You are on page 1of 9

Social Networks 33 (2011) 143–151

Contents lists available at ScienceDirect

Social Networks
journal homepage:

Resource characteristics in social exchange networks:

Implications for positional advantage
David R. Schaefer ∗
School of Social and Family Dynamics, Arizona State University, Box 873701, Tempe, AZ 85287-3701, United States

a r t i c l e i n f o a b s t r a c t

Keywords: Despite the wide range of resources that traverse social networks, social exchange research has focused
Social exchange on only a narrow subset. Notably, prior social exchange research has not considered resources like infor-
Resources mation that have the capacity to diffuse through networks. The current study investigates how differences
between the standard social exchange resource and an information-type resource affect the advantage
provided by one’s network position. Results of a laboratory experiment support predictions and offer two
new insights to the foundations of positional advantage: (1) the location of advantageous positions in a
network differs by resource characteristics, and (2) only in particular situations is a single position able
to experience high levels of both power and exchange frequency.
© 2010 Elsevier B.V. All rights reserved.

1. Introduction on an experimental protocol that does not allow resources to

move through the network (see van Assen (2003) for more than
The social network and social exchange traditions share a 70 such studies). In fact, no prior social exchange research has
concern with understanding how network structures impart com- tested how power emerges when actors exchange resources that
petitive advantages or other forms of benefit to actors in particular have the capacity to diffuse through the network like informa-
positions. Notwithstanding their similar orienting frameworks, tion.
research in the two traditions has largely progressed indepen- One cannot assume that theories developed and tested using
dently. Despite an awareness of the fields’ overlap and early one type of resource hold when other resources are considered. As
investigations of key differences (Cook et al., 1983; Marsden, 1982), social network research has found, differences in resources (e.g.,
with few exceptions (e.g., Burt, 1992), recent research has adopted the content of ties) are associated with differences in network
one perspective or the other. Calls to integrate the two traditions structure (Cross et al., 2001; Lazega and Pattison, 1999; Wellman
have been made (Bonacich, 2009; Cook and Whitmeyer, 1992), but and Wortley, 1990) and the type of network positions that are
the areas have remained separate, so much so that in “mapping” advantaged (Burt, 1997; Dekker et al., 2003; Flap and Volker, 2001;
the field of social network analysis Borgatti placed social exchange Podolny and Baron, 1997; Stokman, 2004). Thus, it is unknown how
research on its own island, isolated from social network research well social exchange theories capture outcomes from the exchange
by a wide gulf (2008). of information-type resources. Broadening the set of resources con-
A major barrier to integrating the two traditions is the form of sidered within social exchange theory is essential to extend the
resources considered. In examining which actors receive resources theory’s explanatory power and increase its relevance to social
the quickest, most directly, or most abundantly, social network network research.
research often presumes there is something “flowing” through The current research considers how networks produce advan-
the network (Borgatti, 2005). In many cases information is the tage for two resource forms: (1) the standard social exchange
resource of interest and its diffusion is either the driving ques- resource, which is constrained to exchange in a single dyad, and
tion or a key theoretical mechanism (Burt, 1992; Granovetter, (2) a resource resembling information that is able to diffuse across
1973). Social exchange theory focuses on the related question the network as it is exchanged. I explain how differences between
of how resource exchanges give some actors power over others. the two resources alter power use and the frequency of actors’
However, social exchange research has focused almost exclusively agreements, with consequences for which positions in the network
are advantaged. Results of a laboratory experiment reveal how
differences in resource characteristics alone produce completely
∗ Tel.: +1 480 727 8332. different distributions of power and exchange frequency across a
E-mail address: network.

0378-8733/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
144 D.R. Schaefer / Social Networks 33 (2011) 143–151

2. Theoretical background A
2.1. Resources
Social network research has acknowledged the diverse nature
of resources and found important differences by type. Given the
same set of actors, networks of information flow, advice, friendship,
and support all display different patterns (Cross et al., 2001; Lazega
and Pattison, 1999; Stokman, 2004; Wellman and Wortley, 1990).
Even within the same structure, resources can move through the B
network along different routes (Borgatti, 2005). Most importantly,
the content of ties is a key factor determining individual outcomes, Fig. 1. Study network.
including which positions in a network structure are optimal. For
instance, supportive ties provide greater benefits when they are
Markovsky et al., 1988). Power use is evidenced by inequality in a
densely connected, while instrumental ties are more rewarding
single exchange or cumulatively across multiple exchange oppor-
when they bridge distinct clusters (Borgatti, 2005; Burt, 1997;
tunities. Exchange frequency reflects how often actors are able to
Dekker et al., 2003; Flap and Volker, 2001; Podolny and Baron,
acquire the valued resources they desire from other actors, regard-
1997; Stokman, 2004).
less of the terms of exchange. The benefits from exchange frequency
Social exchange theory intentionally leaves the nature of
are distinct from dyadic power. For instance, actors can benefit
resources unspecified. Theoretically, resources are any poten-
equally through a high volume of exchange on poor terms or a lower
tial behavior by one actor that would act as reinforcement or
volume of exchange on better terms.
reward for another actor (Emerson, 1972). Empirically, how-
With the standard exchange resource power and exchange fre-
ever, social exchange research has adopted a standard protocol
quency are highly correlated. Thus, prior exchange research has
in which resources are constrained to be non-duplicable and
tended to focus exclusively on advantage in the form of power.
non-transferable (Schaefer, 2007). Resources received from others
However, as discussed below, with information-type resources,
cannot be exchanged further by their recipient (non-transferable)
power and exchange frequency are decoupled, allowing actors to
and their original provider cannot give the same resource to
experience one or the other, but not necessarily both. This decou-
multiple recipients (non-duplicable). Operationally, the experi-
pling necessitates the consideration of both outcomes in order to
menter replenishes the points available for exchange before each
gain a more complete picture of network-conferred advantage.
opportunity. Since only newly available points can be exchanged,
not those previously received, resources are effectively non-
transferable. And, resources cannot be duplicated to facilitate
multiple exchanges because a one-exchange rule constrains sub- 3. Implications of resource type
jects to exchange with only one partner (Willer, 1999).1 In essence,
resources are “used up” during exchange, such that neither the I adopt the following scope conditions of social exchange the-
provider nor the recipient can use the resource in future exchanges. ory: actors control resources that others value; actors pursue
As a resource, information has the capacity to diffuse through the resources they value through exchange; and, actors exchange
a network – a capacity that prior social exchange research has not repeatedly with the same partners (Molm and Cook, 1995). I assume
allowed. Information departs from the standard exchange resource that actors control unique resources and, in line with prior research,
in that both its provider and its recipient control the informa- exchange occurs through a negotiation procedure.3 Thus, actors
tion following exchange. The key conceptual distinction is that can seek all resources available in the network and the acquisi-
the information-type resource is both duplicable and transferable tion of one does not affect the value of others. In addition, I assume
(Schaefer, 2007), which allows either actor to transmit it to some- resources are replenished regularly, but not after every exchange,
one else in a future exchange. The standard exchange resource is which allows information-type resources to diffuse. In discussing
neither duplicable nor transferable and thus cannot leave the dyad. different types of resources, I identify aspects of network structure
The current research investigates how exchange outcomes differ that produce advantages in power and exchange frequency. The
between the standard social exchange resource as described above logic presented below applies to all networks; however, I study
and an “information-type” resource. I assume all other resource Fig. 1 network because it is the smallest network exhibiting the
dimensions remain constant.2 features necessary to observe effects of resource differences on

