2. Finland and Sweden – Different approaches to EU membership

In this section, we start by describing the background of institutional changes in Sweden and Finland after their accession into the EU in 1995. Thereafter, we give an overview of the outcome regarding the institutions working with environmental policy and its implementation.

2.2. Institutional set-up for environmental policy

Between 1987 and 2022, the Ministry of the Environment had the main responsibility for environmental policy in Sweden. The main environmental agencies are the Swedish Environmental Protection Agency (SEPA) and the Agency for Marine and Water Management. The agencies are, relatively speaking, independent of the government. However, there are close informal and formal contacts between the ministries and the agencies, and many civil servants can, over the course of their careers, move back and forth between them (Molander et al., Reference Molander, Nilsson and Schick2002; Hansson, Reference Hansson2020). In addition to the agencies, there are expert authorities. For example, the Swedish Meteorological and Hydrological Institute (SMHI) serves an important role in providing environmental data on air and water quality.

An environmental objective system (EOS), established in 1999, guides policy. The decision to adopt the EOS was made by the Parliament, but it is not a law that can be challenged in court (Swedish Environmental Protection Agency, 2023). Around the same time, an environmental code was also adopted to harmonize with EU legislation (Government Environmental Code Commission, 1996; Duit, Reference Duit2007). According to Swedish Government Bill 1997/98:145, the EOS was a way to reach the government’s vision of a sustainable Sweden. It was also, together with the environmental code, a way to decentralize the work on improving the environment, for example, by allowing the business sector to contribute with their own initiatives. The level of ambition is high: a government report to the parliament (Regeringens skrivelse 1994/95:167, 1995) states that the highest ambition level among EU countries will be the benchmark in Sweden’s work within the EU, and a lowering of standards should not be applied in Sweden. It is also clearly stated that Sweden wants to be a driving force in the development of environmental policy in the EU based on the view that a proactive environmental policy can benefit economic growth.

Currently, the responsibility for achieving the 16 EOS is delegated to 26 government agencies, although at the national level, the SEPA has a coordinating role. Among the agencies that are in some way responsible for the EOS are the Energy Agency, the Board of Agriculture, the Forest Agency, the Agency for Marine and Water Management, the National Heritage Board, the Agency for Economic and Regional Growth, the Transport Agency and the Transport Administration (Sveriges miljömål, 2022). Eight of the agencies are responsible for coordinating the monitoring of specific environmental quality objectives. This delegation has been criticized by the OECD (2007), which states that there are conflicts between environmental goals and other societal goals that are not managed at the central level, and that these conflicts are instead left to subnational levels in the government system to solve. The subnational government entities often do not have either the requisite knowledge or the tools to do this work.

There are 21 counties at the regional level. They are governed both by county administrative boards (a central government agency) and by directly elected so-called regions with their own administrations.Footnote ⁷ The administrations at this level are assigned responsibilities in the EOS.Footnote ⁸ To some extent, this work is guided by the government agencies at the national level. The Energy Agency, for example, provides funding and support to the County Administrative Boards for strategic work with climate and energy, and to the Energy Offices, which often are part of the regions. SEPA, on the other hand, supports the County Administrative Boards in their work with the environmental objectives. Environmental data are also collected by RUS (the regional development and collaboration in the work with Sweden’s environmental goals).Footnote ⁹

Finally, the Swedish Agency for Economic and Regional Growth has the responsibility for developing practices for and supporting government agencies in their work with RIA. The Better Regulation Council, established in 2008 (Dir, 2008, 57), is located in conjunction with this agency. The Council’s task is to examine the impact of new and changed regulations on the regulatory pressure on firms and enterprises. Hence, the Council does not examine whether an RIA has been performed or its quality, but only whether the consequences for businesses have been sufficiently studied ahead of new or changed regulations. The parts of the Swedish administrative system relevant for this paper are depicted in a simplified form in Figure 1.Footnote ¹⁰

Figure 1. An illustration of the main administrations working with environmentally sustainable development in Sweden in 2023. In dark blue are the ministries at the national level, medium blue the government agencies at the national level and light blue at the regional level. Regions are separate, directly elected administrative bodies and therefore marked in dark red. The EOS, in grey, is the responsibility of the Ministry of the Climate and Enterprise. Domestic strategies, depicted in green, are developed by the Regions with support from the County Administrative Boards, the arrow indicating that this work is also related to the EOS. Otherwise, the solid arrows indicate delegation and the dashed lines indirect activities such as support or funding.

In Finland, the Ministry of the Environment is in charge of developing and implementing environmental legislation. The ministry consists of several departments dealing with different kinds of environmental issues: natural environment, built environment and environmental protection (Ministry of the Environment, 2022a). If delegation is necessary, the Ministry can delegate to the 15 Centers for Economic Development, Transport and the Environment (the ELY Centers) (Ministry of the Environment, 2022b). These centers were established in 2010 together with six regional state administrative agencies, each of which covers several regions.Footnote ¹¹ These replaced the 18 county administrative boards that were dismantled in 2010. The ELY centers, together with the Ministry and a research institute, the Finnish Environment Institute (SYKE), provide information on the state of the environment through one source: Environment.fi.

