Information about http://www.rochester.edu/college/psc/primo/primomilyoturnout.pdf

The Effects of Campaign Finance Laws on Turnout,…

Tags: author assistant, buckley v valeo, campaign finance laws, campaign finance reforms, campaign spending, columbia mo, david m primo, department of economics, gubernatorial elections, harkness hall, higher turnout, jeffrey milyo, matt stiffler, organizational contributions, political science university, public affairs university, state campaign finance, truman school, university of rochester, valeo decision,
Pages: 28
Language: english
Created: Tue Jan 31 09:28:46 2006

Display cached document

Page 1

Page 2

Page 3

Page 4

Page 5

Page 6

Page 7

Page 8

Page 9

Page 10

Page 11

Page 12

Page 13

Page 14

Page 15

Page 16

Page 17

Page 18

Page 19

Page 20

Page 21

Page 22

Page 23

Page 24

Page 25

Page 26

Page 27

Page 28

The Effects of Campaign Finance Laws on Turnout, 1950-2000*

David M. Primo
University of Rochester

Jeffrey Milyo
University of Missouri

February 2006

*
We thank Matt Jacobsmeier and Matt Stiffler for research assistance.

Corresponding Author. Assistant Professor, Department of Political Science, University of Rochester, Harkness
Hall 333, Rochester, NY 14627-0146; 585.273.4779; david.primo@rochester.edu; Fax: 585.271.1616

Department of Economics and Truman School of Public Affairs, University of Missouri, 118 Professional
Building, Columbia, MO 65211; 573.882.5572; milyoj@missouri.edu; Fax: 573.882.2697
Abstract

Scholars have proposed many routes by which campaign finance laws may impact turnout. For
instance, laws restricting campaign spending may decrease mobilization, resulting in lower
turnout. Alternatively, such laws might increase the competitiveness of elections, resulting in
higher turnout. Existing studies tend to focus on only one causal pathway, ignoring the net
effects of campaign finance reforms on voter turnout. We exploit the variation in state campaign
finance laws from 1950 to 2000 in order to estimate the reduced-form relationships between
reform and turnout. Using both aggregate and individual-level data, we find that campaign
finance laws on net have little impact on turnout in gubernatorial elections. There are two
exceptions to this finding: Limits on organizational contributions are shown in an individual-
level analysis to increase turnout prior to a sea change in campaign finance ushered in by the
Buckley v. Valeo decision in 1976, while public financing laws are shown to have an equally
large negative impact on turnout in the post-Buckley era. These results strengthens the existing
literature, which finds similarly perverse effects of public financing on the "quality of
democracy," and demonstrates the advantages of reduced-form analysis for understanding the
influence of laws on behavior.
Scholars and practitioners often acknowledge that stricter campaign finance laws impose

an unfortunate tradeoff between the ideals of free speech and association on the one hand and

equality, broadly defined, on the other. Still, campaign finance reform is considered by many to

be a panacea for perceived democratic ills such as low voter turnout, diminished trust in

government, and declining campaign competitiveness. Many political and legal decision-makers

have taken at face value that campaign finance reform will have the intended consequences, and

have made decisions or formed opinions accordingly. For example, in the recent Supreme Court

ruling, McConnell v. FEC, 540 U.S. 93 (2003), the majority asserted that "contribution limits,

like other measures aimed at protecting the integrity of the process, tangibly benefit public

participation in political debate" (540 U.S., at 137). Similar arguments are found in the majority

opinions of two previous landmark Supreme Court cases, Buckley v. Valeo, 424 U.S. 1 (1976),

and Nixon v. Shrink Missouri Government PAC, 528 U.S. 377 (2000).1

Arguably the most basic form of participation in the political debate is voting, and there

is no shortage of claims that turnout would be enhanced by reforming the political system. Often

these claims are made with reference to a single aspect of the political system. For instance,

stricter campaign finance laws will enhance competitiveness, which in turn will increase turnout

(e.g., Teixeira 1992). Alternatively, stricter campaign finance laws that limit negative

advertising would in turn mitigate the demobilizing impact of such ads (e.g., Ansolabehere et al.

1994). The issue with these claims, however, is that they consider one effect of campaign

finance laws at a time. In this paper, we argue that there are in fact several possible effects of

1
In Buckley, the Court argued that public financing for presidential races "is a congressional effort, not to abridge,
restrict, or censor speech, but rather to use public money to facilitate public discussion and participation in the
electoral process, goals vital to a self-governing people" (424 U.S., at 92-93). In Shrink, the majority wrote, "Leave
the perception of impropriety unanswered, and the cynical assumption that large donors call the tune could
jeopardize the willingness of voters to take part in democratic governance" (528 U.S., at 390).
campaign finance laws on turnout, and we argue that an important, but thus far unanswered,

question is what the net effects of campaign finance laws are. Ideally, we would like to

understand the relative impacts of factors like competition, but this approach is perilous. We

argue that our approach allows for a more complete perspective of the impact of the laws than

simply considering the effects piecemeal.

