisors’ and
advisee’s interests are ambiguous as to the extent to which they are
driven by self-interests versus professional interests.
The disclosure penalty and the altruistic signal create the following dynamic: First, awareness of the COI decreases advisees’
trust in their advisors, resulting in a disclosure penalty. The penalty
may be mitigated if advisees are explicitly informed that the origin
of the COI is external, that is, beyond the advisor’s control. A
second independent effect is that COI disclosure can increase
trust if the disclosure reveals that the recommendation sacrificed the advisor’s self-interest. These two independent effects
of COI disclosure— decreasing trust regardless of advice quality (the disclosure penalty) and increasing trust if the advisor
In this article, we focus on cases of disclosure with high-quality
advice. However, to establish a baseline with our paradigm for the
typical COI situations examined in prior research, we first conducted two experiments (reported in the online supplemental material) with poor-quality advice (N ⫽ 456). Similar experiments
have been conducted in past research (Sah et al., 2013, 2019).
Sah, Loewenstein, and Cain’s (2013) article merits extended
discussion here not only for its findings on trust with poor-quality
advice, but also because its methods are the basis for the present
research. Sah et al. (2013) conducted experiments in which advisors instructed advisees to select between two die-roll lotteries (A
or B). The die-roll lotteries resulted in different specific prize sets
(for instance, a $5 Starbucks gift card if a “5” was rolled on Die
A). The prizes for the Die-Roll A were superior in value to those
of Die-Roll B (more than two times the expected value, and
without advice more than 90% of people preferred Die-Roll A over
B). Advisors, however, were subject to a COI in that they would
5
In some situations, it might be obvious that a systemic COI exists (e.g.,
a buyer’s agent in a real estate transaction). This clarity would mitigate the
need for disclosure altogether and eliminate the disclosure penalty if the
systemic external attribution is salient. However, many advisees are not
aware of (or not actively deliberating on) the presence of the advisors’ COI
and take advice at face value (Malmendier & Shanthikumar, 2007). Our
proposition is that advisees will tend to make an internal (personal)
attribution regarding the advisor’s COI unless explicitly informed about the
external (systemic) attribution.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
92
SAH AND FEILER
receive a bonus if the advisee chose the inferior Die-Roll B. The
majority (approximately 80%) of these conflicted advisors gave
poor-quality advice and recommended the inferior Die-Roll B.
Conflicted advisors were randomized to either disclose their COI
in a written statement to advisees or were informed not to disclose
this COI to their advisee. Across six experiments, Sah et al. (2013)
demonstrated that advisees in the disclosure condition are more
likely to report decreased trust in their advisors than advisees in the
nondisclosure condition. Arguably, this is the “intended” purpose
of COI disclosure. Sah et al. (2013) also revealed that when
advisees had to make their choice in front of their advisor, they felt
greater pressure to comply with their advisor’s recommendation with
disclosure than without. This pressure resulted in a perverse effect of
increased compliance with distrusted advice in the presence of a COI
disclosure. Increased compliance occurred even when advisees had
full information of the lottery choices for both die-rolls and thus could
tell that the advice was of low quality. Pressure to comply was
reduced when advisees could make their choice in private away from
their advisors. In the current article, we reduce the pressure to comply
with the advisor by giving the advice in written form online away
from the physical presence of an advisor.
We report our two experiments with poor-quality advice in
detail in the online supplemental material (S1 and S2; N ⫽ 456,
one preregistered). To summarize, COI disclosure decreased trust
in the advisor (conceptually replacing Sah et al.’s, 2013 findings)
and we also found that it increased desire to change advisors. Like
prior research (Sah et al., 2013, 2019), we found that the decrease
in trust from COI disclosure was similar regardless of whether the
disclosure from the advisor was perceived as mandatory or voluntary (Experiment S1) or if the disclosure was directly from the
advisor or made indirectly via a third party (Experiment S2).
Although these findings may be surprising—for example, one may
expect increased trust with voluntary (vs. mandatory) disclosure or
less trust with third-party disclosure (vs. disclosure directly from
the advisor)—as explained in prior research (Sah et al., 2013,
2019), the framing and source of the disclosure do not change the
fact that the information disclosure, however received, alerts the
advisee to the fact that their advisor has a COI.
In the six experiments reported here (N ⫽ 1,766, one preregistered), we examine the impact of COI disclosure with high-quality
advice. The first three experiments examine high-quality advice
when advisor’s professional responsibilities and their self-interests
are aligned; the second three experiments examine high-quality
advice in the presence of misaligned interests and with advisors
that sacrificed their self-interests (see Figure 2). In Experiments 1
to 5, advisees were clients and were given “financial advice”
which consisted of a recommendation to choose one of two lotteries. Like Sah et al. (2013), one lottery had a better expected
outcome for the client—and was preferred by most people in the
absence of advice. While building a set of experiments within a
given experimental paradigm is useful for comparing the robustness of the observed effect and interventions across studies, it is
also useful to conceptually replicate findings in other contexts for
generalizability. Therefore, our final experiment moved to a medical context to examine whether disclosure and self-sacrificing
advice can increase a patient’s trust in his or her doctor.
In each experiment, we predicted how the independent variables
would affect the advisees’ desire to change their advisor for a
second round of advice and their perceptions of their advisors’
High quality advice
(aligned interests;
self-serving advice)
High quality advice
(misaligned interests;
self-sacrificing advice)
Experiments 1, 2 and 3
Experiments 4, 5, and 61
Option for which the
advisor receives a bonus
Superior Option
(Portfolio A)
Inferior Option
(Portfolio B)
The advisor’s
recommendation
Superior Option
(Portfolio A)
Superior Option
(Portfolio A2)
Figure 2. Conflict of interest situations with high-quality advice. Experiments 1, 2, and 3 present a scenario in which the advisor’s professional
and self-interest are aligned. Experiments 4, 5, and 6 present a misaligned
scenario, similar to the common COI scenario examined in prior research.
In all these experiments, advisors recommend the superior option. In the
common COI scenario typically studied in prior research, the advisor gives
low-quality advice: recommends the inferior option (see Experiments S1
and S2 in the online supplemental material). 1Experiment 6 is in the
medical context (there are no portfolio options). 2Experiment 4 examines
advisors recommending either Portfolio A (the superior portfolio) or Portfolio B (the inferior portfolio). Our focus is on advisees’ reaction to
receiving high-quality (Portfolio A) advice but we include low quality
advice as a basis for comparison in this experiment. See the online article
for the color version of this figure.
trustworthiness (as reflected in our preregistrations of Experiments
S2 and 6). We report, but did not make predictions about, investment or treatment choice (whether participants ultimately choose
the option recommended by their advisor). We anticipated that
while choice is materially consequential, it does not accurately
reflect how advisees feel about their advisor and the downstream
consequences of maintaining the relationship with their advisors,
which is our main subject of interest. Problems with interpreting
compliance behavior as an operationalization of trust occur because
although compliance can arise from trust, it can also arise from other
causes (Kelman, 1958; Kramer, 1999). For example, compliance with
an advisor’s recommendation may stem from feeling pressure to
comply in the presence of the advisor (Sah et al., 2013, 2019) or
nonconscious anchoring on the recommendation even when attempting to disregard the advice (Tversky & Kahneman, 1974). Choice
may be of greater interest in studies in which advisors give poorquality advice (to see whether the advisee was manipulated). However, in all our studies, advice was of high quality and in Experiments
1 to 3, there was no information asymmetry, so advisees had full
information to judge advice quality. Advice quality and lottery preference are likely to play a large role in choice whereas our interest is
in whether COI disclosure affects trust and the desire to change
advisors (regardless of compliance).
