Governing structures and structured governing: local political control of public services in Denmark.
By employing a multidimensional conceptualization of policy diversity and devising quantitative indicators, the analysis shows that less policy diversity is found in the two sectors formally controlled by the local governments. The institutions that structure governing, notably the separation of provider and producer roles, are more important than government structure.
In theory as well as in political-administrative practice, the division of responsibilities between central and local governments is a contested issue that pertains to very basic questions
This research is part of a study of the Danish local-govemment reform financially supported by the Danish Democracy and Power Study Project. An earlier version of this article was presented at the tenth Nordic Political Science conference in Uppsala, Sweden, and the fifth Public Management Conference in Barcelona. Spain. We are grateful for comments and suggestions from the participants of the workshops and from our present and former colleagues in Aarhus. We are especially grateful for the subsequent comments from Lotte Bogh Andersen, Peter Bogason, Jens Blom-Hansen. Jorgen Gronnegaard Christensen, Ron King, and Soren Risbjerg Thomsen. Lone Winther has provided invaluable editorial assistance and text revision of the English version of the article, and Brian Larsen has collected and processed data for the labor-market policy analysis.
of the proper size of democracy (Dahl 1970) and what constitutes the "natural" nucleus of a democracy (Sharpe 1970). In federal democracies, local democracy is seen as a way to check and balance central-government power. One of the fundamental rationales for federalism is that the states are guaranteed a right to devise different solutions to local problems (of. Federalist Papers nos. 45 and 46 in Hamilton, Madison, and Jay 1961).
The extent and content of local democracy has also been a major political issue in unitary states. Numerous theories and normative rationales have been suggested to explain and legitimize decentralization of authority, defined here as the devolution of power and responsibilities over policies from central government to local government (Brans 1992; De Vries 2000; Ostrom 1972). However, as recently shown by Michiel De Vries, both theoretically and empirically there is little agreement about the virtues of decentralized government. One of the few (largely) unchallenged advantages of decentralization is its capacity to enhance "the flexibility, responsiveness, and pluriformity of public service delivery" (De Vries 2000, 202). Thus, although equality may suffer, a chief attraction of decentralization is the fact that it increases "the possibility of tailor-made policies" (De Vries 2000, 193).
This article scrutinizes what is possibly the strongest argument for decentralization, namely that decentralized governance allows for adaptation to local needs and wants and therefore leads to more policy diversity than in centralized governance. The idea is that elected local governments have more opportunities and stronger incentives to accommodate local preferences and socioeconomic and demographic challenges (Tiebout 1956). In effect, decentralization increases policy diversity across localities.
Our theoretical point of departure is a neoinstitutionalist approach to public policy. A wide web of formal and informal rules conditions the impact of governing structures (e.g., Lynn, Heinrich, and Hill 2000; Norgaard 1996; Weaver and Rockman 1989). The wider political-institutional context of decentralized policy making shapes policy in ways that may strengthen, inhibit, or even annul the effects of decentralization (Christensen 2000; cf. Milward and Provan 2000). Even within a decentralized government structure, central government regulation, national collective agreements, or professional standards may be so strong that the possibilities for making different policy choices are suppressed. Furthermore, the idea that a decentralized government structure provides sufficiently strong incentives for policy makers to devise policies tailored to local needs and wants may not be valid. Internal decentralization within central government may be equally effective, and the organization of provider-producer roles may be more important than the distribution of formal authority. Decentralization does not automatically lead to more policy diversity.
In this study we empirically analyze the degree of policy diversity in hospitals, high schools, and the labor market in Denmark. Our focus is on system-level variation, that is, the degree of policy diversity across the three sectors. All three policies are public services organized on a regional level, but whereas the former two are county responsibilities, the latter has remained under the auspices of the central government. Within the central, hierarchical structure, corporatist regional councils have been granted wide autonomy in deciding how to implement regional labor market policy for the unemployed. It is more important, however, that the provider and the producer roles have been divided only in regional labor-market policy, and this may make it easier for the corporatist councils to make and implement different policies.
In general, we find more diversity in regional labor-market policy than in the two public services that are core county-council responsibilities. In different ways, central regulation and professional standards reduce the autonomy in the two sectors under county-council control and provide incentives for uniform policy choices across counties. Policy diversity is high only in one aspect of hospital services.
In the following section, we first discuss aspects of the theory of decentralization and other institutional determinants that structure governing in local public services. Second we discuss policy indicators and present a multidimensional, quantitative analysis of policy diversity in high school and hospital services and regional labor-market policy. In the conclusion we summarize the findings and discuss why different aspects of policy diversity display diverse patterns.
DECENTRALIZATION, INSTITUTIONS, AND LOCAL PUBLIC SERVICES
European local governments have been broadly categorized as belonging to a Napoleonic southern or a non-Napoleonic northwestern group (Sharpe 1979). In the former group, functional hierarchies extend from the center to the localities; in the latter group we find functionally stronger and more autonomous local governments (cf. Brans 1992). Profound reforms of local-government size and function have taken place in northwestern Europe only since the 1960s (Brans 1992). One of the important goals of these local-government reforms has been to increase local governments' capacity to tailor their policies according to local preferences and circumstances.
In Denmark amalgamated, larger local governments and a shift from a system of central reimbursement to unconditional block grants were the principal means to empower the local governments. The old system of reimbursement had been conditional on compliance to strict guidelines that were unilaterally decided by central government. The goal of the new regime of central-government subsidies and interlocal-government transfers is to level out local differences in objective spending needs and wealth. Within this new framework, a number of central-government tasks and functions have been transferred to local authorities (Ministry of the Interior 1998).
Nowhere in the family of non-Napoleonic unitary states has the empowerment of local governments since the 1960s gone so far as in Denmark (Blom-Hansen and Pallesen 2001). Two-thirds of total public consumption is now a local-government responsibility, and 80 percent of local expenditures are raised by way of local revenues. High schools and hospitals are core responsibilities of the counties. For historical reasons, Danish labor-market policy has remained a central-government responsibility (Norgaard 1997, 1999). The labor-market organizations have always been closely integrated in the governing structure in labor-market policy through a dense web of corporatist bodies. In recent years labor-market services have been regionalized, and the structure of the regional-state authorities is congruent with the territories of the counties in charge of hospital and high school services.
