Going Interorganisational: Weber Meets Durkheim

Political scientists, including public administrationists, have less enthusiastically taken to organisational and interorganisational concepts than have sociologists and students of management science. We shall offer some penetrating reasons for this abstention. Political science has had its engagements with 'organicism', which may have made large parts of the interorganisation theory literature less attractive, depending, as it does, on definitions of organisational behaviour in terms of consensus on values and goals. But the question is then: why have political scientists, and notably public administrationists, not tried more vigorously to infuse interorganisational research with more political and less organic conceptions?

To ask this question is to beg a prior one. Are interorganisational phenomena important enough to warrant a place in political science research in the first place? We would contend that they are, and especially in that sub-discipline of political science policy analysis called implementation research. There, we would argue, the incorporation of interorganisational perspectives into theory and research is a necessity which comes close to being a hallmark (Hjern & Hull 1982; Hjern 1982). Its interorganisational perspective is the cachet which sets off implementation research from policy analysis in a more behavioural public administration vein.₁

---

* The article was originally prepared as a paper for the 1983 Annual Meeting of the American Political Science Association, Chicago, September 1-4, 1983.

Once a political science interorganisational perspective is adopted, the distinctive remit of implementation research becomes more readily apparent and some of the lingering confusions which have hampered the disciplinary progress of implementation research can be dealt with. The 'Wildavsky problem', for example, a conundrum of long-standing in the short life of latterday implementation research, can be resolved once the need to go interorganisational is acknowledged. Going interorganisational implies, however, a certain break with previous research practice. The paper discusses, in particular, the need to abandon certain methodological devices which much self-styled implementation research has borrowed, without adequate reflection, from public administration.

The paper begins by arguing that implementation research must go interorganisational if it is to recover and develop the potential it once held for political science. There follows a discussion of some methodological issues involved in a political science interorganisational analysis. An approach is then outlined which does not rely on modern interorganisational theory; it draws instead on Max Weber's social interaction method of analysis and derives its hypotheses from Emile Durkheim's contribution to political theory.

A Defense of Implementation Research Against Its Friends

The advent of implementation research in the early 1970's derived from a renewed attempt to take a fresh look at an old theme in the study of politics: how well does the body politic link good representation of societal aspirations ('politics') with their efficient and effective realisation ('administration'). This aspiration of the emergent subdiscipline soon began to go astray, as two of its early proponents noted (Majone & Wildavsky 1978). In an outburst of selfstyled implementation research, the original conception was dismissed by some, distorted through adaptation by others and generally 'annexed' by public administrationists and policy output analysts.2 Before long the distinctive remit of implementation research was lost.

The editors of a recent reader define implementation as 'the carrying out of a basic policy decision, usually made in a statute (although also possible through important executive orders or court decisions)'. For them, 'the crucial role of implementation analysis is to identify the factors that effect the achievement of statutory objectives throughout this entire process' (Mazmanian & Sabatier 1981, 5-6). If such is implementation research, is it not just (behavioural) public administration policy analysis by another name?

One aim of this paper is to start a discussion about the meaning and place
of implementation research in political science — in particular with regard to
behavioural public administration research. The essential argument is that the

interests of political science as a whole require that implementation research — in a meaning which will become apparent — be recognised as an integrative element in a discipline which tends to divide its research attentions between the input and output sides of politics. Thus the remit which we propose for implementation research is its original ambition to explore the links between 'politics' (input) and 'administration' (output). Its integrative role in the overall discipline should be self-evident.

Implementation research in our sense has an evaluative (policy analytic) perspective. It also implies an explicit organisational perspective with its stress on politico-administrative linkages. An evaluative and a certain organisational perspective are, of course, also characteristic of behavioural public administration research. But the perspectives are not the same. The focus of public administration is the relationship between the result of 'politics' — statutes, executive orders and court decisions — and the administrative processes leading to policy outputs or outcomes. There is thus a basic assumption of a stable and sequential relationship between 'politics' and 'administration' as if the intentions of policies can be achieved (at least in significant measure) by the mere addition of clause to legislative clause. Implementation research, by contrast, needs to adopt an initially agnostic attitude as to the nature of the relationship between 'politics' and 'administration'; it attaches particular importance tö the manner in which organisation is added to organisation as the means of linkage between 'politics' and 'administration'.

