Ústřední knihovna FSS MU Brno AU:- ľ" 4240746031 .C ,/,. ''3'Zi MARKETS, HIERARCHIES AND NETWORKS The Coordination of Social Life edited by Grahame Thompson Jennifer Frances Rosalind Levačič Jeremy Mitchell W Published in association 3$«n8y with The Open University ® SAGE Publications London • Thousand Oaks • New Delhi 4240746031 170 John Vickers regulation have often been erected, just as old ones were taken down. Second, deregulation was sometimes deliberately limited in its scope, for example, at sensitive points in the privatization programme. Third, where deregulation in the sense of the removal of legal barriers to entry has occurred, it has often not been complemented by appropriate measures to make'competition effective. Have the frontiers of the state been rolled back in the area of regulation? Should they be? I think that the initial attempts to regulate the privatized telecommunications and gas industries 'with a light touch' were a mistake. A number of the developments referred to above suggest that even the government now thinks so, and its attitude to privatization of the electricity supply industry reflects a rather different approach that does not look much like rolling back the frontiers of the state. The question whether deregulation should go further is perhaps not the right one. Some regulation is unnecessary, but much is required by the market failures that inevitably exist in some industries. Better targeting of regulatory policy on those market failures is not a simple matter of more or less regulation. There are several kinds of regulation, and those which harness and assist market forces should generally be preferred to those which do not. Note This chapter borrows from J. Kay, and J. Vickers, (1988) 'Regulatory reform in Britain', in Economic Policy, 7: 286-351. Also see J. Vickers and G. Yarrow, Privatisation: an Economic Analysis. Cambridge, MA: MIT Press, 1988. Networks Introduction Grahame Thompson The chapters in this section deal with the idea of networks as a means of coordination. Examples of the network approach are drawn from a variety of social science and organizational theory sources. The key feature of networks that they all address, however, is the way cooperation and trust are formed and sustained within networks. In contrast to either hierarchy or market, networks coordinate through less formal, more egalitarian and cooperative means. Exactly how these features operate in a range of different environments and contexts is the subject of each chapter. The chapter by Knoke and Kuklinski draws on the sociological tradition of network analysis to outline the manner in which 'network structures' are formed and operate. They suggest that a range of attributes and relationships underpins network structures, and that they work at different dimensions and levels. On the basis of any particular problem or objective, it is up to the social analyst to decide exactly how the network study is to be set up and the features it will exemplify. A particular type of network operating in the field of economic organization is discussed in Chapter 15. This reports a study into the French engineering industry situated around Lyons in France. Comprising mainly small to medium-sized firms, Lorenz found a surprisingly dense network of subcontracting and main firms, existing in the form of an industrial district, that were very dynamic and innovative in the introduction of new technology. He analyses how cooperation and trust are formed and sustained amongst the network of firms despite the strong competition existing between them. They forgo short-run profit maximization for the longer-term benefits of mutual cooperation. One way of characterizing the kind -• of intra- and inter-firm relationships discussed in Lorenz's case study is as a 'value-adding partnership'. Johnston and Lawrence (Chapter 16) see the typical vertically integrated firm being broken up in the face of the need for flexibility in response to the 172 Grahame Thompson rapidity of market demand changes. The different, and now increasingly organizationally autonomous, parts of the overall productive chain or network they designate as akin to a partnership rather than as a market or hierarchical organization. Policy networks are an example of the informal personal ties that develop between professional groups within the public sector and elsewhere. In Chapter 17 Rhodes draws attention to the way these operate within sub-central government in the UK. He looks at the processes of exchange going on within these kinds of network and at the rules and strategies governing resource translations between the participants. Another example of the way informal relationships can serve to articulate personal networks is demonstrated by Pnina Werbner's analysis of female working and domestic bonds amongst Pakistani immigrant women (Chapter 18). This serves to highlight the strong gender and ethnic character of a good many network structures. It points to the manner in which personal contacts and community values help to reinforce a soldaristic bond so necessary for the operation of local and small-scale network arrangements. Finally in this section we focus on a more overtly political aspect to networks. What is the political form of a typical government structure? In their analysis of the way private interests are governed in the advanced industrial democracies of the West, Streeck and Schmitter (Chapter 19) argue that as well as the market, the state and the community operating as ordering mechanisms, there exists another equally important coordinating arrangement which they term 'associations'. These associations act as a kind of network in which the multifarious interests represented within the private sphere are made manifest and given an expression. They suggest that the network of associations so created provides an important adjunct to the traditional modes of political representation organized around the state. It helps coordinate the political realm by bargaining and negotiating away what might otherwise arise as conflictual and antagonistic social tensions. 14 Network analysis: basic concepts David Knoke and James H. Kuklinski l.. J To appreciate fully the distinctive theoretical underpinnings of network approaches to social phenomena, a comparison with more traditional, individualistic approaches may be useful. In the atomistic perspectives typically assumed by economics and psychology, individual actors are depicted as making choices and acting without regard to the behavior of other actors. Whether analysed as purposive action based on rational calculations of utility maximization, or as drive-reduction motivation based on causal antecedents, such individualistic explanations generally ignore the social contexts within which the social actor is embedded. In contrast, network analysis incorporates two significant assumptions about social behavior. Its first essential insight is that any actor typically participates in a social system involving many other actors, who are significant reference points in one another's decisions. The nature of the relationships a given actor has with other system members thus may affect that focal actor's perceptions, beliefs and actions. But network analysis does not stop with an account of the social behavior of individuals. Its second essential insight lies in the importance of elucidating the various levels of structure in a social system, where structure consists of 'regularities in the patterns of relations among concrete entities' (White et aL, 1976). In the individualistic approach, social structure is seldom an explicit focus of inquiry, to the extent that it is even considered at all. Network analysis, by emphasizing relations that connect the social positions within a system, offers a powerful brush for painting a systematic picture of global social structures and their components. The organization of social relations thus becomes a central concept in analysing the structural properties of the networks within which individual actors are embedded, and for detecting emergent social phenomena that have no existence at the level of the individual actor. [. . .] Adapted from D. Knoke and J.H. Kuklinski, Network Analysis (Beverly Hills: Sage Publications, 1982), pp. 9-21. 