Casting Votes

• Date: 01 December 2011
• Author: Lyn Ragsdale,Jerrold G. Rusk
• Published date: 01 December 2011
• DOI: 10.1177/1065912910379230

Political Research Quarterly

64(4) 840 –857

© 2011 University of Utah

Reprints and permission:

sagepub.com/journalsPermissions.nav

DOI: 10.1177/1065912910379230

http://prq.sagepub.com

Casting Votes: The National Campaign

Context and State Turnout, 1920–2008

Lyn Ragsdale1 and Jerrold G. Rusk1

Abstract

This study examines voter turnout in the American states in U.S. presidential and House elections from 1920 through

2008. A model predicts turnout as the sum of the national campaign context, state autonomy, and electoral continuity.

The national campaign context encompasses conditions that prompt turnout to shift similarly across states. State

autonomy involves state-specific factors that prompt turnout to vary across states. Electoral continuity involves

people voting in successive elections, regardless of other influences. Testing the model finds that national campaign

context effects have increased, but they vary by year, election type, and region and have been mixed since the 1970s.

Keywords

voter turnout, electoral history, nationalization, presidential elections, congressional elections

Studies have long debated the extent to which there has

been a nationalization of American elections. This nation-

alization is conceptualized as a long-term, future-oriented

process that produces a secular increase in the effects of

national factors on election outcomes. Nationalization

is not simply the presence of national effects in a single

election, but instead involves longitudinal change in an

increasingly national direction, including systemic shifts

in party alignments as well as similar responses among

the states to political issues, events, and images. National-

ization also indicates that there is a concomitant decline in

the impact of state and local stimuli on elections, and with

that, parochialism, sectionalism, and localism diminish.1

In his classic work, Stokes (1969, 1975) analyzes the

nationalization of the party vote and voter turnout for U.S.

House elections examined by decade from the 1870s to the

1950s. He finds that among national, state, and constitu-

ency factors, national components have a sizable, increas-

ing effect on turnout and constituency components have

a sharply decreasing effect on turnout and the party vote.

Katz (1973), employing a different method than Stokes,

observes even greater national effects on the party vote. In

contrast, Claggett, Flanigan, and Zingale (1984) see no

trend toward the nationalization of the party vote in House

elections from 1842 through 1970. Vertz, Frendreis, and

Gibson (1987) find that nationalization of the party vote

occurs in presidential elections, but not in House, Senate,

or gubernatorial races. Bartels (1998, 287) observes that

there “are still substantial ebbs and flows in the national-

ization of electoral forces” for the party vote in presidential

elections from 1868 to 1996. To make conclusions even less

clear, Stokes’s original investigation of the national ization of

turnout has not been updated or reexamined. Instead,

turnout scholars focus on a possible structural decline in

the number of people casting votes (e.g., Teixeira, 1987,

1992; McDonald and Popkin, 2001; Patterson, 2002;

Wattenberg, 2002; Rosenstone and Hansen, 2003).

The National Campaign

Context Model of State Turnout

The current study, then, analyzes the degree to which

national factors influence turnout in U.S. presidential

elections and midterm House contests from 1920 through

2008. The question is examined using state aggregate turn-

out measured for the country as a whole and regionally for

the South and the non-South.2 Theoretically, a national

campaign context model is proposed, which rests on three

premises:

1. The extent of change or stability in the national

campaign context determines the degree of

national homogeneity of turnout.

2. State autonomy limits the influence of the

national campaign context on turnout.

3. Turnout in one election will be a strong predic-

tor of turnout in the next, thereby limiting both

national and state effects.