In general Danish local governments have a comparatively high degree of autonomy (Blom-Hansen 1999). Local needs and demands can be translated into local decisions about the scope and organization of public services; however, we can expect different policy choices only if local-political needs and wants vary. In terms of both socioeconomic conditions and political conditions, the Danish regions vary quite a lot (Indenrigsministeriet 1997).
In 1997 the smallest Danish county had some 230,000 inhabitants, whereas almost 630,000 lived in the largest county. Population density varies even more. The unemployment rate varied from 5.8 percent to 9.7 percent in 1997. The frequency of high school students varied between 28.3 percent and 42.7 percent of the relevant age group. County income-tax levels vary less (from 10.2 to 11.7 percent), partially due to the interlocal-government redistribution scheme.
The counties also differ in terms of political leadership. Around half of the counties are usually governed by a coalition headed by the social democrats; the other half have bourgeois mayors. More than half of these mayors have had very long tenures (Pallesen 1998). Therefore, following the classic ideas of centralization and decentralization, it can be expected that Danish counties exploit their opportunities to make policy choices that are informed by the different political wants and needs that they face.
If government structure is decisive for local policy, we expect to find more policy diversity in the public services controlled by the locally elected county councils than in regional labor-market policy. Ceteris paribus, we expect policy diversity to be equally high in the hospital and high school sectors. But other institutions may constrain county autonomy. We discuss the last point first.
Constraints on County Councils: State Regulation, Collective Agreements, and Professional Standards
Kjellberg has argued that local autonomy from the central government is seldom an "either-or" but often a matter of degree and form (1985, 1988; cf. also Bogason 1996; Brans 1992). In a number of ways central government regulates the activities of the localities (Blom-Hansen 1999; Christensen 2000; cf. Lynn, Heinrich, and Hill 2000). Statutes, ordinances, and other kinds of central-government regulations are the most typical form, but rules laid down in collective agreements and professional standards may have the same impact.
Central-government regulation is tighter in high schools than in hospital services. Government regulation of county-hospital policy is quite modest (Pallesen 1999). The Hospital Act merely states that running hospitals is a county responsibility and that hospital treatment is free of charge for all patients. The regulation of county-hospital policy focuses mainly on professional supervision (certification of medical professionals and drugs, general treatment guidelines, etc.). Apart from this, central regulation of county-hospital policy is sparse and sporadic. With very few exceptions (e.g., the right to induced abortion within the end of the twelfth week of pregnancy) there is no central regulation of the intake of patients. The counties have no obligation to provide immediate medical treatment except in the case of acute illness. In effect the counties have considerable autonomy in organizing hospital policy.
Central-government regulation of high schools is denser than regulation of hospitals (Pallesen 1998). Government regulation prescribes that all students who are qualified according to an evaluation by the primary schools have a right to be admitted. The curriculum is also regulated in great detail and so is the minimum number of lessons per student. County autonomy is very limited (Association of County Councils 1997; Ministry of the Interior 1998; cf. Pallesen 1998, 102).
Centralized collective agreements between the parties on the labor market may also have homogenizing effects on service delivery. In the so-called Danish model, much labor-market regulation, for example, working hours, minimum pay, and the right to paid maternal leave, is settled by way of collective agreements (cf. Stroby Madsen 1998). In particular in the public sector, collective agreements reduce the autonomy of the counties. Nationwide collective agreements may be a causal equivalent to state regulation. There is, however, substantial variation in the collective agreements across sectors.
In the hospital sector, collective agreements regulate hospital-staff education, notably for interns and young doctors, in ways that have some implications for work organization. Collective agreements between the various unions and the Association of County Councils also regulate overtime, rest time, time off in lieu of wages, and so on, but they do not regulate the functions performed by the staff. In sum this means that the counties have considerable autonomy in deciding the organization of the work processes in the hospitals.
In the high school sector, the tight central-government regulation is paralleled by an equally strong regulation through the collective agreement between the counties and the national union for high school teachers (Pallesen 1998). In minute details the agreement regulates pay and working conditions as well as all functions and tasks performed by the teachers, for example, the exact time used for preparing lessons and correcting exercises for each subject and level of teaching. The high school collective agreements are thus more detailed and constrain the counties more than those for the hospital sector.
A third factor that may reduce diversity in public services is professional standards (Day and Klein 1987; Roberts and Dietrich 1999). Strong professional standards may result in less local political autonomy and therefore less diversity across regions. By all standards (cf., e.g., Roberts and Dietrich 1999) core hospital services are more professionalized than high school teaching. Whether measured in terms of length of education, degree of specialization, professional organization, informational advantages, or social standing, doctors and the activities they deliver are more professionalized than high school teaching. Other things being equal, the higher degree of professionalization in hospital services leads to less policy diversity in hospitals than in high schools.
Government structure provides an overall opportunity and incentive structure for elected officials in the county councils. But government structure interacts with other constraining institutions, and different institutions may have similar impacts. Due to strong central regulations laid down in collective agreements and central-government rules, we expect less policy diversity in high schools. But strong professional standards in the hospital sector may have a similar constraining impact on county autonomy.
Internal Decentralization and the Organization of Provider and Producer Roles
Returning to the first implication of the classic idea of centralization and decentralization, that more policy diversity is found in the sectors controlled by locally elected county councils than in regional labor-market councils, Kjellberg has cautioned that decentralization can take several forms (1985, 1988). Of particular relevance in this context is the fact that internal decentralization within the central government may be a source of policy diversity equally as strong as external decentralization from the central government to the local-government level.
In recent years, extensive delegation of responsibility has been effected in labor-market policy. In particular, services and job training for the long-term unemployed have been regionalized (Haahr and Winter 1996; Larsen et al. 1996). Since 1993, a number of reforms have tried to secure a flexible and active labor-market policy to meet the individual needs of (long-term) unemployed persons as well as the regional labor demand. Within central rules and guidelines, regional labor market councils (RLMCs) now set targets for the regional effort, and in general they choose the means to reach these targets. Because regional labor-market conditions and the composition of unemployment vary, differences in regional policy can be expected. Case studies of selected counties have shown that regional plans are indeed informed by "specific, regional conditions" (Haahr and Winter 1996, 77; Larsen et al. 1996; cf. also Directorate General for Employment Placement and Vocational Training 1999, appendix). Therefore, internal decentralization in regional labor-market policy may be a causal equivalent to external decentralization in high school and hospital services.