Trust and Trustworthiness
Trust is essential in advisor–advisee relationships, and has been
defined as a psychological state comprising the intentions to accept
vulnerability based on positive expectations of the actions of the
trustee (Rousseau, Sitkin, Burt, & Camerer, 1998). In prior research on COI disclosure, trust has been measured by simply
asking advisees if they believed that their advisor had their best
interests at heart (Sah et al., 2013), or using Mayer, Davis, and
Schoorman’s (1995) tridimensional trustworthiness measure of the
advisor’s integrity, benevolence, and ability/expertise (Sah, Fagerlin, & Ubel, 2016; Sah et al., 2018). These three “trustworthiness”
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
DISCLOSURE PENALTY
characteristics of the advisor are viewed as antecedents to “trust”
and capture suggestions from prior work that trustworthiness perceptions are composed of both the competence and character of
the trustee (Kee & Knox, 1970). Competence and character are
likely to have unique relationships with trust (Colquitt, Scott, &
LePine, 2007). Competence captures the “can do” component of
trustworthiness which corresponds to Mayer et al.’s (1995) ability
(or expertise) measure, and Mayer et al.’s (1995) character variables of integrity and benevolence capture the “will do” component. Colquitt, Scott, and LePine (2007) argue that these two
character facets— benevolence and integrity—may be redundant
with each other in many situations. We hypothesized that advisees’
negative attributions due to their advisor’s self-interest would
affect perceptions of their advisor’s character (Ross & Nisbett,
1991), that is, their advisor’s integrity and benevolence. We do not
focus on, nor make predictions regarding, advisees’ perceptions of
the advisor’s competence or expertise: Indeed, in the first three
experiments, we eliminate advisor’s expertise by giving advisees
full information, so they can assess advice quality.
Desire to change advisors. The main dependent variable for
Experiments 1, 3, 4, and 5 is our behavioral measure of trust which
was the clients’ desire to either keep or change their advisor for a
future investment decision. In Experiment 2, we use a different
behavioral measure of trust which examines the cost clients are
willing to incur to limit their vulnerability to their advisor
(McEvily, Radzevick, & Weber, 2012). In Experiment 6, which
involved a medical context, we measure advisees’ willingness to
listen to the doctor for future decisions. All these measures for trust
incorporate positive expectations, accepting vulnerability, and a
willingness to continue the relationship (Rousseau et al., 1998).
Perceptions of the advisors’ trustworthiness. Our measure
of trustworthiness consisted of responses to five statements that
captured perceptions of the advisor’s character. Our scale was the
averaged responses, measured on a 7-point scale from 1 (strongly
disagree) to 7 (strongly agree), to the following statements: “My
advisor seemed like an honest person,” “My advisor had my best
interests at heart,” “My advisor put my interests first,” “I trusted
my advisor’s recommendation,” and “I did not trust my advisor”
(reverse coded). In Experiments 2 (and S2), we included additional
measures for trustworthiness adapted from Mayer and Davis’s
(1999) tridimensional construct (integrity, benevolence, and competence/expertise) to ensure that these dimensions corresponded
with our measures of character and competence. This, indeed, was
the case.6 Because trustworthiness is an antecedent to trust, we
also examined whether perceptions of the advisor’s character mediated the relationship between disclosure and trust (desire to
change advisors).
Experiment 1: High-Quality Advice Accompanied
With Conflict of Interest Disclosure Reduces Trust
In the first three experiments, advisors (a) recommend the
superior Portfolio A and (b) receive a bonus if clients select
Portfolio A (see Figure 2). Even though professional and selfinterests are aligned for advisors, our theory predicts that merely
knowing that one’s advisor has a self-interest generates skepticism
regarding the advisor’s character, irrespective of advice quality.
Therefore, we predicted that COI disclosure would result in the
disclosure penalty, increasing desire to change advisors, as well as
93
decreasing perceptions of the advisor’s trustworthiness, which
would mediate the relationship between disclosure and desire to
change advisors.
This study was also designed to examine whether distrust from
COI disclosure resulted from increased uncertainty in advice quality due to information asymmetry between the advisor and client
(usually present in advice-giving due to the advisors’ expertise) or
was present even when the uncertainty in advice quality was
removed. When a client has complete information of the lottery
rewards, she can more easily assess the (high) quality of the
advisor’s recommendation. If the advisor is giving objectively
superior advice, this should mitigate negative trustworthiness inferences made from the COI disclosure if advice quality is the
client’s main concern. However, our theory suggests that negative
attributions arise from the presence of the advisor’s self-interest,
irrespective of advice quality. We therefore predicted that trust
should be insensitive to whether the client has complete or incomplete information about the portfolios. If COI disclosure reduces
trust even with complete information, then we can more confidently conclude that the disclosure penalty stems from negative
attributions toward the advisor’s character for merely possessing
the self-interest, as opposed to some feature of, or uncertainty in,
the recommendation itself. The complete information case can be
conceived of as asking for advice from a colleague that does not
necessarily have more expertise than you, but can still provide a
second, independent opinion about what the best course of action
for you may be. As Northcraft and Neale (1987) claim, in the real
world, not every advisor is an expert.
Method
In this, and the following experiments, we report how we
determined our sample size, all data exclusions (if any), all manipulations, and all measures in the study (Simmons, Nelson, &
Simonsohn, 2011). In all our experiments, analyses were conducted only once data collection for that experiment was finished.
Our study received ethics (Institutional Review Board) committee
approval.
Participants and design. We requested 50 –100 participants
per condition from a survey company, ROI Rocket (www.roirocket
.com), and received 374 participants (183 women, 187 men, four
gender unreported, Mage ⫽ 40.2, SD ⫽ 10.6) who were randomly
assigned in a 2 (Disclosure vs. Nondisclosure) ⫻ 2 (Client Information: Complete vs. Incomplete) between-subjects design. All
invited were U.S. residents, between 22 and 60 years of age,
employed full-time, and with a bachelor’s degree. Participants
received $2.25 for completing the study plus an opportunity for a
bonus prize up to $25 in value.
Procedure. Participants played the role of a client deciding to
invest in one of two lotteries: Portfolio A or Portfolio B. The task
was adapted from Sah et al. (2013) and titled “The Investment
Challenge.” In all our experiments within this paradigm, “A” is the
superior portfolio and “B” is the inferior portfolio and the same
6
For all experiments, we also measured responses to four statements
that captured the advisor’s competence or expertise— our “can do” trust
component. All these results are reported in the online supplemental
material. No other items were collected, except where otherwise reported
in this article.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
94
SAH AND FEILER
portfolios were used in each experiment. Both lotteries had six
possible outcomes corresponding to numbers on a die, ranging
from $5 to $25 in value (see Figure 3). Portfolio A had an expected
value of $17 with greater spending flexibility, whereas the expected value for Portfolio B was $13 with less spending flexibility.
The superiority of Portfolio A was also verified in a pilot survey of
93 independent participants from a U.S university (49 women, 44
men, Mage ⫽ 20.1, SD ⫽ 1.0), in which 93% stated that they
preferred Portfolio A over Portfolio B, 2(1) ⫽ 70.55, p ⬍ .001,
⫽ 0.87, when viewing the portfolios with complete information
about possible outcomes. Clients were informed that two participants would be randomly selected upon completion of the study
and awarded a bonus based on their choice of portfolio as shown
in Figure 3 and the outcome of a fair die roll.
Participants were randomly assigned to one of two information
conditions which manipulated how the outcome of a die-roll of “6”
was presented. In the incomplete information condition, the
outcome for rolling a “6” was presented as “???” for both
portfolios (see Figure 3). In the complete information condition,
clients were shown that the sixth outcome was “re-roll” for both
portfolios. Therefore—and in contrast to the incomplete information condition—with complete information there was no
information asymmetry between the advisors and the clients,
providing the client with all the information they need to assess
recommendation quality.
After viewing the portfolios, clients received advice about
which portfolio to choose from a “financial advisor” who had
observed all six of the possible outcomes for each lottery. Clients
read their advisors’ recommendation stated as “I’m looking at all
six of the dice rolls for each portfolio and I recommend that you go
with Portfolio A.”
Participants were randomly assigned to either a disclosure or
nondisclosure condition. In the disclosure condition, the following
sentence preceded the advisor’s recommendation: “I am required
to tell you that I get a $10 bonus if you choose Portfolio A.” This
statement was adapted from disclosure statements used in prior
research (Sah et al., 2013) and conforms to regulations which
Figure 3. Portfolio outcomes for Experiments 1 to 5. Rolling the number
“6” always yielded a reroll of the die for both portfolios. In Experiments 1
to 3, clients were informed of this (the “???” was replaced with “re-roll”),
thus there was no advisor–advisee information asymmetry and clients
could assess advice quality. See the online article for the color version of
this figure.
require COI disclosures to be simple, concise, direct, and conspicuous (Securities and Exchange Commission, 2010). The nondisclosure condition was identical absent the disclosure sentence.
