The new privatized market: A question of ideology or pragmatism within the Swedish addiction treatment system?

Given its traditions of universal welfarism and social democracy, Sweden had already scored unexpectedly high on New Public Management by the 1980s. Health and welfare services remain primarily tax ...

correlated with larger shares of purchasing and purchases from for-profit providers. However, purchasing on a market dominated by for-profit providers has also become the "new normal", regardless of ideology, and recent years have shown a reversed effect of left-wing municipalities purchasing more services than right-wing governments. Pragmatic reasons also influence local-level purchasing. Since the late 1980s, health and welfare systems in many Western countries have undergone dramatic changes regarding modes of production and means of organization and steering. An umbrella term for many of these changes is New Public Management (NPM). As Hood (1995) shows, NPM has been implemented with various motives and different emphases in different countries. NPM typically relies on the effectiveness of market-driven management models also in the public sector, and on control of the output of public services, rather than of processes.
The spread of NPM is often attributed to the advent of new right governments in the 1980s and particularly to the influence of Ronald Reagan and Margaret Thatcher, who sought to roll back state welfarism. NPM reforms were first introduced in English-speaking countries such as the United States, the United Kingdom, Canada, Australia, and New Zealand.
However, Sweden has also been placed into the core of NPM countries (Common, 1998;Hood, 1991Hood, , 1995James & Manning, 1996;Pollitt, 1995). Hood (1995) called Sweden "the most obvious misfit" of NPM seen as an ideologybased reform. NPM is at first glance incompatible with a social democratic regime, with big government and a universalistic welfare state. Research has ascribed the impact of market models in Sweden not only to a strengthened role of the political right, but also to the development within Swedish social democracy itself. In the late 1980s and early 1990s, the Swedish welfare state was criticized for being too bureaucratic, centralized, and paternalistic. From the 1980s, the right, in line with the Anglo-American trends, propagated market models and privatization in production as a new welfare vision to shrink the scope and size of the public welfare state. The social democrats, who lost the 1976 elections and came back in 1982, regarded the shrinking legitimacy of the welfare system as their core political problem. Reform seemed necessary. For the social democrats, municipal steering, with a separation of decisions about allocations and needs assessment from the production of services (purchasing-provider models), was an administrative reform to promote service diversity, citizens' choice, lower costs, and better quality (Stenius, 1999). Already in 1984, the party program proposed a welfare system built on decentralized decision making and freedom of choice for participating citizens (still emphasizing universal rights to welfare) (Gingrich, 2011;Klitgaard, 2007). The social democrats also introduced management by results into Swedish administration (Sundström, 2006). The development was further advanced with different market-type welfare reforms, also promoted by the social democrats (Green-Pedersen, 2002).
Sweden had a large, and thus expensive, public sector and a strong central steering of comprehensive services.
As Hood (1995) points out, it had both economic motives and opportunities to effectively implement these new administrative models. A public procurement legislation was introduced in 1992 (in force as of 1994), later revised by the Swedish Public Procurement Act (LOU 2007:1091and LOU 2016:1145, modeled on EU regulations. The law stipulates that all public purchases above threshold levels be subject to a non-discriminative procurement process. Health care and social services are given more leeway, but for some reason, possibly judicial cautiousness, Sweden has implemented public procurement more rigidly than other European countries also in these sectors (Ahlberg & Bruun, 2010). In the social service sector, public procurement has often resulted in "framework agreements" between suppliers and the contracting authorities. In 2008, the possibility was introduced to steer the service system through service users' choice (Act on System of Choice in the Public Sector, 2008:962 [LOV]). This is, however, a voluntary option for the municipalities, and has so far had little impact on the addiction treatment system (Storbjörk & Samuelsson, 2018). From the 1990s and increasingly during the 2000s, purchaser-provider models, target and performance management, and economization have grown increasingly common in Swedish health and welfare services (Bertling, 2015;Hartman, 2011;Sallnäs & Wiklund, 2018;Zaremba, 2013b).