The regional labor-market councils may even be less constrained than the counties in accommodating policy to local needs and wants. Arguing from a public-choice perspective, the organization of the provider and producer roles is decisive for the effectiveness of policy making. The provider and producer roles can be merged or separated, and there can be more than one producer, If the provider is also producing a service, there is a greater risk that the provider will succumb to producer interests. In contrast the "presence of more than a single producer ... may enable citizens to make more effective choices about the mix of services they prefer.... Public officials may be able to bargain more effectively if alternative public producers are present in the area" (Ostrom 1972, 484). Thus, the separation of producer and provider roles may facilitate a more effective translation of different local preferences.
With a few inconsequential exceptions, hospital and high school services are provided and produced by local governments (Pallesen 1997, 1998). Not only core services (high school teaching and hospital treatment and nursing) but also auxiliary services (catering, cleaning, etc.) are provided in-house by local government employees. In contrast the RLMCs do not produce labor-market services for the unemployed. They buy them from a number of different producers, none of which are employed by the RLMCs. Regional employment offices under the Directorate General for Employment Placement and Vocational Training implement the directives and guidelines of the RLMCs. The RLMCs choose which activation measures to give priority (Haahr and Winter 1996, 58-76), and because different producers supply different services (e.g., education, guidance, practical job training), the councils have a strong influence on the choice of producers. Furthermore, there are typically more producers of the same kind of service.
If public choice is correct in arguing that the organization of provider and producer roles is decisive, we expect that the RLMC can make more efficient and free policy choices and thus that more policy diversity will result. The county councils are more constrained and more likely to pay attention to producer interests, that is, the interests of their own employees.
Summing up the theoretical discussion, decentralization of governing structure is only one of the elements determining local governance (Lynn, Heinrich, and Hill 2000; Milward and Provan 2000). Apart from government structure, a host of other institutions create incentives for the localities not to exploit their formal autonomy. Central guidelines and collective agreements may make it costly to pursue local needs and wants too vehemently. That may very well be the case in the high school sector. But the professional standards and interests of service producers also may make it difficult for county councils to pursue policies that challenge the professions. This may be the situation in the hospital sector. Finally, decentralization also takes place within a central-government structure, and the critical question may be the extent to which there are stronger or weaker incentives to give in to producer interests. The organization of provider and producer roles in regional labor-market policy makes it easier for policy makers to make choices informed by local needs and wants. Thus, it may very well be in this sector that the highest policy diversity is found.
MEASURING POLICY DIVERSITY IN LOCAL PUBLIC SERVICES
So far we have skirted the onerous issue of pinpointing what exactly we mean by diversity and policy variation. There are many dimensions to policy. Hence our selection of policy indicators must be carefully related to the theoretically informed propositions discussed earlier.
Criteria for the Choice of Policy Indicators
As we debate the literature on decentralization, we should give as much leeway to this proposition as possible. The likelihood that policy diversity is high, notably in high schools and hospital services, should be maximized, and the bias of indicators across sectors should be minimized. First and most important, this implies that the aspects of policy we analyze have to be directly attributable to county council and RLMC decisions. Therefore, we focus on policy outputs rather than policy outcomes. Numerous variables intervene and may have an impact on the translation of outputs to outcomes when the issue is complex public services (Day and Klein 1987).
A couple of examples relevant to the cases studied here may illustrate the point. The frequency of caesarean delivery differs substantially in the Danish counties. Although the frequency of caesarean delivery may be an important policy indicator, and indeed important to pregnant women, it is a policy aspect on which local-government decisions have little or no effect. Hence, it is not a reasonable policy indicator according to our criteria. Pupil performance on standardized test scores and long-term employability of formerly unemployed persons are other examples of classic outcome indicators of educational and labor-market policy. But these important indicators are only indirectly attributable to regional labormarket and local-government educational policy, although it is hoped that regional policies will have some impact on these policy outcomes.
Second, we focus on policy aspects, that is, indicators that elected officials and other actors care about. If we choose indicators that professionals but not county officials care about (e.g., high school curricula, the specific content of job-training programs, or division of work between nurses and consultants), it may not be so disquieting that common professional standards have a homogenizing impact on public services across regions. In order to balance the test, the best indicators are those that all relevant actors care and struggle about (i.e., the actors relevant to the institutions discussed earlier). However, because different actors do not feel equally intense about all aspects of policy (cf. Kendall and Carey 1968), more policy indicators should be used. The more policy indicators we find that point toward a certain institutional determinant as decisive, the more confidence we can have in the results.
Third, the indicators we analyze are likely to be sensitive to different local needs and wants. The organization of hospital management is probably not as sensitive to regional socioeconomic differences as the size and number of hospitals. Similarly, the issues of costs and productivity, which influence taxes, is likely to be divisive politically and perhaps across both rich and poor counties.
Policy Indicators in High Schools and Hospitals
We have selected three policy indicators according to the criteria discussed earlier: funding of services, dimensioning of services, and service productivity. The issue of controlling for different task conditions across regions is addressed in the following discussion.
One of the most important decisions the county councils make every year is the level of funding of their health-care and educational services. County-council members focus intensely on the funding and the distribution of the budget across sectors, not least because the public demand for additional and better services without raising taxes and professionals' demand for extra resources vastly exceeds the available resources. Together high schools and hospital services make up 60 percent of the county-council budgets in Denmark. Although health care and high schools are local-government responsibilities, national politicians and the national political parties also take a keen interest in the resource development in these policy areas (Christensen 2000). So far, however, each county has preserved the right to set its own spending (and taxing) priorities without central-government interference (Blom-Hansen and Pallesen 2001).
We measure level of funding in the high school sector as expenditures per enrolled student in each county. In the hospital sector, we have chosen to use expenditures per inhabitant in each county because hospitals also have to provide emergency units and other services not fully related to individual patient treatment.
County-council members are also heavily engaged in decisions about the dimensioning of hospitals and high schools. The main questions are how many high schools and hospitals the counties want to run and how big should they be. These questions are salient because they have ramifications for the location of hospitals and schools and the degree of service centralization within the county. Ceteris paribus, the smaller the average serviceproducing institution is, the more towns in the county can host an institution. The issue tends to ignite strong parochial interests among county-council members.