After reading their advisor’s recommendation, clients made a
choice of either Portfolio A or B, reported their trustworthiness
perceptions of the advisor’s character (Cronbach’s alpha ⫽ .83),
and whether they would want to change advisors.
Results
Desire to change advisors. We used logistic regression to
estimate the effect of disclosure and information completeness on
desire to change advisors (see Figure 4). There was no interaction
between disclosure and portfolio information, b ⫽ 0.20, SE ⫽
0.13, Wald ⫽ 2.38, p ⫽ .12. Consistent with our hypothesis, COI
disclosure increased the desire to change advisors (33%) compared
with nondisclosure (14%), b ⫽ 0.57, SE ⫽ 0.13, Wald ⫽ 18.51,
p ⬍ .001. This was true both within the incomplete information
conditions, b ⫽ 0.73, SE ⫽ 0.37, Wald ⫽ 4.02, odds ratio (OR) ⫽
2.08, 95% CI [1.02, 4.17], p ⫽ .04, and within the complete
information conditions, b ⫽ 1.55, SE ⫽ 0.39, Wald ⫽ 16.22,
OR ⫽ 4.72, 95% CI [2.22, 10.04], p ⬍ .001. Complete portfolio
information (24%) had no significant effect on requests to change
advisors compared with incomplete information (22%), b ⫽ 0.001,
SE ⫽ 0.13, Wald ⫽ 0.01, p ⫽ .99.
Trustworthiness. Similarly, COI disclosure decreased trustworthiness perceptions of the advisor’s character (M ⫽ 4.54, SD ⫽
1.01) compared with nondisclosure (M ⫽ 4.96, SD ⫽ 0.88), F(1,
370) ⫽ 18.68, p ⬍ .001, p2 ⫽ .05. Again, this effect occurred
within both the incomplete information conditions: disclosure
(M ⫽ 4.48, SD ⫽ 1.04) versus nondisclosure (M ⫽ 4.96, SD ⫽
0.91), F(1, 370) ⫽ 11.85, p ⬍ .001, p2 ⫽ .03; and the complete
information conditions: disclosure (M ⫽ 4.59, SD ⫽ 0.98) versus
nondisclosure (M ⫽ 4.96, SD ⫽ 0.86), F(1, 370) ⫽ 7.09, p ⫽ .008,
p2 ⫽ .02. Again, there was no effect of information (complete
information: M ⫽ 4.78, SD ⫽ 1.01 vs. incomplete information:
M ⫽ 4.72, SD ⫽ 1.01), F(1, 370) ⫽ 0.36, p ⫽ .55, p2 ⫽ .001, and
no interaction between disclosure and portfolio information, F(1,
370) ⫽ .34, p ⫽ .56, p2 ⫽ .001.
Bootstrapping mediation analysis (Hayes, 2013; Model 4), using
10,000 random samples with replacement (MacKinnon, Fairchild,
& Fritz, 2007), revealed that trustworthiness perceptions significantly mediated the relationship between disclosure and the desire
to change advisors (0.57, 95% CI [0.27, 0.95]).7
Portfolio choice. Portfolio choice is not the focus of this
article, but we report the analyses nevertheless. There was no
interaction between disclosure and information, b ⫽ 0.20, SE ⫽
0.76, Wald ⫽ 0.07, p ⫽ .79. Clients who received a COI disclosure
were less likely to select the superior Portfolio A (85%)—thereby
rejecting their advisor’s high-quality recommendation—than clients with no disclosure (93%), b ⫽ 1.24, SE ⫽ 0.60, Wald ⫽ 4.29,
p ⫽ .04. This effect was present in both the incomplete information conditions (OR ⫽ 0.35, 95% CI [0.14, 0.90]) and complete
information conditions (OR ⫽ 0.29, 95% CI [0.09, 0.94]). The
amount of information had no effect on portfolio choice (87% with
7
The bootstrapping mediation method provides advantages to, and has
largely replaced, the former Baron and Kenny (1986) mediation step
method (see online supplemental material for more information).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
DISCLOSURE PENALTY
95
Figure 4. The percent of participants who reported they would want to change advisors in Experiment 1, by
condition. In all conditions, the advisor’s self-interest was aligned with what was best for the client and the
advisor recommended the superior portfolio. Disclosure increased the desire to change advisors, relative to
nondisclosure, when the client had incomplete information about the portfolios (advisee-advisor information
asymmetry) and also when the client had complete information (no advisee-advisor information asymmetry).
Error bars are ⫾1 SE. See the online article for the color version of this figure.
incomplete information vs. 91% with complete information selected Portfolio A), b ⫽ 0.64, SE ⫽ 0.64, Wald ⫽ .98, OR ⫽ 1.64,
95% CI [0.84, 3.23], p ⫽ .32.8
ing advisors, and potentially reject valuable advice. This experiment explored the price clients would pay to avoid trusting honest
advisors who disclose a COI.
Discussion
Method
Even when an advisor’s professional and self-interests are
aligned, and the superior option is recommended, COI disclosure
increased desire to change advisors. Put differently, even with
honest advice (no bias produced from the COI) and no information
asymmetry between advisors and advisees, disclosure of a selfinterest significantly decreased trust in the advisor’s character
leading to an increased desire to change advisors.
Importantly, disclosure was damaging for both advisors and
clients. Advisors were trusted less despite giving high-quality
advice. And, clients’ lack of trust in the advisor caused them to
both desire new advisors (which has switching costs for clients in
the real-world) and select inferior options for themselves, even
when they could assess the quality of advice (with complete
information, 13% of clients choose the inferior option with disclosure, compared with 4% with nondisclosure, b ⫽ 1.24, SE ⫽
0.60, Wald ⫽ 4.29, OR ⫽ 3.44, 95% CI [1.07, 11.10], p ⫽ .04).
This increased rejection of high-quality advice may be due to a
wish to “punish” advisors for possessing a COI. People sometimes
forego opportunities for gain or harm themselves if they perceive
others to lack integrity or behave unfairly, as seen in the rejection
of unfair offers in the ultimatum game (Thaler, 1988). Our next
experiment examines if distrust in the advisor’s character generated by the COI disclosure causes people to pay more (i.e., incur
financial costs) to avoid trusting that advisor in the future.
Participants and design. We requested 50 –100 participants
per condition from a U.S. university undergraduate research lab,
and received 164 participants (93 women, 71 men; Mage ⫽ 20.8,
SD ⫽ 1.7) who were randomized into one of two conditions,
disclosure or nondisclosure. Participants received $5 for completing the study, and a potential bonus of up to $25.
Procedure. Clients were informed that they would complete
two tasks with the same advisor. The first task was identical to the
investment portfolio decision in the previous experiment, with all
advisors recommending the superior Portfolio A, and this time all
clients had complete information on the portfolios (die-roll OF “6”
was known to be “re-roll”). In the disclosure condition, clients
received the same disclosure statement from their advisor as in the
previous experiment. In the nondisclosure condition, as previously,
there was no statement. Clients then reported their portfolio
choice, and trustworthiness perceptions (␣ ⫽ .76).9
Clients, however, did not report their desire to change advisors.
Instead, clients proceeded to the second task with the same advisor. This task, adapted from McEvily, Radzevick, and Weber
(2012), served to produce an incentivized behavioral measure of
Experiment 2: Paying to Avoid Honest Advisors Who
Disclose a Self-Interest
Experiment 1 revealed that COI disclosure has a cost—it causes
clients to trust their (honest unbiased) advisors less, desire chang-
8
Portfolio choices for the remaining experiments are reported in the
online supplemental material.
9
Our trustworthiness items use the word “partner” in this study. To
ensure correspondence of this measure with other trustworthiness measures, we also measured the constructs of integrity, benevolence, and
ability adapted from Mayer and Davis (1999). We predicted, and found,
that the integrity and benevolence dimensions revealed the same patterns of
significance as our trustworthiness measure of the advisor’s character, and
the ability/expertise dimension behaved similarly to our competence measure (all these analyses are reported in the online supplemental material).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
96
SAH AND FEILER
trust, which again reflects their willingness to be vulnerable to the
same advisor (Rousseau et al., 1998) but this time with a more
granular financial measure in which we can determine the cost to
clients.