Competition with an increasing role for non-public provision of services-marketization-is a core feature of NPM (Hood, 1995), and is the focus in this article. Non-public service providers have had a strong position in Sweden since the establishment of the alcohol treatment system in the early 1900s (Edman, 2016;Stenius, 1999). After the Second World War, not-for-profit treatment providers were the dominant providers of care. They gradually became only supplements to a growing municipal and county council public service sector in the heyday of the Swedish welfare state. In the late 1980s the trend changed drastically. Purchaser-provider models, procurement with competition between providers, private sector management styles, and decentralized budget responsibility were introduced in addiction treatment. It resulted in new steering models, but also in a growing number of non-public providers, now often incorporated enterprises (Oscarsson, 2000;Sallnäs & Wiklund, 2018;Stenius, 1999). The production of services no longer took place within a corporatist arrangement, but moved onto a competitive quasi-market. 1 Critique (from the left) has repeatedly been raised against the development. Thus the "privatization of the century" in 1980s addiction treatment was viewed as an attack against the universalism of the public sector-dominant welfare system (SoS-rapport, 1991). But there were also leftist advocates of market models. After the economic crises in the 1990s, the benefits were emphasized of cost-efficient public steering with competition and increased user involvement (Edman, 2016;Oscarsson, 2000;Stenius, 1999). Consequently, market steering models for addiction treatment in the 1990s were already also established in social democratic local governments (Edman, 2016;Stenius, 1999).
The last few years have seen the rise of a highly polarized public debate-with partly new arguments-on the role of for-profit providers in the Swedish welfare system. The vast majority of all health and social services in Sweden remain tax-funded. However, compared to the 1980s or 1990s, market models no longer appear as neutral roads to better quality and lower costs. Whereas the total costs for all social services offered by the municipalities increased by 14% between 2009 and 2013, the costs for such care produced by private enterprises increased by 57% during the same period (Statistics Sweden, 2015, p. 9).
Criticism, again primarily from the left, has pointed to opportunist and revenue-seeking big, multinational, providers entering the welfare sector, and that this has had a negative effect on the quality of services for citizens in a weak position; there have been no demonstrated cost savings, but rather a draining of tax money (Agerberg, 2014;Government, 2016;Löfven, 2013;SOU, 2002;Zaremba, 2013a). In 2016, an inquiry set up by the Social Democratic-Green government (SOU, 2016) noted the increasing share of for-profit providers in the 2000s. While most of these providers are small, some are multinational, and some have a very high profit level. The report, supported by social democrats, and the green and left parties, suggested a profit level ceiling (7%) for private, for-profit enterprises in the publicly financed service sector as an attempt to limit the impact of certain for-profit actors. In the discussion following the report, the right-wing and liberal parties defended for-profit provision of care with similar arguments as in the 1990s: private alternatives promote efficiency in service production, improve the quality, staff commitment, and choice for patients (Stenius & Storbjörk, 2003). The attitudes towards private, for-profit provision again seem to differ along ideological lines. Within economy and in "leading-edge" NPM countries, scholars and officials are now talking about the death of this management model (Dunleavy, Margetts, Bastow, & Tinkler, 2006). This can partly be linked to the critique of a new type of privatization and the lack of demonstrated improved economy or outcomes (Hood & Dixon, 2015;Pollitt & Sorin, 2011). In general, NPM reforms have had a substantial impact on the welfare systems, but are yet poorly studied. The few studies conducted internationally on performance measurement in addiction treatment indicate some improved process outcomes such as shorter waiting lists or larger treatment volumes (Hull & Ritter, 2014;McLellan, Kemp, Brooks, & Carise, 2008). But the new market models have also led to unintended negative consequences such as increased fragmentation of treatment systems and less treatment for the most severe cases (Bjerge, 2012;Moore & Fraser, 2013;Nesvåg & Lie, 2010). The impact on the Swedish addiction treatment system has not been thoroughly researched (Creutzer, 2014;Wiklund, 2011). Sallnäs and Wiklund (2018) note a clear increase of for-profit actors during the 2010s, and a recent study on NPM from the service users' point of view indicates improved choice for a small group of substance users, but also limited professional discretion and service user involvement (Storbjörk & Samuelsson, 2018). Given the lack of effect studies, decision makers are not well-equipped to choose adequate steering models.