Their different inter-organisational perspectives is one important feature distinguishing public administration and implementation research approaches. Moreover., the differences are not merely conceptual but also — and crucially so — methodological. In sum, there are, in our view, distinctive public administration and implementation research mind-sets. The difference between them will become clearer in the following discussion of the different approaches to policy analysis taken in implementation and public administration research.

Policy Analysis and the Study of Institutionalisation

Policy may be defined as a set of ideas (goals) and the practical search for institutional arrangements for their realisation. Policy analysis we take to mean describing and evaluating institutions in respect of their contribution to the reliability of policy realisation. Policy analysis, in this definition, may use either a public administration or an implementation research approach. Common to both approaches will be an evaluative or impact analytic perspective, since the aim is to assess the contribution of institutional arrangements to the reliability of policy realisation. Where the two approaches

may differ is in respect of whose ideas (goals) are evaluated and of the particular
institutional arrangements to which outcomes are related.

It is, of course, a truism that political science is about studying institutions, but it is one which bears repetition and reflection when discussing political science policy analysis. Besides general criticism for not having delivered the goods promised on the eve of the revolution, behavioural political scientists are charged specifically with having lost their feel for institutions (cf. Johnson 1975; Lowi 1979), by which latter are meant in the main the formal, juridical and constitutional aspects of the body politic. Anti-formalism, i.e., replacing the legal study of institutions by the examination of process, behaviour and resourcing in and among political groupings, has certainly been a major trend in recent behavioural political science (Wrong 1979, 166). Anti-formalism has broken the grip once held on the discipline by constitutional analysis pure. However, the ensuing interest in the characteristics of the 'pluralist' society opened the door for another formalism in the guise of an explicit — in policy analysis more often implicit — reliance upon mandated organisation as the major ordering principle in theory and research.

Resort to mandated organisation as the ordering principle corresponds to a conceptual and methodological perspective which we would describe as methodological constitutionalism. It is 'constitutionalism' in the sense that the organisations formally accorded responsibility for administering a particular public policy — whether that responsibility is assigned by the written constitution as such or, at one remove, by institutions intended by the constitution for this purpose, e.g., Parliament — are accepted by the researcher as being sometimes the sole but always super-ordinate institutions contributing to the realisation of policy. What this means is that the constitution and its provisions are interpreted not so much as a prescription for a preferred social order than as a description of institutional reality. We would argue that methodological constitutionalism is the predominant principle of research design in public administration policy analysis and, as such, is a characteristic which sets it apart from our understanding of implementation research (cf. the discussion in Hjern & Hull 1982). It is because of its methodological constitutionalism that public administration takes public formal organisations as its units of analysis as unquestioningly as the written constitution was seen as the proper unit of analysis at the beginning of the century.

If policy analysis is to find the institutional arrangements — including potentially organisations unmandated as well as mandated, private as well as public — which increase the reliability of policy realisation, methodological constitutionalism needs to be eschewed. In these wider and less homogeneous sets of organisational arrangements, more attention than hitherto will need to be given to the links between the participating organisations, since it is these links which hold the whole together. And given the remit of policy analysis as

the study of how institutional arrangements contribute to the reliability of
policy realisation, a key issue for theory and research must be the question of
how 'good' institutional arrangements can be institutionalised.

Thus we would argue that the institutionalisation of interorganisational linkages with respect to the reliability of policy realisation should be the core concern of implementation research. The political institutionalisation literature of recent years has tended, by contrast, to focus on structural integration within single (incipient) institutions or organisations (Polsby 1968; Huntington 1969; Thomas et al. 1972) and would seem far too restrictive for application in the analysis of public policies which, so often, involve activities by many actors from many organisations and, indeed, parts of organisations. In particular, the structural integration approach seems to stipulate that an organisation needs to establish relations of hierarchy vis-å-vis the actors in its environment before viable interorganisational links can be institutionalised (Hill 1974, 1076). We shall argue that social interaction analysis provides a more appropriate way of researching whether and how links between organisations are forged and institutionalised. If that should seem modish or faddish, we shall later argue that it is an approach in keeping with ideas about institutionalisation advanced by Weber and Durkheim.

