The Welfare State Upholders: Protests Against Cuts in Sickness Benefits in Sweden 2006– 2019

The power resource approach (PRA) claims that the labour movement continues to be the most prominent defender of the welfare state. The new politics thesis (NPT), on the other hand, claims that the welfare state has created new interest groups in the form of welfare clients who have taken over as the most prominent welfare state upholders. In an attempt to empirically evaluate these claims, we present a study of the extent to which clients and the labour move- ment have been involved in protests against cutbacks in the Swedish sickness benefit from 2006 to 2019. The article contributes to the welfare state literature by studying a most likely case for PRA- style interest group mobilization both in terms of country (Sweden) and policy area (sick-ness insurance). It also tests the claim from PRA scholars that client interests are uncommon in these contexts. Our results show that protest engagement among client groups is greater than the engagement among the labour movement when looking at protests directed specifically against cuts in the sickness benefit programme. However, when broader protests against cutbacks in several transfer programmes are taken into account, the number of protests initiated by clients and by the labour movement is comparable. Overall, our results suggest that both the PRA and the NPT are needed to explain current developments in social democratic welfare states like Sweden.


Introduction
In October 2019, people belonging to the Facebook group 'the National Social Insurance Office Insurgency' placed funeral candles outside of several Swedish Social Insurance Agency offices. Through this action, the activists wanted to draw attention to the group of sick people who have committed suicide due to cutbacks that left them without sickness benefits (Dimblad 2019). In 2008, the trade union Kommunal (representing people hired by municipalities) held frequent public protest meetings in the city of Nyköping. They offered passers-by a cup of soup while also agitating against cutbacks in the sickness benefit programme (Ljungkvist 2009). These are not isolated events; since 2008, when the sickness benefit was reformed in Sweden, hundreds of protest actions have been initiated to stop proposed or implemented cuts -both by people in need of sickness benefits and by organizations belonging to the labour movement.
These protests are not only interesting from a societal point of view given their size and frequency, they are also particularly interesting in relation to the welfare state literature. A central disagreement within this literature is the debate between the supporters of the power resource approach (PRA) and the new politics thesis (NPT). The PRA states that a strong labour movement is a precondition for a strong welfare state and that the labour movement continues to be the main political force defending the welfare state today (Korpi & Palme 2003). The NPT, on the other hand, argues that the welfare state in itself -through its provision of welfare benefits -has created new interest groups, namely clients, who are eager to fight to keep the welfare state intact (Pierson 1994). Many studies have been published that examine which theory better explains our empirical reality with a focus on political parties. However, the interest group aspect of retrenchment politics has largely been neglected by welfare state scholars (Starke 2020).
In this article, the claims put forward by the PRA and the NPT concerning interest group feedback will be investigated. This investigation will be conducted through an examination of the extent to which clients and the labour movement have been involved in the protests against cutbacks in the sickness benefit in Sweden between 2006 and 2019. The study is based on printed media material. Our main focus is on the ability of the two groups to mobilize protests rather than their direct political influence, due to the well-known difficulty of measuring interest group influence. 1 The article contributes to the welfare state literature by studying a most likely case for PRA-style interest group mobilization both in terms of country and policy area. The Swedish labour movement is still one of the strongest in the world, even though there are signs of a general weakening of its bargaining power (Palm 2017;Kjellberg 2018). Adding to this, the labour movement should -according to the PRA -first and foremost defend the transfer programmes that contribute to the decommodification of the labour force, that is, unemployment insurance and the sickness benefit. Previous research on welfare retrenchment in transfer systems suggests that in line with the PRA, the labour movement still constitutes an important defender of these programmes (see Bandau & Ahrens 2019 for a literature review). Some scholars even argue that in social democratic welfare states such as Sweden, client organizations are 'weak' and 'invisible' (Scarbrough 2000;Anderson 2001). A clear limitation of previous research, however, is that apart from Feltenius' (2007) work on pensioners' organizations in 3 Sweden, there is a lack of studies in which client engagement in the defence of transfer programmes, such as the sickness insurance, is directly investigated. This lack of research is particularly clear in regard to studies that have investigated less conventional acts of political participation, such as protest actions. There is therefore a clear need to study interest group feedback in this policy area. If clients are found to be actively defending a transfer programme in Sweden, which could be seen as a most likely case for the PRA, then the NPT is clearly being given more support than before.

