While the referendum works as a veto of citizens against proposals from government and parlia-
ment,6 the initiative allows to propose new policies that are neglected by the political establishment,
perhaps because they are not in their interest. In addition to referenda, initiatives enable citizens to an
unbundling of policy packages created in the parliament via log-rolling agreements such that political
outcomes that are against the interests of a majority of citizens are declined (Besley and Coate
2000). However, initiatives entail additional costs for citizens since they have to collect signatures to
bring an initiative to the ballots. Depending on the signature requirement, the initiative will thus have a
stronger or weaker impact on policy outcomes, the more or less easily issue unbundling may take
place. Referenda and initiatives nevertheless induce a correction of political outcomes in favour of the
preferences of the median voter, even in the case of strategic manipulation or agenda setting.7 As in-
stitutions of high control potential, referenda and initiatives are the more important, the further away
representatives’ decisions are from citizens. While informal control instruments may exist at the local
level, they have to be replaced by formal institutions of control when political competencies are
shifted to higher levels of government. At the EU level, elements of direct democracy can induce their
most beneficial impact to force representatives to follow citizens’ preferences. Aside the necessity of
direct democracy to control representatives at the constitutional level, the first justification for pro-
posing referenda and initiatives at the post-constitutional level in the EU thus follows again from the
control argument.
This also holds despite the existence of asymmetric information in politics where representatives are
better informed than citizens. Extending a model by Aghion and Tirole (1997), Marino and Ma-
tsusaka (2000) study budget procedures used in private and public decision-making. Starting from a
situation where representatives have a bias towards higher spending, they analyse a decision-making
procedure with full delegation in which representatives have full discretion about spending decisions
and another one with partial delegation such that voters can veto or override a spending proposal ex
post by a referendum. They show that in the partial delegation case, representatives have an incentive
to supply biased information to citizens in order to obtain their approval for higher spending levels.
Moreover, the authors show that these information biases can be severe such that full delegation
dominates partial delegation. Since it turns out to be more difficult to distort information in large than
in small projects, the optimal decision-making rule is one in which there is full delegation below a
spending threshold such that routine projects are exclusively decided by representatives, and partial
delegation above the spending threshold such that larger projects can be vetoed by voters.8
Grillo (1997) argues against the working of partial delegation in a polity of a bigger scale and of un-
certain federal/confederate nature such as the EU. Decision-making and information costs are the
more important, the larger a polity such that direct democracy is supposedly more feasible at the lo-
cal or regional than at the national or supra-national level. The costs of conducting referenda and ini-
6. Citizens can use a (binding) referendum to reject government statutes or constitutional amendments. These
can be mandatory or optional. An optional referendum takes place if a certain number of citizens asks for it.
7. For the first rigorous analysis of referenda see Romer and Rosenthal (1979), for a comprehensive analysis un-
der perfect information see Steunenberg (1992). Easily accessible analyses are presented in Mueller (1996, p.
183), Feld and Kirchgassner (2001), Matsusaka (2002) and Feld and Matsusaka (2003).
8. See Kessler (2003) for such a summary of the Marino and Matsusaka model.