64
E-Business and Competitive Strategy
Proposers and responders with a strong social relationship are helpful for both sending and adopting the
invitation. A successful invitation does not only depend on the proposer to send it out, but also its
adoption by the responder as well. Therefore considering both sides of the referral, we hypothesize:
H1c: A successful referral is more likely to occur between friends with small social distance than
friends with large social distance.
Effect of Sense of Fairness on Referral Performance
In our context, the proposer has the right to send the social referral and would assume the risk that the
responder would refuse the proposal due to perceived unfairness, which resembles the “ultimatum game”.
The ultimatum game (Güth et al. 1982; Güth et al. 1990) is a two-player game where Player 1, the proposer,
can offer to divide a fixed total amount, say $10, by giving x amount to Player 2 and keeping $10-x for
himself. Player 2 then decides whether to accept or reject the offer. In the unique sub-game perfect Nash
equilibrium (Gibbons 1989), Player 1 takes the whole amount minus
ε
(
ε
-->0) and Player 2 accepts
ε
,
with an equilibrium solution of (10-
ε
,
ε
). As a matter of fact, if
ε
is given to be 0, multiple equilibriums
would emerge as (10, 0) and (0, 0), and there should be no noticeable difference in probability between
these two equilibrium situations for Player 2. However, numerous experimental studies have shown that
proposers offering less than 30% of the total amount are likely to be rejected, while a fair offer (i.e., 50/50
split) is most likely to be accepted by responders (Güth et al. 1982; Güth et al. 1990). Previous studies
point that 50-50 split is a fair split, which is an ‘obvious’ and ‘acceptable’ compromise, and that “such
considerations are easily displaced by calculations of strategic advantage, once players fully appreciate the
structure of the game”(Binmore et al. 1985; Güth et al. 1990). We follow the definition of fairness in our
study. In online social referral incentive systems, proposer and responder gain the same bonus (5$) from
successful referral which is a fair split (50-50).
Although many referral incentive systems resemble the classic ultimatum game, they have their own
unique features. In the traditional ultimatum game studied in prior literatures, the proposer and the
responder are typically strangers and will not meet in the future. The online social referral incentive
systems take place in situations where the proposer and responder are connected with social networks,
albeit with different social distances. The purpose of an ultimatum game focuses on the split of bonus, but
online social referral incentive systems are based on helping your friends by introducing something new
and useful to them. Different purposes determine people’s different behavior. Although some extant
studies have taken the impact of social relationship into the ultimatum game (Charness et al. 2008;
Macfarlan et al. 2008), based on helping friends, the impact of social relationship in online social referral
incentive systems differs from an ultimatum game.
In this study we use the "Bogardus social distance scale" to measure subjective-affective conception of
social distance (Bogardus 1925). In social distance studies, the center of attention is on the feeling
reactions of people toward other people and toward groups of people. So we also ground our testable
hypotheses based upon theories of social distance (Bogardus 1925), in addition to the sense of fairness in a
quasi-ultimatum game (Güth et al. 1982). With large social distance, the proposer and the responder view
each other as strangers. Therefore, in this respect, online social referral incentive systems resemble
ultimatum games. Based on game theory, people are rational or selfish which means their aim focus on
maximizing their own utility(Güth et al. 1982). Both sides of the referral take fairness into consideration;
both want to maximize their own utility and proposers also fear the responder’s rejection for getting
nothing in the online social referral incentive systems. So fairness is crucial for the success of online social
referral incentive systems with large social distance. Therefore, we propose:
H2: A successful referral is more likely to occur for a fair offer than an unfair offer when proposer
and responder’s social distance is large.
Interaction Effect of Sense of Fairness and Social Distance
In online social referral incentive systems, sending a referral to friends is based on social norms, while the
split of bonuses is based on individual rationality (Camerer et al. 2004). Therefore, the effectiveness of
online social referral incentive systems might be affected by both individual rationality (maximizing
monetary profits based on the fairness rule) and social norms (deriving utility from helping a friend). At
this time, proposer and responder should be taken as social actors in online social referral incentive
53
Shi et al. /Social Distance and Norms in Referral Incentive System
Thirty Fourth International Conference on Information Systems, Milan 2013 7
systems rather than users (Lamb et al. 2003). Online social referral incentive systems should be designed
in terms of a certain context involving social norms and individual rationality.
Individual rationalitys are pervasive in interpersonal and inter-organizational relationships (Heyman et al.
