WAITING FOR NAPOLEON? DEMOCRACY AND RECIPROCITY ACROSS SOCIAL GROUPS

Transcription

1 WAITING FOR NAPOLEON? DEMOCRACY AND RECIPROCITY ACROSS SOCIAL GROUPS Devesh Rustagi and Marcella Veronesi Abstract What explains large and persistent differences in reciprocity across social groups? This paper exploits variation in historical experience of democracy over space and time in Switzerland to highlight its strong positive association with reciprocity today. Individuals from regions that experienced democracy since the Middle Ages display stronger reciprocity than individuals from regions that acquired democracy only after the invasion by Napoleon. Because historical democracy was widespread in Swiss German but limited in Swiss French speaking regions, individuals from these groups differ widely in their reciprocity. The difference, however, disappears when we compare Swiss Germans and Swiss French from regions without historical democracy. These results are not capturing current institutions, beliefs, migration, historical dynasties, language and other group specific characteristics. Further results suggest that the emergence of historical democracy was due to idiosyncratic events and that its effect on reciprocity persists due to intergenerational transmission. JEL: C93, D02, N43, H41, Z10 Keywords: Democracy, reciprocity, social groups, public goods game, intergenerational transmission, Switzerland, von Zaehringen. Rustagi: Goethe University Frankfurt, Faculty of Economics and Business Administration, Theodor W- Adorno Platz 4, D Frankfurt am Main, Germany. rustagi@econ.uni-frankfurt.de. Veronesi: University of Verona, Department of Economics, Via Cantarane 24, Verona, Italy; and Center for Development and Cooperation (NADEL), ETH Zurich. marcella.veronesi@univr.it. We acknowledge financial support from the Professorship of Environmental Policy and Economics at ETH Zurich, funds from Prof. Dr. Willi Zimmermann, and Alfons and Gertrud Kassel-Stiftung. This study has been realized using data from the Swiss Household Panel (SHP), which is based at the Swiss Centre of Expertise in the Social Sciences FORS. We are grateful to seminar participants at Bocconi University, Development Workshop at Mannheim, ETH Zurich, Goethe University Frankfurt, University of Geneva, HEC Lausanne, University of Marburg, LMU Munich, University of Lund, Nagoya City University, Social Norms Workshop at University of East Anglia, 11 th NCBEE Conference, 3rd International Meeting on Experimental and Behavioral Social Sciences, 57 th SIE Conference, 5th AERE Conference, 5th IAERE Conference, and 22nd EAERE Conference for helpful comments. Our special thanks also to Simon Hess, Nicola Fuchs-Schündeln, Saumitra Jha, Michael Kosfeld, Peter Martinsson, Emilia Soldani, Rajesh Ramachandran, Noam Yuchtman, and Marco Zanoli for helpful discussions. 1

2 I. Introduction Reciprocity is viewed as playing an important role for a variety of social and economic outcomes. 1 These include public goods provision, teamwork in organizations, functioning of incentives, charitable giving, tax compliance, and support for redistributive policies. 2 Defined as a tendency to respond to acts perceived to be kind with kindness and those perceived to be mean with meanness, reciprocity differs fundamentally from behavior that is due to benefits from repeated interaction and reputation formation (Fehr and Gächter 2000, Sobel 2005, DellaVigna 2009). 3 Further evidence shows that reciprocity differs widely across groups and forms a key element of social capital that is viewed as fundamental for collective action and economic development (Putnam et al. 1993, Platteau 2000, Bowles and Gintis 2002, Guiso et al. 2011). Yet, our understanding of what social and economic conditions drive differences in reciprocity and why do these differences persist over time remains poor. In this paper, we exploit spatial and temporal variation in historical experience of democracy to study how this associates with reciprocity today across individuals from different social groups. The interplay between democracy and reciprocity is discussed widely in the social sciences (Putnam et al. 1993, Rodrik 2000, Uslaner 2003, Tabellini 2010). According to these studies, democratic institutions are hypothesized to foster reciprocity via allowing for greater participation in decision-making and constraints on the executive. This encourages individuals to interact, deliberate, and learn to compromise, thereby changing the type of people we are by making us less selfish and more public-spirited. As Rodrik (2000) points out when individuals meet and deliberate, they come to understand each other s view-points, develop empathy, recognize the value of moderation, internalize the common interest, and de-emphasize narrow self-interest. Tabellini (2010) and Platteau (2000) further argue that when the power of the elite is constrained by checks and balances, it fosters participation by bringing together individuals from different interest groups for a common cause. In contrast, in autocratic regimes, strong hierarchy of privileges that are accessible only to a small group of people are likely to foster opportunistic free riding as opposed to a disposition for reciprocity. In summary, exposure to democracy is hypothesized to encourage individuals to 1 See for instance Putnam 1993, Ostrom et al. 1999, Platteau 2000, Fehr and Gächter 2000, Sobel 2005, DellaVigna For motivations underlying reciprocity see theoretical contributions by Fehr and Schmidt 1999, Bolton and Ockenfels 2000, Dufwenberg and Kirchsteiger 2004, Falk and Fischbacher For example: Frey and Meier 2004, Bandiera et al. 2005, Rustagi et al. 2010, Bowles and Polonia-Reyes Reciprocity is different from altruism, which is a form of unconditional kindness that does not emerge as a response to altruism received (Fehr and Gächter 2000). Reciprocity is also considered as one of the key motives underlying trust and trustworthiness (Fehr 2009). 2

3 set aside narrow self-interest and instead develop values of reciprocity. These values could then persist over time due to intergenerational transmission (Bisin and Verdier 2011). Although field and laboratory studies show that stronger cooperation emerges when democratic participation is possible (Bardhan 2000, Frey 1998, Dal Bo et al. 2010, Sutter et al. 2010), thus far the connection to reciprocity is only suggestive. We attempt to fill this gap by conducting our study in Switzerland, which provides a unique natural laboratory. Switzerland offers large variation in the experience of historical democracy as well as linguistic groups. We show that individuals from regions that experienced democracy since the Middle Ages display stronger reciprocity today than individuals from regions who experienced democracy only after the invasion by Napoleon (see Figure 1). Because historical democracy was widespread in Swiss German than Swiss French speaking regions, we observe a strong gap in reciprocity across individuals from these two regions today (see Figure 2). Moreover, the gap disappears once we compare Swiss Germans and Swiss French from regions without historical democracy. These results are not capturing current institutions, beliefs, migration, dynasty, language and other group specific characteristics. The Swiss setting is ideal for our study because of differences across space and time in the experience of historical democracy that included in its ambit both urban and rural areas. Starting 13 th century, many rural and urban areas of Switzerland developed early forms of democracy that offered local people opportunities for self-governance, for instance via the election of council members and voting on policies in public assemblies. In contrast, the remaining areas were subjected to feudalistic institutions, where individuals could neither elect their council members nor vote on public policies. Instead, in these areas council members were either appointed or elected from among patrician families, at times even for life. This created a mosaic of landscapes with and without opportunities for self-governance even within a canton whose boundaries today are mostly coterminous with historical cantonal boundaries in the 16 th century. 4 This situation persisted for hundreds of years until Napoleon invaded Switzerland in 1798 and issued the Act of Mediation in 1803 allowing every region the possibility of self-governance. We refer to the experience of democracy in the period preceding 1803 as historical and the period thereafter as modern. While modern institutions of democracy vary only at the cantonal level, historical democracy varies even within cantons, allowing us to use canton fixed effects to separate historical from modern institutions of democracy and to account for historical differences across cantons. 4 A canton is an administrative unit in Switzerland that is equivalent to a state. 3

4 Measuring reciprocity using real world interactions is challenging because behavior that seems consistent with reciprocity might be actually be due to confounding motives arising from repeated interaction and reputation formation benefits (Fehr and Gächter 2000, Sobel 2005, DellaVigna 2009). Moreover, individuals from different groups might have the same disposition for reciprocity and yet behave differently because of differences in their beliefs about the cooperation of others. For instance, reciprocators with optimistic beliefs about the cooperation of others are expected to cooperate but reciprocators with pessimistic beliefs are expected to defect. In such situations, one might wrongly infer a reciprocator with pessimistic beliefs as a free rider resulting in measuring multiple equilibria rather than the difference in reciprocity per se. This concern is accentuated when the comparison is across individuals from diverse groups, as different groups might coordinate on different equilibria. To resolve these challenges, we use a controlled behavioral experiment to measure reciprocity. Following Fischbacher et al. (2001) and Fischbacher and Gächter (2010), we implement a standard one-shot anonymous public goods game in the strategy method to preclude future material benefits as well as beliefs from playing a role. 5 The experiment is conducted online among a representative sample of Swiss Germans and Swiss French. In the experiment, individuals are paired in a two-person experimental group. Each player has to decide on his / her contribution to a public good conditional on a set of contribution decisions of another player. Individuals who are reciprocators are expected to match their contribution in the contribution of the other player, whereas free riders are expected to contribute zero in every decision regardless of what the other player does. Following the literature, we use the Spearman correlation between self and other players contribution as a revealed measure of reciprocity the stronger the Spearman correlation, the stronger is the disposition for reciprocity. The average reciprocity in our sample is 0.65 (s.d. 0.55). Since different areas in Switzerland acquired historical democracy at different points in time, we develop measures of historical democracy at both the extensive and the intensive margin. Our data show that Swiss Germans are nearly four times more likely to be from municipalities with historical democracy than Swiss French and also with a longer duration by over 200 years. Following aforementioned studies (Platteau 2000, Rodrik 2000, Tabellini 2010), postulating a positive association between democracy and reciprocity, we expect Swiss Germans to have stronger reciprocity than Swiss French. Our results confirm this. The propensity for reciprocity among Swiss Germans is 0.73, 5 This experiment has been used in a variety of countries to measure reciprocity and has also been validated in studies by Rustagi et al. (2010) and Kosfeld and Rustagi (2015). 4

5 while that among Swiss French is only The gap remains large in magnitude at 0.29 points and is highly significant even when we control for a rich set of individual, municipal and cantonal level covariates or even canton fixed effects. This means that for each additional 10 Swiss Francs contributed by the other player, Swiss Germans contribute 2.1 Swiss Francs more than Swiss French, which is 57 percent of the contribution by Swiss French. If this gap is due to group specific differences in historical democracy, then it should disappear when historical democracy is lacking. We check this via a falsification test in which we restrict our sample to Swiss Germans and Swiss French who are from municipalities that did not experience historical democracy. The magnitude of the gap in reciprocity drops from 0.29 to 0.04 and is no longer statistically significant. The above results point towards a positive association between reciprocity and historical democracy. However, these results could be capturing the effect of omitted variables and might also be prone to reverse causality. We address these concerns in several steps. First, we include a rich set of individual and municipal / cantonal controls that are hypothesized in the literature to matter for reciprocity. These include age, education, gender, household income, religion, politics, geography, income inequality, GDP per capita, share of foreigners, and an index of business competition. Second, we also consider the historical counterpart of these variables to ensure that we are not capturing differences in historical development and education. Third, we introduce a variety of fixed effects to ensure that our estimates are not capturing differences in canton specific factors (such as modern democracy), historical dynasties with which the municipalities were associated at different points in time, as well as social groups, for instance by dropping Swiss French altogether from our sample. Fourth, we show that our results hold when an epidemiological approach is considered (Fernandez 2010). Fifth, we confirm our findings using an alternative dataset the Swiss Household Panel where reciprocity is measured using data on values such as how justifiable it is to cheat on taxes, lie in own interest, and claim state benefits not entitled to (Guiso et al. 2011). We find that individuals from municipalities with historical democracy display higher reciprocity by 0.42 points. This difference is not just statistically significant at the 1 percent level, but it is also economically significant, especially given that the average reciprocity among individuals from municipalities without historical democracy is Moreover, one standard deviation increase in the duration of historical experience (244 years) is associated with an increase in reciprocity by 0.22 points. This implies that for each additional 10 Swiss Francs contributed by the other player, individuals from municipalities with historical democracy contribute 3.6 Swiss Francs more than individuals from municipalities without, 5

