Patricia Funk, Universitat Pompeu Fabra

This article empirically investigates whether law affects behavior beyond deterrence (‘‘expressive function of law’’). With Swiss panel data, I find that the legal abolition of the voting duty significantly decreased average turnout, even though the fines for not voting have only been symbolic. As for the size of Cantonal turnout reduction, it widely differs between the Cantons and is highly correlated with voter participation before the removal of the voting duty. In contrast to the voting duty, the introduction of postal voting did not affect voter turnout in spite of the substantial decrease in transaction costs. Therefore, in public good areas such as voting, even a sanctionless law targeting at the civic duty might have a bigger impact on behavior than actions which affect the costs of provision for the public good. JEL Classification: K00, H00

1. Introduction The classic ‘‘Law & Economics’’-approach focuses on deterrence: a law enforced by a sanction increases the expected costs of the regulated activity I thank Juan Jose Ganuza, Fernando Gomez, Ali Hortacsu, Peter Kugler, Steven D. Levitt, participants of the 2006 meeting of the European Law and Economics Association and seminar participants of the Universitat Pompeu Fabra for helpful comments and discussions. Special thanks goes to the editor John J. Donohue and two anonymous referees. Financial support from the ‘‘Kommission Nachwuchsfoerderung’’ of the University of Basel is gratefully acknowledged. Send correspondence to: Patricia Funk, Universitat Pompeu Fabra, Department of Economics and Business, Ramon Trias Fargas, 25–27, 08005 Barcelona, Spain; E-mail: [email protected]. American Law and Economics Review doi:10.1093/aler/ahm002 Advance Access publication March 19, 2007  The Author 2007. Published by Oxford University Press on behalf of the American Law and Economics Association. All rights reserved. For Permissions, please e-mail: [email protected]

135

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

Is There An Expressive Function of Law? An Empirical Analysis of Voting Laws with Symbolic Fines

136

American Law and Economics Review V9 N1 2007 (135–159)

1. I follow Dharmapala and McAdams (2003) and summarize law’s effects beyond deterrence as expressive. 2. The literature on the expressive function of law is large and growing (see Ellickson (2001), McAdams and Rasmusen (forthcoming) and Posner (2000) for a general analysis of law and social norms). Specifically, Cooter (1998) and Scott (2000) discuss the dependency of social sanctions on law. McAdams (1997, 2000a) attitudinal theory of expressive law states that law changes the beliefs about which actions people approve/disapprove and has an effect under the following three conditions: law signals popular opinion, individuals update their beliefs about approval patterns, and individuals value approval. The preference-shaping argument has first been made in Dau-Schmidt’s (1990) analysis of criminal law, and later been elaborated by various authors, e.g. Sunstein (1987), Kahan (1996), Lessig (1998) and Pildes (1998). Expressive effects also arise in situations with coordination issues (where law provides a focal point, McAdams, 2000b), or situations where there is uncertainty about a certain behavior (e.g. risks of smoking). Passage of a law such as a smoking ban embodies information about the legislators’ risk estimates and leads to an update of the citizens’ prior beliefs (‘‘informative effect of law’’, Dharmapala and McAdams, 2003). 3. For instance, law might crowd out intrinsic motivation or certain honest types (Bohnet, Frey and Huck, 2001). See also Adler (1999) and Scott (2000) for a more critical view of the expressive function of law.

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

and thereby induces compliance. Since the early nineties, arguments have been put forward after which a law might have an effect independent of the sanction (= ‘‘Expressive Function of Law’’).1 There are different channels by which an unenforced law might work: First, people may obey a law out of civic duty. Secondly, by making a statement about what is right or wrong, law may affect the preferences over the regulated behavior (‘‘preference-shaping’’ effect). Finally, expected social sanctions increase, if citizens are more inclined to enforce norms embodied in the law. The dominant reading of the literature is that through all these channels, the expressive function of law induces compliance.2 However, arguments can be found that work in the opposite direction.3 Whatever the exact underlying mechanism, important policy implications would result if such (beneficial) ‘‘expressive effects’’ existed. First of all, behavior might be changed at lower costs than suggested by the classical ‘‘Law & Economics’’ literature. Merely by passing a law (even if unenforced), citizens might adhere to it. Secondly, the effect of legalizing a certain act might be higher than predicted from the abolition of the sanction. Therefore, the frequently encountered argument that a law which

Voting Laws with Symbolic Fines

137

4. In Switzerland, abortions until 12 weeks pregnancy were legalized in June 2002. One major argument was that the law was no longer enforced anyway. 5. Secondarily enforced seat belt laws only allow the police to fine violators when they are stopped for some other offense. In spite of the low probability of conviction, there appears to be a positive effect on seat belt usage. 6. Law’s (focal) role in solving coordination issues is less disputed and has been quite clearly established (see Bohnet and Cooter, 2005; McAdams and Nadler, 2005).

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

is hardly enforced can equally be abandoned (without causing any change in the respective behavior) would prove to be wrong.4 Although there is scant empirical research on expressive effects of law, legal scholars start taking them for granted. For instance, Hasen (1996: 2167) writes in his analysis on mandatory voting laws: ‘‘[Next to its direct effect], law works in a more subtle way, by shaping preferences (or changing) tastes in much the same way as social norms.’’ In view of the growing faith in the expressive function of law, empirical research on this topic is urgently needed. Up so far, there is anecdotal evidence that ‘‘pooper-scooper laws’’ and ‘‘no smoking’’ signs in U.S. airports have been effective with no or little legal enforcement (see Cooter, 2000). In the case of seat belt legislation, Cohen and Einav (2003) document that barely enforced seat belt laws had an effect on seat belt usage.5 Using a natural experiment, Fisman and Miguel (2006) observe a certain degree of law-adherence in a completely enforcement free zone. The authors study parking violations from diplomats in New York City and take advantage of the fact that diplomats cannot be fined. The goal of this article is to understand law’s expressive power in public goods situations.6 To shed light on the subject empirically, one ideally needed within-country variation in the legal duty to contribute. However, since the goal is to separate law’s expressive effects from its direct penalty effect, the law should have a zero expected penalty, or, as a first approximation, a minimal sanction or low enforcement probability. Furthermore, just comparing contribution levels under these different legal regimes could be misleading, since the enactment of the law is likely to be endogenous (i.e. enactment of a law where people care more about the public good). Therefore, a safer strategy seems to analyze law changes. If, for example, it is found that abolition of a law affects behavior relative to the control group, it points to expressive effects of law.