Clients were informed that their advisor had previously been
asked to divide $10 (anywhere from $0 to $10, in 50 cent increments) and the split would determine the allocation the client
would receive—like a dictator game. Clients stated their preference between receiving a guaranteed bonus of a certain amount $X
(Option A; see Figure 5) or to accept whatever portion of the $10
their advisor had stated would be allocated to them (Option B).
Option A was then presented to them starting from $6 down to $0,
in 50 cent increments and clients reported for each $X, whether
they would prefer to accept that (Option A) or let their advisor
decide the split (Option B). The $X amount at which clients switch
from Option A to B reflects a financially incentivized behavioral
measure of trust in the advisor. Switching to Option B at a
relatively high guaranteed $X amount would reflect a high level of
trust in the advisor. For example, switching to Option B at $4.50
would be consistent with an expectation that the advisor will
provide (at least) a 50/50 split of the $10. On the other hand,
sticking with Option A at even low guaranteed amounts of $X
reflects low trust in the advisor. For example, if a client switches
to Option B at $1.50, this suggests that the client expects the
advisor to provide a split of the $10 that will not favor the client.
We predicted that clients in the COI disclosure condition in the
first task would choose a lower financial dependence on the
advisor on this measure by switching to Option B at lower $X
amounts than clients in the nondisclosure condition.
Results
Financial trust level. As predicted, the highest potential guaranteed amount ($X) at which the client would prefer to trust their
advisor to split the $10 over receiving $X was significantly lower
with disclosure (M ⫽ $3.13, SD ⫽ 1.47) than nondisclosure (M ⫽
$3.69, SD ⫽ 1.27), F(1, 162) ⫽ 6.59, p ⫽ .01, p2 ⫽ .04. In other
words, if their advisor had disclosed a COI in the prior task, they
expected the advisor to leave them a mean of $0.56 (15%) less than
if their advisor had not disclosed a COI in the prior task. This was
despite clients having full information to clearly assess the highquality recommendation given to them, demonstrating a real financial cost of COI disclosure on advisees.
Trustworthiness. Similar results were reflected with our
trustworthiness measure which was significantly lower with disclosure (M ⫽ 3.80, SD ⫽ 0.97) compared with nondisclosure
(M ⫽ 4.13, SD ⫽ 0.78), F(1, 162) ⫽ 5.71, p ⫽ .02, p2 ⫽ .03.
Figure 5. Measure of financial distrust in Study 2. In this example, for guaranteed amounts $6 to $3.50, the
participant preferred Option A, but preferred Option B for lower guaranteed amount, $3 to $0. In this case, the
measure of financial distrust would be $3; the highest amount the participant would prefer to let the partner
decide the split. See the online article for the color version of this figure.
DISCLOSURE PENALTY
Trustworthiness perceptions also mediated the effect of disclosure
on the financial trust level (0.29, 95% CI [0.06, 0.57]).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Discussion
Again, consistent with our hypotheses, with high-quality advice,
COI disclosure decreased trustworthiness perceptions and clients
were willing to accept lower bonuses to avoid trusting their advisor
again. This occurred even though clients had full information and
could, therefore, assess the quality of advice. This demonstrates
clients’ negative reaction to advisors with COI disclosure causes
them to forego opportunities for financial gain in order to avoid
their advisor. Similar findings for our financial measure of trust in
this study and our trust variable in the prior study (desire to change
advisors) help to establish the convergent validity of these results.
Experiment 3: High-Quality Advice Accompanied
With Voluntary Disclosure Still Reduces Trust
The experiments thus far demonstrate that even with high advice
quality and full information, the disclosure penalty creates a negative signal regarding the advisor’s character. To further test the
robustness of the disclosure penalty, in this experiment we used
disclosure in way that conveyed a voluntary sense, using the
words—“I would like you to know that . . .” rather than our
mandatory disclosure used in Experiments 1 and 2 (“I am required
to tell you that . . .”).
Prior research (Sah, Loewenstein, & Cain, 2019) and our Experiment S1 (in the online supplemental materials) have revealed
similar levels of trust when COI disclosures are deemed voluntary
versus mandatory, suggesting the information about the presence
of the advisor’s COI is more influential in assessments of trust than
the framing of the disclosure. That is, voluntary disclosures did not
increase, or even preserve, trustworthiness perceptions of the advisor’s character over mandatory disclosures. To establish if similar effects are present when advice is of high-quality, this experiment utilized the voluntary disclosure framing. With high-quality
advice (and no information asymmetry) and a voluntary disclosure
of a self-interest, clients may perceive advisors as honest and
increase their trustworthiness perceptions of the advisor’s character. However, we predicted that voluntary disclosure would behave
similarly to mandatory disclosure because it should raise the same
concerns over why the advisor has a COI, and thus invoke the
disclosure penalty, negative attributions regarding the advisor’s
character for possessing a COI, rather than rewarding advisors
with increased trust for their honesty and high-quality advice.
Method
Participants and design. We requested 50 –100 new participants per condition from an undergraduate research lab at a private
U.S. university, and received 114 participants (67 women, 46 men,
one gender unreported, Mage ⫽ 20.8, SD ⫽ 1.72) who were
randomly assigned in a two-cell (voluntary disclosure vs. nondisclosure) between-subjects design. Participants received $3 for participating in the experiment with a chance of winning a $25 bonus.
Procedure. Clients followed the same procedure and measurements (including trustworthiness, ␣ ⫽ .83) as in Experiment 1.
The nondisclosure condition was the same as in prior experiments.
97
In the disclosure condition, the following sentence preceded the
recommendation: “I would like you to know that I get a personal
bonus of $10 if you choose Portfolio A.” Again, all clients were
shown that the sixth outcome was “re-roll” for both portfolios.
Therefore, there was no information asymmetry between the advisors and the clients, and the clients could assess the quality of
advice given to them.
Results
Desire to change advisors. As predicted, clients in the voluntary disclosure condition (50%) were significantly more likely
to want to change advisors than those in the nondisclosure condition (27%), b ⫽ 0.50, Wald ⫽ 6.32, OR ⫽ 2.73, 95% CI [1.25,
5.99], p ⫽ .01.
Trustworthiness. Trustworthiness perceptions of the advisor’s character was significantly lower with voluntary disclosure
(M ⫽ 3.89, SD ⫽ 1.01) than without disclosure (M ⫽ 4.53, SD ⫽
0.92), F(1, 112) ⫽ 12.64, p ⬍ .001, p2 ⫽ .10, and this measure, as
predicted, mediated the effect of disclosure on the preference to
change advisors; bootstrap analysis estimated the size of the indirect effect as 0.90; 95% CI [0.36, 1.71].
Discussion
This experiment used voluntary disclosure, as opposed to the
mandatory disclosure implemented in Experiments 1 and 2. Consistent with the previous results with mandatory disclosure, this
experiment documented that COI disclosure decreases trust in
advisors, even when the advisor’s personal and professional interests are not in conflict, the disclosure is deemed voluntary and the
advice quality is high. See also Experiment S2 in the online
supplemental material in which we present evidence showing that
third-party (indirect) disclosure yields similar results to voluntary
disclosure directly from the advisor.
The next three experiments examine the net effect of disclosure
penalty (which decreases trust) and the altruistic signal that occurs
when advisors sacrifice their self-interests (which should increase
trust) to give clients high-quality unbiased advice.
Experiment 4: The Play-Off Between the Disclosure
Penalty and the Altruistic Signal
In the next three experiments, we transition to examining the
effect of disclosure on trust when the advisor’s professional and
self-interests are in conflict (i.e., the advisor receives the bonus if the
client chooses the inferior option). In this experiment, we also randomly manipulate whether advisors give biased poor-quality advice
that satisfies their self-interest or unbiased high-quality advice that
sacrifices their self-interest. If advisors give biased advice, then
disclosure should decrease trust due to the disclosure penalty
(replicating Experiment S1 in the online supplemental material).