6 which is over 100 percent of the contribution by the latter. Similar estimates are obtained when we introduce fixed effects for cantons, historical dynasties, and social groups. Following Oster (2016) we find that selection on unobservables would have to be at least twice as large as selection on observables for our results to be entirely attributed to a selection bias. An open question is why did some areas acquire democracy but others did not. Swiss historiography concurs that the emergence of historical democracy was not systematic but rather due to idiosyncratic events. One event with wider consequences was the abrupt extinction of the Zaehringen dynasty when its last duke died of natural causes without an heir. This led the German king to grant imperial immediacy to areas under the Zaehringen rule. Historians consider this to be a watershed event that gave municipalities the impetus to eventually move towards democracy (Zschokke 1860, McCrackan 1908, Eugster 2015, Schultheiss 2015). The extinction of the Zaehringen line offers plausibly exogenous variation in the emergence of democracy, especially after verifying that areas with and without the Zaehringen rule do not differ in climate and soil suitability indices for agricultural crops in the pre-1500 period, as well as in altitude and rugged land (Galor and Özak 2016). Our results show a large, positive, and significant association of historical democracy with the Zaehringen rule in the Middle Ages. We use the extinction of the Zaehringen as an instrument for historical democracy. The exclusion restriction is that the Zaehringen rule did not have any direct effect on reciprocity. We show that when historical democracy is lacking, areas with and without the Zaehringen rule do not differ in their reciprocity. Also, the association of the Zaehringen rule with reciprocity disappears completely when we explicitly control for historical democracy. The instrumental variables estimates of the effect of historical democracy on reciprocity is also strongly positive, statistically significant, and similar in magnitude to its OLS counterpart. These findings raise questions over the long run nature of the association between historical democracy and reciprocity. We provide suggestive evidence on the role of intergenerational transmission by examining whether the family background of an individual matters in explaining the variation in reciprocity. For this purpose, we collect data on the region of birth and region of stay of parents and grandparents of individuals in our sample. We then build indices to characterize individuals on their degree of Swiss German family background. Our results show that individuals with stronger Swiss German parental or grandparental background display stronger reciprocity. 6

7 Our study connects to the literature showing the long-term effect of important historical events, such as slavery, Jewish persecution in the middle ages, and the introduction of plough in agriculture on trust, violence, and honesty today (Tabellini 2010, Nunn and Wantchekon 2011, Voigtländer and Voth 2012, Alesina et al. 2013, Jha 2013, Grosjean 2014, Becker et al. 2016, Lowes et al. 2017). We contribute to this literature by studying the association between historical democracy and reciprocity. A paper closely related to ours is by Guiso et al. (2016) who examine the association of free city-state experience in Italy in the Middle Ages with the number of non-profit organizations, presence of an organ donation organization, and cheating in math test today. Our paper distinguishes itself from previous studies in several ways. First, unlike the free city experience, which as its name implies was restricted only to cities, historical democracy in Switzerland was comprehensive in covering both rural and urban areas. Second, we go beyond previous studies by conducting not just between but also within state analysis. This allows us to highlight the importance of historical democracy over and above current institutions of democracy, as well as other canton specific differences. Third, in a similar vein, we exploit variation in historical democracy within historical dynasties and social groups to account for dynasty and group specific factors. Fourth, our measure of reciprocity is from a controlled experimental environment, which precludes confounding motives arising from future interaction benefits and beliefs about the contribution of others. As highlighted by Fehr and Gächter (2000), Sobel (2005), and DellaVigna (2009), this step is necessary to distinguish reciprocity from alternative explanations. Fifth, we exploit the plausibly exogenous extinction of the house of Zaehringen to shed some light on the emergence of democracy (see for instance Aidt and Franck 2015). Finally, our study also points towards the role of intergenerational transmission as a plausible channel behind the persistence of historical effects. The paper is organized as follows. Section II describes the field setting and historical background. Section III describes data including our measures of social groups, historical democracy, and experimental measures of reciprocity. Sections IV and V present results on the association of reciprocity with social groups and historical democracy including the role of intergeneration transmission as a plausible channel. Section VI offers concluding remarks. 7

8 II. Field Setting A. Social Groups The Swiss setting allows us to study differences in reciprocity across individuals from different social groups using a within country and also a within canton setting. Switzerland is home to four main native European languages: German, French, Italian, and Romance that comprise 64, 23, 8, and 0.5 percent of the Swiss population (Swiss Federal Statistical Office, see Figure 3) 6. The focus of this study is on the speakers of German and French, which are also the official languages in 25 out of the 26 states of Switzerland that are called cantons. The share of these two groups in the Swiss population is 87 percent and has remained stable both within and across cantons since the foundation of modern Switzerland in 1848 (Swiss Federal Statistical Office). Our focus on linguistic groups as social groups draws from studies in sociology and economics outside of and within Switzerland that show language to be a powerful marker of identity (Gumperz 1982, Giles and Johnson 1987, Clyne 1995, Eugster et al. 2011, Clots- Figueras and Massella 2013). In fact, differences between Swiss Germans and Swiss French are regularly emphasized in the Swiss media. For instance, a newspaper called Südostschweiz (2000) described these two groups as two fronts that stand opposite to each other. Reflecting a similar view, a Swiss weekly magazine Die Weltwoche wrote (2004) Switzerland consists of two halves: On the left hand side is the casual west (Swiss French), on the right hand side the stricter east (Swiss German), and again (2007) It is unequivocal among political geographers that there exists a dividing line which separates our country (Switzerland) into two mental regions. The linguistic diversity of Switzerland dates back to the 7 th century, when Germanic tribes called the Alemanni occupied the northern and eastern part of Switzerland. The westward movement of the Alemanni in the 7-8 th century resulted in the formation of a language frontier between the German and the French speaking parts that is now called Roestigraben. This historically determined language border passes through three bilingual cantons: Valais, Fribourg, and Bern (Figure 3). 7 Notably, this within canton variation in 6 See 7 Within canton variation in language in the remaining cantons is negligible. For instance, in the Swiss Household Panel comprising thousands of Swiss households, the share of French speakers in the most populous German-speaking canton (Zurich) is a mere percent. Similarly, the share of German speakers in the most populous French-speaking canton (Vaud) is only 0.01 percent. The shares are either the same or even smaller in the remaining cantons; in 13 out of 26 cantons, it is even zero. 8

9 linguistic groups is not commensurate with geographical (Eugster et al. 2012) or genetic variation (Buhler et al. 2012). 8 The Swiss setting thus offers us the possibility to study the association of linguistic groups with reciprocity, controlling for country and at times canton specific factors. Note that differences in reciprocity across linguistic groups are unlikely to be due to differences in language per se because both German and French have Tu and Vos differentiation and are also no pronoun drop languages (Galor and Özek 2016). B. Historical Democracy Switzerland offers a unique setting with respect to variation in historical democracy over space and time. This variation exists within historical as well as modern cantonal boundaries, and also within social groups. Moreover, this phenomenon is not restricted to cities but also includes in its ambit rural areas. The first Swiss federal constitution offering cantons the rights to extensive selfgovernance was adopted in However, the historical antecedent of these rights date back to the Middle Ages. Before the 13 th century, several noble houses appointed by the Holy Roman Emperor, such as von Zaehringen, Kyburg, Habsburg, and Savoy, managed large parts of Switzerland as their feudal property. But starting 13 th century, idiosyncratic events, in particular the extinction of the House of Zaehringen, led to many such areas acquiring imperial immediacy (Reichsunmittelbarkeit) from the emperor. 9 This freed many feudal areas from the authority of local nobles and offered its residents limited opportunities for selfgovernance. Over time, some of these areas managed to bolster self-governance by acquiring considerable rights and powers, for instance to levy taxes, possess a seal, administer justice, and appoint their own council and magistrate (Zschokke 1860, McCrackan 1908, Kamm 2017, Landolt 2017a, 2017b, Stadler 2017, Suter 2017, Zahnd 2017). These rights were defended via the formation of an alliance called the Old Swiss Confederacy, which comprised 13 members and several allies. 10 In contrast, the remaining areas of Switzerland 8 The main geographical border formed by the Alps runs along the southwest-northeast axis, whereas the language divide runs along the north-south axis. Similarly, genetic differentiation in Switzerland besides being very small is not across linguistic groups but geographical boundaries created by the Alps. For this reason we do not include Swiss Italians in our sample, as these are found across the Alps. Another reason is that Swiss Italians are largely confined to a single canton (Ticino). 9 In some cases this was the German king who eventually became the emperor in the following years. 10 The 13 members included the rural cantons of Uri, Schwyz, Unterwalden (Obwalden and Nidwalden), Glarus, Zug, and Appenzell (Inner and Outer Rhodes), and the city cantons of Zurich, Bern, Lucerne, Fribourg, Solothurn, Basel (City and Country), and Schaffhausen. 9

10 mostly came to be administered as subjugated territories of the Old Swiss Confederacy. This created a mosaic of areas with and without historical democracy even within cantons, whose boundaries today are mostly coterminous with cantonal boundaries in the 16 th century. While most members and allies of the Old Swiss Confederacy handled communal obligations through deliberation and consensus, barring a few exceptions, most subjugated areas had little say in decision-making. This status remained for the most part uninterrupted for hundreds of years until the invasion of Switzerland by Napoleon in 1798, which resulted in the dissolution of the Old Swiss Confederacy and the liberation of its subjugated territories. The Act of Mediation issued by Napoleon in 1803 made every member and ally of the Old Swiss Confederacy as well as their subjugated territories full members of Switzerland and allowed them to have their own representative governments (Frankhauser 2009). That being said, it would be a mistake to view medieval Switzerland as a place where modern democratic principles, as in equal rights for all, were enshrined. However, as McCracken (1908) notes relative to other areas of Europe during the Middle Ages early Swiss were, nevertheless, the best democrats of their day, unconscious, but practical exponents of the virtues of self-governance and that...the Swiss States, both country districts and towns, were organized upon democratic principles. The emergence of historical democracy in the urban and rural areas of Switzerland was due to idiosyncratic events. In some areas democracy set foot due to the opening of the Gotthard pass connecting Switzerland to Germany and Italy, which made previously remote places strategic transit locations for wars against Lombard cities, pope Innocent IV, and to limit the Habsburg expansion (Heinemann 2003, Schelbert 2014, Stadler 2011, 2017). However, in most areas it was the extinction of the Zaehringen dynasty from the absence of an heir, which led to the emergence of historical democracy (McCrackan 1908, Eugster 2015, Schultheiss 2015). As McCrackan (1908) notes the extinction of the house of Zaehringen came most opportunely, for it is entirely within the range of possibility, that, otherwise, the state they had erected, might have become a principality, or even a monarchy, as enduring as any of those which surround Switzerland today. Broadly speaking, historical democracy in Switzerland was implemented via voting in public assemblies and election of city councils. In rural areas, such as the cantons of Uri and Schwyz, eligible male citizens voted directly on laws and decrees in open-air assemblies called Landsgemeinden (Stadler 2008). This form of democracy was well established by the 14 th century in the rural cantons and even survived into modernity (late 19 th -20 th century) until the introduction of anonymous voting in all but two rural cantons (Stadler 2008). In the 10