138

American Law and Economics Review V9 N1 2007 (135–159)

7. Similar to the American House of Representatives, the ‘‘Nationalrat’’ is the one (of the two legislative chambers), where the number of seats assigned to each region (Canton) corresponds to the population of the region (Canton). The different parties weight is (roughly) determined by the proportion of votes received (proportional representation).

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

Data that come close to this ideal are mandatory voting laws in Switzerland. Half of the 26 Swiss Cantons had a legal obligation to vote. However, the awareness of the legal voting duty was particularly high in 5 Cantons, where symbolic fines for not voting have been enacted (the fines were less than 1 Dollar in most cases). Since 4 out of these 5 Cantons got rid of the legal obligation to vote at different points in time, the empirical strategy will be to analyze, whether removal of this voting duty affected turnout. In order to compare homogeneous voting issues, national parliamentary elections are taken as the subject of investigation (called ‘‘Nationalratswahlen’’; elections are held every 4 years).7 To control for other factors affecting turnout, I include unemployment, education, population and age in the estimations. Also, to account for Canton-specific heterogeneity (language, culture etc.) and common trends in voting behavior, Canton (state) and time fixed effects are estimated. As a main result, I find a significant negative impact of the abolition of the voting duty on voter turnout (VT). Depending on the regression method (OLS or weighted least squares, [WLS]), there was a 6 to 10 percentage point drop in turnout after the minimally fined voting duties have been abolished. However, disentangling the average effect into the different Cantonal effects reveals substantial heterogeneity. While turnout reduction ranges from 5 to nearly 20 percent in the Cantons Aargau, St. Gallen and Thurgau, an insignificant effect is found for the Canton Zuerich. As it turns out, the amount of turnout reduction is to 90 % correlated with the level of turnout prior to the abolition of the voting duty. Therefore, in Cantons where citizens obeyed to the law very much, removal of the voting duty had a larger effect. Overall, the abolition of the voting duty seems to have caused a substantial decrease in VT. As I will argue in the article, the observed reaction is unlikely due to the monetary effect of the sanction, since the fines and the transaction costs associated with paying the fines were negligible. Rather, the legal statement that citizens should vote apparently caused certain citizens to follow, most likely out of civic duty or fear from social sanctions.

Voting Laws with Symbolic Fines

139

8. The twist of the paper is really the minimally fined and later abandoned voting duty. This allows to make a statement concerning an expressive function of law, which is hardly possible with other studies on compulsory voting, which analyze a country’s abandonment of a (heavily) fined voting duty (see Jackman, 2001 for an overview). 9. Most industrialized countries experienced a decrease in voter turnout over the past several decades.

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

As such, the observed turnout drop is interpreted as support for a certain (or at least situational) ‘‘expressive effect of law’’.8 I compare the effect of this voting duty with an other institutional change that took place in Switzerland: the introduction of (optional) mail voting. In contrast to the voting duty, which had a negligible cost effect but entailed a moral message, mail voting lowered the transaction costs of voting. Surprisingly, in spite of the large decrease in transaction costs, there was no effect on turnout. Therefore, the results seem to suggest that a change in the moral message (abolition voting duty) together with a small change in the voting costs (drop fine) can have a bigger effect on VT than a big change in the voting costs with no moral message (option of postal voting). The major contribution of this article consists of adding empirical evidence to the mainly theoretical debate about whether ‘‘expressive effects’’ of law exist. Therefore, new insights on the subject seem valuable even if the data set at hand is quite limited (only four law-changes). While being one of the first empirical tests of expressive law, other studies corroborate my results. In experimental settings, small fines have been found to induce cooperation in public good games, even when the magnitude of the fines was not sufficiently high to alter selfish behavior (Tyran and Feld, 2006). In an entirely different context, Adams and Ferreira (2005) found that directors in companies were more likely to attend meetings when exposed to a (relatively) small fine for nonattendance. Similar to the voting case, the authors see the effectiveness of the small fine not in the monetary incentive, but in the signal, it sends. Next to the article’s contribution on expressive laws, important insights are also given to the literature on voting. Since voting is commonly perceived as a public good (necessary to maintain democracy), policy makers are concerned about decreasing voter turnout.9 One possible remedy is seen in the introduction of postal voting, since it reduces the transaction costs of

140

American Law and Economics Review V9 N1 2007 (135–159)

2. Institutional Background: Cantonal Voting Laws Switzerland is a small federalist country with roughly 7 million inhabitants. The country consists of 26 major districts (called ‘‘Cantons’’), which are further divided into minor districts (‘‘Bezirke’’). The 26 Cantons have their own constitution and legislative power and are free to pass laws, as long as they do not contradict with federal law. As will be discussed in more detail, the Cantons differ with respect to legal regulations of the voting process. Subsequently, I analyze voter participation in national parliamentary elections (‘‘Nationalratswahlen’’) from 1951 to 1999.11 Parliamentary elections are ideal for this study, because the voting subject is unchanged 10. Southwell and Burchett (2000) study an all-mail election in Orgeon. However, the focus of their study is on the composition of the electorate (characteristics of voters) of this mail election compared to the traditional election at the polls. In any case, a comparison of voter turnout (from this All-Mail election versus poll elections) would be difficult to interpret, since there was only one All-Mail election, which completely replaced the polls. In contrast, postal voting was offered as an additional option in Switzerland, which facilitates the interpretation of its effect on voter turnout: voters who go to the polls due to consumption benefits (chatting with neighbors etc.) can still do so and hence, the ‘‘poll-lovers’’ are not crowded out by the introduction of postal voting. 11. Although time series data on Cantonal Voter turnout are available since 1919, I start including data from 20 years or 5 data points before the first major law change in 1971 (national parliamentary elections are held every 4 years). As it turns out, results are very robust with respect to choosing different starting points; regression estimates are available upon request.

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

voting considerably. Although there is sparse empirical evidence about its effectiveness, local governments in the United States start experimenting with this alternative organization of the voting process.10 Hasen (1996), on the other hand discusses whether the legal declaration of voting as a duty might have a positive effect on VT (due to its expressive effect). The effect of both proposals are tested and compared in this article. At least in Switzerland, the voting duty turned out to be more effective. The remainder of the article is structured as follows: Section 2 describes the Cantonal changes in voting laws. Section 3 presents the econometric model and the estimation results. Some robustness checks are conducted in Section 4. The article concludes in Section 5.

Voting Laws with Symbolic Fines

141

12. For instance, if a party decides to engage in tougher competition by increasing advertising expenditures, it affects the perception of this party in all Cantons. 13. See the article in the ‘‘Aargauer Zeitung’’ online from the 14. 1. 2006.