Literature Review
The PRA formed the most prominent strand of literature within welfare state research during the 1980s. According to the PRA, cross-national variations in social provision can largely be attributed to differences in the distribution of life-course risks and power resources among social classes such as workers and employers. The PRA argues that individuals who have a higher number of life-course risks (e.g., illness, work accidents, unemployment and poverty) tend to have a lower number of individual resources available to cope with those risks. Those who are relatively disadvantaged (i.e., the working class) will therefore organize together as a labour movement to demand economic support in the form of social security to be able to cope with their higher number of life-course risks (Korpi 1981). Their efforts will result in what Gøsta Esping Andersen (1990, 25-26) refers to as the 'social democratic welfare state' . Power resources and social class are, according to the PRA proponents, central to understanding both the expansion and the retrenchment of the welfare state (Korpi & Palme 2003). This implies that the strength of the labour movement continues to be the most important explanatory factor in determining the persistence of the welfare state in the longer term.
In his seminal book, Dismantling the Welfare State, Paul Pierson (Pierson 1994) criticizes the PRA and argues that the welfare state is resilient regardless of the strength of the labour movement. Despite the threats of cutbacks, the welfare state has remained remarkably intact through years of increasing austerity measures. The NPT, which Pierson introduces, is built on the following three mechanisms that are arranged in a feedback loop: (i) electoral punishment, (ii) interest group mobilization, and (iii) blame avoidance (Pierson 1994). The politicians currently holding power are, according to Pierson, eager to become re-elected. Due to this desire, they will hesitate to implement cutbacks, as such cutbacks would result in concentrated costs in return for diffuse benefits for voters. A central claim of the NPT is that the welfare state in itself has created policy feedback effects in the form of constituencies who are now dependent on the services that the welfare state has been providing to them. These new interest groups, formed by welfare 4 Scandinavian Political Studies, Vol. 0 -No. 0, 2021 clients, will thus fight for the persistence of their social benefits. Furthermore, as people who experience losses are more likely to let such losses affect their voting decisions, these new interest groups can easily mobilize voters to punish politicians for their retrenchment decisions. To overcome this dilemma, politicians must employ blame avoidance strategies (e.g., implementing less transparent cutbacks) to avoid electoral punishment.
Most of the literature following the ideas put forth in Dismantling the Welfare State has focused on the aspects of the theory that address electoral punishment and blame avoidance, while the interest group mechanism has largely been neglected by scholars (Starke 2020). This is surprising given that interest groups, such as labour unions and client groups, constitute central elements of both the PRA and the NPT, where the two theories are not compatible. This research gap is a result of the general bias in contemporary political science towards election, parties and voters rather than organizational behaviour (cf. Hojnacki et al. 2012;Hacker & Pierson 2014). There exists, however, a small number of studies that have empirically tested the claims of the PRA and the NPT regarding which actors constitute the prime defenders of the welfare state. Most of these studies investigate the relationship between the strength of social democratic parties (and sometimes labour unions) and welfare retrenchment (see Bandau & Ahrens 2019 for a recent literature review). Their focus is primarily on the transfer component of the welfare state: sickness insurance, unemployment insurance and pension. The general consensus in this literature is that 'labour still matters' in regard to these programmes (cf. Bandau & Ahrens 2019). Some of the studies even argue that examples of 'mobilized opposition' among welfare recipients are difficult to find at all in regard to corporatist countries with a strong labour movement (e.g., Scarbrough 2000;Anderson 2001). The reason for this is that 'they have been co-opted into the apparatus of the labour movement' (Anderson 2001). This claim is, however, challenged by Feltenius (2007) in his study of Swedish pensioners' organizations and by Larsson Taghizadeh and Lindbom (2014) in their study of hospital closures (see also Larsson Taghizadeh (2016) for similar results on school closures). These studies show that clients are able to mobilize protests and are capable of stopping welfare retrenchment in these sectors, even in social democratic welfare states. The latter study, however, acknowledges that since it investigated welfare services (healthcare) and not a transfer programme, their case cannot be seen as 'critical' for the PRA (Larsson Taghizadeh & Lindbom 2014).
Hence, the currently available evidence does not seem to support Pierson's claims regarding the importance of client interests versus organized labour, at least not in regard to the transfer component of social democratic welfare states. However, a clear limitation of the retrenchment literature overall is the bias towards institutional forms of political activity, such as the 5 involvement of political parties and bargaining with trade union organizations (cf. Larsson Taghizadeh & Lindbom 2014;Bailey 2015). Conventional political activities, such as being a member of a trade union or a political party, are now on the decline in Western countries (Scarrow 2000), while less conventional forms of political participation, such as participating in public protests and in voluntary advocacy groups, are on the rise (Norris 2002;Peterson 2015). Newer research suggests that such protest activities could even have a more significant impact on the process of welfare reform than that of the institutional forms of contention that have been more frequently studied (cf. Bailey 2015;Bailey & Shibata 2017). Against this background, it is surprising that no studies have, to the best of our knowledge, compared the extent to which client groups and the labour movement have mobilized less conventional protest actions against cutbacks in transfer systems.
