Municipally Owned Enterprises as Danger Zones for Corruption? How Politicians Having Feet in Two Camps May Undermine Conditions for Accountability

The market-inspired reforms of New Public Management have been particularly pronounced in Swedish local government. Notably, municipally owned enterprises (MOEs) have rapidly grown in numbers. Principal-agent theory gives rise to the hypothesis that the massive introduction of MOEs has impacted negatively on the conditions for accountability in Swedish local government. To study this, social network analysis was employed in mapping networks for 223 MOEs in 11 strategically chosen municipalities, covering a total of 732 politicians. The analysis reveals substantial overlaps between principals (representatives of the ultimate stakeholders, citizens) and agents (the boards of the MOEs). Hence, corporatization of public services seems to imply worrisome entanglements between the politicians who are set to steer, govern, and oversee MOEs on the one hand, and the board members of MOEs on the other. The increasing numbers of MOEs may therefore have adverse effects on accountability in important and growing parts of Swedish local government.

Since the mid-1980s, public sector reform in Organization for Economic Co-operation and Development (OECD) countries has been characterized by New Public Management (NPM). Local governments in Sweden have gone comparatively far in implementing these market-inspired reforms (e.g., decentralization, contracting out, and introduction of private-sector styles of management). The creation of municipally owned enterprises (MOEs) is a form of "quasi-privatization" in public sector reform in OECD countries (Christensen & Laegreid, 2003). An important motivation for local governments to create, own, and operate enterprises has been to increase productivity and lower costs in delivering public goods and services. Indeed, a recent literature review found that MOEs are somewhat more effective than traditional public management (Voorn, Van Genugten, & Van Thiel, 2017). However, efficiency is but one value to consider in the public sector. Several scholars have raised concerns that NPM reforms may have unattractive side effects (e.g., undermining political control; accountability; public ethics; and ultimately, democracy) (Andersson & Erlingsson, 2012;Andr e, 2010;Forrer, Kee, Newcomer, & Boyer, 2010;Shaoul, Stafford, & Stapleton, 2012). Against this backdrop, this article asks questions about such unattractive side effects of NPM in general, and the creation of MOEs in particular.
MOEs are not unique to Sweden, but are a rather common organizational form in several other countries as well (Grossi & Reichard, 2008;Voorn et al., 2017). When Dexia Crediop (2004) gathered information about MOEs, they demonstrated that these exist in all EU countries except Luxembourg. However, the number varies considerably, and it is worth noting that after Latvia, Sweden is the EU country with the second most companies per capita in the EU.
What makes Sweden particularly interesting from this point of view, as shown in Figure 1, is that the total number of MOEs in Sweden has increased dramatically over time.
As MOEs have grown in number, they have become ever more important players in their respective local economies. Still, since scholarly attention to this development has been scant, little is known about the consequences of the introduction of MOEs in the operations of local governments; in particular its effects on the conditions for accountability.

MUNICIPALLY OWNED ENTERPRISES AS DANGER ZONES FOR CORRUPTION?
When debating accountability in the Swedish context, the increasing concentration of powers must be considered. As shown in Table 1, in 1999, 10% of Swedish local councilors held three political positions or more-in 2015, 15% of them did. The number of elected politicians with exactly one position has fallen from 60% to 55%. In addition, previous studies have shown that the turnover of Swedish local political elites is low, compared to other countries, and local Swedish elite networks tend to be more socially exclusive than elsewhere. Furthermore, the more influential the political assignment is, the more likely it is to be held by an older man (Eldersveld, Str€ omberg, & Derksen, 1995;Gidlund & M€ oller, 1999;Petersson, Hermansson, Michelletti, & Westholm, 1996;SCB, 2016;SOU, 2006).
The trend toward increasing concentration of powers to a group of socially exclusive individuals is worrying from an accountability perspective, because poor social representationin particular low representation of women-may have implications for policy outcomes (Svaleryd, 2009), and has been associated with increased risks for corruption (Jha & Sarangi, 2018). Parallel with the increasing number of MOEs, where local elites often hold influential positions as members of their boards, the increased concentration of powers in local government makes it pressing to explore the quality of accountability and checks and balances in Swedish local government.
Ultimately, the municipal citizens are the principal stakeholders of MOEs. Their abilities to hold MOEs accountable rest on there being sufficient transparency and separation of powers. Since high societal and economic values are at stake, it is crucial that social scientists ask critical questions about the accountability of these publicly owned and managed organizations. This article contributes to the debate on publicly owned enterprises with new insights by combining social network analysis and answers from a newly conducted survey studying the composition and the role of MOE boards.