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

over time and random shocks on VT (weather conditions etc.) are similar in every Canton, since the voting day(s) are determined on a national level. Focusing on the legislative chamber ‘‘Nationalrat’’ (where parties are chosen according to proportional representation) bears the advantage that supply-side shocks on voter participation occur on a national rather than Cantonal level; therefore, they can be controlled for by time-fixed-effects.12 In contrast, in the other legislative chamber (‘‘Staenderat’’), where each Canton has the right to choose two representatives, the expected closeness of the race, the character of the candidates and other (hardly controllable) Canton-specific supply-side factors might influence VT. Therefore, voter participation for the ‘‘Nationalrat’’ should not remarkably be driven by canton-specific supply-side factors. However, the voting process for the national elections is organized in the different Cantons and therefore subject to the ‘‘Cantonal rules’’. To what extent might differences in Cantonal voter participation rates depend on different regulations of the voting duty? There exist three types of regulations: the strongest one entails a legally prescribed voting duty together with a minimal fine for nonvoting. As can be seen from Table 1, there were five Cantons, which had such a type of regulation in 1951, and one Canton, which still enforces the voting duty today. Eight Cantons had a legal prescription to go to the polls (without any sanctions), and thirteen Cantons did not legally prescribe to go to the polls at all. In this study, I will focus on the Cantons, which had a minimally fined voting duty. First of all, there exist substantial variation over time: four out of five Cantons abolished their duty to go to the polls. Secondly, the citizens were reminded of their voting duty to a much larger extent. For instance, before the voting duty got abolished in the Canton Aargau, police officers were driving through the villages with speakers and invited the citizens to vote.13 In contrast, in the Canton Bern (one of the two Cantons who abolished the unenforced voting duty), the law changed from ‘‘Voting is a civic duty, but not compulsory’’ to ‘‘The right to vote cannot be made compulsory’’. According to governmental officials in the Canton

142

American Law and Economics Review V9 N1 2007 (135–159)

Table 1. Overview of the Cantons’ institutional arrangements of voting Canton

Voting duty Abolished

Amount of fine

Introduction Postal voting

Schaffhausen (SH)

1876

Still enacted

1995

Aargau (AG) St. Gallen (SG) Thurgau (TG) Zuerich (ZH)

1885 1890 1904 1869

1971 1979 1985 1985

Appenzell Ausserrhoden (AR) Bern (BE) Uri (UR) Obwalden (OW) Nidwalden (NW) Glarus (GL) Appenzell Innerrhoden (AI) Tessin (TI) Schwyz (SZ) Luzern (LU) Zug (ZG) Freiburg (FR) Solothum (SO) Basel-Stadt (BS) Basel-Land (BL) Graubuenden (GR) Wadt (VD) Wallis (VS) Neuenburg (NE) Geneve (GE) Jura (JU)

1908

1996

Until 1973: 1 Sfr. After 1973: 3 Sfr. 2–4 Sfr. 2–5 Sfr. 1 Sfr. Until 1955: 1 Sfr. Until 1972: 1–3 Sfr. Until 1985: 5–10 Sfr. —

1921 1888 1902 1913 1887 1847

1981 Still enacted Still enacted Still enacted Still enacted Still enacted

— — — — — —

1991 1995 1995 1994 1995 1979

1830 — — — — — — — —

Still enacted — — — — — — — —

— — — — — — — — —

— — 1994 1995 1995 1980 1995 1978 1995

— — — — —

— — — — —

— — — — —

— — — 1995 1999

1993 1979 1985 1994

1988

Notes: Details on the Cantons’ institutional arrangements are collected until 1999. Information on the Cantons’ legal regulation of the voting duty stems from the Cantonal laws. The Canton Appenzell Ausserrhoden also had a voting fine in early years, but not for the type of elections investigated in this study. The main data source for the Cantons’ postal voting system is a survey called ‘‘Umfrage ueber die briefliche Stimmabgabe’’. The survey was conducted by the Swiss government and can be found at: www.admin.ch/ch/d/pore/va/doku/pdf/enquete bsa.pdf. Only automatic postal voting is counted.

Bern, most of the citizens were not even aware of this change in the law. Therefore, it seems natural to focus on the Cantons, where citizens knew their duty to vote, at the least because of the fine. However, I will empirically

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

Voting duty Enacted

Voting Laws with Symbolic Fines

143

14. Mail voting never replaced the polls, but was offered as a further option. By the end of 1994, a federal law was enacted, which prescribed the Cantons to introduce the option of postal voting in order to facilitate voting for the citizens. From then on, there was only some variation left with respect to the time until the process of mail voting was organized.

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

check whether the change in the completely unenforced laws had an effect as well. Before going to the econometric analysis, I present the raw data. Figure 1 illustrates VT before and after the abolition of the voting duty for the Cantons Aargau (AG) and Zuerich (ZH); the vertical line indicates the last election under the legal obligation to vote. Also depicted is the development of VT for Schaffhausen (SH), the only Canton, where the voting duty still is enforced. As can be seen from Figure 1, there was a general decline in turnout over the past 40 years. However, VT remained comparatively high in the Canton Schaffhausen, where the voting duty never disappeared. In contrast, the Canton Aargau experienced a substantial decrease in VT, once the moral obligation to vote was suspended (middle picture). Since the development of VT before the law change resembles the other Cantons’ development very much, the law change in the Canton Aargau seems to be exogenous with respect to past voting behavior. In the Canton Zuerich, abolition of the voting duty might have resulted endogenously after a large decrease in voter participation. There, no effect of the removal of the voting duty on VT can be observed. The Cantons St. Gallen (SG) and Thurgau (TG) got rid of their voting duty as well. However, they introduced the option of postal voting at the same time, which makes it impossible to graphically (not statistically) separate these two effects. The graphical analysis lets us suspect that an exogenous law change targeting at the moral obligation to vote might have altered voting behavior substantially. It would be interesting to compare the effect of this law change with the effect of an institutional change, which significantly reduced the voting costs: the introduction of postal voting. Similar to the voting duty, the Swiss Cantons differ with respect to the regulation of postal voting. While certain Cantons introduced the option of postal voting already in the 80’s, the majority gave citizens the possibility to vote by mail in the 90’s (see Table 1 last column).14 The Swiss voting

144

American Law and Economics Review V9 N1 2007 (135–159) Voter Turnout (in %)

Schaffhausen Control Group

90

70 60 50 40 30 1951 1955 1959 1963 1967 1971 1975 1979 1983 1987 1991 Year Voter Turnout (in %)

Aargau

Control Group

90 80 70 60 50 40 30 1951 1955 1959 1963 1967 1971 1975 1979 1983 1987 1991 Year Voter Turnout (in %)

Zuerich

Control Group

90 80 70 60 50 40 30 1951 1955 1959 1963 1967 1971 1975 1979 1983 1987 1991 Year

Figure 1. The impact of the legal voting duty on voter turnout.

