Implementation models were suggested in response to the neglect of the various stages of policy execution after the enactment of policies at the governmental level, the missing link as it was called. When using implementation models the basic problem is to design some insti
tutional mechanisms by which implementation may work, if indeed implementation in the sense of goal achievement is at all workable.
Paul A. Sabatier, a pioneer in implementation analysis, raises some fundamental questions about the nature of implementation in a review of the present state of implementation theory (Sabatier, 1986).
Although Sabatier's analysis of the two competing models of imple
mentation - top-down versus bottom-up implementation - as well as his attempt to launch a third model - a kind of coalition model - is perceptive and challenging, it may be argued that it misses a more basic problem concerning the nature of public policy implementation.
It must be recognized that modelling the process of executing public policies - the implementation process - is different from evaluating the extent to which objectives have been accomplished - the imple
mentation assessment. Surveying the plethora of implementation models may clarify some of the problems surrounding implementation analysis. It is based on the distinction between implementation as an outcome and implementation as a process. As long as the concept of implementation remains unexplained, any theory about the conditions for successful implementation will remain ambiguous.
THE CONCEPT OF IMPLEMENTATION
The concept of implementation is characterized by a problematic structure. Webster's Dictionary states that 'implementation' means either the act of implementing or the state of having been implemented; it presents the following key words for 'implement':
to carry out: accomplish, fulfil; to give practical effect to and ensure of actual fulfihnent by concrete measures, to provide instruments or means of practical expression.
'To carry out something' or 'to accomplish something' may sound intelligible and require little explication. A formal definition might be:
(DF1) Implementation = F (Intention, Output, Outcome)
98 THE PUBUC SECTOR
where implementation refers to the bringing about, by means of outputs, of outcomes that are congruent with the original intention(s).
It is readily seen that 'implementation' has a double meaning: 'to give practical effect to' or execution on the one hand, and 'fulfil' or accomplishment on the other. A policy that is executed need not result in the accomplishment of its objectives.
Thus, we have a basic ambiguity in the notion of implementation:
implementation as an end state or policy achievement, and implemen
tation as a process or policy execution. The Oxford English Dictionary
notes the same double meaning: 'to complete, perform; to fulfil'.
The performance of activities need not lead to the fulfilment of objectives. Implementation analysis could be considered a development of public administration whereby the execution of policies is expedited by the addition of evaluation research. Implementation analysis is not confined to what happens after political reforms have been enacted, because it goes beyond the focus on programme execution in public administration as it was traditionally conceived.
The concept of implementation implies assessment; it is made by the actors involved in the implementation process, and one basic task of the implementation analyst is to evaluate the implementation. Given the ends and means of the policy, implementation analysis cannot be confined to a statement of what happens afterwards. The analyst may use the tools of evaluation research in order to arrive at an imple
mentation judgement of the extent of successful implementation, the first major focus in implementation analysis. Success or failure are not the only relevant properties of the implementation of public policies.
The process of enforcing a policy has its own logic, which is the second major focus of the implementation analyst.
Aspects of the implementation process other than the accomplish
ment of the policy objectives that the analyst is interested in include:
the strategies and tactics employed by various parties to the imple
mentation game, the mechanism of delay as a decision parameter, the variety of motives among the participating actors, and the need for coalition building and fixing the game. Implementation requires more than a state of affairs in which there is a policy objective and an outcome (or several objectives and outcomes), since, in addition, the concept of implementation implies that these two entities - objective and outcome - satisfy two different relationships: the causal function and the accomplishment function.
Two ideas are fundamental to the concept of implementation: that the policy programme is the output that brings about the outcomes in such a way that the latter accomplish the objectives of the policy.
Implementation assessment focuses on the operation of a public policy and its consequences. It includes three logically separate activities: (a) clarification of the objectives involved (the goal function), (b) statement of the relationship between outputs and outcomes in terms of causal
IMPLEMENTATION MODELS 99 effectiveness (the causal function), and (c) clarification of the relation between objectives and outcomes in order to affirm the extent of goal achievement (the accomplishment function).
