Do Participatory Democratic Institutions Engender Civic Engagement? Quasi-Experimental Evidence Björn Tyrefors Hinnerich1, Per Pettersson-Lidbom2 and Kåre Vernby3 ***Preliminary. Please do not quote or circulate without permission from the authors*** 1 Department of Economics, Stockholm University, E-mail: [email protected] Department of Economics, Stockholm University, E-mail: [email protected] 3 Department of Government, Uppsala University, E-mail:[email protected] 2 Abstract Do participatory democratic institutions stimulate civic engagement? Or will representative institutions do a better job? We shed new light on this long-standing controversy using data that are uniquely suitable to deal with the potential endogeneity inherent in studying these questions. During the early twentieth century, the Swedish Local Government Act required municipalities with a population above a certain threshold to be governed by representative democracy, whereas municipalities with a population below that threshold were free to choose assembly democracy instead. This allows us to credibly estimate the effect of forms of democracy on two forms of civic engagement: turnout and voluntary associations. Our findings show that participatory democratic institutions did not stimulate civic engagement. Instead, turnout was considerably higher and associational life thrived where local governments were governed by representative democracy. Most contemporary observers see civic engagement as a key element in a well-functioning democracy. The reasons for embracing high levels of citizen participation in political and societal affairs vary depending on scholars’ normative persuasions, but they can essentially be boiled down to the following three distinct propositions: participation makes policy more responsive to the interests of citizens, it furthers the development of citizens’ social and political capacities and it confers legitimacy to public policies (Teorell 2006). Unsurprisingly, therefore, many decry the low levels of civic engagement that characterize modern representative democracies (see, e.g., Barber 1984, Lijphart 1997, Skocpol 1999, Putnam 2000). Since these ills are often perceived to be inherent to representative democracy, one proposed cure has been to entrust participatory institutions with the making of public policy (see, e.g., Barber 1984, Bryan 1999). The movement for direct involvement in policy making has not been confined to the established democracies in Europe and North America, it has also made significant advances globally (Fung and Ohlin Wright 2003). At the same time, there exists a widespread unease, even among some who are highly critical of representative democracy, that participatory institutions will fail to sustain high levels of civic engagement (Mansbridge 1980, Berry et al. 1993, Plotke 1997, Nylen 2002). Despite of the centrality of the issue, there exist relatively few large-scale systematic studies comparing the impact of various forms of direct democracy on civic engagement. Moreover, the findings of those studies that do exist are often difficult to reconcile (see, e.g., Bryan 1999, Zimmerman 1999, Boehmke 2002, Tolbert et al. 2003, Lowry 2005, Freitag 2006, Ladner and Fiechter 2012). In this paper, we present unique quasi-experimental evidence that shed new light on this controversy. After Sweden democratized, during a period of thirty years, municipalities, which had large responsibilities in important policy domains, were variously governed either by assembly or by representative democracy. The suffrage requirements were identical for both types of democracies, but whereas partisan assemblies that were elected every four years governed in representative democracies, assembly democracies were governed by town meetings that were required to meet at least three times a year. To measure civic engagement in this historical context, we have complied data on turnout in elections and town meetings, and on the main voluntary associations. What is most notable, for our purpose, is that the Swedish Local Government Act stipulated that municipalities with 1500 or less inhabitants were allowed to choose between assembly or representative democracy, whereas municipalities with more than 1500 inhabitants were forced to adopt the latter (Tyrefors Hinnerich and Pettersson-Lidbom 2012). We thus obtain an exogenous source of variation in the form of democracy, which can be used to estimate the impact of representative and participatory institutions on civic engagement. Specifically, we can implement a fuzzy regression-discontinuity design (RD) using data on the type of democracy employed in Swedish local governments between 1919 and 1939, and data on political participation, which we define broadly to include turnout in meetings and elections and membership in associations such as unions, temperance lodges and free churches. The fuzzy RD shares the weak identifying assumptions of the well-known sharp regression-discontinuity designs (see, e.g., Bloom 1984, Battistin and Rettore 2000, Hahn et al. 2001 and Lee and Lemieux 2010). In essence, it utilizes the fact that whether the number of inhabitants are a few less or a few more than the threshold above which municipalities are forced to govern themselves by representative democracy can be considered as good as random. We also perform additional analyses using a rich set of controls capturing the political, economic and demographic situation in the municipalities. To the best of our knowledge, our paper is also the first to provide quasi-experimental evidence on the impact of representative and participatory institutions on participation in political and societal affairs. It is therefore advantageous over previous studies for two reasons. First, it ameliorates the problem of reverse causation. Indeed, it is highly plausible that the level of civic engagement in a community will affect the choice of democratic institutions. If this is true, any study attempting the to estimate the impact of institutions on participation without employing a quasi-experimental design would generate severely biased results. Second, our study minimizes the risk of omitted variables bias. In particular, both levels of participation and institutional choices may be outgrowths of more deep-seated and unobservable local characteristics. Under such circumstance, it will be extremely difficult to estimate the impact of participatory institutions on political participation without a quasiexperimental research design. Our findings show that participatory democratic institutions did not lead to more civic engagement. In fact, all our results point in the opposite direction. While voter turnout was relatively high and associational life thrived where local governments were governed by representative democracy, the opposite was true in those municipalities that were governed by assembly democracy. In particular, when comparing municipalities around the population threshold, the results suggest that turnout was somewhere between 20 and 30 percentage points higher in representative democracies. Moreover, the results suggest that whereas associations thrived in representative democracies, they were poorly mobilized in direct democracies. Further analysis shows that this holds for both free churches, temperance associations and labor unions, but that the effects are the largest and most precisely estimated for the latter. The paper is divided into four parts. In the !rst, we describe the previous literature on the impact of participatory and representative democratic institutions on political participation. In the second, we describe the Swedish local government setting during the period we study and place it in context. Our data and research designs are presented in the third and our results in the fourth. In the !fth, we set out our conclusions. The Impact of Participatory and Representative Democratic Institutions on Civic Engagement During the 19th and early 20th century, many contemporary observers feared the ongoing process of democratization and the wide participation in politics and organizational life that was expected to follow in its wake. In modern times, by contrast, most agree that a politically and organizationally active citizenry is at the core of a well-functioning democracy. While there is some disagreement over the direction, magnitude and scope of global trends in the various forms of civic engagement (e.g. Stolle and Hooghe 2004), many scholars decry the low levels of participation in political and societal affairs that characterize modern representative democracies (e.g. Barber 1984, Lijphart 1997, Skocpol 1999, Putnam 2000). What is most interesting from the perspective of this paper is that many perceive low civic engagement as inherent to representative political institutions – some even going as far as claiming that “[r]epresentation kills participation” (Barber 1984, xxxiv) – and that allowing citizens to decide public policy directly would stimulate greater participation in political and societal affairs. Is it true that participatory democratic institutions engender civic engagement? In what ways would the choice between participatory and representative democratic institutions affect civic engagement? Analogously