This article therefore contributes to the welfare state literature by investigating protest actions against cutbacks in Swedish sickness insurance. If protests initiated by clients are clearly less common than labour movement protests, then the PRA is given stronger support. If the results do not point clearly in either direction, the theories are considered to be complementary. Finally, if the protests initiated by clients/client organizations are clearly more common than the protests initiated by the labour movement, then the NPT is given stronger support, even in the least likely case of Sweden. 2 The Case of Swedish Sickness Insurance 2006-2019 First and foremost, Sweden was chosen for the current case study because it can be seen as a possible most likely case for PRA-style interest group mobilization and a least likely case for the NPT. As we previously argued, Pierson's argument that class-based parties have been replaced by client organizations as 'the primary defenders of the welfare state' has been challenged, especially in regard to countries with a strong labour movement. Sweden is the home country of the PRA, and the strength of its labour movement is still evident today. The high trade union density and the historically strong Social Democratic Party have allowed for the development of one of the most extensive welfare states in the world: the social democratic welfare state (cf. Esping Andersen 1990). However, it also needs to be acknowledged that both the Swedish trade unions and the Social Democratic Party have seen decreasing membership rates during recent decades (Kölln 2015;Kjellberg 2018). Moreover, there are signs that trade unions are undergoing change, leading to a general weakening of their bargaining power (Palm 2017). This does not change the fact that when looking at union density, which is a measurement commonly used to compare labour movement strength between countries, Sweden is still one of the top-ranked countries (Kjellberg 2018). Granted, there are always strong reasons to be cautious 6 Scandinavian Political Studies, Vol. 0 -No. 0, 2021 when generalizing results from case studies, as important differences in political systems, political cultures and such are evident. For example, the high levels of citizen participation in political elections and in civil society organizations in Sweden could both result in a stronger labour movement and stronger client organizations than in other countries. However, we still think that the status of Sweden as a possible most likely case for PRA could potentially increase our possibility of generalizing our results to other industrialized countries.
Adding to the above information, the choice to focus on cutbacks in the sickness benefit system further improves the possibilities of seeing the chosen case as a most likely case for PRA-style interest group mobilization, as the labour movement should first and foremost be defending the transfer programmes that are contributing to the decommodification of the labour force against cutbacks. This implies that the sickness benefit should be one of the labour movement's core concerns. One argument against choosing the sickness insurance programme is that illness/diseases can be considered life-course risks rather than class-based ditto, leading to less class-based mobilization (cf. Busemeyer et al. 2009;Jensen 2011). However, according to the PRA, resistance against cutbacks in the sickness insurance should still follow a class-based logic as the lower socio-economic classes face a higher risk of being sick and also have less economic resources to cope with these risks compared to the upper-income classes (cf. Korpi 1981). In line with this, metastudies show a strong relationship between socio-economic status and health as well as the probability of getting serious diseases (see Adler & Ostrove 1999).
The time period of 2006-2019 was chosen because of the number of implemented cuts in the sickness benefit during this time. It has also been indicated in newer published and unpublished research that there was an upsurge of protests against welfare retrenchment during the specific time period (Uba 2019). In 2006, the centre-right coalition (named Alliansen) took office, replacing the Social Democratic Party who had governed from 1994 to 2006. Proposing and implementing cutbacks in the laws regulating the sickness benefit were core political issues for Alliansen. The most substantial and long-lasting changes occurred in 2007 and 2008, when new rules regarding the level and duration of the benefit were implemented. These changes included a lowering of the daily allowance by 3 percent and a limitation of how many days one was eligible for sickness benefits (before this policy implementation, there was no absolute limit). Moreover, a new rehabilitation chain was implemented, meaning that one's ability to work in another position in the workplace needed to be evaluated after 90 days, and one's ability to work at any job available in the labour market needed to be evaluated after 180 days (Försäkringskassan 2019). After the 2014 election, the Social Democrat Party took office again. They removed the absolute 7 time limit, and they slightly raised the income ceiling; otherwise, things remained largely the same. In addition to focusing on a time period during which we observed the largest implemented cuts during the last decades, the chosen time period allows for the comparison of the protest intensity during a right-leaning and a left-leaning government. It is likely that the labour movement will protest less against a left-leaning government compared to a right-leaning government, as they will hesitate to protest against themselves. Studying protests that occur during time periods with right-leaning governments and left-leaning governments therefore increases the possibility of generalizing our results over time.

Methods
This article aims to capture to what extent client groups and the labour movement use public protests to express their opinions against welfare retrenchment. This examination is conducted through a study of protests against proposed or implemented cutbacks made to the sickness benefit based on printed media. The method used in the article could be described as a congruence analysis (cf. Blatter & Haveland 2012). Hence, we evaluate the explanatory relevance of the two theoretical approaches by (1) deducing sets of observable implications from them (cf. page 6), (2) comparing empirical observations with the theoretical implications (in our case, counting protests initiated by clients vs the labour movement) and (3) drawing conclusions regarding the explanatory power of the theories based on the observations.