Professionals are also interested in dimensioning. In general professionals have a preference for stronger county centralization and fewer but bigger service-providing units because this allows them to specialize more and to pursue their professional interests.
Dimensioning of high schools is measured as the average number of classes per school and, in the hospital sector, the average number of beds per hospital. The central government has not issued binding regulations regarding the dimensioning of the hospital sector, but it recommends that the counties reduce the number of hospitals and specialize service provision further (Indenrigsministeriet 1997). The county councils can also decide the dimensioning and the number of service providers in the high school sector, but not quite as freely as in the hospital sector. The Ministry of Education has set up binding minimum standards for the number of teaching hours per student (Pallesen 1998). The required minimum hours of teaching varies with school size. In effect, to run three high schools with nine classes instead of one high school with twenty-seven classes increases county expenditure per student per year by approximately 5 percent.
Finally, from time to time elected officials in the county councils also pay close attention to service productivity. They may not focus as intensely on productivity as they do on overall costs and the location of service-providing institutions (Pallesen 1997), but because of a general demand overload and scarce resources there is a recurring focus on the value for money issue. A number of reforms during the 1980s and 1990s specifically targeted the productivity issue (Pallesen 1997), and recently the central government has tried to compel the counties to make part of budget allocations in the hospital sector activity based. Needless to say, the professionals and their unions also pay much attention to the demand for higher labor productivity, because this implies that they have to work harder and spend less time on activities they cherish (e.g., research and development, curriculum development, training, etc.).
The chosen indicator for service productivity in high schools is the number of lessons per students, which is another expression for the teacher-student ratio. The average number of lessons taught per student is strongly correlated to the teacher preparation time outside the classroom, and it is an essential parameter of the teachers' working conditions. The workload is quite densely regulated in the nationwide collective agreement between the teachers' union and the county-council association. As a general rule, each lesson of forty-five minutes is counted as two working hours (Pallesen 1998).
Therefore the county councils only have a few ways to influence service productivity. They can determine the number and size of the high schools, the number of students per class, and the menu of optional subjects at each school. The Ministry of Education mandates a number of obligatory subjects (math, Danish, English, etc.), whereas a number of subjects (Japanese, Italian, astronomy, etc.) are optional. Ceteribus paribus, a broader menu of optional subjects goes hand in hand with smaller average class sizes and increases the overall number of lessons taught per student.
The chosen service-productivity indicator in hospitals is the number of full-time equivalents per bed. Admittedly, staff per bed is far from being a perfect hospital productivity indicator. But the virtue of the staff per bed ratio is that the county councils take an interest in this figure and issue more-or-less binding overall capacity and overall staffing figures.
Patients and employees are also concerned about the level of service and hospital-labor productivity. Organizations representing patients who experience poor service and too little contact with professionals tend to perceive this as a staffing problem, not a problem of mismanagement. It is not a surprise that hospital professionals have nurtured this conception. The issue of hospital staffing is regulated neither by the central government nor in national collective agreements.
Policy Indicators in Regional Labor-Market Policy
In labor-market policy it is not possible to devise indicators strictly comparable to funding level, dimensioning, and service productivity as measured earlier. First, funding level was not politicized in labor-market policy in the middle and late 1990s (Haahr and Winter 1996). The social-democratic-led government had made combating unemployment a top priority, and resources to regional labor-market policy were abundant. Extra resources were allocated if the RLMCs exceeded the already boosted budgets (cf. Directorate General for Employment Placement and Vocational Training 1999, appendix). Following the criteria for selecting policy indicators, therefore, an analysis of general funding should not be included.
Second, the services in regional labor-market policy are produced by a number of public and semipublic institutions, and they include the option to subsidize public and private employment. In terms of the provider-producer split, the question about location and size of service institutions is not relevant in the case of regional labor-market policy. Instead the important aspect with respect to the dimensioning of policy is the choice among different instruments to activate the long-term unemployed. The RLMC can use various forms of education, guidance, private and public job training, and so on.
In spite of some central guidelines and targets regarding the effect of activation, the RLMCs are fairly free to choose among different instruments. The different organizations that sit on the RLMCs have partially overlapping and partially conflicting preferences for different activation instruments. Just like the location and size of high schools and hospitals, the choice of activation instruments determines the availability of services that the actors care about. For instance, the representatives of local governments in an RLMC may prefer as much public job training as possible because in this way they can externalize some of their labor costs. Representatives of private-sector unions and employer associations in the private sector usually prefer education activities. Job gaining in private businesses is a top priority among the unemployed because it is known to give the best chances for getting an ordinary job (Jepsen, Norgaard, and Vinderslev 2002).
The closest equivalent to dimensioning of high schools and hospitals is therefore the choice of activation instruments. We measure this dimensioning indicator as the regional-budget share of various activation instruments, and we have selected the instruments that have the highest budget shares at the national level. Regional budgets can be calculated in different ways. We measure the dominant instruments' budget share of total appropriations administered by the regions, that is, including the budgets of the employment offices. In this way we construct the most conservative measure of diversity. Results do not vary noticeably no matter which instrument is chosen as the indicator (cf. Jepsen, Norgaard, and Vinderslev 2002). Our indicators tend to underestimate rather than overestimate diversity across the RLMCs.
The RLMCs provide but do not produce activation services themselves and thus do not make staff decisions. As a consequence service productivity cannot be measured in terms of labor productivity; however, three different indicators of cost productivity can be calculated. The total activation pool spending in the regions can be related to (a) the number of full-time people in activation during a year, (b) the number of people in activation, and (c) the number of activation schemes produced. As a measure of value for money, the first indicator is probably the most reasonable and it is the measure for which the central authority holds the RLMCs accountable. We have chosen to present all three indicators to demonstrate that the results are robust no matter what productivity indicator is used.
For all these indicators, we calculate the cross-regional mean. The degree of diversity in service delivery is then measured as the standard deviation, and to ensure comparability across the different indicators, the standard deviation is divided by the mean. Although our measure of policy diversity is complex and multidimensional, we have chosen a simple way of calculating diversity to facilitate comparison within and across sectors. Our measure of diversity is based on population data; that is, it includes the total number of counties and RLMCs. Using population data reduces the need for statistical analysis of the significance of our findings, but in order to indicate the strength of the findings, we have analyzed whether the differences in diversity across sectors and indicators are significant (see appendix).