2004). In most cases, especially in business contexts, people make decisions based on utility maximization
and fairness. On the other hand, social norms capture the behavior of people with different social
distances (Camerer et al. 2004). Therefore, with different social distances, the utility function can be
different for the same person. In this study, we use social distance to determine the social context. Social
distance is one dimension of psychological distance (Liberman et al. 2008). Construal level theory (CLT)
links psychological distance from objects (events) to the mental construal of those objects (events), which
offers an explanation as to why psychological distance would affect the impact of monetary incentive on
the performance of social referrals (Trope et al. 2010). The general idea is that the more distant an object
is from the individual, the more abstract it will seem to them, while the opposite relation between
closeness and concreteness is true as well.
Any event or object can be represented at different levels of construal (high or low). High and low levels of
construal influence people’s different mental construal processes. Lower-level construals are concrete,
relatively unstructured and contextualized representations that include subordinate and incidental
features of events. Higher-level construals are abstract, schematic and decontextualized representations
that extract the gist from the available information. They emphasize superordinate, core features of events
and omit incidental features that may vary without significantly changing the meaning of events. Social
distance as one dimension of psychological distance may influence prediction, evaluation, and action,
inasmuch as these outcomes are mediated by the construal. High-level construal promotes attunement to
what is consistent about an object across multiple contexts such as the split of bonus in the referral. As
social distance increases, the effect of referral bonus split increases. With large social distance, people tend
to care about the social relationship less; therefore extrinsic motivation such as monetary incentive may
dominate the proposer's behavior. With small social distance, psychological proximity triggers low-level
construal, which includes the concrete and contextualized aspects of a referral (the close social
relationship between proposer and responder). Small social distance triggers the close social relationship
context. Facing close friends, social norms determine the behavior of proposer and responder. Although
different cultures have their own social norms (Fiske 1992), friendship is generally viewed similarly across
cultures. A friend is someone sharing similar interests, helping when in need. Close friendship is even
worth spending time and money to maintain. Seeking fairness from close friends would be taken as
neglecting close social relationship and as harmful for a close social relationship. So with small social
distance, proposer and responder would focus on the close social relationship to help friends rather than
fairness of bonus split. At this time, helping friends is the real meaning of a referral, which is good for
social relationship. Based on social norms, the intrinsic affective motivation dominates people’s behavior.
Therefore we propose:
H3a: The split of bonus offered by a proposer is more prone to be a fair offer (5,5) to a responder
with a large social distance than with a small social distance.
H3b: A responder is more likely to refuse an unfair split of bonus offered by a proposer with a large
social distance than with a small social distance.
Research Methodology
We use both a set of controlled laboratory experiments with real monetary incentives and a randomized
field experiment to test our posted hypotheses. We first report the design, data collection, and analysis for
the lab experiment. We corroborate our lab experiment results with a randomized field experiment.
Controlled Lab Experiment
The experiment is divided into two integral parts, one on the proposers and the other on the responders.
Subjects participating in the experiment as proposers are different from those taking part as responders.
To resemble a real-life online social referral system, proposers and responders take part in the experiment
separately and are not allowed to see each other during the experiment. Subjects would be told that the
referral coming from a normal online group-buy website. The online group-buy website contains many
kinds of merchandises, which is shown on the screen for subjects. The seats are randomly assigned to
70
E-Business and Competitive Strategy
proposers or responders in the lab. Related concepts such as social distance(large, medium, small) were
explained to all subjects before the experiment, which is based on the description of Bogardus social
distance (Bogardus 1925); pretests were conducted to make sure subjects could correctly understand the
meaning of the context, task and all questions. Demographics are shown in Table 1. Proposers are asked to
send the referral with the website address to a friend with different social distances by email, IM, or other
social networking website. Responders are asked to answer whether they would accept the referral coming
from their friends with certain social distances. We recruited a total of 720 subjects, undergraduate
students from a large public university. Each subject attending the experiment received $10 as the reward
for participation. According to extant research in IS and marketing, students are active online shoppers
and comprise a representative sample (Sia et al. 2009).