Operational Definitions
Broadly defined, acts of social protest express 'a grievance, a conviction of wrong or injustice' that the protesters are 'unable to correct/…/directly by their own efforts' . This study follows the operational definition outlined above by Peterson (2015, 508). Hence, we coded a range of protests ranging from conventional acts to disruptive and violent ones, including petitions, legal demonstrations, strikes, sit-ins, occupations and violation of property or the creation of other kinds of economic or personal damage (Peterson 2015, 508). The protests that were coded are summarized in Table 1. Verbal protests, such as argumentative texts/letters published in newspapers, were not considered to be a protest activity, as 'the general procedure among protest scholars is not to count verbal protests when trying to capture overall protest activity' (Uba 2019, 8).
There is currently no definition of welfare state retrenchment that has become universal among welfare state scholars. Thus, we used the definition proposed by Starke (2008, 13) that such retrenchment is 'a political decision 8 Scandinavian Political Studies, Vol. 0 -No. 0, 2021 to reduce the level of social protection guaranteed by the state' . This is equivalent to what Pierson calls 'programmatic retrenchment' (Pierson 1994, 15;Starke 2020, 3). According to Peter Starke, policymakers typically have three kinds of retrenchment instruments at their disposal when implementing programmatic retrenchment: (i) reducing benefit levels, (ii) reducing the duration of the benefit or (iii) tightening eligibility by adding more conditions or changing the qualification age for the benefits (Starke 2020, 3). These three kinds of programmatic welfare cutbacks form the operational definition of programmatic welfare retrenchment that we relied on. Hence, our study aimed to include all the protests targeted against these three types of cutbacks in sickness insurance. We also included protests that were, in more general terms, expressed as grievances against 'cutbacks in the sickness benefit' that can be seen as a mixture between the three kinds of programmatic welfare retrenchment above.
We counted both protests that were specifically targeted against cutbacks in sickness insurance and cases where retrenchment in the sickness benefit programme was only one of the grievances put forward (e.g., demonstrations targeting both retrenchment in the sickness benefit programme and in the unemployment benefit programme). In our results, we separated these two types of actions, as the latter was expected to exert less political pressure on decision makers according to the NPT. The demands from protests focusing only on retrenchment in the sickness benefit programme were expected to be specific and clear, while the protests against several kinds of welfare retrenchment were expected to have a longer list of demands, and their protests were consequently expected to be more diffuse. As argued by Pierson (1996, 145), the idea that 'concentrated interests will be in a stronger political position than diffuse ones', which is a 'standard position in political science', is also valid in regard to protests against programmatic welfare retrenchment. Protests that were not clearly aimed at the benefit levels, the duration of the benefit, the eligibility of the benefit in sickness insurance or a combination of these criteria were not included in the dataset. For instance, the pensioners' organization protests against the impact of high tax burden on their retirement income were not included, even though they also mentioned the high tax burden on sickness benefits, as the protests were explicitly directed against differences in tax rates and not welfare retrenchment. These criteria for the inclusion of protests are summarized in Table 2. At least one of the criteria needed to be met for inclusion in the current study.
We operationalized labour movement protests as protests involving the Swedish Trade Union Confederation (Landsorganisationen, LO), LOaffiliated trade unions, the Social Democratic Party and its suborganizations (such as the Federation of Social Democratic Women) or several of these actors. In Sweden, the Social Democratic Party and the LO have strong ties; together, they tend to be described as constituting the Swedish labour movement. Client protests were in turn operationalized as protests organized by recipients of sickness benefits and/or people who claim their right to sickness benefits and/or people who solidarize themselves with the above mentioned individuals in their role as welfare clients (cf. Pierson 1994, 30). To clearly separate labour movement protests from client protests, client protests were defined as engagements that occur outside of organizations, which are created primarily for other purposes (i.e., outside of the labour movement organizations, other political parties, church organizations, etc.) Client protests can either be completely informal, organized by a single actual or potential welfare recipient or formally organized by client organizations. Our operational definitions of labour movement and client protests are summarized in Table 3.

The Process of Collecting and Coding Information about Protest Activities
We followed the standard procedure among scholars who try to capture protest activities (Ortiz et al. 2005) and used newspaper articles to identify protests. More specifically, we used the Swedish Media Archive's Retriever Research platform. The advantage of using this method is that we had a good chance to cover most actions that were highlighted by the Swedish media. Actions not covered by the media were considered less likely to influence public opinion and politicians. Furthermore, the replicability of this method should be seen as fairly high.