State-level laws vary cross-sectionally and over time, unlike federal campaign finance

laws, which have been relatively stable over the past three decades. This makes the U.S. states

an ideal arena for exploring the impact of campaign finance laws on turnout. We utilize both

aggregate-level data from 1950 to 2000 and individual-level data from 1952 to 2000. The

aggregate analysis suggests that reforms such as public financing may have a large and positive

effect on turnout; however, these estimates are not robust to the time period examined. Further,

because individual level characteristics such as age, education, party affiliation and race play an

important role in determining individual turnout decisions, our analysis of aggregate data may

not appropriately control for the differing composition of state populations. For this reason, we

estimate the contextual effects of state campaign finance reforms on individual voter turnout. In

this analysis, which avoids the ecological inference problem, we find that in the pre-Buckley era,

only limits on contributions made by organizations (i.e., corporations, unions or PACs) have a

positive effect on turnout. In the post-Buckley era, however, campaign finance laws have no

positive impact on turnout, and in fact public financing has a large negative impact on voter

turnout. Combined with the findings of Primo and Milyo (2006), who find little impact of state

campaign finance laws on perceptions of government, this paper strengthens the evidence that

campaign finance laws are unlikely to be an effective means for improving the "quality of

2
democracy." It also suggests the value of reduced form analyses for understanding the impact of

political reforms on behavior.

Theory

The health of democracy is often judged by the willingness of citizens to participate in

elections, as indicated by the quotations above and widespread public concern about low voter

turnout in national elections. Institutional reforms are widely thought to be a potential policy

lever for improving turnout. There is a large literature on changes in registration and post-

registration laws (e.g., Wolfinger et al. 2004, Highton 2004, Wolfinger and Rosenstone 1980),

which are best thought of as direct effects. Campaign finance laws, to the contrary, are likely to

have indirect effects on turnout via some other characteristic, whether at the elite or mass level.

When reformers, think tanks, and elected officials discuss the benefits (or drawbacks) of

campaign finance reform for turnout, this is what they often mean. A sampling of such claims is

suggestive.

� Rep. Dennis Moore (D-KS) wrote in an op-ed intended for local constituents,
"current campaign finance law alienates voters," leading to apathy, and in turn,
lower turnout (Moore 1999).

� Sen. Herb Kohl (D-WI) made a clear causal link between reform and
participation: "Whether the presence of unlimited political contributions is
corrupting or whether it just creates the appearance of corruption, the damage is
done. Americans are disaffected with politics and political campaigns and have
voted against the current system with their feet: U.S. voter turnout in elections is
in serious decline... Our representative democracy is harmed by eroding
participation... In response, we should be working to help reconnect the voters
with their elected officials and to invest them in the political debates of the day.
Campaign finance reform, in one form or another, is an important part of that
process" (Congressional Record 1999).

� Speaking about the eventually enacted 2002 Bipartisan Campaign Reform Act
(BCRA), Sen. Charles Schumer (D-NY) remarked, "We have to restore the
system of regulated contributions. If we don't, the cynicism and distrust and lack
of engagement that are already so pervasive will continue to spread. Our citizens

3
are increasingly tuned out from our democratic process. Voter turnout for the
1998 election was 36 percent, the lowest turnout for a nonpresidential election in
56 years. In presidential elections, turnout has declined 13 percent since 1960.
We all know that banning soft money won't cure all of this by itself, but it will
help restore the impression and the reality that politics is more than a game played
by and for only those who can afford to give" (Congressional Record 2002).

� A labor union, in response to attempts to strengthen BCRA, stated, "AFSCME
opposes these new efforts [for stricter laws], not only because they are premature,
but because they would limit the grassroots activities that help increase voter
turnout and they would further diminish the influence of ordinary Americans on
the political process" (AFSCME 2005).

� The Public Policy Institute of California noted, "Spending on [California] state
elections has increased dramatically whereas the turnout of registered voters has
steadily declined, reflecting the public's dissatisfaction with political candidates
and the type of campaign they wage" (Public Policy Institute of California 2004).
The brief went on to note that Californians would be in favor of reforms to
ameliorate these problems.

Scholars have made similar claims:

� Based upon their research finding a demobilizing effect of campaign advertising,
Ansolabehere et al. (1994) and Ansolabehere and Iyengar (1995) discuss potential
reforms to mitigate the impact of negative ads, which the authors find lead to
increased cynicism in the process.

� Teixeira, in a study of declining voter turnout, mentions campaign finance reform
as a possibility for increasing campaign competitiveness and improving
perceptions of government responsiveness (Teixeira 1992).

� In an article noting that compulsory voting is unlikely to have desired effects
unless changes in the political system are made, Franklin states, "[A]rguments for
compulsory voting divert attention from other proposed reforms of the American
electoral process: reforms which would address genuine deficiencies in that
process...Campaign finance reform, for example, by reducing the power of non-
elected bodies should increase the relevance of elected bodies and so raise the
stakes of elections to those bodies and the salience of those elections" (Franklin
1999, 216).

In short, there are ample arguments, both in political science and policy circles, that

campaign finance reform may have a meaningful impact on turnout. However, existing claims

are speculative and are typically based on only one slice of the larger theoretical picture. In fact,

4
we argue that all of the above claims are embedded in a complex system of equations relating

many aspects of the democratic process. In turn, each of these features has some impact on

turnout, as well as on each other. For instance, campaign finance laws may alter the

competitiveness of campaigns directly by enabling challengers to raise more or less money, but

they may also impact competitiveness via increasing or decreasing voter mobilization, which is

in turn directly affected by campaign finance laws. Parsing these effects requires a complex

structural model.