A Defence of Interorganisational Analysis Against Its Critics

Interorganisational analysis is not in the mainstream of political science research methodology; indeed, the concept appears in the titles of just two APSR articles published between 1970 and 1980 (Perry & Levine 1976; McCormick & Kihl 1979). There may be theoretical reasons for this; if organisation theory strikes an acute observer as an elephantine body (Waldo 1978), interorganisational theory would seem to require the daring attempt to straddle simultaneously several wide-bodied animals of the lumbering, encumbering variety. There may also be normative reasons; given that public administrationists are used to creating a 'value-free' platform for their policy analysis by painstakingly reconstructing the authoritatively binding public goals explicitly or implicitly intended by parliamentary decisions, what should they make of a style of analysis which requires them to incorporate still further organisations, some without any formal mandate, some with goals quite divergent from those intended by parliament, and some which may even be incorporated under private rather than- public law? Lastly, there may be methodological reasons, well captured in the passage: 'we have become infatuated with the art/science of organisational "tinkering'¹ without realising that, in fact, the conventional notion of organisation may not be relevant any more' (Gawthrop 1978, 923).

If we are being misled by the simple concept of organisation, any interorganisational
conception can only be so much more misleading.

All three of these reasons for why political scientists (as opposed to organisation analysts in sociology and management science) eschew an interorganisational perspective are worth pursuing further. Brevity requires that we concentrate here on the methodological reasons which directly concern our main topic: the distinction between the behavioural public administration and implementation research modes of policy analysis. Besides, it will be evident from the methodological discussion that organisation theory is not alone in suffering an elephantine problem; much public administration research is riding no less of a Proboscidean mammal. It can also be inferred from the methodological issues that the manner in which behavioural public administrationists in effect circumvent certain central normative issues is only feasible because the canons of replicability and falsifiability in scientific enquiry are (mutely) sidestepped.

We fully take the argument that there are serious methodological problems in wanting to develop an analysis of interorganisational linkages from some concept of 'the organisation'. Our own view is that for the analysis of the institutionalisation of interorganisational linkages the group dynamics of individuals (representing organisations) are probably as important as the strategies of organisations per se (Ohe 1977). Methods for detecting the interpersonal relations associated with interorganisational linkages may prove a less infatuating object of study than organisation as such.

Implementation Research: Going Interorganisational

It has been suggested that it is sufficient for developing a theory of organisations simply to assume that organisations exist, inasmuch as we accept that we recognise them when we see them (March & Simon 1958). This definition of organisation is clearly far too sanguine for use in critical research, for it will inevitably lead to more or less arbitrary operational definitions of organisation. Hence it conflicts with the principles of both replicability and falsifiability in scientific enquiry. It would seem to be not uncommon that these canons of enquiry are violated in organisational — and, by extension, interorganisational — analysis. This is at least true of studies reported in two major journals (ASQ and HBR³), in which it is commonly simply assumed that organisations are distinguishable from their environments and that particular people and phenomena can be readily identified as being inside or outside the organisation (Dunbar 1983, 132, 135).

The problem to which the previous paragraph refers is the problem of

bounding one's unit of analysis. It is a problem which has received some recent attention from organisation theorists (Starbuck 1976), but which is little discussed in political science. In most public administration research the delineation of the unit of analysis is usually taken to be self-evident and so to require little if any reflection: a single agency or type of agency is chosen (e.g., the environmental protection administration). The choice of unit of analysis is selfevident inasmuch as public administration research adheres to the precepts of methodological constitutionalism. Thus the unit of analysis is typically the 'mandated' organisation.

Implementation research, by contrast, should specifically eschew metholological constitutionalism in defining its unit of analysis. Its remit is empirical constitutionalism (Hjern & Hull 1982), i.e., to recast the prescriptions of the written constitution as hypotheses to be tested against the institutional reality of 'the organisational society' (Drucker 1969, 159-75). In the organisational society, where most public activity involves actors from several organisational units, the institutions which connect policy intentions, actions and results have almost certainly only rarely become institutionalised by sole virtue of public mandates, nor need the patterns of linkage correspond to what the public mandates prescribe. The possibility of such 'unconstitutional' behaviour needs to be reflected in research design and, in particular, in the choice of unit of analysis.

