1997 ), and the USA (see O’Toole 1997 ; Salamon 2002 ). There are two broad schools of
thought, depending on how they seek to explain network behavior: power depend-
ence or rational actor. 4
Power Dependence
The power dependence approach treats policy networks as sets of resource-depen-
dent organizations. Their relationships are characterized by power dependence; that
is, ‘‘any organization is dependent on other organizations for resources,’’ and ‘‘to
achieve their goals, the organizations have to exchange resources.’’ So, actors ‘‘employ
strategies within known rules of the game to regulate the process of exchange.’’
Relationships are a ‘‘game’’ in which organizations maneuver for advantage. Each
deploys its resources, whether constitutional-legal, organizational,Wnancial, polit-
ical, or informational, to maximize inXuence over outcomes while trying to avoid
becoming dependent on the other ‘‘players.’’ So, behavior in policy networks is
gamelike, rooted in trust and regulated by rules of the game negotiated and agreed
by network participants. Variations in the distribution of resources and in the
bargaining skills of participants explain both diVerences in outcomes in a network
and variations between networks. Finally, the networks have a signiWcant degree of
autonomy from government (Rhodes 1997 a, ch. 2 ; 1999 / 1981 , ch. 5 ). 5
Rational Choice
The rational choice school explains how policy networks work by combining rational
choice and the new institutionalism to produce actor-centered institutionalism. The
best example is the Max-Planck-Institut’s notion of ‘‘actor-centered institutionalism.’’
For Renate Mayntz, Fritz Scharpf, and their colleagues at the Max-Planck-Institut,
policy networks represent a signiWcant change in the structure of government. They
are speciWc ‘‘structural arrangements’’ that deal typically with ‘‘policy problems.’’
They are a ‘‘relatively stable set of mainly public and private corporate actors.’’ The
links between network actors serve as ‘‘communication channels and for the exchange
of information, expertise, trust and other policy resources.’’ Policy networks have their
own ‘‘integrative logic’’ and the dominant decision rules stress bargaining and
sounding out. So, as with the power dependence approach, the Max Planck school
stresses functional diVerentiation, the linkages between organizations, and depend-
ence on resources (Kenis and Schneider 1991 , 41 – 3 ).
4 Bob Goodin pointed out correctly that theories of complexity are also relevant to the study of
network (personal correspondence). See, for example, La Porte 1975 ; Luhmann 1982 ; Simon 1981 / 1969.
Such ideas exercised some inXuence on the ‘‘governance club’’ research program at Erasmus University,
Rotterdam (see for example Kickert, Klyn, and Koppenjan 1997 ). They have not been a major inXuence
on the rest of the network literature.
5 The analysis of ‘‘power dependence’’ is not limited to the study of networks. More generally see: Blau
1964 ; Emerson 1962 ; Keohane and Nye 1977 , 1987 ; PfeVer and Salancik 1978.
policy network analysis 431