Such a model would need to consider the impact of campaign finance laws on the

following features of politics. If changes in campaign finance law limited the amount of

negative ads, this could have a positive, negative, or minimal impact on turnout, depending on

whether such ads are mobilizing (e.g., Goldstein and Freedman 2002), demobilizing (e.g.,

Ansolabehere and Iyengar 1995, Ansolabehere et. al 1994), or a wash (e.g., Finkel and Geer

1998). More generally, changes in campaign finance law could limit advertising, which would

be likely to negatively impact turnout (Freedman, Franz, and Goldstein 2004). Changes in

campaign finance law could potentially limit mobilization activities, which would hurt turnout

(Gerber and Green 2000, Rosenstone and Hansen 1993). Changes in the sources or size of

contributions could increase trust in government or political efficacy, which in turn could impact

turnout.2 Changes in campaign finance law could also increase or decrease campaign

competitiveness, which would have, respectively, a positive or negative impact on turnout, most

likely through increased mobilization (e.g., Matsusaka and Palda 1993, Matsusaka 1993, Cox

1988, Cox and Munger 1989).

2
This is a common assertion, though the empirical evidence suggests that trust in government does not have a major
impact on turnout (Rosenstone and Hansen 1993, Citrin 1974), but that efficacy seems to have an effect (Rosenstone
and Hansen 1993, Bennett 1986). However, Primo and Milyo (2006) show that campaign finance laws have no
substantively meaningful effect on either trust or efficacy.

5
To illustrate the problems inherent in estimating all of these effects simultaneously, we

will consider just trust, competitiveness, spending, and turnout. We can capture the relationships

with the following system of equations.

Define:

X = a vector of individual characteristics (e.g., age, race, party id, etc.)

Z = a vector of characteristics of state institutions (e.g., campaign finance laws, term

limits, etc.)

TRUST = an individual's perception of trust in government or political efficacy

TURNOUT = an individual's turnout decision

COMPETITION = competitiveness of elections at the time of an individual-level survey

SPENDING = campaign spending by candidates

A stylized structural model for i = (1, ..., n) individuals residing in s = (1, ..., k) states

may then be articulated as:

TURNOUTi = f1(Xi, Zs; TRUSTi, COMPETITIONs, SPENDINGs) (1)
TRUSTi = f2(Xi, Zs; COMPETITIONs, SPENDINGs) (2)
COMPETITIONs = f3(Zs; SPENDINGs) (3)
SPENDINGs = f4(Zs; COMPETITIONs) (4)

Estimating this system of equations would require potentially unavailable data as well as

instruments for the endogenous variables, which are often difficult to find or are subpar. On the

other hand, simply estimating equation 1, with the endogenous variables on the right-hand side,

would lead to biased and inconsistent parameter estimates.3 A better approach is to rewrite

3
The one study that does probe the laws-turnout link, using state-level data, is Gross and Goidel (2003). Using
aggregate data, the authors find that only public financing improves turnout. However, they do not control for year
or state fixed effects in the analysis, and they include endogenous regressors, so we can draw limited conclusions
from these findings.

6
equation one in terms of exogenous variables only. This reduced form can then be estimated

using standard OLS or MLE techniques to assess the net effects of campaign finance laws on

turnout.4 Formally,

TURNOUTi = g(Xi, Zs) (5)

The reduced form model represented by (5), which excludes all endogenous regressors

and includes state and year effects where appropriate, allows us to estimate the reduced form

impact of these laws. In short, then, theory tells us that each campaign finance law may have

many indirect, countervailing impacts on turnout, but these are not estimable given existing data

and methodological constraints. Rather than attempt to identify such a complex structural

system, we propose a reduced-form analysis that captures the net effect of each state campaign

finance law on turnout. This reduced-form approach has the advantages of avoiding errors in

model identification that are frequent in structural analysis while at the same time increasing the

range of years for which data is available.

Data
We analyze both aggregate and individual-level data. Aggregate turnout data is from

America Votes and is measured as total votes cast for governor in an election divided by voting

age population.5 The National Election Studies has the longest time series of questions on

turnout, spanning five decades, so we measure individual-level turnout using the NES. We focus

on self-reported turnout, rather than validated turnout (i.e., where survey researchers verify that

4
For further details, see Maddala (1983) and Kennedy (2003).
5
A better measure of turnout would be total votes case for governor divided by voting-eligible population. As
McDonald and Popkin (2001) have argued, declines in turnout in recent decades have been due to a change in the
denominator, not the numerator. Should data become available that spans our entire time period, we encourage our
study to be replicated using these figures.

7
an individual has actually voted), because validated turnout was used in only limited years,

primarily in the post-Buckley era, and there is little substantive difference in the results when

using validated instead of self-reported turnout. Turnout is dichotomous and is coded one if an

individual responded that he or she voted. Individual-level data is from the 1948-2000 NES

Cumulative Data File.