The archive is the largest digital news archive in the Nordic countries, and it includes a large amount of both printed articles and web articles published since the 1980s (Mediearkivet Retriever 2020). The search string that was used is roughly, when translated to English: (sickness insurance* OR sickness benefits*) AND (protest* OR manifestation* OR demonstration* OR appeal* OR revolt* OR petition* OR strike* or occupation*). 3 The search was restricted to Swedish printed newspaper material, and it resulted in a total of 9 163 unique articles that needed to be carefully scanned to identify those that provide information on protest events against cuts in the sickness benefit programme. The choice of using printed newspaper articles, instead of using articles published online or a combination of both, was made due to technical difficulties with the Retriever online material. Specifically, quite a few of the online articles were not fully available through the Retriever; only the title and approximately two lines of text were visible, and the links to the full articles were often too old to function. When the search string was entered into the archive, we started by going through the articles year by year to sort out those that present information about protest events related to programmatic retrenchment in the sickness benefit. This scanning was performed by reading the title of the article and the part of the article where the keywords are identified. Based on this information and a comparison with the definitions outlined above in Tables 1 and 2, a decision was made either to read further and code the event as a part of the dataset or to discard the article and move on to the next one. To test intercoder reliability, both authors coded 5 percent of the articles (460 in total) independently. While the degree of intercoder reliability was respectable according to well-cited guidelines, we have chosen to formulate our conclusions carefully so that they take our margin of error into account. 4 When coding the protests, our aim was to come as close as possible to an accurate representation of both the number and the character of the protest events. 5 All the protest types listed in Tables 1 and 2 were included in the dataset. The unit of analysis in our dataset consisted of individual events taking place in a well-defined geographic area at a specific point of time. For example, when we encountered information that protests were organized in several cities on the same day and were initiated by the same organization, Organized by recipients of sickness benefits and/or people who claim their right to sickness benefits and/or people who solidarize themselves with the above mentioned individuals in their role as welfare clients. These engagements occur outside of organizations created primarily for other purposes (i.e., outside of the labour movement organizations, other political parties, church organizations, etc.) 11 all the events were coded as individual observations. The full dataset consisted of 487 individual protest observations ranging from 1 May 2006 to 15 November 2019. The majority of the identified protests were demonstrations, public meetings or manifestations (90 percent). Approximately 10 percent of the identified protests were petitions; furthermore, five sit-ins were observed. The collected data did not include any protest strikes or violent protests. The reasons for protesting were almost exclusively regarding worsening conditions in the sickness benefit in general (97 percent); only 3 percent of the protests were clearly directed against a specific kind of retrenchment (cuts in the benefit level, the duration of the benefit or a tightening of the eligibility criteria). Half of the protests were directed against retrenchment in the sickness benefit and against other welfare programmes or services (such as the unemployment benefit). We made sure that all the protests were against cutbacks in the sickness insurance by only searching for and including protests related to 'sickness insurance' (sjukförsäkring) and 'sickness benefit' (sjukpenning). These words are in Sweden synonymous with the state sickness insurance programme that covers loss of income due to illness (rather than due to healthcare costs). Protests regarding the quality of healthcare or cutbacks in the healthcare services were rare in our dataset, and protests only targeted against such cutbacks were not included at all (as our focus is on the sickness insurance and not on healthcare services). When coding the events, one aim was to show what organization or group of people initiated the protest event. The dataset included both events where it was only clear that there was an organization behind the protests and events where the individuals who initiated the protests were represented by name and by what association they belonged to. In both cases, the protests were coded according to the organization or party that was mentioned in the article. When coding the protest events, the initiators were coded into six categories: clients (31 percent), unions (53 percent), 6 the Social Democratic Party (23 percent), the Left Party (5 percent), churchrelated organizations (13 percent) and others (3 percent) (some events were initiated by several of these initiators). In the results, however, only the protest events that involved clients and labour movement organizations (unions and the Social Democratic Party) are presented, following the definitions in Table 3 (404 of 487 observations). Moreover, 15 protest events involving cooperation between clients and the labour movement, as well as other civil society actors, were excluded (3 percent of all observations), 7 as the overall aim of the article is to compare the engagement of clients to that of the labour movement. Most of these collaborations were part of the so-called 'Easter uprising', which means that we cannot rule out that it was a one-off event. While there could be a media bias against mentioning all 12 Scandinavian Political Studies, Vol. 0 -No. 0, 2021 the different organizations/groups that are involved in a protest, cases of cooperation between clients and the labour movement appear to be rather uncommon.
