Experiments-Offering Prosocial Incentives On-Top Do They Sweet

Received: 14 March 2023 | Accepted: 9 November 2023
DOI: 10.1002/mar.21941
RESEARCH ARTICLE
Offering prosocial incentives on‐top: Do they sweeten

the deal or poison the well?
Sven Beisecker | Christian Schlereth
WHU—Otto Beisheim School of Management,

Marketing and Sales Group, Chair of Digital Abstract
Marketing, Vallendar, Germany
Companies often rely on customer feedback to build and improve their business.
Correspondence Customers, in turn, are expected to (i) fill in customer feedback surveys
Sven Beisecker, WHU—Otto Beisheim School (participation) and (ii) provide accurate responses (performance). To encourage
of Management, Marketing and Sales Group,
Chair of Digital Marketing, Burgplatz 2, 56179 active participation and ensure accurate responses, companies traditionally offer
Vallendar, Germany. either self‐benefiting incentives, like lottery prizes, or prosocial incentives, like
Email: [email protected]
charity donations. More recently, some companies have started offering prosocial
incentives on top of self‐benefiting incentives in the hope to “sweeten the deal,” that
is, to improve participation and performance even further. With this research, we
challenge whether the on‐top prosocial incentives are effective. The evidence from
two field experiments and one incentive‐aligned online experiment does not confirm
any such advantage. In contrast, performance can decrease when a low‐amount on‐
top prosocial incentive is offered relative to a pure self‐benefiting setting. This trend
is only reversed once the on‐top incentive amount increases. Furthermore, for
participation, we find that on‐top prosocial incentives are ineffective and, at higher
amounts, even detrimental. Therefore, our empirical insights rather suggest that on‐
top prosocial incentives “poison the well.”
KEYWORDS
affective value, lottery, performance, prosocial incentives
1 | INTRODUCTION Given these advantages, the question arises how best to

incentivize customers to participate in feedback surveys (participa-
Companies increasingly rely on customer feedback to build and improve tion) and provide accurate responses (performance). In this regard,
their business. Feedback allows companies to understand customer extant literature (e.g., Biner & Barton, 1990; Furse & Stewart, 1982;
satisfaction (Griffin & Hauser, 1993), ensure that complaints are directly Hennig‐Thurau et al., 2004; Siuki & Webster, 2021) differentiates
voiced with the company and not through channels outside of its control two types of commonly used incentives: self‐benefiting incentives,
(Richins, 1983), and identify customer pain points to develop innovations which provide a direct benefit to the customer, and prosocial
(Brown, 2008). There is even literature which documents that by voicing incentives, which promise a benefit for others, for example, in the
feedback, a customer's purchase frequency and spend with the company form of a donation. Research outlines that each incentive type affects
increases, also known as the mere measurement effect (Borle et al., 2007; performance and participation in a different way. Prosocial incentives
Morwitz et al., 1993). can raise performance even at a low amount (Khan et al., 2020), but
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium,
provided the original work is properly cited.
© 2023 The Authors. Psychology & Marketing published by Wiley Periodicals LLC.
Psychol Mark. 2023;1–21. wileyonlinelibrary.com/journal/mar | 1

2 | BEISECKER and SCHLERETH
are ineffective at raising participation (e.g., Schwartz et al., 2021). Overall, our results suggest that practitioners should be cautious
Self‐benefiting incentives can decrease performance when the about offering on‐top prosocial incentives. Against the idea of
incentive amount is low (e.g., Gneezy & Rustichini, 2000), but are “sweetening the deal,” we observe that performance either stays
generally suited to raise participation (e.g., Furse & Stewart, 1982). constant or even decreases when a low‐amount on‐top prosocial
However, extant literature lacks an understanding of what to incentive is offered relative to the baseline of no on‐top incentive.
expect when combining different incentive types, that is, when a This trend reverses only once an on‐top prosocial incentive of a
prosocial incentive is offered on top of a self‐benefitting incentive. In higher amount is offered. Hence, the advantages commonly
principle, if utility theory applies (Jensen, 1967), then offering more associated with prosocial incentives—a substantial increase in
incentives could “sweeten the deal,” that is, make it more attractive performance in return for a relatively low incentive amount—do not
to participate and improve performance. Ideally, both incentive types hold in this setting. Besides, the on‐top prosocial incentive is found to
complement each other, such that adding the prosocial incentive be ineffective and, at higher amounts, even detrimental to participa-
could attenuate the drawbacks of the self‐benefitting one. However, tion. Therefore, the additional incentive rather seems to “poison the
the existing literature does not provide guidance regarding the well.” Our paper discusses potential explanations underlying these
effectiveness of this approach. findings.
Companies seem to anticipate such advantages and increasingly
offer an on‐top prosocial incentive in addition to a baseline self‐
benefiting incentive. For example, Procter & Gamble incentivizes 2 | LITE RATURE REVIEW
customers to participate in surveys by promising a self‐benefiting
incentive (e.g., entry into a lottery to win a cash prize) and additionally We first disentangle the different benefits of customer feedback and
a prosocial incentive (e.g., a donation to the P&G Children's Safe thereby outline in which ways participation and performance play a
Drinking Water Fund) under the tagline “Earn rewards. Make an major role. Next, we summarize how self‐benefiting and prosocial
impact.” (Procter & Gamble, 2023). Similarly, mobile survey apps like incentives affect performance when offered in isolation, and
Jagger incentivize respondents to complete surveys in return for self‐ subsequently we do the same for participation. Finally, we describe
benefiting incentives (e.g., cash rewards and prize lotteries) and what is known about the effects of mixing both incentive types.
additionally on‐top prosocial incentives (e.g., planting a tree for every
completed survey) under the tagline “Earn rewards & plant trees”
(Jagger, 2023). 2.1 | Performance and participation when
The present paper tests whether companies can indeed benefit incentivizing customer feedback
from the on‐top prosocial incentives in two field experiments and
one incentive‐aligned online experiment. We investigate how the Customer feedback surveys serve as a crucial marketing tool,
promise of an on‐top donation in addition to a self‐benefiting lottery enabling companies to achieve diverse objectives. The efficacy of
incentive affects customers' performance, that is, the provision of these surveys hinges upon ensuring either robust customer partici-
correct survey responses, and participation, that is, the provision of pation, exemplary performance, or a blend of both, contingent upon
complete survey responses. We then validate our findings in a the specific purpose at hand. Subsequently, we will delineate four
controlled, broader‐designed incentive‐aligned online experiment, in distinct purposes and describe for each whether high performance,
which respondents can win a cash reward in return for solving an high participation, or both is of primary concern.
anagram task (c.f., Goldsmith & Dhar, 2013) within a voluntary post‐ First, feedback surveys allow companies to listen to the “voice of
survey. the customer” and thereby better understand customer satisfaction
While a few extant studies offer insights into the interplay of (Griffin & Hauser, 1993). This understanding can result in benefits like
self‐benefiting and prosocial reasons when it comes to people's moral higher customer retention (Markey et al., 2009), higher gross margin
perceptions and donation behaviors (e.g., Feiler et al., 2012; Newman and sales growth (Agag et al., 2023) and higher average customer
& Cain, 2014), they (i) do not study how they affect customers' spend (Bone et al., 2017). For this purpose, companies benefit if
responses to company surveys in terms of performance and customers have a high participation rate (independent of their
participation and (ii) leave out the incentive amount as a potential satisfaction) and demonstrate a high performance, that is, truthfully
moderator. With the present research, we aim to address these share their level of satisfaction and insights on the reasons.
shortcomings. We observe how customers' performance and Second, companies can use feedback surveys to ensure that
participation differ when a high, a low, or no on‐top prosocial customer complaints are directly voiced with them and not through
incentive is offered, and compare the observed patterns with those channels outside of their control. For example, Richins (1983) reports
expected to hold when a prosocial incentive is offered in isolation, as that dissatisfied customers tend to share their experiences through
described in extant research. This enables us to see whether the self‐ negative word of mouth among family and friends if companies do
benefiting baseline incentive alters the effect of prosocial incentives not encourage raising complaints directly with them. Here, participa-
on performance and participation, and if so, for which donation tion in feedback surveys is of primary importance to provide the
amounts. company with an opportunity to react. If this reaction is fast and the
BEISECKER and SCHLERETH | 3
issue addressed, there is a high chance of transforming a dissatisfied settings, including both survey‐ (Conn et al., 2019; Gritz, 2004;
customer into a satisfied one (Hart et al., 1990; Lovelock & Heerwegh, 2006; Porter & Whitcomb, 2003) and non‐survey tasks,
Wirtz, 2010). Performance is also important, but more so at the such as tasks requiring altruistic behavior (Gneezy & Rustichini, 2000;
later stage of resolving the complaint. Gneezy et al., 2011), work‐related tasks (Charness et al., 2016), and
Thirdly, companies can encourage customers to explain different game‐based tasks (Heyman & Ariely, 2004; Imas, 2014; Khan
aspects of their shopping experience, which provides them with et al., 2020).
useful input to improve this experience (Brown & Katz, 2011; In general, performance mirrors the perceived value respondents
Brown, 2008). In particular processes like design thinking require a derive from the incentive. This value consists of an affective and a
profound understanding of customer touch and pain points with the cognitive component (e.g., Chang & Tuan Pham, 2013; Khan
firm to develop solution‐focused strategies (Celuch & Walz, 2020; et al., 2020; Loewenstein et al., 2001). While affective value
Challagalla et al., 2009). Here, performance is most important describes “the emotional benefit associated with the incentive,”
because companies can only truly understand customer pain points cognitive value describes the “calculative assessment of the tangible
if they are adequately reported. benefits to the self” (Khan et al., 2020; p. 45). For self‐benefiting
Finally, extant research shows that by voicing feedback, a incentives, both value components react in response to the incentive
customer's purchase frequency and spend with the company amount. When the amount is low, respondents derive a negative
increases, also known as the mere measurement effect (Borle affective value because they feel unfairly compensated; at the same
et al., 2007; Morwitz et al., 1993). Dholakia and Morwitz (2002) time, they derive no notable cognitive value because they do not gain
show that this effect persists even 1 year after answering a feedback a lot for themselves (Heyman & Ariely, 2004; Khan et al., 2020). As a
survey. Importantly, a customer's mere participation is sufficient for result, their performance decreases. As the incentive amount
this effect to occur. increases, respondents no longer derive a negative affective value
In summary, companies need to entice customers to take part in because they feel the compensation is adequate. At the same time,
their feedback surveys (participation), provide high‐quality answers they derive a higher cognitive value because they gain more for
(performance), or do both. In the subsequent chapters, we use a themselves (Khan et al., 2020). As a result, their performance
review of extant research to disentangle how different incentive increases again.
types affect performance and participation.
2.2.2 | Effect of prosocial incentives on