In the decentralized Swedish addiction treatment system, the main responsibility for organizing, funding, and providing services, which are based on local needs assessments, is located in the 290 municipalities' social services administrations. In this, the municipalities are assisted by 21 county councils (responsible for medical addiction treatment). All of them have independent political decision-making bodies and independent rights to raise and distribute taxes, and to organize services as they wish (e.g., purchase or deliver in-house). However, it is only the municipal social services that have the right to exercise authority in individual treatment decisions, which implies that some activities cannot be contracted out (see Blomqvist, Palm, & Storbjörk, 2009 on the overall Swedish addiction treatment system). The municipalities have varying political majorities elected every four years. They are also of varying size, with different and more or less predictable needs for addiction treatment (see Table 2).
Analysis of the municipalities' purchases of addiction treatment services, as opposed to in-house service production, can tell us the extent to which the use of private as opposed to in-house addiction services over time has been linked to ideological characteristics within decision-making bodies-if the recurring leftist critique in the national discussion is mirrored at the local level. Or are other conditions more decisive in a pragmatic choice of service production, factors such as the size of the community (small municipalities not being able to produce their own treatment, see SOU, 2011), the affluence of its inhabitants, or a general societal norm enforced by legislation on procurement and central steering? This small and relatively delineated sector of the welfare system (e.g., compared to schools) has previously been neglected in this respect. Addiction treatment has, as noted, historically had a large share of non-public service providers and a tradition of municipal purchasing of primarily residential services. It is also a fairly marginal means-tested welfare sector that is not of obvious interest for the middle class as service users. In both these respects it is similar to child/youth care. But even with its limitations, addiction treatment holds an important position in the Swedish debate on the distribution of common goods to deserving or undeserving citizens-on the practice of universalism. Based on the decentralized care system and a large number of local-level political majorities operating under similar conditions in one country, our study adds to the knowledge of the role of party politics in the marketization of public services.

| Aims
This article examines the degree of purchased interventions, particularly for-profit provision, in this specialized part of the welfare system since the 1970s and with a special emphasis on 1999-2014. We have charted and analyzed national and local developments, and more specifically, we analyze: (1) the national development of purchases of addiction services (size in SEK and share of purchased in-and outpatient services from different types of provider) relative to the total expenses (all in-house and purchased activities); (2) whether political majorities (left or right-wing) have an impact on these developments at local levels; and (3) whether: (i) population size; or (ii) local-level economy has an impact on local-level purchases, i.e., whether pragmatism rules over ideology.

| DATA AND ANALYSES
We used Swedish official statistics to compile all providers of institutional care run as nationally approved Homes for Care and Housing (Hem för vård eller boende [HVB]). Our data series cover 1976,1984,1990,1997,2003, and 2016.
The first four years have been investigated in previous work by Stenius (1999, pp. 92-93, 113) and were complemented for 2003 by Storbjörk for a national inquiry (SOU, 2004). We applied a manual reading of annual catalogs listing and describing the institutions. For the purpose of this article, corresponding information was extracted in March 2016 from the national HVB register that now holds the data. A manual categorization sought to replicate previous analyses by singling out units treating alcohol problems. For 1976-97, units were included if they claimed to treat alcohol misusers. There were few drug units in 1976. Since the 1980s, alcohol and drug treatment have more or less merged into one integrated addiction treatment system (Bergmark, 1998). The categorization for 2003 and 2016 excluded units claiming that they only treated problems with illicit drugs or pharmaceuticals. In 2016, only two units were excluded because of a sole focus on illicit drugs. Across the time series, units were excluded if they had fewer than five beds, provided family care, were shelters and other pure housing facilities, and institutions for youth (up to the age of 25). We counted the number of institutions and beds by type of provider. Our HVB series cover 40 years of institutional care for adults.
In March 2016, we ordered data from Statistics Sweden, which holds information on all municipalities' annual accounts (Statistics Sweden, 2015. The statistics run from 1998, and we used data on each municipality's (social service) total expenditure and purchases of addiction treatment for adults from various types of providers in 1999, 2004, 2009, and 2014. As of 2004, it was possible to distinguish between purchased institutional and outpatient care. We analyze the overall expenditure (measured and presented as Swedish krona, SEK) and costs for purchasing care (instead of producing and delivering it in-house), in total, by type of provider, and by activity and year. We report amounts (SEK) and percentage of purchased care of the overall costs. The data was analyzed in current prices, but Figure 2, reporting on the development over time, used the 2014 monetary value.