LOCAL GOVERNMENT AND MOEs IN SWEDEN
Local governments in Sweden offer transparency and accountability scholars a rich environment. Although a unitary state, local self-government is stressed in the Swedish constitution (Chap. 1, §1), and municipalities enjoy a comparatively high degree of policy-making autonomy and financial independence (Ladner, Keuffer, & Baldersheim, 2016). Additionally, local governments have been made responsible for core areas in the welfare state (e.g., preschool; elementary school; upper secondary school; social services; and care for the elderly). The local level's importance in Sweden is further stressed by the fact that out of a total of 1.3 million publicly employed Swedes (by central, regional, and local governments), 60% are employed by the municipalities. Furthermore, almost two thirds of all public finances in Sweden are spent by local governments.
Sweden is organized into 290 municipalities and 20 regions. The current structure can be traced back to a process that started in 1952 with large-scale municipal amalgamations, which were finalized in 1974. The number of municipalities then fell from 2,498 to 277 (and later, through secessions, increased to today's 290). With this, the number of elected representatives fell from circa 200,000 local politicians to 50,000 in these years, while Sweden's population increased from 7.2 million to 8.2 million. The rationale underlying these changes was to increase the efficiency of local service provision, but a potential side effect, it could be argued, was that these changes paved the way for problems related to concentration of powers and accountability, through the dramatic decrease in politicians per capita. Today, the number of politicians in local government is circa 36,800, while the Swedish population is now close to 10.2 million.
Increased professionalization of local politics is one thing; the quality of accountability is another. As important as municipalities are in Sweden, serious doubts have been raised about the quality of supervision, audit, and hence, accountability, in Swedish local government. Over the past 10 years, almost a third of all local news outlets have been shut down (Nygren & Alth en, 2014). In addition, the quality of local media coverage has been questioned by scholars (Hanberger, Khakee, Nygren, & Segerholm, 2005;Nord & Nygren, 2007). Moreover, the way auditing in Swedish municipalities is designed, and ultimately carried out, has been met with skepticism. This is in part due to the fact that auditors-in relation to the weight of their assignment-work with limited resources (Hanberger et al., 2005;SKL, 2013), and in part because they are elected from a pool of lay politicians that already have preexisting ties to the fellow councilors within the municipality whom they are set to review. Not surprisingly, the independence of local auditors has been called into question (Lundin, 2010). An important backdrop for this analysis is precisely increased professionalization and concentration of powers on the one hand, and the weak mechanisms for accountability on the other.
Before presenting the data, analyses, and results, a few words on MOEs are needed. As documented by the Swedish Agency for Public Management, MOEs are typically responsible for areas producing different kinds of public goods, such as: (1) management of housing, (2) electricity and heating, and (3) water and sewer (Statskontoret, 2012). However, almost 40% of MOEs are active within miscellaneous areas that have been increasingly hard to categorize. Against the background, it is not surprising that the Swedish Competition Authority has criticized MOEs for competing with private firms in already established markets (Konkurrensverket, 2014).
There are strong reasons to examine the role of MOEs from an accountability perspective. First, auditors have complained that it is more difficult to audit and review MOEs compared to public administration proper (SKL, 2013). Second, journalists have reported that it is harder to conduct investigative journalism on MOEs (Hyltner & Velasco, 2009), a fact related to their unwillingness to follow Swedish right to information laws (Erlingsson & Wittberg, 2017;SOU, 2013). Third, MOEs tend to be overrepresented in the corruption scandals that have been exposed in local government (Amnå, Czarniawska, & Marcusson, 2013;Erlingsson, Bergh, & Sj€ olin, 2008;Statskontoret, 2012). The number of MOEs per capita has also been linked to perceived corruption, based on corruption perceptions of roughly 9,000 representatives in local councils (Dahlstr€ om & Sundell, 2014). In sum, there are undoubtedly strong reasons to examine if, and how, the increased willingness of Swedish local governments to create, own, and operate MOEs affects conditions for accountability; and hence, if it can be argued that they are particular danger zones for corruption in a "low corrupt" setting such as the Swedish one.