2.2 | Effect of self‐benefiting and prosocial performance
incentives on performance
When it comes to prosocial incentives, respondents' performance
Hereafter, we discuss the effects of self‐benefiting incentives and, is scope‐insensitive, that is, it changes only in response to
subsequently, of prosocial incentives on performance. Table 1 offers whether the incentive is offered or not but not in response to its
a synopsis of the reviewed literature and encapsulates prevailing amount (Hsee & Rottenstreich, 2004; Khan et al., 2020; Small
findings. Most studies are in the area of surveys and incentivizing et al., 2007): Offering a low prosocial incentive amount tends to
customer feedback (e.g., Cohen et al., 2019; Furse & Stewart, 1982; substantially increase performance relative to no incentive, but
Göritz & Neumann, 2016). Besides, other settings relate to tasks increasing the amount does not result in further heightened
requiring altruistic behavior (e.g., Gneezy & Rustichini, 2000; Yin, Li, performance. This has been found consistently across different
and Singh, 2020) as well as work‐related (e.g., Charness et al., 2016; settings, including work‐related (Charness et al., 2016; Tonin &
Chen et al., 2019) and game‐based tasks (e.g., Imas & Vlassopoulos, 2015) and game‐based tasks (Imas &
Loewenstein, 2018; Khan et al., 2020). Predominantly, these articles Loewenstein, 2018; Imas, 2014; Khan et al., 2020).
explore self‐benefiting and prosocial incentives separately. Again, respondents' performance mirrors the perceived value
they derive from the incentive. Importantly, for prosocial incentives,
only affective value reacts to the incentive amount: It substantially
2.2.1 | Effect of self‐benefiting incentives on increases when a prosocial incentive is introduced, but it does not
performance change when the amount increases (Khan et al., 2020). This effect
can be traced back to literature on donation behavior, which suggests
When it comes to self‐benefiting incentives, respondents' perform- that people derive a positive feeling, also known as “warm glow,”
ance is scope‐sensitive, that is, it changes not only in response to from the mere act of donating, independent of the amount they
whether a self‐benefiting incentive is offered or not but also in donate and what it can achieve for the beneficiaries (e.g.,
response to its amount (Khan et al., 2020): Offering a low amount Andreoni, 1989, 1990, 1993; Barasch et al., 2014; Bezençon
tends to decrease performance relative to no incentive, but et al., 2020; Imas, 2014). Meanwhile, cognitive value does not react
increasing the incentive amount can again heighten performance. to prosocial incentives because they do not offer a personal gain for
This observation has been found consistently across different respondents (Khan et al., 2020).
4
TABLE 1 Literature overview.

|
Self‐
Perfor‐ Partici‐ benefiting Prosocial Mix of Lab or online Field
Setting Task Study mance pation incentives incentives incentives experiment experiment Main finding
Surveys Survey among Cohen ✔ ✔ ✔ No significant effect of prosocial

researchers et al. (2019) incentives on participation
Survey among Conn et al. (2019) ✔ ✔ ✔ ✔ ✔ Higher participation rate for prosocial
professionals joining compared to self‐benefiting
an NPO incentives; no effect on performance
(correctness of attention checks)
Market research panel Göritz and ✔ ✔ ✔ Significantly negative effect of prosocial

survey Neumann incentive on participation
(2016)
General social survey Blohm and ✔ ✔ ✔ Significantly positive effect of self‐

Koch (2013) benefiting incentive on participation
Market research panel Göritz and ✔ ✔ ✔ Significantly positive effect of self‐

survey Luthe (2013) benefiting lottery incentives on
participation
Mail survey on religion Gendall and ✔ ✔ ✔ No significant effect of prosocial

Healey (2010) incentives on participation
Market research panel Brüggen and ✔ ✔ ✔ Respondents’ personality traits can

surveys Dholakia explain differences in performance
(2010) and participation
Survey among Göritz and ✔ ✔ ✔ Significantly positive effect of self‐

university students Wolff (2007) benefiting lottery incentive on
participation
Survey among students Heerwegh (2006) ✔ ✔ ✔ Significantly positive effect of self‐

benefiting lottery incentive on
participation; significantly negative
effect of incentive on performance
(item nonresponse), yet
unsubstantial
Market research panel Gritz (2004) ✔ ✔ ✔ ✔ Response rate higher for bonus points
survey incentive than money lotteries;
performance (number of omitted
items) unaffected by incentive type
Survey among members Bosnjak and ✔ ✔ ✔ Significantly positive effect of self‐

of a professional Tuten (2003) benefiting lottery incentive on
association participation
BEISECKER and SCHLERETH
TABLE 1 (Continued)
Self‐
Survey among college Porter and ✔ ✔ ✔ ✔ No significant effect of self‐benefiting

applicants Whitcomb lottery incentive on participation;
(2003) slightly negative effect of self‐
benefiting incentive on performance

(item nonresponse)
Mail survey on Warriner ✔ ✔ ✔ ✔ Participation increases with self‐

environmental et al. (1996) benefiting incentive amount; no
issues significant effect of prosocial
incentives on participation
Mail survey among Furse and ✔ ✔ ✔ ✔ Participation increases with self‐

microwave owners Stewart benefiting incentive amount; no
(1982) significant effect of prosocial
Altruistic Alumni donations to Feiler et al. (2012) ✔ ✔ ✔ ✔ ✔ Emphasizing both egoistic and altruistic
behavior university; reasons reduces participation
donations to NPO
for children
Donations to charity Yin, Li, and ✔ ✔ ✔ ✔ Monetary incentives increase

Singh (2020) participation in donor acquisition
campaigns, but decrease
performance (average donations)
Blood donations Lacetera and ✔ ✔ ✔ Significantly negative effect of self‐

Macis (2010) benefiting incentive on participation
in blood donations
Donation collection by Gneezy and ✔ ✔ ✔ Small self‐benefiting incentives have a

going from house to Rustichini significantly negative effect on
house (2000) performance, but increase
performance at higher amounts
Rating behavior Newman and ✔ ✔ ✔ ✔ Tainted‐altruism effect: a company's

ethicality; evaluating Cain (2014) charitable efforts are evaluated
an altruistic cause worse when they also bear some
personal benefits
Blood donations, Gneezy ✔ ✔ ✔ Self‐benefiting incentives of a small

volunteer work et al. (2011) amount decrease performance
relative to no incentive; increasing
(Continues)
|
5
6
TABLE 1 (Continued)
|
Self‐
the incentive amount increases

performance
Work Data entry Charness ✔ ✔ ✔ ✔ Self‐benefiting incentives yield higher

et al. (2016) performance than prosocial
incentives for high amounts; the
opposite holds for low amounts
Entry of bibliographic Tonin and ✔ ✔ ✔ ✔ Prosocial incentives increase

records Vlassopoulos performance by 13%
(2015)
Publishing articles on an Chen et al. (2019) ✔ ✔ ✔ ✔ Self‐benefiting incentives increase

online platform average number of articles per
contributor, but not article quality
Customer referrals Gershon ✔ ✔ ✔ ✔ Prosocial incentives increase sign‐up

et al. (2020) rates more than self‐benefiting
incentives
Recycling campaign; Schwartz ✔ ✔ ✔ No significant effect of low prosocial

providing et al. (2021) incentives on participation;
image URLs significantly negative effect of high
prosocial incentives on participation
Games Wall‐sit task, Khan et al. (2020) ✔ ✔ ✔ ✔ Self‐benefiting incentives increase

anagram task performance when the amount is
high; prosocial incentives yield the
same performance independent of
amount; low‐amount self‐benefiting
incentives become more motivating
when affective value is enhanced,
and low‐amount prosocial incentives
become less motivating when
affective value is attenuated
Hand dynamometer Imas and ✔ ✔ ✔ Performance does not react to the size
squeezing task Loewenstein of prosocial incentive in conditions
(2018) of low tangibility
Hand dynamometer Imas (2014) ✔ ✔ ✔ ✔ Performance increases along with the

squeezing task amount of self‐benefiting incentives;
performance stays relatively
constant as the amount of prosocial
Comparing the effects of prosocial and self‐benefiting incentives,
incentives at low amounts and for

the former is usually more effective at raising performance than the
self‐benefiting incentives at high
Self‐benefiting incentives of a small
increasing the incentive amount

amount decrease performance;
incentives increase; individuals
latter when the incentive budget is limited (e.g., Charness et al., 2016).
choose to work for prosocial