Each of the three tasks presents its own peculiar problems. Together they imply that it may be difficult to judge the effectiveness of implementation. The ends and means - the intentions - of policies are formulated and enacted by various kinds of actors in the political process. What is an end or a means is an intentional object to some actor, which means that any definition of implementation must specify the actors involved in the process. These actors may be divided into two sets, the forma tors and the implementors.
Implicit in implementation theory is the idea that the actors who decide on policy are different from the actors who are responsible for the implementation of policy. Though this is far from always the case, the implementation process is built up around an asymmetric relationship between the forma tors of policy and the implementors of policy. The formators may not be the initiators of policy; be that as it may, the theory of implementation assumes that public policy becomes a legitimate concern for implementors once it has been decided upon in formally defined ways. By developing the original implementation formula we are now at a stage where a more powerful and complex concept of implementation may be introduced:
(DF2) Implementation = F (Policy, Outcome, Formator, Implementor, Initiator, Time)
Suppose one asks whether a policy has been implemented. Then one needs information about the extent of congruence between policy objectives and outcomes, but that is not enough. In addition, there has to be a decision concerning the time span that may pass before an implementation judgement can be said to be neither premature nor belated. When is it appropriate to ask whether a programme's objectives have been realized? This is all relevant to the concept of implementation. When we move to the theory of implementation processes, conflicting views crop up. It is easier to introduce a formal concept of implementation than to model the evolving implementation process.
Whereas implementation as an outcome is rather unambiguous - to carry a policy into effect - the implementation process is a more complex phenomenon. Implementation processes involve coalitions, learning, political symbolism, implementation perspectives, as well as control. But this is not enough for the general claims that imple
mentation is advocacy coalitions, is evolutionary learning or hierarchical control. Any kind of mechanism may be used in the implementation process, because of the loose connection between implementation as an outcome and implementation as a process.
100 TIlE PUBUC SECTOR
MODELLING PROCESSES OF IMPLEMENT AnON
The status of so-called implementation models is precarious with regard to the standard public policy models (Dunsire, 1978). They are considered a necessary complement to policy models because, it is argued, these model only the decision-making process, assuming that the enactment of policy implies the execution of policy programmes as well as the implementation of policy objectives (Meter and Horn, 1975).
Or, even worse, the policy models discussed above adhere to a naive assumption about public administration: namely that policies once decided upon will automatically achieve their objectives by means of the policy outputs as if implementation was something utterly simple and automatic.
The argument about implementation deficit implied a radical rejection of this hypothesis. Neglecting the implementation stage could not be considered worse than adhering to a naive theory about public administration and the behaviour of implementors.
Thus, implementation models constitute the missing link between policy decision-making on the one hand and policy execution and policy implementation on the other (Hargrove, 1975). The argument about the missing link may have appeared as a revelation to those who had realized that policy may have one appearance when enacted and a quite different one when put into practice. The basic problem was, however, to explain why this misfit tended to occur more than often - how are we to understand what happens after the formulation and formation of policy? To the extent that this hypothesis about a lack of congruence between policy objectives and policy outcomes is true, it amounted to a rather drastic criticism, if not straightforward rejection, of the standard policy models.
The basic policy models were accused of either neglecting the problematic phase in the policy process or adhering to a naive model of implementation: state the goals, derive the means, execute the programmes and find the outcomes. Beginning with the highly original Pressman and Wildavsky analysis in Implementation (1984), first published in 1973, a large literature poured out warnings against any public policy model that regarded implementation as simple or straightforward.
Every policy model was regarded as incomplete at best or deficient at worst because they lacked a theory about the mechanism of imple
mentation - how programmes should be handled in order that stated objectives could be achieved in terms of positive outcomes. To identify this mechanism of implementation - the key to successful implemen
tation - became the target of a number of new policy models focusing on what happens after the decision stage.
At the same time these new models were open to the same kind of criticism: they by-passed the initial phases of the policy process - that
IMPLEMENTATION MODELS 101 is, they were one-sided. The attempts to extend the implementation models to cover the entire policy cycle have hardly been successful. We are still stuck with a gulf between decision-making models and implementation models in the study of public policy (Sabatier, 1986).