State-level data on political institutions are taken from The Book of the States and

Campaign Finance Law, while demographic data employed in the aggregate analysis are taken

from the Statistical Abstract of the United States.6 Summary statistics for variables used in the

aggregate-level analysis are presented in Table 1; the corresponding figures for the individual-

level analysis are presented in Table 4.

Campaign finance laws have changed dramatically in the states in recent decades. In

1950 few states had restrictions on contributions by individuals, but by 2000 limits on

contributions from both individuals and organizations (i.e., corporations, unions, and PACs)

were the norm. The trend in state reforms mirrors that at the federal level, where a major wave of

changes occurred in the 1970s. State reforms also picked up steam in the 1990s, with more than

one-third of states altering their laws during this period (Malbin and Gais 1998). We are in what

might be called an era of "mature" campaign finance regulation, since all states have disclosure

laws on the books, and most states have some restrictions on contributions.

We focus here on contributions to candidates rather than to parties, as information on the

latter is not readily available for the full time period under study. We consider disclosure laws,

contribution limits on organizations, contribution limits on individuals, the presence of public

financing tied to voluntary expenditure limits, and mandatory expenditure limits in place prior to

the Supreme Court's ruling that such limits were unconstitutional. Figure 1 depicts the number
6
Missing years for state-level demographic variables are linearly interpolated from adjacent years.

8
of states over time with each type of campaign finance law that apply to either legislative or

gubernatorial candidates.

[Insert Figure 1 about here]

While there are several ways to categorize and measure state-level laws, in this case

simpler is better. We measure the presence or absence of particular types of laws, such as

contribution limits and public financing. Using specific dollar amounts leads one into a morass,

in part because states greatly differ in many respects, including cost-of-living, wealth, and the

cost of media markets. Put concretely, does a $1000 limit on individual contributions to a

candidate mean the same thing in Arkansas as it does in California? If not, how would one

compare specific limits across states? Other aspects of campaign finance law, such as

enforcement quality, suffer from similar problems. The presence or absence of particular laws,

on the other hand, can be clearly measured and is directly comparable across states.

Methods

Aggregate-Level Analysis. We examine state-level voter turnout from every

gubernatorial election from 1950 to 2000 (N=756); turnout is measured as the total vote for

governor divided by voting age population. Campaign finance laws are those already in effect for

at least one year at the time of the election and, for the data analysis, refer to gubernatorial

elections only. Five dichotomous campaign finance variables represent the laws in each state.7

7
By defining the variables in this way, we avoid concerns about multicollinearity, especially between limits on
contributions to individuals and organizations. This was verified by conducting a variety of diagnostic tests for the
campaign finance variables. We do not create an index of laws because we do not expect their effects to be additive.
This is borne out in the empirical analysis.

9
These are indicators for the presence of

1. public disclosure of campaign contributions
2. limits on contributions by organizations only
3. limits on contributions by organizations and individuals
4. public subsidies to candidates that abide by expenditure limits
5. mandatory expenditure limits in place prior to the 1976 Buckley v. Valeo decision
outlawing such limits.

We estimate the relationship between voter turnout and campaign finance laws by a

grouped probit (weighted by voting age population), although the main substantive findings

presented here are little different when we estimate a grouped logit or a weighted least squares

model. While our analysis of aggregate turnout data suffers from the familiar ecological

inference problem (Freedman 2001), it nevertheless serves as a useful benchmark for comparison

to the analysis of individual-level turnout data. Our specification includes year fixed effects.

Because both campaign finance laws and voter turnout may be influenced by some unobserved

state-specific factors (e.g., a progressive ideology), state fixed effects are employed to limit the

potential for spurious correlation.

We also include controls for demographic variables and other state political institutions.

The demographic controls include the log of real per capita income, the percent of population

over age 65, percent black, percent with high school degrees, percent with college degrees, and

the Republican vote margin in the most recent presidential election (as a proxy for the partisan

leanings of the population). To account for partisan tides, we also interact the Republican

margin variable with the year indicators. Other state political institutions include indicators for

gubernatorial term limits, direct legislation, poll taxes, literacy tests, and ease of voter

registration (election day voter registration or no voter registration). Because the pre- and post-

Buckley eras differ so much, with voluntary public financing being introduced only post-Buckley

10
and mandatory expenditure limits disappearing due to the same decision, we also estimate these

two time periods separately.

Individual-Level Analysis. Using state-level variables when working with individual-

level variables requires caution. For instance, because the National Election Studies does not

include representative state samples, it is not possible to make claims about a specific state and

how it has changed over time.8 Rather, residing in a state should be viewed as a "treatment" on

the individuals, with state institutional features, including campaign finance laws, representing

treatment effects similar to those we might observe in medical experiments. In this way, we can

ascertain whether living in a state with particular campaign finance laws influences the decision

to vote. Probit models are estimated in this paper, with standard errors corrected for clustering

within state and year. We include only those respondents in states with gubernatorial elections at

the time of the survey (N=16,013).