THEORY: ACCOUNTABILITY AND THE "ARM'S LENGTH" PRINCIPLE
Problems related to the separation of ownership and management in corporate governance have typically been analyzed through the lens of principal-agent theory (PA). The basic model assumes that one party (the principal) delegates an assignment to another party (the agent) (Eisenhardt, 1989). In situations when the principal and the agent have diverging interests, and it is very costly-or even impossible-for principals to monitor the agents, there is a risk that the agent may act in accordance with his/her private interest, rather than representing the principal's. Hence, the asymmetry of information is assumed to benefit the agent. The PA tradition has also been employed within political economy. Here, elected representatives are viewed as "agents," and voters are the "principals" (Maskin & Tirole, 2004;T. Persson & Tabellini, 2000). Voters have varying interests that they want politicians to pursue, such as providing public goods, redistributing income, and so on. Politicians might share such goals, but they might also want to pursue various private interests (e.g., seek side-payments associated with the assignment, engage in rent-seeking, helping their friends and relatives, and similar objectives, to which the voters would object) (Str€ om, Bergman, & M€ uller, 2003).
In this article, PA theory is used to analyze the relationship between voters and politicians, as well as to analyze the relationship between politicians and the MOEs they own. The PA perspective is a suitable tool to highlight conditions for accountability in the particular types of relationships to which MOEs give rise. It is so, because MOEs are hybrid organizations, in the sense that they are simultaneously political organizations and enterprises. As a result, they generate what best can be labeled "nested" or "layered" PA problems. In MOEs, chains of delegations, and hence accountability, become more complex processes than in ordinary firms since they are one additional step farther away from their ultimate stakeholders-the citizens. First, voters elect and are represented by politicians in councils as well as politically appointed boards and committees. These, in turn manage the taxpayer's money, but may have preferences and priorities that differ from those of the voters. Even if elected representatives are assumed to act in complete accordance with the interest of voters, the management of MOEs must still be delegated to managers, thus nesting-or layering-several PA problems in a complex chain of delegation. Every step of delegation involves a transfer of powers and therefore contains risks of deviation from the optimal outcome from the principals' perspective, or what has been dubbed "agency loss" (Lupia, 2003). Ensuring accountability is a way of balancing these risks by enabling punishment of abuse of power. Following Lupia (2003), this study uses the concept of accountability in the sense that an agent is accountable to a principal if the principal can exercise control over the agent. Delegation is not accountable if the principal is unable to exercise control.
The applicability of principal-agent theories to corruption problems has been questioned. Persson, Rothstein, and Teorell (2013) stress that PA theory is not suitable for analyzing cases of systematic corruption in developing countries, for example, and that these cases rather resemble collective action problems. Their main point is that, in the case of systematic corruption, the outcome of an individual's choice is dependent on the behavior of other citizens. For a single individual it is not even rational to try to curb corruption. Therefore, there are no principals in the first place. Another objection to PA theory, as suggested by Caplan (2011) and Achen and Bartels (2017), is that voters lack information or capability to act as principals toward their politicians.
While there is merit to both objections, they do not render the PA perspective irrelevant in the case of analyzing conditions for accountability when MOEs are created and operated by local government. First, while PA theory may be unsuitable for analyzing systematic corruption in developing countries like Uganda, it is indeed more relevant for studies of corruption in developed countries. As noted above, exposing corruption in Sweden typically causes scandals, because it upsets many voters; corruption here is not "the name of the game" but a rather isolated phenomenon which is frowned upon.
With regard to Achens and Bartels' nonrational voter critique (2017), it is indeed correct that it is harder for irrational and/or uninformed voters to act as principals. That observation, however, does not change the fact that the idea that voters can hold politicians accountable is viewed by many as a cornerstone of democracy. If anything, the need for separation of powers and the need to avoid concentration of powers must indeed be seen as far stronger when voters are uninformed or irrational.
Previous research has suggested that the members of MOE boards are mainly politicians which have been elected by a municipal nominating committee on the principle of proportional representation (Sv€ ard, 2016). Thus, memberships on MOE boards are given to political parties in proportion to the electoral result. This principle is based on the assumption that members of the corporate boards function as an "information system" that audits and oversees mangers and hence the PA problem. This is a view that is common when it comes to private firms (Fama & Jensen, 1983). In the specific case of MOEs, the members of the boards are assumed to contribute with knowledge about political objectives and regulations, thus having the capacity to audit the MOEs from a political point of view. In theory, such a structure has several benefits, and would not necessarily have to constitute a problem.
As will become evident in the empirical section, it is common that councilors and members of the municipality board have positions as board members of MOEs. Such overlaps are potentially problematic from the perspective of accountability in at least two ways. First, they may create conflict of interests since the council and the municipality board are representing the owners' interest vis-a-vis the MOEs. The board members of MOEs are continuously involved in strategic decisions, hence, the owners ought to be able to hold them accountable for the overall governance of the MOEs.
From the "information system" approach described above, one could of course argue that politicians are "angels" and that it is therefore not problematic that they simultaneously act MUNICIPALLY OWNED ENTERPRISES AS DANGER ZONES FOR CORRUPTION? as principals and agents; having feet in both camps, at the one and same time being "owners" and "executives" of MOEs. However, such an argument is based on the unrealistic assumption that politicians in these positions never are opportunistic, short-sighted, or look for the occasional side-payment that may come along with the assignment (Roberts, 2011). Therefore, and in line with the so called Alphaville-principle (i.e., "hoping for the best, but expecting the worst"), structural risks should be avoided, and building in separation of powers in set-ups involving e.g., MOEs should be an end in itself. As demonstrated in a recent OECD-report (OECD, 2018), evidence indeed shows that publicly owned companies with a higher proportion of independents on their boards, and correspondingly fewer politicians, are associated with lower risks of corruption.
A second potential problem is that, according to the Swedish Companies Act, the boards are legally bound to be loyal with the enterprise they are representing. However, as politicians, there remains a tricky empirical question to what extent they in real life decision-making situations represent their voters, the municipality board, or the interests of the MOEs. As has been highlighted by Bebchuk and Fried (2004), the independence of board members cannot be taken for granted.
While Bebchuck and Fried are occupied with putting the independence of board members in private enterprises into question (2004), there are no substantial reasons to believe that board members of MOEs are immune against moral hazard and bias. Accordingly, there are clear indications that the roles and loyalties of MOEs in Sweden often blurred. To examine the extent and nature of the problem, a survey was conducted, where a representative sample of board members of MOEs were asked: "When you participate in a MOE board meeting, to what degree do you represent the following interests?" The alternatives given were "the citizens;" "the municipality;" "my party;" and "the MOE." Interestingly, 4 of 10 chose "to a high degree" for no less than three or more of those alternatives. This illustrates-or at least seems to illustrate-a low awareness about the potential conflict of interests inherent in being an elected politician that is also on a MOE board; that is, simultaneously representing the owners (i.e., voters and the municipality) and the firm itself. This topic-conflict of interests, and where the loyalties of board members ultimately lay-is highlighted by Sv€ ard (2016), who pointed out that boards of MOEs always run the risk of being dominated by the CEOs, due to asymmetry in business experience (Thomasson, 2009).
Here, it is important to remember that owner-representation in MOE boards emanates from what amounts to a third-order delegation-problem (i.e., from voter to party ! party to council/municipal board ! council/municipal board to MOE boards). Therefore, owners that are represented on boards of a publicly owned corporation are substantially different compared to owners' representation in private enterprises. Local politicians are agents which may have a personal interests and agendas that may or may not be at odds with the interests of the electorate. Hence, while political representation on MOE boards in theory could be a good thing (e.g., by granting politicians direct insight in and control over the inner workings of MOEs), it does, in practice, not provide any immunity against moral hazard or agency loss. On the contrary, having politicians represented both on the local executive and MOE boards, might well increase informal practices in local politics, and, therefore, reduce transparency and worsen conditions for accountability.

