Appendix A. Compulsory Voting and Turnout in Switzerland, Aggregate-Level Data

Online Appendix

Appendix A. Compulsory voting and turnout in Switzerland, aggregate-level data

Table A.1. Swiss cantons with compulsory voting, for elections since 1971

Canton / Compulsory voting rule / Time period
Aargau / No CV / 1971-2011
Appenzell Inner Rhodes / Non-sanctioned CV / 1971-2011
Appenzell Outer Rhodes / Non-sanctioned CV / 1971-1995
No CV / 1996-2011
Basel-Landschaft / No CV / 1971-2011
Basel-Stadt / No CV / 1971-2011
Bern / Non-sanctioned CV / 1971-1980
No CV / 1981-2011
Fribourg / No CV / 1971-2011
Geneva / No CV / 1971-2011
Glarus / Non-sanctioned CV / 1971-2011
Graubünden / No CV / 1971-2011
Jura / No CV / 1971-2011
Lucerne / No CV / 1971-2011
Neuchatel / No CV / 1971-2011
Nidwalden / Non-sanctioned CV / 1971-2011
Obwalden / Non-sanctioned CV / 1971-2011
Schaffhausen / Sanctioned CV / 1971-2011
Schwyz / No CV / 1971-2011
Solothum / No CV / 1971-2011
St. Gallen / Sanctioned CV / 1971-1978
No CV / 1979-2011
Thurgau / Sanctioned CV / 1971-1984
No CV / 1985-2011
Ticino / Non-sanctioned CV / 1971-2011
Uri / Non-sanctioned CV / 1971-2011
Valais / No CV / 1971-2011
Vaud / No CV / 1971-2011
Zug / No CV / 1971-2011
Zurich / Sanctioned CV / 1971-1984
No CV / 1985-2011

Note: Information retrieved from Funk (2007).

Figure A.1. Aggregate level turnout in Switzerland, by election and by compulsory voting rules

Appendix B. Distributions of citizens’ reluctance to turn out to vote under voluntary voting rules

Figure B.1. Distribution of citizens’ reluctance to turn out to vote under voluntary voting rules in Australia

Figure B.2. Distribution of citizens’ reluctance to turn out to vote under voluntary voting rules in Belgium

Figure B.3. Distribution of citizens’ reluctance to turn out to vote under voluntary voting rules in Brazil

1

Appendix C. Operationalization of variables in the analyses of election studies in Australia, Belgium and Brazil

Gender

Gender of the respondent: code 0 for male voters; code1 for female voters

Age

Age of the respondent in the year of the interview, calculated as the difference between the year of the interview and the year of birth.

Educational level

Australia

Measure of whether or not the respondent obtained a qualification since leaving high school. Respondents are divided in two categories: (1) no qualification since leaving high school, (2) qualification since leaving high school.

Belgium

Measure of the educational level the respondent obtained. Respondents are divided in three categories: (1) none/primary education only/uncompleted secondary education, (2) secondary education, (3) tertiary education.

Brazil

Measure of the educational level the respondent obtained. Respondents are divided in three categories: (1) none/primary education, (2) secondary education, (3) tertiary education.

Social class

Australia

Self-reported social class of the respondent – scale ranging from 1 to 3 in which: (1) signifies working class, (2) middle class, (3) upper class.

Belgium

-Self-reported social class of the respondent – scale ranging from 1 to 4 in which: (1) signifies working class, (2) lower middle class, (3) upper middle class, (4) upper class.

-As self-reported social class was not included in the 2014 questionnaire, income is used as a proxy. Respondents were asked to indicate their total monthly revenue.

Brazil

-Self-reported social class of the respondent – scale ranging from 1 to 6 in which: (1) signifies extreme lower class, (2) lower class, (3) lower middle class, (4) middle class, (5) upper middle class, (6) extreme upper class.