2.2. Positional advantage

In examining the implications of resource characteristics, I focus 3

Unique resources is the simplest assumption one can make (other than identical
on two types of positional advantage: the use of power to receive resources, which would preclude the need for exchange). Unique resources also
resources on more favorable terms and the frequency with which avoids potentially arbitrary decisions about how resources are distributed, which
an actor receives valued resources through exchange. Dyadic power are likely to affect the emergence of power.
Unlike much exchange research, I do not attempt to make precise point
is an actor’s ability to obtain resources at a favorable rate – to get
predictions regarding actors’ payoffs. The goal of this research is to demon-
more at lower cost relative to the partner (Cook and Emerson, 1978; strate the importance of resources for understanding positional advantage. When
information-type resources are considered, the exchange process becomes signifi-
cantly more complex than with the standard exchange resource. Unfortunately, it
goes beyond the scope of this paper to also develop measures that can predict posi-
Prior social exchange research has occasionally departed from the standard tional advantage in any network. Indeed, after 30 years, social exchange researchers
resource as a means to manipulate other aspects of exchange (e.g., Markovsky et al., are still refining measures of power in the relatively simpler situation of the standard
1988; Yamagishi et al., 1988; Willer, 2003), but such research has not focused on exchange resource. Rather, I focus on the logically prior question of where power
resources explicitly and has not considered information-type resources. imbalances will emerge and who will exchange more often under a broader set of
I assume resources are indivisible and have a stable value, and that actors have resource specifications. I identify straightforward principles that can inform future
full information about their value. development of formal measures of positional advantage.
D.R. Schaefer / Social Networks 33 (2011) 143–151 145

3.1. Standard exchange resource of resources is conditional upon providing a different resource in
return. Unfortunately, no ideal measure exists to predict central-
The standard social exchange resource emerged as a means to ity with resources such as information. As Borgatti (2005) argues,
operationalize negative connections between dyads. Two dyads are centrality measures make implicit assumptions about resource
connected when they share a common actor. The connection is movement, none of which correspond with how information
defined as negative when exchange in one dyad inhibits exchange diffuses across a network. For example, betweenness centrality
in the other (Emerson, 1972). Such is the case, for example, when (Freeman, 1979) is calculated using geodesics; but, information is
a seller has two bidders and can only exchange with one of them. not constrained to travel along the shortest path(s) between actors.
The type of connection between two dyads is important because Given the lack of prior research on how information-type resources
it determines how structural power emerges. In negative connec- diffuse through an exchange network, a better default assumption
tions, power imbalances arise due to exclusion (Markovsky et al., is that resource movement is random. Accordingly, I use ran-
1988). Actors are excluded when they have no one to exchange with dom walk betweenness to predict which positions will receive
because their partner(s) exchanged elsewhere. Excluded actors information-type resources more often. Random walk between-
increase their future offers in an effort to be included (Cook et al., ness is medial centrality measure that identifies positions as central
1983; Thye et al., 1997). Thus, powerful positions are those that to the extent they are more often located on random walks between
are least often excluded from exchange (Cook and Emerson, 1978; other actors (Newman, 2005). Within the study network, I expect
Cook et al., 1983; Markovsky et al., 1988). For example, in the sim- actors to be advantaged by exchange frequency when they have
ple A–B–C network, because B can only exchange with one partner, greater random walk betweenness. C will be most advantaged
either A or C will be excluded. A and C will make future concessions (bRW = .83), followed by D (bRW = .7), A and B (bRW = .5), with E expe-
to B in order to be included in exchange, giving B a power advan- riencing the lowest exchange frequency (bRW = .4).
tage. Importantly, for exclusion to produce power imbalances it Power can also arise when actors exchange information-type
must consistently advantage one actor over another. For instance, resources; however the process differs in two ways from the stan-
if A and C could also exchange with one another, then none of the dard exchange resource: (1) exclusion can operate, though in a
three actors could consistently exclude another. Exclusion would different manner than with the standard exchange resource, and
occur on any single exchange, but no structural source of power (2) an additional mechanism – ordering – can lead to power imbal-
imbalance exists to concentrate advantage in one position. ances. Information-type resources also differ from the standard
Positions with access to alternative sources of valued resources, resource in that the conditions necessary for power do not exist
or better alternatives than adjacent positions, are advantaged by initially. That is, every actor can initially exchange with all possi-
exclusion. Because exchange research has focused extensively on ble partners. The conditions necessary for power only emerge as
power in networks with negatively connected dyads, several algo- the distribution of resources changes following exchange. Social
rithms have been developed to predict power use with the standard exchange researchers have not systematically studied the endoge-
resource (Bienenstock and Bonacich, 1992; Bonacich, 1987; Cook nous emergence of power, making it necessary to first identify
and Emerson, 1978; Friedkin, 1992; Markovsky et al., 1988, 1993; the aspects of network structure that produce the preconditions
Yamaguchi, 1996). While differing in the level of power use they for structural power. This is complicated by the fact that with
predict, they tend to agree on which position has greater structural information-type resources, the emergence of power is contingent
power in any dyad (van Assen, 2003; Willer and Emanuelson, 2008). upon the sequence of exchanges across dyads. While it is impossible
One example is the graph-theoretic power index (GPI3 ), which was to predict whether the conditions necessary for power will emerge
developed to predict dyadic power by estimating the probability during a single opportunity, one may be able to predict where they
of positions participating in exchange (Markovsky et al., 1993).5 will emerge more often on average.
Essentially, the algorithm estimates how often exchange would
occur in each dyad if actors chose their partners randomly. Thus, 3.2.1. Power through exclusion
GPI3 has the added utility of being able to estimate exchange fre- The negative connections necessary for exclusion do not exist
quency for each position. The GPI3 values for the study network initially. Actors are not limited in the number of times they can
(Fig. 1) are: A = B = C = .74, E = .77, D = 1.0, suggesting that D will exchange information-type resources under their control; thus, in
complete the most exchanges, followed by A, B, C, and E relatively the study network, D can share the same resource with C in one
equally. In terms of power, positions with the greatest likelihood exchange and E in another. Rather, the capacity for information-
of participating in exchange have a power advantage over neigh- type resources to be “copied” during exchange allows alternative
boring positions who are more likely to be excluded (Lovaglia et al., sources to emerge over time. Negative connections appear as
1995; Markovsky et al., 1988). Comparing GPI3 scores for adjacent resources diffuse across the network to the point where an actor’s
positions reveals that D will develop a power advantage over C and partners control the same resource. For example, referring to Fig. 1,
E because D will never be excluded but will occasionally exclude C should A and B exchange information-type resources with each
and E. All other dyads are expected to be equal power because no other, then they come to share the same resources, making the A–C
position can consistently exclude another. and B–C dyads negatively connected. Were this pattern of exchange
to occur consistently, C would develop a power advantage over A
and B. C need only exchange with A or B to obtain all their resources,
3.2. Information-type resources
which allows C to completely exclude the other.
The negative connections that lead to exclusion arise in net-
Information-type resources differ from the standard resource in
works that contain cycles, such as the A–B–C subgraph of Fig. 1.
that they have the capacity to diffuse across a network. Social net-
When ego is located on a cycle then ego’s partners on the cycle can
work theory suggests that the capacity to acquire resources such
acquire each others’ resources without ego as an intermediary. For
as information is a function of medial centrality, or being located
instance, A and B can exchange and receive each other’s resource
between other actors (Borgatti and Everett, 2006). I expect this is
without going through C. When this occurs, a negative connection
also the case within an exchange framework, where the receipt
can develop between ego’s dyads on the cycle (i.e., the A–C and
B–C dyads). Only a cycle can give an actor alternative sources of a
resource that is valued, making it a necessary structural precondi-
Other measures make identical predictions for the study network. tion for the emergence of power through exclusion.
146 D.R. Schaefer / Social Networks 33 (2011) 143–151