326
In this respect, it is curious that one of the original objectives of New Public Management was to create "arm's length" organizations, where implementation was supposed to be organizationally distanced from the policymakers (Hood, 1991). In line with original NPM intentions, a praxis that has come to be globally recognized in order to reduce agency problems in publicly owned enterprises is to mix independent board members from the private sector with representatives from the public sector (OECD, 2013). The idea is to build in separation of powers, while the interest of the owners is still represented. According to the OECD, there is a growing recognition that board members in publicly owned corporations ideally ought to be at an arm's length from executive power. Furthermore, the OECD has suggested that persons directly linked with the executive powers should not be placed on the boards of stateowned enterprises. In addition, OECD recommends that the boards of state-owned enterprises should be composed so that they can exercise objective and independent judgment and be protected from undue and direct political interference. Similar guidelines regarding the independence of board members are also advocated by the World Bank (World Bank, 2014).
The original intentions of the NPM reforms, as well as the guidelines embraced by the OECD and the World Bank, imply that the traditional way of composing boards of MOEs in Sweden could be put into question. As will be demonstrated, board members in Swedish MOEs are often picked from a pool of local councilors, not seldom coming from the ranks of top politicians already represented in the local executives. This is a deviation from the praxis usually followed by state-owned enterprises globally, as well as the aforementioned policy advocated by both the OECD and the World Bank for governance of state-owned enterprises. As indicated, in Sweden, members of the municipal executive committee are commonly represented as board members in the MOEs. Thus, these individuals are both in power as the owners and board members in the MOEs. In contrast, the OECD guidelines for state-owned enterprises advocate clearly defined roles that ensure the independence of the board directors to ensure effective allocation of responsibilities and accountability in the performance of the enterprise (OECD, 2013).
What has been said thus far motivates and brings us closer to the overarching research question that is the article's focus. If the representatives of the voters-leading elites in local governments-appoint themselves and/or fellow councilors as board members, the increasing numbers of MOEs may undermine checks-and-balances as well as separation of powers in local government. A far-reaching overlap between local politicians and MOE board members could, therefore, lower transparency, and hence, impair conditions for accountability. Subsequently, while having politicians represented in the MOE boards could serve an important purpose as an information channel between MOEs and the council/municipal board, having local elite players represented on MOE boards is associated with risks. In the subsequent sections, we will therefore ask whether there are significant overlaps between the representatives of the voters (i.e., local politicians) and the members of the MOE boards.