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

80

Voting Laws with Symbolic Fines

145

3. Econometric Analysis The focus of this study is on whether an individual’s propensity to vote depends on the legal obligation to vote. Studies based on individual data suggest that an individual’s propensity to vote depends on various other factors such as age, education, gender, income etc.16 Including the 15. Very few Cantons had a system, where it was necessary to send an application first in order to be allowed to vote by mail. Since the reduction in the transaction costs is less evident under this system, these few Cantons were counted as if they didn’t have the option of postal voting. 16. For instance, education, age, income and male sex have been found to have a positive impact on the probability to vote (see Struthers and Young (1989) for an overview of the empirical studies on voting).

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

procedure generally, and also the system of mail voting is very simple. In contrast to the United States, where voters have to register, every eligible Swiss citizen automatically receives the (election) documents per mail. Traditionally, the citizens had to bring the filled-out documents to the poll station. With mail voting, however, a return envelope (with the address of the polling place) is added to the election documents. Therefore, the voter can put his vote in the return envelope and drop it in the nearest letter box. Obviously, this system of (automatic) postal voting brings a large decrease in transaction costs.15 A graphical analysis of the impact of postal voting seems best possible for the Cantons Basel Land (BL), Appenzell Innerrhoden (AI) and Solothurn (SO), which introduced the option of postal voting already between 1978 and 1980. Focusing on these early introducers bears the advantage that there exists a large control group of Cantons, which did not have the option of postal voting at that time. Figure 2 shows the development of VT before and after the option of postal voting was introduced (again, the vertical line is drawn so that after the line, the change is in effect). As becomes evident, the introduction of postal voting does not seem to have increased VT in spite of the considerable reduction in transaction costs. Since VT might depend on many more factors than voting laws and the organization of the voting process, I proceed with the econometric analysis.

146

American Law and Economics Review V9 N1 2007 (135–159) Voter Turnout (in %)

Solothurn

Control Group

90

70 60 50 40 30 1951 1955 1959 1963 1967 1971 1975 1979 1983 1987 1991 Year Voter Turnout (in %)

Basel Landschaft

Control Group

80 70 60 50 40 30 20 1951 1955 1959 1963 1967 1971 1975 1979 1983 1987 1991 Year Voter Turnout (in%)

Appenzell Innerrhoden Control Group

70 60 50 40 30 20 10 1951 1955 1959 1963 1967 1971 1975 1979 1983 1987 1991 Year

Figure 2. The impact of postal voting on early introducer’s turnout.

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

80

Voting Laws with Symbolic Fines

147

institutional factors voting duty (legal obligation to vote) and postal voting into the standard estimation equation yields:

(1)

Vit takes a value of 1, if an individual i was voting in the election year t, and 0 otherwise. Duty is a dummy variable taking a value of 1 if the individual lives in a state s with a minimally fined voting-duty. The dummy variable Postal takes a value of 1 if the individual has the option of postal voting. Zit refers to the individual’s characteristics. Aggregate data on VT are obtained by summing up the number of individual votes in Canton (‘‘state’’) s and dividing them by the population eligible to vote. Since data on aggregate cantonal VT are available, I estimate the following model: VTst = α s + γ t + b1 ∗ Dutyst + b2 ∗ Postalst + b3 ∗ Zst + ust

(2)

VTst denotes VT (in percentage) and Duty and Postal are the Dummy variables described above. As control variables Z, I employ the size of the Canton’s population (in Mio.), the percentage of inhabitants in different age classes, the share of highly educated people and the unemployment rate (for an exact definition of the variables and data sources, see the Data Appendix). Table 2 presents summary statistics. As can be seen therefrom, voter turnout was 54 % on average, with a large variation. Even though elections are held every 4 years (leaving 13 elections in the sample), there are less than 338 observations. First, this is due to the Canton Jura, who was founded in 1977 (and causes 7 missing observations). Secondly, in certain elections, there were not more candidates than seats in a Canton, resulting in acceptance without vote. For these so-called ‘‘Stille Wahlen’’, no turnout data exist. In contrast to the turnout variable, the other data show no missing observations, even though population and age measures have been interpolated, as described in the Data Appendix. Equation (2) is estimated with OLS and WLS. As weights, I use the population older than 19 as a proxy for the eligible population. Note also

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

Vit = b0 + b1 ∗ Dutyst + b2 ∗ Postalst + b3 ∗ Zit +
it

148

American Law and Economics Review V9 N1 2007 (135–159)

Table 2. Summary statistics (1951–1999) Mean

Std. Dev

Min

Max

316

53.9

15.3

17.4

87.7

0 0 0

1 1 1

331 331 331

0.13 0.3 0.18

0.34 0.46 0.38

331 331 331 331 331 331 331 331 331

0.2 29.4 29.4 24 4.7 7.5 4.9 1.9 0.75

0.3 5.3 2.3 2.2 0.5 1.2 1.7 1.3 1.2

0.01 16.2 23.2 19.5 3.2 4.8 2.2 0 0

1.2 40.5 35 30.6 6.3 11.2 10.6 6.5 7

Notes: The table reports summary statistics. Turnout is measured in percentage of the eligible population. The institutional variables are all dummies, where Legal Duty means a legal obligation to vote. The dummy postal takes a value of 1, if the Canton has a system of automatic postal voting, and 0 otherwise. Population measures the number of Cantonal inhabitants in millions. The age structure is captured by the share of residents living in age classes 0–19, 20–39, 40–59, 60–64, 65–74, >75 (in percentage). Education measures the number of high-school graduates among people aged between 15 and 19 (in percentage). The unemployment rate is defined as the percentage of unemployed persons among the active population (i.e., working 6 hours per week or more).

that since time-fixed-effects are estimated, there is no need for including dummies for nation-wide structural breaks.17 Table 3 depicts the estimation results. The first column shows the unweighted regression of Equation (2). As can be seen therefrom, in the Cantons where the minimally sanctioned voting duty was abolished, turnout decreased by as much as 10 percentage points (the effect is slightly smaller in the weighted regression, see column (3)). To check whether abolition of the completely unenforced voting duty had an effect, I include an additional 17. In Switzerland, two institutional changes could have caused a structural break in voter turnout: the introduction of woman suffrage in 1971 and the decrease of the minimal voting age from 20 to 18 in 1991. However, if a dummy for a potential structural break is included, one of the time fixed effects needs to be dropped in order to avoid perfect collinearity. The coefficient of the structural break dummy then equals the dropped time fixed effect.