When using newspaper material, there is a need to acknowledge two kinds of selection biases (following Ortiz et al. 2005). First, there is the selective process of choosing what events need to be reported on by journalists, which among other things is determined by the intensity of the event, by the familiarity/significance of the organization(s) and contextual factors (i.e., whether the chosen protest location is considered to be important). This type of selection bias is unfortunately outside of our control. A likely outcome is that smaller, less significant protests will be omitted in our results, and they will be omitted unevenly; during attention cycles, it is likely that smaller protest events will be reported on instead of ignored. This type of bias is most likely in favour of the labour movement and the PRA, rather than in favour of the NPT. Compared to labour movement protests, client protests are not as well-known and have a more informal character. Furthermore, labour movement organizations often have employees who work on capturing the media's attention in contrast to client groups, who are loosely organized and largely dependent on volunteer work.
The second selection bias is performed by researchers as they decide what qualifies for inclusion in the dataset and what does not; for example, the tendency among scholars to rely on just one or a couple of newspapers. We tried to limit this type of bias by using a media archive platform that covers all nationwide newspapers, as well as a vast amount of the local newspapers printed throughout the country. Obviously, our data collection will not result in a perfect reflection of all protest events against cuts in the sickness benefit that have occurred in Sweden from 2006 to 2019. Rather, the dataset should be seen as an approximation of how the protest events have unfolded, and the results should be able to tell us about the overall trends in protest engagement.

Comparing the Engagement among Clients and the Labour Movement
As shown in Figure 1, where we compare the total amount of client protests and labour protests directed only against retrenchment in the sickness benefit, it is evident that client groups have been engaging themselves to a much higher extent than has the labour movement. Only 41 protest events specifically directed against retrenchment in the sickness benefit have been initiated by the labour movement, while 138 protest events of the same type have been initiated by client groups. Hence, when analysing only the 13 protests that were organized against sickness insurance, the results clearly confirm the theoretical expectations from the NPT (Lindbom 2007) that client groups themselves diverge from the labour movement in an important sense and that their interests are 'specific enough for their priorities to be clear' .
However, when looking at all the protest events in the dataset, including the protests directed against both the cutbacks in the sickness insurance and other kinds of welfare retrenchment such as cutbacks in the unemployment insurance, the picture changes. As shown in Figure 2, the number of protests initiated by the labour movement increases from 41 to 166, while the number of client protests only increases by one. Hence, when looking at both specific and diffuse protests, the results provide slightly stronger support for the PRA than the NPT. Notably, all of the 125 diffuse protests initiated by the labour movement were targeting both cutbacks in the sickness insurance and cutbacks in the unemployment insurance.
However, taking statistical significance as well as bias in the media material into account, the results shown in Figure 2 can also be interpreted as supporting both theories to a similar degree. A paired t-test shows that the difference in the proportion of protests initiated is not statistically 14 Scandinavian Political Studies, Vol. 0 -No. 0, 2021 significant below the 10 percent level (p = .1223). As discussed in the methods section, several factors also indicate that there is a bias in newspaper coverage towards reporting about protests organized by established actors, that is, labour-initiated protests, at the expense of those initiated by organizations that are unknown to the readers, that is, client groups. Moreover, in the process of collecting data, 25 protest events were identified for which the newspaper article stated that there had been protests in several cities, but it was not specified how many protests occurred and therefore several individual protest events could not be registered in the dataset. Nineteen of these 25 protest events were client initiated. Thus, we not only suspect a bias, but we know for sure that the results are biased; the protests initiated by clients were underreported in the media compared to the protests initiated by the labour movement. Therefore, if we had access to all the protest events that occurred during the time period of interest, it is possible that the 27 observation difference would disappear, which would then demonstrate equal support for the NPT and the PRA. 8 Note that Labour Day demonstrations were excluded from the dataset as these occur every year regardless of proposed or implemented cutbacks in the sickness benefit (and are hence not driven by these cutbacks). 9

Comparing the Engagement among Clients and the Labour Movement over Time
As seen in Figure 3, there is also a clear difference in how the protests from the labour movement and client groups are distributed over time. The vast amount of the labour-initiated protests occurred in 2007-2009, while the client groups were more active during the 2010-2012 period, with a clear peak in engagement in 2011. Interestingly, the right-leaning government was forced to make several concessions regarding the level and duration of the sickness benefit and the rehabilitation chain in both these periods, which suggests the possibility that the protests may have been influential to some degree (cf. Dagens Arena 2012; Försäkringskassan 2019).
Most of the protests by the labour movement took place in direct connection to the cutbacks in the sickness insurance that the right-leaning government headed by Alliansen implemented during their first term from 2006 to 2010. More specifically, on 1 July 2008, they decided to implement a limitation in the maximum duration of the benefit and a new rehabilitation chain, meaning that one's ability to work needed to be evaluated after a specific number of days (highlighted as the 'main decision' in Figures 3 and 4, were also made to unemployment insurance which could explain why most of the labour movement protests focused on cuts made in both programs. While it is difficult to know which cutbacks were most important for the protests, their timing (the majority of the protests taking place in 2008 10 ) suggests that their main focus was on the cutbacks in the sickness insurance. The largest cutbacks in the unemployment benefit were implemented already during the first half of 2007 when the government increased the requirements, lowered the benefit level and limited the maximum duration of the benefit.