1Rice University, Houston, TX, USA

Corresponding Author:

Lyn Ragsdale, Department of Political Science, 6100 Main Street,

Rice University, Houston, TX 77005, USA

Email: lkragsdale@rice.edu

Ragsdale and Rusk 841

Elements of the Model

The campaign context model posits that turnout in a given

state in an election year is the sum of three components—

(1) an election-specific component reflecting the shifts in

the national campaign context, (2) an idiosyncratic com-

ponent involving subnational factors within a given state,

and (3) a temporal component that reflects the continuity

in turnout across elections. First, individual elections take

place amid a national campaign context—the set of national

circumstances among the events, conditions, and political

cleavages that confront campaigners and the eligible elec-

torate in the months before Election Day. When circum-

stances change sharply—positively or negatively—from

occurrences in the recent past, the electorate becomes

“more attuned to national level events, personalities, and

issues and hence is more homogeneous in its behavior”

(Claggett, Flanigan, and Zingale, 1984, 77). Thus, turnout

increases (decreases) similarly across the states. Many

paths may produce this outcome. Advances in mass com-

munication may excite people about politics in new ways

that universally shift turnout. A national crisis, such as a

war, may alter citizen’s political interest or trust, produc-

ing greater uniformity in state turnout. Political cleav-

ages may shift, yielding sharp gains for one party over

the other across large parts of the country and creating

more interest in voting. Economic conditions may sharply

improve or worsen from an earlier period prompting

similar increases in voting among the states. Changes in

voting laws may improve people’s ability to vote in cer-

tain states, thereby making turnout more consistent across

all states (Hill, 2006).

Second, state autonomy prompts turnout to vary by

state, as distinctive conditions below the national level—

the success of individual candidates, the hold of state

party organizations and election laws, and the intrusion

of purely state issues—carry the day (Gimpel, 1996).

Renner (1999, 122) writes that “the fifty American states

have their own unique political personalities or political

cultures. These interstate differences reflect divergent

ideological belief systems, party divisions, formal party

organizational strength, demographic characteristics,

and the nature of physical settlement.” Thus, federalism

insures that national effects that do arise may not be per-

manent or stable. To be sure, state autonomy does not

suggest the end of intrastate differences, but instead uni-

formity within a state may be greater than uniformity

across states.

Third, temporal continuity indicates that for a signifi-

cant portion of the eligible electorate, the national campaign

context and state autonomy are of little or no importance.

Many people are habitual voters who go to the polls in

successive elections, regardless of either the national

campaign context or state factors (Green and Shachar,

2000; Plutzer, 2002). There are also habitual nonvoters—

people who are eligible to vote, but who neither register

nor vote. These opposite types of eligibles limit the influ-

ence of national and state factors and create continuity

from one election to the next. Thus, turnout at time t is

likely to strongly predict turnout at time t + 1.

Estimating the Model

These components can be captured in a simple equation

that regresses state turnout in an election year on previous

state turnout plus a constant and an error term. This

approach follows Bartels (1998) who applies it to parti-

san voting. The model is:

TOst = at + b1tTOst-1 + Єst

where TO represents turnout in state s in election year t

and TOst-1, represents turnout in that state in the immediate

past election.3 The intercept at measures the shift in

turnout from the previous election attributable to national

forces in the current election. Thus, for the current election,

the term captures the degree of similarity across state

turnout levels associated with national effects. Positive

values indicate that national factors have increased state

turnout; negative values indicate that national factors have

decreased state turnout. This is an appropriate summary

measure of the various factors that make up the national

campaign context. While the measure cannot assess speci fic

elements—technology, the economy, crises, political

cleavages, or registration laws—it offers a consistent

composite across time and election types. The parameter

b1t estimates the effect of lagged state turnout on current-

year turnout, reflecting continuity. The parameter Єst is a

stochastic term pertaining to state-specific idiosyncratic

factors in election year t. The model assumes that this

standard error term Єst is drawn from a probability

distribution with mean zero and election-specific variance

sn

2. The stochastic variance parameter sn

2 measures the

size of current state forces in a given election. Each of

the terms in the model is measured in percentage points.

The national campaign context model thus predicts that the

impact of national factors will vary by election. National

circumstances paramount in one election may be absent

or of limited importance in the next. Even if national

factors have a predominant effect on turnout in a series of

elections, this periodicity does not necessarily equate with

secular, linear nationalization.

The national campaign context model avoids limita-

tions of the variance components model, first introduced

by Stokes and adopted by others.4 First, the variance

components model is estimated across five elections during