One additional condition must be met for cycles to produce 3.3. Simultaneous advantage
exclusion. Because negative connections only lead to power use
when one actor is consistently able to exclude another, location With the standard exchange resource an actor’s exchange fre-
on a cycle produces power only when some actors have better quency is correlated with power advantage. Actors who avoid
alternatives to exchanging on the cycle than others. Actors with exclusion exchange more often and have a power advantage rel-
access to more resources outside a cycle will initiate exchange on ative to partners who exchange less often (Markovsky et al.,
the cycle less often, leaving actors without off-cycle partners to 1988). However, with information-type resources, power use and
exchange with someone on the cycle. For example, in the study net- exchange frequency are decoupled because different structural
work, chance alone suggests that A and B are more likely to initiate conditions produce each form of advantage. Consequently, posi-
exchange with one another than with C because C will occasion- tions may be advantaged for one outcome but not the other.
ally initiate exchange with D. Thus, A–B exchange will occur most With information-type resources there are two mechanisms
consistently and create a negative connection between the A–C and that can produce power. Actors experience a power advantage from
B–C dyads that advantages C. The A–C and B–C dyads are the only exclusion when they are located on a cycle and have better alterna-
dyads where exclusion will consistently arise and favor one actor tives off the cycle than their partners. Such positions can be more
(C). In general, actors with better alternatives off the cycle will have central in the network than their disadvantaged partners’ posi-
a power advantage over partners on the cycle who have worse or tions, which would create a positive correlation between power and
no off-cycle alternatives. exchange frequency. In the study network, C has high medial cen-
trality, which provides high exchange frequency, and is on a cycle
that confers power through exclusion. Thus, C could experience
3.2.2. Power through ordering both forms of advantage.
Information-type resources allow power to arise through an The same cannot be said for power through ordering. Order-
additional mechanism: ordering. Ordering occurs within the con- ing advantages actors who are in positions whereby others must
text of dyads that are positively connected, where exchange in exchange with them to facilitate their own subsequent exchanges.
one dyad is positively associated with exchange in another dyad Such power advantaged positions can be central to the network and
(Emerson, 1972). For example, consider the positively connected provide high exchange frequency (i.e., D). Alternatively, ordering
I–S–J network in which a student (S) covets a job (J), but must advantaged positions may be located in peripheral network posi-
first complete an internship (I) with another organization and tions that provide low exchange frequency, such as position E. Thus,
receive a letter of recommendation. The I–S and S–J dyads are pos- power advantages through ordering are not necessarily correlated
itively connected because exchange in I–S facilitates exchange in with exchange frequency.
Ordering leads to power imbalances when an actor must
4. Method
exchange in one dyad before exchanging in another (Corra and
Willer, 2002). This is a special case of positive connections, where
To test the effects of resource characteristics on each form
one exchange must temporally precede the other. In the I–S–J net-
of advantage, I devised a laboratory experiment with two condi-
work, the S–J exchange is contingent upon a prior I–S exchange.
tions corresponding to the standard social exchange resource and
This exchange sequence gives a power advantage to I, who pro-
an information-type resource. Participants were undergraduate
vides S with the resource necessary to enter the second exchange
students recruited based on their desire to earn money. Partici-
(Corra and Willer, 2002). The student is dependent upon her super-
pants were randomly assigned to one of the positions shown in
visor and will work hard to meet or exceed expectations, giving the
Fig. 1, where they remained throughout the experiment. Partici-
supervisor greater power.
pants were given detailed instructions on their computer screen,
Because previous research has imposed ordering on a network
followed by practice rounds and the experimental exchange phase.
(Corra and Willer, 2002; Schaefer and Kornienko, 2009), measures
All interactions between participants took place via computer using
to predict where ordering will occur do not exist. In the cur-
a program developed in z-Tree (Fischbacher, 1999). Participants
rent context, power through ordering will emerge whenever ego
had information about their potential partners, which they referred
desires exchange with alter, but ego does not have a resource that
to using a letter, but not other participants or their partners.6 Each
alter values. Ego must then exchange in another dyad to obtain a
experimental condition included 10 cases with five participants
resource that alter values. For example, returning to the study net-
work, when C and D exchange, they acquire each other’s resources
and cannot exchange again. Were C to acquire a resource from
A or B, then D would become interested in exchanging with C. 4.1. Resources
But, D would not have a resource that C valued. D would need
to acquire a resource from E to enable exchange with C. Such To help distinguish between exchange frequency and power
a situation gives E an ordering advantage over D. This type of use, I operationalized resources separately from the payoff from
situation can arise for any position that has at least two part- exchange. Resources were operationalized using indivisible tokens.
ners (anyone but E in the study network). Anyone with multiple Participants were endowed with a single token identified by a
partners may occasionally need to exchange with one partner to unique letter. Exchange between participants in adjacent positions
facilitate exchange with another partner. To the extent ordering was possible if each controlled a token the other had not previ-
consistently favors one position over another, power imbalances ously received. During exchange, participants’ control over tokens
are expected to arise. Given that information-type resources have changed depending upon the experimental condition. With the
not been previously studied within an exchange framework, it is standard exchange resource, tokens were used up during exchange.
an open question what sequences of exchange will transpire and In the information-type condition, participants gained control of
whether ordering will arise in dyads other than D–E. However, their partner’s token during exchange while retaining control of
computer simulations suggest that ordering will appear most often
in the D–E dyad, where it advantages E, and slightly more often
than random in the C–D dyad, where D is advantaged (Schaefer, 6
This reflects the situation in the natural-world where actors are often unaware
2007). of their partner’s other ties.
D.R. Schaefer / Social Networks 33 (2011) 143–151 147

