Better Democracy NZ is a non-partisan, non-profit organisation.

Our mission is to foster the improvement of New Zealand's democratic system and encourage the use of direct democracy through the

Veto, Citizens' Initiated and Recall referendum.


Wednesday, 1 July 2009

DIRECT DEMOCRACY: New Approaches to Old Questions

Professors Arthur Lupia and John G. Matsusaka from the Center for Political Studies put forward an essay to answer old questions by opponents of binding referendums

 Are voters competent?
 What role does money play?
 How does direct democracy affect policy?
 Does direct democracy benefit the many or the few?

KeyWords initiative, referendums, voting, elections, law making

Abstract: Until recently, direct democracy scholarship was primarily descriptive
or normative. Much of it sought to highlight the processes’ shortcomings.We describe
new research that examines direct democracy from a more scientific perspective. We
organize the discussion around four “old” questions that have long been at the heart
of the direct democracy debate: Are voters competent? What role does money play?
How does direct democracy affect policy? Does direct democracy benefit the many
or the few? We find that recent breakthroughs in theory and empirical analysis paint
a comparatively positive picture of the initiative and referendum. For example, voters are more competent, and the relationship between money and power in direct
democracy is less nefarious, than many observers allege. More new studies show that
the mere presence of direct democracy induces sitting legislatures to govern more


For better or worse, the institutions of direct democracy—initiative and referendum—
have become an integral part of American democracy. At present, more than half of the states and cities in the United States provide for the initiative and
referendum, and more than 70% of the population lives in a state or city where
direct democracy is available (Matsusaka 2004). These institutions have been a
part of American government for more than 100 years now, making them older
than universal women’s suffrage and direct election of U.S. senators, and there
is reason to believe that they will become increasingly important in the future.
No state with the initiative or referendum has ever chosen to do away with the
procedure, and states without the procedures are gradually adopting them (at a
rate of about one state per decade since the end of World War II).
Direct democracy is flourishing outside the United States as well. In Western
Europe, ten countries allow initiatives (as do six of the post-Soviet states), most
countries use referendums1 from time to time, and the proposed constitution for
the European Union now includes a provision for initiative and referendum. It has
become almost the norm to use referendums to make important public decisions; 29
elections have been held on European integration alone (see, e.g., Hug & Sciarini
2000, Hug 2002).

Despite its rise, however, direct democracy remains controversial. Most citizens
like it—American and European surveys consistently reveal strong support for direct
democracy. Many journalists and political elites, on the other hand, are deeply
skeptical. Their main points of contention are almost proverbial: Are voters competent to make policy decisions? Is money too important in initiative campaigns?
Does direct democracy allow special interests to subvert the policy process, or
does it empower citizens to counteract special interest influence in the legislature?
These questions were at the heart of the debate over direct democracy during
the Progressive movement a century ago, and they remain the central questions

Scholars also address these questions. In fact, they have written hundreds of
books and thousands of articles about direct democracy. Until recently, however,
this literaturewas almost entirely descriptive or normative. Fewscholars treated the
questions as scientific propositions that could be evaluated theoretically and tested
empirically. The historical obstacles to examining direct democracy scientifically
have been the lack of formal language to filter competing arguments by their
logical consistency and a high cost of computing that impeded all but the simplest
empirical inquiries. In recent decades, these obstacles have eroded. The increasing
sophistication of econometrics and the advent of low-cost computing make it
possible to work with large data sets, isolate key effects, and establish robust
empirical relations. At the same time, the development of powerful new theoretical
tools, including formal modeling, allows more sophisticated examinations of direct
democracy’s strategic, informational, and policy-related aspects.

In this essay,we discuss the research engendered by the fortuitous confluence of
breakthroughs in theory and empirical analysis. After a brief review of definitions
and concepts in the next section, we focus on four “old” questions that have long
been at the heart of the direct democracy debate:
 Are voters competent?
 What role does money play?
 How does direct democracy affect policy?
 Does direct democracy benefit the many or the few?

We address each question in two ways. First, we highlight the methodological
foundations of the new studies in order to distinguish their approach from the
largely descriptive and normative literature that preceded them. Second, we discuss
the new literature’s substantive findings.

1We use “referendums” rather than “referenda” as the plural following the Oxford English Dictionary and the convention of the literature.


Although it is too early to draw definitive conclusions about all aspects of direct
democracy, the new approaches challenge many old answers. The new insights
paint a relatively positive picture of the initiative and referendum. They show
that voters are more competent than many critics allege and that the relationship
between money and power in direct democracy is far less nefarious than many
“prescientific” authors contend. In other words, the new approaches’ results are
more in line with the preponderance of public opinion than they are with the
negative valence of many previous writings.


Our focus here is on new approaches to old questions about direct democracy,
so we do not dwell on legal and institutional details except where they are important
to answering the questions. However, it is useful to define a few terms
and provide a little institutional context before proceeding. Direct democracy is
a broad term that encompasses a variety of decision processes, including town
meetings, recall elections, initiatives, and various forms of referendums. This essay
focuses on the two most important and widely used processes, the initiative and

The referendum is a process that allows citizens to approve or reject laws or
constitutional amendments proposed by the government. Referendums come in
several flavors. Petition referendums allow citizens to challenge laws approved by
the legislature. Legislative referendums are measures referred to the voters by the
legislature for their approval. Some legislative referendums are mandatory. Fortynine
states, for example, require that voters approve constitutional amendments.
Other legislative referendums are less binding. At the extreme are advisory referendums.

Governments use these referendums to gauge public opinion, but the
outcomes of such elections have no direct implications for law or policy.
The initiative is a process that allows ordinary citizens to propose new laws
or constitutional amendments by petition. The main difference between initiatives
and referendums, therefore, is that citizens can write the former whereas only
government officials can draft the latter. Some of the most famous recent uses
of direct democracy, such as California’s Proposition 13 of 1978, were citizensponsored initiatives that became law despite broad opposition in sitting legislatures.