174 David Knoke and James H. Kuklinski Attributes and relations Two basic approaches to viewing and classifying the various aspects of the social world - according to their attributes or their relationships - are often treated as antithetical and even irreconcilable. We need to make clear from the outset how these two approaches to measurement differ. We shall also point out that neither perspective by itself yields satisfactory understandings of social phenomena. Attributes are intrinsic characteristics of people, objects or events. When we think of explaining variance among such units of observation, we almost naturally resort to attribute measures, those qualities that inherently belong to a unit apart from its relations with other units or the specific context within which it is observed. Various types of attribute can be measured: an occupation's average income, a nation's gross national product, a riot's duration, a birth cohort's mean formal schooling, a person's opinion about the president. Persons, objects and events may also be involved in relationships, that is, actions or qualities that exist only if two or more entities are considered together. A relation is not an intrinsic characteristic of either party taken in isolation, but is an emergent property of the connection or linkage between units of observation. Where attributes persist across the various contexts in which an actor is involved (for example a person's age, sex, intelligence, income, and the like remain unchanged whether at home, at work or at church), relations are context specific and alter or disappear upon an actor's removal from interaction with the relevant other parties (a student/teacher relation does not exist outside a school setting; a marital relation vanishes upon death or divorce of a spouse). A wide variety of relational properties can be measured: the strengths of the friendships among pupils in a classroom, the kinship obligations among family members, the economic exchanges between organizations. Many aspects of social behavior can be treated from both the attribute and the relational perspectives, with only a slight alteration of conceptualization. For example, the value of goods that a nation imports in foreign trade each year is an attribute of the nation's economy, but the volume of goods exchanged between each pair of nations measures an exchange relationship. Similarly, while a college student's home state is a personal attribute, a structural relationship between colleges and states could be measured by the proportions of enrolled students coming to each college from each state. [. . .3 The point we are stressing is that, while attributes and relationships are conceptually distinct approaches to social research, they should be seen as neither polar nor mutually exclusive measurement options.. [. . J Relational measures capture emergent properties of social systems that cannot be measured by simply aggregating the attributes of individual members. Furthermore, such emergent properties may significantly affect both system performance and the behavior of network members. For example, the structure of informal Basic concepts 175 friendships and antagonisms in formal work groups can affect both group and individual productivity rates in ways not predictable from such personal attributes as age, experience, intelligence, and the like (Homans, 1950). As another example, the structure of communication among medical practitioners can shape the rate of diffusion of medical innova- * tions in a local community and can determine which physicians are likely to be early or late adopters (Coleman et al., 1966). [. . J Networks Relations are the building blocks of network analysis. A network is generally defined as a specific type of relation linking a defined set of persons, objects or events (see Mitchell, 1969). Different types of relations identify different networks, even when imposed on the identical set of elements. For example, in a set of employees at a workplace, the advice-giving network is unlikely to be the same as the friendship network or the formal authority network. The set of persons, objects or events on which a network is defined may be called the actors or nodes. These elements possess some attribute(s) that identify them as members of the same equivalence class for purposes of determining the network of relations among them. For example, we might stipulate that all payroll employees at plant six of the National Widget Corp. comprise the set of actors among whom an advice-giving network is sought. Additional restrictions on the permissible actors could be imposed (for example only males in managerial jobs), indicating that delimiting network boundaries depends to a great extent upon a researcher's purposes. Our generic definition of a network may imply that only those linkages that actually occur are part of a network. But network analysis must take into account both the relations that occur and those that do not exist among the actors. For example, attending only to the gossip connections in a community and not to the structural 'holes' that occur where links are absent might result in an inaccurate understanding of how rumors spread or evaporate. The configuration of present and absent ties between the network actors reveals a specific network structure. Structures vary dramatically in form, from the isolated structure in which no actor is connected to any other actor, to the saturated structure in which every actor is directly linked to every other individual. More typical of real networks are various intermediate structures in which some actors are more extensively connected among themselves than are others. [. . .1 If network analysis were limited just to a conceptual framework for identifying how a set of actors is linked together, it would not have excited much interest and effort among social researchers. But network analysis contains a further explicit premiss of great consequence: The structure of relations among actors and the location of individual actors 176 David Knoke and James H. Kuklinski in the network have important behavioral, perceptual and attitudinal consequences both for the individual units and f or the system as a whole. In Mitchell's (1969) felicitous terms, 'The patterning of linkages can be used to account for some aspects of behavior of those involved.' For example, a formal organization with a centralized structure of authority among' its various divisions and departments may be most effective (for example enjoying high growth and profitability) in a relatively placid environment, but in a turbulent, rapidly changing environment an organization with a less centralized structure may be more adaptable. [. ■ .1 To illustrate the potential power of a network approach, consider a variety of contemporary social science problems: the sources of homo- phyly of beliefs within a power-elite, the adoption of technological innovations, the causes of corporate profitability, the income earnings of occupational groups, the recruitment processes of social movement organizations, the development of non-traditional sex roles. In each jf these and many other substantive areas, a large research literature can be uncovered that attempts to explain the phenomena as a function of individual or group attributes. Yet in many instances, such characteristics may predict behavior only because of underlying patterns of relations that are often associated with these attributes. [. . .1 Network approaches can more faithfully capture the context of social relations within which actors participate and make behavioral decisions. Research design elements Network analyses take many forms to suit researchers' diverse theoretical and substantive concerns. Four elements of a research design in particular shape the measurement and analysis strategies available to a researcher: the choice of sampling units, the form of relations, the relational content, and the level of data analysis. Varying combinations of these design elements have created a wide diversity among network studies that is evident in the research literature. Sampling units Before collecting data, a network researcher must decide the most relevant type of social organization and the units within that social form that comprise the network nodes. Ordered in a roughly increasing scale of size and complexity are a half-dozen basic units from which samples may be drawn: individuals, groups (both formal and informal), complex formal organizations, classes and strata, communities, and nation-states. A typical design involves some higher-level system whose network is to be investigated with one or more lower-level units as the nodes, for example a corporation with its departments and individual employees as the actors, Basic concepts 177 or a city with its firms, bureaus and voluntary associations as the nodes. [. . .1 Form of relations The relations between actors have both content and form. Content refers to the substantive type of relation represented in the connections (such as supervising, helping, gossiping), and an inventory of content types is presented below. Relational form refers to properties of the connections between pairs of actors (dyads) that exist independently of specific contents. Two basic aspects of relational form are (a) the intensity or strength of the link between two actors, and (b) the level of joint involvement in the same activities (Burt, 1982: 22). Relational content In conjunction with choosing the appropriate sampling units, a network analyst must decide what specific network linkages to investigate. [. . .] Because researchers' capacities to conceptualize and operationalize various types of network are almost unlimited, we can only list the more common types of relational content, citing some representative studies: ® Transaction relations: Actors exchange control over physical or symbolic media, for example in gift giving or economic sales and purchases (Burt et al., 1980; Laumann et al., 1978). & Communication relations: Linkages between actors are channels by which messages may be transmitted from one actor to another in a system (Marshall, 1971; Lin, 1975; Rogers and Kincaid, 1981). 