METHODOLOGICAL APPROACH
Using statistical as well as social network analyses to quantify and visualize various aspects of how MOEs are structured. Social network analysis is a useful way of understanding MUNICIPALLY OWNED ENTERPRISES AS DANGER ZONES FOR CORRUPTION? political connections and power structures (Acemoglu, Johnson, Kermani, Kwak, & Mitton, 2016;Battaglini, Sciabolazza, & Patacchini, 2018;Haselmann, Schoenherr, & Vig, 2018;Kapucu, Hu, & Khosa, 2017). The analyses were conducted on a strategic sample of 11 "medium" to "large" Swedish municipalities. In total, the sample includes 223 MOEs and 732 politicians.
The municipalities were partly chosen based on political-historical criteria, by taking into account if historical power configurations could impact on conditions for accountability. Thus, the municipalities included represent three subcategories: one that historically has been dominated by the Social Democrats; a second which has been dominated by the Moderate party (i.e., the largest center-right party); and a third group for which the political power has shifted frequently. Although the sample is too small to draw general conclusions based on longevity of majorities, the aim was to ensure that different types of power configurations were represented, "time in power for the one and same party" has been viewed as a particular danger-zone for corruption in some of the previous literature (Greene, 2007;Pempel, 1990).
All MOEs for which the municipalities own at least 50% of the votes were included in the sample. The data were primarily collected by closely studying the official websites of the municipalities and by searching the official websites of the MOEs. In order to complement missing data and to get independent validation, alternative sources, such as the firms' annual reports, websites with information on registered companies, and meeting protocols, were also used. (The following websites were frequently used to confirm the data: https://www.allabolag.se; https://www.merinfo.se; https://www.solidinfo.se.) In addition, the MOEs and/or the municipalities were also contacted in order to confirm the data and/or to sort out any ambiguous information.
For each MOE board, the board members name, political party affiliation, and their seats in each company were collected. This allowed the creation of both networks of individual politicians-showing to which other individuals they are connected by sharing the same boards, and firm networks-showing which firms share board members with each other. Only permanent board members were included in the analysis.
The municipalities in this study are presented in Table 2. As is obvious, the selected municipalities are rather heterogeneous. The largest municipality, Stockholm, has a population around 940,000. In contrast, Vellinge, the smallest municipality in the sample, has roughly 35,000 inhabitants. Still, the size of these municipalities should be regarded as medium to large from a Swedish perspective (with the median size of a Swedish municipality being about 15,000). As shown in the third column of Table 1, there is also a large difference in the number of firms where the municipalities own at least 50% of the votes, varying from five firms in Solna to 58 in Gothenburg. The fourth column shows the number of firms per 10,000 inhabitants. A general pattern is that large municipalities typically have few firms in relation to their population. Nevertheless, it is obvious that Piteå and € Osteråker have many firms in relation to their size, even compared to municipalities of their own size.
As mentioned, the sample was partly chosen based on political-historical criteria. The historical power configurations of the municipalities are summarized in Appendix A.1. In Helsingborg, Karlstad, and Stockholm, political power has shifted frequently. Malm€ o, Gothenburg, Piteå, and S€ odert€ alje have been dominated by the Social Democrats, while Solna, Vellinge, and € Osteråker have been dominated by the Moderate party. In Halmstad, power has rarely shifted away from the Social Democrats. Nevertheless, the political dominance in Halmstad was recently shifted to the Moderate party. The Social Democrats ruled Halmstad for 12 straight years between 1994 and 2006, and had ruled Halmstad for most years after universal suffrage was implemented. However, this streak ended in 2006, and a coalition led by the Moderate party has been in power in Halmstad since. Gothenburg is a particularly interesting municipality, being both Sweden's second largest city and historically dominated by the Social Democrats. Gothenburg has also been characterized by a high concentration of political power in a few individuals. In 2010, a large corruption scandal was exposed in Gothenburg, which led to several corruption convictions for civil servants and private CEOs. The concentration of power in Gothenburg has been criticized after these revelations, and measures have been taken in order to reduce the number of seats per politician. Furthermore, Gothenburg has also established a strategy to reduce the number of MOEs, since it has been criticized for its large number of enterprises. Therefore, it is highly relevant to further examine Gothenburg and the structure of its MOEs; by graphing the networks of Gothenburg and other municipalities, comparisons can be made both visually and quantitatively.