Data on local government-whether the municipalities had left-wing and/or social democratic governance; rightwing governance; or coalition governance-following the elections in 1994-2014 were collected from the Swedish Association of Local Authorities and Regions (SALAR). Excel files were downloaded from the SALAR website (SALAR, 2016) and complemented by email correspondence. We focus on the difference between municipalities with left-wing and right-wing governance across the entire period. Although governed by politicians, social services administrations are dominated by treatment professionals and civil servants, and some agreements may be in force across elections. Therefore, organizations change slowly, and stable ideologies appear the most interesting for our study. Additionally, initial analyses of the effects of single elections prior to the studied years showed few correlations. Lastly, we used official data provided online by Statistics Sweden on population size and median income (2014 monetary value) in the adult population of each municipality for each studied year. The municipal divisions have changed slightly over the years. We used the division in force in 2016.
We charted developments in figures and frequency tables, and used multiple regression models to analyze the extent to which local-level purchases of care could be attributed to local government (ideology), population size, wealth, and the total budget for addiction care, in 1999, 2004, 2009, and 2014 when controlling for all other variables. This is a total population study including all Swedish municipalities, but significant levels are indicated for the convenience of the reader (few correlations are significant due to the low number of municipalities).  The public sector dominated in 1984 with 66% of the beds provided by municipalities (and some by county councils). The 1990s represent an equal distribution across providers, but also an increase in enterprises. The (incorporated) enterprises thereafter increased their share of beds from 20% in 1990 to 62% in 2016. Public providers had already stepped back in the 1990s, whereas the drop in beds provided by NGOs is most profound in more recent years. Important factors that may explain this shift are the conversion of state subsidies from earmarked money to the addiction treatment institutions (based on the number of beds) to general subsidies to the municipalities for purchasing different welfare services in the late 1980s and early 1990s (Edman, 2016;Oscarsson, 2000); 2 the enforcement of public procurement law as of 1994; and the pronounced shift from long-lasting social democratic government to a center-right political alliance in the 2006 general election (see Figure 4) government, while critical of the role of for-profit provision, is a minority government, and has a history of ambivalence towards market-type reforms.

| National-level expenditure and purchases of care (HVB) in 1999-2014
By summing the total expenses across all municipalities, we see that the local administrations spent SEK 5.9 billion on addiction treatment in 2014 (Figure 2). The grey lines show a 56% increase in total municipal expenditure between 1999 and 2014. Saving goals do not seem to have been met. Outpatient care costs increased by 84%, whereas the costs of institutional care increased by 30%. In 1999, 62% of the care was purchased (black lines), and this ratio stabilized at around 40% between 2004 and 2014, i.e., an increasing in-house production in the early period with a left-wing/social democratic national government. Around 60% of the institutional care and slightly less than 30% of the outpatient care was purchased in 2004 and 2014. A deviation is the year 2009, when 67% of the institutional care and 19% of the outpatient care were purchased. Overall, the proportion of purchased care relative to the overall expenses (in-house care included) has been quite stable since 2004, but with a greater increase in purchased outpatient care, measured in SEK.
Based on the money spent on purchasing care from different types of providers, Figure 3 displays the distribution across providers in total and by type of care (institutional vs. outpatient). Private enterprises have dominated the purchased care production over the entire study period, increasing their share from 46% in 1999 to 64% in 2014. The most recent increase is attributed to a growing share of enterprises in purchased outpatient care, the sector that also increased in terms of money spent in the most recent years in Figure 2. At the same time, NGOs decreased their share from 17% to 12%, and households-family care and contact persons-almost disappeared. A closer analysis (not shown) revealed that smaller municipalities still in 2014 purchased significantly more from these very small, non-profit private care units compared to more populated municipalities that basically did not purchase such services at all. The state remains an important provider of institutional care due to its monopoly on expensive compulsory care regulated in a special law that compels municipalities to initiate, and thereby purchase, coercive care if the

| The role of ideology and pragmatics in local-level purchases of care in 1999-2014
3.2.1 | Some effect of ideology, but purchasing is the "new normal" Sweden had a social democratic national government in 1982-91. Figure 4 charts the subsequent national political power bases: a (black arrow) center-right-wing government for one three-year term of office, followed by (grey arrow) social democratic rule that changed on both national and local (bars) levels in 2006 when an alliance of center-right-wing parties gained power for two terms. As of 2014, Sweden has been governed by a social democratic-green minority coalition.