However, Khan et al. (2020) also show that this advantage of
increases performance
prosocial incentives can disappear when the incentive is perceived to
have low affect. Their results shed light on the boundaries within
which low‐amount prosocial incentives can increase performance,
thereby offering implications for anyone wishing to maximize
Main finding
amounts
performance on a limited budget.
2.3 | Effect of self‐benefiting and prosocial

experiment
Field
While many studies focus on performance conditional on participa-

tion (Brüggen & Dholakia, 2010; Schwartz et al., 2021), since they are
Lab or online
conducted in a lab environment and cannot record subjects' opt‐in

experiment
decisions, a few studies also examine how incentives affect

participation, for example, in contexts where researchers send survey
✔
invites by mail. Hereafter, we discuss the effects of self‐benefiting

incentives and, subsequently, the effects of prosocial incentives on
incentives
participation.
Mix of
2.3.1 | Effect of self‐benefiting incentives on

incentives
Prosocial
participation
Regarding self‐benefiting incentives, participation tends to increase

along with the incentive amount. This has been found consistently
benefiting
incentives
across different settings including survey‐ (e.g., Blohm & Koch, 2013;
Self‐
Bosnjak & Tuten, 2003; Furse & Stewart, 1982; Göritz & Luthe, 2013;
✔
Göritz & Wolff, 2007; Heerwegh, 2006; Warriner et al., 1996) and
Partici‐
non‐survey tasks, such as tasks requiring altruistic behavior (Yin, Li,

pation
and Singh, 2020) and work‐related tasks, for example, publishing

articles on an online platform (Chen et al., 2019). However, in one
Perfor‐
study about blood donations, Lacetera and Macis (2010) demonstrate

mance
that also the opposite can happen, that is, that providing a cash
✔
incentive can significantly decrease participation. The authors argue

Ariely (2004)
that this occurred because the self‐benefiting incentive reduces the

Heyman and
altruistic behavior's reputational benefits (Ariely et al., 2009; Labroo

& Goldsmith, 2021; White et al., 2020).
Study
2.3.2 | Effect of prosocial incentives on participation

solving arithmetic
computer screen;
Dragging balls on a
Regarding prosocial incentives, offering a low amount usually does

(Continued)
puzzles
not significantly affect participation likelihood, and increasing the

incentive amount can even significantly decrease participation
Task
likelihood. This has been found consistently across settings including

survey‐ (e.g., Cohen et al., 2019; Furse & Stewart, 1982; Gendall &
TABLE 1
Healey, 2010; Göritz & Neumann, 2016; Warriner et al., 1996) and
Setting
non‐survey tasks, in particular, work‐related tasks (Schwartz

et al., 2021).
People perceive a tradeoff between helping others and investing While these papers offer insights into the interplay of self‐
the same time in an activity that benefits them, which can make them benefiting and prosocial reasons when it comes to people's moral
opt out of the task (Berman & Small, 2012; Schwartz et al., 2021). The perceptions and altruistic behavior, they (i) do not study how the
phenomenon of moral wiggle room further argues that people tend combination of self‐benefiting and prosocial incentives affects
to avoid prosocial opportunities because they create feelings of social customers' responses to company surveys in terms of performance
pressure or guilt and can force people to behave more prosocial than and participation and (ii) leave out the incentive amount as a potential
they would like to of their own accord (e.g., Dana et al., 2007; Gneezy moderator. With the present research, we aim to address these
et al., 2014). shortcomings by studying how the offer of prosocial incentives on
An exception from the described patterns are the findings by top of self‐benefiting incentives affects customers' performance and
Gershon et al. (2020), who show in the context of customer referrals participation.
that prosocial incentives (i.e., those which benefit the referred party) Overall, extant literature yields consistently similar results across
are more effective at raising sign‐up rates than self‐benefiting varied settings, as illustrated in Figure 1. The figure demonstrates
incentives (i.e., those which benefit the referring party; Gershon that neither incentive type in isolation is suited to raise performance
et al., 2020). However, this setting is different from the typical and simultaneously participation with a low incentive budget.
donation incentive.
3 | HYPOTHESES
2.4 | Effects of mixing prosocial and self‐benefiting
incentives We subsequently develop hypotheses on how on‐top prosocial
incentives affect performance and participation. We first derive our
Comparatively, little is known about mixing prosocial and self‐ hypotheses related to performance (H1a and H1b) and then our
benefiting incentives. Among the notable exceptions, the study by hypotheses related to participation (H2a and H2b).
Newman and Cain (2014) finds that people evaluate a company's
charitable efforts as worse if the company also derives a personal
benefit (tainted‐altruism effect) due to the belief that the company 3.1 | Hypotheses related to performance
could have behaved more altruistically in the absence of self‐interest.
Similarly, Feiler et al. (2012) find that emphasizing egoistic and We expect respondents to derive a different affective value from an
altruistic reasons reduces people's donation likelihood because they on‐top prosocial incentive, offered in addition to a self‐benefiting
perceive it as a persuasion attempt and show reactive behavior. incentive, than from a prosocial incentive in isolation. Respondents
FIGURE 1 Incentives' effects on performance and participation.

may perceive the fact that they stand to gain something through the H1b: A further increase in the on‐top prosocial incentive
self‐benefiting incentive to be somewhat incongruent with acting increases performance.
altruistically and working for the prosocial incentive. We foresee this
perception to be exceptionally high if the prosocial incentive is
relatively small in amount. It gives respondents the impression of 3.2 | Hypotheses related to participation
working “mostly” for themselves and “just a little” for others. Olivola
and Shafir (2013) provide evidence that people tend to behave more Concerning participation, we argue in analogy to Schwartz et al.
prosocially when the prosocial act requires them to employ (2021) that a low‐amount prosocial incentive will yield participation
substantial effort (or experience pain) in what they call the rates comparable to those in the baseline condition of no incentive.
martyrdom effect. Being promised a personal gain through the self‐ Furthermore, we expect subjects to avoid participation if they are
benefiting incentive might prevent such an effect from unfolding: promised a relatively high prosocial incentive. The reason is that, as
People may perceive their behavior as less of a sacrifice to the the prosocial incentive amount increases, it may start creating an
benefit of others because they directly receive compensation. unwanted perceived urge for subjects to behave more prosocially
Following this argument, the combined incentive will likely than they would like on their own accord (e.g., Dana et al., 2007;
evoke a different set of norms than an isolated prosocial Gneezy et al., 2014). As a result, subjects face a tradeoff between
incentive would. Rather than evoking a norm of social responsi- helping others and investing the same time in an activity that benefits
bility and a tendency to help, the incentive may create an them, which can yield negative emotions and make them opt out of
expectation of reciprocity, that is, of receiving adequate com- the task (Berman & Small, 2012). Apart from that, the combination of
pensation for the effort employed (Groves et al., 1992). In this a prosocial and a self‐benefiting incentive could be perceived by
way, a low on‐top prosocial incentive may be perceived as a weak subjects as a persuasion attempt and make them feel like their
source of affective value and may not increase performance like a behavior is unduly controlled by others (Feiler et al., 2012; Williams
low‐amount isolated prosocial incentive would. The incentivizing et al., 2004). While we believe that offering a low on‐top
party can no longer reap the usual advantages of prosocial prosocial incentive might not yet be perceived as a persuasion
incentives, that is, high performance independent of the incentive attempt, prospective participants may start feeling more controlled as
amount, as described in warm glow theory (c.f., Andreoni, 1989; the incentive amount rises. Therefore, we anticipate that a
Imas, 2014). On the contrary, if an on‐top prosocial incentive in high on‐top prosocial incentive can ultimately be a reason for
this setting is perceived in much the same way as a self‐benefiting prospective participants to simply ignore the task. We hypothesize
incentive, performance may even react negatively to a small the following:
incentive amount because it is viewed as inadequate compensa-
tion (c.f., Gneezy et al., 2011). We summarize this expectation in H2a: Adding a low on‐top prosocial incentive in addition to
the following hypothesis: a self‐benefiting incentive yields the same participation as a
pure self‐benefiting setting.
H1a: Adding a low on‐top prosocial incentive in addition to
a self‐benefiting incentive decreases performance relative to H2b: A further increase in the on‐top prosocial incentive
a pure self‐benefiting setting. decreases participation.
As the amount of the on‐top prosocial incentive increases,

