For those who have been following the debate on policy networks, I would like to
recommend two excellent review articles which address many of the issues raised
by list members and which, to my shame, I did not read before writing the
article for Social Science Japan.

The first is Atkinson, M. and Coleman, W.D (1992): Policy Networks, Policy
Communities and the Problems of Governance. In: Governance, Vol.5 No.2, April,
pp.154-180. The second is Dowding, Keith (1995): Model or Metaphor? A Critical
Review of the Policy Network Approach. In: Political Studies, Vol.43 No.1,
March, pp.136-158

Atkinson and Coleman identify three areas in which policy networks analysis
needs to develop:

(1) The role of the state and of national institutions and regimes. As Adam
Sheingate emphasizes in his post on agricultural policy networks (17 Aug),
policy network analysts developed their framework in reaction to the unwieldy
Katzensteinian emphasis on strong and weak states, which rode roughshod over
obvious differences between policy processes in different domains. However, the
state cannot be written off quite so easily. Atkinson and Coleman draw on an
area of research familiar to them, namely corporatism, to point out that
"researchers have found it difficult to explain cross-national differences in
policy by referring to sectoral variables alone."

National institutional environments DO influence the shape of policy networks.
Policy makers operate in hierarchies as well as in networks. Policy network
analysts therefore have as much to learn from Nobuhiro Hiwatari (7 Aug) and
other historical institutionalists as vice versa. Also, I am currently wrestling
with the problem of how the ideology of technonationalism described by Richard
Samuels and Shigeru Nakayama is expressed in policy networks. If such national
factors are indeed significant - and Samuels and Nakayama are very persuasive -
then policy network analysts must take account of the macro-level after all.

(2) The internationalization of policy networks. A combination of two-level
theory and policy network analysis has great potential, but two-level theory is
limited in its application. For example, where interest associations and
multinational companies engage in policy networks (e.g. by lobbying) in both
countries, international negotiators have less control over bilateral exchanges.

Even in state-funded policy areas there are informal, often professionalized,
policy networks bypassing international negotiations. For example, national
space agency engineers have been known to engage in covert coordination of
programs with their opposite numbers overseas.

(3) Changes in networks over time. Atkinson and Coleman identify three sets of
factors: changes in who is included and who excluded from policy networks (in
Wilks and Wright's usage); use of new ideas to change the structures of
networks; and the effects on processes of changes either in related networks or
in the broader political economy.

From a different perspective, Dowding effectively pins down the limitations of
policy network analysis. For him, the typologies of network types generated by
policy network analysts do not constitute a theory but a system of
classification. (Wilks and Wright, among others, never claimed that they do
more). In some cases the zeal for classification has led to a "lepidopterist"
profusion of categories and sub-categories which ignore causality between one
category and another and generally fail to explain WHY policy networks differ
from each other. Often policy network analysts adopt a weighty new vocabulary to
describe phenomena easily understood by common sense.

To purge this train-spotter approach to political networks, Dowding prescribes a
stiff dose of bargaining models and game theory, justified by Rhodes' original
definition of policy networks as power-dependency relations. In place of policy
network analysis, which concentrates on the properties of the actors, he
proposes sociological network analysis, which concentrates on network
characteristics. However, he rather confusingly adds that actors and their
relations are mutually dependent and not easily distinguished from each other.
Finally, Dowding further dampens his methodological squib by pointing out that
sociological network analysis also has a tendancy to reinvent the wheel: "This
is not an argument against formalism, but it is an argument against too high
expectations from it."

Jonathan Lewis
Research Associate
Institute of Social Science
University of Tokyo