9 Boundary penetration relations: The ties between actors consist of constituent subcomponents held in common, for example, corporation boards of directors with overlapping members (Levine, 1972; Allen, 1974; Mariolis, 1975; Sonquist and Koenig, 1975; Burt, 1982: ch. 8). ® Instrumental relations: Actors contact one another in efforts to secure valuable goods, services or information, such as a job, an abortion, political advice, recruitment to a social movement (Granovetter, 1974; Boissevain, 1974). a Sentiment relations: Perhaps the most frequently investigated networks are those in which individuals express their feelings of affection, admiration, deference, loathing or hostility toward each other (Hunter, 1979; Hallinan, 1974; Sampson, 1969). ® Authority/power relations: These networks, usually occurring in complex formal organizations, indicate the rights and obligations of actors to issue and obey commands (White, 1961; Cook and Emerson, 1978; Williamson, 1970; Lincoln and Miller, 1979). e Kinship and descent relations: A special instance of several preceding generic types of networks, these bonds indicate role relationships among family members (Nadel, 1957; Bott, 1955; White, 1963). 178 David Knoke and James H. Kuklinski Levels of analysis After selecting the sampling units and relational content, a network analyst will have several alternative levels at which to analyse the data collected for a project. Here we consider four conceptually distinct levels of analysis at which an investigation can focus. The simplest level is the egocentric network, consisting of each individual node, all others with which it has relations, and the relations among these nodes. If the sample size is N, there are N units of analysis at the ego-centered level. Each actor can be described by the number, the magnitude, and other characteristics of its linkages with the other actors, for example, the proportion of reciprocated linkages or the density of ties between the actors in ego's first 'zone' (that is the set of actors directly connected to ego). [. . .1 At the next highest level of analysis is the dyad, formed by a pair of nodes. If the sample size is N, there are (N2-N)/2 distinct units of analysis at the dyadic level. The basic question about a dyad is whether or not a direct tie exists between the two actors, or whether indirect connections might exist via other actors in the system to which they are connected. Typical dyadic analyses seek to explain variation in dyadic relations as a function of joint characteristics of the pair, for example the degree of similarity of their attribute profiles. [. . .1 Not surprisingly, the third level of analysis consists of triads. If N is the sample size, there are (-^-) distinct triads formed by selecting each possible subset of three nodes and their linkages. Research using triads has largely concentrated on the local structure of sentiment ties among individual actors, with a particular concern for determining transitivity relations (that is, if A chooses B and B chooses C, does A tend to choose C?). [...]" Beyond the triadic level, the most important level of analysis is that of the complete network, or system. In these analyses, a researcher uses the complete information about patterning of ties between all actors to ascertain the existence of distinct positions or roles within the system and to describe the nature of relations among these positions. Although the sample may consist of N nodes and (N2-N) possible dyadic ties of a given type, these elements altogether add up to only a single system. 1. . J Structure in complete networks One major use of network analysis in sociology and anthropology has been to uncover the social structure of a total system. Systems may be as small as an elementary school classroom and a native village, or as large as a national industry and the world system of nation-states. But for any system, an important step is a structural analysis to identify the significant Basic concepts 179 positions within a given network of relations that link the system actors. The observable actors - be they pupils, organizations or national governments - are not the social structure. The regular pattern of relations among the positions composed of concrete actors constitutes the social structure of the system. Hence identification of positions is a necessary but incomplete prelude in complete network analysis, which requires the subsequent appraisal of the relations connecting positions one to another. Positions, or social roles, are subgroups within a network defined by the pattern of relations (which represent real observable behaviors) that connect the empirical actors to each other. I. . J By occupying positions in a network structure, individual actors have certain connections to other actors, who in turn also occupy unique structural positions. Although empirical actors and their observable linkages provide the data for identifying positions, a network's positions are conceptually distinct from any specific incumbents. For example, in a hospital system the positions defined by patterns of relations between actors - given such conventional labels as doctor, patient, nurse, administrator, paraprofessional, and so forth - persist despite frequent changes in the unique individuals occupying these positions. New positions may be created when an actor(s) establishes a unique set of ties to the pre-existing positions, for example when data-processing specialists are hired to manage the diagnostic and administrative information flow of the hospital. [. . .] In the process, the complexity of the network is typically simplified, reducing a large number of N actors into a smaller number of M positions, since typically several empirical actors occupy the same position (many doctors, many nurses, many patients, and so on). In deciding the basis on which to identify the positions in a complete network and to determine which actors jointly occupy each position, the network analyst has two basic alternatives (Burt, 1978). The first criterion is social cohesion. Actors are aggregated together into a position to the degree that they are connected directly to each other by cohesive bonds. Positions so identified are called 'cliques' if every actor is directly tied to every other actor in the position (maximal connection), or 'social circles' if the analyst permits a less stringent frequency of direct contact. [. . J The second criterion for identifying network positions is structural equivalence (Lorrain and White, 1971; White et al., 1976; Sailer, 1978). Actors are aggregated into a jointly occupied position or role to the extent that they have a common set of linkages to the other actors in the system. Í. . .1 A simple hypothetical example should make these conceptual distinctions clearer. Figure 1 portrays a fictional medical practice network. The lines connecting the actors represent 'frequent contacts on medical matters' (the figure is an unrealistic representation, but useful for 180 David Knoke and James H, Kuklinski Nurses Doctors Patients Receptionists Figure 1 A hypothetical medical practice network illustrative purposes). A social cohesion criterion identifies two distinct cliques, a small one involving just the two receptionists, and a large one containing all three nurses and both physicians. But using structural equivalence criteria, four distinct positions would emerge, corresponding to the four roles labeled in the figure. Nurses and doctors are no longer aggregated because they differ in their patterns of contacts with the other actors (the doctors are linked to the patients but the nurses are not, undoubtedly untrue in a real system). Three of these structurally equivalent positions are also cliques, but the patient position is not a clique because its occupants do not discuss medical matters among themselves. The point of this exercise is that different criteria for identifying structural positions in networks can, and usually do, yield different results. The choice of methods for locating positions in an empirical network ultimately depends, as in the application of any method, on the substantive and theoretical problem the analyst is addressing. For some purposes a clique approach will be preferred, while in other situations a structural equivalence procedure will be more useful. To state a definitive rule about which one to choose that would cover all situations is impossible. E. . J References Allen, M.P. (1974) 'The structure of interorganizational elite cooptation: interlocking