Until implementation research began to break loose from methodological constitutionalism with Pressman & Wildavsky (1973), rather crude distinctions were often made in political science policy analysis between policy intentions, policy action and policy results. Much early policy output analysis, for example, resorted implicitly — to use a metaphor from Max Weber — to a 'tree-felling' reasoning about these interrelationships. No distinction was made between action and result; rather, action was taken to include not only the behaviour of actors (which, however, was not directly observed in research) but also the intended results of that behaviour (which is what was measured). (Our discussion here closely follows Mclntosh 1983.) This conflation of action and result is plausible only if the means of action are fully controlled by the actor. It seems reasonable enough to view the axe as a dispassionate extension of the lumberjack's arm; it would seem rather difficult to make a similar assumption about any particular actor's control over the instruments of some public policy.

From the beginning, the relationships between policy intention, action and result presented a logical conundrum to implementation researchers which is sometimes referred to as the Wildavsky problem (Lane 1983). It is clearly evident in the following quotation from Pressman & Wildavsky (1973, xiii-xiv):

'(But) policies normally contain both goals and the means for achieving them. How, then, do we distinguish between a policy and its implementation? ... Both these meanings of policy rule out the possibility of studying implementation ... We can work neither with a definition of policy that excludes any implementation nor one that includes all implementation!

The quotation reflects a realisation that the individualistic, action-result conception carried over from policy output analysis and lumber jacking is clearly at odds with the reality confronting the researchers. Yet they can see no way out of their dilemma. The way out — and the perspective which implementation research needs to adopt — is to recognise that the relationship between action and result is not the sequential, stable relationship between an activity called 'politics' and one called 'administration', but a complex symbiosis involving many actors. It is because implementation is social interaction that it cannot be the simple extension (arm plus axe) of any one intent. The intent (and behaviour) of any actor — be it parliament, president or judge — is but one factor in the nexus of social interaction out of which policy results are produced.

Implementation research requires studies designed to perform policy analysis of the intentional actions of multiple actors in relation to the ensuing policy results. Because the multiple actors will likely each be pursuing more or less different objectives, it necessarily falls to the researcher consciously to define the policy problem around which his study will be designed. There is no reason why the researcher should not state normative grounds for his choice. The 'objective' strategy of reproducing the enacted policy intent of government or parliament itself involves a value bias and is in no less need of justification (see Bay 1965; Strauss 1962, 326). The methodological imperative of implementation research, working with a model in which the formulation and realisation of policy unfolds in a collectivity of actors in social interaction, then implies less a choosing than a discovery or reconstruction of the proper unit of analysis.

The researcher sets out to reconstruct what actors — with what objectives, strategies and resources — are part of the collectivity. It is crucial that reconstruction of the unit of analysis be pursued by methods which allow replicability and thereby falsifiability. Different researchers should, in principle, under the same general research conditions, be led to reconstruct compositionally similar collectivities for the same policy problem. Moreover — anticipating once again our later references to Weber — the patterns of social interaction so reconstructed need to be described and analysed with as little distortion as possible of the meanings which the actors themselves attach to them. This, in our view, is the sine qua non of implementation research.

Interorganisational Implementation Research: Weber Meets Durkheim

Implementation research can use social interaction analysis to reconstruct the collectivity of individuals which institutionalise linkages between organisations. Thus the organisations engaged in policy formulation and realisation in relation to some problem are identified through the patterns of social interaction among the participating individuals. This, we shall argue, is a very Weberian notion in so far as it is the interactions chosen by individuals which lead us to hypothesise the existence of a collectivity.4 We refer to this collectivity as an 'implementation structure'.

Weber was the declared! enemy of what has been called a form of 'naive collectivism' (Popper 1963, 341), the assumption that social wholes exist empirically. He criticised the use of social wholes as organic wholes rather than as scientific constructs (for the distinction, see Oppenheim 1964, 343). An implementation structure is an analytic construct, a unit of analysis, and has very little in common with organic conceptions of organisation (for a critique of the latter, see Scott 1979)., An important distinction is that whereas organisations presumably 'just' have to be picked by the researcher, implementation structures first have to be reconstructed by the researcher on the basis of how individuals self-select themselves and engage in social interaction with respect to the policy problem under investigation.

The major difference between implementation structures and organic notions of organisation, however, concerns the motives for individual participation. It is at this juncture that the Weberian notions central to the implementation structure approach begin to come close to Durkheimian conceptions. The Durkheimian conceptions introduce a 'social' element into what is otherwise an essentially individualistic perspective.