Several state-level variables are included in the analysis to control for other features of a

state, besides changes in campaign finance law, that may influence turnout. These include

indicators for the presence of the citizen initiative, gubernatorial term limits, whether a poll tax

or literacy tests are necessary to vote, and whether easy voting registration is present (i.e., same

day or no advance registration required).9 We also include state dummy variables to control for

unobserved heterogeneity across states; this is particularly important given that state campaign

finance laws may be passed precisely in those states with chronically low levels of turnout. Year

dummy variables are included in the analysis to control for features of particular survey years

8
NES-provided weights ensure that the analysis of campaign finance laws is accurate with respect to subgroups.
Since in the NES sample the number of individuals surveyed in a state is typically related to that state's population,
we have also verified that there is a negligible relationship between the population of a state and the presence of
campaign finance laws.
9
Poll taxes and literacy tests are of course now outlawed; however, they were present for some years in our sample.

11
that may influence survey responses.10 To investigate whether there is a delay before campaign

finance laws affect turnout, we examine models that include four-year lags of these laws. We

also include four-year leads of laws in some specifications to see whether turnout is related to

future campaign finance laws, which is possible if low levels of turnout usher in reform.

Individual-level controls include education (grade school, high school, some college,

college or more), age, age squared, income (measured in percentiles, with a coding of 1

representing the 1st to 16th percentile, 2 the 17th to 33rd percentile, 3 the 34th to 67th percentile,

4 the 68th to 95th percentile, and 5 the 96th to 99th percentile), unemployment, race, gender, and

the strength of one's partisan affiliation (ranging from 1- 4). We include strength of affiliation

because we expect that individuals with strong partisan ties are more likely to vote, regardless of

whether they are Republicans or Democrats. In addition, we include controls for Republican and

Democratic party identification. We also interact party identification and year to assess whether

there are any national-level partisan trends that may cause Democrats or Republicans to turn out

to the polls to a greater (or lesser) degree. Additional variables include unified state Democratic

government and unified state Republican government, as well as interaction terms for

Democratic respondent living in a unified Democratic state, Democratic respondent living in a

unified Republican state, Republican respondent living in a unified Republican state, and

Republican respondent living in a unified Democratic state. These interactions capture whether

affiliating with an out-of-power party makes one less likely to vote.

10
When working with data of this type, it is also possible to use hierarchical linear modeling (HLM), which enables
the researcher to offer theoretical explanations for differences in behavior across levels. However, HLM requires
many more assumptions than standard regression or maximum likelihood techniques, and it is more sensitive to
measurement error (Bryk and Raudenbush 1992; Steenbergen and Jones 2002). For our purposes, then, HLM is not
the best method.

12
Results

Aggregate-Level Analysis. Summary statistics appear in Table 1. Table 2 reports the

results from the grouped probit estimation. Public financing has a positive and significant

influence on turnout, but most surprising is the large positive and significant effect of laws which

simply require disclosure of the source and size of campaign contributions. On the other hand,

mandatory expenditure limits of the sort outlawed in 1976 have a significant and negative effect

on turnout, while contribution limits also have a negative effect, albeit not significant.

Demographic and socioeconomic variables, such as age and education, have little effect on

turnout, though race and income have the expected effects.

[Insert Tables 1-3 about here.]

In order to better gauge the substantive impact of campaign finance reform, we calculate

the change in voter turnout attributable to a change in law at the mean turnout in the sample

(49%); these marginal effects are listed in Table 3. We also present marginal effects for

estimates obtained when we split the sample of gubernatorial elections around the Buckley

decision in 1976. The results suggest the importance of splitting the sample. For example,

public financing has a large positive effect in the full sample analysis, but when we omit years

prior to 1976, during which no states had implemented public financing, we find no effect.

Similarly, disclosure laws, which were passed in most states by 1980, appear to have no effect on

turnout when we restrict attention to the pre-Buckley period.

Finally, in order to facilitate the comparison of results across the aggregate and individual

analyses, we have re-estimated the model using only those elections that appear in the NES

sample employed in the subsequent individual-level analysis. We lose almost 400 observations

when we restrict ourselves to this NES-matched sample; nevertheless, the estimated coefficients

13
on campaign finance reforms are nearly identical to those obtained for the full sample of

gubernatorial elections from 1950-2000.

Overall, these results merit caution and illustrate the pitfalls of ecological analysis. For

instance, education, which is known to be a major predictor of turnout, fails to attain statistical

significance. This suggests that merely focusing on the aggregate analysis may lead to incorrect

inferences, as the next section demonstrates.

Individual-Level Analysis. Summary statistics appear in Table 4. Table 5 reports the

effects from the individual-level analysis. Year and state-level effects are statistically significant

in joint significance tests, so they are included in this specification. Lags and leads of the

campaign finance variables are not jointly significant, so we report the specification without

these variables included. The inclusion of dummy variables for year and state is important; in

na�ve specifications without these, one can find spurious relationships between campaign finance

laws and turnout.

[Insert Tables 4 and 5 about here]

Demographic variables have the expected impact on turnout. Voting propensity is

increasing in income, except for the very wealthy, who are slightly less likely to vote than those

on the next rung of the income ladder. The probability of voting is increasing monotonically in

education, is quadratic in age, and is lower for nonwhites and women but higher for strong

partisans. The unemployed tend to vote less than the employed. Partisan affiliation has no

impact on turnout.

Institutional restrictions on voting have mixed effects. The poll tax has the expected

negative sign and is statistically significant, but the literacy test has the wrong sign and is not

statistically significant. Easy registration laws, term limits, and direct democracy are statistically

14
as well as substantively insignificant. Similarly, the unified government variables have no

impact on turnout.