Figure 1 indicates which states currently provide for the initiative and

The single most comprehensive source on the law and use of the initiative and referendum in America is the Initiative and Referendum Almanac (Waters 2003). Information on direct democracy in the states also can be found in Magleby (1984), Dubois & Feeney (1998), Matsusaka (2004), and theWeb site of the Initiative & Referendum Institute ( Information on direct democracy in cities can be found in Schmidt (1989) and Matsusaka (2003). European information is reported on the Web site of the Initiative & Referendum Institute Europe (

Figure 1 States with the initiative or referendum, 2003.
The most obvious difference between direct democracy and more traditional
legislative processes is the direct participation of voters. However, scholars have increasingly turned their attention toward other procedural differences whose consequences are potentially important.We focus on three such differences: qualification requirements, the need to run an electoral campaign to pass a law, and the unique challenges of asking legislators and governmental actors to implement laws that they did not create. Before proceeding, we describe each of these factors

To put a policy proposal on the ballot, its advocates are required to demonstrate
that it elicits at least a minimal degree of public interest. Signature requirements
are the most common means for sponsors of initiatives and petition referendums to
demonstrate such interest. If a predetermined number of signatures are collected,
the proposal goes before the voters. The required number of signatures varies
from place to place. In the United States the requirement ranges from a low of
2% of registered voters in North Dakota to a high of 15% of votes cast in the
previous general election in Wyoming. In addition to mandating numbers, some
states require signatures to be collected from multiple regions of the state—to
avoid sponsors soliciting signatures in only a handful of densely populated areas.
Massachusetts, for example, allows no more than 25% of signatures to come from
the Boston area. Another important variable is the time allowed for signature
collection. In California, sponsors must collect all of their signatures within a 150- day window. In Florida, sponsors can take up to four years. As we show below,
qualification requirements are important in the new science of direct democracy
because they affect what kinds of policy proposals get on the ballot and the role
of money in determining ballot access.

Campaigning For many groups using direct democracy, the need to wage an electoral campaign in support of a policy proposal represents an even greater challenge than qualification.

Numerous factors affect the cost of an initiative campaign. In many cases,
reaching large segments of the population requires sizable purchases of advertisement
time on television and radio. Many campaigns also engage heavily in direct
mail. Other factors that affect the cost of waging a successful direct democracy
campaign include the extent of underlying public support (i.e., whether the campaign
seeks to mobilize existing support or change people’s opinions; the latter
is often more expensive) and the resources of the opposition (i.e., whether the
campaign is competitive). When advocates run expensive campaigns, then those
without resources may be increasingly disadvantaged.

If an initiative or referendum wins on Election Day, its impact on policy depends on
howit is implemented. In some cases, implementation is easy (e.g., term-limits legislation often provides simple rules, such as a maximum of three two-year terms).
In most cases, however, new laws require more complex interpretation and more
complicated actions. Some initiatives, for example, offer vague implementation
directions and mandate mild or no sanctions on those who choose not to follow existing directions. Policies passed by direct democracy are not always implemented
as written and do not always affect policy in ways that their supporters anticipate
(Gerber et al. 2001). Therefore, understanding when initiatives and referendums
are most likely to be implemented as they were written is an important part of
understanding direct democracy’s ultimate policy impact.

With basic definitions and concepts in hand,wenowturn to the task of describing
what newapproaches to the study of direct democracy imply about its old questions.
We start with an important question about voters.


Questions about voter competence are acommonfacet of direct democracy debates.
Many people believe that ordinary citizens are incompetent because they base their
political choices on limited factual foundations. Supporting this point of view are
answers to what pollsters call “political information” questions (e.g., “Who is
the Chief Justice of the Supreme Court?”). Citizens routinely provide incorrect
answers to such questions on national surveys (Delli Carpini & Keeter 1996).
With such evidence abundant and widely publicized, it is difficult to imagine
that voters are competent to make the kinds of policy decisions with which direct
democracy confronts them. Yet, a frequent finding in new studies of direct democracy
is that they do. To say that citizens are more competent than many critics
allege is not to say that instances of voter incompetence are trivial or nonexistent.

However, techniques such as game theoretic communication models, laboratory
experiments, and specially designed surveys nowallowscholars to more accurately
distinguish malignant informational contexts from benign ones.

The key insight on which this literature builds is that people base most of their
choices, even complex and important ones, on very simple kinds of information.
Making choices in thisway leaves us vulnerable to mistakes, butwe proceed in this
fashion because we have no other option. Each of us has thousands of choices to
make every day.We cannot know detailed information about, or conduct rigorous
research on, the consequences of every choice we face. Because we cannot know
everything, our choices are—as a matter of necessity—based on the few rules of
thumb or heuristics that we adopt.

In one of the first direct empirical examinations of this phenomenon, Lupia
(1994) demonstrated that voters can use simple pieces of information as substitutes
for the detailed information that political surveys show them to lack. His findings
are the result of an exit poll whose purpose was to determine whether relatively
uninformed voters could use information shortcuts—specifically the identity of
groups who supported or opposed the measures in question—to cast the same
votes they would have cast if able to recite the ballot measures’ details. The exit
poll surveyed California voters who were confronted by five complex propositions
regarding the regulation of the insurance industry. Three of the propositions had
been placed on the ballot by the insurance industry, the other two by trial lawyers
and consumer activists, respectively. Pollsters asked respondents how they voted
on the propositions, some socioeconomic and insurance rate–related questions, and
a series of questions about the propositions. These latter questions were intended
to learn not only what respondents knew about the details of the insurance-reform
debate but also to gauge respondent beliefs about the preferences of persons or
groups who took public stands on the initiatives.

The data showed respondents sorting themselves into three categories. The first
category contained voters who knew neither the answers to the detailed questions
about the propositions nor the insurance industry’s preferences. The second category
contained “model citizens”—voters who consistently gave correct answers to detailed questions about the initiatives and who knew the insurance industry’s preferences. The third category contained respondents who could not answer questions
about the propositions’ details but, like the model citizens, knewthe insurance
industry’s preferences.