If advisors give unbiased high-quality advice, thereby sacrificing
their self-interests, our theory generates competing predictions for
the effect of disclosure on trust (see row 3 in Figure 1). The
disclosure penalty suggests that knowledge of the mere presence of
the advisor’s self-interest should decrease trust. However, with
disclosure, making a recommendation that sacrifices self-interest,
produces the altruistic signal which should increase trust. The net
98
SAH AND FEILER
effect on trust with disclosure depends on which opposing effect—
the disclosure penalty or the altruistic signal—is stronger.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Method
Participants. We requested 100 participants per condition
from ROI Rocket Survey Company. A higher-than-anticipated
response rate (and the absence of a formal quota restriction in their
software), resulted in 541 participants (287 women, 254 men,
Mage ⫽ 42.1, SD ⫽ 10.1) completing the study for $2.25 and a one
in 100 chance to win a bonus up to $25.
Design and procedure. Clients were randomized to a 2 (Nondisclosure vs. Disclosure) ⫻ 2 (Advisor Recommendation: Portfolio A vs. Portfolio B) between-subjects design. The task was
identical to the incomplete information conditions in Experiment
1, where the die roll of six was hidden from the client for both
portfolios (i.e., “???”). The disclosure statement was the same as in
Experiments 1 and 2: “I am required to tell you that . . .” As
previously, clients reported their investment choice, trustworthiness perceptions (␣ ⫽ .85), and desire to change advisors.
Results
Desire to change advisors. There were significant main effects of both disclosure, b ⫽ 0.53, SE ⫽ 0.20, Wald ⫽ 7.05, p ⬍
.01, and portfolio recommendation, b ⫽ 1.17, SE ⫽ 0.20, Wald ⫽
34.03, p ⬍ .001, on desire to change advisors, which were qualified by an interaction, b ⫽ 0.91, SE ⫽ 0.40, Wald ⫽ 5.19, p ⫽ .02
(see Figure 6). When advisors recommended inferior Portfolio B
(consistent with their self-interest), disclosure (56%) significantly
increased the likelihood that clients wanted to change advisors
relative to nondisclosure (32%), b ⫽ 0.99, SE ⫽ 0.26, Wald ⫽
14.72, OR ⫽ 2.68, 95% CI [1.62, 4.44], p ⬍ .001. However, when
advisors recommended superior Portfolio A (sacrificing their selfinterest), disclosure neither increased nor decreased the clients’
likelihood of wishing to change advisors (20% vs. 19%), b ⫽ 0.08,
SE ⫽ 0.31, Wald ⫽ 0.06, OR ⫽ 1.08, 95% CI [0.59, 1.97], p ⫽
.81, suggesting that the disclosure penalty and the altruistic signal
may have canceled each other out.
Trustworthiness. Similarly, there were significant main effects of disclosure, F(1, 537) ⫽ 3.84, p ⫽ .05, p2 ⫽ .01, and
portfolio recommendation, F(1, 537) ⫽ 42.94, p ⬍ .001, p2 ⫽ .07,
on trustworthiness perceptions, qualified by an interaction, F(1,
537) ⫽ 12.85, p ⬍ .001, p2 ⫽ .02. Simple-effect analyses revealed
that for advisors who recommended inferior Portfolio B (consistent with their self-interest), disclosure significantly decreased
trustworthiness (M ⫽ 4.14, SD ⫽ 1.09) compared with nondisclosure, (M ⫽ 4.62, SD ⫽ 0.96), F(1, 537) ⫽ 14.98, p ⬍ .001, p2 ⫽
.03. For advisors who recommended superior Portfolio A, however, disclosure had no significant net effect on trustworthiness
(M ⫽ 5.02, SD ⫽ 1.02 vs. M ⫽ 4.88, SD ⫽ 0.92), F(1, 537) ⫽
1.36, p ⫽ .24, p2 ⫽ .003.
Moderated mediation analyses (Hayes, 2013; Model 7) revealed
that trustworthiness perceptions significantly mediated the effect
of disclosure on the desire to change advisors when the advisor
recommended Portfolio B (0.43. 95% CI [0.20, 0.73]), but, unsurprisingly given the null net effect on trustworthiness perceptions,
not when the advisor recommended Portfolio A (⫺0.13, 95% CI
[⫺0.34, 0.08]).
Discussion
Disclosure decreased trustworthiness perceptions and increased the
desire to change advisors, when advisors gave biased advice that
satisfied their self-interests (recommending Portfolio B). However,
when advisors recommended the superior Option A and sacrificed
their self-interest, disclosure neither increased nor decreased trustworthiness nor desire to change advisors compared with nondisclosure.10
Our theory made predictions of opposing effects when advisors sacrifice their self-interest from both the disclosure penalty decreasing
trust and the altruistic signal increasing trust (see Figure 1). These
effects appeared to negate each other.
Our interpretation that the trustworthiness effects canceled one
another out yields an important implication: If the disclosure
penalty can be muted, then it may be possible for COI disclosure
to increase trust (through the altruistic signal; see Figure 1). Our
theory is that the disclosure penalty is a consequence of an attribution made regarding the advisor’s character. Therefore, the
disclosure penalty could be mitigated if advisees are informed that
the presence of their advisor’s COI is due to an external systemic
cause (i.e., a cause beyond the advisor’s control). The following
two experiments were designed to disentangle the disclosure penalty from the altruistic signal and sought to document an example
in which COI disclosure actually increases trust in the advisor’s
character.
Experiment 5: Salient External Attribution for the
Presence of the COI Mutes the Disclosure Penalty
In the preceding experiments, we found evidence of a disclosure
penalty. If our theory that the mere presence of a COI produces
negative personal attributions (Gilbert & Malone, 1995; Ross &
Nisbett, 1991) regarding the advisor’s character is correct (whether
or not the advisee is cognizant of the personal attribution), then
providing a salient external (systemic) reason for the presence of
the COI should mute the disclosure penalty. If the disclosure
penalty can be mitigated through salient external attribution, then,
due to the altruistic signal, which appears when advisors recommend an option that sacrifice their self-interests, disclosure should
increase trustworthiness. In this experiment, when we inform
clients that the COI was due to an external cause, we expect the
altruistic signal to dominate and trust in the advisor to increase (see
last row in Figure 1).
Method
Participants and design. We requested approximately 100
participants per condition and received 361 new participants from
ROI Rocket (184 women, 177 men, Mage ⫽ 42.8, SD ⫽ 10.7) who
were randomized to one of three conditions: nondisclosure, disclosure, and disclosure with salient external attribution.
Procedure. All advisors recommended the superior Portfolio
A but were awarded a bonus if the client selected the inferior
10
Note that Portfolio A recommendations led to greater trust than
Portfolio B recommendations, even though disclosure did not have a
significant effect on clients who received Portfolio A recommendations
(due to the presence of both the disclosure penalty and the altruistic signal;
i.e., perceptions of advice quality can influence trust in both nondisclosure
and disclosure conditions, but the disclosure penalty is still present irrespective of advice quality when advisors disclose a self-interest).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
DISCLOSURE PENALTY
99
Figure 6. The percent of participants who reported they would want to change advisors in Experiment 4, by
condition. Disclosure increased the wish to change advisors when the inferior portfolio was recommended
(self-serving advice) but had no effect on the wish to change advices when the advice was self-sacrificing
suggesting that the disclosure penalty and altruistic signal may have cancelled each other out. Error bars are ⫾1
SE. See the online article for the color version of this figure.
Portfolio B. Participants could see five possible prizes for each
portfolio but the outcome for a die roll of “6” was shown as “???”
for both portfolios. The disclosure and nondisclosure conditions
were the same as in our prior experiment. In the disclosure with
salient external attribution condition, before viewing the portfolios, clients were informed that “In addition to their base pay, your
advisor can earn a bonus if you choose a certain portfolio. The
structure of this bonus was entirely out of their control; we assigned
the possible bonus to them. They had absolutely no choice in determining which portfolio earns them the bonus.” As in previous experiments, clients reported their investment choice, trustworthiness perceptions (␣ ⫽ .89), and desire to change advisors.