to Riker and Ordershook’s (1968) classic model of the calculus of voting, we may posit a more general calculus of participation, in which a citizen participates if, and only if, the benefit of the implementation of the citizen’s preferred times the probability of affecting the collective outcome exceeds the cost of participation (see, also, Nagel 1987, 26). Along these lines, Berry et al. (1993, 72-73) suggest that participatory institutions may “offer citizens the chance to significantly influence the decisions that affect the quality of life in their communities”, such that these benefits outweigh the “costs in time and energy”, leading more citizens to participate. Echoing this sentiment, Barber (1984, 272) claims that “people refuse to participate where politics do not count – or counts less than rival forms of private activity”. Fung and Olin Wright (2003, 27), too, argue that participatory political institutions seek to “sustain high levels of lay engagement” by offering “the real prospect of exercising state power.” If the policy benefits of participation are higher where participatory institutions are in place, this will lead to more civic engagement in the form of greater citizen political participation and an increased proliferation of voluntary associations. There are, however, a number of arguments against the view that participatory institutions will further civic engagement. Skeptics of participatory political institutions maintain that they increase the costs of participation to prohibitive levels, and that they do not imply that a broad portion of the population will have the potential to influence public policy. Beginning with the cost side, it seems clear that direct citizen participation in policy-making is considerably more demanding and time-consuming and that many see this as one of the main drawbacks of participatory institutions (see, e.g., Berry et al. 1993, Plotke 1997). As Mansbridge (1980, 70-71) writes in her seminal study of town hall meetings in New England, assembly democracy is not only time-consuming, but also generates “fears of making a fool of oneself, of losing control, of criticism, and of making enemies”, which, in turn, can “make participants humiliated, frightened, and even more powerless than before”. This leads many citizens to refrain from attending participatory forums and, of those that do attend, many are wary of participating in the discussion and making their opinion heard. In representative democracies, by contrast, parties fill an important function by acting as political entrepreneurs that accrue the major costs of participation, and therefore stimulate various forms of participation in community affairs (Pomper 1980). As a consequence, we would rather expect civic engagement to be lower in assembly than in representative democracies, because the latter form is less costly and demanding of most citizens. Advocates of participatory institutions may also have overstated the ability of most citizens to influence policy outcomes in such contexts. As Mansbridge (1980, 274) puts it, “a face-to-face assembly lets those who have no trouble speaking in public defend their interests; it does not give the average citizens comparable protection.” A third position argues that which of the two opposing claims regarding the impact of participatory institutions on civic engagement depends on the context. In particular, some observers have argued that participatory institutions may suppress civic engagement in conflictual contexts, such as in cases when socio-economic inequality generates situations of dependency among participants. Mansbridge (1980, 273) argues that “[w]henever interests are in conflict, the greater publicity of one’s own act and the greater sanction of one’s neighbor’s visible disapproval in a face-to-face situation can stimulate conformity to the majority against one’s own real interest” or lead to “nonparticipation in politics”. Barber (1984, 273), for his part, concedes that such outcomes may prevail in “parochialized, hothouse communities”. If this line of reasoning is correct, we would expect representative agents, such as parties, to be especially important in accruing the costs associated with participation in policy making in contexts where the distribution of socio-economic resources makes political participation more costly for significant portions of the population. It is also possible that representative democracy yields a better prospect of influencing public policy for the majority of citizens in this context. As Mansbridge (2003, 515) argues, representatives may “negotiate more perceptively and fight more skillfully than constituents have either the time or the inclination to do.” In unequal and conflictual settings, it is thus possible that representative institutions will stimulate more participation, not only by decreasing the costs, but also by increasing the benefits, in terms of policy influence, of participation. There exists relatively little systematic empirical work on the impact of participatory democratic institutions. Research on the direct democratic institution par excellence, the citizens’ assemblies in Athens in ancient Greece, suggests that, typically, about 20% of citizens attended (Smith 2009, 39). Turning to the New England town meetings of more recent times, which is one of the participatory institutions that most resemble the assembly democracies we are going to study in our paper, show that few of the registered voters attend, and that turnout is considerably higher in ordinary elections (Bryan 1999, Zimmerman 1999). The findings from studies comparing those Swiss municipalities that are governed by citizens’ assemblies to those that are governed by representative democracy are, however, less clearcut (Schaub 2010, Ladner and Fiechter 2012). Studies of closely related, but mote limited, institutions of participatory governance – which typically devolve the making of some public policies, or the task of overseeing public administrators, to citizens – typically yield disappointing results from the perspective of political participation. Participatory governance reforms that have been carried out in the United States, India and Brazil typically have low rates of attendance compared to ordinary elections (Ghatak and Ghatak 2002, Baiocchi 2003, Isac and Heller 2003, Fung 2004). The results of Besley et al. (2005) are more nuanced and show that while participation in village meetings that discuss local resource allocation is generally very low, it varies considerably across Indian states, and is fairly high in India’s most developed and least unequal state (Kerala). In a study of participatory budgeting in Brazil, Baiocchi (2003) argues for a positive relationship between this institution and the proliferation voluntary associations. On the other hand, Goldfrank (2002), who studies participatory reforms in local government in Uruguay, argues that they did not stimulate civic engagement in the form of organizational membership and the number of associations, and that civic engagement fell in the wake of the reforms. Turning to a less costly form of direct democracy – popular initiatives – Tolbert et al. (2003) study respondents in the United States and find that their propensity to vote and contribute to interest groups increase in their exposure to ballot initiatives. Boehmke (2002) finds that states with initiatives have larger (and more diverse) interest groups populations, while Lowry (2005) finds that variations in direct democratic provisions across US states have no impact. In as study of Switzerland, Freitag (2006) studies cantonal variations and finds that engagement in voluntary associations is higher where there are lower barriers to popular initiatives. The Swedish Case The General Societal and Political Setting In the early 20th century, Sweden was an agrarian, poor and unequal society. A large majority resided in rural areas and depended for their livelihood on agriculture and related industries such as fishing and forestry (ref). Sweden was also among the least developed and poorest countries in Europe (Maddison 1995). By present standards, Sweden had a high level of economic inequality, and top income data suggests that it was more unequal than several other West European countries (Atkinson et al. 2011). One important explanation for this inequality was Sweden’s failure to implement the land reforms characteristic of other European countries that had also had a history of feudalism (Dovring 1965). During this period, land was relatively unequally distributed. For example, it was less equal than in the United States and Switzerland, but more equal than in United Kingdom (Vanhanen 2003). Neither did Sweden abolish the institutions of primogeniture and fideicommissum for the landholding aristocracy until 1963 thus facilitating the persistence of existing large estates (Segerstråle 1979). Finally, until they were legally abolished in 1945, Sweden had labor repressive agricultural institutions such as the corvée obligations (“torparsystemet”) and the system with contract-workers (“statarsystemet”) that were mostly paid in kind (e.g., Eriksson and Rogers 1978, Lund and Olsson 2005). As a result, farm workers earned less than 50% of manufacturing workers