11 cities, historical democracy involved widening of the power base via limiting steadily the clout of bishops and aristocrats in the appointment of city councils and via the creation of large assemblies comprising individuals that offered citizens a larger role in governance (Berner 2006, Holenstein 2014). 11 Regardless of the way historical democracy was practiced in Switzerland, it involved elements of both deliberation as well as constraint on the executive. Involving people to either vote directly on policies or offering citizens a larger say in decision making via the institution of greater councils shows stronger presence of deliberation and constraint on the power of aristocrats (Holenstein 2014). For instance, in one of the public assemblies, a referendum forbid powerful monasteries from using the common grazing land for free and ordered them to pay the same tax per cow as local farmers or else be excluded from using the common (McCrackan 1908). Similarly, in many cities council members were prevented from holding on to the municipal offices for life or bequeath these to their children (McCrackan 1908, Schultheiss 2015, Zahnd 2017). Because historical democracy varies within cantons but modern democracy varies at the cantonal level, we are able to disentangle their effects on reciprocity by introducing cantons fixed effects. In addition, there is also within group variation in historical democracy, which allows us to control for group specific differences. In addition, it also lets us construct falsification tests that investigate whether the gap in reciprocity across Swiss Germans and Swiss French exists even when both are from municipalities without historical democracy. III. Data and Descriptive Statistics We collect data from several different sources. Data to measure reciprocity were collected at individual household level using a behavioral experiment. This was followed by an in-depth post-experiment survey with the same individuals to collect data on native language, sociodemographic characteristics, migration, and family background. These individuals were selected in a representative manner by the institute for opinion research LINK, the largest survey agency in Switzerland. We match these individual level data with data we collected on historical democracy at the municipal level using the online Historical Lexicon of Switzerland (e-hls) and a book by Gasser (1932), current and historical municipality / canton specific covariates from the 11 In the Middle ages, the Swiss city states were much smaller than their Italian counterparts. For instance, in the population of Zurich was around 8000, while that of Lucerne was

12 Swiss Federal Statistical Office, data on a competition index by Credit Suisse, data on soil and climate suitability by Swiss Federal Office for Agriculture 12, and data on historical dynasties on the basis of maps by Ammann and Schib (1958) and Putzger (2004). We refer to this dataset as our experimental sample. It comprises 262 Swiss individuals covering 174 municipalities from 23 cantons. 13 We discuss our sample construction strategy in Appendix A.I, whereby we show that individuals and municipalities that are in our sample have similar characteristics as individuals and municipalities that are not (Tables A1-A5). These characteristics include gender, age, education, household income, religious and political orientation, altitude, and Gini index of income inequality. In addition, we also use wave 13 of the Swiss Household Panel (SHP) as an alternative data source. The SHP has 1,859 Swiss individuals from the same municipalities that are covered in our experimental sample. Note that SHP does not have experimental measures of reciprocity but values that are closely connected to reciprocity (see section IV.A). In both the samples, our focus is on Swiss nationals. Below we describe our measures of social group, historical democracy, and reciprocity, followed by descriptive statistics. A. Social Groups We use the native language(s) of an individual to identify the social group to which an individual belongs. This was elicited via the post-experiment survey in which multiple answers were allowed. Individuals could choose from Swiss German, High German, French, Italian, and other; while choosing the latter individuals had to state the language in an empty box. Our sample comprises 195 Swiss Germans (74 percent) and 67 Swiss French (26 percent); these shares reflect the share of the two linguistic groups in the Swiss population excluding Swiss Italians, which are 75 and 25 percent, respectively. 14 To further verify that the native language captures an important dimension of an individual s social group, we find that over 90 percent of our respondents identify with the beliefs and values of their linguistic Other than the canton of Ticino, which we deliberately excluded because it is exclusively Swiss Italian speaking, the remaining two excluded cantons are Appenzell-Innerrhoden and Nidwalden - the smallest cantons in Switzerland by population percent of the individuals in our sample (243 individuals) speak only one native language, either Swiss German or Swiss French. Of the remaining 7 percent (19 individuals), 80 percent (15 individuals) speak Swiss German as well as a non-swiss native language (for example, Russian). We assign these individuals to the Swiss German group. The remaining four individuals speak both Swiss German and Swiss French and are assigned to the group that corresponds with the official language of the canton of birth. Our results are robust even when we assign these individuals to the other group. Note that for 99 percent of the individuals in our sample the native language corresponds with the official language of the municipality. 12

13 group and that these responses do not differ by groups. In another question, we also ascertained the extent to which individuals from the two linguistic groups consider themselves as Swiss and find no difference in this dimension (p-value = 0.75). B. Historical Democracy We measure historical democracy at the municipal level, which is the smallest administrative unit in Switzerland at which individuals live and socially interact. Our primary data sources for historical democracy are the online Historical Lexicon of Switzerland (e-hls) and the study by Gasser (1932). We complement these sources with books on the history of Switzerland by Zschokke (1860) and McCrackan (1908). Our measures consider both whether a municipality experienced historical democracy or not, as well as the duration of this experience. Accordingly, we have two measures: Experience is a dummy variable, which takes the value of one if a municipality allowed individuals stronger opportunities for self-governance from the Middle Ages until 1803, zero otherwise. Table 1 reports summary statistics and shows that 54 percent of the respondents in our sample are from municipalities with historical democracy. Duration is a continuous variable, which is measured as the difference in years between 1803 and the date a municipality acquired historical democracy. For the rural areas, we use the date when Landsgemeinde got established. For the urban areas, we use the date starting which the city could elect its own council. For some subjugated areas these dates are not available, so we use the date around which they came to be administered by the members of the Old Swiss Confederacy or their allies, as this was the time when major changes in selfgovernance took place. Table 1 shows that the average duration of historical democracy in our sample is about 258 years (2.58 hundreds of years). C. Measuring Reciprocity Measuring reciprocity using observational data is challenging. In the field, individuals might behave in a manner that appears to be consistent with reciprocity but is actually due to benefits from repeated interaction and reputation formation (Fehr and Gächter 2000, Sobel 2005, and DellaVigna 2009). Also, it is plausible that individuals from different social groups have the same propensity for reciprocity and yet differ in their behavior because of 13

14 differences in their beliefs about the behavior of others (Fischbacher and Gächter 2010). For example, in a prisoner s dilemma, reciprocators from groups with pessimistic beliefs about the cooperation of others might defect, whereas reciprocators from groups with optimistic beliefs about the cooperation of others might cooperate, resulting in multiple equilibria. This could result in the misclassification of reciprocators with pessimistic beliefs as free riders. These challenges warrant the use of a behavioral experiment. Following previous studies (Fischbacher et al. 2001, Fischbacher and Gächter 2010), we invite individuals to take part in a one-shot anonymous public goods game that is implemented in the strategy method. This allows us to circumvent problems arising from confounding of reciprocity with other motives and beliefs. This method has been used in several field settings (see for instance Kocher et al. 2008, Herrmann and Thöni 2009, Rustagi et al. 2010) and has also been externally validated by Rustagi et al. (2010) and Kosfeld and Rustagi (2015). The experimental instructions and procedures are in Appendix B. Behavioral experiment. In the experiment, two players are randomly assigned to an experimental group. Each player receives an endowment of 100 CHF (Swiss Francs) and has to decide on his/her contribution to the public good. Players could contribute any amount from 0 to 100 in units of 10 CHF. Any amount in the public good is increased by 1.5 times and then distributed equally between the two players, regardless of their contribution. Formally, the payoff of player i, where i {1, 2}, is given by: (1) Πi = 100 Ci (C 1 + C 2 ), where 100 is the endowment received at the start of the game, C i is the contribution of player i to the public good, 0.75 is the marginal per capita return from the public good, and C 1 + C 2 is the total contribution to the public good. Given the marginal per capita return, each Swiss Franc contributed by a player to the public good yields only 0.75 back. Thus, it is individually rational for a player to contribute nothing to the public good. However, because 2*0.75 > 1, it is socially optimal to contribute the entire endowment. This creates a cooperation dilemma. The game involved two decisions: unconditional and conditional. In the unconditional decision, players decided simultaneously on their contribution to the public good and thus beliefs play a role. In the conditional decision, we implemented the strategy method, whereby each player had to decide on his/her contribution for each of the 11 possible contribution 14

15 decisions of the other player, thereby making beliefs redundant. A die was rolled to determine the player for whom the conditional decision is relevant, which is then matched with the unconditional decision of the other player to determine payoffs. Measures of reciprocity. If individuals are motivated only by their own payoff then they should contribute zero in each of the 11 conditional decisions. However, if individuals have a disposition for reciprocity then they should increase their contribution in response to the increasing contribution of the other player. Following Fischbacher et al. (2001) and Fischbacher and Gächter (2010), we use the Spearman correlation between self and the other players contribution in the conditional decision as a measure of an individuals reciprocity - the higher the Spearman correlation the higher is the propensity for reciprocity. The average propensity to reciprocate in our sample is 0.65 points (s.d. 0.55) (panel C, Table 1). We also categorize individuals into types following the guidelines proposed by Fischbacher et al. (2001) and Fischbacher and Gächter (2010). Individuals for whom the p- value of the Spearman correlation is are classified as reciprocators. Individuals who contribute zero in all the 11 decisions are categorized as free riders, those who contribute 100 in all the decisions as altruists, and those who contribute the same amount that is different from zero and 100 in all the decisions as flat contributors. Individuals who do not fall into any of the above three categories are classified as other types. We find that 66 percent of the individuals in our sample are reciprocators, 7 percent are free riders, 4 percent are altruists, 5 percent are flat contributors, and the remaining 18 percent are unclassifiable other types. As an alternative dependent variable, we construct a dummy for reciprocity, which equals one if an individual is a reciprocator and zero otherwise. In addition to the experiment described above, we also conducted a second follow up experiment, which is described in section V.G. One of the two experiments was randomly selected to determine the payoff. 15 After the experiments, individuals took part in an in-depth post-experimental survey. Both the experiment and the post-experiment survey were pretested for respondents understanding of the experiment, questions, and the duration of the study. We also made sure that native language speakers from within Switzerland wrote the instructions. In addition, before the actual decision, individuals had to answer three control questions on game comprehension correctly. We use this to gauge an individuals understanding of the experiment and to ensure that we are not simply capturing noise. 15 We pay a random sub-sample of participants following procedures by Bettinger and Slonim (2007). 40 people were randomly selected for payment, of which 50 percent are Swiss Germans and 30 percent Swiss French. On average, subjects earned CHF