The ELY Centers belong to the administrative branch of the Ministry of Economic Affairs and Employment but are also steered by other ministries since they deal with cross-sectoral issues. In addition to environmental issues, they also play a role in the promotion of regional business policy and transport-related issues (Centre for Economic Development, Transport and the Environment, 2022). They perform environmental impact assessments (Darpö, Reference Darpö, Squintani, Darpö, Lavrysen and Stoll2019). In matters related to environmental permits, the Ministry of the Environment directs the Regional State Administrative Agencies, four of which issue environmental permits (Ministry of the Environment, 2022b). Even SYKE has some administrative tasks, above all concerning issues falling under the Ministry of Agriculture and Forestry, pertaining to the use and quality of water (Ministry of the Environment, 2009). The Forest Agency is responsible for nature conservation (Ministry of the Environment, 2022b). The setup of the Finnish administrative system relevant for this paper is depicted in a simplified form in Figure 2.

Figure 2. An illustration of the main administrations involved in environmental policy in Finland. In dark blue are the ministries, in medium blue the Finnish Environment Institute and an agency at the national level, and in light blue the regional level. The regional councils are separate elected bodies and therefore marked yellow. Data and information about the environment are coordinated at the national level. The strategy for sustainable development is determined by the government. The solid arrows indicate delegation, and the dashed lines indicate indirect activities.

Table 1 compares the Finnish and Swedish public administrations with regard to the delegation of responsibility for a number of environment-, transport- and regional development-related tasks. As is clear from the table, the Swedish public administration is much more fragmented than the Finnish one. We will return to the consequences of these differences in Section 4.

Table 1. A comparison of agency competences in Sweden and Finland

Sources: Centre for Economic Development, Transport and the Environment (2022), Ministry of Environment (2009, 2022b), https://www.regeringen.se/, https://www.sverigesmiljomal.se/.

3. A model of the economy

Given the differences in the setup of national administrations, we will develop a model to analyze the impact of implementation of EU legislation on economic efficiency. The EU’s preparation process for directives usually includes an RIA. Moreover, a directive is designed to allow for the differences that exist between MSs.

Assume an economy consisting of EU and its two MSs, $ i\in \left\{F,S\right\} $ , both of which consist of two regions, $ r\in \left\{1,2\right\} $ . We normalize the geographical area of each region to one and denote the population of each region by $ {n}_{ir} $ , where we assume that $ {n}_{i1}\gg {n}_{i2} $ . Each MS is assumed to be a small open economy, thus taking the price of the traded good, $ y $ , as given: $ {p}_y $ . Moreover, we assume perfect competition in all sectors.

There are two consumption goods in the economy: a numeraire, $ x $ , and the traded good $ y $ . While good $ x $ is produced using labor only, good $ y $ is produced using two production factors, labor, $ l $ , and emissions, $ \epsilon $ . The production set $ Y $ is convex. In each region, there are $ {m}_{ir} $ identical firms that produce good $ y $ . We assume that the number of firms in region 1 greatly exceeds that in region 2: $ {m}_{i1}\gg {m}_{i2} $ .

Labor and emissions are imperfect substitutes to one another.Footnote ¹² We assume wages, $ w $ , to be constant. The shadow price on emissions regulated at a regional level is denoted $ {\lambda}_{ir} $ , and the restriction creating it is $ {E}_{ir}\ge \left({m}_{ir}+\mathrm{E}\left({d}_{ir}\right)\right){\epsilon}_{ir} $ , while the shadow price on emissions regulated at national level is $ {\lambda}_i $ , the restriction being $ {E}_i\ge \left({m}_{i1}+\mathrm{E}\left({d}_{i1}\right)\right){\epsilon}_{i1}+\left({m}_{i2}+\mathrm{E}\left({d}_{i2}\right)\right){\epsilon}_{i2} $ . $ \mathrm{E} $ is a conditional expectations operator and $ {d}_{ir} $ is a uniformly distributed shock in range $ \left[-D,D\right] $ with a mean of zero. The value taken by $ \mathrm{E}\left({d}_{ir}\right) $ is conditional on whether an RIA has been carried out or not. $ {d}_{ir} $ captures uncertainty about optimal emission levels.Footnote ¹³ $ {E}_{i1}+{E}_{i2}={E}_i $ are the emissions allowed for country $ i $ by the EU directive. It is up to the country to allocate the emissions between its two regions. This gives us the profit-maximizing condition for firms producing $ y $ in each of the two countries $ i $ and regions $ r $ :

$$ \max {p}_y{m}_{ir}y\left({l}_{ir},{\epsilon}_{ir}\right)+w{m}_{ir}{l}_{ir}+{\lambda}_{ir}\left[{E}_{ir}-\left({m}_{ir}+\mathrm{E}\left({d}_{ir}\right)\right){\epsilon}_{ir}\right], $$

(1) $$ \mathrm{where}\hskip1em {E}_i\ge {E}_{i1}+{E}_{i2} $$

(2) $$ \max {p}_y{m}_{ir}y\left({l}_{ir},{\epsilon}_{ir}\right)+w{m}_{ir}{l}_{ir}+{\lambda}_i\left[{E}_i-\left({m}_{i1}+\mathrm{E}\left({d}_{i1}\right)\right){\epsilon}_{i1}-\left({m}_{i2}+\mathrm{E}\left({d}_{i2}\right)\right){\epsilon}_{i2}\right] $$