people will start to feel like they are working relatively more for 4 | EMPIRICAL S TUDIE S
others and relatively less for themselves. As a result, they
anticipate attributing a larger share of their effort to the prosocial We conducted three empirical studies: Study 1 and 2 are field
incentive, and the prospect of now making a relatively higher experiments in a Marketing setting (i.e., obtaining customer insights)
sacrifice for the benefit of others could motivate them (Olivola & and Study 3 is an incentive‐aligned online experiment in a broader
Shafir, 2013). As the relative importance of the self‐benefiting non‐Marketing setting. With Study 1, we investigate the effect of a
incentive declines, respondents might view the combined incen- low on‐top prosocial incentive relative to no on‐top prosocial
tive as predominantly prosocial. As a consequence, they will be incentive on performance and participation, allowing us to test H1a
less focused on reciprocity and on maintaining an adequate and H2a. Study 2 extends the field experiment by adding a high on‐
compensation, but instead feel a higher responsibility to help top prosocial incentive, thereby allowing us to test also H1b and H2b.
others (Groves et al., 1992). This, in turn, may enable them to Study 3 is about identifying anagrams. With this experiment, we aim
derive a high affective value from the on‐top prosocial incentive, to investigate the role of affective value as a mediator. We also aim
consequently leading to an increase in performance (e.g., to replicate our field experiments' findings in a controlled setting that
Imas, 2014). This implies that an increase in the incentive amount purposely deviates from the previous studies to obtain indications of
from low to high should raise performance again. We summarize generalizability. Figure 2 provides a summary overview of our
this expectation in the following hypothesis: empirical studies.
4.1 | Study 1 (field experiment) visit or a craft beer set to use at home to instant prizes like a small
selection of sample beers. At the end of the survey, customers
4.1.1 | Aim instantly obtain the information whether they “found” a winning
glass. If so, the restaurant hands out the prize, puts the glass away for
Study 1 is a field experiment conducted in a restaurant. This setting the rest of the week, resets all glasses at the end of a week, and then
offers the advantage of high external validity and enables us to study picks new winning glasses for the coming week.
participation and performance simultaneously. The study focuses on The survey consisted of few short questions and asked
the effect of a low on‐top prosocial incentive on performance and customers for their year of birth, their gender, whether they have
participation, allowing the test of H1a and H2a. used NFC before, when their scan took place (when the glass was full,
more than half full, less than half full, or empty), and what their
primary motivation for scanning was (scanning out of curiosity, to win
4.1.2 | Setting in the lottery, to demonstrate knowledge in front of friends, due to
information from other guests, due to observing other guests, or due
The restaurant serves every ordered drink in a specially designed, to hearing about it before the visit). The survey also asked customers
NFC‐enabled drinking glass (Smartglass), and customers can interact two identification questions, the first four letters of their street name
with it during their visit. Through a unique printing technology, the and their day of birth, which serve the restaurant as an anonymized,
Smartglass is equipped with an NFC chip placed behind a printed unique customer identifier. If a customer orders another drink, they
label (see Figure 3). Customers can “scan” the Smartglass by placing can participate in the survey again to have another chance to find one
their NFC‐enabled smartphone near a label on the glass, which reads of the winning glasses, and the restaurant can use the identifier to
“Connect with NFC here.” The scan triggers an innovative object‐ track such repeat usage. Figure 4 visualizes the experimental setup in
related communication technology, which connects everyday objects a flow diagram.
(here, the glass) to unique digital artifacts (here, the online
questionnaire, which uses the ID of the glass to test whether the
respondent found one of the winning glasses). Customers find a flyer
on the table with instructions on how to scan the glass. They receive
the information that they can win something, however, unless they
engage in a scan, they obtain no information on the incentive or what
happens after scanning. Each scan is automatically captured in the
Cloud, alongside information like the ID of the scanned glass, scan
time, glass size, and type of phone used.
Once customers scan a glass, they learn that they can fill in a
short mobile survey, through which the restaurant asks them to share
some information that it could otherwise not obtain. The manipulated
landing page informs customers about the specific incentive on that
day, which includes lottery prizes ranging from a voucher for the next FIGURE 3 Illustration of Smartglass used in field experiment.
FIGURE 2 Overview of empirical studies.

FIGURE 4 Flow diagram of experimental

setup.
Study 1 spans a time horizon of 4 weeks in 2020, at times with participation is measured by the percentage of customers who
stable COVID regulations in place. The Smartglass was a relatively complied and filled out the survey after learning about the incentive.
new addition to the restaurant at that time. The restaurant was Appendix B examines if the experimental setup influenced scanning
consistently open 5 days per week. frequency, for example, through word of mouth among tablemates.
The analysis results do not demonstrate a statistically significant
confounding effect.
4.1.3 | Experimental manipulation
We manipulated the incentives in the following way: We always 4.1.5 | Sample and descriptives
offered the self‐benefiting incentives. Each day, we manipulated
whether a low on‐top donation was offered (on‐top donation of A total of 6780 beers were served in a Smartglass, of which 24.96%
€0.10 to a local nonprofit organization vs. self‐benefiting incentive were scanned (1,692 scans). Of the 1475 completed survey
alone). Appendix A provides further details. responses (87.17% of all scans), 4.88% contained incorrect data (72
surveys), that is, an objectively wrong street name, day of birth,
or both.
4.1.4 | Outcome measures The majority of survey responses came from male customers
(68.41%) and customers born between 1980 and 1999 (60.07%). As
We examine the effect of our experimental manipulations on in Study 1, most scans were conducted out of curiosity (38.78) or to
respondents' performance and participation. We measure perform- win a prize (32.41%). Customers typically scanned relatively early
ance by testing whether respondents provided answers to our two after ordering a beer, with 66.92% of scans occurring on glasses that
identification questions which are incorrect according to objective were more than half full. 70.31% of all scans were carried out by
standards. A response is deemed incorrect if the provided street respondents who had not used NFC before, indicating Smartglass'
name initials do not match those of any of the street names in a ability to trigger first‐time NFC usage.
comprehensive street list of Germany or if the customer indicated a
nonsensical number as their day of birth. While we acknowledge that
this measure likely underestimates the number of incorrect entries 4.1.6 | Effect of on‐top prosocial incentives on
(and thereby overestimates performance) since an answer can be performance
incorrect in other ways, we view our measure as a lower‐bound
estimate of incorrect responses. We regress respondents' performance onto our experimental
We define participation as the percentage of Smartglass scans manipulation, that is, whether a low on‐top donation was offered
leading to full survey responses. Since we informed customers about or not, while including several control variables related to the setting.
the incentive on the landing page after scanning the glass, the scan Since we conducted a field experiment in a restaurant, factors other
count, not customer count, sets the baseline for participation, that is, than our experimental manipulation varied over time and across
scans. In particular, we control for the respective lottery prizes participation onto our experimental manipulation while controlling
offered over time, guests' prior NFC experience, the type of glass for the afore‐introduced setting‐related variables. Temperature again
scanned (glasses existed in two sizes), the timing of the scan (hour of had to be excluded from the regression due to multicollinearity
the day and whether it happened on the weekend or a weekday), and issues. The results are presented in the right‐hand side of Table 2.
weather conditions (temperature, rain; Muñoz Sabater, 2019). In We do not observe a significant effect of offering a low on‐top
Study 1, the temperature was excluded due to multicollinearity donation on participation relative to offering no on‐top donation
issues. In Table 2, we present a summary of the results. (p > 0.1). This result supports H2a. We conclude that the low‐amount
Offering a low on‐top donation relates negatively to customers' on‐top donation is ineffective at increasing participation.
performance relative to offering no on‐top donation (p < 0.05), which
supports H1a. We also examined the regression results when
including respondent‐related control variables (e.g., gender, year of 4.2 | Study 2 (field experiment)
birth). We found the effect to be robust in direction and level of
significance. 4.2.1 | Aim
Study 2 tests the robustness of our findings from Study 1 and

4.1.7 | Effect of on‐top prosocial incentives on extends the results by also considering the effects of a high on‐top
participation prosocial incentive. Thereby, Study 2 enables us to test H1b and H2b
together with H1a and H2a.
Next, we investigate whether the on‐top donation incentive affects
participation, measured as the percentage of Smartglass scans which
resulted in a complete survey response. To this end, we regress 4.2.2 | Setting
Study 2 is a field experiment in the same restaurant setting as Study 1

T A B L E 2 Logistic regression of performance and participation and spans a time horizon of 8 weeks in 2021, with stable COVID
onto incentive conditions, Study 1.
regulations. By the time of this experiment, at least regular restaurant
DV: Performance DV: Participation guests may have become more acquainted with the Smartglass, such
Independent variables β p β p that we can test whether our main results are robust beyond the
Intercept 3.21 *** 2.04 *** initial introduction of the technology. As in Study 1, the restaurant
was consistently open 5 days per week.
Experimental
manipulation:
Donation (low −0.55 ** −0.18 n.s.

vs. none) 4.2.3 | Experimental manipulation
Setting‐related control
variables: We manipulated daily whether a high on‐top donation (donation of
€0.50), a low on‐top donation (donation of €0.10), or no on‐top
Lottery (nine craft beer 0.70 * −0.27 n.s.
donation was offered. Again, customers learned about this on‐top
sets vs. voucher)
prosocial incentive on the first survey page along with the lottery
Lottery (10 sample −0.36 n.s. 0.06 n.s.
incentive (see Appendix A).
beers vs. voucher)
Lottery (20 sample −0.10 n.s. 0.03 n.s.

beers vs. voucher)
4.2.4 | Outcome measures
Prior NFC experience −0.60 **
(yes vs. no)
We use the same measure for participation and performance as in Study
Glass size (large vs. 0.11 n.s. −0.39 ** 1. We tested whether the experimental manipulation affected scanning
small)
itself, for which we again found no statistical support (Appendix B).
Hour of the day −0.02 n.s. 0.01 n.s.
Weekend (vs. 0.65 ** 0.16 n.s.

weekday) 4.2.5 | Sample and descriptives
Rain [in mm] 0.11 ** −8.54 n.s.
A total of 14,041 beers were served in a Smartglass, of which 15.00%
N 1475 1692
were scanned (2,106 scans). Of the 1,739 completed survey
R2 0.03 0.01
responses (82.57% of all scans), 6.38% contained incorrect data
Note: p < 0.001 ****; p < 0.01 ***; p < 0.05 **; p < 0.10 *; and p > 0.10 n.s. (111 surveys).
The provided answers match those of Study 1: 69.70% male 4.2.7 | Effect of on‐top prosocial incentives on
customers; 58.66% born between 1980 and 1999; 41.29% of the participation
scans happened out of curiosity and 30.25% to win a prize; 61.30%
of scans occurred on more than half‐full glasses; and 66.02% of all We do not observe a significant effect of offering a low on‐top
scans were carried out by respondents who had not used NFC donation on participation relative to offering no on‐top donation
before. (p > 0.1), replicating the results from Study 1 and lending support to
H2a. Besides, we observe a significantly negative effect of offering a
high on‐top donation on participation relative to offering a low on‐
4.2.6 | Effect of on‐top prosocial incentives on top donation (p < 0.05), in line with H2b. The negative effect is in line
performance with the findings by Schwartz et al. (2021) and adds to the discussed
disadvantages of on‐top donations in relation to performance: On‐
The results are summarized in Table 3. We observe a significantly top donations are ineffective at a low amount and even detrimental
negative effect of offering a low on‐top donation on performance to participation at a high amount.
relative to offering no on‐top donation (p < 0.05), replicating the
results from Study 1 and lending further support to H1a. In addition,
the trend reverses once a high on‐top donation is offered: The 4.3 | Study 3 (online experiment)
performance observed when a high on‐top donation is offered
significantly exceeds the one observed when a low on‐top donation is 4.3.1 | Aim
offered (p < 0.05). Thereby, we find empirical support for H1b. We
also examined the regression results when including further Study 3 aims to (i) replicate the findings from our field experiments
respondent‐related control variables (e.g., gender, year of birth) and under conditions of high internal validity, that is, in a more controlled
found the effects to be robust in direction and level of significance. online setting, (ii) investigate the mechanism behind the effect of on‐
top prosocial incentives on performance, and (iii) challenge the
generalizability of the findings from the field experiments by using a
completely different study setting and task. We intentionally
T A B L E 3 Logistic regression of performance and participation
onto incentive conditions, Study 2. diverged from the restaurant's previous marketing and beer setting,
aiming for an incentive‐aligned experimental setup which is based on
DV: Performance DV: Participation
extant studies in the literature, in this case Khan et al. (2020). Since
Independent variables β p β p
extant findings in the literature are largely consistent across different
Intercept −0.72 n.s. 1.75 *** settings that deviate from the classical survey setting (c.f., Table 1),
Experimental we decided to use a game‐based task in Study 3 as a setting‐
manipulation: independent way to study the incentives' effects.
Donation (low −0.48 ** 0.12 n.s.
vs. none)
Donation (high vs. low) 0.68 ** −0.42 ** 4.3.2 | Setting