The implementation models differ as to the nature of the mechanism to be employed in the monitoring of programmes: top-down models versus bottom-up models. A brief survey of the various implementa
tion models follows.
IMPLEMENTATION AS PERFECT ADMINISTRATION
Hood suggests a model of implementation that would 'produce perfect policy implementation' (1976: 6). Such a model would include a unitary administrative system with a single line of authority, enforce
ment of uniform rules or objectives, a set of clear and authoritative objectives implementable on the basis of perfect obedience or perfect administrative control, perfect coordination and perfect information within and between administrative units, absence of time pressure, unlimited material resources for tackling the problem and unam
biguous overall objectives and perfect political acceptability of the policies pursued (Hood, 1976: 6-8).
The model of perfect administration is an ideal-type construct to discover the sources of implementation failure (Hood, 1976: 190-207).
The model approaches implementation from the narrow focus of the characteristics of pure authority relations - hierarchy, obedience, control and perfect coordination - viewed as the mechanism for the accomplishment of successful implementation.
However, empirical work on implementation has resulted in a different finding: namely that mechanisms more symmetrical in nature, such as exchange and negotiation, are more germane to the implementation process than authority and its characteristics (Lipsky, 1980; Barrett and Fudge, 1981). These bargaining mechanisms for the implementation of policy are as important as structures of authority (Bardach, 1977; Pressman and Wildavsky, 1984: 87-124). It is questionable whether the model conditions listed really are conducive to perfect implementation; for example, the model conditions of this top-down version of implementation seldom apply due to intra- or inter-organizational complexity (Hanf and Scharpf, 1978).
IMPLEMENTATION AS POLICY MANAGEMENT
A model of implementation may involve the search for guide-lines for successful implementation. In a 1979 article, 'The Conditions of Effec
tive Implementation: A Guide to Accomplishing Policy Objectives', Sabatier and Mazmanian state:
102 THE PUBUC SECTOR
The program is based on a sound theory relating changes in target group behaviour to the achievement to the desired end-state (objectives). The statute (or other basic policy decision) contains unambiguous policy directives and structures the implementation process so as to maximize the likelihood that target groups will perform as desired. The leaders of the implementing agencies possess substantial managerial and political skill and are committed to statutory goals.
These conditions may at first sight appear innocuous, but when one ponders about the availability of 'sound' causal hypotheses or clear ends and means as well as powerful management strategies, then the risk of circularity in the argument is evident: what works in the implementation process is what accomplishes implementation.
Moreover, there are other equally demanding requirements.
The program is actively supported by organized constituency groups and by a few key legislators (or the chief executive) throughout the implementation process, with the courts being neutral or supportive. The relative priority of statutory objectives is not significantly undermined over time by the emergence of conflicting public policies or by changes in relevant socio
economic conditions that undermine the statute's 'technical' theory or political support. (Sabatier and Mazmanian, 1979: 484-5)
These presumed sufficient conditions for successful implementation do identify crucial factors that affect policy accomplishment: technology, unambiguity of objectives, skill, support and consensus. However, the counter-argument is that this theory begs the question of what a 'sound' policy technology is. Moreover, what is 'substantial' policy skill and 'enough' policy support? And when is a policy 'significantly' undermined by conflict?
IMPLEMENTATION AS EVOLUTION
Wildavsky has introduced the theory of a process of implementation as necessarily resulting not in implementation but in redefinition of objectives and reinterpretation of outcomes - that is, evolution. If implementation processes result in the redefinition of objectives and the reinterpretation of outcomes, then how could there be successful implementation? The evolutionary conception of implementation implies that implementation processes may not be neatly separated from stages of policy formulation, mingling objectives and outcomes. It also implies that implementation is endless: 'Implementation will always be evolutionary; it will inevitably reformulate as well as carry out policy' (Majone and Wildavsky, 1984: 1 16). This is an empirical argument that is open to refutation pending a major survey of pro
gramme accomplishments (see, for example, the analysis of the effects of social policy on British society in the post-war period in George and Wilding, 1984).