16 D. Descriptive Statistics We present our descriptive results in two steps. First we discuss the correlation between social groups and historical democracy, followed by the association of reciprocity with social groups and historical democracy. Our data show that 67 percent of the Swiss Germans are from municipalities with historical democracy, but the corresponding share is only 16 percent for Swiss French. When we consider the duration of historical democracy, we find that it averages 319 years for Swiss Germans and 82 years for Swiss French. This means that Swiss Germans are nearly four times more likely to be from municipalities with historical democracy than Swiss French and also for a longer duration by about 240 years. These differences are highly significant (pvalues < 0.001). The gap in historical democracy across the two groups remains large even when we control for covariates (Table A6 in Appendix A). In line with the literature postulating a positive association between democracy and reciprocity, we expect Swiss Germans to have stronger reciprocity than Swiss French. Accordingly, we first show in Figure 2a that reciprocity is indeed higher among individuals from municipalities with historical democracy (0.83 points) than municipalities without (0.42 points). Figure 2b shows that the propensity for reciprocity is much higher among Swiss Germans (0.73 points) than Swiss French (0.41 points). The magnitude of these differences is large and significant at the 1 percent level. We test econometrically the association of reciprocity first with social groups and then historical democracy in the following sections. IV. Social Groups and Reciprocity We tease out systematically the difference in reciprocity across the two social groups using ordinary least squares (OLS) of the following form: (2)!!"# =!! +!!!!"!"# +!!!"#!! +!!!"!!!+!!!"#, where R imk is the reciprocity of individual i from municipality m in canton k. It is measured as the Spearman correlation between own and other players contribution in the conditional decision of the public goods game. Our key variable of interest is SG, which is measured as a binary variable that equals one if an individual i s native language is Swiss German and zero otherwise (Swiss French). To ensure that our Swiss German dummy is not capturing the 16

17 effect of other plausibly correlated variables, we include a number of control variables that are hypothesized to matter for reciprocity in the literature. X is a vector of individual sociodemographic characteristics, such as age, education, gender, household income, religion, and politics. M is a vector of municipality specific variables including altitude and Gini coefficient for income inequality, and cantonal level GDP per capita.!!"# is an idiosyncratic error term clustered at the cantonal level. Owing to the small number of clusters (n = 23), the standard errors could be biased downwards leading to a higher probability of Type I error. Therefore, we also estimate standard errors using the wild cluster bootstrap-t procedure by Cameron et al. (2008). A. Main Results Table 2 reports results from the estimation of equation 2 (Table A7 in Appendix A shows coefficients on controls). For the coefficient on the Swiss German dummy, we additionally report in square brackets p-values obtained using the wild cluster bootstrap procedure. Column 1 is without control variables and shows that the Swiss German dummy has a positive coefficient (0.32), which is significant at the 1 percent level. In column 2, we exploit our rich dataset and control for a host of important individual level covariates. The coefficient on the Swiss Germany dummy retains its magnitude as well as significance. In column 3, we additionally account for differences in geography, inequality and economic environment at municipal and cantonal level. As before, the coefficient on the Swiss Germany dummy remains remarkably robust. With respect to Swiss French as the benchmark category, Swiss Germans display stronger reciprocity by 0.29 points. This difference is economically large, especially given that the average reciprocity among the Swiss French is 0.41 points. It is also nearly one-half of the average reciprocity in our full sample (0.65 points). In monetary terms, for each additional 10 Swiss Francs contributed by the other player, Swiss Germans contribute 2.1 Swiss Francs more than Swiss French, which is 57 percent of the contribution by Swiss French. 16 We implement a number of robustness checks on the association between social groups and reciprocity. We first show that the gap is not due to differences in cantonal level institutions. As mentioned before, because within canton variation in linguistic groups occurs only in the three Roestigraben cantons, we exploit this sub-sample and introduce canton fixed 16 This is computed from a regression of contribution by an individual on the contribution by the other player in the public goods game, after controlling for the full set of covariates. The average contribution for Swiss Germans and Swiss French are 6.1 and 3.7 Swiss Francs, respectively. 17

18 effects to control for cantonal specific differences in institutions. The results reported in column 1 of Table A8 shows that the Swiss Germany dummy has a positive coefficient, which is significant at the 5 percent level. In column 2, when we drop these three cantons from our sample, the coefficient on the Swiss German dummy turns out to be 0.31 (s.e. 0.07), which is similar in magnitude and significance to our main result. To mitigate further the scope of influential cantons, we show that our results hold when we drop the highly urbanized cantons of Zurich in column 3 and Geneva in column 4. In addition, when we drop one canton at a time the coefficient on the Swiss German dummy varies from 0.26 to 0.33, and remains highly significant. We also reproduce our main results using alternative measures of reciprocity and social groups in Table A9. We start by using as our dependent variable a dummy for whether an individual is a reciprocator or not. Column 1 reports the results and shows that Swiss Germans are, on average, 18 percentage points more likely to have a disposition for reciprocity than Swiss French. We then change the way we measure social groups by taking into consideration the extent to which an individual identifies as belonging to one s own and other linguistic groups. Following Aron et al. (1991), we show respondents five Venn-like diagrams with different degrees of overlap between the respondent and the social group, ranging from 1 (no overlap) to 5 (complete overlap) (see Figure A1, Appendix A). Each respondent faces four such diagrams, one each for Swiss German, Swiss French, Swiss Italian, and foreigner. We focus on the degree of Swiss German identity and study how this associates with reciprocity using the degree of Swiss French identity as a benchmark, after controlling for the degrees of Swiss Italian and foreigner identity. Column 2 reports the results and shows that the coefficient on the degree of Swiss German identity is 0.052, which is significant at the 5 percent level. A one standard deviation increase in the degree of Swiss German identity (1.14) is associated with an increase in reciprocity by 0.06 points. We also show in Tables A10 and A11 that we are not capturing other cantonal and municipal level differences. These include current share of foreigners, an index of business competition, and proxies of historical development and human capital. Lastly, we add several individual level controls in Table A12, including a dummy variable for naturalized Swiss citizen and whether an individual has migrated within Switzerland or not 17, Big Five personality traits, and game comprehension, measured as the number of correct responses to 17 Migration in our sample occurs only within Switzerland and within a linguistic group, and is therefore not problematic, as individuals are assigned to social groups on the basis of their native language. The only exception is three individuals who have residence across the linguistic groups. Dropping these individuals does not change our results. 18

19 control questions. Thus far, our measure of reciprocity is unable to distinguish between free riders, altruists, and flat contributors, as the Spearman correlation is zero for these three types. We show that our results hold when we drop altruists and flat contributors in column 5 or introduce dummy variables for these two types as additional controls in column 6. We also reproduce our main results using the wave 13 of the Swiss Household Panel (SHP) as an alternative data source. We do not have experimental measures of reciprocity for respondents in this sample, so we use data on values that are associated with reciprocity, such as cheating on tax declaration, lying in own interest, and claiming state benefits not entitled to (Knack and Keefer 1997, Guiso et al. 2011) 18. Individuals are asked to report the extent to which they think that each of the three items can be justified on a scale from 0 to 10, where 0 means never justified and 10 always justified. For the ease of interpretation, we invert the scale so that a higher score reflects stronger values. Because the three responses are correlated, we extract their underlying principal component and regress this on the Swiss German dummy controlling for the same set of covariates as in equation 2. The results are reported in columns 1-3 of Table A13 and show that without or with controls Swiss Germans have values that are over 40 percentage points higher than Swiss French and that these differences are also statistically significant at 1 the percent level. B. Falsification Tests If the gap in reciprocity between the social groups is capturing differential experience of historical democracy then it should disappear when historical democracy is altogether lacking. We test this by comparing Swiss Germans and Swiss French who are from municipalities without the historical experience of democracy. Column 1 in Table 3 shows that the gap in reciprocity between Swiss Germans and Swiss French drops from 0.29 (column 3, Table 2) to 0.04 and is no longer statistically significant (p-value = 0.66). 19 We ensure that this result is not due to some peculiarity of our sub-sample in column 2. When we replace Swiss Germans from municipalities without historical democracy with Swiss 18 Swiss Household Panel does not include the full list of questions on values asked in the World Values Survey, such as accepting a bribe, speeding over the limit, and throw away litter in a public space. 19 We also conduct an analogous test where we compare Swiss French from municipalities with historical democracy but Swiss Germans from municipalities without. If our hypothesis that historical democracy shaped reciprocity is correct then the sign on the Swiss German dummy should flip from positive to negative. We find that this is indeed the case: the coefficient on Swiss German dummy is and is even statistically significant. The absolute magnitude of the difference is comparable to the gap we obtain in our main result (Column 3, Table 2). This can be interpreted as suggesting that democracy has a positive association with reciprocity regardless of the social group. 19

20 Germans from the same canton but from municipalities with historical democracy, the gap in reciprocity reemerges and is also statistically significant at the 1 percent level. Together these results confirm that the strong gap in reciprocity across social groups is most likely due to their differential historical experience of democracy. V. Historical Democracy and Reciprocity We examine econometrically the association of reciprocity with historical democracy. Towards the end of the section, we also discuss reasons behind the persistence of this association, the emergence of democracy, and alternative interpretations. The exposure to historical democracy might come from an individuals municipality of residence as well as the municipality of birth. Since for over 90 percent of the respondents (both Swiss Germans and Swiss French) these two overlap, we present our main results using measures of historical democracy from the municipality of residence. Subsequently, while conducting robustness checks, we use the sub-sample of Swiss migrants and present results that consider measures from the municipality of birth using an epidemiological approach. Following previous studies that establish the association between historical events and behavior today (see for instance Tabellini 2010, Nunn and Wantchekon 2011, Voigtländer and Voth 2012, Alesina et al. 2013, Becker et al. 2016, Guiso et al. 2016), we investigate the association between historical democracy and reciprocity using the following OLS specification: (3)!!"# =!!!+!!!!"!" +!!!"#!! +!!!!!!+!!"!! +!!"#, where R is the reciprocity of individual i from municipality m and canton k, measured as the Spearman correlation between own and other players contribution in the conditional decision of the public goods game. HD is a measure of historical democracy from the resident municipality. The coefficient of interest is!!, which captures the association between an individual s reciprocity today and historical democracy from his / her resident municipality. We present results using both the experience of historical democracy and the duration of this experience. X and M are vectors of individual and municipality / canton specific controls that are listed in panel E of Table 1 and are hypothesized to matter for reciprocity in the literature. These include at the individual level age, education, gender, household income, religion and politics; altitude and Gini of income inequality at the municipal level; and GDP per capita at 20