during most of the period up to WWII (Elmer 1963). The period studied in this paper consists of the decades that followed immediately after Sweden’s democratization. Following a period of intense social unrest – arguably, the massive strikes that preceded democratization were perceived as a revolutionary threat – universal male suffrage to the second chamber of the Parliament was introduced in 1909 and in 1918 women gained the right to vote. Up until 1918, votes in municipal elections were weighted according to income and wealth. Thus, in some municipalities, one single individual could have more than 50 percent of the total votes. For example, in 1871, there were 54 local governments where one individual had a majority of votes. However, the maximal voting power was limited to 10 percent in 1900. The reform of 1918 implied that women and men that were at least 23 years old had one vote each and that companies and corporate bodies not longer had voting rights.4 4 Those who were in bankruptcy or in the permanent care of poverty relief institutions were not eligible to vote. Neither were prisoners, foreign citizens, people with unpaid taxes and people with cognitive disabilities. Assembly and Representative Local Governments Local governments have historically played an essential part of the Swedish society. In the Local Government Act of 1863, municipalities were, for the first time, given independent income taxation rights.5 During our period of investigation, 1919-1938, the average local income tax rate was about 10 percent. Local governments had control over important public goods and services within the areas of social welfare and education. Specifically, municipalities were formally responsible for the following five spending programs: basic compulsory education, social welfare or poor relief, child welfare, basic pensions and health care. Basic compulsory education was the largest spending program constituting more than 40 percent of total spending while social assistance to the poor was the second largest program with about 20 percent of total spending. During the period investigated in this paper, municipalities were divided into three categories. Urban local governments were called cities (“städer”), and semi-urban were called boroughs (“köpingar”) and rural were called rural local governments (“landskommuner”). In 1950, there existed 133 cities, 84 boroughs and 2281 rural local governments. As we explain below, we only focus on rural local governments in this paper. Historically, Swedish local governments were governed by face-to-face assemblies (“kommunalstämmor”) that consisted of all eligible voters. Until 1918, government by representative democracy (“kommunfullmäktige”) was voluntary for rural municipalities, whereas it was mandatory in cities with more than 3000 inhabitants. In practice, however, only a tiny fraction of rural municipalities chose this option. Following a change in the Local Government Act in December 1918, which was a part of the major constitutional reform that extended equal and universal suffrage, national law required that municipalities with a 5 This section is based on the Swedish Code of Statutes (Svensk författningssamling, SFS). SFS 1862:13, SFS 1918: 1026 and SFS 1930:251. population of more than 1500 people have representative democracy.6 Those with populations below this threshold, on the other hand, could choose between direct and representative democracy. As noted above, we will exploit the population threshold in order to carry out a fuzzy regression discontinuity design. Table 1 shows the number of municipalities that were governed by representative democracy, and whether they had chosen it voluntarily or not, as well as the number that were governed by assembly democracy, for the regular election years 1919, 1922, 1926, 1930, and 1934. As is clear, the bulk of municipalities with a population size below the 1500 threshold chose to keep direct democracy. However, Table 1 also shows that the number of municipalities that voluntarily choose to switch to representative government quadrupled between 1919 and 1934. [Table 1 about here] In municipalities with representative democracy, elections were proportional with closed party lists in multi-member districts and were held every fourth year. The main parties were the Communists and the Social Democrats on the left and the Agrarian Party, the Liberal Party, and the Conservative Party on the right. However, there were also a fairly large number of elections that were characterized by Statistics Sweden as nonpartisan because there were no lists with party political labels. The elected council was required to hold at least three meeting per year, and national law required that a majority (for some decisions, a supermajority) of the council be present to constitute a quorum. The chairman of the council and the vice-chairman were elected on a yearly basis and the “executive” of the local government (“kommunalnämnden”) was required to have 5 to 11 members and was elected by the council. The number of council members ranged from 15 to 40 depending on the population size. 6 For a thorough discussion of the debate in Parliament over assembly and representative democracy that preceded the changes in the Local Government Act of 1918, see Strömberg (1974) and Wallin (2007). Importantly, none of these population thresholds were close to the threshold of 1500 used in this paper.7 For municipalities with assembly democracy, the Local Government Act was identical, except that assemblies of citizens rather than elected politicians made decisions.8 Decision-making by the citizens’ assembly had to proceed as follows. After the discussion of an item on the agenda had been concluded, the chairman made a motion. The assembly then took a voice vote (“yes” or “no”) after which the chairman declared the outcome. If someone requested a second vote, this vote could either be an open (a roll call vote) or closed vote (a secret ballot). Any attendant at town meetings, thus always had the option of requiring a secret ballot. In addition, eligible voters attending the town meeting were entitled to represent another voter (one, at most) if she had the power of attorney do so. Civic Engagement in Sweden 1919-1939 In the parliamentary elections that followed the introduction of female suffrage about 55% of the eligible voters participated. This figure continued to grow during our period of study, and in the 1940 parliamentary elections turnout stood at 70%. Looking at local elections in those municipalities that practiced representative democracy, turnout was lower than in parliamentary elections in the beginning of our period of study, but gradually increased and was close to that for national elections at the end (Ersson and Wide 2001). Unfortunately, the attendance rate at town meetings was not recorded by Statistics Sweden. However, using a database that digitizes minutes from local governments in six out of Sweden’s then 24 counties, we have collected the minutes of the town meetings for a set of local governments 7 The closest threshold was at 2000, and stated that in municipalities with a larger population than this, the council had to have, between 15 and 25 members. 8 Also, the chairman and the vice-chairman of the assembly had to be at least 25 years old and were elected for a four-year term rather than on a yearly basis. after democratization, and these contain information about attendance rates in case there was a second vote (described above).9 Between 1919 and 1938, there were a total of 567 meetings with a second vote. Consequently, there was a second vote in 19% of all required meetings. We therefore have data on 74 local governments, and the average attendance rate was 14% (Petterson-Lidbom and Tyrefors Hinnerich 2012).10 Table 2 summarizes turnout rates in local elections for rural municipalities by the type of government during our period of study. [Table 2 about here] Voluntary associations were also an important and growing component of civic engagement during this period. Membership in the three main popular movements at the time – free churches, temperance associations and unions – had been growing steadily since the second half of the 19th century and mobilized a large portion of the population during the period we study. For instance, overall estimates suggest that in 1920 about 25% of Swedish adults were members of one of these three main voluntary associations (Lundkvist 1977, 67). Historians have pointed to their vital importance for Sweden’s democratization (see, e.g., Lundkvist 1977, Ch. 6). The union movement was an early champion of universal suffrage, together with a large contingent of free churches and temperance associations. These popular movements were also deeply involved in the political process at the local level. They mobilized voters, held election rallies and, and in some cases, even printed ballots, in order to affect election outcomes. Finally, they were “schools of democracy”, supplying members and activists with education and organizational skills, and engaging in political socialization and consciousness building (Johansson 1954, Ch. 6-7, Lundkvist 1977, Ch. 6). 