Equation (1) is applicable if the emissions are differentiated at the local level and equation (2) if there is a uniform regulation for the entire country. With no environmental policy in place, $ {\lambda}_{ir}={\lambda}_i=0 $ . Differentiating equations (1) and (2) with respect to emissions, we can solve for the expected impact of local or national policy, respectively:

(3) $$ \mathrm{E}\left({\lambda}_{ir}\right)={p}_y\frac{m_{ir}}{m_{ir}+\mathrm{E}\left({d}_{ir}\right)}\frac{\partial {y}_{ir}}{\partial {\epsilon}_{ir}}, $$

(4) $$ \mathrm{E}\left({\lambda}_i\right)={p}_y\frac{m_{i1}}{m_{i1}+\mathrm{E}\left({d}_{i1}\right)}\frac{\partial {y}_{i1}}{\partial {\epsilon}_{i1}}={p}_y\frac{m_{i2}}{m_{i2}+\mathrm{E}\left({d}_{i2}\right)}\frac{\partial {y}_{i2}}{\partial {\epsilon}_{i2}}. $$

If the government does an impact assessment and finds the value of $ {d}_{ir} $ , it can set policy with certainty about the shadow cost (i.e., it can set policy so that $ {d}_{ir}=0 $ ), otherwise the policy will differ from optimal. Knowing the expected cost of environmental policy, we can solve for labor demand and emissions in sector $ y $ as a function of the shadow price of emissions: $ {l}_{ir}\left(\mathrm{E}\left({\lambda}_{ir}\right)\right) $ and $ {\epsilon}_{ir}\left(\mathrm{E}\left({\lambda}_{ir}\right)\right) $ , respectively. We assume that both labor demand and emissions fall in the price of emissions.

Since production is assumed to take place in each of the four regions, emissions, too, arise from all regions. We denote damages from pollution by $ \delta (e) $ . Damages are assumed to be exponentially increasing in pollution: $ {\delta}^{\prime }(e)>0,{\delta}^{\prime \prime }(e)\ge 0 $ . They are either local (only affect the region where emissions take place) or global (i.e., affect the entire EU, $ i\in \left\{F,S\right\} $ ). This means that there may be spillovers of emissions from one region to another. We denote spillovers $ \kappa \in \left[0,1/4\right] $ , with $ \kappa =0 $ denoting no spillovers (a local pollutant) and $ \kappa =1/4 $ a uniformly mixing (global) pollutant (Besley and Coate Reference Besley and Coate2003).Footnote ¹⁴ We assume $ \kappa $ to be constant for all four regions for each respective pollutant.

Normalizing the price of the numeraire good, $ x $ , to one, the representative citizen in each region has continuous, quasi-linear preferences:

$$ {U}_{ir}={x}_{ir}+{p}_y{y}_{ir}^c-\delta \left[\left(1-3\kappa \right)\left({m}_{ir}+{d}_{ir}\right){\epsilon}_{ir}+\kappa \left({m}_{i,-r}+{d}_{i,-r}\right){\epsilon}_{i,-r}+\kappa \sum \limits_{-i,r=1}^2\left({m}_{-i,r}+{d}_{-i,r}\right){\epsilon}_{-i,r}\right]. $$

$ {y}_{ir}^c $ denotes the consumption of good $ y $ in region $ r $ of country $ i $ . Utility then equals utility from consumption minus the disutility from own emissions, adjusted for the emissions “exported” to the other regions, and minus the disutility from “imported” emissions. In order to keep the notation simple, we denote the downfall of emissions in each region (i.e., pollution) by $ {e}_{ir}=\left(1-3\kappa \right)\left({m}_{ir}+{d}_{ir}\right){\epsilon}_{ir}+\kappa \left({m}_{i,-r}+{d}_{i,-r}\right){\epsilon}_{i,-r}+\kappa {\sum \limits}_{-i,r=1}^2\left({m}_{-i,r}+{d}_{-i,r}\right){\epsilon}_{-i,r} $ .

Citizens get labor income from working in one of the two sectors. Labor demand in sector $ y $ in each region is given by $ {l}_{ir}\left[\mathrm{E}\left({\lambda}_{ir}\right)\right] $ , and labor demand in sector $ x $ is assumed to be a constant: $ {l}_{ir}^x $ . A stricter environmental policy lowers $ {l}_{ir}\left[\mathrm{E}\left({\lambda}_{ir}\right)\right] $ . Income in region $ r $ is then given by $ w\left[{l}_{ir}^x+{l}_{ir}\left[\mathrm{E}\left({\lambda}_{ir}\right)\right]\right] $ , and we can write the representative citizen’s budget constraint as $ w\left[{l}_{ir}^x+{l}_{ir}\left[\mathrm{E}\left({\lambda}_{ir}\right)\right]\right]\ge {x}_{ir}+{p}_y{y}_{ir}^c $ . Disregarding the constant $ w{l}_{ir}^x $ , we can write the indirect regional utility function as

(5) $$ {V}_{ir}=w{n}_{ir}{l}_{ir}\left[\mathrm{E}\left({\lambda}_{ir}\right)\right]-{n}_{ir}\delta \left[{e}_{ir}\left[\mathrm{E}\left({\lambda}_{ir}\right)\right]\right]\;\mathrm{s}.\mathrm{t}.\hskip0.5em \left({m}_{ir}+{d}_{ir}\right){\epsilon}_{ir}\le {E}_{ir},\hskip0.36em {E}_{i1}+{E}_{i2}\le {E}_i. $$