Much recent literature considers that organisational units attain (internal) participation mainly through their members' consent to a domain of major values and goals (admitting that side-payments of various kinds are also usually mentioned). The 'domain consensus' idea has since been extended to interorganisational situations. This super-ordinate goal approach (March & Simon 1958; Thompson 1967) relies, of course, on an integrating mechanism very similar to one of the basic types proposed by Durkheim, viz. the 'mechanical solidarity' by which the reinforcement of common values and goals binds individuals to a social unit (Durkheim 1964, 256-257).

A reinforcement of consensus on values and goals could be something which commits individuals to engage in social interaction leading to a verifiable implementation structure. But this need not be the only or major integrating mechanism. Durkheim himself proposed, in addition, 'organic solidarity' as an integrative mechanism in societies with an increasing population of social

units but without a shared set of common values and goals (Durkheim 1964,
275-280). 'Organic solidarity' integrates through the need to accommodate
interdependencies among the greater number of social units.

Implementation structures, therefore, may form through organic solidarity. If this could be verified empirically, the working of implementation structures would be more amenable to a political science 'model' of institutionalisation than if mechanical solidarity were found to prevail (Scholl 1981). Under organic solidarity participants may be conceived of as interdependent claimants, each arguing for and contributing to courses of action as long as their demands are sufficiently met, regardless of how well intermittent outcomes satisfy any superordinate goal or plan.

Either or both of Durkheim's notions of mechanical and organic solidarity, therefore, may provide a useful individual-level motivational underpinning for why and how implementation structures form. To repeat, however, our interest is in social interaction analysis as a method for reconstructing interorganisational units of analysis rather than in the motives for participation as such. Indeed, the social-interaction, individual-level method for reconstructing implementation structures can afford to be perfectly agnostic as to why individuals engage in joint activity. The principal political science interest in interorganisational relations concerns the extent and processes of their institutionalisation and, of course, the consequent effects for policy activity. As we shall see, where Weber has much to contribute to method, Durkheim offers useful ideas about the role of intermediaries in institutionalising interorganisational relations.

Although Weber and Durkheim differed in their methods of research, they shared a common interest in institutions and institutionalisation. Weber's interest was more oriented towards social interaction than organisation analysts brought up on translations of his notion of bureaucracy usually recognise. And whereas Durkheim probably could have lived quite well with the 'reification of the organisation' common in modern organisation theory, this would have created methodological problems for Weber. For Durkheim the existence of social units is not so much an hypothesis to be verified as a fait social; the task of the researcher is rather to demonstrate the effects of their existence (Benoit-Smullyan 1948; for the complexity of Durkheim's position, however, see König 1973, 8-9). For Weber, social wholes like organisations exist only in so far as they are taken into account by individuals in action (Weber 1951, 439).

Weber's approach to understanding social interaction is reconstructive. His concepts provide tools for establishing a significant and reliable 'narrative structure' for analysing the intentional actions of individuals (Rehberg 1979, 204-205), where 'intentional' should be understood to mean 'accountable' rather than the (social) psychological motives for action. Weber requires, first,

that the actors under study be treated as if individually responsible for recognising and choosing between alternative courses of action and, second, that the researcher refrains from applying standards of evaluation legitimized by the prevalent institutional arrangements of the day.

Methodological constitutionalism, therefore, is far removed from Weber's approach to research, and his reconstructive method of enquiry differs sharply from system theoretic interpretations of social interaction based on anthropological and/or psychological 'constitution-of-man' parameters (cf. Winckelmann 1964, 196, 220; Tenbruck 1975, 692-693). Weber's treatment of types of authority (including bureaucracy) is derived in terms of the arrangements of social interaction initiated by the individuals involved (see Rehberg 1979, 213-215).

It can, therefore, be quite misleading to assert as a Weberian notion that bureaucratisation 'is a process of the depersonalisation of administrative relationships' (Selznick 1949, 50). Rather, the ideal-typical bureaucracy relies on social interaction such that the actors are able successfully to anticipate each other's reasons for and instruments of action with a higher probability than under other arrangements. Moreover, for Weber the rationalisation of social interaction inherent in the ideal-typical bureaucracy opened the way to breaking up older forms of social organisation in which the individual could steal away from personal responsibility for his own actions. Thus Selznick's remark about 'depersonalisation' would seem to miss Weber's idea that bureaucracy creates a potential for individuals to define courses of rational action without having to orient themselves to the prescribed ends of one prevalent organisation (Weber 1951, 150).