21 the cantonal level. FE is a vector of canton, dynasty, and group fixed effects that capture canton, dynasty, and group specific factors that may affect reciprocity. Given the potential for within canton and spatial correlation of errors, we report standard errors that are clustered on the canton including p-values using the wild cluster bootstrap procedure (Camerer et al. 2008), as well as standard errors adjusted for spatial correlation (Conley 1999). A. Benchmark Estimates Table 4 report benchmark estimates without fixed effects on the association of reciprocity with historical democracy using experience in panel A and duration in panel B (for coefficients on covariates see Table A14a-b, Appendix A). Below the coefficient on reciprocity, we additionally report p-values from the wild cluster bootstrap procedure (in square brackets) and standard errors adjusted for spatial clustering (in curly brackets). Column 1 is without control variables and shows that the coefficient on experience is 0.41 while that on duration is 0.085, which are significant at the 1 percent level. The coefficients remain robust in magnitude and significance when we introduce individual level covariates in column 2 and municipal level variables alongside cantonal level GDP per capita in column 3. Given that the average reciprocity among individuals from municipalities without historical democracy is 0.42, the coefficient on experience in column 3 implies that individuals from municipalities with historical democracy display twice the reciprocity of individuals from municipalities without. Moreover, one standard deviation increase in duration (2.44) is associated with a rise in propensity for reciprocity by 0.22 points. 20 These estimates are not due to influential cantons, for when we drop one canton at a time in Table A15, the coefficient on experience ranges from 0.38 to 0.45 while that on duration from to 0.093; they always remain highly significant. These results imply that for each additional 10 Swiss Francs contributed by the other player individuals from municipality with historical democracy contribute 7.1, whereas individuals from municipalities without contribute 3.6 Swiss Francs, reflecting a difference of 103 percent over the base. We examine the possibility that other cantonal level differences today as well as in the past are driving the strong positive association between reciprocity and historical democracy. Following Alesina et al. (1999) and Francois et al. (2011) who highlight the importance of 20 We also tested for a specification with a quadratic term of duration. While the coefficient on duration remains large and statistically significant, the coefficient on duration squared is very small and statistically insignificant. We conducted similar tests for other cantonal and municipal level variables, such as historical population. As before, our results remain robust whereas the coefficients on the squared terms are small and statistically insignificant. 21

22 ethnic diversity and competition, we introduce in Table 4 share of foreigners (column 4) and an index of business competition (column 5). 21 Tabellini (2010) shows that values associated with reciprocity are affected by historical development and historical education. Accordingly, we consider proxies of these variables from the time when modern Switzerland was found. Because we do not have measures of GDP per capita from the historical period, we use as our main proxy the unemployment rate in 1888 (column 6) and consider additional proxies in Table A16 of Appendix A. These include population from the period before the invasion of Switzerland by Napoleon in 1798, and measures from 1850 on population and population density. For proxies of historical education, we include in column 7 of Table 4 enrollment in primary and secondary schools in Finally, in column 8, we include the index of modern democracy constructed by Stutzer (1999) and extended by Fischer (2009), acknowledging that it is endogenous. 22 The motivation is to test whether the effect of historical institutions on reciprocity today is channelized via modern institutions of democracy or not. The coefficients on added covariates, such as competition index, unemployment rate, and modern democracy have the expected sign and are also statistically significant. Nonetheless, the coefficients on experience and duration remain remarkably robust in both magnitude and significance. These results suggest that historical institutions of democracy matter for reciprocity over and above modern institutions despite the fact that the historical institutions themselves are long gone, and so must have left their mark by shaping values that were transmitted over time across generations. B. Fixed Effects Estimates The above results show a strong positive association between reciprocity and historical democracy, which is robust to many contemporaneous and historical controls. However, we cannot rule out that this association is not capturing the effect of a number of other factors that vary across cantons. In addition, it could be that our estimates are capturing differences in historical dynasties with which our municipalities were associated in the past. For instance, Becker et al. (2011) show that even if whole empires themselves perish, they can have longlasting effects on trust and corruption. Another possibility is that we are capturing some systematic differences across social groups. We mitigate these concerns in Table 5 by 21 We use data on locational quality index prepared by Credit Suisse on the attractiveness of Swiss Cantons and economic regions to business (see Holdener et al. 2013). 22 The modern democracy index scores institutions of direct legislation in all the 26 cantons of Switzerland. It is based on a number of requirements on constitutional initiative, statutory initiative, fiscal referendum, and statutory referendum. The index varies from 1-6, where 1 means weak and 6 strong. 22

23 introducing fixed effects for cantons in panel A, historical dynasties in panels B-C, and social groups in panels D-E. Each column reports the coefficient on historical democracy from a separate regression using experience in columns 1-3 and duration in columns 4-6. Canton fixed effects Panel A shows the association of reciprocity with historical democracy after controlling only for canton fixed effects. The coefficient on experience and duration without any additional individual or municipal level controls are 0.40 and 0.081, respectively, both of which are significant at the 1 percent level. When we augment this specification with individual level controls in column 2 and municipal level controls in column 3, the coefficients increase slightly to 0.42 and 0.09 but remain highly significant. These estimates are very similar to the ones without canton fixed effects (Table 4) and suggest that canton specific factors are not driving the association between reciprocity and historical democracy. This is not surprising after having seen in Table 4 that the coefficients on measures of historical democracy are robust to the introduction of a variety of important cantonal level variables. Dynasty fixed effects We consider two different dynasty fixed effects corresponding with two different time periods: Middle Ages and the period just before the invasion of Switzerland by Napoleon. In the 13 th century, the Holy Roman Emperor placed most areas in Switzerland under the authority of one of the four major noble houses: Zaehringen, Kyburg, Habsburg, and Savoy. The remaining areas were either under lesser-known nobles or prince bishoprics or under a common league / tithing. We assign municipalities in our sample to each of these five categories using maps by Ammann and Schib (1958) and Putzger (2004) and then investigate whether the association between reciprocity and historical democracy exists even after introducing fixed effects for these dynasties. Panel B reports the estimates and shows that without or with controls the coefficients on experience and duration are positive and highly significant. In the specification with the full set of controls, the size of the coefficient on experience is 0.41 (column 3) while that on duration is (column 6). These estimates are, as before, similar to those reported in Table 4 and clearly suggest that our estimates are not capturing dynasty specific differences from the Middle Ages. By 1415, Switzerland was mostly free of major noble dynasties. Instead, from this period onwards, the members and allies of the Old Swiss Confederacy developed a complex system of dependencies, which were managed either by a single member or jointly by two, 23

24 three, and at times even eight members. This system lasted until the invasion of Switzerland by Napoleon in 1798, which led to the dissolution of the Old Swiss Confederacy. We match each municipality in our sample with the original member or ally with which it was associated until Municipalities that were jointly administered by two or more members are assigned to a category of their own. Panel C shows that in the specification with the full set of controls the coefficients on experience and duration are 0.38 and and retain their high level of significance. We also consider a sub-sample of municipalities that were under the rule of the city-state of Bern (the largest city state north of the Alps covering nearly 25 percent of the Swiss area) for over 250 years (from ). Column 1 of Table A17 shows that the coefficients on experience (0.58) and duration (0.12) turn out to be slightly larger than our previous estimates but remain statistically significant at the 1 percent level. Our results also hold when we confine our sample to individuals from municipalities within the modern cantonal boundaries of Bern (column 2) or when we drop the city of Bern (column 3). Group fixed effects If our estimates are capturing some systematic differences across social groups then the scope of this concern is plausibly larger when the variation in historical democracy that we exploit is between groups but smaller when it is within groups. Accordingly, we report within group estimates obtained after controlling for social groups in panel D. In the specification with full set of controls, the coefficients on experience and duration are 0.38 and 0.081, which are significant at the 1 percent level. These estimates are remarkably similar to those reported in Table 4 and suggest that our results are not capturing group specific differences. Panel E shows that we obtain similar results when we restrict our sample to Swiss Germans only, which is akin to introducing a Swiss German fixed effect. Thus far, we have considered each fixed effect at a time. When we introduce all the three main fixed effects at the same time (canton, dynasty ca. 1200, and social group), the coefficients on experience and duration turn out to be 0.36 (s.e. 0.07) and (s.e ); these are highly significant. These results confirm that our benchmark estimates in Table 4 are not capturing differences across cantons, dynasties, and social groups. C. Additional Municipal and Individual Level Controls It is assuring that our results are robust to the introduction of a variety of cantonal level controls and fixed effects. However, this does not rule out the possibility that some other 24

25 municipality or individual level difference is driving the positive correlation of reciprocity with historical democracy. We now test whether our results are robust to the inclusion of additional municipal and individual level control variables. For this purpose, we use the specification in column 3, Table 4. However, all of our results hold even when we choose alternative specifications that control for canton or dynasty or group fixed effects. Table A18 includes additional controls at the municipal level such as the share of foreigners and an index of business competition, followed by proxies for current and historic levels of economic development. Because there is no data on GDP per capita at the municipal level today or in the past, we use current unemployment rate and population density from the time of founding of modern Switzerland (1850) as proxies. 23 Finally, we also consider the distance of a municipality from the local cantonal capital that was the center of power in the Middle Ages. The motivation is that during the Middle Ages, the authority of the state in areas further away from the capital could have been weaker allowing such areas to exercise more self-governance. Table A19 includes additional controls at the individual level like a dummy for naturalized Swiss citizen, a dummy for migrant from within Switzerland, Big Five personality traits, and a measure of game comprehension (number of control questions answered completely). In both the tables, the coefficients on experience and duration remain all throughout robust in magnitude and significance. The coefficient on experience is above 0.40 while that on duration stays above These results suggest that the positive association between reciprocity and historical democracy is most likely not capturing the effect of other municipal and individual level characteristics. D. Other Robustness Checks Epidemiological approach. In Table A20 we present results using measures of historical democracy from the birth municipality for our sub-sample of migrants. Using an epidemiological approach (Fernandez 2010, Algan and Cahuc 2014), we show that migrants from birth municipalities with historical democracy exhibit stronger reciprocity than migrants from birth municipalities without, after accounting for resident canton fixed effects, individual and municipal level controls. The coefficients on experience and duration are 0.56 and 0.12, which are slightly larger than the estimates we obtain using the full sample, but are also highly significant. 23 The oldest population data we have is from , but it is available only for a sub-set of municipalities. When we use this measure, the coefficients on experience and duration remain positive and highly significant. 25

26 Sub-sample analysis. In the Middle Ages, historical democracy was not practiced in a uniform manner in Switzerland but differed largely along the lines of rural and urban areas. Another distinction that warrants attention is that only male citizens were allowed to participate in decision-making. Although these differences could have affected reciprocity, Table A21 shows that the coefficients on experience and duration do not differ by rural-urban (columns 1-2) or male-female (columns 3-4) categories. In Table A22, we show that our results hold when we restrict our sample to the three Roestigraben cantons. As before, the difference is larger than what we obtain in the full sample (column 1) but survives even when these cantons are dropped (column 2). Finally, we also show in column 3 that altruists and flat contributors are not driving our results by dropping them altogether from the sample. Alternative dependent variable. We also reproduce our main results using a dummy variable for reciprocity as an alternative dependent variable. Table A23 reports the results and shows that the coefficients on both experience and duration turn out to be positive and highly significant. Individuals from municipalities with historical democracy are over 36 percentage points more likely to display a propensity for reciprocity than individuals from municipalities without. Moreover, one standard deviation increase in duration is associated with a rise in the likelihood of being a reciprocator by over 19 percentage points. Given that the share of reciprocators in our sample is 65 percent, these magnitudes are large. Selection on unobservables. In addition for the specification with the full set of controls (column 3, Table 4), we perform a bounding exercise to consider the sensitivity of the findings to unobserved selection as proposed by Oster (2016). We find that the selection on unobservables would have to be over two times as large as the selection on observables to zero out the estimated effect. E. Results using Swiss Household Panel In Table A24, Appendix A we reproduce our main results from Table 4 using data from the Swiss Household Panel. Note that we use as our dependent variable the principal component of values related to reciprocity, such as whether it is justifiable to cheat on tax declaration, lye in own interest, and claim state benefits not entitled to. Our results show that without or with controls there exists a strong positive and statistically significant association of the principal component of values with experience (panel A) and duration (panel B) of historical democracy. The magnitude of the coefficient on experience implies that with respect to 26