corporate directors.' American Sociological Review 39: 393-406. Basic concepts 181 Boissevain, J. (1974) Friends of Friends: Networks, Manipulators, and Coalitions. New York: St Martin's. ßott, E. (1955) 'Urban families: conjugal roles and social networks.' Human Relations 8: 345-83. Burt, R.S. (1978) 'Cohesion versus structural equivalence as a basis for network subgroups.' * Sociological Methods & Research 7: 189-212. Burt, R.S. (1982) Toward a Structural Theory of Action: Network Models of Social Structure, Perceptions and Action. New York: Academic. Burt, R.S., Christman, K.P. and Kilburn, Jr H.C. (1980) 'Testing a structural theory of corporate cooptation: interorganizational directorate ties as a strategy for avoiding market constraints on profits.' American Sociological Review 45: 821-41. Coleman, J.S. and MacRae, Jr D. (1960) 'Electronic processing of sociometric data for groups up to 1000 in size.' American Sociological Review 25: 722-7. Coleman, J.S., Katz, E. and Menzel, H. (1966) Medical Innovation: A Diffusion Study. Indianapolis: Bobbs-Merrill. Cook,K.S. and Emerson, R.M. (1978) 'Power, equity and commitment in exchange networks.' American Sociological Review 43: 721-39. Granovetter, M. (1974) Getting a Job: A Study of Contacts and Careers. Cambridge, MA: Harvard University Press. Hallinan, M.T. (1974) The Structure of Positive Sentiment. Amsterdam: Elsevier. Etonians, G. (1950) The Human Group. New York: Harcourt, Brace, Jovanovich. Hubbeil, C.H. (1965) 'An input-output approach to clique identification.' Sociometry 28: 377-99. Hunter, J.E. (1979) 'Toward a general framework for dynamic theories of sentiment in small groups derived from theories of attitude change', pp. 223-38 in Paul W. Holland and Samuel Leinhardt (eds) Perspectives on Social Network Research. New York: Academic. Laumann, E.O., Galaskiewicz, J. and Marsden, P.V. (1978) 'Community structure as interorganizational linkages', pp. 455-84 in A. Inkeles et al. (eds) Annual Review of Sociology, Vol. 4, Palo Alto, CA: Annual Reviews. Levine, J. (1972) 'The sphere of influence.' American Sociological Review 37: 14-27. Lin, N. (1975) 'Analysis of communication relations', in Gerhard J. Hannemann and William J. McElwen (eds) Communication and Behavior. Reading, MA: Addison-Wesley. Lin, N. (1976) Foundations of Social Research. New York: McGraw-Hill. Lincoln, J.R. and Miller, J. (1979) 'Work and friendship ties in organizations: A comparative analysis of relational networks.' Administrative Science Quarterly 24: 181-99. Lorrain, F. and White, H.C. (1971) 'Structural equivalence of individuals in social networks.' Journal of Mathematical Sociology 1: 49-80. Mariolis, P. (1975) 'Interlocking directorates and control of corporations.' Social Science Quarterly 56: 425-39. Marshall, J.F. (1971) 'Topics and networks in intra-village communication.' pp. 160-6 in Steven Polgar (ed.) Culture and Population: A Collection of Current Studies. Monograph 9. Chapel Hill, NC: Carolina Population Center. Mitchell, J.C. (1969) 'The concept and use of social networks', pp. 1-50 in J. Clyde Mitchell (ed.) Social Networks in Urban Situations. Manchester: Manchester University Press. Nadel, S.F. (1957) The Theory of Social Structure. London: Cohen & West. Rogers, E.M. and Kincaid, D.I. (1981) Communication Networks: Toward a New Paradigm for Research. New York: Macmillan. Sailer, L.D. (1978) 'Structural equivalence: meaning and definition, computation and application', Social Networks 1: 73-90. Sampson, S.F. (1969) 'Crisis in a cloister.' PhD dissertation, Cornell University, Ithaca, New York. Sonquist, J. and Koenig, T. (1975) 'Interlocking directorates in the top US corporations: a 182 David Knoke and James H. Kuklinski graph theory approach.' Insurgent Sociologist 5: 196-229. White, H.C. (1961) 'Management conflict and sociometric structure.' American Journal of 1 Sociology 67: 185-7. White, H.C. (1963) An Anatomy of Kinship. Englewood Cliffs, NJ: Prentice-Hall. White, H.C, Boorman, A. and Breiger, R.L. (1976) 'Social structure from multiple networks 1. Blockmodels of roles and positions.' American Journal of Sociology 81: 730*80. Williamson, O.E. (1970) Corporate Control and Business Behavior. Englewood Cliffs, NJ: Prentice-Hall. 15 Neither friends nor strangers: informal networks of subcontracting in French industry Edward H. Lorenz Economists as a rule have attached little importance to the role of such social ties as trust and friendliness in market exchange. As Albert Hirschman (1982) has observed, this can be explained by the fact that the ideal market upon which claims of allocative efficiency rest involves large numbers of price-taking anonymous buyers and sellers supplied with perfect information. With such markets there is no room for bargaining, negotiation or mutual adjustment, and the operators that contract together need not enter into a recurrent or continuing relationship. This chapter considers a case which does not conform to the economist's competitive ideal, that of continuing and recurrent relations between French firms and their subcontractors. These are relations involving mutual dependency, where each firm's actions influence the other. The situation by its very nature calls for cooperation, and it is reasonable to ask whether trust plays a role in this process. in 1985 I began a study of the introduction of new technology in small and medium French engineering firms.1 This was prompted by a number of intriguing bits of evidence. From 1975 firms in this category had improved their performance relative to large firms in terms of profitability and rates of growth of output and employment. Further, in terms of the latter two criteria, the smaller firms in this category (between 10 and 100 employees) had outperformed the larger (Delattre, 1982). Secondary sources also showed that small and medium firms had been some of the most dynamic investors in advanced computer-based technology, primarily NC and CNC machine tools (Cavestro, 1984). This picture of comparatively rapid growth and technological sophistication contradicted established views of the role of small firms in the French economy. [. . .] Preliminary visits to firms with 200 to 500 employees revealed that most had substantially reduced their employment levels since 1980. The value Adapted from Gambetta, D. (ed.) Trust: Making and Breaking of Cooperative Relations (Oxford: Basil Blackwell, 1989), pp. 194-210. 202 Russell Johnston and Paul R. Lawrence sanction, of course, is to terminate the partnership. In the Prato area, this could happen if, for instance, an impannatore failed to pass orders back to the weaver who had supplied the fabric design that was being sold. It seems clear that, for at least some value-added chains, a value-adding partnership is a viable and advantageous means of achieving the benefits of vertical integration. By observing the characteristics of and the processes followed by successful partnerships, executives can determine whether VAPs might pay off for their organizations. Business relationships premised on the need to achieve bargaining power may be more aggressively competitive than is in their best interest. Remember that the examples cited earlier - US automobiles, Italian textiles, and drug distribution - all evolved from competitive, sometimes acrimonious relationships. Notes 1. For other discussions of the new organizational forms, see Raymond Miles and Charles Snow, 'Network organizations: new concepts for new forms', California Management Review, Spring 1987, p. 62; Robert G. Eccles, 'The quasifirm in the construction industry', Journal of Economic Behavior and Organization, December 1981, p. 335; Calvin Pava, 'Managing the new information technology: design or default?', in HRM Trends and Challenges, ed. Richard E. Walton and Paul R. Lawrence (Boston: Harvard Business School Press, 1985); and Andrea Larson, 'Networks as organizations', unpublished manuscript, 1987. 2. This description draws heavily on Michael J. Piore and Charles F. Sabel, The Second Industrial Divide (New York: Basic Books, 1984) and on Gianni Lorenzoni, Una Politico Innovative (Milan: Etas Libri, 1979). 3. The facts about Massimo Menichetti are excerpted from the HBS case, Massimo Menichetti (B) 686-135, revised October 1986, prepared by Ramchandran Jaikumar. 