their own token. In all conditions, participants’ computer dis- 4.3. Exchange outcomes
plays indicated any tokens they had previously received and which
tokens were currently controlled by themselves and their part- Exchange frequency was calculated as the average number of
ners. exchanges per period (the total number of completed exchanges
I operationalized the payoff from exchange using points that divided by the number of exchange periods). To measure power at
were converted into money after the experiment. Any time two the dyadic level, I calculated the proportion of points exchanged in
participants could exchange tokens they were given 24 points to a dyad that actors in each position received. For example, A’s power
divide. Participants did not know they were dividing 24 points; relative to B was:
rather, they were able to request 1–24 points from one another. PointsAB
Requests were then transformed into an offer to the partner. When PowerAB =
PointsAB + PointsBA
participants reached an agreement on how to split the points, each
received their share of the points and their control over the tokens PointsAB is the total number of points A received from B and
changed according to their experimental condition. Payoffs were PointsBA is B’s points from A. Since A’s power over B is directly
based solely on the points earned during exchange, not the con- related to B’s power over A (PowerAB = 1 − PowerBA ) power ratios
trol of tokens (which only determined who could exchange with are reported once for each dyad. Power-balanced dyads are indi-
whom). cated by power use equal to .5. To facilitate the comparison of
For example, when two participants exchanged the standard exchange frequency and power as forms of advantage, I also
resource, each lost control of their own token which prevented measure power for each position. Positional power was cal-
them from exchanging with anyone else. Participants did not culated by averaging a position’s power across all dyads (i.e.,
acquire their partner’s token because the tokens were “used up” PowerA = (PowerAB + PowerAC )/2). Because power use takes time to
during exchange. Rather, each participant received the points develop in a network (Cook and Emerson, 1978), the measures of
agreed upon during exchange. By contrast, with the information- power in the following analyses were based on the latter half of
type resource, participants retained control of their own token and exchange periods.
gained control of their partner’s token (e.g., a copy was made). To measure whether each power mechanism created the struc-
This allowed each participant to use either their original token or tural conditions for power use, I calculated how often exclusion
their partner’s token in future exchanges. In addition, participants and ordering appeared in a dyad and which actor was advantaged.
received the agreed-upon points. Exclusion occurred when two actors could have exchanged but did
not because ego exchanged elsewhere, leaving alter without an
4.2. Structure of exchange exchange partner. For each dyad, I first calculated the proportion
of periods in which each actor excluded the other. Then, because
The exchange phase of the experiment contained several peri- only imbalances in exclusion create power imbalances, I computed
ods, each with multiple rounds of negotiation. Participants knew the net difference in the proportion of periods the actors excluded
the number of rounds per period but not the total number one another (i.e., the absolute value of the difference between ego’s
of periods. At the beginning of each period participants were exclusion of alter and alter’s exclusion of ego). Exclusion was high
endowed with a token. Throughout the period, participants could in a dyad if one actor consistently excluded the other and low if two
exchange one or more times depending upon experimental condi- actors excluded one another at equal rates.
tion. Ordering occurred when ego’s acquisition of a resource made
Within a period, participants had a fixed number of rounds for exchange in another dyad possible. I coded ordering as occur-
negotiations and exchange. Negotiation consisted of a series of ring, and disadvantaging ego, when such situations arose and ego
offers and counteroffers over how to split the 24 points. Within completed the requisite exchange. This measure does not con-
a round, participants simultaneously requested points from each sider whether ego actually used the acquired resource in a second
possible partner while partners made requests from them. After exchange, because ordering can produce power regardless (so long
the round, participants learned what others had offered to them. as a second exchange occurs sometimes). As with exclusion, I first
If the requests within a dyad summed to 24 or less an exchange computed how often each actor in a dyad was advantaged over the
automatically took place, in which case participants received the other, which I then used to calculate the net difference in advantage
agreed-upon points and their computer screens noted which token within the dyad.
had been forfeited and/or received.7 Participants who did not reach
an agreement could accept a partner’s offer, repeat a previous offer, 5. Results
or submit a counteroffer in the next round.
After an exchange, another round would begin if any two adja- The following analysis proceeds in three steps. I begin by
cent participants controlled tokens that each other valued. If no examining whether the pattern of exchange frequency across posi-
more exchanges were possible or the maximum number of rounds tions differed by resource type. Second, I test whether power
had been reached then a new period began: participants were given emerged in different dyads depending upon resource type. I inves-
their initial token and the record of tokens received was emptied. tigate whether the mechanisms of power emerged in each dyad
To keep all conditions under 2 h, the number of rounds and periods as expected and led to a power imbalance in the dyad. Third,
varied across conditions.8 I test whether positions simultaneously experienced power and
exchange frequency advantages for each resource type.