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

Dependent variable Voter turnout (in %) Institutional variables Legal duty (with fine) Legal duty (without fine) Dummy postal Control variables Population (in 1,000,000) Age 0 to 19 (%) Age 20 to 39 (%) Age 40 to 59 (%) Age 60 to 64 (%) Age 65 to 74 (%) 75 and Older (%) Education (%) Unemployment rate (%)

Observations

Voting Laws with Symbolic Fines

149

Table 3. The effect of the legal voting duty on turnout (1) Legal duty with minimal fine Legal duty without fine Dummy postal Age 0 to 19 (%) Age 20 to 39 (%) Age 40 to 59 (%) Age 60 to 64 (%) Age 65 to 74 (%) Population (in Mio.) Unemployment Education Canton-fixed effects Time-fixed effects Observations R-squared Econometric method

1.28 (1.83) 2.03 (1.89) 0.84 (1.87) 1.27 (1.67) 1.32 (2.99) 0.53 (2.97) −19.56 (16.95) −0.38 (1.05) 0.50 (0.90) Yes Yes 316 0.99 OLS

(3)

(4)

10.22 (3.52)∗∗∗ 0.24 (2.20) 1.28 (1.84) 2.03 (1.90) 0.83 (1.88) 1.26 (1.69) 1.30 (3.01) 0.52 (2.99) −19.49 (16.76) −0.37 (1.05) 0.49 (0.93) Yes Yes 316 0.99 OLS

6.08 (3.26)∗

6.19 (3.47)∗ 0.50 (1.59) −0.04 (1.47) 0.24 (1.28) −1.60 (1.46) −0.97 (1.45) 1.20 (3.41) −2.78 (2.47) −16.84 (9.71)∗ −0.20 (0.72) 1.07 (0.70) Yes Yes 316 0.99 WLS

−0.08 (1.44) 0.32 (1.19) −1.51 (1.37) −0.90 (1.37) 1.32 (3.33) −2.70 (2.40) −16.98 (9.51)∗ −0.21 (0.72) 1.14 (0.58)∗ Yes Yes 316 0.99 WLS

Notes: Dependent variable is Cantonal turnout at the federal elections between 1951 and 1999. Legal duty with minimal fine is a dummy variable taking a value of 1, if the legal voting duty has a symbolic fine and zero otherwise. Legal duty without fine takes a value of 1, if the Canton had a nonfined legal voting duty, and 0 otherwise. The dummy postal takes a value of 1, if the Canton has a system of automatic mail voting, and 0 otherwise. Population is measured in millions, and age structure, education and unemployment in percentage. All estimations include Canton—and election fixed effects. Standard errors clustered at the Canton level depicted in parenthesis.

dummy as for whether the Canton had an unenforced law. As depicted in columns (2) and (4), there was no significant turnout reduction. However, this is not surprising since the abolition of the voting duty was hardly discussed in the two respective Cantons. The result that the removal of the symbolically fined voting duty caused a turnout drop between 6 and 10 percentage points is particularly strong in view of the fact that the inclusion of Canton- and Time-Fixed Effects captures a large part of the variance of the Dummy Duty. A simple

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

10.20 (3.48)∗∗∗

(2)

150

American Law and Economics Review V9 N1 2007 (135–159)

18. Similarly, the controls are highly correlated with the fixed effects and therefore lose significance. 19. Note that poll voting was the only option at the times when the voting duty was fined. As such, the transaction costs from going to the polls were certainly much higher than the fine for not voting. Removal of the fine should not have altered the decision to vote out of monetary considerations.

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

regression of the Dummy Duty on Canton- and Time-Fixed Effects shows that 77 percent of the variance of the Voting-Duty-Dummy can be explained by these Fixed Effects. Therefore, the finding that the coefficient of the Dummy Duty remains significant even with Fixed Effects underlines the statistical importance of the voting duty.18 The natural interpretation of this result is that the abolition of the voting duty decreased VT due to a change in the perception of voting. Before the law change, it was a duty to go to the polls and afterwards, the duty was removed. Nevertheless, it could also be possible that the law change partly affected behavior due to the drop of the fine as well as the transaction costs related with the payment of the fine. As for the transaction costs associated with the payment of the fine, it turns out that they were very close to zero. From telephone calls with people working for the Cantonal governments (in ZH, AG, TG, SG and SH), I learnt that the fines were (or still are) primarily collected by (yearly) bills or by adding the fines to the taxes/communal charges. In very recent times, where the districts were very small (the collection of the fine occurs on the district level), there even existed some district officers (called ‘‘Waechter’’), which came to the defaulting citizens’ houses to collect the fines. Secondly, as can be seen from the Table 1, the amount of the fine has always been very low. The most frequently employed fine was 1 Swiss Frank, which is less than 1 Dollar (the current exchange rate is roughly 0.8 Dollar per Swiss Frank). Therefore, it seems unlikely that the abolition of the fine was responsible for the observed drop in VT.19 Furthermore, if voters react to small changes in the voting costs such as a 1 Dollar amount (drop of the fine), then voters should react to a much larger extent to substantial changes in the voting costs. However, the sign of the Dummy

Voting Laws with Symbolic Fines

151

4. Robustness 4.1. Outliers Since there were only five Cantons with an enforced voting duty, the estimation results might be driven by ‘‘outliers’’. Two Cantons might be particularly influential for the estimation results: The Canton Aargau, because it experienced a particularly large decrease in turnout after abolition of the voting duty (see Figure 1), and the Canton Zuerich, because it is the largest in terms of population and receives the most weight in the weighted regressions. Table 4 depicts the impact of the voting duty, once the Cantons Aargau and Zuerich are dropped. As expected, the estimated impact of the voting duty diminishes, as soon as the Canton Aargau is omitted. On the other hand, the coefficient increases if the Canton Zuerich is dropped from the sample.