By comparing our diversity measure on comparable indicators across policy sectors we suggest that the best way to probe diversity is to make it a relative standard. Policy is rarely, if ever, totally uniform, and an absolute standard of much or little diversity seems arbitrary; rather, policy in one sector may be characterized by more or less diversity than in another sector. It would be ideal if all indicators in the same sector were the same size but different from the other sectors. In that case we can draw the most robust conclusions about the importance of different institutional determinants of local policy. However, different aspects of a policy may display different degrees of diversity because different causal mechanisms are at work.
Finally, the choice of method reflects the idea that policy diversity should be seen as an effect of institutional determinants whatever the root cause of the diversity. Hence we have argued that there are a number of reasons why regions should opt for different policies. Following decentralization and any other institutional theory, institutions are conditioning variables. We fully expect that there are a number of variables that will work through or be conditioned by the institutions and that these variables may explain differences in policy across counties. But we are not interested in explaining the differences in policy across regions. Our focus is differences in diversity across sectors and the scope of this diversity. Institutions may facilitate or impede the forces potentially fostering diversity. For these reasons, the differences we detect among sectors will not be explained away by other variables. The comparisons of policy diversity can be made "without the usual laundry list of statistical controls," to quote an article with a similar argument (Chubb and Moe 1988, 1071).
There is one lacuna in this argument: How can we tell whether differences in diversity are caused by differences in institutional context rather than differences in task conditions? Or, put another way, although the three sectors are comparable (regionally provided public services), they are not similar. It may be that the need for diversity is bigger in regional labor-market policy and that this need would manifest itself no matter what the institutional organization of the sector is. Arguments of this kind can rarely be totally refuted, but in order to minimize the risk that differences in task conditions are the cause of different degrees of diversity rather than institutional setting, we have controlled all our measures of diversity for the socioeconomic indicator that explains most of the variance in policy across regions (last row of tables 1 and 2). (1)
DOES DECENTRALIZATION VOUCH FOR POLICY DIVERSITY?
In the following section we discuss the differences and similarities in diversity in the two sectors under county-council authority, high schools and hospitals, and the institutions that may account for this pattern. Then we compare policy diversity in high schools and hospitals with the policy diversity in regional labor-market policy.
An isolated look at the high school sector shows the scope of diversity to be roughly the same across different indicators and when differences in county-socioeconomic differences are taken into account (see bottom line of table 1). Diversity is somewhat higher on the dimensioning indicator (.11), but in comparison with dimensioning in the hospital sector (.39), it is significantly lower (see appendix). Within the high school sector, the difference in diversity between the productivity (.04) and the dimensioning indicator (.11) is significant, but it is much smaller than the comparable difference in diversity in the hospital sector (.08 and .39). In the hospital sector, diversity in funding level (.06) and productivity (.08) is much smaller than in dimensioning (.39). On five of six relevant comparisons, the variance in the high school policy indicators is not significantly different from the variance in hospital funding and productivity. In contrast, hospital dimensioning stands out as the only indicator with substantial and significantly higher diversity than all other policy indicators in the high school and hospital sectors, also after controlling for the most relevant socioeconomic difference between the counties.
We can draw at least two conclusions. First, in general the differences in the degree of central regulation in the two sectors do not have an impact. In the less-regulated hospital sector, the diversity in funding level and productivity is not significantly higher than in the high school sector. This suggests that professional standards or interests may be an equally strong constraint on county autonomy as binding central regulations. Both may have a powerful homogenizing impact, and it may be difficult for the counties to disregard the demands of strong producer interests no matter their source.
Second, in both the high school and the hospital sectors, the largest diversity is found with respect to dimensioning, but diversity is much larger in the hospital sector. In general, professional arguments for a more uniform structure with fewer and bigger hospitals are disregarded or at least eluded by some county councils. The comparatively high degree of diversity underscores that the issue of location, which is reflected in the size indicator, is an essential local political issue. A reasonable interpretation is that elected county officials determine size and location of hospitals with a keen eye to the question of service proximity to their voters.
Parochialism is probably also at work in the high school sector. But it is easier to accommodate the parochial interest in high school policy than in hospital policy because there are more high schools than hospitals in all counties. The diversity in dimensioning is thus much smaller than in the hospital sector.
Compared to both services provided by the county councils, regional labor-market policy provided by the RLMCs is characterized by much more diversity across regions.
Diversity in RLMC policy is relatively large. For all dimensioning measures the set variation is above .25. The RLMCs use highly different activation instruments, although in all regions, education and, in particular, wage subsidies for public employment constitute large budget shares.
Variation in the dimensioning of labor-market policy is one of the factors leading to differences in service productivity, because the cost of using different instruments varies. No matter which indicator is used for measuring cost productivity, there is more regional diversity in labor market policy productivity than in either of the services provided by the counties. The diversity is smallest on the indicator that we argued was the best indicator of value for money (.13).
Still, the level of diversity in RLMC policy is more correctly assessed in comparison to the high school and hospital sectors, keeping the policy dimension constant. On the dimensioning indicators in labor-market policy, diversity is at the same high level and not significantly different from the diversity in the hospital sector. But on all indicators, labor-market diversity is significantly higher than high school dimensioning.
All indicators of service productivity show that diversity in labor-market policy is higher than in the services provided by the county councils. This result only gains strength if the diversity in labor market is controlled for differences in the RLMC essential-task conditions, that is, the general level of unemployment and the share of long-term unemployment (more than two years' unemployment; cf. bottom line of table 2). This suggests that some of the labor-market policy diversity has been suppressed by the unweighed measures of diversity.
First, the results suggest that a decentralized government structure is neither a necessary nor a sufficient cause of policy diversity. Due to other institutional constraints, policy diversity is generally low in the high school sector. Furthermore internal decentralization may provide an equally (or more) hospitable institutional context for regional policy autonomy as decentralization to local governments. The dimensioning of the hospital sector and labor-market policy exhibits comparably high levels of policy diversity.