17 Policy networks and sub-central government R.A.W. Rhodes [...] Varieties of network Following Benson (1982: 148), a policy network can be defined as a 'complex of organizations connected to each other by resource dependencies and distinguished from other [. . .] complexes by breaks in the structure of resource dependencies'. Rhodes (1981: ch. 5, 1985, and 1986: ch. 2) elaborates this definition, arguing that networks have different structures of dependencies, structures which vary along five key dimensions: » Constellation of interests - the interests of participants in a network vary by service/economic function, territory, client group and common expertise (and most commonly some combination of the foregoing). • Membership - membership differs in terms of the balance between public and private sector; and between political-administrative elites, professions, trade unions and clients. • Vertical interdependence - intra-network relationships vary in their degree of interdependence, especially of central or sub-central actors for the implementation of policies from which, none the less, they have service delivery responsibilities. • Horizontal interdependence - relationships between the networks vary in their degree of horizontal articulation: that is, in the extent to which a network is insulated from, or in conflict with, other networks. • The distribution of resources - actors control different types and amounts of resources, and such variations in the distribution of resources affect the patterns of vertical and horizontal interdependence. Although the available research on British policy networks is limited, none Adapted from R.A.W. Rhodes, Beyond Westminster and Whitehall: The Sub-central Governments of Britain (London: Unwin Hyman, 1988), pp. 77-85. 204 R.A.W. Rhodes Policy networks and sub-central government 205 the less it is possible to identify some of the main varieties in sub-central government (SCG).1 Thus it is possible to distinguish, at a minimum, between policy and territorial communities on the one hand and issue, professionalized, intergovernmental and producer networks on the other. Policy communities are networks characterized by stability of relationships., continuity of a highly restrictive membership, vertical interdependence based on shared service delivery responsibilities and insulation from other networks and invariably from the general public (including Parliament). They have a high degree of vertical interdependence and limited horizontal articulation. They are highly integrated. The distinction between policy and territorial communities refers, rather obviously, to differences in their constellation of interests. Policy communities are based on the major functional interests in and of government - for example education, fire (Richardson and Jordan, 1979; Rhodes, 1986: ch. 8) - whereas territorial communities encompass the major territorial interests - for example in Scotland, Wales and Northern Ireland (Keating and Midwinter, 1983; Rhodes, 1986: ch. 7). Other networks differ in that they are less integrated. The least integrated form is the issue network. The distinctive features of this kind of network are its large number of participants and their limited degree of interdependence. Stability and continuity are at a premium, and the structure tends to be atomistic (Heclo, 1978). Commonly, there is no single focal point at the centre with which other actors need to bargain for resources. The prime example in British government seems to be the field of leisure and recreation. Seven central departments have responsibilities in this area, and at the sub-central level there are many non-departmental and intermediate organizations as well as all the tiers of local government (see Travis et al., 1978, esp. p. 27a). Professionalized networks are characterized by the pre-eminence of one class of participant in policy-making: the profession. The most cited example of a professionalized policy network is the National Health Service, wherein the power of the medical profession is substantial. The water service provides a further example wherein the constraints on water engineers seem particularly weak (Keating and Rhodes, 1981; Gray 1982; Saunders, 1983: 34-7)2. In short, professionalized networks express the interests of a particular profession and manifest a substantial degree of vertical independence whilst insulating themselves from other networks. The analysis of the influence of professions cannot be confined to the distribution of resources but must also cover ideology. Dunleavy (1981a: 10) suggests that professions with operational control in peripheral agencies will develop a national-level ideological system. Consequently, trends in national professional opinion constrain or influence the centre, and the national professional association will both periodically formalize professional opinion and continuously disseminate information on best professional practice. Peripheral agencies see the national-level system as a source of ideas; it sets the parameters to their decision-making. Finally, the rotation of professions between peripheral agencies coupled with the usual traits of a profession, such as training and qualifications, serve to reinforce the national-level ideology: to present a unified 'view of the world' based on common ideas, values and knowledge. And Dunleavy's (1981b) case study of high-rise housing illustrates the operation of one" such national-level ideological structure. Professional influence is exercised in traditional interest group activities (for example lobbying) it is institutionalized in policy networks; and it sets the parameters to decisionmaking through national-level ideological structures. Intergovernmental networks or, in the case of England, the 'national community of local government' (Rhodes, 1986: 11-16 and ch. 3) are the networks based on the representative organizations of local authorities. Their distinctive characteristics are topocratic membership ( and the explicit exclusion of all public sector unions); an extensive constellation of interests encompassing all the services (and associated expertise and clients) of local authorities; limited vertical interdependence because they have no service delivery responsibilities but extensive horizontal articulation or ability to penetrate a range of other networks. The intergovernmental networks differ between the four nations. In England, there are a large number of organizations acting on behalf of local authorities in some capacity. This set of organizations speaks for disparate interests but manifests a high degree of interdependence, hence the appellation 'national community'. However, their links with individual local authorities - their members - are sporadic. In Scotland, Wales and Northern Ireland there is no equivalent to the English national community. Local authorities in each nation have their representative organizations, and only in Wales is there more than one such body. Moreover, the reduced scale of the networks means that they operate informally with far less reliance on explicit consultative mechanisms and far greater exchange with their members (James, 1982; Connolly, 1983; Rhodes, 1986, 256-67). Given its topocratic membership, it might be anticipated that the national community of local government would conflict with the technocratic policy communities. However, national-level ideological structures are not limited to the service-specific policy communities and their associated professions. There is a 'national local government system' - which incorporates not only the national community and the policy communities for local government services but also the territorial intergovernmental networks - and this set of organizations defines the national role and state of opinion in local government as a whole (Dunleavy, 1981b: 105; Dunleavy and Rhodes, 1983: 121-2; Rhodes, 1986: 31-2, 36, 416). It is a key means by which local government can convey a wide variety of different views to Whitehall and it also provides a framework within which any individual local authority can situate its problems, concerns and strategies. Local authority actors do not decide policies for their area in isolation; they look to the national local 206 R.A.W. Rhodes government system for guidance about what standard of service to provide, for ideas to imitate or avoid, for ways of tackling common problems and for justifications or philosophies of particular strategies. Some councils are innovators across a wide field of policy, but they are rather exceptional. Most councils of the time follow national trends in the local government world, or national trends in their kind of authority facing their kind of general problem under their kind of political control. Each of them will innovate from time to time in one issue area or another, adding their own small contribution to the national picture. But most of the time local decisions are made within nationally defined parameters of what counts as good policy, rather than helping to redefine those parameters. If policy networks represent the all-pervasive functionalism in the organization of British government, then the national local government system is a mechanism of ideological integration. Producer networks are distinguished by the prominent role of economic interests (both the public and the private sector), in policy-making; their fluctuating membership; the dependence of the centre on industrial organizations for delivering the desired goods and for expertise; and the limited interdependence of the economic interests. Thus Tivey (1982) describes the development of the Nationalized Industries Chairmen's Group, its links with the Treasury and how it