Resumé

Policy-making and implementation have usually been treated as two distinct disciplines. In this article it is argued that implementation is affected by the prior policy-making process. Hypotheses regarding such impacts are derived from the policy-making theories resting on a) the rational decision-making model, b) the conflict-bargaining model, and c) the garbage-can model: Implementation failures are more likely 1) if goals are absent or vague and if alternatives and their consequences have not been considered; 2) if the policy-making process involves participants with conflicting interests and compromising; 3) if there are many and changing participants with limited attention and if symbols are important in the policy-making process. These hypotheses are tested and discussed in relation to a reorganization case, the decentralization of the disablement pension administration in Denmark in 1976.

'... Presidents are more likely to be punished for not making promises of administrative reform than
for not implementing them! (March & Olsen 1983, 291)

How Are Policy-Making and Implementation Related?

Policy-making and implementation processes have been studied very intensively by political scientists - policy-making for several decades and implementation mostly within the last decade. Generally, however, policy-making and policyimplementation been treated as two separate phenomena by scholars divided by specialization. Scholars of legislative behaviour and policy-making have rarely been interested in what happened after the laws were passed, and implementation scholars have typically restricted their interests to what happened after the passing of a law. When implementation gaps - which is the typical finding - are identified, implementation scholars tend to explain such gaps by examining the implementation process and structure, looking - inter alia - at the diverging interests of implementing actors or their lack of capacity to implement the laws.

---

* This article is a revised version of a paper which was presented at the IPSA Conference in Paris, July 15-20, 1985, in a panel on 'The Process of Extended Legislation and Its Scope'.

Nevertheless, implementation studies have often been inspired by policymaking for example by the application of policy-making concepts and theories of conflict and bargaining to implementation processes and the conceptualization implementation as continued policy-making in a (slightly) changed (Pressman & Wildavsky 1973, Bardach 1977, Sætren 1983, Elmore 1978). In addition, some polincal scientists avoid a sharp distinction between policy-making and implementation and prefer to treat them as one continuous policy-making process (March & Olsen 1983; Sætren 1983). However, in all of these cases policy-making models and theory are applied to the analysis of implementation rather than to the linkages between policy-making or legislative processes and implementation.

In this paper we shall discuss some hypotheses about the impact of policymaking on implementation, and we shall discuss these hypotheses in relation to a case-study on reorganization policy: the decentralization of the Danish disablement pension scheme from the central to the regional level in 1976.

In the absence of any ready-made theoretical basis for analyzing how policymaking implementation, we shall try to see if it is possible on the basis of well-known theories on policy-making to develop some hypotheses about a relationship between policy-making and implementation. This simultaneous application of different decision-making theories is inspired by Allison (1971), J.P. Olsen (1972) and Elmore (1978). The applied theories are a) the rational decision-making model, b) the conflict-bargaining model, and c) the garbagecan

Our main hypothesis derived from the rational model is that implementation failures are more likely if the goals are unclear or absent in the policy-making phase and if relevant alternatives and their consequences have not been considered. to the conflict-bargaining perspective, policy decisions are often compromises resulting from conflict and bargaining between actors with different goals and resources. Official goals in laws etc. may therefore be conflicting, or legitimizing solutions that are motivated by other latent goals of the various actors. This suggests the hypothesis that conflicts of interest make the successful implementation of official goals less likely. According to the garbage-can policy-making is characterized by limited attention from multiple and changing part-time actors with competing claims for attention to other tasks. There are no 'direct connections between the purposes and intentions the actors and the events we want to explain' (Olsen 1972, 48). Symbolic aspects play an important part for the decision process and outcome. Arguing that there is no direct link between goals and policy decision, this model suggests the hypothesis that successful implementation of policy goals may be reduced in proportion to the extent to which the policy-making process is characterized by 'garbage-can' traits.

These hypotheses and models are elaborated below in connection with their
application to our empirical case. However, first we shall present it: the decentralization
the Danish Disablement Pension Administration.

Since the models are probably more supplementary than mutually exclusive, we expect that no single model can offer a satisfactory explanation of the relationship between policy-making and implementation, but that all models may contribute to our understanding of that relationship. The important question the paper is not how well each policy-making model fits our data on the making of a Danish decentralization reform, but if and how, from each model, we can derive fruitful hypotheses regarding the relationship between policymaking implementation.

A Decentralization Case

Structural Changes

The case-study of this paper focuses on reorganization. It examines a Danish decentralization reform aimed at moving the authority to grant disablement pensions and a few other welfare benefits from the central level to 15 new independent regional Boards of Rehabilitation and Pension. The secretariats of the boards are, however, staffed by the social welfare administration of the county in each region. The main task of the boards is the granting of disablement pensions, but they also grant other kinds of pension for people under the age of 67, motor vehicles for handicapped pensions and rehabilitation aid for the establishment of private undertakings. The earlier administrative structure was decentralized in two ways. First, three central boards were discontinued and their decision-making functions transferred to 15 new regional boards. Second, the municipalities were assigned a much greater role in the preparation of cases than before.

Objectives

It is difficult to find precise statements in the decentralization law or other relevant official documents about its goals. And indeed, evaluation and implementation have often experienced problems in identifying precise goals that can be used as a standard against which to measure success or failure. In much legislation statements of goals are absent, vague or conflicting, and latent goals may be more important than the manifest official ones, thus making evaluation very difficult (Cronbach et al. 1981, 129-33).

Though some scholars have proposed to stop using official legislative goals as an evaluation standard (Hjern 1978, 1982, Elmore 1979-80), we think it is still relevant to study these goals. F^rrom a democratic perspective, it is appropriate to examine whether the goals of government are realized. An additional considerationis the goals have, after all, been articulated in the legislative process in

order to convince the electorate and perhaps also parliamentary coalition
partners.

As mentioned above, the official goals of the Danish decentralization reform have not been easy to identify. One clearly stated goal was, however, to shorten decision-making time. Examining bills, parliamentary debates and official reports from commissions and committees that prepared the reform, it is possible - with a little ingenuity - to identify additional goals as indicated by statements of purpose, expected favourable consequences related to the reform, as well as expectations to avoid unfavourable side effects. Among such other goals uncovered by our analysis were a step-by-step reduction of administrative costs and preserving the 'necessary' uniformity of decision-practice. Other likely objectives were to involve the applicants more in the decision-making process and to increase expertise in the decision-making by encouraging the introduction of more relevant information about the clients and local labour-market conditions.

Implementation

Reorganization studies reveal an abundance of failures. Most often reorganization are defeated (March & Olsen 1983, Seidman 1975), and when reorganization does take place, most studies of such reforms report results not satisfying prior expectations of, for instance, a decrease in administrative costs (March & Olsen 1983, 28, Meier 1980, Morgan & Pelissero 1980).