[Insert Table 6 about here]

Campaign finance laws also have neither a statistically nor a substantively significant

impact on turnout over the entire timer period. The first column of Table 6 displays the marginal

effects on the laws for a typical respondent. As we noted earlier, one critique of this analysis is

that our time period is too long. For instance, mandatory expenditure limits were outlawed by

Buckley, which means there is no variance in this variable from 1976 on, and public financing

does not emerge until the 1980s and 1990s, meaning that there is no variance in the pre-Buckley

era. If we split the sample into pre-1976 and 1976-2000, however, some fascinating patterns

emerge. For the period prior to Buckley, limits on organizational contributions increase the

probability of voting by 9 percent, a non-trivial amount. This effect is erased, however, once a

state adopts an individual limit as well. For 1976-2000, only public financing is statistically

significant, and its impact is negative and substantively large. The second and third columns of

Table 5 describe the effects for these two time periods. In short, then, the only two campaign

finance variables that attain statistical significance in the entire analysis are organizational

contributions in the years prior to 1976, and public financing for the modern era. The effects of

the two move in opposite directions but in nearly equal magnitudes. This suggests that early

reforms may have had a positive effect on turnout, but that such have effects have since

dissipated. Even more importantly, in an era of "mature" campaign finance reform, further

efforts, such as public financing, may have deleterious effects on turnout.

15
Discussion
This paper makes three important contributions. First, it offers a theoretical foundation

for understanding the ways in which campaign finance laws may impact turnout. Most studies

consider these effects piecemeal, but we present a more complete picture. Future work may

want to disentangle these effects, but we think that understanding the net impact of the laws is an

important first step for thinking about public policy.

Second, it makes a simple yet often overlooked methodological point. Reduced-form

analysis, while perhaps not as "sexy" methodologically as a structural analysis, provides a solid

foundation for subsequent study and has the virtue of simplicity. To analyze the impact of laws

that have multiple routes of influence on a variable of interest requires either a complicated

structural model that is likely to lead one down the wrong path or a willingness to focus on net

effects via a reduced form analysis. We choose the latter in this paper and are able to get

leverage on an important question. Importantly, by excluding endogenous variables from the

right-hand side of our estimations, we avoid endogeneity bias. Moreover, we demonstrate

(again) the pitfalls of ecological analysis.

Third, it contributes to the literature on campaign finance, and more broadly on the

"quality of democracy," by suggesting that campaign finance laws are unlikely to have a

significant and positive effect on turnout, and may in some cases have unintended negative

consequences. This is of both theoretical and practical interest. Theoretically, it reinforces

existing findings that using laws to impact democratic participation is difficult. On the practical

side, it offers policy analysts and jurists a scientific study that can take the place of speculation

with respect to the impact of campaign finance laws on turnout.

16
References
AFSCME. 2005. "AFSCME Legislation Fact Sheet: Campaign Finance Reform." Accessed
10/22/2005 at http://www.afscme.org/action/legfs17.htm.

Ansolabehere, Stephen, and Shanto Iyengar. 1995. Going Negative. New York: Free Press.

Ansolabehere, Stephen, Shanto Iyengar, Adam Simon, and Nicholas Valentino. 1994. "Does
Attack Advertising Demobilize the Electorate?" American Political Science Review
88:829-838.

Bennett, Stephen Earl. 1986. Apathy in America. Dobbs Ferry, NY: Transnational Publishers.

Book of the States. Various Years. Lexington, KY: Council of State Governments.

Bryk, Anthony S., and Stephen W. Raudenbush. 1992. Hierarchical Linear Models. Newbury
Park, CA: Sage.

Buckley v. Valeo. 424 U.S. 1 (1976).

Citrin, Jack. 1974. "Comment: The Political Relevance of Trust in Government." American
Political Science Review 68:973-988.

Congressional Research Service. Various Years. Campaign Finance Law. Washington, DC:
National Clearinghouse on Election Administration.

Congressional Record. 1999. Vol. 145, No. 139, S12804, October 19.

Congressional Record. 2002. Vol. 148, No. 33, S2096, March 20.

Cox, Gary W. 1988. "Closeness and Turnout: A Methodological Note." Journal of Politics
50:768-775.

Cox, Gary W., and Michael C. Munger. 1989. "Closeness, Expenditures, and Turnout in the
1982 U.S. House Elections." American Political Science Review 83:217-231.

Finkel, Steven E., and John G. Geer. 1998. "A Spot Check: Casting Doubt on the Demobilizing
Effect of Attack Advertising." American Journal of Political Science 42:573-595.

Franklin, Mark N. 1999. "Electoral Engineering and Cross-National Turnout Differences: What
Role for Compulsory Voting?" British Journal of Political Science 29:205-216.

Freedman, David A. 2001. "Ecological Inference and the Ecological Fallacy." International
Encyclopedia for the Social and Behavioral Sciences 6:4027-4030.

17
Freedman, Paul, Michael Franz, and Kenneth Goldstein. 2004. "Campaign Advertising and
Democratic Citizenship." American Journal of Political Science 48:723-741.