5.1. Exchange frequency
When requests summed to less than 24 the remaining points went unclaimed.
On average, agreements yielded a summed payoff of 23.17 points. ANOVA results
indicate no differences in summed payoffs over time (four trial blocks), by condition, I begin by analyzing differences in exchange frequency by posi-
or by dyad. tion and resource type using a two-way ANOVA (with position
The standard resource condition had 80 periods with four rounds each while as a within-subjects factor). Results indicate significant differ-
the information-type resource condition had 32 periods of eight rounds each. To
give participants the opportunity to exchange resources to their furthest extent, the
ences by condition (F[1,18] = 361.5, p < .001) as expected. With the
information-type condition had more rounds, which necessitated fewer periods in standard exchange resource, participants exchanged .76 times per
order to keep the length of the experiment the same. period on average, compared to 2.20 times per period with the
148 D.R. Schaefer / Social Networks 33 (2011) 143–151

Table 1 power advantage produced by exclusion is evidenced by a corre-

Mean exchange frequency and power by position and resource type.
lation of .35 between exclusion and dyadic power (p = .009) across
Position Standard resource Information-type resource all dyads. Thus, as expected, D had a significant power advantage
over C and E. No other dyads exhibited exclusion that consistently
Exchange Positional Exchange Positional
frequency power frequency power favored one position over the other or imbalanced power.
With information-type resources, both exclusion and order-
A .62b (.15) .51b (.05) 2.05c (.32) .47b (.09)
B .71b (.15) .48b (.05) 2.25bc (.38) .47b (.04) ing have the potential to create imbalanced power. Exclusion
C .88a (.06) .50b (.03) 3.36a (.45) .55ab (.12) was expected to occur among actors on the A–B–C cycle, where
D .91a (.08) .61a (.10) 2.39b (.36) .45b (.11) exchange creates alternative sources for the same resource. A and
E .70b (.13) .39c (.11) .94d (.10) .60a (.11) B were expected to exchange more often and become alternatives
Note: Standard deviations in parentheses. Positions with different superscripts have to C because C could exchange off-cycle with D. As shown in Table 2,
different means based on paired samples t tests. C was consistently able to exclude positions A and B, while A and
B did not consistently exclude other actors. Ordering also occurred
in the network, though in different dyads than exclusion. Ordering
information-type resource. Differences by position existed over-
favored E over D, D over C, and A over C. Ordering occurred in the
all (F[4,72] = 96.1, p < .001) and by resource type, as indicated by the
other dyads, but did not consistently favor one actor over the other.
significant position × condition interaction (F[4,72] = 68.6, p < .001).
Power imbalances are expected to develop when either exclu-
As shown in Table 1, with the standard exchange resource, C and D
sion or ordering consistently advantages one actor over another.
exchanged most often, followed by the other positions. A series of
To test this relationship, dyadic power was regressed on exclu-
follow-up t tests reveal that C and D exchanged significantly more
sion and ordering. Results indicate that dyadic power was stronger
often than A, B, and E. This is largely consistent with predictions
when ego excluded alter (ˇ = .359, p = .009) or ego had an ordering
based on theGPI3 , with the exception that C exchanged more often
advantage over alter (ˇ = .286, p = .036). Consequently, as shown in
than expected.
Table 2, power imbalances were observed in the dyads that dis-
A different pattern of exchange was observed with the
played the strongest exclusion (A–C and B–C) and the strongest
information-type resource. Follow-up t tests reveal significant dif-
ordering (D–E).
ferences between all positions except A and B. The most central
In only one dyad was a power mechanism not positively asso-
position, C, exchanged more often than D, who exchanged more
ciated with a power advantage. D’s ordering advantage over C did
often than A and B. The most peripheral position E experienced
not translate into power over C. One possibility is that ordering
the fewest exchanges of all. The rank ordering of exchange fre-
did not occur frequently enough to elicit a power imbalance. It
quency across positions is consistent with expectations based on
is likely there is a threshold, whereby a power mechanism must
the random-walk betweenness of positions.
exist a minimum amount of time to have consequences for actors’
behavior. Unfortunately, no prior research has examined the effect
5.2. Power use
of variability in the presence of power mechanisms on the emer-
gence of power use. Another possibility is that C’s and D’s other
I next test how the pattern of power use in the network differed
dyads affected their behavior in the C–D dyad. Recent research has
by resource type (see Table 1). I begin with a two-way ANOVA,
shown that the value of external dyads affects negotiation behav-
with position as a within-subjects factor. Average power was nearly
ior in a focal dyad (Schaefer and Kornienko, 2010). C was power
equal with either resource type (.50 with the standard resource
advantaged in its two external ties to A and B, which can raise
and .51 with the information-type resource). Although the main
C’s expectations regarding exchange outcomes. In contrast, D was
effect of position was not significant (F[4,72] = 1.2, p = .33), results
disadvantaged in its external tie to D, which can lower D’s expecta-
indicate a significant position × condition interaction (F[4,72] = 9.9,
tions. In combination, this has the potential to offset D’s potential
p < .001). Follow-up t tests indicate that with the standard resource,
for power advantage through ordering.
D was the most powerful position in the network, but with the
information-type resource E was the most powerful.
Further insight is gained by examining the occurrence of each 5.3. Simultaneous advantage
power mechanism and ensuing results for power. With the stan-
dard exchange resource, the only power mechanism that can The final question is whether a position’s ability to experience
operate is exclusion. Exclusion was expected to occur in the C–D advantage in the form of power and exchange frequency differs by
and D–E dyads, and advantage D. As shown in Table 2, the C–D resource type. To address this, I compare how positions fared rela-
and D–E dyads displayed exclusion that consistently favored D. The tive to one another on each outcome and with each resource type.