4.2. Cantonal Reactions to the Abolition of the Voting Duty The variability of the estimated voting-duty coefficient (see Table 4) indicates a differing effectiveness of the voting duty between the Cantons. In order to asses the Cantonal reactions to the abolition of the voting duty, I estimate the Cantonal effects instead of an average effect. Precisely, I split up the Dummy ‘‘Voting Duty’’ into its individual Cantonal components.21 Columns 1 and 2 in Table 5 display the estimated coefficients for the Cantonal voting duty dummies. As can be seen therefrom, having a legal 20. However, as explained in Funk (2006), postal voting did not only alter the voting costs, but also the social pressure to vote. As such, the relationship between the effect of postal voting and voters’ sensitivity to voting costs is not entirely clear-cut. 21. For instance, the Dummy AG takes a value of 1, if the Canton is Aargau and the voting duty is enforced in this year, and 0 otherwise.

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

Postal is insignificant (see Table 3, Row 3). Therefore, voters do not seem to be very sensitive to voting costs.20 The analysis conducted thus far suggests that the abolition of the voting duty decreased VT significantly, and that it was not due to the elimination of the fine. Therefore, the legal statement that ‘‘citizens are supposed to vote’’ seemed to be effective.

152

American Law and Economics Review V9 N1 2007 (135–159)

Table 4. Robustness check with respect to outliers

Age 0 to 19 (%) Age 20 to 39 (%) Age 40 to 59 (%) Age 60 to 64 (%) Age 65 to 74 (%) Population (in Mio.) Unemployment Education Excluded Canton Canton-fixed effects Time-fixed effects Observations R-squared Econometric method

(2)

(3)

(4)

7.25 (2.83)∗∗ 0.67 (1.86) 1.96 (1.99) 0.85 (1.93) 1.41 (1.80) 1.25 (3.04) 0.53 (3.04) −14.76 (17.17) −0.46 (1.04) 0.53 (0.94) Aargau Yes Yes 303 0.99 OLS

3.23 (1.96) −0.88 (1.50) 0.71 (1.37) −0.72 (1.43) −0.08 (1.43) 1.31 (3.06) −1.61 (2.46) −17.16 (12.66) −0.19 (0.62) 1.49 (0.55)∗∗ Aargau Yes Yes 303 0.99 WLS

12.78 (3.54)∗∗∗ 1.75 (1.93) 2.36 (2.02) 1.39 (2.01) 1.60 (1.78) 0.94 (3.15) 1.22 (3.06) −24.61 (32.74) −0.50 (1.07) 0.86 (0.92) Zuerich Yes Yes 303 0.99 OLS

11.78 (3.91)∗∗∗ 0.46 (1.76) 0.46 (1.41) −0.70 (1.53) −0.46 (1.51) −1.58 (2.90) −1.25 (2.47) −3.41 (26.90) −0.78 (0.71) 2.00 (0.51)∗∗∗ Zuerich Yes Yes 303 0.99 WLS

Notes: Dependent variable is Cantonal turnout at the federal elections between 1951 and 1999. Legal duty with minimal fine is a dummy variable taking a value of 1, if the legal voting duty has a symbolic fine and zero otherwise. The Dummy Postal takes a value of 1, if the Canton has a system of automatic mail voting, and 0 otherwise. Population is measured in millions, and age structure, education and unemployment in percentage. All estimations include Canton- and Election Fixed Effects. Standard errors clustered at the Canton level depicted in parenthesis.

obligation to vote lead to a 16-18 percent higher voter turnout in the Canton Aargau, but had no significant effect in the Canton Zuerich - the effects in the Cantons Thurgau and St. Gallen lie between 5 and 10 percent. Therefore, only for the Canton Zuerich, no significant positive coefficient can be observed. Two explanations seem plausible for this ‘‘aberrant’’ Canton. First, the Canton Zuerich was the only Canton who increased the amount of the fine substantially (from 1 Swiss frank to 10 Swiss franks; see Table 1), as long as the duty was enforced. This conversion from a symbolic fine into a real fine might have allowed to ‘‘buy’’ not voting and removed

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

Legal Duty with minimal fine Dummy postal

(1)

Voting Laws with Symbolic Fines

153

Table 5. Individual effects and check for endogeneity (1) Dummy Aargau

Dummy St. Gallen Dummy Zuerich Lead-Dummy Aargau Lead-Dummy Thurgau Lead-Dummy St. Gallen Lead-Dummy Zuerich Controls Canton-fixed effects Time-fixed effects Observations R-squared Econometric method

18.43 (2.30)∗∗∗ 7.92 (3.41)∗∗ 10.75 (2.48)∗∗∗ 1.51 (2.20)

16.49 (2.02)∗∗∗ 4.68 (2.18)∗∗ 7.43 (1.59)∗∗∗ 0.71 (1.53)

Yes Yes Yes 316 0.99 OLS

Yes Yes Yes 316 0.99 WLS

(3)

(4)

19.09 (2.64)∗∗∗ 8.79 (3.55)∗∗ 13.43 (2.74)∗∗∗ 1.95 (2.72) −1.40 (2.20) −2.47 (2.69) −16.24 (2.87)∗∗∗ −2.98 (3.27) Yes Yes Yes 316 0.99 OLS

17.70 (2.17)∗∗∗ 5.66 (2.08)∗∗ 9.97 (1.66)∗∗∗ 1.75 (1.72) −2.14 (1.08)∗ −3.26 (1.67)∗ −10.86 (1.58)∗∗∗ −5.24 (1.99)∗∗ Yes Yes Yes 316 0.99 WLS

Notes: Dependent Variable is Cantonal turnout at the federal elections between 1951 and 1999. The individual dummies (Dummy Aargau, Dummy Thurgau, Dummy St. Gallen, Dummy Zuerich) take a value of 1, if the respective Cantons have a voting duty and 0 otherwise. The postal voting dummy and the controls are the same as in Tables 3 and 4 and are not listed. All estimations include Canton- and election-fixed effects. Standard errors clustered at the Canton level in parenthesis.

the feelings of guilt of not voting. Gneezy and Rustichini (2000) found a very similar effect in a completely different context: introducing fines for parents, who picked up their children from school too late, increased the number of late-coming parents. Their interpretation for this result was that parents perceived the fine as a price, which allowed to ‘‘buy’’ being late. For sure, the fact that the Canton with the highest (nominal as well as real) fine experienced the smallest turnout decrease after its removal underlines my conjecture that it is not the monetary drop of the fine that caused people to vote less. If so, one should have seen the biggest drop in the Canton Zuerich. Secondly, removal of the voting duty might have had the smallest effect in Zuerich since this Canton abandoned the voting duty as the latest, and turnout before abolition was the lowest as well. Simple correlations

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

Dummy Thurgau

(2)