The national indirect utility (welfare) function is

(6) $$ {W}_i=w\sum \limits_{r=1}^2{n}_{ir}{l}_{ir}\left[\mathrm{E}\left({\lambda}_i\right)\right]-{n}_i\delta \left[{e}_i\left[\mathrm{E}\left({\lambda}_i\right)\right]\right]\;\mathrm{s}.\mathrm{t}.\hskip0.5em \left({m}_{i1}+{d}_{i1}\right){\epsilon}_{i1}+\left({m}_{i2}+{d}_{i2}\right){\epsilon}_{i2}\le {E}_i. $$

EU’s welfare function is given by

(7) $$ W=w\sum \limits_{i=F}^S\sum \limits_{r=1}^2{n}_{ir}{l}_{ir}\left[\mathrm{E}\left(\lambda \right)\right]-\sum \limits_{i=F}^S\sum \limits_{r=1}^2{n}_{ir}\delta \left[{e}_{ir}\left[\mathrm{E}\left(\lambda \right)\right]\right]\;\mathrm{s}.\mathrm{t}.\hskip0.5em {E}_F+{E}_S\le E. $$

The EU determines the aggregate level of environmental policy, which is differentiated across the MS in a manner that promotes cost efficiency. Optimal policy is determined by maximizing equation (5) for local pollutants and equation (7) for global pollutants with respect to $ {\lambda}_{ir} $ and $ \lambda $ , respectively, subject to $ E\ge {\sum}_{ir}{m}_{ir}{\epsilon}_{ir} $ – we assume that EU makes its policy under certainty, having done an RIA ahead of its prosal. Moreover, a national government may use equation (6), differentiated with respect to a uniform national policy, $ {\lambda}_i $ , to set a common national policy. This yields for local, national and EU policy, respectively:

(8) $$ w{n}_{ir}{l}_{ir}^{\prime}\left[\mathrm{E}\left({\lambda}_{ir}\right)\right]={n}_{ir}\left(1-3\kappa \right)\left({m}_{ir}+{d}_{ir}\right){\epsilon}_{ir}^{\prime }{\delta}^{\prime}\left[{e}_{ir}\left[\mathrm{E}\left({\lambda}_{ir}\right)\right]\right], $$

(9) $$ w\sum \limits_{r=1}^2{n}_{ir}{l}_{ir}^{\prime}\left[\mathrm{E}\left({\lambda}_i\right)\right]=\left(1-2\kappa \right){\delta}^{\prime}\left[{e}_i\left[\mathrm{E}\left({\lambda}_i\right)\right]\right]\sum \limits_{r=1}^2{n}_{ir}\left({m}_{ir}+{d}_{ir}\right){\epsilon}_{ir}^{\prime}\left[E\left({\lambda}_i\right)\right], $$

(10) $$ w\sum \limits_{i=F}^S\sum \limits_{r=1}^2{n}_{ir}{l}_{ir}^{\prime}\left[\mathrm{E}\left(\lambda \right)\right]={\delta}^{\prime}\left[e\left[\mathrm{E}\left({\lambda}^{\ast}\right)\right]\right]\sum \limits_{i=F}^S\sum \limits_{r=1}^2{n}_{ir}\left({m}_{ir}+{d}_{ir}\right){\epsilon}_{ir}^{\prime}\left[E\left(\lambda \right)\right]. $$

If an RIA has been made it becomes possible to set $ {d}_{ir}=0 $ .Footnote ¹⁵ The left-hand side in equations (8) to (10) reflects the marginal costs of emission reduction in terms of reduced employment, while the right-hand side captures the marginal benefits. Policy then fulfills the textbook definition of optimal policy. For a local pollutant, that is, in the absence of spillovers ( $ \kappa =0 $ ), policy is optimally determined for each region, and the optimal shadow price of policy may vary between regions depending on the respective sizes of the marginal costs and marginal benefits of emission control. For a global external effect, the shadow price on emissions should be equalized for both MSs and all four regions.

4. Policy choice under different institutions

The process of doing or augmenting national RIAs as the EU’s preparation process for legislation proceeds has been described by Hammes and Nerhagen (Reference Hammes and Nerhagen2023) for Finland. Based on Uusikylä et al. (Reference Uusikylä, Ahonen and Takanen2015), Hammes and Nerhagen note that an RIA can be done in one or, preferably, several of the five identified windows in which EU’s decision-making proceeds. Thus, in window 0, when the government obtains unofficial information that the EU Commission is thinking about possible new legislation, a baseline analysis could be done. In window I, the Commission publishes its work program, a road map or a green or white book. At this point, a fuller national impact assessment of possible policy options becomes available based on the Commission’s early studies. In window II, when the Commission’s official proposal is laid out, the baseline analysis or RIA done in windows 0 or I can be updated at a short notice to reflect the final proposition. In window III, the Commission’s proposal is negotiated in the European Parliament and the Council of Ministers. At this point, the co-legislators can suggest different compromises, which can be numerically assessed within the framework of the previous RIAs. Finally, in window IV, when the issue comes to national implementation, the RIA can be used as a basis for post-evaluations. Given that a national RIA along these lines has been made, the correct implementation of EU legislation in national legislation will be easier. Hammes and Nerhagen note that no such process exists in Sweden and that the country’s approach to influencing EU legislation should be reorganized.Footnote ¹⁶

Figure 3 shows a stylized model of the government structure in Finland and Sweden – a decentralized governance structure for Sweden and a centralized one for Finland. The figure is an adaptation of the model of decision-making in firms that has been used, for example, by Milgrom and Roberts (Reference Milgrom and Roberts1992) and by Martin (Reference Martin1994). The administrative system of Finland is then seen as a “vertically integrated” one, while the Swedish system is a “delegating” one. Reasons given for vertical integration include increased efficiency, that is, negotiations with many actors entail large transaction costs, bounded rationality, that is, a better control over the enterprise and increased flexibility and hindering opportunism, that is, inducing the employees to act in the respective organization’s interests (Martin, Reference Martin1994).