Table 2
Exclusion, ordering, and dyadic power by dyad and resource type.

Dyada Standard resource Information-type resource

Exclusion Dyadic power Exclusion Ordering Dyadic power

A/B −3.90 (11.14) .53 (.06) .20 (.79) .20 (1.48) .50 (.06)
C/A 3.40 (8.28) .52 (.08) 8.30** (6.67) 1.80** (1.40) .57† (.12)
C/B −.50 (9.29) .52 (.04) 6.70** (6.43) .90 (1.60) .58* (.12)
C/D −22.60*** (6.04) .44** (.07) – −4.00*** (1.94) .52 (.13)
E/D −8.30*** (4.32) .39** (.11) – 6.60** (4.22) .60** (.11)

Note: Standard deviations in parentheses. t tests were used to compare the balance of power mechanism in each dyad to 0, representing the absence of a consistent advantage,
and dyadic power to .5, representing balanced power.
p < .05.
p < .01.
p < .001.

p < .10.
The means for each ego/alter dyad correspond to ego’s advantage relative to alter.
D.R. Schaefer / Social Networks 33 (2011) 143–151 149

Table 3 Whereas social network research often considers resources that

Average z scores for exchange frequency and power by position and resource type.
move across a network (Borgatti, 2005), social exchange research
Standard resource Information-type resource has rarely considered resources that can be transmitted between
dyads, and never considered resources like information that have
Position Exchange Positional Exchange Positional
frequency power frequency power the capacity to diffuse across a network. To help bridge this divide
I examined an information-type resource from within a social
A −.92 .17 −.17 −.32
B −.31 −.19 .06 −.38 exchange framework. I contrasted the standard resource used in
C .70 .02 1.37 .37 social exchange research with a resource embodying the charac-
D .91 1.09 .22 −.52 teristics of information. Differences between resources affect how
E −.39 −1.08 −1.49 .85 they move through a network, whether they can be used in multiple
Note: Scores are standardized within condition. exchanges, and their availability to others. Results from a laboratory
experiment demonstrate the multiple ways that resources affect
which network positions are advantaged.
To facilitate comparisons, I standardized each measure of advan-
Going beyond the standard exchange resource revealed a decou-
tage for the resource type (see Table 3; scores are standardized
pling of power use and exchange frequency that enabled an
within columns).
examination of their association across network positions. This
I first examine whether positions experienced the same form of
leads to the first major implication of this study: only in certain
advantage for each resource type. For exchange frequency, the same
circumstances does a position provide advantages in both power and
positions are advantaged with either resource type, though the rel-
exchange frequency. When the same mechanism produces power
ative advantage of positions C and D switched between resource
and exchange frequency, a single position can experience a simul-
types. For power, however, the relative advantage of positions
taneous advantage. This is the case with the standard exchange
differed more dramatically between the two resource forms. The
resource, where those who avoid exclusion have a power advantage
biggest difference is that with the standard exchange resource, D
and greater exchange frequency. However, with information-type
was most power advantaged and E was least advantaged. However,
resources, power and exchange frequency emerge independently
with information-type resources, D was least power advantaged
and mechanisms can favor different positions for each form of
and E was most advantaged. This is because the power mechanisms
advantage. Consequently, power-advantaged positions may or may
not only differ by resource type, but emerge in different dyads in
not also exchange most often. Only in select circumstances do
the network. While D can exclude E with the standard resource, D is
resources and network structure concentrate power and exchange
subject to an ordering disadvantage to E with the information-type
frequency advantages in the same position.
This research follows Burt’s call for more research on “trans-
Next, I compare whether the same positions were advantaged
mission mechanisms responsible for the competitive advantage
on power and exchange frequency within each resource type.
of network brokerage,” (2008, p. 966) and offers insight to the
With the standard exchange resource, the same position was most
debate over the advantages provided by structural holes (Reagans
advantaged on both outcomes. D was able to exchange most often
and Zuckerman, 2008). In the study network, C and D occupy
and enjoyed the most power. This was expected because the same
structural holes between disconnected alters. Results of the exper-
principle – avoiding exclusion – is ultimately responsible for both
iment indicate that only in limited circumstances were C and D
outcomes. By contrast, with information-type resources different
advantaged in terms of power and exchange frequency. With the
positions were advantaged for each outcome. C exchanged most
standard exchange resource, D experienced both forms of advan-
often, while E was the most powerful. C’s position provides the
tage; however, with the information-type resource, D was power
medial centrality necessary to receive a variety of resources trav-
disadvantaged. C’s position provided consistently high exchange
eling across the network and C was power advantaged in some
frequency, but C was not as powerful as other positions in the net-
dyads, but not all. By contrast, E was the most power advan-
work with either resource form. Still, C was able to avoid other
taged because E was located in a positions that experienced
actor’s power use, thereby experiencing the autonomy, rather than
power through ordering. Although E’s peripheral position allowed
control, that Burt’s more recent work has emphasized (Burt, 2005).
E to experience power, the position did not allow E to exchange
Thus, structural holes are better positions than average, but no
often, making E the most disadvantaged on exchange frequency.
guarantee for maximizing power and exchange frequency for all
These results indicate that with information-type resources power
resource types.
and exchange frequency are not commensurate forms of advan-
The second major implication of this research is that: no sin-
gle network position is necessarily advantaged in either power use
Lastly, it is informative to look across the columns in Table 3
or exchange frequency for all resource types. Altering resource char-
to observe how each position’s outcomes vary depending upon
acteristics completely changed the outcomes experienced by all
the resource exchanged and the form of advantage considered. For
actors in the network. Because the mechanism(s) producing each
all positions but C, z scores are both positive and negative. This
form of advantage differ as resource characteristics change, the dis-
indicates that each position’s relative standing varied depending
tribution of advantage across the network changes. Thus, I suggest
upon the resource type and form of advantage. Even C, who was
it is unlikely for a position’s advantage to be robust across resource
never disadvantaged, was only advantaged under particular cir-
cumstances. In sum, the ability of a position to offer advantage is
This research helps explain findings that positional advantage
highly contingent upon the resources actors exchange and the form
depends upon relational content. Underlying differences between
of advantage considered.
resources help explain why brokerage positions are more valu-
able when network content consists of impersonal information
6. Discussion rather than socio-emotional resources (e.g., Flap and Volker, 2001;