Results
Desire to change advisors. See Figure 7 for a depiction of the
results. As in the previous experiment, when advisors recommended the superior Portfolio A, there was no statistical difference
between the nondisclosure (14%) and disclosure (18%) conditions
in the desire to change advisors, b ⫽ 0.31, SE ⫽ 0.36, Wald ⫽
0.77, OR ⫽ 0.73, 95% CI [0.37, 1.47], p ⫽ .38. However, disclosure with external attribution decreased the likelihood of requesting to change advisors (7%) relative to disclosure alone, b ⫽ 1.03,
SE ⫽ 0.42, Wald ⫽ 5.95, OR ⫽ 0.36, 95% CI [0.16, 0.82], p ⫽
.01, and to nondisclosure, although this latter effect was not
Figure 7. The percent of participants who reported they would want to change advisors in Experiment 5, by
condition. Disclosure alone had no effect on the wish to change advices (with self-sacrificing advice). Muting
the disclosure penalty with a salient external attribution for possessing a conflict of interest, lead to increased
trust from the altruistic signal, that is, less desire to change advisors. Error bars are ⫾1 SE. See the online article
for the color version of this figure.
SAH AND FEILER
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
100
significant, b ⫽ 0.72, SE ⫽ 0.43, Wald ⫽ 2.74, OR ⫽ 0.49, 95%
CI [0.21, 1.14], p ⫽ .10.
Trustworthiness. Similarly, trustworthiness perceptions differed significantly across the three conditions, F(2, 358) ⫽ 7.40, p ⬍
.001, p2 ⫽ .04. As in Experiment 4, disclosure alone (M ⫽ 5.14,
SD ⫽ 1.17) did not increase or decrease trustworthiness relative to
nondisclosure, (M ⫽ 4.97, SD ⫽ 0.93), t(358) ⫽ 1.33, d ⫽ 0.16, p ⫽
.18, but disclosure with external attribution (M ⫽ 5.47, SD ⫽ 0.97)
significantly increased trustworthiness compared with both nondisclosure, t(358) ⫽ 3.80, d ⫽ 0.53, p ⬍ .001, and disclosure alone,
t(358) ⫽ 2.41, d ⫽ 0.31, p ⫽ .02. Mediation analysis (with a
multicategorical independent variable) revealed trustworthiness to be
a significant mediator (0.59, 95% CI [0.27, 1.02]) between disclosure
with external attribution and the desire to change advisors.
Discussion
Only when (a) advisors recommend an option sacrificing
their self-interest—thereby sending the altruistic signal—and
(b) advisees are informed that the source of the advisor’s COI
(self-interest) is due to an external cause (beyond the advisor’s
control)—thereby muting the disclosure penalty— does COI
disclosure increase the client’s trust in the advisor. If no salient
external attribution for the COI is present, replicating the previous experiment, the disclosure penalty which decreases trust
and the altruistic signal which increases trust appear to negate
each other leaving a null net effect on trust.
The attribution given for the presence of an advisor’s COI has
important implications. Although it appears we have strong tendencies to attribute nefarious qualities to advisors who possess COIs, this
penalty may be overcome if advisees “rationalize” or “normalize” the
existence of their advisors’ COIs to systemic issues. If all advisors
possess COIs, for example, physicians paid via a fee-for-service
model, then the disclosure penalty may be muted. Our next study
moves to the medical (fee-for-service) context to investigate if the
altruistic signal leads to increased trust in physicians who give selfsacrificing advice in the presence of a COI.
Experiment 6: The Altruistic Signal in a
Medical Context
Our final experiment moves to a medical scenario. Prior experiments have revealed that, consistent with the disclosure penalty,
physicians who possess a COI and give self-serving poor-quality
advice with simple salient disclosures (that recipients can deliberate on) are trusted less than those that do not disclose (Sah et al.,
2019). In this experiment, we investigate how advisee trust is
affected when a physician who gives self-sacrificing high-quality
advice discloses his or her fee-for-service COI. U.S. hospitals
often have a fee-for-service compensation model (payment for
each treatment or procedure performed) and therefore an external
reason for the physicians’ COIs exists. This external attribution
may mute the disclosure penalty if made salient to the patient as
patients will then interpret the conflict as being beyond the advisor’s control. If the disclosure penalty is muted, then the altruistic
signal may dominate and lead to increased trust as in the prior
experiment (as depicted in the last row in Figure 1).
Method
Participants and design. We randomly assigned 212 participants from MTurk (114 women, 98 men, Mage ⫽ 36.0, SD ⫽ 10.0;
all with at least a bachelor’s degree) to one of two conditions:
disclosure with salient external attribution and nondisclosure. This
study was preregistered on the Open Science Framework.11
Procedure. The scenario was based on the real event described
in the introduction that took place in a U.S. emergency room (This
American Life, 2009). Participants were asked to imagine that they
were the parent of a 14-year-old girl who had been involved in a
minor car accident. They read that after examination the doctor had
recommended that the patient be discharged but you, the parent, had
requested a CT scan. The doctor explained the risks (radiation exposure) and benefits (closer look) of a CT scan and still recommended
no scan. In the disclosure with salient external attribution condition,
there was one additional sentence: “I also should let you know that,
like most hospitals, we operate on a fee-per-service model, so I
actually receive a payment for each CT scan I conduct.”
Participants then decided between “CT scan to check for potential neck fracture” and “No CT scan to avoid radiation exposure.”
After making this decision, on a 7-point scale, participants indicated whether they would listen to the doctor for future decisions.
Participants also completed the trustworthiness measure, this time
with respect to the doctor (␣ ⫽ .96).12
Results
Listen to the doctor in the future. As predicted, participants
in the disclosure condition (M ⫽ 5.54, SD ⫽ 1.23) were more
likely to indicate that they would listen to this doctor for future
decisions than those in the nondisclosure condition (M ⫽ 4.97,
SD ⫽ 1.60), F(1, 210) ⫽ 8.69, p ⫽ .004, p2 ⫽ .04.13
Trustworthiness. Also, as predicted, trustworthiness perceptions of the advisor’s character was significantly higher with
disclosure (M ⫽ 5.69, SD ⫽ 1.15) than without (M ⫽ 5.29, SD ⫽
1.44), F(1, 210) ⫽ 5.17, p ⫽ .02, p2 ⫽ .02,14 and this measure
mediated the effect of disclosure on willingness to listen to the
doctor for future decisions (0.37; 95% CI [0.05, 0.70]).
Discussion
Like the prior experiment, COI disclosure (with external attribution) and with self-sacrificing advice led to significantly increased
trust in the advisor compared with nondisclosure. Even though prior
work has shown a decrease in trust with simple salient COI disclosure
in medical contexts (Hwong et al., 2017; Sah et al., 2019), the
disclosure penalty appears to be muted in this context for fee-forservice COIs which can easily be attributed to factors beyond the
advisor’s control (systemic attribution). Self-sacrificing advice (when
accompanied with COI disclosure) in this context, therefore, leads to
an increase in trust due to the altruistic signal.
11
See https://osf.io/uj8yp/. Due to a pricing increase at ROI Rocket, we
ran the study through MTurk.
12
At the end of the study, we also asked participants for any comments
they may have with regards to their impression of the doctor.
13
The difference is also significant via t-test, t(210) ⫽ 2.95, d ⫽ 0.41,
p ⫽ .004.
14
The difference is also significant via t-test, t(210) ⫽ 2.27, d ⫽ 0.31,
p ⫽ .02.
DISCLOSURE PENALTY
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
General Discussion
Conflicts of interest (COIs) are ubiquitous among advisors
across professions and disclosure is a popular remedy. However,
prior studies examining the effect of disclosure on trust have often
confounded the presence of a COI with poor-quality advice. In this
article we focus on cases when advisors give high-quality advice.
That is, when the advisor recommends the best option for the
advisee. We find that COI disclosure with high-quality advice
damages the advisor–advisee relationship and increases the desire
in advisees to switch advisors.