Second, the public choice argument that a split between provider and producer roles facilitates local decision makers' implementation of different policies and accommodation of local demands is partially confirmed. In three out of four matched comparisons (productivity in both high schools and hospitals and high school dimensioning), diversity levels are higher in regional labor-market policy. However, the results may also be interpreted in another way. A decentralized-government structure in which the producer and provider roles are fused may produce the same level of diversity or more than one in which provider and producer roles have been separated if local political attention is high and citizen preferences for a certain policy are intense and concentrated. The large diversity in the dimensioning of hospitals is an example of this.
In theory governing structure has an impact on policy, but it is not the only institution that matters. Even the most fundamental attribute of governing structure, the fact that decentralization increases the prospects of local adaptation and thus policy diversity, cannot be taken for granted. Other institutions may strengthen, inhibit, or even annul the plausible effects of decentralization.
The results in this study contest the idea of central government control being tantamount to conformity and uniformity and decentralization to local authorities being conducive to diversity and local adjustment.
First, it has been demonstrated that other institutions may impose so many constraints on local autonomy that local governments are not free to choose local policy even on those dimensions of policy that they most plausibly control. Centrally devised institutional constraints on local autonomy can still be so severe that the impact of government structure is too weak to have an effect on policy. These constraining institutions take many forms, as the cases of high schools and hospital services demonstrate: Central government regulation, nationwide collective agreements, and strong professional interests may all have a powerful homogenizing impact. That internal decentralization may be as effective as or even more effective than decentralization to another government level may simply be due to the fact that the externally imposed institutional constraints are more lenient in the former case.
Second, the sources of preferences for diversity and thus the central concerns of various decision makers have to be taken into account. When locally elected officials face strongly organized public employees with intense producer interests that to some extent are reflected in central regulation, collective agreements, and professional standards, they have to weigh the preferences and demands of these constituents against the broader but usually more diffuse interests of the electorate. For elected officials to challenge intense producer interests there must be plausible electoral gains to harvest. In most circumstances it is unlikely that voters feel strongly about the detailed organization of public services.
If an aspect of policy is clear to voters, the likelihood increases that elected officials will dare producer interests. The dimensioning indicator taps into the issue of size and thus localization and proximity of service institutions to users-cum-voters. In high schools and especially in hospitals, the level of diversity is higher on this dimension, even though the strong medical professions have an intense and outspoken interest in a more uniform hospital structure. Detailed central regulation of high schools does not preclude diversity in size, but the county councils face similar economic and professional incentives not to run very small high schools.
In the hospital sector the consequences are more uncertain, and there are more ways to organize the provision of services. If it is decided to run small hospitals in sparsely populated areas, perhaps out of parochial concerns, the small hospitals may not include all specialties or around-the-clock services. Sometimes small hospitals are highly specialized organizations offering few elective services. The scope of autonomy in deciding the organization and the mix of services is higher in the hospital sector. When there are more alternatives, diversity increases.
There is a third reason why internal decentralization to a regional, corporatist body under the auspices of central government does not automatically guarantee more policy diversity and adjustment to local priorities: It is plausible that it is not this attribute that mateters. In contrast to the high school and hospital sectors, the provider role has been separated from the producer role only in regional labor-market policy. Moreover, RLMCs have several service producers to choose among. It has been theoretically shown that exactly these attributes are decisive for "effective governing" at the local level (Milward and Provan 2000). If this aspect of policy organization is what really matters, there is hope also for the traditional decentralized government structure. A recent Danish study that compares regional labor-market policy by the RLMC with labor-market policy by the municipalities--the government layer below the county level--supports this proposition (Jepsen, Norgaard, and Vinderslev 2002). In both cases the provider role has been separated from the producer role, and in both cases policy diversity is much higher than in the county cases studied here.
If local governments provide but not produce, they may also be able to escape the crosscutting incentives due to different roles. For this reason they will be less constrained in making choices offending intense and highly outspoken interests.