-As self-reported social class was not included in both 2006 and 2002 questionnaires, a classification based on household items and level of education of the breadwinner, on the one hand, and income, on the other hand, is used as a proxy in 2006 and 2002, respectively. Such a classification is either available in the BES dataset (as in 2002) or is performed according to the Instituto Brasileiro de Geografia e Estatistica’s (IBGE) parameters (IBGE, 1998).[1]

Religious attendance

Australia

Self-reported religious attendance – scale ranging from 1 to 6 in which: (1) signifies never, (2) less than once a year, (3) at least once a year, (4) several times a year, (5) at least once a month, (6) at least once a week.

Belgium

Self-reported religious attendance – scale ranging from 1 to 6 in which: (1) signifies never, (2) only on social occasions/very seldom, (3) only on major holidays/a few times a year, (4) about once a month, (5) several times a month, (6) every Saturday or Sunday.

Brazil (2014, 2002)

Self-reported religious attendance – scale ranging from 1 to 6 in which: (1) signifies never, (2) rarely, (3) few times a year, (4) one or two times a month, (5) once a week, (6) two or more times a week.

Voluntary turnout

Australia

Measure of whether the respondent would have turned out if it would not have been compulsory. Respondents could indicate continued engagement on a scale ranging from 1 to 5, in which (1) signifies would definitely not have voted, (2) would probably not have voted, (3) might, might not have voted, (4) probably would have voted, (5) definitely would have voted. This variable has been standardised to run from 0 to 1 so that a one-unit increase signifies an increase from the most reluctant to the most willing voter.

Belgium

Measure of whether the respondent would still turn out to vote if it were not compulsory anymore. Respondents could indicate continued engagement on a scale ranging from 1 to 4, in which (1) signifies never, (2) sometimes, (3) generally, (4) always. Standardised to run from 0 to 1 so that a one-unit increase signifies an increase from the most reluctant to the most willing voter.

Brazil

Measure of whether the respondent would still turn out to vote if it were not compulsory anymore. Respondents could indicate continued engagement on a scale ranging from 1 to 3, in which (1) signifies no, (2) maybe/it depends, (3) yes. Standardised to run from 0 to 1 so that a one-unit increase signifies an increase from the most reluctant to the most willing voter.

Distance to party

Australia

Ideological distance of the voter to each party respectively. First, the party’s positions are calculated by taking the mean ideological placement of that party by all highly educated voters (=qualification since leaving high school). The distance to the party is the absolute value of difference between the respondent’s self-placement and the party’s position.

Belgium

Ideological distance of the voter to each party respectively. First, the party’s positions are calculated by taking the mean ideological placement of that party by all highly educated voters (=tertiary educated). The distance to the party is the absolute value of difference between the respondent’s self-placement and the party’s position.

Brazil

Ideological distance of the voter to each party respectively. First, the party’s positions are calculated by taking the mean ideological placement of that party by all highly educated voters (=tertiary educated). The distance to the party is the absolute value of difference between the respondent’s self-placement and the party’s position.

Strength of party identification

Australia

Measure that indicates the extent to which a voter feels attached to a party. If the voter does not identify with a party, this amounts to 0. For the party the voter identifies with – if any – it indicates the strength of the identification, ranging from 1 (not very strong supporter) over 2 (fairly strong supporter) to 3 (very strong supporter).

Belgium

Measure that indicates the extent to which a voter feels attached to a party. If the voter does not identify with a party, this amounts to 0. For the party the voter identifies with – if any – it indicates the strength of the identification, ranging from 1 (not close at all) over 2 (not close), 3 (close) to 4 (very close).

Brazil

Measure that indicates the extent to which a voter likes a party more than others. If the voter does not like a party more than others, this amounts to 0. For the party the voter likes more than others it indicates the strength of such a liking, ranging from 1 (likes a little) over 2 (likes somewhat) to 3 (likes a lot).

1

Appendix D. Reported turnout in Swiss election surveys

Table D.1. Actual and reported turnout rates in Swiss election surveys (1971-2011)

Election / Actual turnout level (%) / Reported turnout in election sample (%)
1971 / 56.4 / 60.8
1975 / 52.4 / 58.8
1979 / 48.1 / 73.1
1987 / 47.4 / 64.5
1991 / 46.0 / 63.7
1995 / 42.3 / 62.1
1999 / 43.2 / 61.8
2003 / 45.2 / 68.1
2007 / 48.3 / 68.9
2011 / 49.1 / 74.3

Sources: IDEA turnout level data ( Selects cumulative data file, 1971-2011.