Figure 3. Two governance structures: a decentralized (fractured) and a centralized one, and the paths to information about the local level these lead to. Adapted from Milgrom and Roberts (Reference Milgrom and Roberts1992).

Figure 3 is used to illustrate the decision to carry out an RIA or not – a decision made at first at the ministry level. In the decentralized governance structure, the decision is made a second time, too, this time at a government agency level. We do not have enough information to assign any probabilities for the “yes” and “no” decisions by the government ministry or the agency. However, as an illustration we could assume that the probability of carrying out an RIA is 50 per cent at each node. Thus, in the centralized country, the ministry will with a 50 per cent probability perform an RIA. In the decentralized country, the probability of doing this is 50 per cent times 50 per cent, that is, 25 per cent, since both the ministry and the agency will make the decision with a 50 per cent probability. Thus, ceteris paribus, the decentralized country is much less likely than the centralized country to carry out an RIA.

We will interpret Figure 3, that is, examine the consequences that performing or not performing an RIA has on the economic efficiency of implementing EU legislation, especially directives, and the impact of delegating the RIA to a one-issue agency in the following sections. EU directives, unlike EU regulations, which are binding legislative acts, leave leeway for national implementation. We will start from the centralized governance structure, that is, Finland, and then proceed to examine the Swedish case.

4.1. Centralized governance structure

In the centralized government structure, the government ministries have capacity to perform, or order, an RIA. In this process, they can turn to the local agencies, which are directly placed under them in the organization, for information about the local circumstances. In Finland, the 15 ELY centers deal with economic, transportation and environmental issues and are able to provide information about at least these aspects of a legislation. This gives the government the information needed to carry out multi-dimensional RIAs.

Completing an RIA means obtaining information to determine the value of $ {d}_{ir} $ in equations (1) and (2), and setting policy without uncertainty. In addition to performing an RIA, the ministry decides whether to obtain more information about the regional consequences of legislation. In case, the government does not obtain more information, it sets a uniform policy for the entire country in line with equation (9). This is denoted by the policy having a shadow price, $ {\lambda}_i $ , that is equal for both regions.

We start by examining a local pollutant and summarize the findings in the following proposition:

Proposition 1. A government, doing a national-level RIA and setting a uniform environmental policy to regulate a local pollutant, chooses a policy that is too lax for region 1 and too strict for region 2.

Proof: The optimal policy for a local pollutant is obtained by maximizing equation (5), while the government maximizes equation (6) to set a national policy. We prove Proposition 1 by showing that a national shadow price cannot exceed the shadow price in region 1, and that the shadow price in region 2 can never exceed the national shadow price. For a local pollutant ( $ \kappa =0 $ ), we start with region 1 and set $ {\lambda}_{i1}<{\lambda}_i $ , which indicates that the damages are greater in region 1 than the national average: $ {\delta}_{i1}^{\prime}\left[{e}_{i1}\left[E\left({\lambda}_{i1}\right)\right]\right]>{\delta}_i\left[{e}_i\left[E\left({\lambda}_i\right)\right]\right] $ . Assuming that both the marginal supply of labor and marginal emissions are linear in $ \lambda $ , we solve for the marginal damages from equations (8) and (9):

(11) $$ {\delta}^{\prime}\left({e}_{i1}\right)=\frac{w{n}_{i1}{l}_{i1}^{\prime}\left[E\left({\lambda}_{i1}\right)\right]}{n_{i1}\left({m}_{i1}+{d}_{i1}\right){\epsilon}_{i1}\left[E\left({\lambda}_{i1}\right)\right]}>\frac{w{\sum}_{r=1}^2{n}_{ir}{l}_{ir}^{\prime}\left[E\left({\lambda}_i\right)\right]}{\sum_{r=1}^2{n}_{ir}\left({m}_{ir}+{d}_{ir}\right){\epsilon}_{ir}^{\prime}\left[E\left({\lambda}_i\right)\right]}={\delta}^{\prime}\left({e}_i\right). $$

Simplifying equation (11), setting $ {d}_{ir}=0 $ since we assume that the government has done an RIA yields

$$ 1>\frac{m_{i1}\left({n}_{i1}+{n}_{i2}\right)}{m_{i1}{n}_{i1}+{m}_{i2}{n}_{i2}}. $$

But we have assumed that $ {m}_{i1}\gg {m}_{i2} $ , meaning that the right-hand side is greater than one. Equation (11) thus cannot hold and it must be that $ {\lambda}_{i1}>{\lambda}_i $ . It can similarly be shown that it must be that $ {\lambda}_{i2}<{\lambda}_i $ . The government’s chosen policy is thus lower than optimal for region 1 and higher than optimal for region 2: $ {\lambda}_{i2}^{\ast }<{\lambda}_i<{\lambda}_{i1}^{\ast } $ .