Since 1999, the proportion of municipalities that do not purchase any care (≤5% of the treatment was purchased) has decreased from 12% to 3%, and the decreasing standard deviations indicate greater similarity across municipalities over time (Table 1). The most drastic drops are found in 1999-2004 and in 2009-2014. The middle period appears stable. A redistribution (not shown in the table) from purchased institutional to purchased outpatient care also took place: the share of municipalities not purchasing institutional care (≤5%) increased from 10% to 17% in 2004-14, whereas those not purchasing outpatient care decreased from 50% to 39%. This redistribution was most prominent between 2009 and 2014.
It is first and foremost municipalities with a tradition of left-wing and coalition governments that do not purchase care. Right-wing municipalities generally have high percentages of purchased care (relative to the overall costs; medians) across the study period. In 2014, all municipalities, including left-wing municipalities, present higher average percentages of purchased care than during previous years. The shift was quite remarkable between 2009 and 2014, following a period with bourgeois national government, revised procurement law (LOU), and a new system of choice law (LOV). On the other hand, the differences grew again in shares of purchased care between the left-and rightwing municipalities: bourgeois municipalities purchased more once again. Smaller municipalities more often report that they do not purchase any services. The standard deviations are smaller for larger municipalities, indicating greater similarity in levels of purchasing in more populated municipalities. The dispersion is also smaller across the richest municipalities. Better-off municipalities bought more services in 1999, but the pattern was reversed in 2014, when the poorest municipalities reported the highest average purchase levels. In terms of total expenditure on addiction care per inhabitant, those municipalities that spent the least reported higher levels of non-purchase. This pattern remained stable across the time series. Table 2 used linear multiple regression to study the correlations between municipal characteristics and local purchase of care (0-100% of total expenditure was purchased relative to the total net costs) in total for all addiction Municipalities with greater overall expenses for addiction services purchased less treatment than the municipalities spending the least in 1999, also when controlling for population size and wealth, but the tendency was reversed thereafter. The pattern is mixed for purchases of institutional care (Models B) that fluctuates more on the local level depending on needs. Right-wing municipalities purchased more institutional care than did left-wing municipalities in 2004 and 2014, richer municipalities purchased less than the poorest municipalities, and municipalities spending more on addiction services per inhabitant also purchased more institutional care than did those spending the least. Models C, lastly, shows that right-wing local governments in 1999, 2004, and 2009 bought more from enterprises than left-wing municipalities, while in 2014 left-wing municipalities bought more than did bourgeois municipalities.
Due to highly skewed distributions, we used multiple logistic regression (instead of multiple linear regression) to analyze the associations between municipal characteristics and purchasing of outpatient services (Table 3, Models D). Purchasing more than 5% of the outpatient care was coded as 1 (purchasing ≤5% coded 0). Right-wing local governments were more inclined than left-wing governments to purchase outpatient services in 2004; there was no difference between the two in 2009; and in 2014 the condition was reversed with right-wing municipalities having lower odds than left-wing municipalities to purchase outpatient services. Larger municipalities, populationwise, would more often purchase outpatient care over the entire time period. This was most pronounced in 2014. Richer municipalities had higher odds of purchasing outpatient care than their less well-off counterparts.
In 2004 and 2009, municipalities that spent the most on addiction services also purchased more outpatient care as compared to those spending the least. In 2014, it was those spending the least that were more inclined to purchase outpatient care.

| DISCUSSION
This article set out to study what the Swedish addiction treatment system-a market-looks like in terms of division of labor between different types of providers, development over time, and whether one could distinguish any effects on purchasing services as opposed to producing treatment within the public administration that could be linked to ideological differences between municipalities with left and right ideologies, or if other, more pragmatic factors were at play. Our analysis elaborates on statistical associations. In this discussion we therefore outline some possible explanations to be tested in future studies.