27 municipalities without the historical experience of democracy, values are higher on average by 32 percentage points in areas with experience. Similarly, a one standard deviation increase in duration is associated with an increase in the principal component of values by 0.15 percentage points. These results hold even when we introduce different fixed effects. F. Discussion Why did some areas acquire democracy but others did not? As discussed in section II, historical circumstances under which the municipalities acquired democracy were often idiosyncratic. As we discuss below, these cast doubts on reverse causality from reciprocity to democracy. In particular, one historical circumstance the extinction of the House of Zaehringen offers us quasi-experimental variation in historical democracy. We use this event to report intention to treat (ITT) and instrumental variable estimates (IV), whereby we use the extinction of the Zaehringen as an instrument for historical democracy. The Zaehringen were among the powerful noble dynasties in the Middle Ages that controlled large swathes of land in Switzerland. But, the dynasty came to an abrupt end in 1218 when its last duke died of natural causes without an heir. This led the German king to grant imperial immediacy to areas under the Zaehringen rule. Historians consider this to be a watershed event that gave municipalities the impetus to move towards democracy (Zschokke 1860, McCrackan 1908, Eugster 2015). For instance, Eugster (2015) points out that the extinction of the Zaehringen served as an essential prerequisite for the tendency towards more municipal autonomy of the 13 th and 14 th century. Though the timing of the extinction of the Zaehringen is plausibly exogenous, the Zaehringen rule in itself is not. Therefore, any pre-existing differences across areas with and without the Zaehringen rule could induce selection bias. The bias is expected to be weaker if the Zaehringen were assigned these areas than when these were selectively acquired by the Zaehringen themselves. Although historical records concur that emperors Henry IV and Lothair assigned Zaehringen the territories, this argument holds water as long as the emperors themselves did not grant the Zaehringen more productive areas. We believe this is highly unlikely because favoring Zaehringen over other contemporary dynasties could have resulted in a rift with the nobility, whose support the emperors needed in the Middle Ages. Nevertheless, we test for the scope of pre-existing differences by comparing municipalities that were under the Zaehringen rule with municipalities that were under other dynasties that ruled concurrently on the Swiss plateau. Bearing in mind that in the Middle 27

28 Ages, pre-existing differences are likely to be based on factor endowments in agriculture, we follow Alesina et al. (2013) and Galor and Özak (2016) and consider a number of proxies reflecting crop suitability. These include municipal level climate and soil suitability indices for a variety of cereal and fodder crops that could have been grown in Switzerland in pre period. The former captures differences in cultivation options on the basis of climatic conditions, while the later captures different elements of the landscape characterized by bedrock, slope and gradient. The data for these indices were obtained from the Swiss Federal Office for Agriculture. Table A25 reports the results and shows that there is no difference in either climate (column 1) or crop suitability (column 2) across the dynasties. We further test these results by comparing two additional municipal level characteristics: average altitude and share of waste / rugged land. Columns 3-4 report the results. As before, there is no difference in any of these characteristics between municipalities from different dynasties. This evidence can be interpreted as suggesting that pre-existing differences in agricultural productivity and geographical attributes across areas with and without the Zaehringen rule were most likely lacking. We start by reporting intention-to-treat (ITT) estimate in Table A26. Column 1 includes our full set of controls and shows that the coefficient on the Zaehringen dummy turns out to be 0.12 and is significant at the 5 percent level. 24 The size of the coefficient implies that individuals from municipalities that were associated with Zaehringen rule in the past have higher reciprocity by 0.12 points than individuals from municipalities that were associated with other dynasties. The instrumental variables estimates of the effect of historical democracy on reciprocity require that the Zaehringen rule did not have any direct effect on reciprocity. We offer two pieces of evidence in support of this. First, once we directly control for historical democracy via experience in column 2 and duration in column 3 of Table A26, the coefficient on Zaehringen dummy drops considerably in magnitude from 0.12 (column 1) to (column 2) and (column 3); it is also now no longer statistically significant. In contrast, the coefficients on experience and duration remain robust in magnitude and significance. Second, if being under the Zaehringen rule itself led to higher reciprocity then we should observe a positive and significant association between the Zaehringen dummy and reciprocity even when historical democracy is lacking. Instead, as Figure A2 shows the difference in reciprocity across dynasties is small and statistically insignificant. A regression- 24 Our results hold even when we control for other dynastic dummies. In this case, the coefficient on Zaehringen dummy is 0.15 and is significant at the 10 percent level. 28

29 based test confirms this and shows that the coefficient on Zaehringen dummy is not significantly different from the coefficients on Kyburg and Habsburg dummies (p-value = 0.64), casting doubts that the Zaehringen rule was special relative to other dynasties. Panel A of Table A27 reports first-stage estimates of the association of Zaehringen dummy with experience in columns 1-2 and duration in columns 3-4. Without or with controls there is a strong positive association between historical democracy and the Zaehringen dummy, which is also highly significant. Municipalities with Zaehringen rule are nearly 30 percentage points more likely to have historical democracy and for a longer duration by nearly 160 years than municipalities without. Given that 54 percent of the municipalities in our sample were exposed to historical democracy for 258 years, these differences are large in magnitude. The F-statistics are also large (> 19), confirming that the instrument is relevant. Panel B reports second-stage estimates of the effect of historical democracy on reciprocity. Both experience and duration have positive coefficients that are also statistically significant at the 5 percent level. The coefficients in specifications with full set of controls are 0.45 on experience and on duration, which are remarkably similar to their OLS counterparts in column 3 of Table 4. These results confirm the positive association of historical democracy with reciprocity. G. Alternative Interpretations We rule out the alternative interpretation that the differences in reciprocity across social groups that we observe is actually due to Swiss French playing an out-group game. Although in expectation Swiss French are more likely to be matched with Swiss Germans because the latter form the majority group in Switzerland, our data suggests that this explanation is not at work. These data comes from a second experiment, which is similar to the experiment described in section III except for one key difference we now reveal the linguistic group of the other player. Specifically, we randomly matched an individual from a linguistic group to another individual either of the same (in-group) or a different linguistic group (out-group). For example, some Swiss French were randomly matched with other Swiss French, while the remaining were randomly matched with Swiss Germans. All individuals who took part in the first experiment also took part in the second experiment. Our results reveal no difference in reciprocity when an individual is randomly assigned to an in-group or an out-group treatment, either among Swiss French (p-value = 0.95) or among Swiss Germans (p-value = 29

30 0.87). A related concern that Swiss French display lower reciprocity because they are a minority in Switzerland is also ruled out, as we do not find any difference in reciprocity across Swiss French from cantons where they are in absolute majority (Vaud, Neuchatel, Geneva, and Jura) to cantons where they are in relative majority (Fribourg and Valais) or where they are in minority (Bern) (p-value = 0.69). H. Plausible Channel What could have led to the persistence of the effect of historical democracy on reciprocity even until today? One plausible channel is the intergenerational transmission of values, which could have led to the formation of distinct group specific cultures (Bisin and Verdier 2011, Algan and Cahuc 2014). Our results from the epidemiological approach already suggest that reciprocity persists even when individuals move. In this section, we investigate the role of family background in the transmission of reciprocity. Our data does not allow us to test directly for intergenerational transmission, as we do not observe the reciprocity of parents and grandparents in the experiment. So, we investigate whether our measure of reciprocity changes in response to changes in the degree of parental and grandparental linguistic background. Bisin and Verdier (2011) posit three distinct processes for the transmission of values from parents to children: direct socialization of children by parents, indirect socialization via parents choosing the social environment of their children (school and neighborhood), and assortative mating. We focus on the combined effect of the three processes because of high correlation among them. 25 To proxy for the family background of an individual, we use two dimensions: the linguistic region in which individuals parents or grandparents were born (region of birth) and spent most of their time (region of stay). 26 Data were collected via the post-experimental survey, where individuals could choose from the following options: Swiss German, Swiss French, Swiss Italian, and other (outside Switzerland). We build a quantitative score for each individual to measure the strength of parental / grandparental background an individual is exposed to without taking the individuals language into consideration. For the parental background, the scoring rule is such that each individual can have from zero to four points 25 For instance, parents / grandparents region of birth and region of stay can be used as proxies for direct and indirect socialization, respectively. In our sample, the correlation between region of birth and region of stay is 0.89 for grandparents, and 0.84 for parents. Similarly, by comparing mothers and fathers regions of birth and stay, we could also infer assortative mating. The correlation is 0.99 between their regions of stay and 0.76 between their regions of birth. 26 Unfortunately, we do not have these data on the municipality of parents and grandparents. 30

31 (maximum two points each from mother s and father s regions of birth and stay). For the grandparental background, each individual can have from zero to eight points (maximum four points each from maternal grandparents and paternal grandparents regions of birth and stay). The higher the score, the higher is the exposure an individual has to the linguistic background of parents and grandparents. Because of little overlap, the Swiss German and Swiss French family backgrounds are strongly negatively correlated with each other. 27 In line with our previous analysis, we report results using the Swiss German parental and grandparental background and hypothesize these to have a positive association with reciprocity. An additional advantage of this approach is that it allows us to address the concern arising from differences in pure and mixed heritage across linguistic groups. We estimate the association of reciprocity separately with parental and grandparental background. Columns 1-2 in Table 6 report the results using parental and grandparental background respectively. Because parental and grandparental estimates use different scales, we also report standardized coefficients for easy comparison in the row below the unstandardized coefficients. We find that the stronger the Swiss German family background, the stronger is reciprocity among individuals. A one standard deviation increase in Swiss German parental background (1.71) and Swiss German grandparental background (3.30) is associated with an increase in reciprocal behavior by 0.10 and 0.07 points, respectively. 28 These results can also be broadly interpreted as demonstrating the role of intergenerational transmission in the persistence of reciprocity over time. VI. Conclusions Evidence in economics suggests large variation in reciprocity across social groups and its implications for a variety of economic outcomes. Yet, our understanding of the determinants of reciprocity and how it is fostered is still poor. We investigate the association of historical democracy with reciprocity today and also shed some light on plausible mechanisms that could explain the emergence of historical democracy in the Middle Ages and the persistence of its effect on reciprocity over time. 27 Only nine individuals have an overlap in Swiss German and Swiss French parental background, and 33 individuals in grandparental background. 28 We believe that this difference in transmission of reciprocity across generations is due to differences in historical democracy. Consequently, we study whether the association still holds when we restrict our sample to those municipalities that were not exposed to historical democracy. The results show that the coefficients on parental as well as grandparental background disappear altogether and are close to zero in magnitude. This is despite the fact that in this sample Swiss Germans have a stronger Swiss German grandparental and parental background by 5.6 and 3.38 points than Swiss French (out of a maximum possible is 8 and 4). 31