The result in Proposition 1 is unsurprising and well established in the literature; see, for example, Baumol and Oates (Reference Baumol and Oates1988). It states that despite performing an RIA, a government that does not differentiate policy to regulate a local pollutant but relies on a national policy instead, chooses a suboptimal policy.

Proposition 1 does not hold for a global pollutant, however. For these pollutants, it is cost efficient to set policy so that the marginal cost of emissions reduction is equal everywhere and for all firms. Thus, it falls on the EU to propose emission regulation that equalizes the marginal cost of regulation in both countries, ceteris paribus. The centralized country, in turn, chooses the optimal policy by maximizing equation (6) subject to its emissions not exceeding the EU’s target emissions: $ {e}_{i1}+{e}_{i2}\le {E}_i $ . The solution is shown in equation (9). If an RIA has been made, the optimal policy can be chosen. Without an RIA, the policy outcome is random, and the policy is set at a suboptimal level.

4.2. Decentralized governance structure

In a decentralized governance system, the government has very little capacity for performing RIAs. Such systems have been examined in an extensive literature on delegating, since the information-search activity may be delegated to government agencies (Epstein & O’Halloran, Reference Epstein and O’Halloran1999; Bendor et al., Reference Bendor, Glazer and Hammond2001). However, should the government decide not to delegate, it would choose policy with no information about $ {d}_{ir} $ . In this case, the government chooses an inefficient policy with a very high likelihood: the probability of choosing exactly the correct policy given the continuously distributed $ {d}_{ir} $ is zero. Whether the policy outcome under- or overshoots the optimal policy cannot be determined at the present level of generality; the probability of either happening is equally large. This is the same outcome as for the government that does not performing an RIA in the previous section.

Most of the time, then, the government of Sweden delegates to a government agency. These agencies usually have a one-dimensional policy space within which they work, however. Thus, if the issue is delegated to an agency dealing with environmental issues, it is these aspects that the agency will concentrate on. If it is delegated to an agency tasked with, for example, promoting regional growth, the agency will concentrate on this aspect.

If the agency decides not to obtain more information about the local/regional impact of a policy, it will propose a uniform national policy for the government, based on an RIA at this level. Instead of maximizing welfare as given by equation (6), it will maximize one of the following welfare functions, depending on which aspect of the policy it works with:

(12) $$ {W}_i^e=-\sum \limits_{r=1}^2{n}_{ir}\delta \left[{e}_{ir}\left[\mathrm{E}\left({\lambda}_i^e\right)\right]\right]\;\mathrm{s}.\mathrm{t}.\hskip0.5em {m}_{i1}{\epsilon}_{i1}+{m}_{i2}{\epsilon}_{i2}\le {E}_i $$

(13) $$ {W}_i^l=w\sum \limits_{r=1}^2{n}_{ir}{l}_{ir}\left[\mathrm{E}\left({\lambda}_i^l\right)\right]\;\mathrm{s}.\mathrm{t}.\hskip0.5em {m}_{i1}{\epsilon}_{i1}+{m}_{i2}{\epsilon}_{i2}\le {E}_i $$

Equation (12) is used by an agency concentrating on environmental issues and equation (13) by one dealing with economic growth and/or employment; however, not with regional growth here, since the agency does not differentiate its policy between regions. Equations (12) and (13) can be used to prove the following proposition:

Proposition 2. Solving a unidimensional environmental policy problem, regardless of whether the environmental problem is a global or a local one, results in policy that deviates from the optimal.

Proof: We prove Proposition 2 for a national-level policy. The solution for a local pollutant is analogous. Maximizing equations (12) and (13) with respect to $ {\lambda}_i $ and rearranging yields for the environmental agency and the economic growth agency, respectively:

(14) $$ \left(1-2\kappa \right){\delta}^{\prime}\left[{e}_i\left[\mathrm{E}\left({\lambda}_i^e\right)\right]\right]\sum \limits_{r=1}^2{n}_{ir}\left({m}_{ir}+{d}_{ir}\right){\epsilon}_{ir}^{\prime }=0, $$

(15) $$ -w\sum \limits_{r=1}^2{n}_{ir}{l}_{ir}^{\prime}\left[\mathrm{E}\left({\lambda}_i\right)\right]=0. $$

Comparing equation (14) with the quasi-optimal policy for a global pollutant in equation (9) indicates that the environmental agency, ignoring the labor market consequences of the policy, will set a stricter than optimal policy, and in fact strengthens the environmental policy up to a point where the marginal benefit from a reduction in emissions is zero. Consequently, $ {\lambda}_i^e>{\lambda}_i $ , that is, the policy that is set taking only the environmental damages into account is stricter than a policy considering all effects. The economic growth agency in turn will struggle to internalize the external effect since it only considers the negative impact on employment, thus setting a policy that is too low, $ {\lambda}_i^l<{\lambda}_i $ .∎

A corollary of Proposition 2 is that the country with a decentralized governance structure will set a stricter policy governing the global pollutant than is optimal from a global point of view. Thus, the policy determined by the EU is no longer optimal, ceteris paribus.