If the rationalisation of relations inherent in bureaucratic organisations potentially helps individuals to engage in social interaction outside a consensus on prescribed ends, a Weberian method of enquiry requires that such actions first be analysed in their own right and not as activities undermining the prescribed aims of the organisation. On the principle that new institutions emerge from the manner in which interaction patterns develop, and not vice versa, social interaction outside that consensus may indicate the institutionalisation of new interorganisational linkages to meet the demands of old or new interdependencies. Here we return to Weber and Durkheim.

The Weberian notion of social interaction is well integrated with the manner in which he conceives various types of institutional arrangements (Verband, Bureaucracy, Staat) to have developed over time. But he has little guidance to offer about the interrelationships between these types of institutional arrangements in modern society. At this juncture Durkheim contributes the useful notion of organic solidarity whereby the demands of interdependencies between an increasing number of social units in modern society necessitate a degree of functional integration between them. Durkheim does not consider that this

integration can occur spontaneously, but rather that it may have to be institutionalised by intermediaries. Thus the implementation structure approach relies heavily for method on the Weberian notion of an individualistic, reconstructive approach for understanding how social interaction patterns may lead to new institutions, and on Durkheimian ideas for its hypotheses about the role of intermediaries in establishing and maintaining linkages between organisations.

True to his research interest in the representation collective, Durkheim conceived of intermediaries as occupational social units, as organisations, mediating between the state and the agents of economic life (see the preface in Durkheim 1964; we follow closely the review in Barnes 1920). According to Durkheim, the state needs to regulate economic life. However, it can secure competent legislation only on the basis of general principles; its massive and slowmoving machinery is ill adapted to dealing with the highly specialised industrial activities of modern society. As a result of the state's lack of adaptability there is constant oscillation between an excess of inexpert regulation and a condition of partial anarchy.

The remedy as Durkheim saw it was to have organisations interpolated between the population and the state. He understood these organisations as being separate from the state, as having sufficient plasticity and adaptability to complex and diverse social conditions and as being capable of enlisting the loyalty and respect of the populations they serve. The interpolated organisations should, on the one hand, have sufficient generality of purpose to allow the state to regulate their actions intelligently and, on the other hand, should possess the detailed knowledge and the flexibility to enable them to comprehend and administer expertly the diverse needs of specialised industrial and related interests.

Durkheim's conception should be understood as something different from interest group theories of the state and from modern corporatism theory. Durkheim did not ridicule the notion of the state as government-by-interestgroup theories can do (Bentley 1949, 222), and he seems to refer to local institutional arrangements rather than the national corporatism of the recent theories of the same name.

Durkheim's description of the problems of state regulation seems entirely modern. In order to see whether his solution to this problem works, it needs to be recast as a hypothesis of implementation policy analysis: the more institutionalised are the intermediary functions among actors in local interorganisational public policy fields, the better served are the affected population. In order to test this Durkheimian and other similar hypotheses, implementation research needs to go interorganisational with Weber.

In Praise of Implementation Research

The question of the linkage between 'politics' and 'administration' is a traditional, major field of political science research. In the past the emphasis was on how these links can be structured by statute: how can clause be added to legal clause in order to forge, maintain and better the links between 'politics' and 'administration'? Since the advent of the 'organisational society', however, the structuring of these links by statute can no longer be assumed to be the sole, and probably not even the major, source of linkage between 'politics' and 'administration'. The 'new' issue is how organisation is added to organisation, and it is an issue of considerable significance for the legitimation and reinforcement of democratic constitutions.

In particular, the more policy-oriented kind of research which has gathered pace in the recent past provides political science with good scope for continuing its age-old tradition of formal, constitutional analysis, but fructified by behavioural research in the 'political' and 'administrative' parts of the polity. Great importance then attaches to perspectives for (re-)integrating the 'political' and 'administrative' parts of constitutional life by analysis of the organisational linkages connecting the two. It is to this question that implementation research is specifically addressed. So, too, of course is behavioural public administration. It is because of their common research focus that the special remit of implementation research has become blurred, by obscuring the very different methodological foundations upon which the two sub-disciplines build.

The purpose of policy analysis is to identify, describe and evaluate the reliability of currently practised institutional arrangements with respect to the realisation of specific policy purposes. Political science policy analysis frequently resorts to the principles of 'structuring by statute' in the way in which it designs its research. We have argued that implementation research must eschew this 'methodological constitutionalism' if it is to tackle, in the necessary 'empirical constitutionalist' vein, the question of the linkage between 'politics' and 'administrations'.