RESULTS
Social network analysis is a useful tool for quantifying various aspects of networks, as well as mapping characteristics of each member-in this case each board member or politicianin a network. For an accessible introduction to network analysis for social scientists see e.g., Borgatti, Everett, and Johnson, (2018) and Wasserman and Faust (1994). Some indicators are simple and intuitive. Table 3 shows the overlap of representatives in percent between individuals representing the owners (principals) on the one hand, and those on the boards of MOEs (agents) on the other. As is evident, it is common that municipal council members also are appointed as board members in MOEs. The share varies from 16% (Solna) to 45% Note: a Municipally owned enterprises, defined as active firms in which the municipality owns at least 50% of the votes.

MUNICIPALLY OWNED ENTERPRISES AS DANGER ZONES FOR CORRUPTION?
(Stockholm). The share of members of the municipal executive committee who also are board members of MOEs is even higher. In one of the cases in this sample, Piteå, every single member of the municipal executive committee is also a MOE board member: there is a perfect overlap between "principals" and "agents" in this specific municipality. Table 4 shows the number of individuals with seats on three or more MOE boards. Column 2 shows the number of individuals with seats on at least three boards among all politicians. In general, this type of overlap is not all too common, although Gothenburg is an outlier, with 34 individuals who are on three or more boards.
Column 3 and 4 in Table 4 show the number of representatives from the local councils and municipal executive committees who are on three or more MOE boards. In S€ odert€ alje, nine local council members and five members of the municipal board are on three or more  MOE boards. Both Piteå and Stockholm have several municipal council members and municipal executive committee members who are on three or more MOE boards. Up to this point, it has been shown that overlaps between representatives of the owners on the one hand, and boards of MOEs on the other, are common. Furthermore, it has been shown that several positions are concentrated in the hands of just a few individuals in some of the surveyed municipalities. Differences regarding these aspects may give an indication of the degree of separation of powers and checks and balances, and conversely, concentration of powers in the hands of a few in these municipalities. In order to achieve a clearer and more holistic picture, social network analysis is employed to create graphs for each municipality included in the study. Two networks were created for each municipality: one of individual politicians, showing which other individuals they are connected to via the same board; and one showing which MOEs share board members with other firms. Graphs visualizing these networks are found in the Appendix, and the resulting network statistics are presented in Tables 5 and 6.   Table 6 presents statistics for the individual politician networks and includes the average number of boards seats per board director. For most of the municipalities, this figure is in an interval between 1.20 and 1.70 seats per board member, but there is one particular outlier. In € Osteråker, every member of the municipal executive committee occupies circa three seats on   average, which is a remarkable figure. One explanation for this is the odd fact that the € Osteråker municipality operates a real estate corporate group that includes no less than 10 different companies, with close to identical board member composition in every single one of them.
Graph density, as shown in both Tables 5 and 6, is a highly interesting variable. A high graph density (a value close to 1) indicates that most nodes in the network are interconnected with most of the other nodes in the network. In this context, high graph density should be interpreted as an indication of low degrees of separation of powers. The high graph density in the individual network means that most individuals are interconnected with each other through at least one board, which, in line with the theoretical points of departures, might foster informality and impair conditions for accountability.
The graph density measure does not account for the strength of network connections, but does merely tell us how many interconnections exist. A municipality like Solna, which has comparatively few MOEs, has a relatively high graph density although most board members only are seated on one board. Graph density is strongly negatively correlated with the size of the networks, meaning that large municipalities with many MOEs have lower graph density FIGURE 4 Firms that share board members in Gothenburg (number of firms: 58). Source: Authors' own collection based on public information from the official home pages of the local government and the MOEs, as well as from the sites merinfo.se, allabolag.se and solidinfo.se. Data were collected during the period June-August 2017. Note: Size of nodes determined by degree centrality.