Within Swedish institutional addiction treatment (HVB), where we have the longest statistics series, NGOs produced most of the treatment in the 1970s, while in 1984 municipalities had become the largest producer group. In 1990, NGOs and municipalities had about equal shares, but a third provider type appeared: the for-profit incorporated enterprises. During the social democrat government (1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006) 1999, 2004, 2009, and 2014; percentage not purchasing (bought less than 5%), and central tendencies for percentage of purchased care of total costs for addiction services Note. a n refers to 2014. a new economic actor-strong in terms of offering diverse interventions and handling procurement processes-has demonstrated its interest in addiction treatment (see also Sallnäs & Wiklund, 2018). The growing size of the institutions and the decreasing number of providers may indicate that bigger enterprises are buying smaller ones (SOU, 2016). The declared goal in the 1990s was to increase producer diversity (Edman, 2016). The recent development could be a sign of concentration in service production and thus an opposite development.
The total national net costs for Swedish municipal addiction treatment, both outpatient and inpatient care, has increased substantially between 1999 and 2014 (56%; Figure 2). The cost increase is steady over time with no visible impact of changes of government ideologies. The increase is most pronounced in outpatient care (85%, as compared to 30% for institutional care; Figure 2). The total costs for purchased care was high in 1999 (relative to the total net costs), dropped in 2004, and the share of purchased services has since remained stable, around and just above 40%. To summarize: nationally, the total turnover has increased, and particularly for outpatient treatment. The share of services purchased appears stable, but the increased total expenditure implies that more money is spent also on purchased care; the market is growing. In addition, the distribution of producers within purchased care has changed. Forprofit providers are now the dominant provider group that municipalities buy from. It is not possible to distinguish any breaches in the general growth of the market, nor in the growing share of for-profit providers. However, there are some signs that the market has recently entered a new phase with an increasing role for for-profit providers in outpatient treatment (a market that is also increasing in financial turnover), new kinds of private providers associated with clear profit interests, a shrinking role for NGOs and, as a consequence, a new, ideological debate.
Research has showed how procurement and marketization have challenged the previous corporative and municipal social work logic (Alvehus & Andersson, 2017;Forkby & Höjer, 2008;Freidson, 2001;Meagher & Szebehely, 2013;Sallnäs & Wiklund, 2018). While social workers purchasing addiction services often stress previous experiences, established collaboration with providers, and clients' wishes, LOU stresses the non-discrimination of producers and tender-based ranking order (Storbjörk & Samuelsson, 2018;Storbjörk & Stenius, in press). The Norwegian government has chosen to protect the NGO engagement in community matters and the tradition of NGO-public sector cooperation by preventing incorporated enterprises from entering procurement processes and addiction treatment markets (Bogen & Backer Grønningsaeter, 2016). We believe that the increasingly centralized and formalized procurement practices in Sweden at least partly explain why we see bigger players on the addiction treatment arena while NGOs have difficulties managing in this new landscape (Storbjörk & Stenius, in press). All in all, we cautiously hypothesize that a "New Privatized Market" is emerging in Swedish addiction treatment.
Moving to the local-level analyses that proved valuable for understanding the role of ideology in developments over time, we note that today only a very small minority of municipalities adhere solely to in-house services. Purchasing addiction care has become the norm, in right-wing as well as left-wing municipalities, while it was less so in the late 1990s. One explanation could be the somewhat slow but firm implementation of public procurement legislations leading to greater conformity. The difference in purchasing behavior between small and big municipalities has also diminished over time. It is possible that almost mandatory procurement protocols, guided by overarching EU directives, weaken the role of local ideology in the organization of care.
Our analysis shows that during the social democrat national government (1999 and 2004), the gap in purchasing behaviors between left-wing and right-wing municipalities disappeared. The fact that the left started buying more implies that purchasing services became normalized rather than being an ideological question (Table 1). On the other hand, between 2004 and 2014, with a right-wing government and at a time when the market also slightly changed, we see again a widening gap. The right-wing municipalities today buy more than the left-wing ones (confirmed in the multiple regression analyses of Table 2). The picture becomes more complicated when we move on to purchases of out-and inpatient treatment. Right-wing municipalities bought more outpatient treatment and less inpatient care than the left-wing municipalities 15 years ago, and previously bought more from private enterprises, whereas today the ratio is reversed. We can only speculate about the reasons: one hypothesis is that a bourgeois ideology was more decisive for purchasing care before public procurements had become the "new normal", with left-wing local governments just lagging behind.