1

Appendix E. Matching, Propensity Score Matching Results by year

Table E.1. Matching results logit analysis (voluntary versus compulsory voting)

Full / Matched
1971
Treatment units / 207 / 149
Control units / 157 / 149
Mean distance treatment group / 0.583 / 0.573
Mean distance control group / 0.549 / 0.559
Improvement in balance / 60.27%
1995
Treatment units / 704 / 700
Control units / 2,110 / 700
Mean distance treatment group / 0.266 / 0.264
Mean distance control group / 0.245 / 0.263
Improvement in balance / 93.78%
1999
Treatment units / 208 / 204
Control units / 810 / 204
Mean distance treatment group / 0.256 / 0.251
Mean distance control group / 0.191 / 0.248
Improvement in balance / 95.99%
2003
Treatment units / 428 / 427
Control units / 1,496 / 427
Mean distance treatment group / 0.232 / 0.231
Mean distance control group / 0.220 / 0.231
Improvement in balance / 96.38%
2007
Treatment units / 344 / 343
Control units / 1,230 / 343
Mean distance treatment group / 0.246 / 0.245
Mean distance control group / 0.211 / 0.243
Improvement in balance / 93.83%
2011
Treatment units / 341 / 340
Control units / 1,148 / 340
Mean distance treatment group / 0.246 / 0.246
Mean distance control group / 0.224 / 0.244
Improvement in balance / 94.26%

Table E.2. Matching results logit analysis (compulsory voting with sanctions versus any other type of voting)

Full / Matched
1971
Treatment units / 140 / 127
Control units / 157 / 127
Mean distance treatment group / 0.483 / 0.475
Mean distance control group / 0.461 / 0.468
Improvement in balance / 67.43%
1995
Treatment units / 301 / 299
Control units / 2,110 / 299
Mean distance treatment group / 0.132 / 0.131
Mean distance control group / 0.124 / 0.131
Improvement in balance / 98.54%
1999
Treatment units / 9 / 9
Control units / 810 / 9
Mean distance treatment group / 0.013 / 0.013
Mean distance control group / 0.011 / 0.014
Improvement in balance / 80.02%
2003
Treatment units / 227 / 225
Control units / 1,496 / 225
Mean distance treatment group / 0.143 / 0.142
Mean distance control group / 0.130 / 0.142
Improvement in balance / 99.69%
2007
Treatment units / 64 / 64
Control units / 1,230 / 64
Mean distance treatment group / 0.054 / 0.054
Mean distance control group / 0.049 / 0.054
Improvement in balance / 96.99%
2011
Treatment units / 37 / 37
Control units / 1,148 / 37
Mean distance treatment group / 0.040 / 0.040
Mean distance control group / 0.031 / 0.040
Improvement in balance / 95.94%

Table E.3. Matching results conditional logit analysis (voluntary versus compulsory voting)

Full / Matched
1971
Treatment units / 207 / 149
Control units / 157 / 149
Mean distance treatment group / 0.583 / 0.573
Mean distance control group / 0.549 / 0.559
Improvement in balance / 60.27%
1995
Treatment units / 701 / 698
Control units / 2,099 / 698
Mean distance treatment group / 0.266 / 0.265
Mean distance control group / 0.245 / 0.264
Improvement in balance / 94.77%
1999
Treatment units / 207 / 203
Control units / 802 / 203
Mean distance treatment group / 0.255 / 0.250
Mean distance control group / 0.192 / 0.250
Improvement in balance / 99.17%
2003
Treatment units / 426 / 424
Control units / 1,493 / 424
Mean distance treatment group / 0.232 / 0.219
Mean distance control group / 0.231 / 0.230
Improvement in balance / 93.57%
2007
Treatment units / 342 / 341
Control units / 1,224 / 341
Mean distance treatment group / 0.246 / 0.245
Mean distance control group / 0.211 / 0.243
Improvement in balance / 94.78%
2011
Treatment units / 338 / 337
Control units / 1,145 / 337
Mean distance treatment group / 0.245 / 0.244
Mean distance control group / 0.223 / 0.243
Improvement in balance / 94.29%