Setting‐related control
variables: Study 3's questionnaire was inspired by the study in Khan et al.
(2020). It started with a filler task, in which respondents evaluated
Lottery (forty vs. five 0.75 **** −0.29 **
sample beers) several words regarding their perceived pleasantness on a 7‐point
Likert scale (1: very unpleasant; 7: very pleasant), followed by
Prior NFC experience 0.11 n.s.
(yes vs. no) demographic questions (age, gender, occupation). Upon completion,
respondents were informed that they now already earned their cash
Glass size (large vs. −0.41 * −0.06 n.s.
small) reward and that the official survey had ended.
After the filler task, our experiment began on a voluntary basis:
Hour of the day 0.14 *** 0.01 n.s.
We asked respondents whether they want to participate in an
Weekend (vs. 0.30 n.s. −0.06 n.s. optional post‐survey, in which they would be shown five jumbled
weekday)
words, that is, anagrams. Their task was to do their best in trying to
Rain [in mm] 0.02 n.s. 0.05 * rearrange the anagrams' letters such that a sensible word would
Temperature [in °C] 0.09 *** −0.01 n.s. emerge. The example “ETKBAS” ‐> “BASKET” illustrated the task. We
N 1,739 2,106
adapted five of the six anagrams from Goldsmith and Dhar (2013)
and Khan et al. (2020). Three anagrams were rather easy (“OOLSCH”
R2 0.04 0.01
‐> “SCHOOL”; “SEUMO” ‐> “MOUSE”; “DINSLA” ‐> “ISLAND”) and
Note: p < 0.001 ****; p < 0.01 ***; p < 0.05 **; p < 0.10 *; and p > 0.10 n.s. two anagrams were rather difficult (“FABELY” ‐> “LABEFY”;
“UDARIVMIQU” ‐> “QUADRIVIUM”). Including the latter two that a respondent invested into identifying a solution to the
anagrams is a crucial design choice because it allows us to measure anagrams. Theoretically, an alternative performance measure could
how long respondents persist in trying to solve very difficult have been the number of correctly solved anagrams (see Appendix
anagrams. To encourage respondents to do their best, they were C). However, Goldsmith and Dhar (2013) argue against its use
promised a self‐benefiting incentive in the form of a lottery: For because the experiment was designed so that most respondents
every correct anagram, respondents had a chance to win £1, that is, a could solve the three easy but not the two difficult anagrams.
maximum of five chances, and every 50th correctly identified
anagram won (expected reward value of a correct anagram: £0.02).
Besides the lottery, we manipulated whether respondents 4.3.5 | Sample and descriptives
obtained an on‐top prosocial incentive. After reading about the
offered incentive, we asked respondents whether they wanted to A total of 603 respondents completed the survey, of which 91.87%
participate and to indicate how they feel about the incentive on a (554) participated in the post‐survey. These respondents solved an
seven‐point Likert scale (1: strongly disagree; 7: strongly agree; “I average of 2.94 anagrams and had a median post‐survey completion
would feel good working toward this incentive.”, “Thinking about this time of 178.5 s. The sample is balanced in terms of gender (50.25%
incentive would make me feel good.”, “I have positive feelings toward male), is UK‐based, and has a mean age of 40.48 years. The majority
this incentive.”). This scale was proposed by Khan et al. (2020) to of respondents are employees (64.56%).
measure an incentive's affective value. We did not include a measure
of cognitive value since extant literature suggests that it is unaffected
by the amount of a prosocial incentive and is mainly related to the 4.3.6 | Manipulation check
self‐benefiting reward (Khan et al., 2020). Besides, leaving out this
construct allowed us to keep the survey concise. If a respondent Before deciding whether to participate in the post‐survey, respon-
chose not to participate, they were immediately redirected to the dents evaluated the statement “I perceive the maximum donation
survey provider, whereas if they chose to participate, they first amount as relatively high.” on a 7‐point Likert scale (1: strongly
submitted their answers to the anagram task and were then disagree; 7: strongly agree). A two‐sided t‐test reveals that respon-
redirected to the survey provider to signal survey completion. To dents in the condition with a maximum donation amount of £0.50
attract respondents, we used Prolific. We implemented and executed perceive the on‐top prosocial incentive to be significantly higher than
the questionnaire using the online survey platform DISE (Schlereth & those in the condition with a maximum donation amount of £0.10
Skiera, 2012). (x£0.10 = 3.43; x£0.50 = 4.24; t(401) = 5.36, p < 0.001). We conclude
that our manipulation was successful.
4.3.3 | Experimental manipulation

4.3.7 | Effect of on‐top prosocial incentives on
We randomly assigned participants to one of three survey versions, performance
which deviated in whether the respondents received no, a low
(donation of £0.02 for every correct anagram), or a high (donation of We regress our experimental conditions, that is, whether a high, a
£0.10) on‐top prosocial incentive. In versions two and three, we had low, or no on‐top donation was offered, on the natural logarithm of
respondents evaluate the statement “I perceive the maximum performance, measured by the time respondents persisted in trying
donation amount as relatively high.” on a seven‐point Likert scale to solve the anagrams while controlling for respondents' gender, age,
(1: strongly disagree; 7: strongly agree) before deciding whether they and occupation (Table 4). We take the natural logarithm of time since
would like to participate. a quantile plot indicated that this transformation better approximates
the time variable to the normal distribution. The results show that
introducing a low on‐top prosocial incentive does not significantly
4.3.4 | Outcome measures affect performance (p > 0.1); therefore, we do not find support for
H1a in the online experiment. However, once the amount of the on‐
We measured performance and participation similar to Goldsmith top prosocial incentive increases further, the performance shows a
and Dhar (2013) and Khan et al. (2020). Participation is the significant increase (p < 0.1), which is in support of H1b and replicates
percentage of respondents completing the voluntary post‐survey, our findings from Study 2. Together, the results suggest that on‐top
independent of how many anagrams they solved. For performance, prosocial incentives are not suited to increase performance when the
we followed Goldsmith and Dhar (2013), who argue that with two of budget is low and only become beneficial for higher incentive
the anagrams being very difficult to solve, performance captures to amounts.
what extent respondents comply with our instructions to do their We go onto test whether the effect of on‐top prosocial
best possible and persist in solving the task despite its difficulty. This incentives on performance is mediated by affective value (Cronbach's
means that we measure performance as the effort, that is, the time α = 0.95). To this end, we conduct a mediation analysis using
PROCESS Model 4 (n = 10,000) by Hayes and Preacher (2014), with p < 0.1), while affective value does not significantly differ between
the on‐top prosocial incentive amount as multicategorical indepen- the conditions of no and a low on‐top prosocial incentive (a1 = −0.07,
dent variable (low amount as reference condition), affective value as p > 0.1). This observation suggests that a warm glow effect, known to
mediator, gender, age, and occupation as control variables, and the occur for prosocial incentives offered in isolation (e.g.,
natural logarithm of performance as dependent variable, following a Andreoni, 1989; Imas, 2014), does not hold in the case of on‐top
procedure similar to Invernizzi et al. (2022). The model results are prosocial incentives. When regressing the on‐top prosocial incentive
summarized in Figure 5. conditions and affective value simultaneously on performance,
Our results reveal a significant indirect effect of a high relative to affective value has a significantly positive effect on performance
low on‐top prosocial incentive on performance through affective (b = 0.09, p < 0.01), whereas the effect of the high on‐top prosocial
value (indirect effect = 0.02; 90% bootstrap confidence interval [CI]: incentive relative to the low incentive is no longer significant
[0.0009, 0.0385]). At the same time, there is no significant indirect (c'2 = 0.12, p > 0.1). Taken together, these results suggest that the
effect of no relative to a low on‐top prosocial incentive on effect of a high relative to low on‐top prosocial incentive on
performance through affective value (indirect effect = −0.01; 90% performance is fully mediated by affective value.
CI: [−0.0241 to 0.087]).
A high on‐top prosocial incentive yields a significantly higher
affective value than a low on‐top prosocial incentive (a2 = 0.18, 4.3.8 | Effect of on‐top prosocial incentives on
participation
TABLE 4 Regression of performance onto incentive conditions, We regress participation onto our experimental conditions while
Study 3. controlling for gender, age, and occupation (see Table 5). We do not
ln(Performance) observe a significant effect of offering a low on‐top donation on
Independent variables β p participation relative to offering no on‐top donation (p > 0.1), lending
support to H2a. Additionally, we observe a significantly negative
Intercept 4.82 ****
effect of offering a high on‐top donation on participation relative to
Experimental manipulation:
offering a low on‐top donation (p < 0.05), in line with H2b. Thereby,
Donation (low vs. none) −0.00 n.s. we replicate our results from Study 1 and 2.
Donation (high vs. low) 0.13 *
Control variables:
5 | D IS CU SS IO N A N D CO N CL US I O N
Gender (male vs. female) 0.14 **
Age 0.01 *** Companies have traditionally offered either self‐benefiting or