32 Our study takes place in Switzerland, whereby we exploit variation in historical democracy over space and time and match this with variation in linguistic social groups. Reciprocity is elicited using a one-shot anonymous public goods game in the strategy method to control for benefits from repeated interaction, reputation formation, as well as beliefs about the contribution of others. We find that the linguistic social groups differ in their exposure to historical democracy: Swiss Germans are four times more likely to be from municipalities with historical democracy than Swiss French and also for longer duration. In line with this difference, our first set of results show a large gap in reciprocity across social groups: Swiss Germans display significantly higher reciprocity than Swiss French. Moreover, the gap disappears completely when we compare Swiss Germans and Swiss French from municipalities without historical democracy. We find a strong positive association between historical democracy and reciprocity, which holds even when we control for a variety of fixed effects including cantons, dynasties, and social groups. In addition, we exploit the extinction of the Zaehringen dynasty from the absence of an heir as a plausibly exogenous shock to show a strong positive association between historical democracy and the Zaehringen rule, as well as between historical democracy and reciprocity today. Lastly, we also offer suggestive evidence on the role of intergenerational transmission in the persistence of these effects by showing that stronger Swiss German parental and grandparental background is associated with stronger disposition for reciprocity. Our findings are in line with previous studies that document, inter alia, a strong and persistent effect of historical experiences on preferences and beliefs today. 32

33 REFERENCES Aidt, T. S., and R. Franck Democratization Under the Threat of Revolution: Evidence From the Great Reform Act of Econometrica, 83, Alesina, A., R. Baqir, and W. Easterly Public Goods and Ethnic Divisions. Quarterly Journal of Economics, 114, Alesina, A., P. Giuliano, and N. Nunn On the Origins of Gender Roles: Women and the Plough. Quarterly Journal of Economics, 128(2), Algan, Y., and P. Cahuc Trust, Growth and Well-being: New Evidence and Policy Implications, In S. N. Durlauf and P. Aghion (eds.), Handbook of Economic Growth, Volume 2, Ammann, H., and K. Schib Historischer Atlas der Schweiz, Aarau. Aron, A., E. N. Aron, M. Tudor, and G. Nelson Close Relationships as Including the Other in the Self. Journal of Personality and Social Psychology, 60, Bandiera, O., I. Barankay, and I. Rasul Social Preferences and the Response to Incentives: Evidence from Personnel Data. Quarterly Journal of Economics, 120(3), Bardhan, P Irrigation and Cooperation: An Empirical Analysis of 48 Irrigation Communities in South India. Economic Development and cultural change, 48, Becker, S., O. K. Boeckh, C. Hainz, and L. Woessmann The Empire Is Dead, Long Live the Empire! Long-Run Persistence of Trust and Corruption in the Bureaucracy. Economic Journal, 126(590), Berner, H Grosser Rat (Ancien Regime). Historisches Lexikon der Schweiz (31 January). Bettinger, E. and R. Slonim Patience in Children: Evidence from Experimental Economics. Journal of Public Economics, 91(1-2), Bisin, A. and T. Verdier The Economics of Cultural Transmission and Socialization. In J. Benhabib, A. Bisin, and M. Jackson (eds.), Handbook of Social Economics, Volume 1A, The Netherlands: North-Holland. Bolton, G. E. and A. Ockenfels ERC: A Theory of Equity, Reciprocity and Competition. American Economic Review, 90, Bowles, S., and H. Gintis The Inheritance of Inequality. Journal of Economic Perspectives, 16(3),

34 Bowles, S., and S. Polania-Reyes Economic Incentives and Social Preferences: Substitutes or Complements? Journal of Economic Literature, 50, Buhler, S., J. M. Nunes, G. Nicoloso, J. M. Tiercy, and A. S. Mazas The Heterogeneous HLA Genetic Makeup of the Swiss Population. PLOS One, 7, Cameron, C., J. B. Gelbach, and D. L. Miller Bootstrap-baased Improvements for Inference with Clustered Errors. Review of Economics and Statistics, 90(3), Clots-Figueras, I. and P. Masella Education, Language and Identity. Economic Journal, 123, F332 F357. Clyne, M The German Language in a Changing Europe. Cambridge: Cambridge University Press. Conley, T. G GMM Estimation with Cross Sectional Dependence. Journal of Econometrics, 92, Dal Bo, P., A. Foster, and L. Putterman Institutions and Behavior: Experimental Evidence on the Effects of Democracy. American Economic Review, 100, DellaVigna, S Psychology and Economics: Evidence from the Field. Journal of Economic Literature, 47, Dufwenberg, M. and G. Kirchsteiger A Theory of Sequential Reciprocity. Games and Economic Behavior, Eugster, E Zähringen, von Historisches Lexikon der Schweiz (1 January). Eugster, B., R. Lalive, A. Steinhauer, and J. Zweimuller The Demand for Social Insurance: Does Culture Matter? Economic Journal, 121, Fehr, E On the Economics and Biology of Trust. Journal of the European Economic Association, 7, Falk, A. and U. Fischbacher A Theory of Reciprocity. Games and Economic Behavior, 54, Fehr, E., and S. Gächter Fairness and Retaliation: The Economics of Reciprocity. Journal of Economic Perspectives, 14, Fehr, E. and K. Schmidt A Theory of Fairness, Competition, and Cooperation, Quarterly Journal of Economics, 114, Fernandez, R Does Culture Matter? In J. Benhabib, A. Bisin and M.O Jackson (eds), Handbook of Social Economics, Volume 1, Fischbacher, U. and S. Gächter Social Preferences, Beliefs, and the Dynamics of Free Riding in Public Goods Experiments. American Economic Review, 100, Fischbacher, U., S. Gächter, and E. Fehr Are People Conditionally Cooperative: 34

35 Evidence From a Public Goods Experiment. Economic Letters, 71, Fischer, J Development of Direct Democracy in Swiss Cantons between 1997 and MPRA Paper no Francois, P., T. Fujiwara, and T. van Ypersele Competition Builds Trust. Working Paper. Frankhauser, A Mediationsakte. Historisches Lexikon der Schweiz (12 August). Frey, B. S Institutions and Morale: The Crowding-out Effect. In A. Ben-Ner and L. Putterman (eds.), Economics, Values, and Organization, New York: Cambridge University Press. Frey, B. S., and S. Meier Social Comparisons and Pro-social Behavior: Testing Conditional Cooperation in a Field Experiment. American Economic Review, 94, Galor, O. and Özak, O The Agricultural Origins of Time Preferences. American Economic Review, 106, Gasser, A Die Territoriale Entwicklung der Schweizerischen Eidgenossenschaft Aarau: Sauerlaender. Giles, H. and P. Johnson Ethnolinguistic Identity Theory: A Social Psychological Approach to Language Maintenance. International Journal of the Sociology of Language, 68, Grossjean, P A History of Violence: The Culture of Honor and Homicide in the US South. Journal of the European Economic Association, 12, Guiso, L., P. Sapienza, and L. Zingales Civic Capital as the Missing Link, In J. Benhabib, A. Bisin, and M. Jackson (eds.), Handbook of Social Economics, Volume 1A, The Netherlands: North-Holland. Guiso, L., P. Sapienza, and L. Zingales Long Term Persistence. Journal of the European Economic Association, 14(6), Gumperz, J. J Language and Social Identity. Cambridge: Cambridge University Press. Heinemann, H Friedrich II, Historical Lexicon of Switzerland (May 19). Herrmann, B. and C. Thöni Measuring Conditional Cooperation: A Replication Study in Russia. Experimental Economics, 12, Holdener, V., F. Hürzeler, T. Rühl, A. Schnell, and N. Sydow Locational Quality of the Swiss Cantons and Regions. Credit Suisse Global Research. Special Issues Regions, September. 35

36 Holenstein, A Räte. Historisches Lexikon der Schweiz (9 December). Jha, S Trade, Institutions and Ethnic Tolerance: Evidence from South Asia. American Political Science Review, 107, Kamm, R Glarus (Kanton): Kommunale Bewegung und Staatlichkeit, Section 2.2.4, Historisches Lexikon der Schweiz (30 May). Knack, S. and P. Keefer Does Social Capital Have an Economic Payoff? A Cross- Country Investigation. Quarterly Journal of Economics, 112, Kocher, M. G., T. Cherry, S. Kroll, R. J. Netzer, and M. Sutter Conditional Cooperation on Three Continents. Economics Letters, 101, Kosfeld, M., and D. Rustagi Leader Punishment and Cooperation in Groups: Experimental Field Evidence from Commons Management in Ethiopia. American Economic Review, 105, Landolt, O. 2017a. Schwyz (Kanton): Die Entstehung des Landes Schwyz, Section 2.2, Historisches Lexikon der Schweiz (26 June). Landolt, O. 2017b. Schaffhausen (Gemeinde): Herrschaft, Politik, und Verfassung vom Hochmittelalter bis zum Ende des 18 Jahrhundert, Section 2, Historisches Lexikon der Schweiz (26 June). Lowes, S., N. Nathan, J. A. Robinson, and J. Weigel The Evolution of Culture and Institutions: Evidence from the Kuba Kingdom, Econometrica, 85(4), McCrackan, W. D The Rise of the Swiss Republic: A History. New York: Henry Holt and Company. Nunn, N., and L. Wantchekon The Slave Trade and the Origins of Mistrust in Africa. American Economic review, 101, Oster, E Unobservable Selection and Coefficient Stability: Theory and Evidence, Journal of Business & Economic Statistics, Ostrom, E., J. Burger, C. B. Field, R. B. Norgaard, and D. Policansky Revisiting the Commons: Local Lessons, Global Challenges. Science, 284, Platteau, J.-P Institutions, Social Norms, and Economic Development. Academic Publishers & Routledge. Putnam, R., R. Leonardi, and R. Y. Nanetti Making Democracy Work: Civic Traditions in Modern Italy. New York: Simon & Schuster. Putzger Historischer Weltatlas, Schweizer Ausgabe, Berlin. Rodrik, D Do Democratic Transitions Produce Bad Economic Outcomes. American Economic Review Papers and Proceedings, 95,

37 Rustagi, D., S. Engel, and M. Kosfeld Conditional Cooperation and Costly Monitoring Explain Success in Forest Commons Management. Science, 330, Schelbert, L Historical Dictionary of Switzerland. Second Edition. Rowman & Littlefield, UK.. Schultheiss, M Zürich (Gemeinde): Die Stadt wird Selbständig, Section 2.1, Historisches Lexikon der Schweiz (25 January). Sobel, J Interdependent Preferences and Reciprocity. Journal of Economic Literature, XLIII, Stadler, H Landsgemeinde Historisches Lexikon der Schweiz (11 November). Stadler, H Schöllenen. Historisches Lexikon der Schweiz (11 April). Stadler, H Uri (Kanton): Die Entstehung des Landes Uri und die Eidgenossenschaft im Spätemittelater, Section 2.2, Historisches Lexikon der Schweiz (22 May). Stutzer, A Demokratienindizes für die Kantone der Schweiz. Working Paper 23, University of Zurich. Suter, M Zürich (Kanton): Herrschaft, Politik, und Verfassung vom Hochmittelalter bis zum Ende des 18 Jahrhundert, Section 2, Historisches Lexikon der Schweiz (30 May). Sutter, M., S. Haigner, and M. Kocher Choosing the Carrot or the Stick: Endogenous Institutional Choice in Social Dilemma Situations. Review of Economic Studies, 77, Tabellini, G Culture and Institutions: Economic Development in the Regions of Europe. Journal of the European Economic Association, 8, Uslaner, E Trust, Democracy and Governance: Can Government Policies Influence Generalized Trust? In M. Hooghe and D. Stolle (eds.), Generating Social Capital: Civil Society and Institutions in Comparative Perspective, Houndmills: Palgrave Macmillan. Voigtländer, N., and H. J. Voth Persecution Perpetuated: The Medieval Origins of Anti-Semitic Violence in Nazi Germany. Quarterly Journal of Economics, 127, Zahnd, U. M Bern (Kanton): Kommunale Bewegung und Territorialbildung im Spätemittelalter, Section 2.2, Historisches Lexikon der Schweiz (29 May). Zschokke, H History of Switzerland. Translated by Shaw, F.G. New York: Hurst and Company Publishers. 37