The study’s central finding is that voters in the second and third categories voted
in very similar ways, whereas voters in the first category voted quite differently.
Voters in the third category voted as if they were model citizens. Such evidence
supports the claim that voters with apparently low levels of political information
can use information shortcuts to emulate the voting behavior they would have
exhibited if they were as informed as the best-informed persons in the survey.
In a more expansive empirical analysis, spanning many states and elections,
Bowler & Donovan (1998, p. 168) reveal that although voters are not “fully
informed” about the details of most of the initiatives and referendums on which
they vote, many of these voters “appear able to figure out what they are for and
against in ways that make sense in terms of their underlying values and interests.
Failing that, others appear to use a strategy of voting no when information is lacking or when worries about general state conditions are greatest. Just as legislators do, these voters make choices purposefully, using available information.”
No scholar working in this area regards information shortcuts, such as interest
group endorsements, as a panacea for the potential problems that limited information
can cause for direct democracy. Results such as those offered above, however,
suggest that voters are more competent than commonly portrayed. The lingering
question is “When are such voters competent?”

New theoretical research (e.g., Lupia&McCubbins 1998) examines conditions
under which voters who use information shortcuts cast the same votes they would
have cast if they were better informed. These models draw from social psychological
research on persuasion (e.g., Eagly & Chaiken 1993, Petty & Cacioppo 1986)
and economic models of strategic communication (see, e.g., Farrell & Gibbons
1989, Farrell & Rabin 1996, Milgrom & Roberts 1986). This kind of research
begins with the premise that a voter may lack the information she needs to vote
competently. Although relying on others for information often requires less of the
voter’s time and effort than does conducting her own investigation of the issues at
hand, such dependence can also entail substantial peril. Not all people from whom
voters can seek advice are trustworthy or knowledgeable.
Theoretical models built from such premises (Lupia & McCubbins 1998,
Ch. 3–5) and associated laboratory experiments (Lupia & McCubbins 1998, Ch.
6–8) show that voters need not know many details of a ballot measure in order to
vote as if they did. If voters have access to reliable advisers, then they can vote
competently despite not knowing many details about the choice at hand. Indeed,
voters have an incentive to seek advice from people who are credible and to avoid
information providers who provide vague or misleading reports. But how do they
do this? Theory shows that political institutions can do some of the work for them.
If these institutions generate the forces that clarify others’ incentives in the eyes of voters (e.g., truth-in-advertising laws, perjury penalties, or incentives to be known as trustworthy, each of which can minimize the range of false statements made about a particular initiative), then voters can do a much better job of choosing good advisers. Therefore, voters can be competent while lacking information about those who provide them with information. Seen in this light, the requirements for voter competence in direct democracy can be quite minimal.

As Lupia & Johnston (2001) argue, it may also be true that polities encounter
issues for which experts are themselves not well informed. In this case, there are
two possible scenarios. In the first scenario, no one in the polity is sufficiently
well informed. In such a case, letting the voters choose policy by direct democracy
rather than leaving it to the legislature causes no loss in competence. In the second
scenario, some elites are knowledgeable, but the electorate’s experts are not. This
scenario can occur in a closed society, where the channels of communication are
centrally controlled, or in a society where effective channels of communication
do not exist. In such a situation, letting voters make policy by direct democracy
can do considerable damage. European and North American states, however, are
not closed societies. Each has access to modern forms of communication and
competitive political environments. Therefore, if someone in a direct democracy
debate has the opportunity to expose the opposing side’s weaknesses, the competitive
nature of politics provides a strong incentive to do so publicly. In such
cases, it is possible, but unlikely, that competing elites will conspire to withhold
important information from potential supporters in the electorate. In sum, if
there are people who are willing to inform voters and if voters can use environmental
factors such as institutions to better understand the motives of the people
they listen to, then voters can cast the same votes they would have cast if more

Research specifically devoted to questions of voter competence in direct democracy
is a relatively newphenomenon, but much of the theory and evidence produced
by political scientists so far reveal that common stereotypes about voter incompetence rely on shaky theoretical and empirical foundations.


Contemporary discussions of American democracy are preoccupied with the role
of money. When it comes to elected representatives, the main concern is that
campaign contributions lead to corruption and cause officials to adopt policies
favored by contributors but contrary to the public interest. Corruption is a less
significant concern with the initiative and referendum because voters only approve
or reject a specific law; they do not place an official in a position to dispense
favors to contributors. So when it comes to direct democracy, the concern is that
wealthy interests “buy” legislation directly by placing measures on the ballot
and campaigning for their approval. However, many new studies suggest that
whatever the capacity of money to influence ballot proposition elections, it does
not give narrow special interests any greater advantage than they already enjoy in
the legislature, at least with regard to fiscal policy or the social policies that have been studied.

There have been numerous attempts to measure the effect of money on ballot
proposition elections. The earliest studies (e.g., Lowenstein 1982; Magleby 1984,
Ch. 8; Owens&Wade 1986) compared passage rates for small samples of initiatives
in which one side heavily outspent the other to passage rates where spending was
about equal. The general findingwas that heavy spending against a measure tended
to lead to the measure’s defeat, whereas heavy spending in favor of a measure had
a minimal effect.

That basic finding has held up well in studies using more sophisticated techniques
(Garrett&Gerber 2001). For example, Gerber (1999) estimated regressions
for a sample of 125 propositions in eight states to measure the effect of spending
for and against a measure on the measure’s success, controlling for the type
of spending and other variables. She found that spending against a measure had
a large and statistically significant negative effect on the probability of passage,
whereas spending in favor had a small and insignificant effect.