The result from Proposition 2 carries over to a situation where the government agency obtains more information from the local/regional impact of regulation. In a similar way as in the proposition, an environmental agency obtaining only environmental information from the county administrative boards sets a too strict policy for both regions: $ {\lambda}_{ir}^e>{\lambda}_{ir} $ . The only question that remains is, whether an agency dealing with regional growth, but which obtains more information about a policy’s impact on a regional level from the county administrative boards, which mainly deal with environmental issues, might manage to balance the two opposite impacts affecting general welfare. As the Swedish Agency for Economic and Regional Growth is not an agency primarily tasked with environmental policy, we consider this point moot and conclude that the Swedish environmental policy, due to the way in which the governance system is constructed, tends to impose stricter-than-optimal environmental policies.

5. Relevance and empirical evidence for the model

In this section, we will give some examples of how Finland and Sweden have implemented environmental EU directives. We will also discuss their frameworks related to RIA. The latter is interesting since it is a means to analyze the efficiency of (environmental) policy. We start by commenting on the reasonableness of the model of the Swedish government, however.

The Swedish approach, delineated in Proposition 2, may seem very extreme. However, from previous research we know that RIA is not an established practice in Swedish policy-making (Fors, Reference Fors2017; Nerhagen et al., Reference Nerhagen, Forsstedt and Hultkrantz2017; Nerhagen & Forsstedt, Reference Nerhagen and Forsstedt2019; Better Regulation Council, 2023). According to Radaelli (Reference Radaelli2010), this can be explained by the specific culture and tradition of policy-making in Sweden, that is, by the use of government inquiries and a referral process. Regarding environmental policy, it was only in 2010 that the SEPA was commissioned to develop the use of benefit-cost analysis in its work with the EOS (Government Proposition 2009/10:155, 2010). Since then, some evaluations of how the method is applied have been made (Söderholm, Reference Söderholm2014; Wallström & Söderqvist, Reference Wallström and Söderqvist2017; Swedish Environmental Protection Agency, 2020).Footnote ¹⁸ A change of practice might however be underway based on a proposal for a new law presented by the Ministry of Finance (2022).

A couple of examples can be used to illustrate that Proposition 2 is relevant for the Swedish policy-making context. The first is the above-mentioned evaluation by Söderholm (Reference Söderholm2014, 10), who gives a number of examples where the direct costs of an action (e.g., to achieve a goal of reducing noise indoors, or of reducing energy consumption in buildings by a certain amount) have been quantified by government agencies. However, the problem he identifies is that these estimates only weakly reflect the true costs of the actions. These analyses focus on market prices and do not include all relevant societal effects such as external costs of the investments, for example noise emissions from windmills. Moreover, they disregard questions such as which policy instruments must be implemented, the cost of implementing these policy instruments and how high de facto costs the actors face due to the policy instruments, for example, transaction costs or losses of consumer or producer surplus. Moreover, the quantification of the benefits of regulation is also of poor quality.

The second example is from a government inquiry that proposed a framework for climate policy in Sweden (SOU, 2016, 21). A goal of zero net emissions of greenhouse gases in year 2045 was proposed, 85 per cent of which will have to be reached in Sweden and 15 per cent can be bought from abroad in the form of emissions reductions there.Footnote ¹⁹ The inquiry quantified the costs using the TIMES-Sweden model, but only after it had already decided its conclusions, i.e. the cost estimates in no way influenced its recommendations. Our third example is from a statement made by the Ministry of the Environment in preparation for the planned revision of the EU’s air quality directive (Ministry of the Environment, 2022). The pro memorandum states, as a matter of fact, that an RIA of the possible consequences for Sweden has not been carried out. Instead, it claims that the benefits and costs, that is, the outcome of the RIA performed for the entire EU, also apply to Sweden, and concludes that the benefits will exceed the costs. This conclusion is in contrast to information presented by the European Environment Agency showing that zero per cent of the population in urban areas are exposed to concentrations above EU air pollution standards in SwedenFootnote ²⁰.

In Finland, the policy preparation and negotiation phases ahead of EU legislation were studied in 2014 (Uusikylä et al., Reference Uusikylä, Ahonen and Takanen2015), and the Finnish Council of RIA was created in conjunction with the Prime Minister’s office and began work in 2016. According to the OECD (2019), an RIA nowadays is regularly performed ahead of strategically important negotiations. The Government adopts an annual plan for analysis, assessment and research that underpins policy decision-making and steers analysis, assessment and research activities toward specific priority areas selected by the Government. Under the leadership of the Prime Minister’s Office, the Government working group for the coordination of research, foresight and assessment activities oversees the formulation of the plan. The working group includes experts from all administrative branches. There are also joint analysis, assessment and research activities coordinated by the Government, which generate information that supports decision-making procedures, work practices and management by knowledge (Prime Minister’s Office, 2022).

The Finnish Council of Regulatory Impact Analysis makes a yearly synthesis of its work. In its latest summary (Finnish Council of Regulatory Impact Assessment, 2023), it notes that during 2022, it presented 43 statements altogether, covering about 12 per cent of all government proposals.Footnote ²¹ The Council has made statements both of government proposals that are of broad societal and economic importance, and of proposals of less broad importance that nevertheless have societal and economic impacts. The Council notes that while the quality of the RIAs has increased over time, its recommendations are still not followed to a high enough degree. It recommends that more resources at the ministries should be directed toward preparing legislation in multidisciplinary teams, and that the government’s program, instead of dictating the means to reach a goal, should concentrate on defining the goal. Finally, the Council wants its position to be written into legislation, instead of the present ad hoc arrangement.