9 According to our calculations, there was a second vote in 14% (608 meetings) of all required assembly meetings (4440) and a secret ballot was used in 40% of these cases. 10 The 74 local governments for which we have attendance rates are simply those that are available in the database at the time of writing this. In the future, those working with the database have plans to extend coverage. The free churches and temperance movements were the earliest mass popular movement in Sweden. The free churches started gaining momentum during the first half of the 19th century and grew even more after Sweden took some steps towards religious freedom in the mid 19th century. At the outset, these revivalist churches were, by virtue of their struggle for religious freedom, revolutionary protest movements. Historical data also suggests that they, to a large extent, mobilized workers, although there were also important contingents of upper class members. As the 19th century came to its close, however, the share of farmers and artisans was increasing at the expense of workers (Lundkvist 1977, 47-50, 104-105, 117). The period of democratization in the early 20th century did, however, bring deep splits within the movement. Although the majority of free churches were most closely tied to the Liberal Party, a large minority was tied to the anti-democratic Conservative party. During the period after democratization that we study in this paper, this pattern of association persisted. The free churches and their papers were important forums for political debate; although they rarely took official political stands after the issue of religious freedom had been settled (Johansson 1952, 90, 119-120, 136ff, 166-172). Research shows that free church members’ local political activity was mainly channeled through temperance associations and, in some cases, trade unions. This was a conscious strategy, and allowed members of different free church congregations, whose views on theology as well as other issues diverged, to join forces on the specific political issues that united them (Lundkvist 1977, 168, 171). The early 19th century also saw the emergence of organized temperance associations. The associations were with few exceptions, religious, and there was often a considerable membership and leadership overlap with the revivalist movement. But the temperance associations also suffered from membership competition with the free churches. Towards the end of the century, following the establishment of the International Order of Good Templars (IOGT), the movement started to grow at a quicker pace and the total number of members in free churches stood at about 250 000 in 1900. At this point, the movement was increasingly turning their focus away from personal salvation to campaigning politically for local and national referenda on prohibition. The campaign resulted in the national referendum in 1922, in which the side favoring prohibition lost by a narrow margin (Johansson 1954, 26, 29-30, Lundkvist 1977, 50-52, 67). Although the temperance movement was initially most closely associated with the Liberals, but also to some extent the Conservatives, the 20th century saw their gradual orientation towards the left of the political spectrum. Investigations of the membership structure of temperance associations indicate that their recruitment base was similar to that of free churches; a large group was workers, like in the free churches, but the contingent of businessmen was smaller (Lundkvist 1977, 119). Whereas some lodges and associations associated with the Conservatives were skeptical of universal suffrage, many others gradually came to be advocates. One important reason was the political goal of prohibition, an issue that became closely tied to that of universal suffrage during the first decade of the 20th century.11 This development culminated in 1918, when the grand lodge of IOGT, the largest of the temperance movements, declared its support of women suffrage and its opposition to the system of income-weighted voting in municipalities. There were also clear personal links between the temperance movement and the Social Democratic party. For instance, teetotalers made up between 40 and 50% of the members of the second chamber of the Swedish parliament during the period of our study, and a large majority of these were Social Democrats. During the 1920s, many of the Communist parliamentarians were teetotalers (Johansson 1952, 109-110, 118, 125, 157, 166-172). At the local level, temperance 11 In particular, the temperance movement became advocates of allowing municipalities to hold local referenda with universal suffrage on whether or not to prohibit the sale of alcoholic beverages. Having taken the step of advocating universal suffrage on matters of utmost importance to local public finances – taxes from the sale of alcoholic beverages could amount to as much as 12% of local revenues – the step to advocating universal suffrage was not far. The movement for this local veto against the sale alcoholic beverages was therefore perceived as threatening by leading Conservatives (Lundkvist 1977, 184-185). associations mobilized voters and supplied political candidates for all parties, but primarily the Liberals and the Social Democrats (Lundkvist 1977, 186). The largest popular movement during our period of study was the trade unions. They were also the youngest, only beginning to take form in the late 20th century. Despite considerable employer resistance it managed to grow quickly during the first part of the 20th century (Johansson 1954, 31-42). The massive loss of members that resulted from the defeat in the general strike of 1909 was only temporary and, in 1920, about 400 000, or 31%, of workers were union members. This was a lower than in many other West European countries, such as the United Kingdom, but after 1920 the Swedish union movement continued to grow rapidly. In 1940, it organized almost a million workers and well over 100 000 white collar workers. This corresponded to a unionization rate of 66% among workers and 58% among all wage earners. During this entire period, however, workers in agriculture were very poorly organized in comparison to other industries, even if the unionization rate grew considerably within this category as well (Kjellberg 1983, 269-272). The movement was an early champion of universal suffrage, as was evidenced by the political general strike of 1902 which mobilized over 100 000 workers. The vast majority of unions were closely allied with the Social Democratic party, although parts of the movement had close ties to the Communists. Of national parliamentarians, between 20 and 40% were union members during our period of study, all of which belonged to the Social Democrats and Communists (Johansson 1952, 166172). At the local level, trade unions mobilized voters and supplied candidates, primarily for the Social Democrats (Lundkvist 1977, 186). In Table 3, we show the development of associational memberships for the three types of organizations included in our study by the type of government during our period of study. [Table 3 about here] Data and Research Design In order to study the impact of the participatory democratic institutions that existed in some Swedish rural communities during the period following democratization on civic engagement, we have collected a new and comprehensive data set. Given the unique features of the Local Government Act of 1918 that we have described above, we can use this data set to generate credible estimates of the causal effect of assembly democracy on civic engagement using a fuzzy regression discontinuity design. These data and the design are described in the following. Data Our data is a new and comprehensive panel consisting of about 2500 local governments during the period 1918-1938. Our data on civic engagement comes from the Popular Movement Archive and contains annual membership figures for all unions, free churches and temperance associations during this period. These data were originally collected by Carl Göran Andrae and Sven Lundkivst at the Department of History, Uppsala University (see, e.g. Lundkvist 1977) and have previously been used in several quantitative studies of the emergence and spread of popular movements in Sweden during the first half of the 20th century (see, e.g., Hedström 1994, Sandell 2001, Edling and Liljeros 2003). All remaining data comes from both published and unpublished material from Statistics Sweden, the government agency responsible for producing official statistics about Sweden. The unpublished material is kept in the National Archives of Sweden. We have also collected 21 baseline or pre-treatment characteristics, i.e., variables dated before the introduction of the two treatments—direct or representative democracy—in 1919. One set of variables consists of characteristics of social-welfare programs which include the baseline outcome—aggregate welfare spending—together with the number of total recipients including children, the number of adults, the number of children directly supported, the number of children indirectly supported, the number of people receiving full support, the number of people boarded out, the number of people in public institutions (i.e., the number of poorhouses), the number of public institutions and the number of slots available in the public