uses its knowledge of its industries to compete with the private sector for government resources. However, the effectiveness of the Group is constrained by competition between its members: competition which extends beyond the distribution of public money between the industries to the market-place and the sale of their respective products. Similarly, Dunleavy (1982: 191 and 192 fig. 11.2) suggests that private industry has been a major influence on the development of policy in the nuclear network - a network in which, for example, GEC is firmly embedded. Of late, the analysis of economic producer groups has been dominated by corporatist theory. Apart from the conceptual inadequacies of this theory, it has not fared well when applied to policy-making in British government. Leaving aside the bi-partite 'Social Contract' - the archetypical case which has been over-cited and remains a bad example -the case of industrial policy offers little solace. Thus, Hogwood (1979) and Grant (1982) argue that an industrial policy community exists but its boundaries are imprecise and, in spite of a degree of informal contact, it remains loosely integrated. Clarity is served if this network is distinguished from the highly integrated policy communities, but it is clear that producer networks have few if any corporatist characteristics. It is possible that the concept of corporatism could retain some utility but only if limited to government-imposed integration/regulation: 'state corporatism' in Schmittens (1979: 20) formulation. When so restricted, it does at least refer to a specific type of network relationship. The distinction between the public and private sector does not refer solely to industry; it is' also relevant in the analysis of professions. Policy networks and sub-central government 207 Dunleavy (1982: 193-5) argues that, when a profession is split between the public and the private sector, the latter tends to have a higher status within a unified profession. When the public and private sectors work together, for example in research and development, the flow of influence will be from the professionals in the private sector to those in the public sector. A profession can be a key channel of influence for the private sector. Thus: the concentration of nuclear engineers in these governmental bodies (i.e. UK Atomic Energy Authority, Atomic Energy Commission of America) working very closely with nuclear power plant manufacturers, has distorted their conceptions of the public interest on nuclear power. (Dunleavy, 1982: 197) In yet another form, professional influence emerges as a key element in policy networks. The variety of networks is potentially much greater than the examples discussed above. However, the most important conclusion to be drawn from the examples concerns the need to compare networks. There is no one pattern of relationships for all policy areas. The definition of networks and the discussion of characteristics and types have suggested a basis for such comparison and illustrated the known variety. Two topics remain to be explored: the relationship between networks and the national government environment and relationships within and between networks. To this point, the analysis of networks has been static, an exercise in definition and typology. It is also necessary to explain changes in the context and in the relationships of policy networks. Policy networks and the national government environment Within a unitary institutional structure, the centre is the fulcrum of policy networks. Allied to the tradition of executive authority, central government cannot be treated as one more group; its role is constitutive. It can specify unilaterally, substantive policies, control access to the networks, set the agenda of issues, specify the rules of the game surrounding consultation, determine the timing and scope of consultation, even call a network into being. Whilst it may prefer, and on occasion be constrained, 'to create a nexus of interests so that co-operation flows from a sense of mutual advantage' (Richardson and Jordan, 1979: 105), it retains the option of coercion. Through the substantial resources it controls, the centre has the luxury of choice between the many available strategies. Policy networks are not necessarily a constraint on government but can be manipulated by government in its own interest; the relationship is asymmetric. This general point to one side, it is necessary to recognize that the centre has a multiplicity of interests. Policy networks may be based on a 208 R.A.W. Rhodes department or even a section of a department, each of which can have a distinct style. Relationships within a network are shaped by the 'depart-1 mental philosophy' or the 'store of knowledge and experience in the subjects handled, something which eventually takes shape as a practical philosophy' (Bridges, 1971: 50). This observation is unremarkable, but it is difficult to explain variations in style if the central department is treated as a unitary. If the era of the 'giant department' (Clarke, 1971) has passed, none the less large, multi-functional departments persist. Thus, the Department of the Environment (DoE) is composed of major divisions including (at various times during the 1970s) water, transport, local government, housing, planning and construction. By no means all of these divisions are at the heart of a function-specific network, but equally there is no single DoE policy network. It is inadequate, therefore, to search for a departmental philosophy. It is also necessary to search for variations within departments to determine whether or not a single department has several distinct styles: a possibility rendered all the more probable when it has been created from several previously separate departments. The separate organizational arrangements devised for transport and construction during the 1970s attest to their distinctiveness. The terms 'central government' and 'the centre' have to be understood, therefore, as shorthand for a diverse collection of departments and divisions. It is only to be expected that this diversity is matched by the range of interests within central government. At its simplest, it is possible to distinguish between the 'guardians', or the Treasury, concerned to restrain public expenditure and 'advocates' or the service spending departments (Wildavsky, 1975: 7). However, a further two distinctions are necessary. The 'advocates' comprise those departments (and policy networks) which have a direct involvement with the services of SCG and those which have no such involvement. The latter will be at least neutral in, for example, any argument with the Treasury involving local expenditure, and more probably they will have a healthy interest in local authorities bearing the brunt of any reductions in expenditure. Last but by no means least is the DoE, which, as the areal department responsible for local government, acts both as guardian in the negotiations about government grant and as 'advocate' for spending on those services for which it has responsibility. And this characterization of the interests within central government is general, omitting the particular interests associated with, for example, a specific policy initiative. To the range of policy networks, therefore, it is necessary to add a parallel and profuse range of interests. Second, the analysis of policy networks presupposes that they have a key impact on policy content. However, as Lowi (1972) has argued, the 'politics determines policy' axiom can be turned on its head; 'policies determine polities'. It is no mere coincidence that the Home Office, responsible for policies on police, fire and prisons, should be repeatedly characterized as authoritarian, secretive and directive. Lowi's reversal of conventional axioms has'the virtue of pungent argument but the problem Policy networks and sub-central government 209 of overstatement; policy is both a dependent and an independent variable. But leaving such complexities to one side, it is clear that the analysis of policy networks cannot be limited to an analysis of process; it must encompass policy content. The second feature of the British political tradition which conditions * the operation of policy networks is the two-party system. Ministers face in two directions. They are the heads of the bureaucracies at the heart of the policy networks but they are also the leaders of the majority party. Policy networks have not supplanted party political channels of communication and influence. Party is at times a complementary and at other times a rival channel of influence. The effects of party are pervasive. It spans levels of government and communicates a range of interests. Most important, it spans the policy network. If policy networks are closed, then party is one of the means for prising them open. Rhodes (1986: 387-9) concluded from his study of the national community of local government that party was the grit in the molluscs of Whitehall-based policy networks, capable