Gerber, Alan S., and Donald P. Green. 2000. "The Effects of Canvassing, Telephone Calls, and
Direct Mail on Voter Turnout: A Field Experiment." American Political Science Review
94:653-663.

Goldstein, Ken, and Paul Freedman. 2002. "Campaign Advertising and Voter Turnout: New
Evidence for a Stimulation Effect." Journal of Politics 64:721-740.

Gross, Donald A., and Robert K. Goidel. 2003. The States of Campaign Finance Reform.
Columbus, OH: Ohio State University Press.

Highton, Benjamin. 2004. "Voter Registration and Turnout in the United States." Perspectives
on Politics 2:507-515.

Kennedy, Peter. 2003. A Guide to Econometrics. Cambridge, MA: MIT Press.

Maddala, G.S. 1983. Limited-Dependent and Qualitative Variables in Econometrics. Cambridge:
Cambridge University Press.

Malbin, Michael J., and Thomas L. Gais. 1998. The Day After Reform: Sobering Campaign
Finance Lessons from the American States. Albany, NY: Rockefeller Institute Press.

Matsusaka, John G. 1993. "Election Closeness and Voter Turnout: Evidence from California
Ballot Propositions." Public Choice 76:313-334.

Matsusaka, John G., and Filip Palda. 1993. "The Downsian Voter Meets the Ecological
Fallacy." Public Choice 77:855-878.

McConnell v. FEC. 540 U.S. 93 (2003).

McDonald, Michael P., and Samuel Popkin. 2001. "The Myth of the Vanishing Voter."
American Political Science Review 95:963-974.

Moore, Dennis. 1999. "Campaign Finance Reform is Needed." Olathe Daily News, October 22.

Nixon v Shrink Missouri Government PAC. 2000. 528 U.S. 377

Primo, David M., and Jeffrey Milyo. 2006. "Campaign Finance Laws and Political Efficacy:
Evidence From the States." Election Law Journal 5.

Public Policy Institute of California. 2004. "Research Brief: Voters' Views of Politics in
California: Dissatisfaction, Distrust, and Withdrawal." Issue # 96, November.

18
Rogers, William. 1993. "sg17: Regression standard errors in clustered samples." Stata Technical
Bulletin 13:19-23.

Rosenstone, Steven J., and John Mark Hansen. 1993. Mobilization, Participation, and
Democracy in America. New York: Macmillan.

Steenbergen, Marco R., and Bradford S. Jones. 2002. "Modeling Multilevel Data Structures."
American Journal of Political Science 46:218-237.

Teixeira, Ruy A. 1992. The Disappearing American Voter. Washington, DC: Brookings
Institution Press.

United States Census Bureau. Various Years. Statistical Abstract of the United States.
Washington, DC: GPO.

White, Halbert. 1980. "A Heteroskedasticity-Consistent Covariance Matrix Estimator and a
Direct Test for Heteroskedasticity." Econometrica 48:817-830.

Wolfinger, Raymond E., Benjamin Highton, and Megan Mullin. 2005. "How Postregistration
Laws Affect the Turnout of Citizens Registered to Vote." State Politics and Policy
Quarterly 5(1):1-23.

Wolfinger, Raymond E., and Steven J. Rosenstone. 1980. Who Votes? New Haven, CT: Yale
University Press.

19
Table 1. Summary Statistics for Aggregate-Level Analysis
Variable Mean Standard
Deviation

Gubernatorial turnout (%voting-age population) 49.25 14.50
Public disclosure of contributions .81 .40
Contribution limits (organizations only) .38 .49
Contribution limits (orgs. and ind.) .28 .45
Public funding .05 .22
Mandatory expenditure limits .33 .47
Log(real per-capita income) 9.73 0.37
High School 52.57 22.96
College 13.17 6.63
Age65+ (% Population) 10.45 2.47
Black (%Population) 8.36 9.34
Republican % of the Presidential Vote 50.87 11.16
Citizen initiative .47 .50
Gub. term limits .47 .50
Poll tax .04 .20
Literacy test .12 .33
Easy registration .03 .18
Off-year election .08 .26
Data are for all state gubernatorial elections from 1950-2000; N=756.

20
Table 2. Results from Aggregate-Level Analysis
Variable Impact on Turnout
Coefficient t-statistic

Public disclosure of contributions .062 2.55
Contribution limits (organizations only) -.022 0.97
Contribution limits (orgs. and ind.) -.037 1.58
Public funding .108 4.11
Mandatory expenditure limits -.053 2.46
Log of real per-capita income .705 5.85
High school .000 0.02
College -.002 0.55
Age65+ (% population) .010 1.18
Black (% population) -.029 6.38
Republican % of the presidential vote .135 0.85
Citizen initiative -.016 0.42
Gub. term limits .025 1.07
Poll tax -.287 7.54
Literacy test -.043 1.62
Easy registration .076 1.71
Off-year election -1.26 2.24
Constant -6.51 5.81

Grouped probit specification (weighted by voting age population) includes state and year dummy variables, as well
as a set of interaction variables for (year X republican share of presidential vote). Absolute values of t-statistics
listed.; N=756.