Using a very modest criterion of success, our case represents a relative success in so far as the decentralization did take place, in contrast to most reorganization proposals reported in the literature. However, when the administrative practice is compared with the goals of the reform, the case seems to fit nicely into the general pattern of failures. After the launch of the reform, decision-making time has increased substantially, the staff has doubled, and uniformity among regions has declined.

Speeding Up the Administrative Procedure

Apparently the most important goal of the decentralization reform was to reduce the time spent on handling cases (Folketingstidende 1974/75, A, 4136). For several years severe criticism had been levelled at the slow administrative procedure of the central Disablement Insurance Tribunal (DIT). Because of the minimal involvement of the municipalities in the administrative procedure before decentralization took place, the quality of the social and medical information in the applications was not satisfactory. Lacking personal contact with the applicants, DIT had to solicit supplementary information by procuring large numbers of documents from various specialists, with periods of delay as a consequence.

It was, then, expected that the reform would speed up the administrative

procedure. This expectation rested in particular on the belief that decisionmakingboards receive better and more complete information on each case from the municipalities, which were to be responsible for providing a concerted evaluation of the social arid medical situation of the clients by social workers, municipal medical advisers, and local politicians. The time spent on each case in the municipalities was also expected to become reasonably short because most applicants were well known to municipal agencies from earlier contact.

In Table 1 the average decision-making time before and after the reform is presented. Before the reform, the average total time from application to decision was 5 to 6 months, while the average time in 1979 and 1982 was one year. The decision-making period was expected to fall - but actually it has been doubled.

Cutting Down the Administrative Costs

Both the decentralization bill and the subsequent debate in Parliament reveal the expectation that the reform would lead to a gradual reduction of administrative costs. It seems fair to consider that expectation as another policy goal of the decentralization reform. The argument was that productivity would be enhanced by the concentration of several related functions in the social welfare administrationof counties, which would now be able to coordinate the work and to choose among alternative solutions for each applicant without having to start a new case in another agency (Folketingstidende 1974-75, A, 4139-40). The more careful and detailed preparatory case-work of the municipalities was also expected to ease the way for the regional decision-makers (Betænkning 1975, 17-20). Because of anticipated start-up problems - particularly due to inexperience the reduction of staff was, however, expected to take place only

gradually. In Parliament only one opposition party - the Progress Party, which
is a conservative, tax-protest party - voiced scepticism. It appears that the
Progress Party was right!

In the spring of 1977 - one year after initiation of the reform - the number of social workers and heads in the new regional offices exceeded the previous number of personnel by 50 to 73 per cent. At the same time, the number of cases processed per month was smaller than before the reform, representing a fall in productivityl of approximately 60-65 per cent (Winter 1980a, 47-54). Admittedly, that time the number of cases has grown but so has the number of professional staff (although the latter rate is not known).

In sum, the staff productivity was expected to rise in the long run (which may be more than one year after the reform), but actually it has been reduced by more than 50 per cent! Besides, there is no doubt - irrespective of an absence of hard data - that the municipal staffs employed to prepare the cases for the regional boards have grown significantly in size.

Preserving Uniformity

One probable reason why the authority of the boards was not further decentralized moving it to the municipal level was a wish to preserve uniformity of decisions. Such uniformity would, of course, be much more difficult to preserve if decision-making authority were spread among the 278 different municipalities. was expected, however, that the 15 regional boards could ensure the 'necessary' uniformity of decision-practice with some help from the central authorities (Betænkning 1969, 107).

As an indicator of uniformity we compare the distribution of different decision outputs in each region. The decisions may be: a) 'no pension', b) 'pension at lowest level', c) 'pension at medium level' or d) 'pension at highest level'. If the relative share of each decision type is the same in all regions, the decision practice is very uniform. The degree of uniformity has been measured by the weighted standard deviation (SD) of the relative share of each kind of decision.

Already before the reform there was a slight variation in the decision distributionamong regions, partly because of differences in demographic, occupationaland market conditions. However, since the reform was introduced, uniformity has been greatly reduced. For instance, the SD on the percentage of applicants getting 'no pension' has grown from 2.2 in 1975/76 to 7.3 in 1978. The increase in spread can hardly be explained by a change in demographic variables, as these are constant for the period in question. The possibility cannot be excluded that the local and regional authorities have paid greater attention to differing labour market conditions, but unemployment indicators suggest that this is not the major reason. The lack of uniformity in client treatment must

instead be attributed - at least to a significant degree - to different decision
practices among the boards and their staffs (Winter 1980a, 1980b).

Other Goals

Apparently a larger involvement of the clients in the decision-making process was another goal. Furthermore, the official documents indicate - however vaguely - that improvement of the information collected on each client was expected to be an important benefit of the reform. Though no data about these phenomena have been collected, it seems very likely that particularly client - involvement has increased and that also the information on clients has improved. However, to the extent these assumptions are correct, the improvements must primarily be attributed to the more complete preparation of cases in the municipalities have direct client contact, and only to a very little extent to the transfer of decision-making authority from the central to the regional level.

Conclusion

The decentralization reform must be evaluated as another case of policy failure. In implementing the reform, the expected outcome of reductions in decisionmaking and costs did not materialize. On the contrary: The results were increasing delays and increasing costs. In addition, uniformity was substantially reduced. In the following we shall see if the policy-making process² may help to explain why the policy goals were not implemented.

Application of the Rational Decision-Making Perspective

In its pure form the rational decision-making model presumes a choice situation characterized by clear preferences, the listing of all possible alternatives, calculationof consequences of each alternative, and a choice based on comparing the consequences of these alternatives in light of the preferences (Simon 1976, Meyer 1973). The model has been one of the most criticized theoretical notions in the social sciences, the main criticism being that it is an unrealistic description of decision-making. Few adherents to the model would deny this, but employ it as an ideal type for descriptive purposes in analyzing actual decision-making and/or as a normative theory of how or in what direction actual decisionmakingought change (Meyer 1973). For our purpose we may also regard models with limited or bounded rationality - implying the listing of fewer alternatives and calculating fewer consequences than the pure rational model demands (Simon 1976, March & Simon 1958) - as a part of the rational perspective.The is not necessarily static; preferences as well as environmental

factors may change during the decision-making process to make the invention of
new alternatives and/or new calculations of consequences necessary.

Though few would claim that this model is a realistic description of actual decision-making, it might, nevertheless, be fruitful in generating hypotheses on why decisions are implemented or fail to be implemented. Inspired by the rational model, we find it reasonable to believe that implementation failures are more likely if preferences are unclear or absent in the policy-making phase, if not at least a few relevant alternatives have been considered, and if no qualified attempts have been made to estimate the consequences of the alternatives or, at a minimum, of the preferred alternative. Also, we expect that implementation failures are more likely if changes of relevant conditions do not lead to the reconsideration of alternatives and their consequences.

At first glance the DIT decentralization bill could convey an impression of rationality. Its goals were relatively clear, and it was based on preparations stretching over more than a decade, from 1962 to 1975, and involving consideration several commissions and committees.³ However, a closer analysis raises doubts about the rational character of the policy-making process.