Twelve of the labour-initiated protests directed specifically against cuts in the sickness insurance and 16 of their diffuse protests were organized in the 3 months before 4 June 2008, which was when the law was voted through in the Swedish parliament. This event is highlighted as the 'main decision' in Figures 3 and 4. Thus, part of their protest mobilization back in 2008 was deliberately trying to prevent the proposed legislative changes.
Client mobilization, on the other hand, started to increase in 2010 and peaked in 2011, which may imply that their political action is largely motivated by the consequences of retrenchment rather than the threat of retrenchment, 17 as no major legislative changes resulting in cutbacks in the sickness benefit programme were implemented during the 2010-2012 period. Most of the protests took place after the first clients met the time limit and lost their sickness benefit at the turn of the year 2009/2010 (highlighted in Figures 3  and 4). The peak of the demonstrations initiated by clients occurred the same year as the 'Easter Uprising' (highlighted in Figures 3 and 4), where numerous groups protested against the implemented cutbacks in the sickness insurance that had at this point in time resulted in thousands of citizens losing their insurance. However, the vast amount of client-initiated protests did not occur directly as a part of the Easter Uprising -only three of the 66 client-initiated protests held in 2011 were part of this specific protest wave. Furthermore, it is evident in the material that several client groups were already formed at the beginning of the investigated study period. The time that it takes to build client organization networks and the like could therefore be a reason for why the client protest peak occurs later than the labour movement peak. A clear upsurge in client engagement can also be observed in 2019. These protests were again not a consequence of legislative proposals; rather, the overall time trend indicates that clients seem to sporadically engage themselves to a larger extent than the labour movement does. One possible explanation for the lack of engagement from the labour movement during this time period could be the change to a social democratic-led government. However, as there are few observations overall after the client peak in 2011, more data from a period with a social democratic government are needed to further test this idea.

Conclusion
The NPT claims that the welfare state has created new interest groups in the form of welfare clients who have taken over from the labour movement the position of the most prominent welfare state upholders. However, this claim has been called into question, especially in regard to the transfer component of social democratic welfare states such as Sweden. In an attempt to determine whether client groups or the labour movement are the more important 'defenders of the welfare state', we studied the extent to which the labour movement and client-initiated protests against cuts were made to the sickness benefit programme in Sweden between 2006 and 2019.
Our results are mixed. When considering the protests in which only grievances against cuts in the sickness benefit programme were put forward, our results show that client groups are clearly engaging themselves to a greater extent than the labour movement. However, if all protests are counted, including protests targeted both against cutbacks in the sickness insurance and several other types of welfare retrenchment (e.g., in the unemployment insurance), the labour movement is shown to be engaged to a slightly greater 18 Scandinavian Political Studies, Vol. 0 -No. 0, 2021 extent than are client groups. Potential biases in the newspaper material and intercoder reliability suggest that we should be careful to give the credence to PRA over NPT based on the total number of reported actions alone. The chosen material is not only suspected but also shown to be biased in favour of the labour movement. In the dataset, it is evident that when protest events have been organized in numerous cities during the same day, there is a bias towards reporting the exact number of locations in regard to labour movement organizations. These events, which occur in several cities but have unknown locations or numbers, have not been coded due to the lack of information provided. It is also likely that labour movement organizations, who are well-established formal actors, are covered by the media to a greater extent than small rather informal client groups and networks. For instance, the data material only covers seven internet-based protests; six of these protests are client initiated. This might suggest that client groups are protesting online to a higher extent than the labour movement and consequently that their protests are omitted in the media to a higher extent than are the labour movement protests.
A limitation of our study was that we were not able to directly study the political influence of the different protests mobilized by the labour movement and client groups, nor was we able to study inside lobbying activities. 11 Therefore, we cannot draw any safe conclusions in this regard. However, allowing ourselves to speculate, there are both theories and empirical research that support the notion that protests targeted at specific cutbacks could indeed be more likely to stop retrenchment proposals (cf. Larsson Taghizadeh & Lindbom 2014). A central assumption in the NPT is that concentrated interests will be in a stronger political position than diffuse ones (Pierson 1994). One of the reasons for this is that their members' interests are specific enough for their priorities to be clear (Lindbom 2007). On the other hand, our analysis of the distribution of protests over time suggests that most of the client protests took place after the cutbacks were implemented, which could have reduced their potential influence. Furthermore, trade unions should have opportunities for inside lobbying that client groups do not have (especially in traditionally corporatist countries like Sweden). It should be said, however, that the cutbacks were implemented by a rightwing government, reducing the potential political influence of labour movement lobbying.