Gerber’s empirical analysis is motivated by a model of interest-group behavior
based on optimization principles from economic theory. She observes that different
groups have different resources available to influence policy (e.g., citizen
groups may have access to more volunteer labor, whereas business groups have
more cash), and their efforts have different productivities (e.g., business groups
may find it difficult to persuade voters if their motives are more suspect than those
of certain citizen groups). As a result, campaign spending by some groups is likely
to be more effective than spending by other groups. Consistent with this idea,
Gerber finds that spending by business groups is particularly ineffective in gaining
approval for measures but is particularly potent in defeating measures. In contrast,
spending by citizen groups appears more effective in passing measures (Gerber
defines citizen groups to include unions and other noncorporate public interest

Such studies imply a role for money in direct democracy that cuts against
popular stereotypes. Consider, for example, the implications of the role that money
plays in groups’ abilities to satisfy states’ signature requirements. A fact beyond
dispute is that qualifying a measure for the ballot can be expensive. Using paid
signature gatherers to qualify an initiative in California, for example, now costs
well over $1 million. Professional signature collectors have always been a part
of the initiative process (Magleby 1984, Ch. 4), but they are becoming nearly
indispensable in many states. A popular stereotype is that the growing role of paid
signature gatherers allows wealthy interests greater advantages in buying policy.
Gerber’s research shows that the stereotype is not quite right. Business groups
regularly fail to pass the initiatives they sponsor. However, Gerber also finds that
the ability to get on the ballot helps them in other ways. It can trigger a legislative response to stave off the initiative, and even if the measure fails, it may attract enough votes to signal the existence of a large constituency that might prompt the legislature to take some action on the issue—though these outcomes are likely only if the policy in question has substantial public support. Indeed, a survey of business groups reported by Gerber (1999) reveals that their primary reason for promoting initiatives is to send a signal to the legislature, not to achieve passage of the measure.

New studies of voter decision making in direct democracy also reinforce
Lowenstein’s (1982) and Magleby’s (1984) early results. As noted above, Bowler
&Donovan (1998) showthat when voters are uncertain about the likely policy consequences of a ballot proposition, they tend to vote against it. Voters prefer to stick with policies whose consequences they have experienced, namely the policies that continue when initiatives lose, rather than risk voting for a new initiative whose consequences might be very bad. Thus, spending vast sums of money to defeat
an initiative may make voters sufficiently confused and uncertain that they vote
against it. Convincing voters that an initiative represents an improvement over the
known status quo, by contrast, requires more than money. As Lupia (1994), Gerber
(1999), and Lupia & Johnston (2001) show, it also requires the endorsements of
well-known public figures and evidence of broad grass-roots support.

There are two upshots of these considerations. First, the deep pockets of business
groups do not allow them to “buy” favorable policy, but they do provide leverage
in fighting off measures the groups oppose. Second, citizen groups do appear to
be able to use their wealth to gain approval for measures. Thus, it seems that
money matters, but in a nuanced way. The spending evidence also suggests why
the initiative process does not lead to policies contrary to the wishes of the majority (see below):Without preexisting public support, the financial resources of business groups are ineffective in changing the status quo, and the financial resources of most citizen groups are too scarce to bring about much change.


Policies and Performance

The most important question that might be asked about direct democracy is whether
it affects public policy—and, if so, how? If direct democracy does not affect policy
outcomes, then questions about factors such as voter competence and the power
of money lose much of their relevance. A tricky issue in answering this crucial
question is how to measure the effect of an initiative or referendum on policy.
The earlier descriptive and normative literatures approached the issue by studying
the content of measures that were actually approved by the voters and making a
judgment call on their importance. Such an approach relies on several assumptions
that are not generally valid, among them that (a) direct democracy affects policy
only through the passage of measures, (b) policies brought about by the initiative
would not have come about otherwise, and (c) measures approved by the voters
are actually implemented.

Game theory has been instrumental in clarifying how to measure the impact
of direct democracy on policy. The key insight is that the institutions of direct
democracy can affect policy even if they are never used. Intuitively, the threat of an initiative by an interest group may be enough to induce a change in behavior of the legislature. The legislature, for example, may partially accommodate an interest
group to stave off an initiative it considers to be even worse. If the legislature
responds to a threat by altering its policy choice, then the initiative has changed
the policy outcome even though a measure never appeared on the ballot. The
most transparent development of this idea is that of Gerber (1996), who builds on
the agenda-setter model pioneered by Romer & Rosenthal (1979). Gerber (1998)
shows that the possibility is more than theoretical by discussing real cases in which
the threat of an initiative did engender a legislative response. In short, the initiative has both a direct effect (measures approved by the voters) and an indirect effect (changes in legislative behavior), so its effect on policy cannot be measured by examining only the propositions that actually pass.

Theory implies that the mere availability of direct democracy can affect policy.
Agrowing body of empiricalwork tests this claim directly. Some scholars use panel
data for American states and cities and Swiss cantons and local governments. They
then compare policy outcomes in states with and without the initiative process. A
common approach is to regress a policy variable on a set of control variables and a
dummy variable that equals 1 for states with the initiative process. If policy differences remain after controlling for other known determinants of policy outcomes,
such as demographics and political variables, the differences are ascribed to the
availability of the initiative process.

Such studies generally find that the initiative and referendum matter across a
wide range of policies. Matsusaka (1995, 2000, 2004) and Matsusaka & McCarty
(2001) find differences in tax and spending policies in American states and cities.
Feld & Matsusaka (2004) and Schaltegger & Feld (2001) find differences in tax
and spending policies among Swiss cantons and municipalities. Gerber (1996,
1999) finds differences in parental abortion notification and capital punishment
policies, and Gerber & Hug (2002) find differences in minority rights policies
across American states. Moreover, and again consistent with the theory described
above, the studies generally find that the policy impact of initiative and referendum
processes in states diminishes as the cost of using the devices rises (e.g., as the
signature requirement increases).

Beyond observing that direct democracy matters, it is interesting to know how
it matters. For example, does it tend to bring about conservative or liberal policies?

When the initiative was first proposed by the Progressives early in the twentieth
century, it was opposed by conservatives on the grounds that it would become the
tool of socialists and other left-leaning groups: “If socialism is ever established in the United States—so far as it can be established—it will be through the initiative and referendum” (Boyle 1912, p. 29). More recently, the initiative has been associated with conservative causes, especially tax limitation. The older descriptive literature searched for ideological tendencies in direct legislation by subjectively classifying measures and then counting up the number of conservative and liberal measures that passed. The results were generally ambiguous (Cronin 1989). As discussed above, the counting-up approach is not generally valid because it ignores the indirect effects of initiatives and referendums. It also assumes that the measures are faithfully implemented—an assumption to which we return below.