We pit a dispositional attribution account against an informationprocessing account and find evidence favoring the attribution theory
perspective. The attribution account emphasizes the default tendency
of advisees who are aware of their advisors’ COI to make negative
inferences about their advisors’ character rather than attribute the
existence of the COI to an external systemic cause (unless informed
otherwise). This disclosure penalty which increases the desire to
change advisors arises even when advisors give high-quality advice
(and clients have complete information to assess the (high) quality of
the advice). It also arises regardless of whether the advisor’s professional and personal interests are aligned or actually in conflict. Finally, the penalty can be muted if advisees are informed that the
presence of their advisor’s COI is due to an external cause (i.e., a
cause beyond the advisor’s control). The default tendency to assume
poor character traits in the advisor due to the presence of a self-interest
reveals the unforgiving nature of COI disclosures.
However, we also find that trust can increase through COI
disclosure under a particular circumstance: When advisors recommend an option that sacrifices their self-interest, disclosure reveals
the self-sacrifice and generates a positive altruistic signal. Even
then, trust gain from the altruistic signal competes with trust loss
from the disclosure penalty and the net effect depends on the
stronger of the two signals. If the disclosure penalty is muted, for
example with a salient external attribution for the presence of the
COI, then disclosure can lead to increased trust overall.
These results suggest that, although transparency is often desired,
COI disclosure may lead to essential damage to the advisor–advisee
relationship and an increased desire to change advisors if the disclosure penalty is strong. That is, in practice, an unintended consequence
of disclosure may be that even honest, unbiased advisors who disclose
a self-interest may find their clients looking elsewhere and perhaps
even their valuable advice being rejected, unless there is a clear
external attribution that can be made for the existence of the COI.
Theoretical Implications
The current research draws on attribution theory and contributes
to the literature on COI disclosure and trust. We aimed to contribute to this conceptual space to further inform the dialogue on COI
disclosure and extend the literature to domains in which advisors
with COIs are honest and give high-quality advice. Theoretically,
COI disclosure is said to alert consumers to potential (low) quality
advice (Healy & Palepu, 2001), and thus lead to less favorable
judgments of the advisor. Our findings (using simple, salient
disclosures in which advisees are likely to notice and deliberate
on) reflect this pattern but also extend it further than the prior
theory would predict. While prior studies often confounded the
presence of a COI with poor-quality advice, we disentangle these
variables in a series of controlled experiments which uncover that
101
even when there is no uncertainty as to the quality of advice—in
fact, even when the quality of advice can be judged objectively as
being high—advisees still report decreased trust in their advisor
with COI disclosure. These findings are better explained by an
attribution theory perspective rather than a pure informationprocessing account of disclosure (i.e., a judgment correction process). The mere presence of an advisor’s self-interest, regardless of
whether it biases the advisor or affects the advice quality, results
in advisees attributing negative characteristics to their advisors.
Only when advisees are informed that the presence of their advisor’s COI is due to an external cause (i.e., a cause beyond the
advisor’s control), and the advisor sacrifices any self-interest to
place the advisee first, does trust in the advisor increase compared
with a nondisclosure situation.
It is important to note that advisees in our studies had little
information competing with the disclosure information—in rich
environments in which advisees may be focused on other information (such as which product to buy) rather than deliberating on
the disclosure, prior research has shown an increase in perceptions
of the advisor’s expertise which leads to increased persuasion (Sah
et al., 2018). This effect—named disclosure’s expertise cue—
occurs only when COI disclosures are processed automatically
(without much conscious awareness) rather than deliberatively.
When disclosures are processed deliberatively, consistent with our
findings, Sah et al. (2018) find evidence of decreased trust in the
advisor (the disclosure penalty). In this article, in order to systematically manipulate and examine aspects of the COI landscape such
as the quality of advice and whether the advice is self-serving or
self-sacrificing, we focus on simple, salient disclosures in which
the advisees are likely to deliberate on.
In sum, this research demonstrates the robustness of the COI
disclosure penalty which is present regardless of the quality of
advice. Such distrust in reaction to awareness of a COI emerges
even with honest advisors, and even when advisors advise an
option sacrificing their personal interests, causing consumers to
potentially discount valuable advice. The “cost” of possessing
COIs is therefore borne by both advisors and consumers regardless
of whether advisors are influenced by the COI or not.
Managerial, Legal, and Policy Implications
In addition to contributing to theory, this research has implications for managers, advisors, regulators, and policymakers. Our
findings add to, and help to refine, a growing body of literature
demonstrating the complexity of using disclosure as a remedy for
COIs. Although disclosure may “work” in the sense of reducing
trust in the advisor, this benefits advisees only when the advice
given is actually biased. In many other situations, disclosure damages the advisor–advisee relationship and could lead to valuable
advice being ignored. While disclosure will and should remain an
important aspect of addressing the problem of COIs, we should not
overestimate the extent to which it is an all-encompassing solution.
Policies and laws that help advisors eliminate COIs should remain
the first priority.
Advisors often declare that because they do not succumb to bias,
they do not possess a COI (McCoy & Emanuel, 2017). Certainly
the presence of a COI does not mean the presence of biased or
poor-quality advice (Lo & Ott, 2013; Pretty v. Prudential Insurance Company of America, 2010; Rosenbaum, 2015). These fac-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
102
SAH AND FEILER
tors, however, are largely irrelevant with regards to advisees’
perceptions of their advisors’ character. A penalty exists for merely
possessing a COI regardless of whether the advisor succumbs to
the conflict or not. Given the robust disclosure penalty finding,
managers, regulators, and individual advisors may want to avoid or
eliminate COIs wherever possible, so they can signal the absence
of any conflicts. For example, managers should carefully consider
the incentive systems for professional advisors. Our findings suggest that disclosure of these incentives (which is often required)
may have adverse consequences for the advisor–advisee relationship. Importantly, it could make clients look elsewhere for advice.
Removing COIs from an organizations’ incentive or reward structures should have substantial effects on advisee (and consequently,
public) trust (Brennan et al., 2006; Criminal Code, Criminal penalties for acts involving Federal health care programs, 2011;
United States v. Goss, 2004). Greater trust should also lead to
greater positive reputational effects. This would lead other organizations and institutions to make efforts to eliminate COIs so they
can declare the absence of any conflicts (Sah & Loewenstein,
2014).
Our results also highlight that a critical determinant of whether
trust decreases with disclosure is due to the tendency to attribute
the COI to the advisors’ poor character rather than external factors.
However, if COIs are the norm in particular industries, advisees
may perceive such conflicts as systemic rather than reflective of
the advisors’ character, leading to a decrease in the disclosure
penalty. Even so, our studies suggest this external systemic cause
of the COI must be salient to the advisee at the time of the COI
disclosure. Furthermore, especially if the disclosure penalty is
muted, advisors may gain trust if they sacrifice their self-interests
to give good quality advice, as our final two experiments demonstrate.
To be clear, the authors are not advocating to eliminate disclosure mandates. The findings in this study, however, do show
situations in which COI disclosure may hurt the advisor–advisee
relationship more than is intended. Adding to a body of work on
unintended consequences of COI disclosure (Cain et al., 2005;
Grady et al., 2006; Loewenstein, Sah, & Cain, 2012; Sah et al.,
2013, 2018, 2019), and the difficulty to assess bias in advice, laws
that consider eliminating conflicts of interest may have greater
benefit than those that merely require conflicts of interest to be
disclosed.
Limitations and Future Directions
The research presented here has some limitations that should be
considered for future research. First, in each of our experiments the
disclosure provided new information to the advisee that he or she
was previously unaware of. Our conversations with practitioners
suggest that this is the more common situation in which disclosure
occurs because the intended purpose of disclosure is to reveal
something that was previously unknown. However, the effects of
disclosure may be different when it provides information that the
advisee is already aware of.
Second, in all our experiments, advisees had little information
about their advisors. In the field, advisees may have additional
information on their advisors and other signals that could indicate
high or low trustworthiness. For example, a long-standing relationship with an advisor may cause advisees to instinctively attri-
bute or rationalize the advisor’s disclosed COI to a cause beyond
his or her control. On the other hand, an advisor revealing a
self-interest within an ongoing relationship with an advisee could
be perceived as an even greater violation of trust.