This research identified resources as a previously unstudied

exchange element with implications for the advantages provided 9
Some positions may be more robust than others. For instance the central position
by network position. Differences in resources mark a key barrier in a star network will always experience the greatest resource breadth, but will only
to the integration of social network and social exchange theories. have a power advantage with non-duplicable, non-transferable resources.
150 D.R. Schaefer / Social Networks 33 (2011) 143–151

Podolny and Baron, 1997) or explicit rather than tacit knowledge Acknowledgments
(e.g., Dekker et al., 2003). Brokerage positions are valuable for
acquiring transferable resources like information or explicit knowl- This research was supported by the National Science Foundation
edge, but offer no advantage for non-transferable resources like (#SES-0425431) and the Social and Behavioral Science Research
socio-emotional resources and tacit knowledge, which must be Institute at the University of Arizona. Special thanks go to Linda
acquired directly from their source. Molm, Miller McPherson, Ronald Breiger, Joeseph Galaskiewicz,
Investigating how social exchange research can better repre- and the members of the Social Psychology Discussion Group for
sent phenomenon of interest to social network researchers paves their valuable comments. I gratefully acknowledge the research
the way for studies of outcomes beyond power and exchange assistance provided by Rebecca Arnold, Ryen Hanna, and Seth
frequency. Recent social exchange research has considered many Wright.
outcomes of interest to social network researchers, including the
development of trust and cohesion in dyads (Lawler et al., 2008;
Molm et al., 2009) as well as the stability of networks (Doğan References
et al., 2009). Social network theory only stands to benefit by bet-
ter access to social exchange theories detailing when cohesion will Bienenstock, E.J., Bonacich, P., 1992. The core as a solution to exclusionary networks.
be isolated to “pockets” within a network (Lawler, 2002), when Social Networks 14, 231–243.
Bonacich, P., 1987. Power and centrality: a family of measures. American Journal of
instrumental relationships evolve into affective ties (Molm et al., Sociology 92, 1170–1182.
2007), and how a network-induced perception of control affects Bonacich, P., 2009. Using social networks for evil. In: Keynote Address at SUNBELT
commitment to a partner (Lawler et al., 2006). Understanding XXIX International Network for Social Network Analysis Conference. San Diego,
how the processes behind these outcomes differ under a wider Borgatti, S.P., 2005. Centrality and network flow. Social Networks 27, 55–71.
range of resource conditions that more closely represent those Borgatti, S.P., 2008. Network reasoning. In: Keynote Address at SUNBELT XXVIII
found in social network research is a useful next step (cf. Schaefer, International Network for Social Network Analysis Conference. St. Pete’s Beach,
2009). Borgatti, S.P., Everett, M.G., 2006. A graph-theoretic perspective on centrality. Social
Given this research is a first step in understanding the impor- Networks 28, 466–484.
tance of resources for positional advantage, several limitations and Burt, R.S, 1992. Structural Holes: The Social Structure of Competition. Harvard Uni-
versity Press, Cambridge.
future directions remain. To begin, although this research extended
Burt, R.S., 1997. A note on social capital and network content. Social Networks 19,
the nature of resources beyond that found in social exchange 355–373.
theory, resources were still oversimplified. It would be useful to Burt, R.S., 2005. Brokerage and Closure: An Introduction to Social Capital. Oxford
University Press, Oxford.
expand this research by systematically manipulating other facets
Burt, R.S., 2008. Information and structural holes: comment on Reagans and Zuck-
of resources and their distribution. For example, the location of erman. Industrial and Corporate Change 17, 953–969.
power advantaged positions may be different if the initial distribu- Cook, K.S., Emerson, R.M., 1978. Power, equity and commitment in exchange net-
tion of resources provides alternatives for some actors (Marsden, works. American Sociological Review 43, 721–739.
Cook, K.S., Emerson, R.M., Gillmore, M.R., Yamagishi, T., 1983. The distribution of
1983), if resources are complementary (Yamaguchi, 1996), or if power in exchange networks: theory and experimental results. American Jour-
actors can receive commissions during exchange (Marsden, 1982). nal of Sociology 89, 275–305.
Results may also differ depending upon whether resources depre- Cook, K.S., Whitmeyer, J.M., 1992. Two approaches to social structure:
exchange theory and network analysis. Annual Review of Sociology 18,
ciate (or appreciate) in value and whether they are operationalized 109–127.
using pools of points or endowments (Dijkstra and van Assen, Corra, M., Willer, D., 2002. The gatekeeper. Sociological Theory 20, 180–207.
2008). The consequence of resource differences within such dif- Cross, R., Borgatti, S., Parker, A., 2001. Beyond answers: dimensions of the advice
network. Social Networks 23, 215–235.
ferent exchange contexts remains to be seen. Dekker, D., Stokman, F., Franses, P.H., 2003. Effectiveness of brokering within account
Second, an important next step is to develop measures to pre- management organizations. Erasmus Research Institute of Management, The
dict advantage across a wide range of networks. Several challenges Netherlands.
Dijkstra, J., van Assen, M.A.L.M., 2008. Transferring goods or splitting a resource pool.
confront such an endeavor because moving beyond the standard
Social Psychology Quarterly 71, 17–36.
exchange resource allows the distribution of resources to change Doğan, G., van Assen, M.A.L.M., van de Rijt, A., Buskens, V., 2009. The stability of
during exchange and power mechanisms to arise endogenously as exchange networks. Social Networks 31, 118–125.
Emerson, R.M., 1972. Exchange theory, Part II: exchange relations and networks. In:
resources diffuse. While cycles are necessary for structural power
Berger, J., Zelditch Jr., M., Anderson, B. (Eds.), Sociological Theories in Progress,
through exclusion with information-type resources, the length of vol. 2. Houghton-Mifflin, Boston, pp. 58–87.
cycles will affect resource diffusion and hence the time required Flap, H., Volker, B., 2001. Goal specific social capital and job satisfaction: effects of
for the structural conditions that enable exclusion to appear. In different types of networks on instrumental and social aspects of work. Social
Networks 23, 297–320.
addition, predicting power through exclusion will be more difficult Fischbacher, Urs., 1999. z-Tree – Zurich Toolbox for Readymade Economic Experi-
when networks contain multiple, possibly overlapping cycles. Chal- ments – Experimenter’s Manual. Working Paper No. 21, Institute for Empirical
lenges predicting power through ordering arise when networks Research in Economics, University of Zurich.
Freeman, L.C., 1979. Centrality in social networks: conceptual clarification. Social
contain positions with degree greater than two, who may or may Networks 1, 215–239.
not be subject to ordering depending upon the resources available Friedkin, N., 1992. An expected value model of social power: predictions for selected
through each partner. exchange networks. Social Networks 14, 213–230.
Granovetter, M.S., 1973. The strength of weak ties. American Journal of Sociology
Lastly, it may be useful to consider whether an actor’s aware- 78, 1360–1381.
ness of network structure affects behavior. Though awareness of Lawler, E.J., 2002. Micro social orders. Social Psychology Quarterly 65, 4–17.
the network has not been found to have an effect with the stan- Lawler, E.J., Thye, S.R., Yoon, J., 2006. Commitment in structurally enabled and
induced exchange relations. Social Psychology Quarterly 69, 183–200.
dard exchange resource (Molm et al., 1999), with information-type
Lawler, E.J., Thye, S.R., Yoon, J., 2008. Social exchange and micro social order. Amer-
resources, or transferable resources in general, structural power ican Sociological Review 73, 519–542.
arises endogenously based on actors’ exchange decisions. If actors Lazega, E., Pattison, P.E., 1999. Multiplexity, generalized exchange, and cooperation
in organizations: a case study. Social Networks 21, 67–90.
are aware of the structure they may pursue exchanges that do not
Lovaglia, M., Skvoretz, J., Willer, D., Markovsky, B., 1995. Negotiated exchanges in
create the prerequisites for structural power. Moreover, awareness social exchange networks. Social Forces 74, 123–155.
of the network may allow actors to strategically choose partners Markovsky, B., Skvoretz, J., Willer, D., Lovaglia, M., Erger, J., 1993. The seeds of weak
who provide better access to distant resources, possibly increasing power: an extension of network exchange theory. American Sociological Review
58, 197–209.
their diffusion across the network, with consequences for oneself Markovsky, B., Willer, D., Patton, T., 1988. Power relations in exchange networks.
and others. American Sociological Review 53, 220–236.
D.R. Schaefer / Social Networks 33 (2011) 143–151 151