of stimulating change. Of course, British government cannot be reduced to the simplistic duality of party versus bureaucracy. But the fluctuating relationship between the two is central to understanding the sources of inertia and innovation in the policy process. I . J The emphasis throughout the post-war period has fallen on the extension of functional politics at the expense of territorial representation. The dual polity was created, local elites were marginalized, and uniform service provision prevailed over regional/national differences. These developments are only half the story. Paradoxically, the extension of functional politics also served to politicize SCG. Policy networks may reflect many features of the national government environment but they also changed that environment. The most obvious reactions to the extension of functional politics were the re-emergence of nationalism and the emergence of the topocratic professions and the intergovernmental lobby to counter the influence of the technocrats and the function-specific policy networks. An intermediate tier of representation supplanted direct contact with local political elites but functional politics also led to the modernization of SCG and the attendant spread of party politics. It generated sectoral cleavages and contributed to class de-alignment. SCG became the locus of conflicts rooted in multiple social cleavages, and the politicization of local government began to pervade central - local relations with the onset of economic decline. SCG politics became the politics of Westminster and Whitehall. The extension of functional politics was an important factor in the erosion of the dual polity and the politicization of SCG. A number of features of this trend warrant further comment. First local government witnessed the revival of municipal socialism in new clothes.3 The 'new urban Left' (Gyford, 1983a, 1983b, 1984, 1985) rejected the legacy of a centralized, reformist socialism. As Beer (1982: 167) notes, the riiS-— w. 210 R.A.W. Rhodes Labour Party was 'wholeheartedly democratic, but the democracy to which it adhered in theory and practice was not participatory, but ': ' deferential, representative, indirect and centralized. Populism was as foreign to it as localism and individualism.' The romantic radicalism of the new urban Left rejected the responsibility ethic as but deference in a different guise. Rather local government was to be the means for resisting the 'cuts' but also an example of what socialism could achieve. The bases of support for this programme of radical activism were diverse, encompassing party and community activists, radical elements in the local government professions and socialist councillors. The 'new alliance' embraced the women's movement, black organizations, environmentalists and CND: indeed, the spectrum of social movements with their origins in the 1960s (Boddy and Fudge, 1984a: 7-9). And local government's 'responsibility ethic' was anathema to the new urban Left. The politics of confrontation saw new stars in the firmament with Ken Livingstone and David Blunkett becoming national figures. Conflict over the GLC's 'fare's fair' policy and the Liverpool budget were not isolated incidents but illustrations of the new style in SCG. Furthermore, the landscape of local politics changed markedly with the rise of the SDP/Liberal Alliance. After the 1985 shire county council elections, twenty-five out of forty-six English and Welsh parties had no overall control. The Alliance was the largest single party on two councils and formed the minority administration on five councils. As yet this change has had its most marked effects on council procedures; but, with the onset of the budgetary process: The newly hung counties can certainly expect a new period of uncertainty, with protracted negotiations, an increasingly delicate officer role in terms of confidential briefings and information distribution, and committee and council meetings of quite unprecedented length. (Leach and Stewart, 1986: 15) The rise of the Alliance in local government not only fosters its parliamentary aspirations but destabilizes local politics at a time of unprecedented instability and holds out the prospect of complex coalition politics. Third, politicization was not a feature of local government alone. Public sector unions, disillusioned by fifteen years of pay policy, reacted angrily to the 'cuts' in public expenditure, privatization and a government which sought to limit drastically union power. The ever-present threat of unemployment may have exercised a restraining influence on some unions, but militancy was the order of the day in the NHS, the nationalized industries and the civil service. The government confronted, therefore, an increasingly turbulent sub-central system, but the policy networks, as part of the national government environment, now constrained the ability of the centre to respond to the changes in SCG. The very existence of the networks caused certain policy-making processes and outcomes. A product of the welfare state, Policy networks and sub-central government 211 Table 1 Growth of public expenditure by programme in cost terms % changes 1980-81 to 1983-84 Defence 14.2 Overseas aid and services 4.4 Net payments to EEC 83.9 Agriculture 23.9 Industry, etc. -3.8 Arts 6.0 Transport -5.1 Housing -49.3 Other environmental services 0.6 Law and order 19.0 Education 0.0 Health 5.8 Social security 25.5 Other public services -16.1 Common services 29.9 Scotland 3.3 Wales 0.0 N. Ireland 7.9 Asset sales -63.5 Planning total 7.0 Planning total, excl. asset sales 7.7 Planning total, excl. asset sales and net sales of land and buildings 8.4 Net interest 26.7 Total expenditure, incl. interest 9.2 Note: Figures are adjusted for reduction in National Insurance surchange and changes in treatment of housing and sickness benefits and of Property Services Agency. Source: Ward, 1984, table 7, p. 6. they had a vested interest in, and helped to fuel, its continued expansion. In an era of economic decline, they resisted political pressure for cuts: a bulwark of inertia. As the centre sought to control SCG, its bureaucratic strategies foundered on the disaggregation of policy systems, politicization and the multiplicity of interests in and of the centre. Thus, local government current expenditure rose in real terms between 1979 and 1983 (see HMSO, 1983). Nor was this pattern exceptional. Total public expenditure continued to rise and in spite of repeated cries of anguish, NHS expenditure rose by 14 per cent in real terms in the same period (Social Trends, 1985: 122). Indeed, as Ward (1984: 26) demonstrates, only housing of the major welfare services experienced a 'cut' in expenditure in real terms (see Table 1), although a focus on resources is unable to demonstrate whether or not there has been marked deterioration in service levels. Thus, although the NHS had a 17 per cent increase in volume expenditure (1979--84), this figure reduced to 7.2 per cent (4 per cent for hospitals, etc.) when the relative price effect (or higher costs of the NHS) was taken into 212 R.A.W. Rhodes account. It was further estimated by the Department of Health and Social Security that demographic and technological pressures required an increase in expenditure of approximately 6 per cent. Consequently, the hospital and community health sector experienced problems in meeting demands (Social Services Committee, 1984: x-xi). As the committee commented, the NHS needs to live at the same rates of pay and price inflation as the rest of us. None the less the alleged dismantling of the welfare state remains some way off. Policy networks have become as central a feature of the national government environment as some of the hoary old chestnuts of constitution, less prominent and debated but a more determinant influence. They lie at the heart of one of the major problems of British government: policy messes, or the non-correspondence of policy systems and policy problems. The failure to appreciate that service delivery systems are complex, disaggregated and indeterminate has led to the failure of policies. The process of differentiation in government requires not only policies on substantive problems but also policies on the procedures for managing differentiation (or institutionalizing indeterminacy). Moreover, these comments are a critique not of functional differentiation in itself but of the failure to recognize that it is a central feature of the policy process; substance and procedure have to be endlessly traded off in the internally differentiated or pluralized system of SCG. Policy networks in all their variety are a defining characteristic of SCG. Exploring this variety requires an examination of relationships within networks, of the process of exchange and the rules and strategies governing resource transactions. [. . J Notes 1. This classification is an empirical one, restricted to SCG. Benson, 1982, pp. 154-8, distinguishes between all governmental networks in terms of their 'types of structural interests': i.e. demand (or client) groups; support groups (which provide needed resources for the public sector organizations); administrative groups (or those occupying positions of administrative control); provider groups (which deliver services); and co-ordinating groups (or those responsible for rationalization within and between programmes). Networks will vary, therefore, as the configuration of interest varies. There are a number of problems with this approach. First, the constellation of interests is only one relevant dimension of network structure. Second, in the context of British SCG, it is difficult to distinguish demand from support groups and administrative groups from provider groups. Third, economic functional groups are omitted as such, forming part of (presumably) support or provider groups. Equating manufacturing industry with either environmental groups on the one hand or the medical profession on the other seems unhelpful. However, given the current state of research on networks any classification must be treated with caution. The listing employed here is tentative. 