38 TABLE 1: SUMMARY STATISTICS Sample Mean Standard Deviation Panel A: Social groups Swiss German dummy Panel B: Historical democracy Experience (dummy) Duration (hundreds of years) Panel C: Reciprocity Reciprocity Panel D: Family background Parental Swiss German background Grandparental Swiss German background Panel E: Covariates Individual level Age University degree Male Log household income Catholic Protestant Left wing Center Municipality / canton level Gini income Altitude GDP per capita Observations 262 Notes. Swiss German dummy refers to the proportion of individuals whose native language is Swiss German. Experience is a dummy variable equal to one if an individual lives in a municipality that experienced historical democracy before 1803, zero otherwise. Duration refers to the number of hundreds of years an area was exposed to historical democracy before Reciprocity is measured using a public goods game. The parental Swiss German background is a score from zero to four, reflecting the degree to which the parental region of birth and region of stay is Swiss German speaking. Similarly, grandparental Swiss German background is a score from zero to eight, reflecting grandparental region of birth and region of stay. Age is measured in number of years. Male is a dummy variable equal to one if the individual is male, zero otherwise. University degree is a dummy variable equal to one if the individual has a university degree, zero otherwise. Log household income is the log of annual household income / Catholic and protestant are dummy variables indicating the religion of an individual. Left wing and center are dummy variables referring to the political orientation of an individual. Altitude is the mean elevation of the municipality in meters / 100. Gini income is a measure of income inequality at the level of resident municipality. GDP per capita is at the cantonal level and is measured in 1000s of Swiss Francs. The number of observations for the variable parental Swiss German background is 253 and grandparental Swiss German background is 249 because of missing answers to the corresponding questions. 38

39 TABLE 2: SOCIAL GROUPS AND RECIPROCITY Dependent variable: Reciprocity No controls Individual controls (1) (2) Full set of controls (3) Swiss German 0.319*** 0.316*** 0.291*** (0.070) (0.069) (0.068) [0.000] [0.000] [0.000] Individual controls No Yes Yes Municipal controls No No Yes GDP per capita No No Yes Observations R-squared Notes. OLS coefficients with robust standard errors clustered at the cantonal level in parentheses. Square brackets report p-values from the wild cluster bootstrap procedure proposed by Cameron et al. (2008) with 1,000 replications. Table A7 in Appendix A reports coefficients on covariates. *** indicate significance at the 1% level. TABLE 3: SOCIAL GROUPS AND RECIPROCITY: FALSIFICATION TEST Swiss French and Swiss Germans from municipalities without historical democracy (1) Dependent variable: Reciprocity Swiss French from municipalities without but Swiss Germans with historical democracy (2) Swiss German *** (0.095) (0.061) [0.637] [0.000] Individual controls Yes Yes Municipal controls Yes Yes GDP per capita Yes Yes Observations Number of cantons R-squared Notes. OLS coefficients with robust standard errors clustered at the cantonal level in parentheses. Square brackets report p-values from the wild cluster bootstrap procedure proposed by Cameron et al. (2008). Column 1 compares differences in reciprocity across Swiss Germans and Swiss French from municipalities without historical democracy. Column 2 replaces Swiss Germans from municipalities without historical democracy with Swiss Germans from municipalities with historical democracy but from the same canton. Individual and municipal level controls include the variables listed in Panel E of Table 1. *** indicates significance at the 1% level. 39

40 TABLE 4: HISTORICAL DEMOCRACY AND RECIPROCITY No controls (1) Historical democracy 0.410*** (0.045) [0.000] {0.064} Individual controls (2) 0.405*** (0.055) [0.000] {0.070} Municipal controls (3) 0.424*** (0.056) [0.000] {0.070} Dependent variable: Reciprocity Share of Comp. foreigners index (4) (5) Panel A: Experience 0.433*** (0.062) [0.000] {0.074} 0.396*** (0.056) [0.000] {0.069} Unemployment 1888 (6) 0.407*** (0.057) [0.000] {0.074} Schooling 1888 (7) 0.416*** (0.061) [0.000] {0.074} Modern democracy (8) 0.409*** (0.055) [0.000] {0.068} Share of foreigners (0.004) Competition index 0.056*** (0.019) Unemployment ** (0.969) Primary school (0.170) Secondary school (0.429) Modern democracy 0.028* (0.014) R-squared Panel A: Duration Historical democracy 0.085*** (0.008) [0.000] {0.013} 0.084*** (0.011) [0.000] {0.015} 0.089*** (0.011) [0.000] {0.014} 0.091*** (0.012) [0.000] {0.015} 0.084*** (0.011) [0.000] {0.014} 0.086*** (0.011) [0.000] {0.015} 0.087*** (0.012) [0.000] {0.015} 0.086*** (0.010) [0.000] {0.014} Share of foreigners (0.004) Competition index 0.046** (0.018) Unemployment * (1.091) Primary school (0.176) Secondary school (0.448) Modern democracy 0.034** (0.013) R-squared Individual controls No Yes Yes Yes Yes Yes Yes Yes Municipal controls No No Yes Yes Yes Yes Yes Yes GDP per capita No No Yes Yes Yes Yes Yes Yes Observations Notes. OLS coefficients with robust standard errors clustered at the cantonal level in parentheses. We additional report in square brackets p-values from wild cluster bootstrap procedure (Cameron et al. 2008) and in curly brackets standard errors adjusted for spatial clustering with a threshold of 100 Km (Conley 1999). Our results are robust to alternative thresholds at 10, 25, 50, and 200 km. Columns 4-8 introduce additional controls at the cantonal level. These are current share of foreigners in column 4 and competition index in column 5; unemployment rate in 1888 in column 6, school enrollment in 1888 in column 7, and index of modern democracy in column 8. Data on share of foreigners, unemployment rate and school enrollment in 1888 are from the Swiss Federal Statistical Office, and on competition index by Credit Suisse (see Holdener 2013). The index of modern democracy is from It varies from one to six and is based on studies by Stutzer (1999) and Fischer (2009). Individual and municipal controls include the variables in Panel E of Table 1. Tables A14a-b in Appendix A report coefficients on covariates. ***, **, * indicate significance at the 1%, 5%, and 10% levels, respectively. 40

41 TABLE 5: HISTORICAL DEMOCRACY AND RECIPROCITY: FIXED EFFECTS ESTIMATION (CANTON, DYNASTY, AND SOCIAL GROUP) Dependent variable: Reciprocity Experience Duration No controls (1) Individual controls (2) Full set of controls (3) No controls (4) Individual controls (5) Full set of controls (6) Panel A: Canton fixed effects Historical democracy 0.405*** 0.412*** 0.418*** 0.081*** 0.082*** 0.088*** (0.052) (0.063) (0.074) (0.010) (0.012) (0.013) Panel B: Dynasty fixed effects ca Historical democracy 0.369*** 0.366*** 0.414*** 0.078*** 0.077*** 0.089*** (0.038) (0.058) (0.066) (0.006) (0.011) (0.012) Panel C: Dynasty fixed effects ca Historical democracy 0.396*** 0.401*** 0.377*** 0.083*** 0.083*** 0.082*** (0.060) (0.068) (0.073) (0.009) (0.012) (0.012) Panel D: Social group fixed effects Historical democracy 0.356*** 0.353*** 0.384*** 0.074*** 0.073*** 0.081*** (0.032) (0.047) (0.052) (0.006) (0.009) (0.010) Panel E: Swiss German fixed effect Historical democracy 0.359*** 0.320*** 0.364*** 0.074*** 0.066*** 0.075*** (0.032) (0.052) (0.053) (0.006) (0.010) (0.010) Individual controls No Yes Yes No Yes Yes Municipal controls No No Yes No No Yes Fixed effects Yes Yes Yes Yes Yes Yes Notes. OLS coefficients with robust standard errors clustered at the cantonal level in parentheses. Each cell reports the coefficient on historical democracy from a separate regression using experience in columns 1-3 and duration in columns 4-6. All columns include fixed effects of the type listed in the panel headline. In panel A, we include fixed effects for cantons. In panel B, we include fixed effects for the dynasties with which our municipalities were associated ca In panel C, we include fixed effects taking into consideration cantonal boundaries before the invasion of Switzerland by Napoleon in In panel D, we control for the Swiss German dummy. In panel E, we exclude Swiss French from the sample and exploit variation in historical democracy only in our Swiss German sub-sample. Individual and municipal controls include the variables in Panel E of Table 1. All panels control for GDP per capita except for panel A because of canton fixed effects. The number of observations in each panel except for panel E is 262. Panel E excludes Swiss French therefore the number of observations is 195. All our results hold when we adjust our standard errors using either the wild cluster bootstrap procedure proposed by Cameron et al. (2008) or spatial clustering following Conley (1999). *** indicates significance at the 1% level. 41

42 TABLE 6: FAMILY BACKGROUND AND RECIPROCITY Dependent variable: Reciprocity Parental background (1) Grandparental background (2) Family background 0.060*** 0.021** (0.017) (0.009) Standardized coefficient Individual controls Yes Yes Municipal controls Yes Yes GDP per capita Yes Yes Observations R-squared Notes. OLS coefficients with robust standard errors clustered at the cantonal level in parentheses. Swiss German background refers to the linguistic region in which an individual s parents or grandparents were born as well as spent most of their time. Swiss German parental background is a score ranging from zero to four; while Swiss German grandparental background is a score ranging from zero to eight. Individual and municipal level controls are the same as in Panel E of Table 1. ***, ** indicate significance at the 1% and 5% level, respectively. Residuals of reciprocity on covariates Residuals of historical democracy on covariates (experience) FIGURE 1: RECIPROCITY AND HISTORICAL DEMOCRACY Notes. Y-axis plots residuals from a regression of reciprocity on covariates. X-axis plots residuals from a regression of historical experience of democracy on the same covariates. Each point represents an individual. 42

43 Reciprocity Historical democracy (experience) Reciprocity Swiss German Swiss French FIGURE 2A FIGURE 2B FIGURE 2: RECIPROCITY BY HISTORICAL DEMOCRACY AND SOCIAL GROUPS Notes. Y-axis plots reciprocity, measured as the Spearman correlation between self and other players contribution in the conditional decision of the public goods game. FIGURE 3: GEOGRAPHICAL DISTRIBUTION OF SWITZERLAND S OFFICIAL LANGUAGES Source: Swiss Federal Statistical Office (2000) 43