Table 1. Demographics
Gender
Age
Online Shopping
Experience
Proposer
Male 58.86%
21.35(1.414)
1.93(0.607)
Responder
Male 55.32%
21.26(1.343)
2.059(0.53)
The first factor in the design is social distance. The experiments for proposers had three groups ranging
from large to small social distance as a trichotomy (Bogardus 1925; Bogardus 1933). For large social
distance, subjects are asked to imagine friends meeting a few days ago by Internet. You have never met in
reality. The Internet is the only way to communicate with your friends. For medium social distance,
subjects are asked to imagine friends as your classmate or co-worker. You would meet every day, but your
topics always focus on the business rather than a private party. For small social distance, subjects are
asked to imagine friends having a familial relationship or intimate relationship with each other. You have
your own private party every week, and your topics are quite private. Each group had 60 proposers. The
second factor is split of referral bonus. We had three different split schemes: (0, 10), for which the
proposer would receive $0 and the responder would receive $10; (5, 5), for which both the responder and
proposer would receive $5; (10, 0), for which the responder would receive $0 and the proposer would
receive $10. The proposer could choose one of the three splits of bonus under one of the three social
distance scenarios. Since there are three levels of social distance, and three different referral bonus split
schemes, we employed two 3*3 full factorial designs for the responders. Subjects were randomly selected
into each group (60 subjects). We performed the analysis of statistical power (Cohen 1992), and our
sample size is shown to have adequate statistical power (>80%) to detect an effect.
Table 2. 3*3 Full Factorial Experimental Design
(0, 10)
Small social distance
(5, 5)
Small social distance
(10, 0)
Small social distance
(0, 10)
Medium social distance
(5, 5)
Medium social distance
(10, 0)
Medium social distance
(0, 10)
Large social distance
(5, 5)
Large social distance
(10, 0)
Large social distance
Before subjects received treatments, they were told the duty of the responder (register for a website and
make a purchase) and the purpose of the referral with bonus, respectively within each group. Subjects in
different groups were not allowed to communicate about the study. After respondents received treatments,
they were asked to complete a questionnaire. Subjects in each treatment were also informed that the
experiment was anonymous and they were not allowed to communicate in any form during the experiment.
We used a 7-point Likert-type scale to measure the tendency of proposers to send referrals to friends. We
were able to test the tendency of proposers to send referrals to friends with different social distances.
Proposers were given three choices: give the total referral bonus of $10 to the responder, keep the whole
referral bonus of $10, or divide it equally. Responders were asked whether they would accept the referral
with a different split of bonus coming from friends with different social distance. A manipulation check is
performed to ensure the respondents have received the treatments.
39
Shi et al. /Social Distance and Norms in Referral Incentive System
Thirty Fourth International Conference on Information Systems, Milan 2013 9
Controlled Lab Experiments Data Analyses
To test our hypotheses, independent sample t-tests, one- and two-way ANOVA analyses were employed.
We had three bonus split schemes in our lab experiment study (10,0), (5,5) and (0,10). A proposer’s choice
of (10,0) means $5 deviation from a fair offer (5,5), and the same goes for (0,10). To measure the deviation
from a fair offer (5, 5), we treat (10,0) in the same way as (0,10). The choice (5,5) is marked as 5. The
choice (10,0) (responder gets nothing) and (0,10) (responder gets the whole amount) are both marked as
10. The average amount offered by the proposer deviates from the fair offer (5,5) with decreasing social
distance (F=14.343, p=0<0.05), supporting H3a.
The intention to send referrals to friends increases as social distance decreases (F=73.73, p<0.001),
supporting H1a. Also, with decreasing social distance, the percentage of responders who accepted the
referral increased, supporting H1b (F=19.94, p<0.05). We measured referral success by multiplying the
value of proposers’ intention to send referrals to friends with a given social distance (e.g., large) times the
percentage of responders who accepted the referral with the same social distance, respectively. We
observed that the successful referral with small social distance was higher than either large or medium
social distance. All comparisons were statistically significant. Therefore Hypothesis H1c was supported.
We then calculated the percentage of responders who accepted the offer in all treatments. With small
social distance, the average percentage of adoption of responders getting nothing (10, 0), $5 (5, 5) and $10
(0, 10) was 0.895, 0.912 and 0.933, respectively (Figure 1). The one-way ANOVA analysis revealed an
insignificant effect (F=0.274, p=0.761>0.1). Therefore, a monetary incentive did not seem to enhance
referral performance for friends with small social distance. The results indicate that close friends are more
affected by social norms of accepting a close friend’s recommendation (referral) than by individual
rationality (monetary incentive).