institutions. The other set of variables consists of three geographic variables: total area and land area, three economic variables: arable land, income tax base and economic structure (percent of the economy based on agriculture), population size, and four variables capturing the political characteristics of the community: the number of eligible voters at the parliamentary elections in 1917, the turnout at the parliamentary elections in 1917, the proportion of left-wing voters at the parliamentary elections in 1917, the number of union members and an indicator for having at least one labor union in the municipality. Table 4 contains descriptive statistics for the main variables and the pre-treatment control variables used in this paper. [Table 4 about here] The Fuzzy Regression Discontinuity Design The fuzzy regression discontinuity design differs from the more-well known sharp regression discontinuity design in that crossing the threshold used in the application does not necessarily cause the probability of being treated to jump from zero to one. All the fuzzy RD design requires is that the probability of treatment jumps at the threshold (Lee and Lemieux 2010). In our particular application, those municipalities with a population over the threshold of 1500 inhabitants are forced to govern themselves by representative democracy whereas those remaining were free to choose between assembly and representative government. In particular, we will estimate a regression of the following form: (1) !!" ! ! ! !!!" ! ! !!"!! ! !!" where i indexes municipalities and t indexes years. !!" is the outcome variable (i.e. one of our measures of civic engagement), !!"!! !is the forcing variable (population size at t-1), and !!! is an indicator variable taking the value 1 if a local government has direct democracy and 0 if it has representative democracy. The parameter of interest is !, which is the effect of having direct rather than representative democracy. Since our RD design is fuzzy, we will use the eligibility rule!!!" ! !!!!"!! ! !"##! as an instrumental variable for!!!" . It is important to stress that we have a compliance problem. However, since the treatment rule is only binding for those municipalities with populations above the threshold, the problem is only one-sided (e.g., Bloom 1984, Battistin and Rettore 2008). Therefore, our estimate of the parameter ! corresponds to the effect of the treatment on the treated. One aspect of the Local Government Act complicates our analysis, namely the status-quo bias in favor of representative democracy. In particular, this problem arises because a municipality that has representative democracy cannot switch to assembly democracy until after five years, even if its population falls below 1500. We solve this by redefining the assignment variable as (2) !!" ! !"# !!"!! ! !!!"!# where !!!"!# !is the population size in 1918. The treatment rule can then be defined as !!" ! !!!!" ! !"##! and we estimate the following specification: (3) !!" ! ! ! !!!" ! ! !!"!! ! !!" where we use !!" !as an instrument for !!" . To approximate ! !!"!! !we use a number of different order of polynomials of the forcing variable (1st up to 3th order). Moreover, we use a number of different window sizes around the population threshold of 1500 ranging from ±500 to ±63. When narrower windows are used, ! !!"!! !can be approximated with lower order polynomials. Finally, as can be seen above, our data has a panel structure, but we follow Lee and Lemieux (2010) who strongly recommend that one treat panel data as a pooled-cross-section so as to avoid imposing a specific dynamic structure, which would not, in any case, improve our prospects of identifying the causal effect of interest. However, in order to avoid the potential mischiefs that may result from autocorrelation, we follow their suggestion to use municipality-clustered standard errors. Results To gauge whether participatory or representative democratic institutions do a better job of stimulating civic engagement, we begin by showing the fuzzy-RD results for the impact of these institutions on turnout and total membership in voluntary associations. These are then followed by auxiliary analyses where we assess the robustness and credibility of our results. Finally, we present some extensions to our main analysis. Main Results Figure 1 shows the first-stage relationship between the forcing variable and direct democracy and Figures 2 and 3 show the reduced form relationship between the forcing variable and voter turnout in municipal elections and total associational memberships. We have plotted conditional means where observations have been grouped into bins of size 50 and we have graphed the outcome variables as a 3rd order polynomial function of the forcing variable on each side of the threshold. Beginning with the first-stage relationship in Figure 1, it shows that there is a large leap in the probability of having direct democracy as one crosses the 1500 threshold. In particular, the probability of having direct democracy increases by about 50 percentage points. As can also be seen, there is near perfect compliance on the right-hand side of the population threshold. The reason why there exists a few observations that have population figures higher than 1500 and still have direct democracy is not that these municipalities were excepted from the rule. Rather, it is a consequence of the fact that we had to redefine our forcing variable because of the status-quo bias in favor of representative democracy that was built into the Local Government Act (see equation (2)). For all practical purposes, however, we have perfect compliance among the municipalities whose value of the forcing variable exceeds 1500.12 Turning to the reduced form relationships between the forcing variable and our measures of civic engagement in Figures 2 and 3, we see that there are clear discontinuities at the thresholds in both figures. Moreover, the sizes of the jumps are large. Judging from Figure 1, mean voter turnout almost doubles at the threshold. And in Figure 2, we see that average associational memberships jumps by 60, or approximately 75%, when passing the threshold. [Figure 1 about here] [Figure 2 about here] [Figure 3 about here] Turning to the RD results, they are in Table 5 and 6. In each table, we present three different estimates: (i) The reduced-form effect, i.e., Yit = a + bGit + g(Wit) + !it, (ii) The first-stage effect, i.e., Dit = c + dGit + h(Wit) + "it, (iii) The instrumental variable (Wald) estimate, i.e., the ratio between the estimates of the reduced-form effect and the first-stage effect. 12 The estimated treatment effect with imperfect compliance on both sides is a local average treatment effect (LATE) but it converges to the treatment effect on the treated (TOT) when one moves towards perfect one-sided compliance (see, e.g., Bloom 2009). For all practical purposes, therefore, this paper estimates the TOT rather than the LATE of direct democracy on civic engagement. Panel A of Tables 5 and 6 shows the reduced-form estimates, Panel B shows the first-stage estimates, and Panel C displays the Wald estimates. [Table 5 about here] [Table 6 about here] Beginning with the reduced form estimates in Panel A, the estimates suggest that civic engagement is adversely affected by direct democracy. Depending on the exact model specification, the results in Table 5 suggest that voter turnout drops by between 12 and 15 percentage points when crossing the 1500 threshold. Given that average voter turnout for our sample during this period was 28.5 percent, this is indeed a large effect. The reduced form estimates for associational memberships are in Table 6, vary somewhat more than those for turnout. They suggest that the number of memberships drop by between 70 and 140 when comparing those that could choose direct democracy to those who could not. Like was the case with voter turnout, this too is substantively large effect given that the average number of association members for our sample is 192.5. The first-stage estimates that are in Panel B of Tables 5 and 6 show that the impact of the threshold on the proportion of municipalities that had direct democracy. For the sample used to analyze voter turnout, the share of municipalities having direct democracy increases by between 43 and 55 percentage points depending on the bandwidth. As can be seen from Table 6, the corresponding figures for the sample used to analyze associational memberships are 35 and 41 percentage points. All first-stage effects in Tables 5 and 6 are precisely estimated. Panel C of Tables 5 and 6 shows the estimated causal effect of direct democracy on our two measures of civic engagement. This is analogous to the Wald estimate, which is the reduced form estimate divided by the first-stage estimate. As can be seen from Table 5, the treatment effect on the treated of direct democracy on turnout is large and negative. Our estimates suggest that direct democracy lowers turnout by between 26 and 31 percentage points. The estimates for associational memberships are also large but somewhat more varying. Depending on the bandwidth, they suggest that direct democracy lowers the number of members by somewhere between 180 and 370. Robustness Checks As mentioned