First, though some goals can be identified in the final decentralization bill, much of the preparation of the reform seems to have taken place without any reference to specific goals. The decentralization was decided by Parliament in two decisions: 1) In 1970 the organizational structure of the whole social welfare system was changed, including the establishment of the regional rehabilitation and pension boards. 2) In 1975 the functions for the new boards were transferred from the earlier central agencies. In the first legislation no goals can be identified in the bill, and even in the reports that prepared the bill the important goals of reducing decision-making time and costs cannot be found. And there is only very vague reference to uniformity and client-involvement considerations. The goals of reducing delay and cost were first introduced in the final bill in 1975. The solution or means, decentralization, seems to be more of a guiding-star than the goals. We shall discuss that later in connection with the garbage-can perspective.

Second, the consideration of alternatives seems to have been very limited. In 1966 the Welfare Reform Commission discussed a large number of alternative models for reorganizing the total welfare system. But as far as disablement pensions and other pensions were concerned, all models proposed decentralization a regional authority. The only alternatives considered in the Commission concerned the question of where among several regional authorities the functions should be placed: the prefect, the County Council, or new independent Also, decentralization to the municipalities was considered for some of the functions. The Commission did not, however, consider the alternative a central authority.

In this connection we must, however, point to a general data problem. Even if
the documents containing proposals to, and reports from, the meetings of the

Commission and other committees do not indicate clear statements regarding
goals, the listing of alternatives, and calculation of consequences, one cannot
preclude the possibility that these were considered (e.g. in the secretariat).

However, a central alternative was considered in a report by another committee internal reorganization of DIT, which in 1968 discussed how to reduce delay in DIT. The report states that keeping the authority to grant pensions with the central DIT but reorganizing it to increase contact with local authorities and practitioners would both be too costly and fail to solve the problems the reform was intended to solve. Thus the central alternative may have been considered, but it received very little attention in the course of the policy-making process. Later a successful experiment was undertaken which involved the municipalities in the preparation of cases with the assistance from DIT advisers while DIT still made the decisions. However, this experiment did not lead to any reconsideration of the matter.

It also seems that estimates regarding the consequences of the alternatives were very vague and imprecise. Some efforts were made to calculate the regional need for staffing, and it was estimated that the required number of personnel would not exceed the pre-reform number (Betænkning 1975). The Association of Counties was apparently unable to estimate personnel needs with any kind of precision (Socialcentrene 1972). In addition, no attention was paid to the consequences terms of municipal costs in spite of the fact that the role of the municipalities the preparation of cases was substantially enlarged by the reform.

From a rational perspective it is especially strange that no attempt was made to estimate the consequences of the reform with regard to the primary goal of reducing decision-making time. It was uncritically assumed that decentralization automatically would bring about the desired result. In the decentralization bill and in the parliamentary debate in 1975, the governmental representatives did not say anything about what the average time devoted to each case actually was and to what level it was expected to fall, although statistics on the decisionmaking of DIT had been compiled for several years. Only one member of Parliament (a medical specialist who personally had worked for DIT) made any reference to the actual decision-making time, which she estimated to be almost one year (Folketingstidende 1974/75, F, 4926). Nobody challenged that estimate in spite of the fact that the actual average time devoted per case was less than 6 months, of which DlT's portion was 3 to 4 months.

Also, the actual time devoted to each case in DIT had fallen substantially during the late 1960s in connection with internal reorganization of that agency. A huge number of extra cases - about 28,000, more than a whole year's normal case load - had piled up as a result of a disablement pension reform in 1965 which introduced higher pensions. This big back-log of cases had, however, been removed by the end of 1967, facilitating a speed-up in the decision-making. The average decision-making time at the central level was reduced from 7.9 months

in April/May 1967 to 3.5 months in June 1968. Out of these 3.5 months only 2.3 months were internal to DIT, while the rest was attributable to external medical specialists and (perhaps) the municipalities. Later in 1968 and 1969 the decisionmakingtime reduced even more, to less than 3 months per case! This analysis demonstrated that it was in fact possible to reduce the central decision-making time considerably, and from the perspective of rational decision-making it might, therefore, have made sense to consider more carefully whether the decision-making time really could be further cut by decentralization. However, it was simply assumed that such a cut would automatically follow (Betænkning 1968).

In comparison, during the period of 1978 to 1982, after the reform was in place, the average time was about 5 months. Furthermore, while in 1967 the share attributable to the municipalities was 1.3 months, it was 6 to 8 months in 1978-82. retrospect it seems clear that it was very unrealistic to expect the total decision-making time to drop below about 4.5 months under a decentralized structure that also called for an increase in municipal involvement in the preparation cases. After the reform, during the period of 1979 to 1982, the average time devoted to each case was one year!

Another factor that was not taken into account was the fact that some of the functions originally intended for the regional boards were placed with the municipalities. of the original arguments in favour of decentralization pointed to the advantage of coordinating the granting of disability pensions with the granting of rehabilitation aid, sickness allowances and expensive auxiliary facilities. most of these other functions were in the end transferred to the municipalities. In spite of the fact that some of the original reasons for introducing reform had disappeared, it was not reconsidered.

In sum, by applying the rational decision-making perspective to an analysis of the linkage between policy-making and implementation, a number of deviations from the rational model may help explain parts of the implementation failure. Although the policy-making process took more than a decade, and although the official goals of the legislative output were relatively clear - at least in the end - it appears that the policy-makers largely failed to consider alternatives and their consequences. And even though it was demonstrated that it indeed was possible to reduce the decision-making time of a central agency considerably and successfully the municipalities in the processing of cases, this did not inspire the policy-makers to consider the consequences and advantages of different alternatives. same holds true for changing circumstances such as decreasing delays and a reduction in the functions assigned to the new boards. The following conflict/bargaining and garbage-can models may both help explain why the policy-making process did not match the rational ideal and point to other linkages between policy-making and implementation.

Application of the Conflict-Bargaining Perspective

The conflict-bargaining model is critical of the rational decision-making model for assuming that all participants in the decision-making processes share the same goals (March & Olsen 1984, 6; Lindblom 1959). According to the conflictbargaining decision- and policy-making processes are characterized by conflict among many different actors (individuals, organizations, public institutions, involved. Each participant follows his own interests, and agreement about goals is rare. Yet the participants will usually be better off if a decision is made in spite of the fact that any single alternative is unlikely to fully satisfy all participants. The participants engage in a bargaining process (Olsen 1972, 47, Pressman & Wildavsky 1973, Bardach 1977, Elmore 1978, Sætren 1983). The choice they arrive at will depend on the interests of the participants, the intensity of these interests, the resources of the participants and their ability to form coalitions. Also, the compromises reached in the bargaining process focus more on means than on goals.