It is a part of methodological constitutionalism that the policy problems to which research is addressed are problems as defined and addressed by the formal political system; in implementation research, by contrast, the policy problems of research need to be problems as defined and addressed by relevant (which it is for the researcher to define) social actors in a much wider definition than just the actors of the formal political system (who, however, may be included among the relevant actors). This is to say that the unit-of-analysis problem is a central issue in, and how it is tackled a defining characteristic of, implementation research.

If implementation research is to fulfil its remit of empirical constitutionalism,
it needs to analyse the polity using organisational and methodological constructs

which are agnostic to those of the formal constitution. Implementation research analyses policy activity in terms of how actors choose to organise and coordinate their activities; in this way it is able to analyse, in their own right, how links between 'politics' and 'administration' become institutionalised.

For method, we have looked to Weber. Social interaction analysis, in his sense, means treating actors as individually accountable for the alternative courses of action which they take. In this view, action which, in some strict sense, is unconstitutional need be lacking neither in accountability nor in subjection to democratic controls.

The living constitution corresponds variably in time and space to its written counterpart. As the gap between the two widens, the codified links between 'politics' and 'administration' will become increasingly tenuous and the actual linkages between 'politics' and 'administration' may become inefficient in their workings. The Durkheimian diagnosis seems pertinent: because of the state's lack of adaptability a constant oscillation between excess regulation and a condition of partial anarchy prevails. The more this diagnosis seems to apply, the more implementation research is needed as a complement to other types of political science policy analysis.

---

NOTES

1. Empirical research using such an implementation approach is reported in Hull & Hjern (1982 and 1983) and in Hjern & Hull (1983).

2. Dismiss: Dunsire 1978, 69f; Dunsire equates concepts like implementation and coordination with one another, which seems to confuse issues in practical research; cf. Jordan 1982. Adapt: Under this label we refer to attempts to use the concept of implementation synonymously with various forms of 'metapolicy making'; Allison 1980 and Bardach 1978. Annex: This is the most frequent 'technique' to avoid having to deal with the special remit of implementation research; cf. articles in Mayntz 1980.

3. ASQ: Administrative Science Quarterly; HBR: Harvard Business Review.

4. if 1 call myself a sociologist (and that is what my employment contract says), there is no other reason than to bar the still haunting practise of using social wholes for analytic purposes. In other words: Sociological analysis needs to be founded on the action of the one, the few or the many actors, e.g. strictly 'individualized' methodologically'. Weber in a letter 9.3. 1920. Cited from Max Weber und die Sociologie heute, 1965, 137. (B.Hj. translation).

REFERENCES

Allison, G.T. 1980. implementation Analysis', in L. Lewin and E. Vedung (eds.), Politics as
Rational Action, Dordrecht and Boston: D. Reidel Publishing Company.

Bardach, E. 1978. The Implementation Game. Cambridge and London: The MIT Press.

Barnes, H.F. 1920. 'Durkheim's Contribution to the Reconstruction of Political Theory', Political
Science Quarterly 35, 236-254.

Bay, C. 1965, 'Politics and Pseudopolitics: A Critical Evaluation of Some Behavioral Literature',
A PSR 59, 39-51.

Benoit-Smullyan, E. 1948. 'The Sociologism of Emile Durkheim and His School', in H.E. Barnes
(ed.), An Introduction to the History of Sociology, Chicago, 111.: University of Chicago
Press.

Bentley, A.F. 1949. The Process of Government. Evanston, 111.: The Principia Press of Illinois.

Gawthrop, L.C. 1978. 'The Political System and Public Service Education', American Behavioral
Scientist 21, 917-936.

Hill, L.B. 1974. 'Institutionalization, the Ombudsman and Bureaucracy, APSR 68, 1075-1085.

Hjern, B. 1982. 'Implementation Research — The Link Gone Missing', Journal of Public
Policy 2, 3, 301-308.

Hjern, B. & Hull, C. 1982. 'Implementation Research as Empirical Constitutionalism', European
Journal of Political Research 10, 105-115.

Hjern, B. & Hull, C. 1983. 'Small Firm Employment Creation: An Assistance Structure
Explanation', discussion paper from the International Institute of Management, Science
Center Berlin.