The effect sizes in these studies were small to moderate. For
example, for the disclosure penalty, the odds ratio of disclosure on
the desire to change advisors with high-quality advice was approximately 3.18 and the average p2 for the same effect on advisor
trustworthiness was .06. Importantly, we propose that eliminating
COIs are likely to have a much larger effect on improving advice
quality and protecting consumers than policies such as disclosure
or mandatory second options (Sah, 2018).
The effects may also differ in situations in which advisees have
little or no opportunity to choose a different advisor. If the time
and financial cost of seeking another advisor are prohibitive,
advisees may engage in motivated reasoning to increase their trust
in their advisor, thereby muting the disclosure penalty. In this vein,
prior research has shown that the greater the monetary cost to seek
a second opinion, the more likely advisees are to report trusting
their primary advisor (Sah & Loewenstein, 2015). Moreover, some
professionals or occupations are trusted more than others. Physicians, for example, are trusted more than financial advisors (Gallup Poll, 2015). A different baseline for trust can lead to advisees’
rationalizing the presence of their advisor’s COI, again muting the
disclosure penalty (Rose et al., 2019).
A conceivable, albeit less plausible, situation that we did not
empirically examine in this article is the case in which (similar to
our Experiments 1 to 3) the best option for the client is aligned
with the advisor’s self-interest, but this time the advisor sacrifices
her self-interest to recommend an inferior option to the client. If
signaling trustworthiness supersedes all other objectives, conceivably an advisor may benefit from such an approach—recommending an option that is knowingly inferior but visibly demonstrates
self-sacrifice, thereby sending the altruistic signal. If the long-term
benefit of the trust gain is greater than the short-term benefit from
the self-interest and any benefits from giving high-quality advice,
then this approach may benefit the advisor at the advisees’ expense. However, the disclosure penalty will still be present in this
situation (the differing effects on trust would be like Situation 3
shown in Figure 1), mitigating any increase in trust and making it
unlikely that this could be an optimal strategy for the advisor. Such
strategic behavior might be consistent with Cialdini’s classic work
on “baiting,” in which, for example, a waiter points a customer to
the less expensive appetizer or wine (self-sacrificing advice), so as
to be perceived as doing a favor for the customer, but then goes on
to recommend more expensive main courses or desserts (Cialdini,
2001). Future research could investigate these possibilities.
Finally, our article focuses on advisees rather than advisors. It is
possible that advisors may behave differently if they have to
disclose a COI, either by eliminating the COI if they are able to do
so (Sah & Loewenstein, 2014), or increasing or decreasing the bias
in their advice (Sah, 2019). Reputational concerns may also encourage advisors to reduce bias in their advice and repeated
interactions may model these concerns (see Koch & Schmidt, 2010
and Experiment 4, Sah & Loewenstein, 2015). However, regardless of how advisors behave and the quality of their advice, the
findings in this article demonstrate that advisors who disclose a
COI face a disclosure penalty, even if their advice is unbiased and
their interests are not actually in conflict.
DISCLOSURE PENALTY
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Conclusion
Across a range of professions such as medicine, finance, and
law, disclosure policies seek to inform and protect consumers from
advice that may have been compromised by an advisor’s conflict
of interest. However, disclosing conflicts of interest has complex
effects on both advisors and advisees. The findings in this article
give us an improved understanding of the trust dynamics involved
when advisors disclose conflicts of interest along with high-quality
advice. Disclosure leads to an increased desire to change advisors
even when advice quality is high: an effect we call the disclosure
penalty. If advisors sacrifice their self-interests, disclosure produces a competing positive signal: an effect we call the altruistic
signal. The net effect on perceptions of the advisor’s trustworthiness depends on which effect—the disclosure penalty or the altruistic signal—is stronger. The disclosure penalty highlights a substantial hazard which is borne not only by biased advisors but also
by honest unbiased advisors and their advisees. Even when advice
is in the advisees’ best interest, conflict of interest disclosure can
damage the advisor–advisee relationship, lead advisees to change
advisors, and may drive advisees away from valuable advice. Laws
that eliminate conflicts of interest may provide greater benefit to
advisees than those that merely require conflicts of interest to be
disclosed.
References
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable
distinction in social psychological research: Conceptual, strategic, and
statistical considerations. Journal of Personality and Social Psychology,
51, 1173–1182. http://dx.doi.org/10.1037/0022-3514.51.6.1173
Boatright, J. R. (2000). Conflicts of interest in financial services. Business
and Society Review, 105, 201–219. http://dx.doi.org/10.1111/0045-3609
.00078
Brennan, T. A., Rothman, D. J., Blank, L., Blumenthal, D., Chimonas,
S. C., Cohen, J. J., . . . Smelser, N. (2006). Health industry practices that
create conflicts of interest: A policy proposal for academic medical
centers. Journal of the American Medical Association, 295, 429 – 433.
http://dx.doi.org/10.1001/jama.295.4.429
Cain, D. M., Loewenstein, G., & Moore, D. A. (2005). The dirt on coming
clean: Perverse effects of disclosing conflicts of interest. The Journal of
Legal Studies, 34, 1–25. http://dx.doi.org/10.1086/426699
Campbell, M. C., & Kirmani, A. (2000). Consumers’ use of persuasion
knowledge: The effects of accessibility and cognitive capacity on perceptions of an influence agent. The Journal of Consumer Research, 27,
69 – 83. http://dx.doi.org/10.1086/314309
Cappola, A. R., & FitzGerald, G. A. (2015). Confluence, not conflict of
interest: Name change necessary. Journal of the American Medical Association, 314, 1791–1792. http://dx.doi.org/10.1001/jama.2015.12020
Carson, T. L. (1994). Conflicts of interest. Journal of Business Ethics, 13,
387– 404. http://dx.doi.org/10.1007/BF00871766
Church, B. K., & Kuang, X. (2009). Conflicts of interest, disclosure, and
(costly) sanctions: Experimental evidence. The Journal of Legal Studies,
38, 505–532.
Cialdini. (2001). Influence: Science and practice (4th ed.). Boston, MA:
Allyn & Bacon.
Colquitt, J. A., Scott, B. A., & LePine, J. A. (2007). Trust, trustworthiness,
and trust propensity: A meta-analytic test of their unique relationships
with risk taking and job performance. Journal of Applied Psychology,
92, 909 –927. http://dx.doi.org/10.1037/0021-9010.92.4.909
Criminal Code, Criminal penalties for acts involving Federal health care
programs., 42 U.S. C. 1320a–7b § (2011).
103
Friestad, M., & Wright, P. (1994). The persuasion knowledge model: How
people cope with persuasion attempts. The Journal of Consumer Research, 21, 1–31. http://dx.doi.org/10.1086/209380
Gallup Poll. (2015). Honesty/ethics in professions. Retrieved from http://
www.gallup.com/poll/1654/honesty-ethics-professions.aspx
Gilbert, D. T., & Malone, P. S. (1995). The correspondence bias. Psychological Bulletin, 117, 21–38. http://dx.doi.org/10.1037/0033-2909.117.1.21
Grady, C., Horstmann, E., Sussman, J. S., & Hull, S. C. (2006). The limits
of disclosure: What research subjects want to know about investigator
financial interests. The Journal of Law, Medicine & Ethics, 34, 592–599.
http://dx.doi.org/10.1111/j.1748-720X.2006.00073.x
Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. New York, NY:
Guilford Press.
Healy, P. M., & Palepu, K. G. (2001). Information asymmetry, corporate
disclosure, and the capital markets: A review of the empirical disclosure
literature. Journal of Accounting and Economics, 31, 405– 440. http://
dx.doi.org/10.1016/S0165-4101(01)00018-0
Hwong, A. R., Sah, S., & Lehmann, L. S. (2017). The effects of public
disclosure of industry payments to physicians on patient trust: A randomized experiment. Journal of General Internal Medicine, 32, 1186 –
1192. http://dx.doi.org/10.1007/s11606-017-4122-y
Johns, M. Z. (2007). Informed consent: Requiring doctors to disclose
off-label prescriptions and conflicts of interest. The Hastings Law Journal, 58, 967–1024.
Kee, H. W., & Knox, R. E. (1970). Conceptual and methodological considerations in the study of trust and suspicion. The Journal of Conflict Resolution, 14, 357–366. http://d...
Purchase answer to see full
attachment