Actors with different interests and political values can form a coalition, whereas one actor cannot endorse the goals peculiar to each of the other ones. As a contrast to the rationalist remarks by the former Danish Prime Minister Hilmar Baunsgaard, who said that it is necessary to agree about goals before discussing means, C. Schultze maintained: 'The first rule of the successful political is "Don't force a specification of goals or ends'" (Cronbach et al. 1981, 130). Though the interests of the actors involved may differ, the goals may be convergent. This implies that one means may satisfy different goals (Haas 1964). Convergent interests are often the basis of compromise.

Each participant may seek to act rationally in maximizing his own values, but the decision-making process as a whole, with its pattern of conflicting participants values, does violence to the assumption of clear preferences in the rational theory. According to the conflict-bargaining model, goals stated during and after the decision-making process may have a strategic function that is different from that attributed to them by the rational perspective. Rather than being real goals, statements of goals may have an instrumental role in persuading other actors and in legitimizing positions and decisions aimed at quite different and unofficial goals. Also, participants often want to play down or hide a conflict after a compromise has been reached. Accordingly, they may be motivated rationalize the compromise by referring to other (and often very abstract or vague) values than their various core values.

This description of policy-making from a conflict-bargaining perspective is well known from both legislative and implementation studies. However, when implementation studies describe failures and causes of failures, they often seem to forget the character of the policy-making process that preceded the decision that is to be implemented.

Applications of the conflict-bargaining model to the linkage between policymakingand

makingandimplementation make evaluation and implementation studies more complicated. If laws and other policy-decisions are results of conflict and bargaining among different interests, how, then, can implementation success or failure be measured? Official goals may represent only a minor share of the goals of the important participants or may even be misleading or irrelevant. Neverthelesswe chosen to evaluate the implementation against the official goals. But we find the notions of conflicting interests and variable influence useful in explaining the adopted programme and its stated objectives.

Applying the conflict-bargaining perspective for the purpose of generating hypotheses about links between policy-making and implementation, the following hypotheses seem plausible: The conflicting interests of decisionmaking may lead to policy programmes that have no clear objectives. goals may even be inconsistent with each other, and the official goals may be quite different from the latent but real goals of the participants. Because of compromising, a given programme may even end up being so far from the interests of the policy-makers involved that it is impossible to discover any of the objectives originally aimed at. In such cases it would be difficult indeed to identify any implementation success. Compromising may also lead to programmes are technically impossible to implement. In sum, the probability that it is possible to identify successful implementation is likely to be negatively related to the degree of conflict in the policy-making phase.

The main conclusion of our case study is, however, that there was surprisingly little conflict about that particular reform. First of all, the main principle of decentralization was apparently not questioned by anyone participating in the policy-making process. Nevertheless the actors had quite different goals, some of which were quite different from the official ones.

In the following we shall try to identify different groups of actors and interests. To some of the original initiators of the decentralization reform, the primary goal was to change the decision practice. The decentralization idea was probably born by Bent Rold Andersen, a scholar of welfare. He had done a study on disablement and labour market participation, which made him very critical about the predominantly medical criteria for assigning disablement pensions. He found that these criteria were only weakly related to actual labour market participation and salaries. Therefore, he proposed that social criteria should be given greater weight so that a disablement pension should be granted when further rehabilitation was considered unrealistic. He thought that a decentralization DIT could prepare the way for changing the criteria in evaluating the merit of applicants by getting the casework away from the medically dominated DIT, by introducing more local information about the clients and local labour market conditions, and by coordinating the granting of pensions with the granting of rehabilitation and sickness allowance.

Rold Andersen's idea was supported by a young public official in the Ministry

of Welfare, Ole Høeg. Though his formal position was low, he had a substantial influence on the decentralization. He was a member of the secretariat for a committee which prepared a disablement pension reform with higher pensions from 1962 to 1964. The committee supported decentralization, bud did not make any detailed proposal about that. Later Ole Høeg was an influential member of the secretariat for a Welfare Reform Commission, which prepared a reform of the total Danish social welfare system, including the decentralization of DIT. Also Bent Rold Andersen had a considerable impact on that reform from a position as a consultant to the chairman of the commission.

It is remarkable that for these initiators the decentralization was a means to change the pension criteria whilst they did not expect any decrease in decisionmaking or costs. Nor did they anticipate that the uniform standard for decision-making would be preserved. Other participants, mainly bureaucrats but also some politicians, did not want to abandon the medical criteria for evaluating cases but were in favour of facilitating the introduction of more local - level knowledge of clients and their situation in order to strengthen the role of those social criteria already suggested in the disablement pension law.

To many politicians - particularly from the liberal and conservative parties, but to a considerable extent also from the Social Democratic Party and the Social Liberal Party - decentralization was considered a value in its own right; decentralization the code-word for 'getting closer to citizens'. At the same time decentralization was thought of as an instrument conducive to speeding up) decision-making, making decisions more fair, and decreasing costs. Apparently, most politicians did not go beyond this general way of reasoning in connection with the reorganization of DIT.

In addition, the decentralization of DIT was only a tiny part of a huge reorganization the entire Danish social welfare system. The main principle for the reform was decentralization implying that as many services as possible should be delivered by the municipalities while the residual client directed services - which were not suited for municipal administration - should be decided by regional authorities. Only overall regulation and planning should be performed by the central level.

With regard to speeding up decision-making, most political parties expected that decentralization would bring about the desired results. This expectation was shared by some of the officials in the Ministry of Welfare, by the Head of DIT (cf. below), and by the municipalities as represented by the Association of Municipalities. The municipalities were very critical of the slow pace of casework in DIT. Generally favouring decentralization, they were in particular interested in transferring as many functions as possible to the municipal level. Therefore the Association of Municipalities demanded that some of the functions intended for the regional Rehabilitation and Pension Boards should be placed with the municipalities.

This Association was very influential and more powerful than the Association of Counties, which, naturally, was interested in keeping as many functions as possible at the regional level. The staff of the Association of Municipalities was highly efficient, and the large majority of Parliament voted with favour on the claims for municipal responsibility. At first glance it might appear to have been a compromise solution, therefore, when in the end the functions originally intended for the regional Board were split between the counties and the municipalities. truth seems to be, however, that the municipalities did not want the responsibility assigned to the counties. The small municipalities in particular feared the pressure from local citizens applying for the very expensive benefits involved, as well as the envy and opposition the granting of such benefits could generate in a small community.

The handicap organizations did not fully agree with the decentralization of DIT. Particularly the organizations representing the smaller groups of disabled people, e.g. the deaf and the blind, were sceptical (though the chairman of the latter group who was also the chairman of the Welfare Reform Commission was in favour of decentralization). These groups feared that decentralization would reduce uniformity of decision-making considerably and that the expertise of DIT would be lost. The organizations for the major groups of handicapped (the mentally retarded and the physically impaired) were more in favour of decentralization. the end the League of Handicap Organizations accepted the decentralization DIT, partly because of the internal disagreement, partly because the League generally supported the overall decentralization principles of the new social welfare reform. (After the reform, however, the League has found the outcome the decentralization reform very disappointing.)