The new research, which is based on theoretically justified estimates of the
full effects (direct plus indirect), provides a fairly clear picture of the ideological tendency of the initiative and referendum. For fiscal policies over the past several decades, the evidence points almost uniformly to a conservative effect of the initiative (see Matsusaka 2004 for evidence and a review of the literature). Initiative states taxed and spent about 4% less than noninitiative states (all else equal), adopted less redistributive financing schemes (relying more on user charges and fees and less on broad-based taxes), and decentralized expenditure from state to
local governments. Similarly, evidence from Swiss cantons shows that the initiative
and referendum reduce spending (Feld & Matsusaka 2004, Schaltegger & Feld 2001). For social policies, Gerber (1999) finds that initiative states are more
likely to require parental notification when minors seek abortions and are more
likely to use capital punishment. It is easy to find examples of particular initiatives that promote liberal causes, such as funding for mass transit, but the statistical evidence shows that these are isolated cases that run against a more general trend:

Over the past four decades, the initiative has tended to bring about more fiscally
and socially conservative policies at the state level than would occur otherwise.
The reverse was true earlier in the twentieth century. Matsusaka (2000) documents
that the initiative increased state spending during the period 1902–1942,
particularly spending on public education and welfare programs. The initiative also
appears to have driven up spending in cities over the past few decades (Matsusaka
2004). Thus, itwould be wrong to conclude that direct democracy is inherently conservative.

Rather than thinking of direct democracy as ideologically predisposed
in a particular direction, it is better to think of it as a “median-reverting” institution that pushes policy back toward the center of public opinion when legislatures move too far to the right or left.

Amore common question is whether the initiative and referendum lead to better
or worse policy decisions. Although many people offer opinions on this matter,
their conclusions usually tell us more about their own policy preferences than about
the desirability of the resulting policies as seen from broader or different normative perspectives. Of course, it is difficult to come up with an objective definition of a good or bad policy, and the new research, with its positive focus, has tended to shy away from normative assessments of the institutions of direct democracy. Yet some interesting steps in this direction have been taken.
One criticism of direct democracy is that voters are myopic: They will approve
new spending programs while at the same time cutting their taxes. The evidence
generally rejects the notion that voters do not understand that budgets must balance
over time. Although initiatives tend to bring about both lower taxes and
lower spending, neither initiatives nor referendums have a significant effect on the
amount of debt issued (Matsusaka 1995, 2004; Kiewiet & Szakaly 1996). At least
in this respect, the initiative process does not lead to prima facie irrational public policies—particularly when you compare such results to the deficit spending
patterns of many professional legislatures.

A scattering of studies has attempted to measure the quality of public decisions
using economic indicia. These studies face the familiar problems associated with
nonexperimental data, and the techniques to resolve those problems are still being
developed, so the findings should be viewed as preliminary. An early study by
Pommerehne (1983) found that municipal services such as trash collection are
more efficient (output for a given level of expenditure) in cities with direct democracy than in cities without direct democracy. Feld & Savioz (1997) estimate an
aggregate production function for the economies of Swiss cantons and find that
those with direct democracy enjoy a higher total output for a given level of labor
and capital inputs than those without. One explanation could be that direct
democracy cantons are using their public funds more productively than cantons
without direct democracy. Blomberg et al. (2001) also adopt a macroeconomic
approach and find that American states with the initiative experience faster growth
in output per capita than those without. They also suggest that public spending
is less wasteful in states with the initiative than in those without. Frey & Stutzer
(2000) approach a similar question in a different way. They study survey data in
which people indicate their “happiness” (subjective well being) and find happier
people in cantons with more direct democracy, after controlling for income and
other demographics.

Other potential benefits of direct democracy have been highlighted recently.
Mendelsohn&Cutler (2000), in a study of a Canadian referendum, find that voters
pay more attention to the media and learn more as campaigns progress, even among
the most poorly informed segments of the electorate. Tolbert and her coauthors
(2003, p. 23) use 1996, 1998, and 2000 American National Election Studies to show
that “exposure to ballot initiatives increases the probability of voting, stimulates
campaign contributions to interest groups, and enhances political knowledge.”
Again, each of these studies should be viewed with some caution because they use
data from specific contexts, but they do lay down a few tracks toward measuring
the quality of public policies under direct democracy. It seems worth noting that
every such study to date points to higher-quality (or no worse) decisions when the
institutions of direct democracy are available.

The Implementation Problem

The literature documenting that direct democracy changes policy is complemented
by another strand of research that investigates the channels though which the policy
changes occur. New research integrates delegation models, econometric analyses,
and case studies to clarify what happens in direct democracy’s implementation
stage. It shows that implementation is a particularly thorny issue for initiatives.
The focal empirical result is that great variation exists in how legislators, bureaucrats, and other government employees react to winning initiatives. Some
measures, once passed, take full effect, whereas others are reinterpreted or ignored
(Gerber et al. 2001). These variations occur because the people who create and
support winning initiatives are not authorized to implement and enforce them.
Instead, they must delegate these tasks to legislatures and bureaucrats.
New research (see, e.g., Gerber et al. 2001, Bali 2003, Gerber et al. 2004) addresses
the question, “When will these governmental actors use their delegated powers to weaken the impact of laws they dislike?” It reveals that initiatives face more difficult implementation problems than many policies passed by professional
legislatures. The initiative process is a very expensive way to seek policy change;
it is cost effective only for people who seek policy changes that the government refuses to provide (Gerber 1999). It is not surprising, therefore, that many initiatives propose policy changes that reflect dramatic departures from the status quo (e.g., overhauling a state’s property tax system), that legislators are unwilling to impose on themselves (e.g., term limits and certain types of campaign finance reform), that the major parties dislike (e.g., open primaries), that cut across existing political cleavages (e.g., bilingual education), that offend important legislative constituencies (e.g., tort reform), or that are considered too hot to handle (e.g., immigration policy and gay marriage).