The focus of our study was, however, on comparing protest mobilization. Perhaps the safest conclusion that can be drawn from the results here is that both the labour movement and client groups seem to play an important role in protesting against cuts made to the transfer component of the welfare state covering life-course risks. The fact that client groups managed to mobilize a comparable number of protests to that of the labour movement against cutbacks in general and that they organized a majority of the 19 protests directly targeted at the cutbacks in sickness insurance clearly questions the claims from some supporters of the PRA that client interests are generally 'invisible' and 'uncommon' in social democratic welfare states (cf. Scarbrough 2000;Anderson 2001). Based on the limited evidence available as well as our own data, 12 this claim may very well be accurate when it comes to transfer programmes covering risks that are strongly linked to social class or are administered by the labour movement such as the unemployment insurance (cf. Anderson 2001;Larsson Taghizadeh & Lindbom 2014). Protests organized by the unemployed outside the labour unions do seem to be weak and uncommon in social democratic welfare states such as Sweden (cf. Lahusen 2013). However, the claim from the PRA scholars clearly does not seem to be accurate when it comes to larger programmes that to a greater extent cover life-course risks such as the sickness insurance or pension where client interests appear to be important (cf. Feltenius 2007).
From this perspective, our results support the newer more pragmatic line in welfare state research that both the PRA and the NPT are needed to explain current developments in Western welfare states (cf. Gingrich 2015). Both theories are needed, even in regard to explaining mobilizations against cutbacks in the social insurance system in Sweden, which is the home country of the PRA. As client groups are shown to be significant defenders of one of the labour movement's core concerns in Sweden, it is possible that such groups also constitute important actors in other countries as well. Most studies on interest organizations avoid the question of political influence because it is so difficult to answer (Dür 2008, 560;Hojnacki et al. 2012, 9.7).

2.
Formulating testable expectations leads, inevitably, to sacrificing parts of the theoretical complexity. However, as this article aims to conclude which one of the theories that best describes the chosen case, simplification is necessary and the expectations from the theories need to clearly separated from each other. 3.
The actual search string used, in Swedish, is (sjukförsäkring* OR sjukpenning*) AND (protest* OR manifestation* OR demonstr* OR upprop* OR uppror* OR namnlist* OR namninsamling* OR namnunderskrift* OR underskrift* OR protestlist* OR strejk* OR ockup*). The asterisks are used to make sure that all suffixes will show up.

4.
Across all 460 articles reviewed by both authors (most of them not containing protests), the coding of the protests and their initiators matched in 98.9 percent of the articles.
As for the small number of articles where any of the authors found protests, the coding matched in 70.6 percent of the articles. For percentage agreement approaches, there is no universally accepted threshold for what percent agreement that indicates acceptable reliability. According to the often-cited guidelines by Landis and Koch (1977), values between 61 and 80 percent should be seen as indicating substantial agreement.
Neuendorf (2002), on the other hand, argues that there is considerable disagreement among researchers about any values below 80 percent. 5.
The full list of variables, including coding guidelines and mean values, is available in the Appendix. Eight of the protests involving collaborations were client initiated, three were initiated by both clients and the labour movement, two were supported by both groups and two were initiated by the labour movement. 8.
The same is true if we count the actions mobilized by the Swedish church as client protests. 9.
For more information, see the Appendix. 10.
Thirty-nine diffuse protests were initiated by the labour movement in 2007, 79 in 2008, eight in 2009 and one in 2011. 11.
The use of newspaper data also resulted in a lack of information regarding the number of participants in the vast majority of the coded protest events. If more information regarding the number of participants were available, clearer conclusions regarding the actual political importance of the protests could have been drawn. 12.
None of the client protests against the cutbacks in the sickness insurance in our dataset was larger than the labour movement petition against the cutbacks in the unemployment insurance, which was signed by as many as 240,000 people (2.6 percent of the population) (cf. Larsson Taghizadeh & Lindbom 2014). Furthermore, our data also show that client groups were generally less likely to initiate and participate in protests that were both targeted against the cutbacks in the sickness insurance and the unemployment insurance. Why: Sickness Benefits in general 1 when the protest is concerned with cuts in the sickness benefits in general, but it is not specified if it is the benefit level, the duration of the benefit or changing eligibility criteria that is the main concern. Otherwise, 0 The only numerical variable in the dataset, missing data on 377 rows. As this variable both measures the number of signatures on a petition, and participants in a demonstration, the number is high due to two instances of extreme values (105,000 and 65,000).