MUNICIPALLY OWNED ENTERPRISES AS DANGER ZONES FOR CORRUPTION?
in general. Even with these limitations, graph density provides interesting information. Among the firm networks, S€ odert€ alje and € Osteråker have a high graph density in comparison to other municipalities with the same number of firms. This is also true for Solna compared to Vellinge. However, Solna has only five firms, so this is likely a result of the "size effect" mentioned above. Solnas networks does not stand out by any other measure. Furthermore, € Osteråker also stands out regarding graph density in its individual network.
Network modularity is an additional dimension of interest with respect to accountability. In a modular network, nodes within a module are highly connected to each other but sparsely connected to nodes in other modules. Network modularity is intuitively understood by using graphical visualization, as shown in Figure 2. In municipal networks with higher modularity, it is easier to identify the area of responsibility for each politician, and thus lower modularity could be argued makes it harder to identify who should be held accountable. Network graphs for all studied municipalities are shown in Figures 3 to 24.
In general, modularity is high-both for politicians and MOEs, but again, there are significant variations between municipalities. In sum, the implications of this article's findings, are, in theory, if the arm's length principle was respected, the massive introduction of MOEs could have potential to enable municipalities to increase separation of powers by populating

342
MOE boards with individuals who otherwise would be hesitant to engage in politics. In practice, however, this is not what the representatives (politicians) of the ultimate stakeholders (citizens) have done. The evidence thus suggests that the group of politicians who create MOEs often appoint themselves and their peers to the boards of the enterprises they own.

FIGURE 22
Firms that share board members in Vellinge (number of firms: 4). Source: Authors' own collection based on public information from the official home pages of the local government and the MOEs, as well as from the sites merinfo.se, allabolag.se and solidinfo.se. Data were collected during the period June-August 2017. Note: Size of nodes determined by degree centrality.

FIGURE 23 Individual politicians' connections in €
Osteråker (number of politicians: 26). Source: Authors' own collection based on public information from the official home pages of the local government and the MOEs, as well as from the sites merinfo.se, allabolag.se and solidinfo.se. Data were collected during the period June-August 2017. Note: Size of nodes determined by degree centrality.

MUNICIPALLY OWNED ENTERPRISES AS DANGER ZONES FOR CORRUPTION?
Hence, it seems evident that MOEs in Sweden are not structured in such a way that transparency and accountability are ensured.
While individual politicians may have sufficiently high personal ethical standards to avoid temptations that can follow from the aforementioned nested principal-agent relations, the current setup is a departure from the original intents of the NPM-reforms-that implementation should be distanced from policymakers. Hence, it is also in violation of World Bank and OECD guidelines for publicly owned enterprises.
In addition, politicians who also are board members seem oftentimes unable to realize that being a MOE board member might put their different loyalties under pressure. As demonstrated above, in the survey that asked, "When you participate in a MOE board meeting, to what degree do you represent the following interests?" The alternatives given were "the citizens;" "the municipality;" "my party;" and "the MOE." Interestingly, 41% of the surveyed board members choose "to a high degree" for three or more of those alternatives. This indicates local politician on the boards of MOEs may be unable, unwilling, or lack the adequate competence, to make tough decisions in situations where conflict of interests arise. The survey was sent to all board members in 30 strategically chosen Swedish municipalities. The response rate was 47%, and a total of 120 individuals answered the survey. The survey also found other illuminating results from the accountability perspective. Asking the primary representatives of the owners (citizens)local councilors, more than 50% believe that transparency is impaired when services are corporatized, and 40% believe it is "unsuitable" that politicians from the municipal executive boards are also members of MOE boards.

FIGURE 24
Firms that share board members in € Osteråker (number of firms: 12). Source: Authors' own collection based on public information from the official home pages of the local government and the MOEs, as well as from the sites merinfo.se, allabolag.se and solidinfo.se. Data were collected during the period June-August 2017. Note: Size of nodes determined by degree centrality.