Table E.4. Matching results conditional logit analysis (voluntary versus compulsory voting with sanctions)

Full / Matched
1971
Treatment units / 140 / 127
Control units / 157 / 127
Mean distance treatment group / 0.483 / 0.475
Mean distance control group / 0.461 / 0.468
Improvement in balance / 67.43%
1995
Treatment units / 299 / 297
Control units / 2,099 / 297
Mean distance treatment group / 0.132 / 0.131
Mean distance control group / 9.124 / 0.130
Improvement in balance / 92.22%
1999
Treatment units / 9 / 9
Control units / 802 / 9
Mean distance treatment group / 0.014 / 0.014
Mean distance control group / 0.011 / 0.014
Improvement in balance / 92.26%
2003
Treatment units / 227 / 225
Control units / 1,493 / 225
Mean distance treatment group / 0.143 / 0.142
Mean distance control group / 0.130 / 0.142
Improvement in balance / 99.99%
2007
Treatment units / 63 / 63
Control units / 1,224 / 63
Mean distance treatment group / 0.054 / 0.054
Mean distance control group / 0.049 / 0.054
Improvement in balance / 95.93%
2011
Treatment units / 37 / 37
Control units / 1,145 / 37
Mean distance treatment group / 0.040 / 0.040
Mean distance control group / 0.031 / 0.040
Improvement in balance / 99.07%

Appendix F. Variables included in analyses on Selects data

Female: 0 = male respondents; 1 = female respondents

Age: Age of the respondent

Education: 1 = lower educated (primary school, compulsory education, basic vocational training); 2 = middle educated (vocational education, diploma school, high school); 3 = higher educated (higher vocational training, vocational college, university)

Religious attendance:0 = never, 1 = rarely, 2 = sometimes, 3 = often

Partisan:0 = no party attachment; 1 = party attachment

- The strength of party attachments was included in less election studies, implying the estimation sample would have been smaller. Therefore, it was decided to rely on a dichotomous variable distinguishing between partisans and non-partisans in Switzerland).

- For the alternative specific conditional logit models, partisanship is included as an alternative specific-variable, implying that what is included is whether a respondent identifies with a particular party (not whether or not a citizen identifies with a party in general).

Left-right:Self-placement (and party placements) on a 0-10 left-right scale

Vote choice:Harmonized variable distinguishing between different parties that obtained vote in elections between 1971 and 2011

Appendix G. The impact of compulsory voting on choosing the most proximate party in Australia (full models)

AUS 2001 / AUS 2004 / AUS 2007 / AUS 2010 / AUS 2013
b
(s.e.) / b
(s.e.) / b
(s.e.) / b
(s.e.) / b
(s.e.)
Would not vote / -0.488* / -0.833* / -0.550* / -0.450* / -0.083
(0.243) / (0.367) / (0.277) / (0.224) / (0.170)
Female / 0.047 / -0.270* / -0.085 / 0.008 / -0.095
(0.118) / (0.132) / (0.111) / (0.099) / (0.075)
Age / -0.013** / -0.011** / -0.004 / 0.003 / 0.006**
(0.004) / (0.004) / (0.004) / (0.003) / (0.002)
Middle educated (ref : low) / --- / --- / --- / --- / ---
Higher educated (ref: low) / -0.005 / 0.258 / -0.195 / -0.087 / -0.015
(0.125) / (0.142) / (0.123) / (0.106) / (0.084)
Partisan / 0.103 / 0.460*** / -0.032 / 0.099 / 0.220***
(0.070) / (0.079) / (0.062) / (0.058) / (0.045)
Constant / 0.111 / -0.887 ** / 0.522* / -0.363 / -0.766***
(0.273) / (0.297) / (0.255) / (0.242) / (0.178)
N / 1,269 / 1,144 / 1,338 / 1,699 / 2,924
pseudo R2 / 0.015 / 0.053 / 0.005 / 0.005 / 0.011

Note: Logit coefficients and standard errors (in parentheses) are reported. Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.