Hull, C. & Hjern, B. 1982. 'Helping Small Firms Grow: An Implementation Analysis of Small
Firm Assistance Structures', European Journal of Political Research 10, 187-198.

Hull, C. & Hjern, B. 1983. 'Policy Analysis in Mixed Economy: An Implementation Approach',
Policy and Politics 11, 3, 295-312.

Huntington, S. 1969. Political Order in Changing Societies. New Haven: Yale University Press.
Johnson, N. 1975. 'The Place of Institutions in the Study of Polities', Political Studies, 271-292.

Jordan, G. 1982. 'The Moray Firth Working Party: "Performance", Without "Conformance"',
European Journal of Political Research 10, 117-131.

König, R. 1973. 'Emile Durkheim: Suicide', Kolner Zeitschrift fur Sociologie und Sozialpsychologie
25, 1-23.

Lane, J.-E. 1983. 'The Concept of Implementation', Statsvetenskaplig Tidskrift 86, 17-40.

Majone, G. & Wildavsky, A. 1978. 'Implementation as Evolution', in Policy Studies Annual
Review 2, London and Beverly Hills: SAGE Publications.

March, J.G. & Simon, H.A. 1958. Organizations. New York: Wiley.

Mayntz, R. 1980., (eds.), Implementation politischer Programme, Konigstein/Ts.: Verlag
Athenäum.

Mazmanian, D.A. & Sabatier, P.A., (eds.) 1981. Effective Implementation. Lexington: D.C.
Heath and Company.

McCormick, J.M. & Kihl, T.W. 1979. 'Intergovernmental Organizations and Foreign Policy
Behavior: Some Empirical Findings, APSR, 73, 494-505.

Mclntosh, D. 1983. 'Max Weber as a Critical Theorist', Theory and Society 12, 69-109.

Ohe, W. 1977. 'Macht — Abhängigkeit — lnterdependenz', discussion paper series, dp/77-104,
International Institute of Management, Science Center Berlin.

Oppenheim, F.E. 1964. 'Rationalism and Liberalism', World Politics 16, 342-361.

Perry, J.L. & Levine, C.H. 1976. 'An Interorganizational Analysis of Power, Conflict and
Settlements in Public-Sector Collective Bargaining', APSR 70, 1185-1202.

Polsby, N.W. 1968. 'The Institutionalization of the U.S. House of Representatives', APSR
62, 144-169.

Popper, K.R. 1963. Conjectures and Refutations; The Growth of Scientific Knowledge. New
York: Basic Books.

Pressman, J.L. & Wildavsky, A. 1973. Implementation. Berkeley: University of California Press.

Rehberg, K.-S. 1979. 'Rationales Handeln als großbiirgerliches Aktionsmodell', Kölner Zeitschrift
fiir Soziologie und Sozialpsychologie 31, 199-236.

Scholl, R.W. 1981. Än Analysis of Macro Models of Organizations', Administration and Society
13, 271-298.

Scott, W.G. 1979. 'Organicism: The Moral Anaesthetic of Management', Academy of Management
Review 4, 21-28.

Selznick, P. 1949. Än Approach to a Theory of Bureaucracy', APSR, 47-54.

Strauss, L. 1962. 'Epilogue', in J.J. Storing (ed.), Essays on the Scientific Study of Politics.
New York: Holt, Rinehardt & Winston.

Starbuck, W.H. 1976. 'Organizations and their Environments', in M.D. Dunnette (ed.), Handbook
of Industrial and Organizational Psychology: 1069-1123. Chicago: Rand McNally.

Tenbruck, F.H. 1975. 'Das Werk Max Webers', Kölner Zeitschrift fur Sociologie und Sozialpsychologie
27, 663-702.

Thomas, D.W., Potter, H.R., Miller, W.L. & Aveni, A.F. (eds.). 1972. Institution Building: A Model
for Applied Social Change, Cambridge, Mass.: Schenkman Publishing Co.

Thompson, J.D. 1967. Organizations in Action. New York: McGraw-Hill.

Weber, M. 1951. Gesammelte Aufsätze zur Wissenschaftslehre, (2., durchges. u. erg. Aufl.
von Johannes Winckelmann), Tubingen: Mohr.

Winckelmann, J. (ed.) 1964. Max Weber-Gedächtnisschrift der Maximillians-Universität
Miinchen. Miinehen.

Wrong, D. 1979. Power. New York: Harper and Row.