The challenge for supporters of such initiatives, a challenge that many laws
passed by legislative majorities do not face, is that the same governmental actors
who once blocked the policies from proceeding through traditional legislative
channels may be in a position to influence, or even determine, the extent of
their postelection implementation and enforcement. In other words, laws passed
by voters against the wishes of legislative majorities or governors face powerful
postpassage opposition that laws passed by these government entities do not.
Moreover, when a professional legislature passes a law, rarely is its next move to
disband. Although some legislatures pass laws just before a newbatch of legislators
is sworn in, the legislative body itself continues. Therefore, it is in a relatively good position to oversee those charged with implementing its edicts. Organizations that pass initiatives, by contrast, often disband soon after the election. They cease
raising money, they cease recruiting supporters, they shut down their offices—
they cease to exist. Compared with professional legislatures, such entities are in a
relatively bad position to oversee those charged with implementing their edicts.
Agame theoretic model that builds in these postelection features of the initiative
process shows that the policy changes that are most likely to prevail as initiatives
(as opposed to prevailing in a standing legislature) are less likely to be implemented and enforced, all else staying constant (Gerber et al. 2004). This result is important because many public and scholarly observers of the initiative process believe that a victory at the polls implies a direct policy change. The new research turns this common wisdom on its head. Much of what is important to understanding the initiative process’s policy consequences occurs after Election Day.



The Progressives promoted the initiative and referendum as ways to counteract
special interest influence in state legislatures. One of the most prominent criticisms of these processes, then and now, is that they actually have the reverse affect: They allow wealthy, organized interests to subvert the policy process (e.g., Broder 2000). The evidence in support of the subversion criticism is largely circumstantial.

Special interests are active in qualifying measures for the ballot and in campaigning
for their approval. It is expensive to collect the signatures required to place a
measure on the ballot (more than $1 million in California), making it difficult for
groups without access to significant resources to make proposals.

However, the circumstantial argument rests on assumptions that are not generally
consistent with evolving theory. For one thing, the dominance of the proposal
process by narrow groups does not imply the process works to the disadvantage
of a majority of citizens. In the agenda-setting models of Gerber (1996) and
Matsusaka & McCarty (2001), only policies favored by the majority will be approved
if they come to the ballot. If voters are sufficiently well informed, they will
reject any policies that they find worse than the status quo. In such cases, voters are always better (or noworse) off having an initiative or referendum available, even in the extreme case where proposals are entirely monopolized by special interests. Intuitively, the voters are better off if the interest group offers a proposal that the voters prefer to the status quo provided by the legislature, and they are no worse off if the interest group offers an inferior policy because they will simply reject it. Put differently, a majority of voters can gain when they have the opportunity to choose between two policy options instead of having to accept the decision of the legislature.

The conclusion that the majority is always better off having the initiative and referendum available is a fairly general property of complete information models but
does not necessarily hold with incomplete information. For example, Matsusaka&
McCarty (2001) show that when the legislature is uncertain about the preferences
of the majority of voters, the voters might be worse off having the initiative available because the legislature will cater to special interests in an effort to avoid the risk of an initiative campaign. Similarly, Gerber & Lupia (1995) show that when voters are uncertain about the policy implications of a referendum and are also
unwilling or unable to seek the advice of credible experts, then direct democracy
can produce policy outcomes worse than the status quo. In sum, when voters or
legislators have information sufficient to make simple binary comparisons such as
“better” and “worse,” then direct democracy can be an effective way to improve
policy—and the studies cited in our earlier discussion of voter competence suggest
that voters have sufficient information more often than critics allege.

Another problem with the circumstantial argument is that it neglects the influence
of special interests in the legislature. It is probably fair to say that wealthy,
organized interests are more influential than disorganized interests both in the
legislature and in initiative campaigns. The question, then, is whether special interests are relatively more influential in the legislative or in the initiative process.

It could be, for example, that the direct democracy playing field is tilted in favor
of special interests but less so than the playing field of more traditional legislative processes.

Neither the circumstantial evidence nor theory provides a clear answer as to
whether special interests gain from the initiative process. The only way to find the
answer is by examining the evidence of how direct democracy has functioned in
practice. The most extensive attempt to answer the question is that of Matsusaka
(2004), who studies data spanning the entire twentieth century at both the state
and local level and asks whether the initiative has brought about policies favored
by special interests or by the majority. (A special interest is defined as an interest contrary to the majority’s interest.) The core evidence in Matsusaka’s book concerns fiscal policy of state and local governments over the past several decades.
The research strategy is first to document how the initiative changed fiscal policy
and then to compare those changes with public opinion surveys to see if a majority
of people opposed the changes, as the special interest subversion view implies, or
favored the changes. Over the past three decades, three fiscal effects of the initiative are documented (Ch. 3): (a) The initiative cut spending and taxes, (b) the initiative decentralized spending from state to local governments, and (c) the initiative shifted revenue out of broad-based taxes and into user fees and charges for services.

Matsusaka (Ch. 4) examines a variety of opinion data in which people expressed
their preferences about spending, taxes, the division of spending between state and
local governments, and revenue mix. The opinion evidence consistently shows that
a majority of people favored the changes brought about by the initiative. In no case
was there a majority opposed to the changes. The idea that the initiative allows
special interests to subvert the policy process to the detriment of the public is not
supported by any of the evidence.

Gerber’s (1999) finding on parental abortion notification and death penalty
policies points in the same direction. Although she examines only a single year,
and her analysis does not focus on this point directly, it can be shown that her
estimates imply that policies in initiative states were more likely to be consistent
with majority opinion than policies in noninitiative states (see Matsusaka 2004,
Appendix 4).

Lascher et al. (1996) and Camobreco (1998) also adopt a scientific approach
to the problem. Both studies attempt to compare the responsiveness of policy
to public opinion in initiative and noninitiative states across a number of policy
dimensions. They find little effect. However, their results are generally statistically insignificant, preventing strong inferences from being drawn, and in any case rest on a problematic methodology (Matsusaka 2001).