when discussing the historical background to our case, before 1919 municipalities with more than 3000 inhabitants were forced to have representative democracy. The remaining municipalities could choose direct democracy, which is something that almost all of them did. The reform that was undertaken in 1919 and which underlies our estimation strategy changed the threshold to 1500. Our data includes a rich set of pre-reform control variables for each municipality, which makes it possible to evaluate whether these variables, which were determined before 1919, are balanced on either side of the 1500 threshold. In a previous paper, we have shown that there is little evidence for an effect of the threshold on a large set of 21 economic, geographic and political baseline characteristics, providing strong support for the notion that whether the number of inhabitants in a municipality are a few less or a few more than 1500 can be considered as good as random. In addition, we can test whether including a large number of baseline characteristics in our RD regressions affects our conclusions. Table 7 displays the results when we include the abovementioned 21 pre-determined covariates as well as time dummies and the total number of association memberships for 1918. The sample is slightly reduced when adding the pretreatment controls. Therefore, we show both the results with and without controls for the sample for which we have data on pre-treatment controls. As can be seen when comparing Panels A and B, the results for voter turnout are hardly affected at all as the magnitude and significance of the coefficients are very close, irrespective of bandwidth. Turning to associational memberships, and comparing the results in Panel C in Table 7 to those in Panel D, we see that the magnitude of the coefficients tends to decrease upon the inclusion of the pretreatment controls. However, the effects are still substantively large and statistically significant. [Table 7 about here] One outstanding issue concerns our forcing variable: population size. If there were any possibilities for municipalities to strategically misreport it, this would render our design problematic. Fortunately, from our perspective, the municipalities themselves did not administer the population registers. Instead, the Swedish State church was charged with recording births, deaths as well as migrations in and out of each municipality. In practice, this task fell on the clergyman of each parish or one of his assistants. Therefore, municipalities could not misreport their population size. However, although it does perhaps seems farfetched, one can not a priori rule out the possibility that municipalities might actively encourage in- and out-migration with the objective of staying below (or above) the 1500 threshold. This would, again, render municipalities on either side of the threshold incomparable, and thereby invalidate our RD approach. To test for sorting around the threshold, we use the density test by McCrary (2008). According to this test, there is no significant change in the density when crossing the 1500 threshold. The results from the McCrary density test are shown graphically in Figure 4 where the plot shows local densities with bin size 19 and the density as a non-parametrically estimated function of the forcing variable together with 95% confidence intervals. [Figure 4 about here] Extensions Thus far, we have only distinguished between two forms of civic engagement: voter turnout and membership in associations. However, we have information on three types of association membership in our data. This feature of our data can be utilized to indirectly gain some insight into whether direct democracy suppressed civic engagement across all socio-economic strata or just some. This seems especially important given that a great deal of the interest in participatory democratic institutions is motivated by the belief that they will empower marginalized groups (see, e.g., Fung and Olin-Wright 2003, Smith 2009). Table 8 presents our disaggregated analysis. To conserve space, we only show the secondstage results with pre-treatment controls, but the same basic pattern emerges when we do not include the controls. As can be seen, direct democracy has a negative impact on all three types of association. However, the size of the effect is clearly largest for unions, and it is only for this outcome that the effect is statistically significant across all specifications. This means that our results on aggregate membership are primarily driven by union membership. Since unions were the only associations in which all members were working-class, the results suggest that direct democracy first and foremost suppressed participation among the lower socio-economic strata. [Table 8 about here] Conclusion Beginning with the student movements of the 1960s, calls have been sounded for creating arenas where citizens can engage more directly in politics and policy making. One important argument has been that such reforms would help us deal with the low civic engagement that allegedly plagues representative democracy. Little empirical research exists, however, on the basis of which we can evaluate the veracity of such assertions. The dearth of such empirical evidence becomes even more troubling when one considers the fact that experiments with more direct forms of citizen involvement in politics have become increasingly common in developing countries. The few systematic studies that do exist are correlational and their conclusions are incoherent. Thus, the fact of the matter is that we still do not know whether direct democracy stimulates civic engagement or whether representative democratic institutions will do a better job. This study provides the first quasi-experimental evidence on this issue. Using a panel of Swedish local governments 1919-1938, we exploit a unique feature of the Swedish Local Government Act to generate credible estimates of the causal effect of assembly democracy on civic engagement. The fact that municipalities with less than or equal to 1500 inhabitants could choose between adopting assembly or representative democracy, whereas others had to adopt representative democracy, allows In our empirical analysis, we find that participatory democratic institutions has a large dampening effect on citizen participation in political and societal affairs. In particular, when comparing municipalities around the population threshold, our results suggest that turnout was somewhere between 20 and 30 percentage points higher in representative democracies. Moreover, the results suggest that whereas associations thrived in representative democracies, they were poorly mobilized in direct democracies. This is especially true of unions, suggesting that participatory institutions have a particularly adverse impact on the civic engagement of lower socio-economic strata. All our results are robust to the inclusion of a large number of controls and we find no evidence of sorting around the population threshold. What, then, are the broader implications of our results? First, and most obviously, they strongly suggest that it might not be possible to solve the problem of low civic engagement by introducing participatory democratic institutions. Second, they are especially disheartening given the recent drive for reforms aiming at introducing participatory forms of governance in developing countries. During the period when assembly democracy was practiced, Sweden was a newly democratized developing country characterized by high levels of economic inequality and poverty. Moreover, it still retained features of a labor repressive agricultural system that had been put in place long before democratization. Going all the way back to Rousseau (1762, III:4), theorists of participatory institutions has suggested that such institutions are unlikely to work well in these kinds of societies. Our findings do nothing to assuage these fears. On the contrary, we find that civic engagement suffers when direct democracy is introduced in a context where socio-economic inequality is such that it generates situations of dependency among participants. One possible retort to our findings is that many proposals for increasing participatory governance do not advocate doing away with existing representative institutions, but merely seek to open up new additional venues for direct participation. Unless there is a spillover effect, so that low participation in the participatory channels suppress participation in the representative channels, adding new venues for participation might, in theory, increase overall civic engagement. Without more and better research on this topic, however, we will not know. Another possible retort is that the form of direct democracy studied here – popular assemblies – is not representative of other forms of direct democracy such as referenda. Given that ours is the first study of the impact of direct democracy on civic engagement that uses a credible research design, this is, of course, an open question. And irrespectively of whether there is something to such and objection or not, our findings regarding the impacts of assembly and representative democracy on civic engagement go to the heart of democratic theory. As Smith (2009, 30) writes, “the idea of a popular assembly where citizens engage in face-to-face interactions and decision-making is arguably the touchstone for much writing on citizen participation”. Our results also shed new light on scholarly controversies regarding the impact of direct democracy on public policy. As we said in the introduction, many value civic engagement because they believe that participation makes policy more responsive to the interests of citizens. A large number of studies have shown that direct democracy leads to less government spending. Some have interpreted this in light of accountability, arguing that voters are more fiscally conservative than their representatives, and that direct democracy then limits ’excessive’ spending (see, e.g., Bails and Tieslau 2000). An alternative interpretation is that in a direct democracy, it is easier for the well-to-do elite to monopolize the policy-making process (Tyrefors Hinnerich and Pettersson-Lidbom 2012). Our results speak strongly in favor of the latter explanation in that they show that various forms of societal and political participation are substantially lower in direct democracy. References Abadie, Alberto. 