Occupation (employed vs. other) −0.18 *** prosocial incentives to motivate customers to participate in feedback
surveys (participation) and provide accurate responses (performance).
N 554
More recently, several companies have started offering an on‐top
R2 0.06
prosocial incentive and a baseline self‐benefiting incentive in the
Note: p < 0.001 ****; p < 0.01 ***; p < 0.05 **; p < 0.10 *; and p > 0.10 n.s. hope to achieve the advantages in performance and participation of
F I G U R E 5 Effect of no and high (relative to low) on‐top prosocial incentive on performance through affective value. p < 0.001 ****;
p < 0.01 ***; p < 0.05 **; p < 0.10 *; and p > 0.10 n.s.
T A B L E 5 Logistic regression of participation onto incentive For marketeers, our paper outlines different purposes of
conditions, Study 3. customer feedback surveys and describes for each purpose whether
Participation it is more important to focus on participation, performance, or both
Independent variables β p jointly. For example, if a company's goal is to leverage the mere
Intercept 2.68 **** measurement effect (Borle et al., 2007), it will be relatively more
interested in raising participation, whereas if its goal is to understand
customer pain points (Brown, 2008), it will primarily want to raise
Donation (low vs. none) 0.23 n.s.
performance. A company's particular goal may then decide which
Donation (high vs. low) −0.75 ** incentive is most suitable. Depending on the goal, Figure 1 guides
Control variables: managers whether prosocial or self‐benefitting incentives are better
Gender (male vs. female) −1.07 *** suited.