To summarize, it appears that direct democracy has not led to special interest
subversion when it comes to fiscal policies and certain social policies over the past
several decades. There may be other policy dimensions not yet studied in which
the special interest subversion view is correct. However, the fact that none of the
existing evidence supports that view argues for its rejection until confirmatory
evidence appears.

The picture that emerges is of the initiative as a majoritarian device. As discussed
above, the initiative has pushed policy in a conservative direction over the past several decades. If we put the facts together, the implication is that, over the past several decades, legislatures have tended to deliver policy to the left of majority opinion, and the initiative has provided the majority with a way to correct what they see as the mistakes of their representatives. Matsusaka (2004, Ch. 7) reports additional evidence that supports this view. It turns out that the antitax effect of the initiative was especially strong in the 1970s and 1990s, two periods during which public opinion was drifting in a conservative direction (e.g., Stimson 1999), but the initiative did not affect taxes in the 1980s, a period when public opinion was shifting in a liberal direction. One interpretation is that direct democracy causes policy to react more swiftly to changes in public opinion than does policy making by legislatures.

The view of the initiative as a way to correct nonmajoritarian policies of the legislature can also explain why the initiative drove up spending in the early twentieth century. In the decades surrounding the turn of the nineteenth century, the economy underwent a dramatic shift from a rural to an urban economy. Urban voters
wanted increased government spending on education, roads, sewers, welfare, oldage
pensions, and so on.However, rural interests controlled the legislatures because
district lines tended to overweight the countryside (this was before the one-person
one-vote principle was established). In the face of recalcitrant legislatures, the
initiative allowed the new urban majority to directly approve its new spending


This essay describes a new approach to research on direct democracy that has
emerged in recent years in the wake of breakthroughs in theory and empirical
analysis. We have not attempted to be comprehensive. Indeed, to highlight what
the new science of direct democracy can do, we have also chosen to highlight
findings that are in stark contrast to the largely negative conclusions of writers in
previous decades. Much of the newresearch paints a comparatively positive picture
of the initiative and referendum. For example, voters are more competent and the
relationship between money and power is less nefarious than many observers
allege. Other evidence suggests that the mere presence of direct democracy makes
policy more responsive to public opinion.

The same research also demonstrates, however, that direct democracy is no
panacea. If voters cannot find effective information shortcuts, then direct democracy
may end up producing outcomes that they later regret. Despite substantial
evidence that the mere presence of direct democracy induces more medianoriented
policy outcomes (whether through direct means or the indirect means
identified by scholars such as Gerber and Matsusaka), median outcomes may
not always be a good thing. Indeed, much of the Constitutional apparatus of
American government is intended to prohibit median outcomes when the majority
threatens the rights of the minority. In addition, the historical record reminds
us that the process can also be abused if run in conjunction with intimidation
or manipulation at the ballot box (see, e.g., Lupia & Johnston 2001 for an examination of the litany of abuses associated with referendums in the Napoleonic

To limit such abuses and to increase the likelihood that direct democracy serves
the public well, it is essential to understand howitworks. Scientific approaches can
help. Empirical research can help weed out theories that sound reasonable but are
contradicted by the data, while strengthening theories that survive empirical evaluations.

Theoretical research can help distill seemingly persuasive but logically
inconsistent conclusions from claims that truly follow from concrete observations.
Collectively, such endeavors can help societies more accurately understand
and more effectively adapt to the challenges posed by direct democracy’s “old