Appendix H. The impact of compulsory voting on choosing the most proximate party in Belgium (full models)

BEL-F 1991 / BEL-F 1999 / BEL-F 2014 / BEL-W 1991 / BEL-W 1999 / BEL-W 2014
b
(s.e.) / b
(s.e.) / b
(s.e.) / b
(s.e.) / b
(s.e.) / b
(s.e.)
Would not vote / -0.464** / -0.597** / 0.213 / -1.050*** / -0.207 / -0.435
(0.166) / (0.179) / (0.294) / (0.212) / (0.205) / (0.306)
Female / 0.023 / 0.079 / -0.027 / -0.062 / -0.093 / -0.206
(0.127) / (0.123) / (0.215) / (0.147) / (0.150) / (0.210)
Age / 0.006 / -0.011* / -0.001 / -0.007 / 0.003 / 0.005
(0.005) / (0.004) / (0.007) / (0.005) / (0.005) / (0.007)
Middle educated (ref: low) / 0.178 / 0.252 / -0.223 / 0.079 / -0.048 / 0.462
(0.164) / (0.168) / (0.290) / (0.186) / (0.201) / (0.291)
Higher educated (ref: low) / 0.477** / 0.507** / -0.023 / 0.285 / 0.034 / 0.561*
(0.169) / (0.163) / (0.298) / (0.185) / (0.192) / (0.281)
Partisan / --- / --- / 0.102 / --- / --- / 0.023
(0.077) / (0.074)
Constant / -1.962*** / -0.920** / -1.818*** / -0.261 / -0.591 / -1.725***
(0.243) / (0.276) / (0.520) / (0.279) / (0.316) / (0.470)
N / 1,904 / 1,629 / 679 / 933 / 787 / 536
pseudo R2 / 0.017 / 0.026 / 0.005 / 0.032 / 0.003 / 0.017

Note: Logit coefficients and standard errors (in parentheses) are reported. Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.

Appendix I. The impact of compulsory voting on choosing the most proximate party in Brazil (full models)

BRA 2002 / BRA 2006 / BRA 2010 / BRA 2014
b
(s.e.) / b
(s.e.) / b
(s.e.) / b
(s.e.)
Would not vote / -0.253 / -0.058 / -0.091 / -0.130
(0.131) / (0.233) / (0.159) / (0.124)
Female / -0.030 / 0.171 / -0.195 / -0.088
(0.127) / (0.228) / (0.153) / (0.116)
Age / 0.002 / -0.013 / -0.007 / 0.007
(0.005) / (0.009) / (0.006) / (0.004)
Middle educated (ref : low) / 0.207 / 0.310 / -0.081 / 0.254
(0.144) / (0.281) / (0.178) / (0.139)
Higher educated (ref: low) / 0.235 / 1.017** / 0.552* / 0.562***
(0.190) / (0.345) / (0.221) / (0.159)
Partisan / 0.065 / 0.040 / -0.011 / -0.016
(0.053) / (0.102) / (0.063) / (0.052)
Constant / 0.136 / -0.125 / 0.491 / -0.372
(0.248) / (0.448) / (0.304) / (0.251)
N / 1,330 / 417 / 713 / 1,371
pseudo R2 / 0.007 / 0.033 / 0.012 / 0.009

Note: Logit coefficients and standard errors (in parentheses) are reported. Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001. We follow the recommended use of the Brazil election studies data, weights were applied to correct for the overrepresentation of voters in Sao Paolo in the dataset and for the under- or overrepresentation of socio-demographic groups in the 2002 (PESO_NAC variable), 2006 (peso_1 variable) and 2014 (FPOND variable) election studies.

Appendix J. Distribution of correct voting in the elections studies included

Election study / % respondents voting for
the most proximate party
Australia 2001 / 37.44
Australia 2004 / 33.31
Australia 2007 / 50.70
Australia 2010 / 46.75
Australia 2013 / 46.99
Belgium (Flanders) 1991 / 15.84
Belgium (Flanders) 1999 / 21.39
Belgium (Flanders) 2014 / 15.76
Belgium (Wallonia) 1991 / 29.12
Belgium (Wallonia) 1999 / 38.15
Belgium (Wallonia) 2014 / 22.63
Brazil 2002 / 54.46
Brazil 2006 / 41.76
Brazil 2010 / 54.75
Brazil 2014 / 51.19

1

Appendix K. The moderating impact of willingness to turn out to vote on proximity voting in Australia

Table K.1. Alternative specific conditional logit model explaining the vote choice in the Australian 2001 elections

Party (ref. = Liberal Party) / Australian Labor Party / National Party of Australia / Australian Democrats / Australian Greens / One Nation Party
b
(s.e.) / b
(s.e.) / b
(s.e.) / b
(s.e.) / b
(s.e.)
Distance to party / -0.467***
(0.075)
Strength of partisanship / 1.654***
(0.070)
Distance to party x would not vote / 0.880***
(0.229)
Would not vote / 1.426** / 1.311 / 0.565 / 0.447 / 1.326*
(0.016) / (0.971) / (0.595) / (0.595) / (0.674)
Female / 0.016 / 0.016 / 0.807* / 0.734* / 0.361
(0.228) / (0.415) / (0.330) / (0.361) / (0.454)
Age / 0.009 / -0.017 / 0.006 / 0.002 / -0.026
(0.008) / (0.013) / (0.011) / (0.012) / (0.015)
Middle educated (ref: low) / --- / --- / --- / --- / ---
Higher educated (ref: low) / 0.127 / 0.138 / 0.490 / 0.568 / 0.634
(0.240) / (0.428) / (0.358) / (0.403) / (0.490)
Social class / -0.082 / -0.154 / 0.455 / 0.942** / -0.443
(0.065) / (0.402) / (0.315) / (0.350) / (0.459)
Religious attendance / 0.073 / 0.038 / -0.063 / 0.063 / 0.139
(0.065) / (0.110) / (0.097) / (0.100) / (0.118)
Constant / -1.471 / -0.540 / -2.962** / -3.930*** / -1.140
(0.900) / (1.006) / (0.910) / (1.024) / (1.090)
N / 6,786
AIC / 1,574.103
BIC / 1,833.362
Log Likelihood / -749.05129

Note: Logit coefficients and standard errors (in parentheses) are reported. Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.

Table K.2. Alternative specific conditional logit model explaining the vote choice in the Australian 2004 elections

Party (ref. = Liberal Party) / Australian Labor Party / National Party of Australia / Australian Democrats / Australian Greens / One Nation Party
b
(s.e.) / b
(s.e.) / b
(s.e.) / b
(s.e.) / b
(s.e.)
Distance to party / -0.406***
(0.063)
Strength of partisanship / 1.523***
(0.051)
Distance to party x would not vote / 0.508*
(0.205)
Would not vote / 0.732 / -0.739 / -6.734 / 0.420 / 0.699
(0.467) / (0.976) / (4.054) / (0.592) / (1.508)
Female / 0.326 / 0.373 / -0.065 / 0.557 / -1.759
(0.231) / (0.410) / (0.606) / (0.321) / (1.170)
Age / -0.005 / -0.023 / -0.013 / -0.001 / -0.060*
(0.007) / (0.013) / (0.018) / (0.010) / (0.027)
Middle educated (ref: low) / --- / --- / --- / --- / ---
Higher educated (ref: low) / -0.151 / -0.860* / -0.958 / 0.716 / -1.015
(0.250) / (0.426) / (0.650) / (0.394) / (0.832)
Social class / 0.111 / 0.236 / 1.128 / 0.457 / 0.489
(0.224) / (0.383) / (0.657) / (0.313) / (0.767)
Religious attendance / 0.054 / 0.123 / 0.128 / -0.239* / -0.172
(0.064) / (0.107) / (0.134) / (0.105) / (0.241)
Constant / -0.605 / -0.788 / -3.221 / -2.043* / 0.372
(0.595) / (1.023) / (1.665) / (0.889) / (1.977)
N / 6,192
AIC / 1,258.001
BIC / 1,513.78
Log Likelihood / -591.00057

Note: Logit coefficients and standard errors (in parentheses) are reported. Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.

Table K.3. Alternative specific conditional logit model explaining the vote choice in the Australian 2007 elections

Party (ref. = Liberal Party) / Australian Labor Party / National Party of Australia / Australian Greens
b
(s.e.) / b
(s.e.) / b
(s.e.)
Distance to party / -0.395***
(0.061)
Strength of partisanship / 1.510***
(0.068)
Distance to party x would not vote / 0.125
(0.216)
Would not vote / 0.041 / 0.230 / 0.756
(0.482) / (0.816) / (0.598)
Female / -0.468* / -1.247** / -0.276
(0.224) / (0.422) / (0.312)
Age / -0.004 / -0.022 / 0.002
(0.007) / (0.012) / (0.010)
Middle educated (ref: low) / --- / --- / ---
Higher educated (ref: low) / -0.121 / -0.819* / 0.854*
(0.249) / (0.410) / (0.399)
Social class / -0.134 / 0.126 / 0.154
(0.209) / (0.104) / (0.290)
Religious attendance / 0.013 / 0.126 / 0.107
(0.064) / (0.104) / (0.088)
Constant / 0.883 / 0.204 / -2.086*
(0.556) / (0.976) / (0.843)
N / 4,816
AIC / 1,213.682
BIC / 1,369.195
Log Likelihood / -582.84125

Note: Logit coefficients and standard errors (in parentheses) are reported. Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.

Table K.4. Alternative specific conditional logit model explaining the vote choice in the Australian 2010 elections

Party (ref. = Liberal Party) / Australian Labor Party / National Party of Australia / Australian Greens
b
(s.e.) / b
(s.e.) / b
(s.e.)
Distance to party / -0.387***
(0.055)
Strength of partisanship / 1.411***
(0.056)
Distance to party x would not vote / 0.329
(0.183)
Would not vote / -0.057 / -0.080 / -0.014
(0.387) / (0.664) / (0.436)
Female / 0.778*** / 0.494 / 0.335
(0.194) / (0.331) / (0.236)
Age / -0.006 / 0.003 / -0.016*
(0.006) / (0.010) / (0.007)
Middle educated (ref: low) / --- / --- / ---
Higher educated (ref: low) / 0.139 / -0.208 / -0.031
(0.211) / (0.345) / (0.264)
Social class / -0.228 / -0.094 / 0.269
(0.187) / (0.311) / (0.233)
Religious attendance / -0.135* / -0.065 / -0.102
(0.056) / (0.094) / (0.071)
Constant / 0.439 / -1.462 / 0.144
(0.523) / (0.872) / (0.636)
N / 6,188
AIC / 1,754.512
BIC / 1,916.041
Log Likelihood / -853.25598

Note: Logit coefficients and standard errors (in parentheses) are reported. Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.

Table K.5. Alternative specific conditional logit model explaining the vote choice in the Australian 2013 elections

Party (ref. = Liberal Party) / Australian Labor Party / National Party of Australia / Australian Greens
b
(s.e.) / b
(s.e.) / b
(s.e.)
Distance to party / -0.401***
(0.039)
Strength of partisanship / 1.380***
(0.042)
Distance to party x would not vote / 0.234
(0.131)
Would not vote / 0.360 / 0.214 / -0.195
(0.290) / (0.481) / (0.378)
Female / -0.057 / 0.490* / 0.290
(0.142) / (0.250) / (0.190)
Age / -0.010* / 0.016* / -0.006
(0.004) / (0.008) / (0.006)
Middle educated (ref: low) / --- / --- / ---
Higher educated (ref: low) / (0.163) / -0.145 / 0.189
0.245 / (0.271) / (0.232)
Social class / (0.138) / 0.120 / 0.552**
-0.140*** / (0.231) / (0.186)
Religious attendance / (0.041) / -0.230** / -0.191**
0.096 / (0.074) / (0.059)
Constant / -0.264 / -2.541*** / -1.393**
(0.389) / (0.684) / (0.531)
N / 10,664
AIC / 2,877.744
BIC / 3,052.335
Log Likelihood / -1,414.8722

Note: Logit coefficients and standard errors (in parentheses) are reported. Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.

Appendix L. The moderating impact of willingness to turn out to vote on proximity voting in Belgium

Table L.1. Alternative specific conditional logit model explaining the vote choice in the Belgian 1991 elections (Flanders)

Party (ref. = Christian-Democratic Party) / Green party / Liberal / Social-Democratic Party / Extreme Right Party / Flemish Nationalist Party
b
(s.e.) / b
(s.e.) / b
(s.e.) / b
(s.e.) / b
(s.e.)
Distance to party / -0.526***
(0.041)
Strength of partisanship / ---
Distance to party x would not vote / 0.296***
(0.067)
Would not vote / -0.007 / 0.420* / 0.468* / 0.685** / 0.058
(0.244) / (0.177) / (0.195) / (0.238) / (0.231)
Female / 0.218 / -0.248 / -0.228 / -0.565** / -0.253
(0.187) / (0.140) / (0.158) / (0.196) / (0.181)
Age / 0.056*** / 0.026*** / 0.008 / 0.045*** / 0.012
(0.008) / (0.005) / (0.006) / (0.007) / (0.007)
Middle educated (ref: low) / 0.330 / -0.137 / -0.002 / 0.181 / 0.153
(0.254) / (0.188) / (0.209) / (0.243) / (0.244)
Higher educated (ref: low) / 0.621* / -0.152 / -0.180 / -0.580 / 0.341
(0.287) / (0.216) / (0.261) / (0.337) / (0.276)
Social class / -0.175 / 0.376*** / -0.407*** / -0.296* / 0.035
(0.134) / (0.101) / (0.117) / (0.141) / (0.131)
Religious attendance / -0.416*** / -0.356*** / -0.433*** / -0.590*** / -0.316**
(0.088) / (0.079) / (0.081) / (0.095) / (0.098)
Constant / -2.372*** / -0.951* / 1.564** / -0.496 / -0.572
(0.589) / (0.454) / (0.473) / (0.597) / (0.568)
N / 10,254
AIC / 5,267.022
BIC / 5,570.91
Log Likelihood / -2,591.511

Note: Logit coefficients and standard errors (in parentheses) are reported. Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.

Table L.2. Alternative specific conditional logit model explaining the vote choice in the Belgian 1999 elections (Flanders)

Party (ref. = Liberal Party) / Green party / Social-Democratic Party / Extreme Right Party / Flemish Nationalist Party / Christian-Democratic Party
b
(s.e.) / b
(s.e.) / b
(s.e.) / b
(s.e.) / b
(s.e.)
Distance to party / -0.482***
(0.035)
Strength of partisanship / ---
Distance to party x would not vote / 0.339***
(0.073)
Would not vote / -0.393 / -0.419 / -0.372 / -0.284 / -0.027
(0.275) / (0.254) / (0.282) / (0.302) / (0.219)
Female / 0.717*** / 0.003 / -0.035 / 0.229 / 0.229
(0.189) / (0.193) / (0.211) / (0.213) / (0.160)
Age / -0.013 / 0.031*** / 0.022** / -0.008 / 0.018**
(0.008) / (0.008) / (0.008) / (0.009) / (0.006)
Middle educated (ref: low) / 0.581* / 0.136 / -0.098 / 0.191 / -0.265
(0.285) / (0.242) / (0.264) / (0.290) / (0.216)
Higher educated (ref: low) / 1.183*** / -0.268 / -0.136 / 0.406 / 0.098
(0.298) / (0.298) / (0.313) / (0.323) / (0.234)
Social class / -0.547*** / -0.980*** / -0.712*** / -0.619*** / -0.436***
(0.140) / (0.144) / (0.154) / (0.158) / (0.119)
Religious attendance / 0.024 / -0.222** / -0.176* / 0.158* / 0.412***
(0.068) / (0.071) / (0.075) / (0.072) / (0.052)
Constant / 0.198 / 0.547 / 0.528 / -0.101 / -1.206**
(0.503) / (0.475) / (0.496) / (0.528) / (0.396)
N / 8,682
AIC / 4,300.165
BIC / 4,597.063
Log Likelihood / -2,108.0825

Note: Logit coefficients and standard errors (in parentheses) are reported. Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.