Marsden, P.V., 1982. Brokerage behavior in restricted exchange networks. In: Mars- Schaefer, D.R., Kornienko, O., 2009. Building cohesion in positively connected
den, P.V., Lin, N. (Eds.), Social Structure and Network Analysis. Sage, Beverly Hills, exchange networks. Social Psychology Quarterly 72, 384–402.
CA, pp. 201–218. Schaefer, D.R., Kornienko, O., 2010. Comparison processes in social exchange net-
Marsden, P.V., 1983. Restricted access in networks and models of power. American works. In: Shane, R., Thye, Edward, J., Lawler (Eds.), Advances in Group Processes,
Journal of Sociology 88, 686–717. vol. 27. Emerald, Bingley, UK, pp. 185–205.
Molm, L.D., Cook, K.S., 1995. Social exchange and exchange networks. In: Cook, K.S., Stokman, F.S., 2004. What binds us when with whom? Content and structure in
Fine, G.A., House, J.S. (Eds.), Sociological Perspectives on Social Psychology. Allyn social network analysis. In: Keynote Address at the SUNBELT XXIV International
and Bacon, Boston, pp. 209–235. Network for Social Network Analysis Conference, Portorož (Slovenia).
Molm, L.D., Peterson, G., Takahashi, N., 1999. Power in negotiated and reciprocal Thye, S., Lovaglia, M., Markovsky, B., 1997. Responses to social exchange and social
exchange. American Sociological Review 64, 876–890. exclusion in networks. Social Forces 75, 1031–1049.
Molm, L.D., Schaefer, D.R., Collett, J.L., 2007. The value of reciprocity. Social Psychol- van Assen, M.A.L.M., 2003. Exchange networks: an analysis of all networks up to
ogy Quarterly 70, 199–217. size 9. In: Thye, S.R., Skvoretz, J. (Eds.), Advances in Group Processes: Power and
Molm, L.D., Schaefer, D.R., Collett, J.L., 2009. Fragile and resilient trust: risk and Status, vol. 20. Elsevier, Oxford, England, pp. 67–103.
uncertainty in negotiated and reciprocal exchange. Sociological Theory 27, 1–32. Wellman, B., Wortley, S., 1990. Different strokes from different folks: community
Newman, M.E.J., 2005. A measure of betweenness centrality based on random walks. ties and social support. American Journal of Sociology 96, 558–588.
Social Networks 27, 39–54. Willer, D., 1999. Network Exchange Theory. Praeger, Westport, CT.
Podolny, J.M., Baron, J.N., 1997. Resources and relationships: social networks and Willer, D., 2003. Power-at-a-distance. Social Forces 81, 1295–1334.
mobility in the workplace. American Sociological Review 62, 673–693. Willer, D., Emanuelson, P., 2008. Testing ten theories. Journal of Mathematical Soci-
Reagans, R., Zuckerman, E., 2008. Why knowledge does not equal power: ology 32, 165–203.
the network redundancy trade-off. Industrial and Corporate Change 17, Yamagishi, T., Gillmore, M.R., Cook, K.S., 1988. Network connections and the dis-
903–944. tribution of power in exchange networks. American Journal of Sociology 93,
Schaefer, D.R., 2007. Votes, favors, toys and ideas: the effect of resource character- 833–851.
istics on power in exchange networks. Sociological Focus 40, 138–160. Yamaguchi, K., 1996. Power in networks of substitutable/complementary exchange
Schaefer, D.R., 2009. Resource variation and the development of cohesion in relations: a rational-choice model and an analysis of power centralization.
exchange networks. American Sociological Review 74, 551–572. American Sociological Review 61, 308–332.