CONCLUSIONS
Although effects of NPM as well as the movement from "government" to "governance" have been widely debated (e.g., Bellamy & Palumbo, 2010;Pierre, 2000), effects on accountability and democracy have not typically been at the core of this debate. However, a small but growing body of aligned literature has asked whether the downplaying of formal institutions-and a corresponding growth of informal networks-has had a deteriorating impact on transparency and accountability in the public sector (Aars & Fimreite, 2005;Kersbergen & Waarden, 2004;Rhodes, 1997;Sørensen & Torfing, 2005). The aim here has been to contribute to this literature by addressing a specific aspect of this development: the increased willingness of local governments in Sweden to create and operate enterprises of their own.
In order to assess whether the increased willingness of municipalities to create, own, and operate public enterprises (MOEs) affect accountability, we deduced our expectations from the existing debate about corporate boards, gathered from theories on principal-agent theory, and corporate governance in the private sector. This study maintains that if local governments, and its top politicians, choose to appoint top-politicians as board members, MOEs will give rise to layered or nested principal-agent relations. These, in turn, enhance risks associated with moral hazards, simultaneously weakening local checks and balances, diminishing transparency, and, by implication, worsening conditions for accountability. If this is what is observed from data, it would imply that municipally owned enterprises could be particular danger zones for corruption.
What did this article find then? Employing social network analysis, it uncovered dense social networks and potentially worrisome entanglements between representatives of the owners (members of municipal executive committees and councils), on the one hand, and the boards of the MOEs, on the other. As suggested in previous literature, breaching the arm's length principle is disturbing for at least two reasons. First, it invites political interests and meddling into the affairs of the enterprises; second, and perhaps more worrying, it facilitates informality in decision making, hence, weakening conditions for accountability.
The findings provide tentative insights into the mechanisms that may help us understand why municipalities with comparatively many MOEs also tend to experience more problems with corruption (Dahlstr€ om & Sundell, 2014;Statskontoret, 2012). The results are policy relevant, especially when read in light of a heated and ongoing debate in Sweden about the composition of MOE boards-in particular, concerning how the boards ultimately are appointed. Are the motives for appointments political, or are they in the interest of the ultimate stakeholders/principals-the citizens?
A first observation relates to the composition of MOE boards. Everywhere in Sweden where political parties are responsible for electing and appointing boards for political institutions at different levels (parliament, councils, and committees in local and regional government) and boards of state-owned companies, female representation is systematically above 40%. However, on the boards of MOEs, the mean figure is 25%, which is even below female representation found on boards of private enterprises in Sweden. In a previous study (Erlingsson, Fogelgren, Olsson, Thomasson, & € Ohrvall, 2014), by conducting interviews with centrally placed politicians and surveying local councilors, being appointed as a board member was sometimes regarded as a "spoil" or "payback" for long and faithful service to MUNICIPALLY OWNED ENTERPRISES AS DANGER ZONES FOR CORRUPTION? one's party. Being elected and appointed as a MOE board member is considered an honor, and in addition to this, a financial remuneration follows the assignment. Against this backdrop, it came as no surprise that the survey results underlying the aforementioned study showed that the processes leading up to the appointment of boards are viewed as nontransparent, and that a majority of the respondents did not think that appointment-processes guarantee that the best-suited individuals get a place on the MOE boards. That this is considered as a real political problem also from the central government's point of view is underscored by the fact that in 2015, a government commission suggested that it should be made easier for municipalities to appoint board members on other grounds than today's strong and almost exclusive emphasis on party-political affiliation (SOU, 2015).
A second observation relates to the fact that in Sweden, several experts have criticized the composition of the boards from yet another perspective. They claim that members of MOE boards too often do not know what it entails to be a member of a board of an enterprise, and hence, are not always appropriate candidates for the assignment (Andersson, Dansell, Hubendick, & Sen, 2012;Selling, 2012). Others have claimed that it is precisely the incompetence of board members that may explain the high incidence of MOEs occurring in Swedish corruption scandals (Sandler & Forsgård, 2013;Statskontoret, 2012). Commentators have cited a survey that showed that 60% of MOE board members believe they do not have enough knowledge about how one should act as a board member (Sandler, 2012). This analysis has had an explorative character, in the sense that 11 municipalities are few compared to the total number of municipalities in Sweden (290). Still, the high concentration of powers and what seems to be weak conditions for accountability-as quantified, for example, by the low modularity of the MOE networks (in all 11 municipalities), are enough to serve as a warning light regarding the situation in Swedish municipalities. Our findings, therefore, should not only encourage future research to see whether the patterns hold for MOEs in Sweden as a whole, but also to explore whether similar observations are observed in MOEs in other countries as well.

A.2. Politician Ruling and Mandates
In Tables 7-12, the ruling party and mayor have been listed for the municipalities that are studied.