According to the conflict/bargaining perspective and also according to most studies on reorganization, DIT should have been expected to play a major role in the policy-making process. One of the most general observations offered by the relevant literature is that public institutions are characterized by an impulse to expand and that they fight - most often successfully - against any efforts to bring about their dissolution. If they accept termination, it is typically because they expect to get better chances of survival and growth within another institutional

The dissolution of DIT seems, however, to deviate from this pattern. The decentralization reform meant that the functions and most likely also the personnel of a central public institution in Copenhagen were to be spread all over the country. The personnel got no guarantees of future employment. Although many of the staff members were old enough to be eligible for a pension, it appears that the majority was against decentralization. However, the available data show no indication of any kind of either internal or external protest from the institution or its staff.

Externally DIT was represented only by its director, Mr. F.M. Hartmann - a

very unusual administrative actor who held many visions for the welfare system and initiated a number of reforms. As early as in 1960 he had initiated the decentralization of rehabilitation functions from DIT to ten new rehabilitation centres. When Hartmann first was confronted with the idea of decentralizing DIT, he voiced his opposition. However, he was persuaded to change his mind - probably by the prospects of improving the quality of social information on the clients, securing coordination of affiliated functions, and speeding up the decision-making process. Hartmann was such a strong leader that he could permit himself to come out in support of the decentralization reform even if the majority of his staff was against it. The absence of protests from the agency personnel might be attributed in part to loyalty toward a strong manager and in part to apathy. (We find it reasonable to believe, however, that similar behaviour would be less likely today due to the introduction of both formal and informal channels of influence for the personnel of public agencies.)

According to the conflict/bargaining perspective on policy-making and its effect on implementation, successful implementation is likely to be negatively related to the degree of conflict in the policy-making phase. In our case, however, the decentralization solution enjoyed broad consensus. Whatever opposition there was to the reform was only weakly articulated - if voiced at all. On the other hand, the motives for supporting the reform varied greatly. The main official goal of speeding up decision-making enjoyed broad support, but many of the participants in the policy-making process showed little interest in reducing costs. And some participants placed special emphasis on the importance of improving the quality of case-work and allowing for social circumstances to be taken into account. The presence of a variety of interests and values, both official and unofficial, manifest and latent, makes the analysis of implementation difficult. is doubtful whether the motivating force behind policy decisions can be found in officially stated goals. Accordingly, their validity as evaluation standard is brought into question.

The analysis supports our hypotheses to the effect that the relationship among policy-making, policy-goals, and implementation of such goals becomes more uncertain and less consistent if the policy-making phase is characterized by conflicting interests. However, since in our case surprisingly little conflict has been identified, the conflict-bargaining model has only limited value in explaining relationship between policy-making and implementation.

Application of the Garbage-Can Perspective

The 'garbage-can model' is substantially more complicated than the previous ones, but it also claims to offer a more satisfactory description of many kinds of decision-making. In contrast to the instrumental view of both the rational and the conflict-bargaining models, the garbage-can model does 'not specify any

direct connections between the purposes and intentions of the actors and the events we want to explain' (Olsen 1972, 48). The model presumes that decisionmakingsystems often characterized by very loosely coupled relationships among participants, problems, solutions, and choice opportunities, all of which are mixed in a manner similar to that of refuge dumped into a garbage can (Olsen 1972, March & Olsen 1983). These elements are related in a temporal rather than consequential order. Automatic or unconscious aspects are dominating,and choice is seen as 'a product of certain processes having dynamics of their own, which by their interactions generate outcomes which are not intended by anyone' (Olsen 1972, 48).

Decision-making participants are normally only parttime participants with limited capacity and attention. Their behaviour in any decision-making process is therefore influenced by their concurrent roles. Consequently the attention given to any specific decision will depend both on the character of that decision and on the alternative claims for the attention of the actors eligible for participation 1972, March & Olsen 1984, 18). Also, the longer the decisionmaking and the more participants who enter the process, the more questions likely to be 'dumped into the can', with a corresponding decrease in attention likely to be paid to any given issue. Even if a given political decision may be regarded as extremely important by the outside observer, it is likely to be considered quite unimportant by many of the participants (March & Olsen 1984, 17). Participants also come and go, and this turnover increases with the length of the decision-making period.

Each choice opportunity is regarded as an ambiguous stimulus. 'What is being decided will itself be determined through the course of deciding it'. Organizations not only an instrument for problem solving, they are 'also a collection of choices looking for problems, issues and feelings looking for decision-inprocess which they can be mediated, and solutions looking for questions' 1972, 49). As a consequence, the decision-making process is very much affected by timing.

The garbage-can model also emphasizes the symbolic aspects of decisionmaking.Organizations their individual members feel a need to convince their observers that the decisions they are making are legitimate. Legitimacy can be obtained by referring both to appropriate objectives reached by the decision, or to the way the decision has been made. Often the latter kind of legitimization is the easiest one. It seems that for some participants it is more important to participate in the decision-making process than actually influencing the outcome.The itself and participation in it may have a considerable symbolic value. By the same token decisions may also be taken for symbolic reasons. Demonstrating a will to change things may for some actors be more important than actually implementing their proposals or decisions. Because of the dominationof and instrumental values in our culture, reorganization of

public institutions may have a symbolic value in demonstrating a will to solve
problems (March & Olsen 1983, 1984, Sætren 1983, 14-15).

Especially when attention is a scarce resource, participants may also be attracted by solutions that have an immediate symbolic and compelling appeal to the participants. In reorganization, decentralization may represent such a symbol. Decision-making rituals may also require certain symbols, depending on the type of policy-area, to be used as legitimization. As an example, reorganization seem to be heavily dependent on the symbols of economy, efficiency, and expeditious administration of public business (March & Olsen 1983, 283).

If the logic of garbage-can model is applied to an analysis of the linkage between and implementation, the connections seem to be even more fragile than suggested by the conflict-bargaining model. If policy-making is performed like garbage dumping, there does not seem to be much hope for documenting implementation success, at any rate if success is defined in an instrumental way. If the function of policy-making is primarily ritualistic and symbolic, it becomes difficult to even raise the question of what is to be implemented. How can symbols be implemented? In addition, to the extent the garbage-can model is descriptive of both policy-making and implementation, the two processes are ordered in a temporal rather than in a consequential manner. And with a predominance of new participants in the implementation phase, hardly any substantial link can be expected to exist between policy-making and implementation. Besides, if values and attention to issues are changing all the time, the policy-makers cannot be expected to care much about implementation. fact, the model may help explain why politicians pay so little attention to implementation (March & Olsen 1983, Sætren 1983).

The garbage-can model may be useful in accounting for the apparent consensus the relatively little interest accorded to the decentralization of DIT. Most participants in the policy-making phase were part-time participants. Their attention to the specific problem of reorganizing DIT was dependent on other claims to their attention, especially in view of the extreme length of the policymaking For example, when the question of changing the pension levels was placed on the agenda of the Disablement Pension Committee in the early sixties, that issue attracted much more interest than did the reorganization of DIT. When the question was subsequently treated in the Welfare Reform Commission, reorganization of DIT was regarded as a very minor problem in relationship to the enormous task of changing both the content and the organization the whole welfare system. With the possible exception of representatives the handicap organizations, the members of the Commission could, therefore, hardly be expected to pay major attention to the reorganization of DIT. An additional factor in this connection is that the DIT issue was also far less controversial than many other issues.

In addition, there was a substantial turn-over of participants in the decisionmaking as the decentralization issue moved from one decision-arena to another (cf. note 3). The effect of a heavy turn-over may be changes in policy, but it may also very well be uncritical acceptance of almost any proposals stated in the beginning of the policy-making phase. The solutions are transferred from one decision-arena to another, but the coupling between solution and goals becomes more and more tenuous.

Timing is also very important in relation to the advancement of reform proposals. Perhaps the likelihood of getting the decentralization proposal accepted politically increased over time exactly because the decentralization of DIT was not treated as an isolated question, but as a relatively minor part of a larger and very important and complicated reform.

The policy-making process was probably also characterized by the use of symbols. For example, the belief in decentralization described above was so strong that it was generally accepted that central welfare authorities in the future should have no hand in dealing with individual welfare cases. It was simply taken for granted that as much responsibility as possible should be placed with the municipalities, and the rest with the authority second closest to the citizens, namely the counties. It seems likely, therefore, that decentralization of DIT represented a symbol of such importance to the policy makers that it precluded any serious consideration of alternatives and their consequences.

As suggested, the garbage-can model does not specify any direct connections between the goals of the actors and the event to be explained. We have been surprised to observe that in our case the purposes of decentralization are hardly mentioned in the reports and debates concerning the issue. Only in the final bill of 1975 do we find any specific references to goals. Examining the drafts of this bill, it becomes apparent that goals and purposes were articulated late in the course of the drafting of the bill. The goals of reducing decision-making time was, in fact, added as one of the last corrections in the final draft. Although this procedure should not be overinterpreted, it suggests that goals may not be the guiding-star of every human and organizational behaviour. In our case, certainly the means (the articles in the bill about the decentralization) seem more important the goals. The articulation of goals may also have had a legitimizing function since rationalist rituals require that actions and laws can only be justified if they are purposeful.

Another and related aspect of the garbage-can model is that goals change over time. Mapping backward in time from the passing of the decentralization bill in 1975 (cf. note 3), we find reasons to suspect that the most important goals for both the bureaucrats in the Welfare Ministry and the politicians were not the official ones, but to fulfill an earlier promise to decentralize DIT. Similarly, the task of a special committee on decentralization of DIT was - from 1972 to 1975 - not to reconsider goals and means, but to fulfill an obligation (dating from the

legislation in 1970 regarding the reorganization of the welfare system) to specify how decentralization could take place. By the same token, the 1970 legislation was nothing more than a confirmation of a compromise reached by most of the political parties in the Welfare Reform Commission.

If this interpretation is correct, the substantive goals are traceable to the Welfare Reform Commission in spite of the fact that these goals apparently were not explicitly articulated by the Commission, and the decentralization was discussed only for a few minutes. More likely, the Welfare Reform Commission had just adopted the solution from the earlier Disablement Pension Committee, which in its 1964-report in only two sentences had supported decentralizations of DIT. When first in 1968 the decentralization of DIT had been accepted politically a part of a compromise about the entire social welfare system, it seems that in each of the subsequent phases of the policy-making process operationalizing legitimizing earlier decisions and commitments were more important than considering policy goals. The solution was regarded as a given precondition.

Particularly in the last phase of the legislative process leading to decentralization DIT, legitimization appears to have been a primary concern. Here we find the first situation where the Parliament discussed that specific reform separately. By that time new participants had entered both the bureaucratic and the political scene, and it may not have been easy for them to indentify goals that could explain the reform. At first, we thought that the official goals of reducing decision-making time had been invented at that time, but on closer examination we have concluded that the source probably is a 1968-report from the committe on internal reorganization of DIT that has been mentioned above (Betænkning 1968, 14-15). It seems, however, that compared to this report the bill focusses less on the goal of improving the quality of decision-making by means of increasing the level of local-level information on the clients and their circumstances as well as by individualizing the case-work. Improvement of the level of information and individualized case-work are, rather, seen as instruments to shorten decisionmaking This different focus on values may be accidental, but it raises questions regarding how much importance can be attached to officially formulated goals and their priority.

On the basis of our data it is not possible to find all the elements contained in the garbage-can model reflected in the particular policy-making process we have taken as our case. It does not display quite the accidental character predicted the model. But its notions regarding limited attention, importance of symbols and the loose coupling of solutions and goals have proven useful.

These aspects of policy-making presumed by the garbage-can model are also helpful in explaining why implementation did not fulfill the policy goals. The chances of successful implementation of official policy goals must be expected to decrease to the extent that the policy-making process concerning one particularproblem

cularproblemis characterized by the involvement of many part-time participants,by attention, changing values, symbolic politics, and ritual legitimizationof through the tactic of referring to or inventing appropriate goals and consequences.

Conclusion

This study represents an attempt to integrate the study of policy-making/legislation policy-implementation. Only a few scholars of policy-making and legislative have been interested in what happens after the passage of a bill, and most implementation scholars restrict their interest to the very implementation where they also seek their explanations for the frequent implementation

The thesis of this study is that implementation problems often are caused not only by behaviour in the implementation phase, but also by the character of the policy-making process prior to the passage of a law or other kind of political decision. In the absence of any body of theory on how policy-making affects implementation, the study attempts to derive hypotheses about such linkages from three well-known models of policy-making: the rational decision-making model, the conflict-bargaining model, and the garbage-can model. These theoretical are applied to a case study of the decentralization of the Danish administration of disablement pensions and other aid to disabled persons in 1975.

Most studies on reorganization show that reorganization proposals are rarely implemented in the sense that the reorganization actually takes place. If it does take place, it rarely fulfills prior expectations and goals. Though in our case reorganization did take place, it was a failure in terms of producing the intended results. The main structural element of the reform was the decentralization of the national Disablement Insurance Tribunal (DIT) which, along with a few other agencies at the national and regional levels, was replaced by 15 new regional Rehabilitation and Pension Boards assisted by the municipalities. However, the reform has not reached its official goals of speeding up the processing of cases, reducing administrative costs, and ensuring a uniform standard of decisionmaking. fact, decision-making time has more than doubled, the productivity of staff has decreased by more than 50 per cent, and the decision-making practice seems to differ substantially from region to region.