Appendix Pairwise Analysis of the Policy Indicators Variance (Levene's Test Equality of Variances) High Scholl Policy Indicators Funding Service Level Productivity (Expenditure Dimensioning (Lessons per per Student) (School Size) Pupil) High school policy indicators Funding level (expenditure 0 0 0 per student) Dimensioning (school size) + 0 Service productivity (lessons + per pupil) Expenditure per inhabitant Hospital size Staff per bed Information and guidance Education Wage subsidy, private employment Wage subsidy, public employment Expenditure per number of full-time persons Number of persons Hospital Policy Indicators Expenditure Staff per Hospital per Inhabitant Size Bed High school policy indicators Funding level (expenditure + 0 + per student) Dimensioning (school size) + 0 + Service productivity (lessons + 0 + per pupil) Expenditure per inhabitant + 0 + Hospital size 0 0 Staff per bed + Information and guidance Education Wage subsidy, private employment Wage subsidy, public employment Expenditure per number of full-time persons Number of persons Labor-Market-Policy Indicators (Dimensioning, Share of Budget) Wage Wage Subsidy, Subsidy, Information Private Public and Guidance Education Employment Employment High school policy indicators Funding level + + + 0 (expenditure per student) Dimensioning + + + 0 (school size) Service productivity + + + 0 (lessons per pupil) Expenditure per + + + 0 inhabitant Hospital size 0 0 0 0 Staff per bed + + + + Information and 0 + + + guidance Education 0 0 + Wage subsidy, private 0 + employment Wage subsidy, public + employment Expenditure per number of full-time persons Number of persons Service Productivity Expenditure per Number Expenditure Expenditure of Full-Time per Number per Number Persons of Persons of Schemes High school policy indicators Funding level (expenditure 0 0 per student) Dimensioning (school size) 0 0 Service productivity (lessons + + per pupil) Expenditure per inhabitant 0 0 Hospital size + + Staff per bed + + Information and guidance + + Education + + Wage subsidy, private + + employment Wage subsidy, public + + employment Expenditure per number of 0 0 full-time persons Number of persons 0 Sources: Arbejdsmarkedsstyrelsen 1999, Virksomhedsregnskab 1998, Bilag bd. 1 og 2; Bestandsstatistik over AF's indsats i gennemforelsen af lov om en aktin arbejdsmarkedspolitick og lov og orlov: Aktivering (01.01.1998-31.12.1998), korselsdato 22. januar 1999; Kommunalstatistisk Database (KSDB); Undervisningsministeriet,Nogletal. Gymnasier og 2-arige HF-kurser, 1995-96; Amtsradsforeningen 1998, Amternes okonomi: Budget '98; Sundhedsstyrelsen 1997, Virksomheden ved sygehusene 1995; Table 1 Indicators of High School and Hospital Policy, 1996 High-Schools Funding Level Dimensioning Service (Expenditure (School Size Productivity per Student in Number (Lesson per County in DKK) of Classes) Pupil in Hours) Copenhagen 56,400 20.1 1.63 Frederiksborg 52,100 22.6 1.51 Roskilde 45,100 25.8 1.59 Vestsjaelland 55,900 25.8 1.63 Storstrom 59,800 20.0 1.59 Fyn 54,900 22.3 1.68 Sonderjylland 53,900 25.8 1.47 Ribe 48,100 20.7 1.51 Vejle 51,100 22.8 1.52 Ringkobing 49,600 19.7 1.55 Arhus 50,500 23.8 1.61 Viborg 52,900 22.8 1.64 Nordjylland 50,800 23.5 1.58 Average 52,392 22.7 1.58 SD 3,865 2.2 0.06 SD/average 0.07 0.09 0.04 SD/average 0.05 0.11 0.04 (adjusted for (general needs) (general needs) (general needs) socio-economic differences) Hospital Funding Level Dimensioning Service (Expenditure (Hospital size Productivity per Ihabitant in Beds) (FTE per Bed) County in DKK) Copenhagen 5500 631 3.1 Frederiksborg 4720 430 2.9 Roskilde 5100 314 3.3 Vestsjaelland 5500 187 3.1 Storstrom 5400 210 3.2 Fyn 5800 206 3.3 Sonderjylland 5100 178 2.8 Ribe 5100 157 3.1 Vejle 4800 228 2.7 Ringkobing 4500 185 3.1 Arhus 4800 229 3.5 Viborg 5500 165 3.2 Nordjylland 5400 257 3.2 Average 5171 260 3.1 SD 384 128 0.22 SD/average 0.07 0.49 0.07 SD/average 0.06 0.39 0.08 (adjusted for (health-care-needs) (population density) (health-care socio-economic needs) differences) Sources: Kommunalstatistisk Database (KSDB); Undervisningsministeriet, Nogletal. Gymnasier og 2-arige HF-kurser, 1995-96; Amternes okonomi: Budget '98 Sundhedsstyrelsen 1997, Virksomheden ved sygehusene 1995; Indenrigsministeriet, Kommuale Nogletal 1997. Note: FTE = full-time equivalents. Table 2 Indicators of High School and Hospital Policy, 1996 Dimensioning (a) Wage Wage Information Subdidy Subdidy Region and Guidance Education (Private (Private Employee) Employee) Copenhagen 3.3 13.8 4.7 24.9 Frederiksborg 2.8 11.6 3.5 20.2 Roskilde 1.5 23.7 3.7 17.5 Vestsjaelland 1.6 8.9 2.5 26.0 Storstrom 0.5 24.4 2.4 16.6 Fyn 2.6 8.5 1.9 17.3 Sonderjylland 2.1 12.1 5.7 24.5 Ribe 2.6 11.6 3.7 25.4 Vejle 3.9 8.6 5.0 16.4 Ringkobing 4.4 11.3 5.1 15.5 Arhus 5.0 12.2 4.5 11.8 Viborg 6.6 6.4 3.5 19.0 Nordjylland 1.5 15.8 3.8 24.3 Average 2.8 12.6 3.8 21.0 SD 1.4 5.3 1.1 5.6 SD/average 0.61 0.42 0.29 0.27 SD/average 0.58 0.56 0.28 0.30 (adjusted for (general (general (general (+2 years' socio-economic unemployment) unemployment) unemployment) share of differences) unemployment) Service Productivity No. of Full-Time No. of No. of County Persons Persons Schemes Copenhagen 124.0 51.4 33.4 Frederiksborg 107.8 50.9 30.7 Roskilde 111.3 48.3 29.2 Vestsjaelland 121.0 55.1 33.6 Storstrom 125.6 59.7 38.1 Fyn 97.8 46.0 29.7 Sonderjylland 114.6 48.2 30.5 Ribe 124.8 62.0 33.8 Vejle 90.3 42.4 25.3 Ringkobing 98.1 42.7 28.1 Arhus 112.5 51.7 32.7 Viborg 103.2 39.7 26.2 Nordjylland 119.2 54.0 33.8 Average 112.2 51.2 32.0 SD 10.9 7.2 4.4 SD/average 0.10 0.14 0.14 SD/average 0.13 0.19 0.30 (adjusted for socio- (+2 years' (+2 years' (+2 years' economic share of share of share of differences) unemployment) unemployment) unemployment) Source: Arbehdsmarkedsstyrelsen 1999, Virksomhedsregnkab 1998, Bilag bd. 1 og 2; Benstadsstatistik over AF's indsats i gennemforelsen af lov om en aktiv arbejdsmakedspolitik og lov og orlov: Aktivering (01.01.1998-31.12.1998), korselsdato 22.januar 1999. (a) Most important instruments share of budget (percent). (b) Expenditure per indicator (DKK 1,000).
(1) The most important socioeconomic variable conditioning high school, health-care, and labor-market policy has been identified by a number of regression analyses where each of the areas' policy indicators have been regressed against a set of independent variables, measuring general and area-specific needs variables (e.g., general social needs, population density, specific health-care needs, level of [long-term] unemployment).
Association of County Councils (Amtsradsforeningen). 1997. Opgavekommissionen. Amtsradsforeningens oploeg (The taks commission. The proposals of the Association of County Councils). Copenhagen: Amtsradsforeningen.
Association of County Councils (Amtsradsforeningen). 1998. Amternes okonomi--Budget '98 (The economy of the counties: The 1998 budget). Copenhagen: Amtsradsforeningen.
Blom-Hansen, J. 1999. Avoiding the "joint-decision trap." Lessons from intergovernmental relations in Scandinavia. European Journal of Political Research 35:35-67.
Blom-Hansen, J., and T. Pallesen. 2001. The fiscal manipulation of a decentralized public sector: macroeconomic policy in Denmark. Environment and Planning C: Government and Policy 19:607-23.
Bogason, P. 1996. The fragmentation of local government in Scandinavia. European Journal of Political Research 30:65-86.