2003. “Semiparametric Instrumental Variable Estimation of Treatment Response Models,” Journal of Econometrics 113: 231-263. Acemoglu, Daron and James Robinson. 2006. Economic Origins of Dictatorship and Democracy, New York: Cambridge University Press. Angrist, Joshua D. and Guido W. Imbens. 1995. “Two-Stage Least Squares Estimation of Average Causal Effects in Models with Variable Treatment Intensity.” Journal of the American Statistical Association 90: 431-442. Bails, D. and Tieslau, M.A. 2000. “The impact of fiscal constitutions on state and local expenditures.” Cato Journal 20: 255-77. Baiocchi, Gianpaolo. 2003. “Participation, Activism and Politics: The Porto Alegre Experiment” in Archon Fung and Erik Olin-Wright. Deepening democracy: Institutional innovations in empowered participatory governance. London: Verso. 45-76. Barber, Benjamin. 1984. Strong Democracy: Participatory Politics for a New Age. Berkeley: University of California Press. Banerjee, Abhijit V. and Rukmini Banerji and Esther Duflo and Rachel Glennerster and Stuti Khemani. 2010. "Pitfalls of Participatory Programs: Evidence from a Randomized Evaluation in Education in India," American Economic Journal: Economic Policy 2: 1-30. Battistin, Erich and Enrico Rettore. 2008. “Ineligibles and Eligible Non-Participants as a Double Comparison Group in Regression-Discontinuity Designs.” Journal of Econometrics 142: 715-730. Berry, Jeffrey J. and Kent E. Portney and Ken Thomson. 1993. The Rebirth of Urban Democracy. Washington D.C.: The Brookings Institution. Besley, Timothy, Rohini Pande and Vijayendra Rao. 2005. “Participatory Democracy in Action: Survey Evidence from India." Journal of the European Economics Association 3:648657. Bloom, Howard S. 1984. “Accounting for No-Shows in Experimental Evaluation Designs.” Evaluation Review 8: 225–246. Bloom, Howard S. 2009. Modern regression discontinuity analysis. New York, NY: MDRC Boehmke, Frederick J. 2002. “The Effect of Direct Democracy on the Size and Diversity of State Interest Group Populations” Journal of Politics 64: 827-844. Bryan, Frank M. 1999. “Direct Democracy and Civic Competence: The Case of Town Meeting” in Stephen L. Elkin and Karol Edward Soltan. Citizen Competence and Democratic Institutions. University Park, Pennsylvania: Pennsylvania State University Press. Edling, Christofer and Fredrik Liljeros. 2003. “Spatial diffusion of social organizing: Modelling trade union growth in Sweden.” Geography and Strategy 20: 267–192. Freitag, Markus. 2006. “Bowling the State Back in: Political Institutions and the Creation of Social Capital” European Journal of Political Research 45: 123-152. Fung, Archon and Erik Olin-Wright. 2003. Deepening democracy: Institutional innovations in empowered participatory governance. London: Verso. Fung, Archon. 2004. Empowered Participation: Reinventing Urban Democracy. Princeton: Princeton University Press. Ghatak, Maitreesh and Maitreya Ghatak. 2002. “Recent Reforms in the Panchayat System in West Bengal: Towards Greater Participatory Governance?’ Economic and Political Weekly of India January: 45–58. Goldfrank, Benjamin. 2002. “The fragile flower of local democracy: A case study of decentralization/participation in Montevideo.” Politics & Society 30: 51-83. Hahn, Jinyong and Petra Todd and Wilbert Van der Klaauw. 2001. “Identification and Estimation of Treatment Effects with a Regression-Discontinuity Design,” Econometrica 69: 201-209. Hedström, Peter. 1994. “Contagious Collectives: On the Spatial Diffusion of Swedish Trade Unions, 1860–1940.” American Journal of Sociology 99:1157–79. Isaac, Thomas M. and Patrick Heller. 2003. ”Democracy and development: decentralized planning in Kerala, India.” In Deepening Democracy: Institutional Innovations in Empowered Participatory Governance, ed. Archon Fung and Erik Olin Wright, pp. 77–110. London: Verso. Johansson, Hilding. 1952. Folkrörelserna och det demokratiska statsskicket i Sverige. Karlstad: Gleerups. Johansson, Hilding. 1952. Folkrörelserna. Örebro: Folkbildningsförlaget AB. Ladner, Andreas and Julien Fiechter. 2012. “The influence of direct democracy on political interest, electoral turnout and other forms of citizens’ participation in Swiss municipalities”, Local Government Studies, forthcoming. Lee, David and Thomas Lemieux. 2010. “Regression Discontinuity Designs in Economics,” Journal of Economic Literature 48: 281-355. Lundkvist, Sven. 1977. Folkrörelserna i det svenska samhället . Uppsala: Almkvist & Wiksell. Lijphart, Arend. 1997. “Unequal Participation: Democracy’s Unresolved Dilemma” American Political Science Review 97: 1-14. Lowry, Robert C. 2005. “Explaining the Variation in Organized Civil Society across States and Time.” The Journal of Politics 67: 574-594. Mansbridge, Jane J. 1980. Beyond Adversary Democracy. New York: Basic Books. Mansbridge, Jane J. 2003. “Rethinking representation.” American Political Science Review 97: 515–28. McCrary, Justin (2008), “Manipulation of the Running Variable in the Regression Discontinuity Design: A Density Test,” Journal of Econometrics, Volume 142 , Issue 2, Nagel, Jack. 1987. Participation. New Jersey: Prentice-Hall. Nylen, William R. 2002. “Testing the Empowerment Thesis: The Participatory Budget in Belo Horizonte and Betim, Brazil.” Comparative Politics 34: 127-15. Pateman, Carole. 1970. Participation and Democratic Theory. Cambridge: Cambridge University Press. Plotke, David. 1997. “Representation is Democracy.” Constellations 4: 19-34. Pomper, Gerald. 1980. “The Contribution of Political Parties to American Democracy” in Gerald Pomper. Party Renewal in America. New York: Praeger Publishers. Putnam, Robert. 2000. Bowling Alone: The Collapse and Revival of American Community. New York: Simon and Schuster. William H. Riker and Peter C. Ordeshook, ”A Theory of the Calculus of Voting.” American Political Science Review 62: 25-42. Rosenstone, Steven J. and John M. Hansen. 1993. Mobilization, Participation, and Democracy in America. New York: MacMillan. Rousseau, Jean-Jaques. 1947[1762]. The Social Contract. New York: Hafner Rueschemeyer, Dietrich. 1998. “The Self-Organization of Society and Democratic Rule: Specifying the Relationship.” in Dietrich Rueschemeyer and Marylin Rueschemeyer and Björn Wittrock. Participation and Democracy East and West. New York: M.E. Sharpe. 9-25. Sandell, Rikard. 2001. “Organizational Growth and Ecological Constraints: The Growth of Social Movements in Sweden, 1881 to 1940.” American Sociological Review 66:672–94. Skocpol, Theda and Marshall Ganz and Ziad Munson and Bayliss Camp and Michele Swers and Jennifer Oser. 1999. “How Americans Became Civic” in Theda Skocpol and Morris P. Fiorina. Civic Engagement in American Democracy. Washington, DC: Brookings Institution Press, and New York: Russell Sage Foundation. 27-80. Stolle, Dietlind and Marc Hooghe. 2004. “Inaccurate, Exceptional, One-Sided or Irrelevant? The Debate about the Alleged Decline of Social Capital and Civic Engagement in Western Societies.” British Journal of Political Science 35: 149-167. Teorell, Jan. 2006. “Political Participation and Three Theories of Democracy: A Research Inventory and Agenda.” European Journal of Political Research 45: 787–810 Tolbert, Caroline J. and Ramona S. McNeal and Daniel A. Smith. 2003. “Enhancing Civic Engagement: The Effect of Direct Democracy on Political Participation and Knowledge” State Politics and Politics Quarterly 3: 23-41. Tyrefors-Hinnerich, Björn and Per Pettersson-Lidbom. 2012. “Democracy, Redistribution and Political Participation: Evidence from Sweden 1919-1938.” Working Paper, Department of Economics, Stockholm University. UN. 2008. Participatory Governance and the Millennium Development Goals. New York: United Nations. Verba, Sidney and Kay Lehman Schlozman and Henry E. Brady. 