The Annual Review of Political Science is online at
Bali V. 2003. Implementing popular initiatives:
What matters for compliance? J. Polit. 65:
Blomberg SB, Hess GD, Weerapana A. 2001.
The impact of voter initiatives on economic
activity.Wellesley CollegeWork Pap. 2001–
Bowler S, Donovan T. 1998. Demanding
Choices: Opinion, Voting, and Direct
Democracy. Ann Arbor: Univ. Mich. Press.
198 pp.
Boyle J. 1912. The Initiative and Referendum:
Its Folly, Fallacies, and Failure. Columbus,
OH: Smythe
Broder DS. 2000. Democracy Derailed: Initiative
Campaigns and the Power of Money.
New York: Harcourt. 300 pp.
Camobreco JF. 1998. Preferences, fiscal policies,
and the initiative process. J. Polit.
Cronin TE. 1989. Direct Democracy: The Politics
of Initiative, Referendum, and Recall.
Cambridge, MA: Harvard Univ. Press. 289
Delli Carpini MX, Keeter S. 1996. What Americans
Know About Politics and Why It Matters.
New Haven, CT: Yale Univ. Press. 397
Dubois PL, Feeney F. 1998. Lawmaking by Initiative:
Issues, Options and Comparisons.
New York: Agathon. 268 pp.
Eagly AH, Chaiken S. 1993. The Psychology
of Attitudes. FortWorth, TX: Harcourt Brace
Jovanovich. 800 pp.
Farrell J, Gibbons R. 1989. Cheap talk with two
audiences. Am. Econ. Rev. 79:1214–23
Farrell J, Rabin M. 1996. Cheap talk. J. Econ.
Perspect. 10:103–18
Feld LP, Matsusaka JG. 2004. Budget referendums
and government spending: evidence
from Swiss cantons. J. Public Econ.
Feld LP, Savioz MR. 1997. Direct democracy
matters for economic performance:
an empirical investigation. Kyklos 50:507–
Frey BS, Stutzer A. 2000. Happiness, economy
and institutions. Econ. J. 110:918–
Garrett E, Gerber ER. 2001. Money in the initiative
and referendum process: evidence of
its effects and prospects for reform. In The
Battle over Citizen Lawmaking, ed. MDWaters,
pp. 73–96. Durham, NC: Carolina Acad.
294 pp.
Gerber ER. 1996. Legislative response to the
threat of popular initiatives. Am. J. Polit. Sci.
Gerber ER. 1998. Pressuring legislatures
through the use of the initiatives: two forms
of indirect influence. In Citizens as Legislators:
Direct Democracy in the United
States, ed. S Bowler, T Donovan, CJ Tolbert,
pp. 191–208. Columbus: Ohio State Univ.
Press. 316 pp.
Gerber ER. 1999. The Populist Paradox: Interest
Group Influence and the Promise of Direct
Legislation. Princeton, NJ: Princeton Univ.
Press. 171 pp.
Gerber ER, Hug S. 2002. Minority rights and
direct legislation: theory, methods, and evidence.
Work. pap., Dep. Polit. Sci., Univ.
Gerber ER, Lupia A. 1995. Campaign competition
and policy responsiveness in direct
legislation elections. Polit. Behav. 17:287–
Gerber ER, Lupia A, McCubbins MD. 2004.
When does government limit the impact
of voter initiatives? The politics of implementation
and enforcement. J. Polit. 66:43–
Gerber ER, Lupia A, McCubbins MD, Kiewiet
DR. 2001. Stealing the Initiative: How State
Government Responds to Direct Democracy.
Upper Saddle River, NJ: Prentice Hall. 120
Hug S. 2002. Voices of Europe: Citizens, Referendums,
and European Integration. Lanham,
MD: Rowan & Littlefield
Hug S, Sciarini P. 2000. Referendums on European
integration: Do institutions matter in
the voter’s decision? Comp.Polit. Stud. 33:3–
Kiewiet DR, Szakaly K. 1996. Constitutional
limitations on borrowing: an analysis of state
bonded indebtedness. J. Law Econ. Org.
Lascher EL Jr, Hagen MG, Rochlin SA. 1996.
Gun behind the door? Ballot initiatives, state
policies, and public opinion. J. Polit. 58:760–
Lowenstein DH. 1982. Campaign spending and
ballot propositions: recent experience, public
choice theory and the First Amendment.
UCLA Law Rev. 29:505–641
Lupia A. 1994. Shortcuts versus encyclopedias:
information and voting behavior in California
insurance reform elections. Am. Polit. Sci.
Rev. 88:63–76
Lupia A, Johnston R. 2001. Are voters to
blame? Voter competence and elite maneuvers
in public referendums. In Referendum
Democracy: Citizens, Elites, and Deliberation
in Referendum Campaigns, ed.
M Mendelsohn, A Parkin, pp. 191–210.
Toronto: MacMillan/St. Martin’s
Lupia A, McCubbins MD. 1998. The Democratic
Dilemma: Can Citizens Learn What
They Need to Know? New York: Cambridge
Univ. Press. 282 pp.
Magleby DB. 1984. Direct Legislation: Voting
on Ballot Propositions in the United States.
Baltimore, MD: Johns Hopkins Univ. Press.
270 pp.
Matsusaka JG. 1995. Fiscal effects of the voter
initiative: evidence from the last 30 years. J.
Polit. Econ. 103:587–623
Matsusaka JG. 2000. Fiscal effects of the voter
initiative in the first half of the 20th century.
J. Law Econ. 43:619–48
Matsusaka JG. 2001. Problems with a methodology
used to test the responsiveness of policy
to public opinion in initiative states. J.
Polit. 63:1250–56
Matsusaka JG. 2003. I&R in American cities:
basic patterns. See Waters 2003, pp. 51–58
Matsusaka JG. 2004. For the Many or the
Few: The Initiative Process, Public Policy,
and American Democracy. Chicago: Univ.
Chicago Press. In press
Matsusaka JG, McCartyNM.2001. Political resource
allocation: benefits and costs of voter
initiatives. J. Law Econ. Organ. 17:413–48
Mendelsohn M, Cutler F. 2000. The effect
of referendums on democratic citizens: information,
politicization, efficacy, and tolerance.
Br. J. Polit. Sci. 30:669–98
Milgrom P, Roberts J. 1986. Relying on the information
of interested parties. Rand J. Econ.
Owens JR, Wade LL. 1986. Campaign spending
on California ballot propositions, 1924–
1984: trends and voting effects. West. Polit.
Q. 39:675–89
Petty RE, Cacioppo JT. 1986. Communication
and Persuasion: Central and Peripheral
Routes to Attitude Change. New York:
Springer-Verlag. 262 pp.
Pommerehne WW. 1983. Private versus
oeffentliche muellabfuhr—Nochmals betrachtet.
Finanzarchiv 41:466–75
Romer T, Rosenthal H. 1979. Bureaucrats versus
voters: on the political economy of resource
allocation by direct democracy. Q. J.
Econ. 93:563–87
Schaltegger C, Feld LP. 2001. On government
centralization and budget referendums. CES
Work. Pap. No. 615, Cent. Econ. Stud., Munich,
Schmidt DD. 1989. Citizen Lawmakers: The
Ballot Initiative Revolution. Philadelphia:
Temple Univ. Press. 352 pp.
Stimson JA. 1999. Public Opinion in America:
Moods, Cycles, and Swings. Boulder, CO:
Westview. 208 pp. 2nd ed.
Tolbert CJ, McNeal RS, Smith DA. 2003.
Enhancing civic engagement: the effect of
direct democracy on political participation
and knowledge. State Polit. Policy Q. 3:23–
Waters MD. 2003. Initiative and Referendum
Almanac. Durham, NC: Carolina Acad. 654
A NEW GENERATION, Virginia Sapiro 1
LATINO POLITICS, Rodolfo O. de la Garza 91
AGENT-STRUCTURE DEBATE, Kate O’Neill, J¨org Balsiger,
and Stacy D. VanDeveer 149
and Malcolm E. Jewell 177
IN THE UNITED STATES, Kenneth Goldstein and Travis N. Ridout 205
G. Bingham Powell, Jr. 273
Richard L. Hasen 297
Michael X. Delli Carpini, Fay Lomax Cook, and Lawrence R. Jacobs 315
Leslie Vinjamuri and Jack Snyder 345
Howard Gillman 363
Vincent L. Hutchings and Nicholas A. Valentino 383
and Nolan McCarty 409
Arthur Lupia and John G. Matsusaka 463
Subject Index 507
Cumulative Index of Contributing Authors, Volumes 1–7 523
Cumulative Index of Chapter Titles,

No comments: