in representative democracies. Theoretically, and on the basis of piecemeal evidence, it is fully open
whether interest group influence is stronger or less intense in direct than in representative democra-
cies. Given this uncertainty, it is worth noting that there is systematic evidence, for the U.S. and Swit-
zerland, that political outcomes are closer to citizens’ preferences under direct than under represen-
tative democracy. Using aggregate data on Swiss cities in 1970, Pommerehne (1978) shows that the
median voter model performs better in jurisdictions with referenda and initiatives. Gerber (1996,
1999) provides evidence for the U.S. strongly supporting Pommerehne’s result. For two political is-
sues, parental consent laws and capital punishment, she shows that initiatives correct policy out-
comes towards the preferences of the median voter. Gerber (1996a) illustrates that the deviation of
policy outcomes from citizens’ preferences can be traced back to the influence of interest groups on
representatives. Moreover, Feld and Schaltegger (2002) find evidence for the Swiss cantons that the
influence of the state administration on federal matching grants is reduced by a fiscal referendum. This
evidence supplements the control argument emphasised in favour of elements of direct democracy
for the EU constitution. There is no reason to assume that referenda and initiatives should not simi-
larly induce outcomes of EU policy to be closer to the preferences of EU citizens.
2.3 Expressive Voting, Direct and Representative Democracy
This may indeed be the problem because the information problem in democracies is a fundamental
one. It is not simply captured by asymmetric information between representatives and voters. Be-
cause an individual citizen faces a low probability to influence voting outcomes, his/her instrumental
benefit of participating in elections or referenda is much lower than the expected costs. From a ra-
tional choice perspective, voter turnout should therefore be low. The expected costs of participating
in democratic decisions are the higher the better informed citizens have to be in order to obtain re-
flected political decisions. These information costs can be expected to be lower in representative
than in direct democracy although incentives for supply and demand of information are higher in di-
rect than in representative democracy (Kirchgassner, Feld and Savioz 1999, Feld and Kirchgassner
2000).
Despite the lack of incentives to show up at the ballots, voter turnout is considerably high in direct
and representative democracies. In Britain, voter turnout amounted to an average of 74.1 percent in
the 7 elections between 1970 and 1995 without any decreasing trend. While it was still 70.9 percent
in 1997, it fell strongly to 59.4 percent in 2001. In Germany, it was 77.8 percent in 1990, 79.0 in
1994, 82.2 percent in 1998 and 80.6 percent in 2002.9 In Switzerland, turnout in referenda and ini-
tiatives decreased since the 1950’s when still 51 percent of citizens participated. Since the 1970’s, it
appears to have stabilised at around 40 percent on average. Elections of representatives at the Swiss
federal level were at a 46.3 percent turnout on average in the last 20 years.10 In the U.S., turnout
9. In the seventies voter turnout was even higher than 90 percent in Germany, in the eighties, it was higher than
80 percent. See: Statistisches Jahrbuch 1997 für die Bundesrepublik Deutschland, pp. 90. For the British
data see: Annual Abstract of Statistics, 1997, p. 75.
10. The exact figures are: 1979, 48.0, 1983: 48.9, 1987: 46.5, 1991: 46.0, 1995: 42.2. Source of the data: Schweizeri-
sches Bundesamt für Statistik and W. Seitz, Nationalratswahlen 1995: Übersicht und Analyse, Bundesamt
für Statistik, Bern 1997, p. 115.