From a theoretical perspective, our findings highlight that the
Age 0.01 n.s.
predictions of utility theory (Jensen, 1967) may not hold in the
Occupation (employed vs. other) 0.36 n.s. context of on‐top prosocial incentivization. While utility theory
N 603 would predict that adding more incentives should increase custom-
R 2
.05 ers' utility and, as a result, raise their performance, our empirical
results do not support this. Our findings suggest that adding a
Note: p < 0.001 ****; p < 0.01 ***; p < 0.05 **; p < 0.10 *; and p > 0.10 n.s.
different type of incentive can change the overall perception of
incentives and alter how respondents process them. Respondents do
not evaluate each part of the incentive individually, but instead,
both incentives. The present paper investigates the effectiveness of depending on the incentive mix, they view it as either predominantly
this approach. self‐benefiting or prosocial. This predominant view, in turn, deter-
In principle, such an incentive combination could “sweeten mines whether respondents derive a high affective value from the
the deal” because the two incentive types offer a certain incentive and, ultimately, whether the incentive can raise
complementarity. Prosocial incentives are suited to raise per- performance.
formance even with a limited incentive budget (Khan et al., 2020) The present work gives rise to several potential avenues for
but are ineffective at raising participation (Schwartz et al., 2021). future research. First, we encourage future research to conduct
On the other hand, self‐benefiting incentives can decrease further field experiments to test the generalizability of our findings.
performance when the incentive amount is low (Gneezy & For example, future research could investigate what happens when
Rustichini, 2000). However, they are generally suited to raise the baseline self‐benefiting incentive is a performance‐dependent
participation (Furse & Stewart, 1982). Combining both incentive cash reward and not, as in the case of our lottery, a probabilistic
types could attenuate the drawbacks of either incentive and outcome. This distinction could be important from a theoretical point
create a situation in which both performance and participation of view because whether the self‐benefiting incentive is probabilistic
react positively to the offered incentive amount. could alter its relative perceived importance depending on partici-
However, our results from two field experiments and one pants' risk preferences. Earlier, we argued that the perception of the
incentive‐aligned online experiment do not reveal any such advan- incentive may change from predominantly self‐benefiting to pre-
tage. Performance stays constant or decreases when a low‐amount dominantly prosocial as the amount of the on‐top prosocial incentive
on‐top prosocial incentive is offered relative to the baseline of no on‐ increases. Suppose respondents are risk averse and heavily discount
top incentive. This trend only reverses once an on‐top prosocial the self‐benefiting incentive in light of its uncertain payout. In this
incentive of a higher amount is offered. Regarding participation, the case, the switch in perception might happen much faster than if the
on‐top prosocial incentive is found to be ineffective and, at higher self‐benefiting incentive had a certain payout. Further, this distinc-
amounts, even detrimental. Therefore, the additional incentive seems tion is important from a practical point of view because out of the
to “poison the well.” companies that offer on‐top prosocial incentives, some indeed
Our study provides essential managerial and theoretical contri- provide a cash reward as a baseline self‐benefiting incentive. For
butions. Managerially, our findings call into question the continuation these companies, it is essential to know whether our findings
of incentivization schemes recently adopted by companies such as generalize to their specific incentive scheme to make an informed
P&G. In particular, managers should be cautious about changing decision about the potential discontinuance.
existing customer incentivization schemes by adding new elements in Second, future research could test our findings' robustness using
the form of on‐top prosocial incentives because such initiatives can alternative performance operationalizations within field studies. Our
have unintended consequences. While managers often have a bias field studies focus on the correctness of data shared in a survey
for action (Patt, 2000), our findings suggest that maintaining the setting. It would be interesting to see whether other types of
status quo regarding customer incentivization may be the superior performance react similarly to on‐top prosocial incentives. Our
choice. literature review provides a helpful overview in this regard,
summarizing several different tasks that have been studied in extant scholarship. Open Access funding enabled and organized by
research, including tasks related to altruistic behavior, work‐related Projekt DEAL.
tasks, or game‐based tasks. For example, Khan et al. (2020) examine
subjects' persistence in a physically demanding task. CONFLIC T OF INTEREST STATEM ENT
Since Study 3 focuses on a game‐based task, future research The author declare no conflict of interest.
could run further online experiments that look at other participation
and performance tasks in a marketing‐ and survey‐related setting. For DATA AVAILABILITY STATEMENT
example, understanding customer pain points is essential to support The data that support the findings of this study are available from the
companies in the ideation phase to develop innovations (e.g., corresponding author upon reasonable request.
Brown, 2008). Many companies use design thinking, a holistic
approach to addressing customer problems, and in this setting, need ORC I D
to gain insights into customer experiences. Thus, it is essential to Sven Beisecker http://orcid.org/0000-0003-1297-7620
motivate customers to share their challenges, especially those often
overlooked. Future studies could ask experienced customers to detail RE F ER EN CES
unpleasant scenarios in their interactions with a company and Agag, G., Durrani, B. A., Shehawy, Y. M., Alharthi, M., Alamoudi, H., El‐
suggest improvements. Participation metrics might include the Halaby, S., Hassanein, A., & Abdelmoety, Z. H. (2023). Understanding
the link between customer feedback metrics and firm performance.
response rate, while performance could gauge the depth and time
Journal of Retailing and Consumer Services, 73(C), 103301.
spent detailing experiences and ideas. Andreoni, J. (1989). Giving with impure altruism: applications to charity
In the present paper, we test a potential mechanism behind the and ricardian equivalence. Journal of Political Economy, 97(6),
effect of on‐top prosocial incentive amounts on performance by 1447–1458.
Andreoni, J. (1990). Impure altruism and donations to public goods: a
looking at affective value. While we find evidence that affective value
theory of warm‐glow giving. The Economic Journal, 100(401),
indeed mediates the relationship, future research should test 464–477.
alternative process hypotheses and thereby rule out the potential Andreoni, J. (1993). An experimental test of the public‐goods
influence of other mediators. Ultimately, this could inform the crowding‐out hypothesis. The American Economic Review, 83(5),
1317–1327.
development of appropriate intervention strategies. For example,
Ariely, D., Bracha, A., & Meier, S. (2009). Doing good or doing well? Image
potential intervention strategies for future research to test include motivation and monetary incentives in behaving prosocially.
using a more emotionally charged message framing or increasing the American Economic Review, 99(1), 544–555.
salience of the on‐top prosocial incentive by highlighting what even a Barasch, A., Levine, E. E., Berman, J. Z., & Small, D. A. (2014). Selfish or
selfless? on the signal value of emotion in altruistic behavior. Journal
low donation amount can achieve for the beneficiaries.
of Personality and Social Psychology, 107(3), 393–413.
As more and more customers get used to being offered on‐top
Berman, J. Z., & Small, D. A. (2012). Self‐interest without selfishness: the
prosocial incentives, we cannot preclude that they might gradually hedonic benefit of imposed self‐interest. Psychological Science,
become more receptive to this kind of incentivization. Therefore, 23(10), 1193–1199.
future research should aim to disentangle to what extent our findings Bezençon, V., Girardin, F., & Lunardo, R. (2020). When does an ethical
attribute matter for product evaluation? The role of warm‐glow
are driven by the mere novelty of the incentive scheme as compared
feelings for low‐rated products. Psychology & Marketing, 37(11),
to deep‐rooted consumer psychology and preferences. Similarly, the 1571–1585.
proposed incentive scheme's medium‐ and long‐term consequences Biner, P. M., & Barton, D. L. (1990). Justifying the enclosure of monetary
should be evaluated. Companies are arguably less interested in the incentives in mail survey cover letters. Psychology & Marketing, 7(3),
153–162.
answers to a single customer feedback survey but more in developing
Blohm, M., & Koch, A. (2013). Respondent incentives in a national face‐to‐
answers over time and customers' continued engagement. Longitudi- face survey,“ methods. data, analyses, 7(1), 89–122.
nal field studies could capture such effects. Bone, S. A., Lemon, K. N., Voorhees, C. M., Liljenquist, K. A.,
Fombelle, P. W., Detienne, K. B., & Money, R. B. (2017). ‘Mere
measurement plus’: how solicitation of open‐ended positive feed-
A C KN O W L E D G E ME N T S
back influences customer purchase behavior. Journal of Marketing
We thank Connect One Digital AG, particularly their founders Research, 54(1), 156–170.
Carsten Kehrein and Thomas Nieraad, for granting us access to their Borle, S., Dholakia, U. M., Singh, S. S., & Westbrook, R. A. (2007). The
IoT communication technology solution in the field experiments as impact of survey participation on subsequent customer behavior: an
empirical investigation. Marketing Science, 26(5), 711–726.
well as Klaus Brockhoff for raising our attention on Smartglasses. We
Bosnjak, M., & Tuten, T. L. (2003). Prepaid and promised incentives in web
also thank Maximilians Brauwiesen GmbH and Family Ohlig for surveys: an experiment. Social Science Computer Review, 21(2),
conducting the field experiments in their restaurant. Further, we are 208–217.
grateful to Prof. Dr. Bernd Skiera and his doctoral students for their Brown, T. (2008). Design thinking. Harvard Business Review, 86(6), 84–92.
Brown, T., & Katz, B. (2011). Change by design. Journal of Product
valuable feedback during several doctoral colloquia and to Prof. Dr.
Innovation Management, 28(3), 381–383.
Walter Herzog for his guidance as second PhD supervisor. Finally, we
Brüggen, E., & Dholakia, U. M. (2010). Determinants of participation and
would like to thank the Konrad‐Adenauer‐Stiftung e. V. for financially response effort in web panel surveys. Journal of Interactive
and intellectually supporting the first author through a PhD Marketing, 24(3), 239–250.
Celuch, K., & Walz, A. M. (2020). The role of active identification in driving online access panels. International Journal of Market Research, 46(3),
retail customer feedback. Services Marketing Quarterly, 41(2), 327–345.
163–181. Groves, R. M., Cialdini, R. B., & Couper, M. P. (1992). Understanding the
Challagalla, G., Venkatesh, R., & Kohli, A. K. (2009). Proactive postsales decision to participate in a survey. Public Opinion Quarterly, 56(4),
service: when and why does it pay off? Journal of Marketing, 73(2), 475–495.
70–87. Hart, C. W., Heskett, J. L., & Sasser, Jr WE (1990). The profitable art of
Chang, H. H., & Tuan Pham, M. (2013). Affect as a decision‐making system service recovery. Harvard Business Review, 68(4), 148–156.
of the present. Journal of Consumer Research, 40(1), 42–63. Hayes, A. F., & Preacher, K. J. (2014). Statistical mediation analysis with a
Charness, G., Cobo‐Reyes, R., & Sánchez, Á. (2016). The effect of multicategorical independent variable. British Journal of
charitable giving on workers' performance: experimental evidence. Mathematical and Statistical Psychology, 67(3), 451–470.
Journal of Economic Behavior and Organization, 131(A), 61–74. Heerwegh, D. (2006). An investigation of the effect of lotteries on web
Chen, H., Hu, Y. J., & Huang, S. (2019). Monetary incentive and stock survey response rates. Field Methods, 18(2), 205–220.
opinions on social media. Journal of Management Information Hennig‐Thurau, T., Gwinner, K. P., Walsh, G., & Gremler, D. D. (2004).
Systems, 36(2), 391–417. Electronic word‐of‐mouth via consumer‐opinion platforms: what
Cohen, A. J., Washington, S., Butler, C., Kamal, P., Patino, G., Tresh, A., motivates consumers to articulate themselves on the Internet?
Mena, J., Ndoye, M., & Breyer, B. N. (2019). Altruistic donation to Journal of Interactive Marketing, 18(1), 38–52.
improve survey responses: A global randomized trial. BMC Research Heyman, J., & Ariely, D. (2004). Effort for payment: a tale of two markets.
Notes, 12(1), 113. Psychological Science, 15(11), 787–793.
Conn, K. M., Mo, C. H., & Sellers, L. M. (2019). When less is more in Hsee, C. K., & Rottenstreich, Y. (2004). Music, pandas, and muggers: on
boosting survey response rates*. Social Science Quarterly, 100(3), the affective psychology of value. Journal of Experimental Psychology:
1445–1458. General, 133(1), 23–30.
Dana, J., Weber, R. A., & Kuang, J. X. (2007). Exploiting moral wiggle room: Imas, A. (2014). Working for the “warm glow“: On the benefits and limits
experiments demonstrating an illusory preference for fairness. of prosocial incentives. Journal of Public Economics, 114, 14–18.
Economic Theory, 33(1), 67–80. Imas, A., & Loewenstein, G. (2018). Is altruism sensitive to scope? the role
Dholakia, U. M., & Morwitz, V. G. (2002). The scope and persistence of of tangibility. AEA Papers and Proceedings, 108(May), 143–147.
mere‐measurement effects: evidence from a field study of customer Invernizzi, A. C., Bellucci, M., Acuti, D., & Manetti, G. (2022). Form and
satisfaction measurement. Journal of Consumer Research, 29(2), substance: visual content in CSR reports and investors' perceptions.
159–167. Psychology & Marketing, 39(5), 974–989.
Feiler, D. C., Tost, L. P., & Grant, A. M. (2012). Mixed reasons, missed Jagger. (2023), Best paid survey app that plants trees. Accessed
givings: the costs of blending egoistic and altruistic reasons in November 17, 2023. https://jaggerapp.com/
donation requests. Journal of Experimental Social Psychology, 48(6), Jensen, N. E. (1967). An introduction to bernoullian utility theory: I. utility
1322–1328. functions. The Swedish Journal of Economics, 69(3), 163–183.
Furse, D. H., & Stewart, D. W. (1982). Monetary incentives versus Khan, U., Goldsmith, K., & Dhar, R. (2020). When does altruism trump
promised contribution to charity: new evidence on mail survey self‐interest? The moderating role of affect in extrinsic incen-
response. Journal of Marketing Research, 19(3), 375–380. tives. Journal of the Association for Consumer Research, 5(1),
Gendall, P., & Healey, B. (2010). Effect of a promised donation to charity 44–55.
on survey response. International Journal of Market Research, 52(5), Labroo, A. A., & Goldsmith, K. (2021). The dirty underbelly of prosocial
565–577. behavior: reconceptualizing greater good as an ecosystem with
Gershon, R., Cryder, C., & John, L. K. (2020). Why prosocial referral unintended consequences. Journal of Consumer Psychology, 31(3),
incentives work: the interplay of reputational benefits and action 417–428.
costs. Journal of Marketing Research, 57(1), 156–172. Lacetera, N., & Macis, M. (2010). Do all material incentives for pro‐social
Gneezy, U., Imas, A., & Madarász, K. (2014). Conscience accounting: activities backfire? The response to cash and non‐cash incentives for
emotion dynamics and social behavior. Management Science, 60(11), blood donations. Journal of Economic Psychology, 31(4), 738–748.
2645–2658. Loewenstein, G. F., Weber, E. U., Hsee, C. K., & Welch, N. (2001). Risk as
Gneezy, U., Meier, S., & Rey‐Biel, P. (2011). When and why incentives feelings. Psychological Bulletin, 127(2), 267–286.
(Don't) work to modify behavior. Journal of Economic Perspectives, Lovelock, C. H., & Wirtz, J. (2010). Services Marketing. Prentice Hall
25(4), 191–210. International.
Gneezy, U., & Rustichini, A. (2000). Pay enough or don't pay at all. Markey, R., Reichheld, F., & Dullweber, A. (2009). Closing the customer
Quarterly Journal of Economics, 115(3), 791–810. feedback loop. Harvard Business Review, 87(12), 43–47.
Goldsmith, K., & Dhar, R. (2013). Negativity bias and task motivation: Morwitz, V. G., Johnson, E., & Schmittlein, D. (1993). Does measuring
testing the effectiveness of positively versus negatively framed intent change behavior? Journal of Consumer Research, 20(1), 46–61.
incentives. Journal of Experimental Psychology: Applied, 19(4), Muñoz Sabater, J. (2019). ERA5‐Land hourly data from 1981 to present,
358–366. Copernicus Climate Change Service (C3S) Climate Data Store (CDS)
Göritz, A. S., & Luthe, S. C. (2013). Lotteries and study results in market [data set]. https://doi.org/10.24381/cds.e2161bac
research online panels. International Journal of Market Research, Newman, G. E., & Cain, D. M. (2014). Tainted altruism: when doing some
55(5), 611–626. good is evaluated as worse than doing no good at all. Psychological
Göritz, A. S., & Neumann, B. P. (2016). The longitudinal effects of Science, 25(3), 648–655.
incentives on response quantity in online panels. Translational Issues Olivola, C. Y., & Shafir, E. (2013). The martyrdom effect: when pain and
in Psychological Science, 2(2), 163–173. effort increase prosocial contributions. Journal of Behavioral Decision
Göritz, A. S., & Wolff, H.‐G. (2007). Lotteries as incentives in longitudinal Making, 26(1), 91–105.
web studies. Social Science Computer Review, 25(1), 99–110. Patt, A. (2000). Action bias and environmental decisions. Journal of Risk
Griffin, A., & Hauser, J. R. (1993). The voice of the customer. Marketing and Uncertainty, 21(1), 45–72.
Science, 12(1), 1–27. Porter, S. R., & Whitcomb, M. E. (2003). The impact of lottery incentives
Gritz, A. S. (2004). The impact of material incentives on response quantity, on student survey response rates. Research in Higher Education,
response quality, sample composition, survey outcome and cost in 44(4), 389–407.
Procter & Gamble. (2023). Sign up. Earn reward. Make an impact. a Canadian incentives experiment. Public Opinion Quarterly, 60(4),
Accessed November 17, 2023. https://www.pggoodeveryday.com/ 542–562.
Richins, M. L. (1983). Negative word‐of‐mouth by dissatisfied consumers: White, K., Habib, R., & Dahl, D. W. (2020). A review and
a pilot study. Journal of Marketing, 47(1), 68–78. framework for thinking about the drivers of prosocial
Schlereth, C., & Skiera, B. (2012). DISE: Dynamic intelligent survey engine. consumer behavior. Journal of the Association for Consumer
In A. Diamantopoulos, W. Fritz, & L. Hildebrandt (Eds.), Quantitative Research, 5(1), 2–18.
marketing and marketing management‐festschrift in honor of Udo Williams, P., Fitzsimons, G. J., & Block, L. G. (2004). When
Wagner (pp. 225–243). Gabler Verlag. consumers do not recognize “benign“ intention questions as
Schwartz, D., Keenan, E. A., Imas, A., & Gneezy, A. (2021). Opting‐in to persuasion attempts. Journal of Consumer Research, 31(3),
prosocial incentives. Organizational Behavior and Human Decision 540–550.
Processes, 163(March), 132–141. Yin, B., Li, Y. J., & Singh, S. (2020). Coins are cold and cards are
Siuki, H., & Webster, C. M. (2021). Social good or self‐interest: caring: the effect of pregiving incentives on charity perceptions,
incentivizing consumer social media engagement behaviour for relationship norms, and donation behavior. Journal of Marketing,
health messages. Psychology & Marketing, 38(8), 1293–1313. 84(6), 57–73.
Small, D. A., Loewenstein, G., & Slovic, P. (2007). Sympathy and
callousness: the impact of deliberative thought on donations to
identifiable and statistical victims. Organizational Behavior and
Human Decision Processes, 102(2), 143–153. How to cite this article: Beisecker, S., & Schlereth, C. (2023).
Tonin, M., & Vlassopoulos, M. (2015). Corporate philanthropy and Offering prosocial incentives on‐top: Do they sweeten the
productivity: evidence from an online real effort experiment.
deal or poison the well? Psychology & Marketing, 1–21.
Management Science, 61(8), 1795–1811.
Warriner, K., Goyder, J., Gjertsen, H., Hohner, P., & McSpurren, K. (1996). https://doi.org/10.1002/mar.21941
Charities, no; lotteries, no; cash, yes: main effects and interactions in
A P P E N D IX A
See Table A1.
T A B L E A1 Experimental design of the field experiments, Studies 1 and 2.
Donation Donation
Week Day Restaurant's lottery prizes condition Restaurant's lottery prizes condition
1 1 One chance to win a voucher for a free None Five chances to win a None
meat platter selection of sample beers
2 Low (€0.10) Low (€0.10)
3 None None
4 Low (€0.10) Low (€0.10)
5 None None
2 1 Twenty chances to win a selection of None Forty chances to win a None