154

American Law and Economics Review V9 N1 2007 (135–159)

4.3. Endogeneity The decision to abolish the voting duty might be endogenous. If the Cantons chose to abolish the voting duty due to a particularly strong decrease in VT, then a lower VT after the abolition of the voting duty might reflect a Cantonal trend, which already started before the law change. One crude way for testing the endogeneity of law changes is to include lead dummies in the regressions.22 Columns three and four in Table 5 indicate that endogeneity might be a problem for the Canton St. Gallen, since a large drop in turnout preceded the law change. A straightforward way to control for such Cantonal differences in the development of VT is to include Canton-specific time trends. Unfortunately, the Canton-specific time trends (together with Canton- and Time-Fixed Effects) explain nearly the whole variance of the voting duty dummies. Therefore, I do not have sufficient variation to include Cantonal trends. As such, I observe that the abolition of the voting duty caused a drop in average turnout. The drop was particularly large in the Canton Aargau, where turnout rates were very high at the time of abolition. From a theoretical perspective, it is not so surprising that the biggest drop in VT occurred in the Canton Aargau. Since the removal of the voting duty may have changed the perception of the social value/esteem of voting, it potentially affected all the law-adherents’ behavior, which were 80 percent in the Canton Aargau, but only between 40 and 50 percent in the other Cantons. Additionally, the update in the beliefs about the value of voting is likely to be higher in a community where most of the people go to the polls than in a community where many people don’t vote in spite of their duty to do so.

22. See Friedberg (1998) for a similar approach in her analysis on divorce laws.

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

between the size of the ‘‘expressive effect’’ (coefficient before Cantonal voting dummy, weighted and unweighted) and turnout before removal of the duty are higher than 90 percent in both specifications. As such, abolition of an unenforced law seems to have a particularly strong effect if people adhere to the law to a high degree.

Voting Laws with Symbolic Fines

155

5. Conclusions

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

This article empirically analyzes the effect of a hardly sanctioned law on behavior. Using panel data on VT, I find that the legal abolition of the (minimally fined) voting duty significantly decreased VT. Since the drop in VT was very unlikely caused by the drop of the fine or the transaction costs associated with paying the fine, this result is consistent with the existence of a certain (or at least situational) expressive function of law. As for the ‘‘size’’ of the expressive effect, it differed between the Cantons. The abolition of the voting duty had the biggest impact on VT in the Canton Aargau, where VT was the highest at the time of the law change (strong voting norm). A substantial, but lower turnout drop is found in the Cantons St. Gallen and Thurgau. In the Canton Zuerich, where the voting duty was abolished at a time of low law-obedience, the effect was insignificant. Therefore, removal of an unenforced law might have a particularly strong effect on behavior in a situation where people generally adhere to the law. Although it is implausible that the observed drop in VT resulted due to the change in the ‘‘monetary’’ incentives (drop fine), I cannot rule out that the existence of minimal sanctions helped maintaining the social voting norm strong. Firstly, because not-voting was minimally sanctioned, the citizens were at least yearly reminded of their voting duty. Secondly, the amount of informal sanctioning might have been weakly connected to the enforcement of the sanction, since the persons responsible for collecting the fines knew the ‘‘sinners’’ for certain. As such, even if passing a law affected behavior with no formal sanctions, a minimal or random enforcement (at the beginning) might strengthen its ‘‘expressive’’ effect. However, since the data set is quite limited (identification stems from four law changes), it is too early to draw definite conclusions about the expressive function of law. Although (panel) data on unenforced (or hardly enforced/sanctioned) laws are hard to find, there is definite need for more research in this area. A panel with more law changes would allow to more carefully address the conditions, under which ‘‘expressive effects’’ of law exist. For instance, the present analysis on mandatory voting laws suggests that the impact of a law change (abolition voting duty) might depend on whether it is truly exogenous or results after a remarkable change in behavior.

156

American Law and Economics Review V9 N1 2007 (135–159)

Data Appendix Data on VT at federal elections were obtained from the Swiss Federal Statistical Office. Cantonal time series start at 1919. Missing data points occur

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

Next to the effect of mandatory voting laws, I was able to test the effect of (optional) postal voting on VT. The regressions show that this institutional change did not lead to a significant increase in VT in spite of the large reduction in transaction costs. However, while the reduction in voting costs should have increased the incentives to vote, the newly introduced nonverifiability of the voting act might have undermined the very same (Funk, 2006). The results from this study are most likely applicable to other situations with public good character (blood donations, garbage removal etc.). In particular, the analysis suggests that governmental actions which appeal to the civic duty (i.e. by legally prescribing it) may have substantial effects. In contrast, as the example of postal voting shows, actions which target at reducing the costs of provision of the public good might be less effective in certain situations. While a legal prescription may motivate citizens to contribute to public goods, a harsh enforcement of law might be counterproductive in this area. In Switzerland, there was one Canton (Zuerich) which raised the fine for nonvoting from 1 Dollar to roughly 8 Dollars and therefore turned the ‘‘symbolic’’ fine into a ‘‘real’’ fine. Surprisingly, the data indicate a negative correlation between the amount of the fine and VT. While the amount of the fine might be endogenous, the increase in the fine might also have ‘‘legitimated’’ nonvoting in the eyes of certain citizens: while conscious citizens certainly experienced feelings of guilt for not voting under a fine as low as 1 Dollar, they might have no longer felt guilty under a substantial fine and chosen to ‘‘pay’’ for not going to the polls. In sum, the study provided some insights about how citizens reacted to changes in the (only minimally fined) legal obligation to vote (expressive function of law), and to a reduction in the transaction costs of voting. Since this is one of the first (Real World) Experiments, which studies the contribution to a public good under varying ‘‘incentive schemes’’, more research is highly warranted for gaining a broader understanding of this topic.

Voting Laws with Symbolic Fines

157

References Adams, Renee, and Daniel Ferreira. 2005. ‘‘Do Directors Perform for Pay?’’ CEI Working Paper Series No. 2005–2. Adler, Matthew. 1999. ‘‘Expressive Theories of Law,’’ Working Paper No. 275, University of Pennsylvania Law School.