Brans, M. 1992. Theories of local government reorganization: An empirical evaluation. Public Administration 70:429-52.
Christensen, J. G. 2000. The dynamics of decentralization and recentralization. Public Administration 78, no. 2:389-408.
Chubb, J. E., and T. M. Moe. 1988. Politics, markets, and the organization of schools. American Political Science Review 82, no. 4:1065-87.
Dahl, R. 1970. After the revolution. New Haven, Conn.: Yale University Press.
Day, P., and R. Klein. 1987. Accountabilities: Five public services. London: Tavistock Publications.
De Vries, M. 2000. The rise and fall of decentralization: A comparative analysis of arguments and practices in European countries. European Journal of Political Research 38:193-224.
Directorate General for Employment Placement and Vocational Training (Arbejdsmarkedsstyrelsen). 1999. Arbejdsformidlingen: Virksomhedsregnskab 1998 (The labor exchange office: Concern accounts, 1998).Copenhagen: Arbejdsmarkedsstyrelsen (herunder Bilag, bind 1 og 2).
Haahr, J. H., and S. Winter. 1996. Den regionale arbejdsmarkedspolitik. Planloegning mellem centralisering og decentralisering (Regional labor market policy: Planning between centralization and decentralization). Aarhus: Systime.
Hamilton, A., J. Madison, and J. Jay. 1961. The federalist papers. New York: Mentor, Penguin.
Indenrigsministeriet. 1997. Kommunale nogletal (Local government key figures). Copenhagen: Indenrigsministeriet.
Jepsen, M. B., A. S. Norgaard, and J. D. Vinderslev. 2002. Forskrifter, forhindringer og farlige fristelser: Aktivering af ledige i stat og kommuner (Prescriptions, constraints, and dangerous temptations: State and local government activation of unemployed). In Kommunale patologier, ed. J. BlomHansen, F. Bruun, and T. Pallesen. Aarhus: Systime.
Kendall, W., and G. W. Carey. 1968. The "intensity" problem and democratic theory. The American Political Science Review 62 (March):5-24.
Kjellberg, F. 1985. Local government reorganization and the development of the welfare state. Journal of Public Policy 5:215-39.
1988. Local government and the welfare state: Reorganization in Scandinavia. In The dynamics of institutional change: Local government reorganization in Western democracies, ed. B. Dente and F. Kjellberg. London: Sage.
Larsen, F., C. Hansen, H. Jorgensen, and M. Lassen. 1996. Implementering af Regional Arbejdsmarkedspolitik (Implementation of regional labor market policy). Aalborg: Carma.
Lynn, L. E., C. J. Heinrich, and C. J. Hill. 2000. Studying governance and public management: Challenges and prospects. Journal of Public Administration Research and Theory 10, no. 2:233-61.
Milward, H. B., and K. G. Provan. 2000. Governing the hollow state. Journal of Public Administration Research and Theory 10, no. 2:359-79.
Ministry of the Interior (Indenrigsministeriet). 1998. Fordelingen af opgaver i den offentlige sektor (The allocation of tasks in the public sector). Betaenkning nr. 1366. Copenhagen: Indenrigsministeriet.
Norgaard, A. S. 1996. Rediscovering reasonable rationality in institutional analysis. European Journal of Political Research 29:31-57.
1997. The politics of institutional control: Corporatism in Danish occupational safety and health regulation & unemployment insurance, 1870-1995. Aarhus: Politica.
1999. Arbejdsmarkedspolitikken: Korporatisme til alle tider og alle sider (Labor market policy: Corporatism at all times and in all directions). In Magtens organisering: Stat og interesseorganisationer i Danmark, ed. J. Blom-Hansen and C. Daugbjerg. Arhus: Systime.
Ostrom, E. 1972. Metropolitan reform: Propositions derived from two traditions. Social Science Quarterly 20:474-93.
Pallesen, T. 1997. Health care reforms in Britain and Denmark: The politics of economic success and failure. Arhus: Politica.
1998. De danske amter--politikere eller administratorer (The Danish county councils--politicians or administrators: The case of the high school sector). Gymnasieskolen som eksempel. In Offentlig og effektivt? ed. J. Blom-Hansen J. Christensen, D. Gronnegaard, J. Bejer, P. Nannestad, T. Pallesen, and L. D. Pedersen.. Copenhagen: Gyldendal.
1999. Dansk Sundhedspolitik--Hvad udad tabes ma indad vindes (Danish health policy--what is lost externally must be gained internally). In Magtens organisering: Stat og interesseorganisationer i Danmark, ed J. Blom-Hansen, and C. Daugbjerg. Arhus: Systime.
Roberts, J., and M. Dietrich. 1999. Conceptualizing professionalism: Why economics needs sociology. American Journal of Economics and Sociology 58, no. 4:977-98.
Sharpe, L. J. 1970. Theories and values of local government. Political Studies 18:153-74.
1979. Decentralist trends in Western democracies: A first appraisal. In Decentralist trends in Western Europe, ed. L. J. Sharpe. London: Sage.
Stroby Madsen, C. 1998. Arbejdsmarked og europoeisk integration (Labor market and European integration). Copenhagen: FAOS.
Sundhedsstyrelsen. 1997. Virksomheden ved sygehusene 1995 (The activities at the hospitals, 1995). Copenhagen: Sundhedsstyrelsen.
Tiebout, C. M. 1956. A pure theory of local expenditure. Journal of Political Economy 64:416-24.
Undervisningsministeriet. 1996. Nogletal. Gymnasier og 2-arige HF-kurser (Key figures. High schools and 2-year higher preparation courses). Copenhagen: Undervisningsministeriet.
Weaver, R. K., and B. A. Rockman, eds. 1989. Do institutions matter? Government capabilities in the United States and abroad. Washington, D.C.: Brookings Institution.
Asbjorn Sonne Norgaard
University of Aarthus
|Printer friendly Cite/link Email Feedback|
|Author:||Norgaard, Asbjorn Sonne; Pallesen, Thomas|
|Publication:||Journal of Public Administration Research and Theory|
|Date:||Oct 1, 2003|
|Previous Article:||Policy diffusion through institutional legitimation: state lotteries.|
|Next Article:||Are Federal Trust Fund Locked Boxes a Lock?|