1996. Voice and Equality: Civic Voluntarism in American Politics. Cambridge Mass.: Harvard University Press. Zimmerman, Jospeh F. 1999. The New England Town Meeting: Westport, CT: Praeger Publishers Democracy in Action. Table 1. Number of local governments with representative and assembly democracy Election year Representative democracy Assembly democracy Mandatory Voluntary 1919 870 67 1469 1922 889 117 1398 1926 887 147 1377 1930 873 192 1354 1934 867 274 1273 Table 2. Average voter turnout (%) in the rural local government elections and assemblies Election year Representative Assembly 1919 1922 1926 1930 1934 52 28 42 51 58 7 9 11 10 18 Notes: Turnout rates for assembly democracies are calculated on the basis of meeting attendance and the number of eligible voters in the municipality in question Table 3. Average number of association memberships in the rural representative and assembly democracies Election year Representative Assembly Panel A. Union memberships 1919 1922 1926 1930 1934 Panel A. Free church memberships 1919 1922 1926 1930 1934 Panel A. Temperance association memberships 1919 1922 1926 1930 1934 96.96 125.6 163.0 185.7 195.6 6.591 6.471 9.051 25.60 14.67 91.38 94.02 94.36 92.82 93.35 17.00 18.52 18.38 17.90 18.00 154.9 149.4 142.8 112.1 110.9 26.92 19.85 17.22 14.74 15.10 Notes: Turnout rates for assembly democracies are calculated on the basis of meeting attendance and the number of eligible voters in the municipality in question Variables Table 4. Descriptive Statistics for all rural local governments Mean St.Dev. min max Panel A. Outcome variables 1919-1938 Voter turnout Total number of association memberships Number of unions members Number of free church members Number of members in temperance associations 0.285 192.5 80.18 49.75 62.52 0.216 423.9 248.7 113.2 158.8 0.0005 0 0 0 0 0.81 12199 11036 8171 7102 Panel B: Treatment determining or forcing variable 1919-1938 Population size 1706 2008 90 26491 Panel C: Baseline or pre-treatment characteristics as measured in 1917 or 1918 Sum of aggregate social welfare spending Number of total recipients including children Number of adult recipients Number of children directly supported Number of children indirectly supported Number of people receiving full support Number of people boarded out Number of people in public institutions Number of public institutions Number of slots available in public institutions Total area (km") Land area (km") Arable land (km") Income tax base Economic structure (% agriculture) Population size Number of eligible male voters at the parliamentary elections in 1917 Number of voters at the parliamentary elections in 1917 Proportion of left-wing voters at the parliamentary elections in 1917 Number of union members Indicator for having at least one labor union 4763 59 38 7 14 21 7 13 0.77 19 18163 17033 1573 204645 49.6 1706 359 7891 104 59 15 38 28 14 20 0.58 24 81553 75877 1195 446346 22.1 1997 373 0 0 0 0 0 0 0 0 0 0 0 15 0 3713 0 110 0 119291 1714 1090 289 581 295 139 196 8 200 1.947e+06 1.814e+06 13524 6.691e+06 98.5 21648 4373 228 234 0 3003 0.29 0.20 0 1.00 22 0.16 87 0.37 0 0 1047 1 Note: All nominal values are deflated with CPI with 1914 as the base year. Table 5. Estimates from the regression-discontinuity design on voter turnout (1) (2) (3) (4) (5) Panel A: Reduced form relationship between voter turnout and the instrument Git= 1[Wit ! 1500] Reduced form effect -0.144*** (0.047) -0.123* (0.066) -0.151*** (0.051) -0.132* (0.070) -0.130** (0.065) Panel B: Reduced form relationship between direct democracy and the instrument Git= 1[Wit ! 1500] First-stage effect 0.552*** (0.077) 0.430*** (0.093) 0.537*** (0.083) 0.431*** (0.098) 0.452*** (0.093) Panel C: The causal effect of direct democracy versus representative democracy on voter turnout Direct democracy =1 -0.260*** (0.075) -0.285** (0.137) -0.282*** (0.083) -0.307** (0.143) -0.287** (0.128) Observations 3,502 1,704 1,704 790 413 250 3 250 2 125 2 63 1 Bandwidth 500 Degree of polynomial in 3 the forcing variable Wit Notes: Standard errors clustered at both the municipality level and the running variable Wit are within parentheses. The forcing variable Wit it is defined as max{Xit-1, Xi1918} where X is population size. Mean for the indicator of having at least union club is about 0.43 for local governments with representative democracy near the threshold. * Significant at the 10 percent level, ** Significant at the 5 percent level, *** Significant at the 1 percent level. Table 6. Estimates from the regression-discontinuity design on the number of association memberships (1) (2) (3) (4) (5) Panel A: Reduced form relationship between the number of association memberships and the instrument Git= 1[Wit ! 1500] Reduced form effect -72.957** (31.417) -109.169** (43.877) -76.213** (31.775) -141.325*** (47.148) -101.275** (42.667) Panel B: Reduced form relationship between direct democracy and the instrument Git= 1[Wit ! 1500] First-stage effect 0.409*** (0.084) 0.352*** (0.103) 0.409*** (0.089) 0.378*** (0.108) 0.385*** (0.102) Panel C: The causal effect of direct democracy versus representative democracy on the number of associations memberships Direct democracy =1 -178.337** (83.800) -310.209* (160.841) -186.144** (86.787) -373.924** (172.432) -262.800* (137.391) Observations 14,253 6,986 6,986 3,327 1,773 250 3 250 2 125 2 63 1 Bandwidth 500 Degree of polynomial in 3 the forcing variable Wit Notes: Standard errors clustered at both the municipality level and the running variable Wit are within parentheses. The forcing variable Wit it is defined as max{Xit-1, Xi1918} where X is population size. Mean for the indicator of having at least union club is about 0.43 for local governments with representative democracy near the threshold. * Significant at the 10 percent level, ** Significant at the 5 percent level, *** Significant at the 1 percent level. Table 7. Estimates from the regression-discontinuity design on civic engagement for the sample with with pre-treatment controls (1) (2) (3) (4) (5) Panel A: The causal effect of direct democracy on voter turnout without pre-treatment controls Direct democracy =1 -0.270*** -0.299** -0.293*** -0.321** -0.301** (0.076) (0.136) (0.082) (0.142) (0.127) Panel B: The causal effect of direct democracy on voter turnout with pre-treatment controls Direct democracy =1 -0.256*** -0.256** -0.270*** -0.283** (0.069) (0.119) (0.071) (0.127) -0.285*** (0.086) Observations 408 3,474 1,693 1,693 785 Panel C: The causal effect of direct democracy on the number of association memberships without pre-treatment controls Direct democracy =1 -187.353** -325.656* -194.957** -389.618** -274.018* (88.416) (170.495) (89.696) (181.041) (143.597) Panel D: The causal effect of direct democracy on the number of association memberships with pre-treatment controls Direct democracy =1 -159.207** -220.557** -95.335** -216.490* -161.552** (65.554) (104.018) (47.893) (110.560) (62.917) Observations 14,133 Bandwidth 500 Degree of polynomial in 3 the forcing variable Wit 6,936 6,936 3,302 1,752 250 3 250 2 125 2 63 1 Notes: Standard errors clustered at both the municipality level and the running variable Wit are within parentheses. The forcing variable Wit it is defined as max{Xit-1, Xi1918} where X is population size. Mean for the indicator of having at least union club is about 0.43 for local governments with representative democracy near the threshold. * Significant at the 10 percent level, ** Significant at the 5 percent level, *** Significant at the 1 percent level. Table 8. Estimates from the regression-discontinuity design on the number of memberships by type of association with pre-treatment controls (1) (2) (3) (4) (5) Panel A: The causal effect of direct democracy versus representative democracy on the number of union memberships Direct democracy = 1 -77.125** (39.091) -136.516* (77.099) -73.290** (36.888) -133.590* (79.782) -97.501** (43.513) Panel B: The causal effect of direct democracy versus representative democracy on the number of free church memberships Direct democracy = 1 -18.088 (11.245) -23.598 (16.857) -16.383 (11.216) -36.476* (19.842) -30.314** (13.489) Panel C: The causal effect of direct democracy versus representative democracy on the number of temperance lodge memberships Direct democracy =1 -10.640 (16.221) -41.180 (26.652) -11.001 (15.374) -21.257 (23.899) -27.427 (20.851) Observations 14,153 6,946 6,946 3,307 1,755 250 3 250 2 125 2 63 1 Bandwidth 500 Degree of polynomial in 3 the forcing variable Wit Notes: Standard errors clustered at both the municipality level and the running variable Wit are within parentheses. The forcing variable Wit it is defined as max{Xit-1, Xi1918} where X is population size. Mean for the indicator of having at least union club is about 0.43 for local governments with representative democracy near the threshold. * Significant at the 10 percent level, ** Significant at the 5 percent level, *** Significant at the 1 percent level. Figure 1. The “first-stage” relationship between direct democracy and the forcing variable 1 Direct_democracy .8 .6 .4 .2 0 -500 0 500 Figure 2. The reduced form relationship between the total number of members in the popular movement and the forcing variable 180 160 Nr of members 140 120 100 80 60 -500 0 500 Figure 3. The reduced form relationship between turnout and the forcing variable Turnout in percent of eligible voters .5 .4 .3 .2 .1 -500 0 500 Figure 4. The McCrary density test .0005 .0004 .0003 .0002 .0001 1000 1200 1400 1600 Population t-1 Notes: Estimation based on full data using McCrary (2008) test. 1800 2000
© Copyright 2025 Paperzz