Bigger and wealthier municipalities vary less in their purchasing behavior than smaller and poorer municipalities.
(Note: there was no relation between size and political majority.) The impact of size seems feasible in that a small municipality has less possibility to produce treatment, and the demand for treatment can be more difficult to foresee.
Bigger municipalities buy in general relatively less treatment, and produce more themselves, regardless of political majority (SOU, 2011). However, towards the end of the study period, if one controls for political majority, they actually buy more institutional and outpatient treatment. The pattern is most clear for outpatient care. One hypothesis for the latter, in line with our emerging market thesis above, is that private providers establish themselves where they can find sufficient local markets (Tables 2 and 3). Sallnäs and Wiklund (2018) explain the expansion of for-profit providers in outpatient care by the fact that this field is not strictly controlled by supervisory bodies. It is thus easy to establish such units.
Wealthier municipalities also buy consistently less inpatient and more outpatient services than their less wealthy counterparts. Addiction problems may be less of an issue in wealthier areas, given that the treatment population in Sweden is highly marginalized (Storbjörk & Room, 2008). It could also reflect that the overall resources are rationally steered towards outpatient care in a population with less need for intensive interventions. Notably, budgetary issues combined with the questioning of the effectiveness of institutional care also lead many municipalities to implement internal guidelines prioritizing outpatient care over institutional treatment (Storbjörk & Samuelsson, 2018).

| CONCLUSIONS
The share of public provision of addiction services which can be locally decided has remained fairly stable during the last 15-20 years. However, the surrounding landscape has changed. Procurement law entails that who municipalities buy from is primarily determined by market forces. Competent bidders are advantaged. Incorporated enterprises have gained an increasing role among those providing purchased care, also within the expanding outpatient treatment. New kinds of multinational enterprises have entered the market. The role of NGOs has diminished. This points to a new type of privatized production, different from that of the 1980s.
Which has been more important for this development: ideology or pragmatism? It is clear that very small and very big municipalities, and poor and rich ones, have somewhat different purchasing behaviors. Differences can be explained by pragmatic reasons, such as the possibility to plan for unexpected new needs, the feasibility of launching public municipal production for a local market, the possibility to have long-standing relations with non-public providers, and the attractiveness of a big local market for private providers.
On the role of ideology, our data seem to confirm the picture presented in earlier research. The left-wing governments or municipalities do not consistently present any clear alternative in their purchasing behavior. The acceptance of arguments favoring market incentives from the left was also mirrored in a rapid privatization of a large addiction treatment clinic in Stockholm in the early 2000s (Stenius & Storbjörk, 2003). Our analysis shows that in 1994-2006 the share of purchased care in left-wing municipalities grew and reached the level of right-wing municipalities. Purchases did not seem to be a very ideological question. Thereafter, when the government again became bourgeois, and the role of for-profit providers reached a higher level, the purchasing behavior diversified once more and became relatively more important in right-wing municipalities than in left-wing ones. But the difference is not dramatic. To some extent, it looks as though the left has only slowed down the privatization development, not stopped it. Hood (1995) argues that NPM as party politics may suit both left-and right-wing sides, with arguments that can attract the floating middle-ground voters. Addiction treatment, of small interest to middle-class voters, has sometimes been caught up in reforms for larger welfare areas such as elder care. Also, non-public provision has a very long tradition in Swedish addiction treatment. All these conditions, the legislative changes and EU directives, and previous ambivalence within the social democratic party impede their and their allies' chances to develop feasible alternatives to market models. The path of the present government is to try to regulate this market. The recent proposal (SOU, 2016) to cut profit returns marks an attempt to eliminate outright speculation in the welfare market.

| Limitations
The study is limited to services provided by the municipal social services, which hold the major responsibility for addiction treatment. The role of the regional-level health care system, responsible for medical treatment, has increased somewhat over the years, but the annual accounts available for the regional county councils are limited to "specialized psychiatric care" and do not allow us to distinguish dependence care from the much broader field of general psychiatry.