Finally, two examples show that Finland and Sweden have implemented EU directives differently when they could be expected to be similar. The first example is the air quality directive. Finland implemented the limit values in line with the directive (Kukkonen et al., Reference Kukkonen, Salmi, Saari, Konttinen and Kartastenpää1999), while Sweden chose a more strict implementation (SOU, 2015, 27). For example, Finland identified and deducted the contribution from sanding and salting of icy roads in the measurement of particulate matter (PM 10) while Sweden did not, despite sanding and salting in the winter being used in both countries.

The second example concerns the action plan that all MSs had to submit in 2016 following Directive 2014/94/EU on the deployment of alternative fuels infrastructure. In the evaluation done by the EU (SWD 365), it was concluded that while Finland had fulfilled all obligations, Sweden had not and needed to send in a revised and updated version (Hansson, Reference Hansson2020; Nerhagen et al., Reference Nerhagen, Jussila Hammes and Pyddoke2021).

6. Summary and conclusions

This paper describes how EU membership has shaped the government systems in two Nordic countries Finland and Sweden, and the possible implications of this on the efficiency of EU policies. An overview of the actions taken by the two countries to adapt their state administrations to their accession to the EU leads us to conclude that from the beginning the strategies that the countries chose differed greatly. While Sweden had the ambition of changing the Union, not least regarding its environmental policy, Finland wanted to learn “the rules of the game” as fast as possible. Regarding the administrative system, Sweden has kept the administrative model established centuries ago, which implies much more delegation of preparatory work to government agencies. In contrast, Finland has modified its government system compared with the pre-EU era, which has resulted in more government control and enabled strategic work on EU matters. Early on, it also established processes for cooperation between the government and the parliament on EU matters, thus enhancing the democratic legitimacy of its EU policy.

From the literature, it is clear that the organization of a government influences policy- and law-making. To study the impact of institutional design on policy outcomes, we analyze a model of a small open economy with emissions resulting from the production of a traded good. The emissions can have either local or global impacts. We conclude that the administrative system in a country can influence the effectiveness of environmental policies. The reason is that with a higher degree of delegation and specialization, it is less likely that a thorough analysis that includes more than one or a few of the consequences of legislation is performed. Inefficiencies in our model can arise from two distinct sources: first, a government imposing a uniform national policy when a pollutant is a local one generates inefficient policies when the regions constituting a country differ from one another and the marginal cost of emissions thus varies spatially, and second, imperfect information, whether it arises from government ministries too small to perform the analysis themselves or from delegating analysis-making to agencies with too narrow (unidimensional) agendas. A third possible source of inefficiency, not included in the model, is that the EU’s policy may be wrong to begin with, possibly due to insufficient input from the MSs.

The main consequence of ineffective environmental policies is a drag on the economy. Therefore, a logical way of examining its impacts would be to look at the development of some measure of economic welfare in the countries concerned. However, it is unclear from the literature whether environmental policy so far has had a detrimental effect on economic growth, that is, evidence for the industrial flight hypothesis remains inconclusive. Our results also illustrate that deviations from an optimal policy in one country can have implications for the effectiveness of a policy at the EU level.

Moreover, our model highlights that a fragmented government system with much decentralization can have implications for policy-making and how well the implementation of EU directives is adapted to country-specific circumstances. As noted, the aim of directives is to allow for some flexibility in the implementation of EU legislation. We have illustrated with some examples that there appears to be a difference in the actual implementation between Sweden and Finland, possibly resulting from Finland having adopted the management practices of the EU and being proactive in the legislative process. For future research, in order to gain a better understanding of the differences in outcome, it could be interesting to systematically study some cases from the infringement database, or to catalogize the RIAs that have been made in both countries vis-à-vis EU directives, and count the exceptions requested by both countries.

Regarding policy, we make no claims of having identified an optimal level of government centralization or decentralization. However, we note that there might be a case for Sweden to strengthen its ministries similar to what Finland has done. As noted, this is an issue that has been raised previously in the literature and more recently also by the Swedish Climate Policy Council in its yearly report from 2022 (Swedish Climate Policy Council, 2022). Moreover, we recommend moving the responsibility for overseeing RIAs to a ministry, preferably the Prime Minister’s Office, but possibly even the Ministry of Finance would suffice. This institutional change would have to be set in law. This change would ensure the domestic use of the methodology for RIAs used in the EU. In addition, the government and parliament need to take greater responsibility for the implementation of EU directives and the negotiations that precede them.

Currently, we do not have any recommendations as to how Finland might enhance its system of RIAs other than that we support all three proposals by the Finnish Council of Regulatory Impact Assessment (2023) for its continued work delineated above. Above all, we support its proposition to write its status in law, instead of the present ad hoc arrangement. Future research, for example, on the attitudes and competences of civil servants in both countries, might yield additional insights, however. Other questions for future research include the causes of the differences between Finland and Sweden, and the reasons for the shortcomings in Sweden: Is there a lack of understanding of how environmental policy is designed by the EU, which leads to incomplete implementation? What might cause such a lack of understanding? And how does a decentralized versus a centralized government system impact capacity building and organizational learning?