Figure 1. Adoption of Referrals with Different Referral Bonuses and Social Distances
With medium social distance, the average percentage of adoption of the responders getting nothing (10, 0),
$5 (5, 5) and $10 (0, 10) was 0.64, 0.86 and 0.90, respectively (Figure 1). The percentage of adoption for
responders who got nothing was lower than for responders who got $5 (t=-2.759, p=0.007<0.05). The
percentage of adoption for the responders who got nothing was lower than for responders who got $10
(t=-3.458, p=0.001<0.05). The percentage of adoption for the responders who got $5 was not statistically
significant in relation to the responders who got $10 (t=-0.668, p=0.506>0.1). The one-way ANOVA
analysis showed F=7.5, p=0.001<0.05. Therefore, the effect of monetary incentive for medium social
distance was stronger than the effect for small social distance.
With large social distance, the average percentage of adoption of the responders getting nothing (10, 0), $5
(5, 5) and $10 (0, 10) was 0.42, 0.84 and 0.71, respectively (Figure 1). Comparing the percentage of
responders who accepted the referral, the group with a fair split (5, 5) was higher than the group that got
nothing (t=-4.982, p=0.001<0.05). For the percentage of responders who accepted the referral, the group
that got the whole pie (0, 10) was higher than the group that got nothing (t=-3.374, p=0.001<0.05).
Surprisingly, the group with a fair split (5, 5) was higher than the group that got the whole pie, but
60
E-Business and Competitive Strategy
insignificant (t=1.441, p=0.152>0.1). The one-way ANOVA analysis showed F=12.947, p=0<0.05. In sum,
H3b was supported.
The analysis of Hypothesis 3a implied that proposers are prone to choose a fair split of referral bonus
when facing “friends” with a large social distance. To test this formally, we calculated the percentage of
successful referrals based on the multiplication of the percentages of proposers who proposed a fair split of
bonus (e.g., both get $5) with the percentage of adoption by responders with the same split. The result was
statistically significant and pointed out that a fair split of referral bonus was the most successful incentive
scheme for friends with large social distance. Therefore, we further support Hypothesis 2.
Additionally, using two-way ANOVA, we tested the interaction effect of social distance and bonus split on
responder adoption, and detected a significant effect (F=4.3, p<0.01).
Randomized Field Experiment
According to the lab experimental study, the effect of split of bonus was impacted by social distance
between proposers and responders in online social referral systems. Aiming at achieving external validity
and gaining more insights, we corroborate our lab experiment with a randomized field experiment. We
collaborated with 08ticket (http://www.08piao.com/), one of the world’s largest online ticketing
merchants to conduct the study. This online merchant is a company whose major business is online
ticketing locating in various provinces of China. Multiple types of tickets are sold on the 08ticket website,
with the major type being singers’ solo vocal concerts tickets and scenic spot tickets. With the steady
development of online ticketing business, they have recently extended their business to include mobile
commerce. The company offered us access to their customer data, and assisted the randomized field
experiment process.
Design and Process of the Randomized Field Experiment
Based on prior literatures, 30% is usually a cutting threshold at which offers are accepted by the responder
in an ultimatum game (Güth et al. 1982; Güth et al. 1990). If responders are only able to obtain less than
30% of the whole amount, most responders tend to refuse offers as a way of punishing proposers’ greed.
So we changed our split of bonus from (10,0), (5,5) and (0,10) in our experimental study into (7,3), (5,5)
and (3,7) in our field study. We had three different distributional splits: (7,3), for which proposers will
receive $7 and responders will receive $3; (5,5), for which proposers and responders will receive $5 each
(fair split); (3,7), for which proposers will receive $3 and responders will receive $7. Proposers and
responders would not get the bonus until a responder accepts the referral and makes a purchase. We also
set a control group as (0,0), for which proposer and responder will receive nothing.
To highlight the impact of social distance, we used two levels of social distance (large and small). Large
social distance means the friends only have contact information or the friends are similar as the
workmates only communicate for business. Small social distance means the friends have a familial
relationship or intimate relationship with each other. Our social distance was based on social distance of
Bogardus (Bogardus 1933). So our field study had eight treatment and control groups (Table 2).
Table 3. Factorial Design of Field Study
(7,3)
Large social distance
(5,5)
Large social distance
(3,7)
Large social distance
(0,0)
Large social distance
(7,3)
Small social distance
(5,5)
Small social distance
(3,7)
Small social distance
(0,0)
Small social distance
The process of the field study was as follows:
(1) We randomly selected current customers from the online ticketing company (08ticket) as proposers.
(2) On behalf of the online ticketing company, we randomly assigned a current customer to one of our
eight treatment or control groups (the proposers) by sending an email including explanations of the
company’s online social referral incentive system and the intended split of bonus given a certain social
distance.
Documents you may be interested
Documents you may be interested