sample beers selection of sample beers
2 Low (€0.10) Low (€0.10)
3 None None
4 Low (€0.10) Low (€0.10)
5 None None
3 1 Nine chances to win a craft beer set None Five chances to win a None
selection of sample beers
2 Low (€0.10) High (€0.50)
3 None None
4 Low (€0.10) High (€0.50)
5 None None
4 1 Ten chances to win a selection of None Forty chances to win a None

sample beers selection of sample beers
2 Low (€0.10) High (€0.50)
3 None None
4 Low (€0.10) High (€0.50)
5 None None
5 1 Five chances to win a None

2 Low (€0.10)
3 None
4 Low (€0.10)
5 None
6 1 Forty chances to win a None

2 Low (€0.10)
3 None
4 Low (€0.10)
5 None
7 1 Five chances to win a None

2 High (€0.50)
3 None
4 High (€0.50)
5 None
TABLE A1 (Continued)
Donation Donation
Week Day Restaurant's lottery prizes condition Restaurant's lottery prizes condition
8 1 Forty chances to win a None

2 High (€0.50)
3 None
4 High (€0.50)
5 None
A P P E N D IX B APPENDIX C
Table B1 summarizes the scan rates in each experimental condition, Table C1 provides the results of regressing the number of correctly
separately for Study 1 and 2. For each study, we ran an ANOVA to solved anagrams onto the incentive conditions. Similar to the results
test whether there are any significant differences in the daily scan of Goldsmith and Dhar (2013), we find no significant effect of the
rates across conditions. The results confirm that there are no experimental manipulations (p > 0.1). This is simply an artifact of our
significant differences, neither in Study 1 (F(1, 18) = 0.15, p > 0.1) experimental design, which was set up with the expectation that
nor in Study 2 (F(2, 37) = 1.54, p > 0.1). We conclude that the decision most respondents should be able to solve the three easy, but not the
to scan is not affected by the provided incentives and, therefore, our two difficult anagrams (Goldsmith & Dhar, 2013).
field studies constitute proper experiments.
T A B L E B1 Scan rates by experimental condition, Studies 1 T A B L E C1 Regression of number of correct anagrams onto
and 2. incentive conditions, Study 3.
Scan rates Number of correct anagrams

On‐top prosocial incentive Study 1 Study 2 Independent variables β p
None 23.75% 13.77%
Intercept 3.13 ****
Low 25.09% 16.66%
High ‐ 16.76%
Donation (low vs. none) −0.11 n.s.
Donation (high vs. low) 0.18 n.s.
Control variables:
Gender (male vs. female) 0.20 **
Age −0.01 *
Occupation (employed vs. other) −0.19 *
N 554
2
R .02
Note: p < 0.001 ****; p < 0.01 ***; p < 0.05 **; p < 0.10 *; and p > 0.10 n.s.

Experiments-Offering Prosocial Incentives On-Top Do They Sweet

Uploaded by

Copyright:

Available Formats

Experiments-Offering Prosocial Incentives On-Top Do They Sweet

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Experiments-Offering Prosocial Incentives On-Top Do They Sweet

Uploaded by

Copyright:

Available Formats

Received: 14 March 2023 | Accepted: 9 November 2023

Offering prosocial incentives on‐top: Do they sweeten

Sven Beisecker | Christian Schlereth

WHU—Otto Beisheim School of Management,

1 | INTRODUCTION Given these advantages, the question arises how best to

Psychol Mark. 2023;1–21. wileyonlinelibrary.com/journal/mar | 1

2.2.2 | Effect of prosocial incentives on

TABLE 1 Literature overview.

Surveys Survey among Cohen ✔ ✔ ✔ No significant effect of prosocial

Market research panel Göritz and ✔ ✔ ✔ Significantly negative effect of prosocial

General social survey Blohm and ✔ ✔ ✔ Significantly positive effect of self‐

Market research panel Göritz and ✔ ✔ ✔ Significantly positive effect of self‐

Mail survey on religion Gendall and ✔ ✔ ✔ No significant effect of prosocial

Market research panel Brüggen and ✔ ✔ ✔ Respondents’ personality traits can

Survey among Göritz and ✔ ✔ ✔ Significantly positive effect of self‐

Survey among students Heerwegh (2006) ✔ ✔ ✔ Significantly positive effect of self‐

Survey among members Bosnjak and ✔ ✔ ✔ Significantly positive effect of self‐

Survey among college Porter and ✔ ✔ ✔ ✔ No significant effect of self‐benefiting

benefiting incentive on performance

Mail survey on Warriner ✔ ✔ ✔ ✔ Participation increases with self‐

Mail survey among Furse and ✔ ✔ ✔ ✔ Participation increases with self‐

Donations to charity Yin, Li, and ✔ ✔ ✔ ✔ Monetary incentives increase

Blood donations Lacetera and ✔ ✔ ✔ Significantly negative effect of self‐

Donation collection by Gneezy and ✔ ✔ ✔ Small self‐benefiting incentives have a

Rating behavior Newman and ✔ ✔ ✔ ✔ Tainted‐altruism effect: a company's

Blood donations, Gneezy ✔ ✔ ✔ Self‐benefiting incentives of a small

the incentive amount increases

Work Data entry Charness ✔ ✔ ✔ ✔ Self‐benefiting incentives yield higher

Entry of bibliographic Tonin and ✔ ✔ ✔ ✔ Prosocial incentives increase

Publishing articles on an Chen et al. (2019) ✔ ✔ ✔ ✔ Self‐benefiting incentives increase

Customer referrals Gershon ✔ ✔ ✔ ✔ Prosocial incentives increase sign‐up

Recycling campaign; Schwartz ✔ ✔ ✔ No significant effect of low prosocial

Games Wall‐sit task, Khan et al. (2020) ✔ ✔ ✔ ✔ Self‐benefiting incentives increase

Hand dynamometer Imas (2014) ✔ ✔ ✔ ✔ Performance increases along with the

Comparing the effects of prosocial and self‐benefiting incentives,

incentives at low amounts and for

self‐benefiting incentives at high

Self‐benefiting incentives of a small

increasing the incentive amount

choose to work for prosocial

2.3 | Effect of self‐benefiting and prosocial

While many studies focus on performance conditional on participa-

conducted in a lab environment and cannot record subjects' opt‐in

decisions, a few studies also examine how incentives affect

invites by mail. Hereafter, we discuss the effects of self‐benefiting

2.3.1 | Effect of self‐benefiting incentives on

Regarding self‐benefiting incentives, participation tends to increase

non‐survey tasks, such as tasks requiring altruistic behavior (Yin, Li,

and Singh, 2020) and work‐related tasks, for example, publishing

study about blood donations, Lacetera and Macis (2010) demonstrate

incentive can significantly decrease participation. The authors argue

that this occurred because the self‐benefiting incentive reduces the

altruistic behavior's reputational benefits (Ariely et al., 2009; Labroo

2.3.2 | Effect of prosocial incentives on participation

Regarding prosocial incentives, offering a low amount usually does

not significantly affect participation likelihood, and increasing the

likelihood. This has been found consistently across settings including

non‐survey tasks, in particular, work‐related tasks (Schwartz

Donation (high vs. low) 0.68 −0.42 4.3.2 | Setting