yt 1 Wyt
Friedkin & Johnson (1990) generalized this model to include a matrix X of
independent variables and a column vector b of their regression coefficients:
yt 1 Wyt Xb
Many research studies yield results that are consistent with the social
network influence model’s hypothesized effects (e.g., Friedkin 2004).
Coleman’s Collective Action Model
In The Mathematics of Collective Action (1973), James Coleman
modeled legislative vote-trading within a market of perfect
information on policy preferences, and resulting prices (power).
P = PXC
P: each legislator’s equilibrium power, following all vote exchanges
X: their interests over a set of legislative events (bills) to be decided
C: their control over each event (i.e., one vote per actor on each bill)
Marsden’s Network Access Model
Peter Marsden (1983) modified Coleman’s market exchange
model so that network relations restrict access to vote transfers.
In contrast to Coleman’s market model allowing every
legislator to trade votes with all others, Marsden
assumed varied opportunities for dyadic vote trades.
Compatibility of interests – based on trust, ideology,
or party loyalty – may restrict the subset of actors
with whom a legislator would prefer to log roll votes.
P = PAXC
A: aij =1 if vote exchanges are possible; aij = 0 if no exchange access
PXA = WK*
K*: equilibrium control matrix (L actors control the votes of K agents)
Resource deployment model operationalized actors’ control as
confirmed policy communication network, measuring “self-
control” as the N of orgs not confirming the sender’s information
exchange offers (i.e., indicator of independence in the system).
Legislative Outcome Predictions
Predicting pass/fail of labor policy bills, U.S. better exemplified a
resource mobilization process, while German and Japanese data
better fit a resource deployment model (Knoke et al. 1996:181).
Dynamic Access Models
Frans Stokman’s stage models of dynamic access: (1) actors’
form policy preferences, influenced by the preferences of actors
who have access to them; then (2) officials cast votes based on
preferences formed during that prior stages of influence activity.
Networks & policy preferences exert mutually formative
influences; then votes cast on fixed preferences.
• Power-driven actors seek access to most powerful
players
• Policy-driven: interaction of power & policy
positions
C = RA X = XCS O = XV
C: control over events R: actors’ resources A: access to other actors
X: preferences on events (interests) S: salience of event decisions
V: voting power of the public officials O: expected outcomes
Amsterdam Policy Outcomes
Stokman & Berveling (1998) compared dynamic policy network
models to real outcomes of ten Amsterdam policy decisions.
Policy Maximization model performed better
than either Control Maximization or Two-
Stage models. Policy-driven actors “accept
requests selectively to ‘bolster’ their own
preferences as much as possible.”