21
Table 3. Marginal Effects of Campaign Finance Laws on Turnout
(Estimates from Aggregate-Level Analysis)

NES-matched
Variable Entire sample Pre-1976 1976-2000
Mean of dependent variable .49 .49 .52 .45

Change in probability from the presence of campaign finance laws
Public disclosure .025** .016 .011 -.008
of campaign contributions
Limits on contributions -.009 -.016 .042** -.019
from organizations only
Limits on contributions from -.015 -.022* .020 -.005
organizations and individuals
Public funding of candidates .043*** .043*** n/a -.010
conditional on expenditure limits
Mandatory expenditure limits -.021** -.023** -.020 n/a
(pre-Buckley)
The change in probability of a favorable response from the implementation of a particular law (or set of laws) is
derived from the estimated coefficients of the probit models, where all changes are calculated at the mean of the
dependent variable. *p < .10, ** p < .05, ***p < .01.

22
Table 4. Summary Statistics for Individual-Level Analysis
Variable Mean Standard
Deviation

Did you vote? .64 .48
Public disclosure of contributions .87 .33
Contribution limits (organizations only) .45 .50
Contribution limits (orgs. and ind.) .28 .45
Public funding .07 .26
Mandatory expenditure limits .25 .43
Income=2 .17 .38
Income=3 .32 .47
Income=4 .29 .45
Income=5 .05 .22
Unemployed .06 .24
High School .48 .50
Some College .19 .39
College .16 .37
Age 45.07 16.95
Age Squared 2318.62 1681.63
Nonwhite .15 .36
Female .55 .50
Partisan Strength 2.88 .97
Democrat .54 .50
Republican .35 .48
Unified Dem. Govt. .33 .47
Unified Rep. Govt. .12 .33
Dem. x Unified Dem. .21 .40
Dem. x Unified Rep. .06 .24
Rep. x Unified Rep. .05 .22
Rep. x Unified Dem. .09 .29
Citizen Initiative .48 .50
Gub. Term Limits .44 .50
Poll Tax .04 .19
Literacy Test .07 .25
Easy Registration .02 .15
The NES was administered every two years beginning in 1948, with the exception of 1950. The turnout question
was not asked in 1954 or 1962, and the unemployed question was not asked in 1954 or 1966, so those years are
omitted from the sample. 1948 is omitted because many questions were not asked in that initial year. N=16,013. The
following states were not included in the analysis due to a lack of observations, if responses did not vary within the
state, or if they held off-year elections: Alaska, Delaware, Hawaii, Kentucky, Mississippi, Montana, Nevada, New
Jersey, North Dakota, Rhode Island, Vermont, and Virginia.

23
Table 5. Results from Individual-Level Analysis
Variable Impact on
Turnout
Coef. z-stat
Public disclosure of contributions -.05 .61
Contribution limits (organizations only) .07 1.00
Contribution limits (orgs. and ind.) -.03 .49
Public funding -.05 .73
Mandatory expenditure limits -.02 .33
Income=2 .21 5.38
Income=3 .37 9.99
Income=4 .52 12.38
Income=5 .49 6.52
Unemployed -.25 4.12
High School .36 9.93
Some College .76 16.13
College .95 19.43
Age .07 15.53
Age Squared .00 10.92
Nonwhite -.12 3.46
Female -.09 3.86
Partisan Strength .27 14.25
Democrat -.06 .12
Republican .08 .20
Unified Dem. Govt. -.11 1.27
Unified Rep. Govt. -.05 .45
Dem. x Unified Dem. .16 1.62
Dem. x Unified Rep. .02 .23
Rep. x Unified Rep. .17 1.26
Rep. x Unified Dem. .10 .82
Citizen Initiative -.05 .37
Gub. Term Limits -.01 .11
Poll Tax -.38 3.71
Literacy Test .03 .28
Easy Registration .00 .03
Constant -.05 .61

Probit specification includes state and year dummy variables with robust standard errors adjusted for clustering
within state and year, as well as interaction terms for year and party identification. Absolute values of z-statistics are
presented in the table. The dependent variable is coded 1 for a response indicating that an individual voted, and 0
otherwise. N=16,013.

24
Table 6. Marginal Effects of Campaign Finance Laws on Turnout
(Estimates From Individual-Level Analysis)

Variable Did You Vote? Did You Vote? Did You Vote?
(Entire Sample) (Pre-1976) (1976-2000)
Mean of dependent variable .64 .69 .58

Change in probability from the presence of campaign finance laws
Public disclosure -.019 -.06 n/a
of campaign contributions
Limits on contributions .026 .093* -.06
from organizations only
Limits on contributions from -.010 .010 -.026
organizations and individuals
Public funding of candidates -.02 n/a -.087**
conditional on expenditure limits
Mandatory expenditure limits -.008 -.041 n/a
(pre-Buckley)
The change in probability of a favorable response from the implementation of a particular law (or set of laws) is
derived from the estimated coefficients of the probit models, where all changes are calculated at the mean of the
dependent variable. *p < .10, ** p < .05.

25
Figure 1: Campaign Finance Laws in the U.S. States, 1950-2000

25
Individual Limits
Corp./Union Limits
20
Public Financing

Number of States with Law
Disclosure Laws
15 Mandatory Expenditure Limits

50 952 954 956 958 960 962 964 966 968 970 972 974 976 978 980 982 984 986 988 990 992 994 996 998 000
19 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2
Year