From the perspective of the rational decision-making model, we have derived the hypothesis that implementation failures may be more likely if a) preferences are unclear or absent in the policy-making phase, if b) at least a few relevant alternatives have not been considered, and if c) no qualified attempts have been made to estimate the consequences of the alternatives or, at a minimum, the preferred alternative. In addition, an implementation failure is more likely if

changing conditions during the policy-making process do not lead to a revision
of the evaluation of alternatives and consequences.

Although in our case the policy-making process took more than a decade and in spite of the fact that the official goals in the legislative output were relatively clear, it appears that very little consideration of alternatives and their consequences place. Also, some important circumstances changed during the decision-making process, but these were not taken into consideration.

According to the conflict-bargaining model, policy-making is characterized by conflict among actors with different interests and resources. Most decisions are compromises, and they focus on the means rather than the values. This makes a clear definition of values difficult. New values may also be invented in order to legitimize a compromise where the goals of the participating actors differ. From this model we have derived the following hypothesis: the probability that official policy goals will be successfully implemented is likely to be negatively related to the degree of conflict in the policy-making phase.

It seems, however, that in our decentralization case there was a considerable degree of consensus. Only two groups did not favour reform: the handicap organizations (which did not protest because they regarded the decentralization of DIT as a part of a larger package with which they otherwise were satisfied); and the majority of the personnel of DIT (they were not consulted, and, surprisingly, did not protest). But while there was a high degree of consensus about the means, the actors that participated in the policy-making process did differ with regard to goals and their priority. These findings support the notions that one can hardly take official goals at their face value when searching for an evaluation or implementation standard. Conflicting interests probably render the relationship among policy-making, policy-goals, and implementation of such goals more uncertain. Since, however, surprisingly little conflict has been identified, the conflict-bargaining model may have only limited value in our case.

The garbage-can policy-making model emphasizes the loose connection among participants, values, problems, choice opportunities, and solutions. The model does not assume any consistency between the purposes of the actors and their behaviour. One of the most important notions of the model is that of limited attention. Most of the participants in the decision-making process are only part-time participants with competing claims for attention from other roles. Also, participants are often exchanged with new ones. The level of attention furthermore, affected by the length of the decision-making period, the number of participants, and the number of competing issues on the agenda. Values change over time. Symbols are as important as results. These characteristics garbage-can decision-making imply that any direct relationships and internal consistency among policy-making, stated policy-goals, and implementation be expected.

Some aspects of the model can be applied to our case even if the policymakingprocess

makingprocesswas not quite as accidental as the model suggests. The notion of limited attention is important to an understanding of the process that took place. The decentralization of DIT was considered not as an isolated reform but as part of a massive reform of the entire social welfare system and the local government system. Other issues contained in this larger package attracted most of the attention of the many different people who participated in the policy-making process in the course of the 13 years it lasted. In fact, it appears that only one page of paper and only a few minutes of debate were devoted to the DIT decentralizationreform the fora most directly responsible for the content of the decision; and it seems that values were not an issue before the final debate. And the unproblematic passing of the bill was rather a confirmation of earlier decisions than a new one.

With regard to the importance of symbols in the policy-making process, 'decentralization' enjoyed wide and uncritical support in the sixties and seventies. was regarded as both a democratic value in its own right and as an instrument to improve and individualize public services to citizens, to speed up decision-making, and to reduce costs. The decentralization of the disablement fitted beautifully into this pattern of belief. But symbols are difficult to implement!

In our analysis we have not employed any single theoretical framework for dealing with the relationship between policy-making and subsequent implementation. by Allison (1971) and Elmore (1978), we have found it useful to apply different models of policy-making. Each has proven relevant in identifying different factors in the policy-making process which may impede successful implementation. The rational decision-making perspective shows that irrational behaviour is likely to cause implementation failures, and the conflict-bargaining and garbage-can models do not only help to explain why decision-making often is irrational, but also suggest how conflicting interests, limited attention and the intervention of symbols held dear by the policy makers reduce or even obstruct consistency among policy-making, policy, and implementation.

These findings have important implications for evaluation and implementation Implementation successes and failures and programme effects can only be interpreted in an instrumental way. A success implies that a programme has proven to be a useful means to obtain certain goals. For practical and democratic most evaluation and implementation studies focus on the fulfilment official policy goals, but this study as well as many others demonstrate that policy-making processes are often far from the instrumental logic of the rational model and that official goals may not be taken at their face value.

This places the evaluation and implementation researcher in a dilemma when he is trying to identify a reasonable evaluation standard. If he focuses on official goals he will often be unable to explain the goal achievement without investigatingwhat in the prior policy-making phase. Another tempting

research strategy more in line with the popular conflict-bargaining perspective is to evaluate the goal-fulfilment in relation to the interests of the various actors that have participated in the policy-making process - and maybe also in relation to the actors of the implementation process (Hjern & Hull 1982, Sætren 1983). But this strategy tends to disregard the official goals stated in Parliament and the question of democratic accountability. In addition, even the conflict-bargaining perspective relies on assumptions of instrumentality, intensity of interest and attention in the policy-making phase that are not always realistic. How can we evaluate a programme implementation if the policy adoption was a symbolic action or if the policy-making actors did not care particularly about the policy solution?

In any case, however, the relevance of evaluation and implementation research is likely to increase considerably if both the policy-making and implementation processes are studied (Winter 1985), and this should be an obvious task for political scientists.

---

NOTES

1. Productivity is defined as the number of cases processed per social worker/head in one year. Although we recognize that it is often problematic to measure productivity in administrative organizations, we find it justified to employ this rough indicator here, because of the very high number of cases (now 27,000 a year versus 45,000 before). Also, the cases remain rather similar, and the criteria for granting a pension have not been changed with the reform.

2. The study of the policy-making process before the decentralization reform is basedon data from 1) official documents, i.e. bills, parliamentary debates, reports from parliamentary committees, reports from commissions and committees dealing with this issue between 1962 and 1975, annual reports from DIT; 2) unofficial written material about this and related issues from the same period available from the archive of the Ministry of Welfare, including reports on the preparation bills and proposals to, and reports from, the meetings of different commissions, committees subcommittees. Among the data are also interviews made in 1985 with 13 persons who participated in the policy-making process, which lasted from 1962 to 1975. The quality of the interview data varies substantially, as must be expected when asking about events which took place 10-23 years ago. Besides, for many of the interviewees, this particular reform played only a minor role in their professional career. Also due to the time that has passed, some of the key participants in the reorganization discussions had died before the interviews were made. The interviewees are treated anonymously and no references to single interviews are made.