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

for elections, where the number of candidates did not exceed the number of seats in a Canton. In these cases, no votes took place but the candidates were elected automatically (‘‘Stille Wahlen’’). Information on the Cantons’ institutional details on the voting duty stems from Cantonal laws. A summary on the Cantons’ laws on postal voting can be found in the document ‘‘Umfrage ueber die briefliche Stimmabgabe’’ at: www.admin.ch/ch/d/pore/va/doku/pdf/enquete bsa.pdf. Most of the control variables stem from the population census. Population is measured in million number of inhabitants (per Canton and year). Population data are collected in the population census at a roughly decennial basis. Intermediary values were obtained by linear interpolation. Age is measured as percentage of inhabitants in the following age classes: 0–19, 20–39, 40–59, 60–64, 65–74, > 75. The data stem from the population census as well and missing data were obtained by linear interpolation. Higher education is approximated by the number of ‘‘high-school degrees’’, in percentage of the number of 15 to 19 year old people. High-school is put in quotation marks, because the Swiss school system is different from the American one. After 6 years of primary school (commonly attended from 6 to 12 years), there are three options: the ‘‘Realschule’’ (lowest level), the ‘‘Sekundarschule’’ (intermediate level) and the ‘‘Gymnasium’’ (highest level, denoted as ‘‘highschool’’). While completion of the first two types of education takes between 2 and 3 years, ‘‘high-school’’ lasts 5 years. Therefore, ‘‘high-school’’ is commonly completed at age 18 and the number of ‘‘high-school’’-degrees per number of 15 to 19 year old teenagers represents an adequate indicator for the frequency of attendance of higher education. The data are available on an annual basis from the Statistical Yearbooks of Switzerland. The unemployment rate in Switzerland is defined as the number of unemployed persons per active population. The active population consists of individuals working more than 6 hours per week. Unemployment Rates in Switzerland are measured in relation to the active population. Data are available on an annual basis from the State Secretariat for Economic Affairs (seco).

158

American Law and Economics Review V9 N1 2007 (135–159)

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

Bohnet, Iris, and Robert D. Cooter. 2005. ‘‘Expressive Law: Framing or Equilibrium Selection?’’ Working Paper No. 138, University of California-Berkeley Public Law. Bohnet, Iris, Bruno S. Frey, and Steffen Huck. 2001. ‘‘More Order with less Law: On Contract Enforcement, Trust, and Crowding,’’ 95 American Political Science Review 131–44. Cohen, Alma, and Liran Einav. 2003. ‘‘The Effects of Mandatory Seat Belt Laws on Driving Behavior and Traffic Fatalities,’’ 85(4) Review of Economics and Statistics 828–43. Cooter, Robert. 1998. ‘‘Expressive Law and Economics,’’ 27 Journal of Legal Studies 585–608. Cooter, Robert. 2000. ‘‘Do Good Laws Make Good Citizens? An Economic Analysis of Internalized Norms,’’ 86 Virginia Law Review 1577–601. Dau-Schmidt, Kenneth. 1990. ‘‘An Economic Analysis of the Criminal Law as a Preference-Shaping Policy,’’ 1 Duke Law Journal 1–38. Dharmapala, Dhammika, and Richard McAdams. 2003. ‘‘The Condorcet Jury Theorem and the Expressive Function of Law: A Theory of Informative Law,’’ 5 American Law and Economics Review 1–31. Ellickson, Robert C. 2001. ‘‘The Market for Social Norms,’’ 3 American Law and Economics Review 1–49. Fisman, Ray, and Edward Miguel. 2006. ‘‘Cultures of Corruption: Evidence from Diplomatic Parking Tickets,’’ NBER Working Paper No. 12312. Friedberg, Leora. 1998. ‘‘Did Unilateral Divorce Rates Raise Divorce Rates? Evidence from Panel Data,’’ 88(3) American Economic Review 608–27. Funk, Patricia. 2006. ‘‘Modern Voting Tools, Social Incentives and Voter Turnout: Theory and Evidence,’’ Paper presented at the annual meeting of the American Economic Association, Chicago, IL. Gneezy, Uri, and Aldo Rustichini. 2000. ‘‘A Fine is a Price,’’ 29(1) Journal of Legal Studies 1–18. Hasen, Richard L. 1996. ‘‘Voting without Law?’’ 144 University of Pennsylvania Law Review 2135–79. Jackman, Simon. 2001. ‘‘Compulsory Voting,’’ In International Encyclopedia of the Social and Behavioral Sciences. Oxford, UK: Elsevier. Kahan, Dan M. 1996. ‘‘What Do Alternative Sanctions Mean?’’ 63 University of Chicago Law Review 591–653. Lessig, Lawrence. 1998. ‘‘The New Chicago School,’’ 27(2) Journal of Legal Studies 661–91. McAdams, Richard. 1997. ‘‘The Origin, Development and Regulation of Norms,’’ 96 Michigan Law Review 338–433. McAdams, Richard. 2000a. ‘‘An Attitudinal Theory of Expressive Law,’’ 79 Oregon Law Review 339–90.

Voting Laws with Symbolic Fines

159

Downloaded from http://aler.oxfordjournals.org/ at Aarhus Universitets Biblioteker / Aarhus University Libraries on October 17, 2012

McAdams, Richard. 2000b. ‘‘A Focal Point Theory of Expressive Law,’’ 86 Virginia Law Review 1649–729. McAdams, Richard, and Janice Nadler. 2005. ‘‘Testing the Focal Point Theory of Legal Compliance: Expressive Influence in an Experimental Hawk/Dove Game,’’ 2 Journal of Empirical Legal Studies 87–123. McAdams, Richard, and Eric B. Rasmusen. Forthcoming. ‘‘Norms in Law and Economics,’’ The Handbook of Law and Economics. Pildes, Richard H. 1998. ‘‘Why Rights are not Trumps: Social Meanings, Expressive Harms and Constitutionalism,’’ 27(2) Journal of Legal Studies 725–63. Posner, Eric A. 2000. Law and Social Norms. Cambridge, MA: Harvard University Press. Scott, Robert E. 2000. ‘‘The Limits of Behavioral Theories of Law and Social Norms,’’ 86 Virginia Law Review 1603–47. Southwell, Priscilla, and Justin Burchett. 2000. ‘‘Does Changing the Rules Change the Players? The Effect of All-Mail Elections on the Composition of the Electorate,’’ 81 Social Science Quarterly 837–45. Struthers, John, and Alistair Young. 1989. ‘‘Economics of Voting: Theories and Evidence,’’ 16 Journal of Economic Studies 1–42. Sunstein, Cass. 1987. ‘‘On the Expressive Function of Law,’’ 144 University of Pennsylvania Law Review 2021–53. Tyran, Jean-Robert, and Lars Feld. 2006. ‘‘Achieving Compliance when Legal Sanctions are Non-deterrent,’’ 108(1) Scandinavian Journal of Economics 135–65.