IMPLEMENTATION MODELS 103 IMPLEMENTATION AS LEARNING
Wildavsky has outlined yet another interesting interpretation of the nature of the implementation process (Browne and Wildavsky, 1984).
Implementation is modelled as an endless learning process where the implementors through continuous search processes come up with improved goal functions and more reliable programme technologies.
There is no natural end to the process of policy implementation because each stage means an improvement in relation to earlier stages where, over time, the original objectives are bound to become transformed and the initial means replaced.
The theory that implementation is learning may be regarded as an optimistic explanation of the hypothesis that implementation is evolution. The kind of implementation process conceived of in the various versions of a top-down approach - naive implementation, perfect administration, a hierarchical model, conditions for successful implementation - is considered suboptimal because of its assumption of a one-shot implementation process.
IMPLEMENTATION AS STRUCTURE
The events constituting a process of implementation are typically approached as pieces forming a whole. How can one separate what is part of an implementation process and what is not (demarcation)?
What are the basic components of a process of implementation (identification)? In 'Implementation Structures: A New Unit for Administrative Analysis', Hjem and Porter state: 'An implementation structure is comprised of subsets of members within organizations which view a programme as their primary (or an instrumentally important) interest' (1981: 216).
Obviously, an implementation structure consists of sets of actors, but which sets of actors constitute one and only one implementation structure? Is it enough that these actors are members of organizations and have a 'primary interest' in a programme? Is it not necessary for people who have a 'primary' interest in a policy also to wish or attempt to put it into effect?
The approach of an implementation structure follows from an emphasis on properties of processes of implementation other than those of the top-down perspective: organization complexity, self
selection of participants, multiplicity of goals and motives, local discretion. Hjem and Porter state:
Implementation structures are not organizations. They are comprised of parts of many organizations; organizations are comprised of parts of many programmes. As analytic constructs, implementation structures are conceptualized to identify the units of purposive action which implement programmes. They are 'phenomenological administrative units', partly defined by their participating members. (1981: 222)
104 THE PUBUC SECTOR
The description of implementation structures as compnsmg units that implement programmes is of little help as it is circular. Either an implementation structure is a construct - simply a 'unit for admin
istrative analysis', or 'implementation structures are administrative entities' (Hjem and Porter, 1981: 219), but not both.
The concept of an implementation structure is relevant for the analysis of implementation processes, but one has to be aware of the fallacy of reification or misplaced concreteness. No wonder that implementation is described as difficult, as it is hard to find out how an implementation structure is to be demarcated and identified (Hjem and Hull, 1982).
IMPLEMENTATION AS OUTCOME
Fudge and Barrett state that a theory of the implementation process follows from a particular concept of implementation. If implementation is not 'putting policy into effect', Fudge and Barrett state, then: 'The emphasis . . . shifts away from a master/subordinate relationship to one where policy-makers and implementers are more equal and the interaction between them becomes the focus for study' (1981: 258).
The concept of implementation and the concept of an implemen
tation process should be kept analytically separate. Why could not organizational complexity or autonomy, just like exchange and negoti
ation, be conducive to, or compatible with, implementation as 'putting a policy into effect'? Similarly it could be the case that the perfect administration model may only achieve a state describable as 'getting something done'. To analyse what must obtain in order to apply the concept of implementation is different from producing a model as to how implementation - in particular, successful implementation - comes about.
IMPLEMENTATION AS PERSPECTIVE
Walter Williams (1982) has argued strongly in favour of taking a special perspective as the starting point for policy execution. Is this so
called implementation perspective some kind of practical science of administration, a body of knowledge that policy-makers and imple
mentors could draw upon as they approach the implementation of policies? For Williams, the implementation perspective is the perspec
tive of the practitioners.
It does not follow from the fact that actors are participating in something which they label an 'implementation process' that imple
mentation is really going on. As several implementation studies have testified, actors may execute policies believing that their actions eventually bring about implementation, but they may be wrong. In order to state the extent to which an implementation perspective meets