3. The Decision-Making Process and Decision Points Concerning the Decentralization of DIT Decision Decision Points Main Output Time 1. 1962-64 Committee on Disablement Proposal for increasing disablement Pension pensions for the most disabled persons. Decentralization of DIT briefly supported 2. 1964-68 The Welfare Reform Proposal for a reorganization and Commission decentralization of the total Danish social welfare system, including decentralization of DIT 3. 1965-68 Committee on internal Reorganization proposals. Support for reorganization of DIT to decentralization of DIT reduce delay in case-work 4. 1970 Parliament Passing a reorganization bill basedon the principles of decision no. 2. The bill established the regional rehabilitation pension boards (RPB) 5. 1970-72 The Welfare Reform Proposal for the material content of Commission the social welfare laws and more detailed division of labour between the new organizational units, also in relation to the RPBs 6. 1972-75 Committees on the decentrali- More detailed decentralization propozation DIT and related sal. Estimation of personnel needs functions 7. 1975 Parliament The passage of a bill decentralizing the DIT and related functions

3. The Decision-Making Process and Decision Points Concerning the Decentralization of DIT Decision Decision Points Main Output Time 1. 1962-64 Committee on Disablement Proposal for increasing disablement Pension pensions for the most disabled persons. Decentralization of DIT briefly supported 2. 1964-68 The Welfare Reform Proposal for a reorganization and Commission decentralization of the total Danish social welfare system, including decentralization of DIT 3. 1965-68 Committee on internal Reorganization proposals. Support for reorganization of DIT to decentralization of DIT reduce delay in case-work 4. 1970 Parliament Passing a reorganization bill basedon the principles of decision no. 2. The bill established the regional rehabilitation pension boards (RPB) 5. 1970-72 The Welfare Reform Proposal for the material content of Commission the social welfare laws and more detailed division of labour between the new organizational units, also in relation to the RPBs 6. 1972-75 Committees on the decentrali- More detailed decentralization propozation DIT and related sal. Estimation of personnel needs functions 7. 1975 Parliament The passage of a bill decentralizing the DIT and related functions

REFERENCES

Allison, G. 1971. Essence of Decision: Explaining the Cuban Missile Crisis. Boston: Little Brown.

Bardach, E. 1977. The Implementation Game. Cambridge: The MIT Press.

Beretning fra Invalideforsikringsretten 1971. Beretning fra Invalideforsikringsretten for finansårene
til 1969/70. Copenhagen.

Beretning K.L. 1973. Beretning over Kommunernes Landsforenings virksomhed 1972/73.
Copenhagen.

Beretning S.A.S.U. 1973. Beretning fra sammenslutningen af sociale udvalg 1971/72. Copenhagen.

Betænkning 1965. Betænkning om invalidepension m.v. Bet. nr. 371/1965. Copenhagen.

Betænkning 1968. Betænkning vedr. Invalideforsikringsrettens administration. Copenhagen.

Betænkning 1969. Socialreformkommissionens 1. Betænkning. Det sociale tryghedssystem.
Struktur og dagpenge. Bet. 543/1969. Copenhagen.

Betænkning 1972. Socialreform kommissionens 2. betænkning. Det sociale tryghedssystem.
Service og bistand, Bet. 664/1972. Copenhagen.

Betænkning 1973. Betænkning vedr. mulighederne for at decentralisere invalideforsikringsrettens
førtidspensionsudvalgets afgørelser. Copenhagen.

Betænkning 1975. Afsluttende betænkning vedr. decentralisering af invalideforsikringsrettens
og førtidspensionsudvalgets afgørelser. Copenhagen.

Cronbach, L.J. et al. 1981. Toward Reform of Program Evaluation. San Francisco: Jossey-Bass
Publishers.

Elmore, R.F. 1978. 'Organizational Models of Social Program Implementation', Public Policy
26, 185-228.

Elmore, R.F. 1979-80. 'Backward Mapping: Implementation Research and Policy Decisions',
Political Science Quarterly 94, 601-16.

Folketingstidende (different annual volumes).

Hass, E.B. 1964. Beyond the Nation State: Functionalism and International Organization.
Stanford: Stanford University Press.

Hjern, B. 1978. Implementation and Network Analysis, Discussion Paper, Berlin: lIM, Wissenschaftzentrum.

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

Hiibbe, P. 1978. Invalidepensionsundersøgelserne4. Ansøgere til invalidepension. Copenhagen:
Socialforskningsinstituttet.

Jeune, B. & Sabroe, S. 1980. 'Three-Year-Incidence of Decided Disability Pensions in Denmark
Before Decentralization', Scandinavian Journal of Social medicine, suppl. 16, 8-16.

Lindblom, C.E. 1959. 'The Science of Mudding Through', Public Administrative Review 19,
79-88.

March, J. & Simon, H.A. 1958. Organizations. N.Y.: John Wiley.

March, J.G. & Olsen, J.P. 1983. 'Organizing Political Life: What Administrative Reorganization
Tells Us about Government', American Political Science Review 11, 281-96.

March, J.G. & Olsen, J.P. 1984. Garbage Can Models of Decision Making in Organizations.
Paper.

Meier, K.J. 1980. 'Executive Reorganization of Government: Impact on Employment and
Expenditures', American Journal of Political Science 24, 396-412.

Meyer, P. 1973. Systemic Aspects of Public Administration. Copenhagen: Gad/Aarhus: Politica.

Morgan, D.R. & Pelissero, J.P. 1980. 'Urban Policy: Does Political Structure Matter?', American
Political Science Review 74, pp. 999-1006.

Olsen, J.P. 1972. 'Public Policy-Making and Theories of Organizational Choice', Scandinavian
Political Studies 1, 45-62.

Pressman, J.L. & Wildavsky, A. 1979. Implementation, 2nd ed., Berkeley: UCLA Press.

Seidman, H. 1975. Politics, Position and Power. The Dynamics of Federal Organization, 2nd ed.,
N.Y.: Oxford University Press, London.

Sikringsstyrelsen 1978. Siksringsstyrelsens undersøgelser, nr. 3. Afsluttende statistik fra invalideforsikringsretten
førtidspensionsudvalget. Copenhagen.

Sikringsstyrelsens pensionsstatistik 1976/77, 1977, 1978, 1979, 1982.
Simon, H.A. 1976. Administrative Behaviour, N.Y.: The Free Press.

Socialcentrene 1972. Forslag til socialcentrenes organisatoriske opbygning, Copenhagen: Amtsrådsforeningen.

Sætren, H. 1983. Iverksettning av offentlig politik. Oslo: Universitetsforlaget.

Winter, S. 1980a. Decentraliseringen af førtidspensionerne. Forventninger og resultater. Århus:
Århus Amtskommune, Social- og Sundhedsforvaltningen.

Winter, S. 1980b. Goals and Effects of Decentralizing the Implementation of Disablement
Pension Policy in Denmark. Paper, The International Working Group on Policy-Implementation,

Winter, S. 1983. 'Studying Implementation of Top-Down Policy From the Bottom Up. Implementation Danish Youth Employment Policy', Ch. 5. in Rist, Ray C, (ed.), Finding Work: Cross-National Perspectives on Employment and Training. London: Falmer Press Ltd., The Toylar & Francis Group, Forthcoming 1986.

Winter, S. 1985. 'Iværksættelsesbarrierer', Politica 17, 467-87.