2. Since this was written the regional water authorities have been privatized and are now run as private companies. ledsl 3. Although space precludes a detailed discussion, my survey of the available theories omits Policy networks and sub-central government 213 a large but diffuse socialist literature. A useful preliminary survey is provided by Sancton, 1976, and some more recent contributions are summarized and evaluated in Gyford, 1985. Not only is there a need for a historical account of socialist thinking on decentralization, the areal division of powers and local government, but there are also a number of more specific gaps - for example, Herbert Morrison and Harold Laski's , democratic centralism and its implications for local government. The liberal theory of local government has been,better served by commentators than the socialist theory, and it is time that the inequity was redressed. My thanks to Peter Richards (Southampton) for prompting these reflections. References Barker, A. (1982) Quangos in Britain. London: Macmillan Beer, S.H. (1982) Britain Against Itself. London: Faber & Faber. Benson, J.K. (1982) 'A framework for policy analysis', in Rogers et al. 137-76. Boddy, M. and Fudge, C. (1984a) 'Labour councils and the new left alternative', in Boddy and Fudge: 1-21. Boddy, M. and Fudge, C. (eds) (1984b) Local Socialism! London: Macmillan. Bridges, Sir Edward (1971) 'Portrait of a profession', in Chapman and Dunsire: 44-60. Chapman, R.A. and Dunsire, A. (eds) (1971) Style in Administration. London: Allen & Unwin. Clarke, Sir Richard (1971) New Trends in Government. London: HMSO. Connolly, M. (1983) Central-Local Relations in Northern Ireland. (Final Report to Social Science Research Council). Drucker, H., Dunleavy, P., Gamble, A. and Peele, G. (eds) (1983) Developments in British Politics. London: Macmillan. Dunleavy, P. (1981a) 'Professions and policy change: notes towards a mode! of ideological corporatism',Public Administration Bulletin, 36: 3-16. Dunleavy, P. (1981b) The Politics of Mass Housing in Britain, 1945-75. Oxford: Clarendon Press. Dunleavy, P. (1982) 'Quasi-governmental sector professionalism: some implications for public policy-making in Britain', in Barker: 181-205. Dunleavy, P. and Rhodes, R.A.W. (1983) 'Beyond Whitehall', in Drucker et at. (eds): 106-33. Grant, W.P. (1982) The Political Economy of Industrial Policy. London: Butterworths. Gray, C. (1982) 'The regional water authorities', in Hogwood and Keating: 143-67. Gyford, J. (1983a) 'The implications of local socialism', Local Government Studies, 9(1): 13-17. Gyford, J. (1983b) 'The new urban Left: a local road to socialism?', New Society, 21 April: 91-3. Gyford, J. (1984) 'From community action to local socialism', Local Government Studies, 10(4): 5-10. Gyford, J. (1985) The Politics of Local Socialism. London: Allen & Unwin. Heclo, H. (1978) 'Issue networks and the executive establishment', in King: 87-124. HMSO (1983) Rates. London: HMSO. Hogwood, B. (1979) 'Analysing industrial policy: a multi-perspective approach', Public Administration Bulletin, 29: 18-42. Hogwood, B. and Keating, M. (1982) Regional Government in England. Oxford: Clarendon Press. James, M. (1982) 'The Welsh local authority associations', paper presented to the Conference of the Political Studies Association, Queen's University, Belfast, September. Keating, M. and Midwinter, A. (1983) The Government of Scotland. Edinburgh: Mainstream. 214 R.A.W. Rhodes Keating, M. and Rhodes, R.A.W. (1981) 'Politics or technocracy? The regional water authorities'. Political Quarterly, 55: 78-84. 'King, A. (ed.) (1978) The New American Political System. Washington, DC; American Enterprise. Leach, S. and Stewart, J.D. (1986) The Hung Counties. Luton: Local Government Training Board (mimeo). Lowí,;T.J. (1972) 'Four systems of policy, politics and choice', Public Administration Review, 32: 298-310. Rhodes, R.A.W. (1981) Control and Power in Central-Local Government Relations. Alder-shot: Gower. Rhodes, R.A.W. (1985) 'Corporatism, pay negotiations and local government', Public Administration, 63: 287-307. Rhodes, R.A.W. (1986) The National World of Local Government, London: Allen & Unwin. Richardson, J.J. and Jordan, G. (1979) Governing under Pressure. Oxford: Martin Robertson. Rogers, D., Whitten, D. and Associates (eds) (1982) Interorganizational Coordination. Ames, Iowa: Iowa State University Press. Sancton, A. (1976) 'British socialis,t theory of the division of power by area', Political Studies, 24: 158-70. Saunders, P. (1983) 'The "regional state": a review of literature and agenda for research', Paper presented at the Sociai Science Research Council conference on Political Theory and Intergovernmental Relations, Nuffield .College, Oxford, September. Schmitter, P.C. (1979) 'Still the century of corporatism?', in Schmitter and Lehmbruch: 7-52. Schmitter, P.C. and Lehmbruch, G. (eds) (1979) Trends towards Corporatisi Intermediation. London: Sage. Social Services Committee (1984) Public Expenditure on the Social Services, Fourth Report, Session 1983-4. HC395. London: HMSO. Tivey, L. (1982) 'Nationalized industries as organized interests', Public Administration, 60: 42-55. Travis, A.S., Veal, A.J., Duesbury, K. and White, J. (1978) The Role of Central Government in Relation to the Provision of Leisure Services in England and Wales. Birmingham University, Centre for Urban and Regional Studies, Research Memorandum 86. Treasury and Civil Service Committee (1984) Third Report. The Government's Expenditure Plans 1984-85 to 1986-87. Session 1983-4, HC285, London: HMSO. Ward, T. (1984) Memorandum by Mr. Terry Ward, specialist adviser to the committee, in Treasury and Civil Service Committee: Appendix 1. Wildavsky, A. (1975) Budgeting: A Comparative Theory of the Budgetary Process. Boston: Little Brown. 18 Taking and giving: working women , and female bonds in a Pakistani immigrant neighbourhood1 Pnina Werbner Women-centred networks The focus on practices of female seclusion and veiling in Islamic societies has sometimes obscured the importance of extra-domestic networks sustained by women within their 'separate world'. Yet such extra-domestic, women-centred networks have important bearings on gender relations, conjugal roles and the external support women can draw upon. Purdah, as Papanek points out in a seminal paper (Papanek, 1973), is both a system of task allocations and an expression of male and family status. In the latter sense, purdah is non-complementary. It rests on the conception of an active male, an achiever in the public domain, and a passive female, secluded within the domestic domain, the object of male protection. [. . .] Once we examine the 'world of women' not simply as a world of domestic chores or idle gossip, but as the complex world of extra-domestic female relationships, we are able to shift from the presentation of purdah as a cultural logic to a sociological analysis of variations in conjugal relations as these obtain in purdah societies. As Rosaldo hypothesized at the outset of the current debate on gender relations: Women's status will be the lowest in those societies where there is a firm differentiation between domestic and public spheres of activity and where women are isolated from one another and placed under a single man's authority, in the home. Their position is raised when they can challenge those claims to authority, either by taking on men's roles or by establishing social ties, by creating a sense of rank, order and value in a world of their own. (Rosaldo, 1974: 36) In Manchester, Pakistani migrant women living in the central residential enclave initiate and sustain widely ramifying women-centred networks. Through the contacts they forge with other women, they extend the family Adapted from S. Westwood and P. Bhachu (eds) Enterprising Women: Ethnicity, Economy and Gender Relations (London: Routledge, 1988), pp. 177-202.