Issue Scales, Information Cues, and the Proximity and Directional Models of Voter Choice

• Author: Kevin K. Banda,Jonathan Kropko
• DOI: 10.1177/1065912918760729
• Date: 01 December 2018
• Published date: 01 December 2018
• Subject Matter: Articles

https://doi.org/10.1177/1065912918760729

Political Research Quarterly

2018, Vol. 71(4) 772 –787

© 2018 University of Utah

Article reuse guidelines:

sagepub.com/journals-permissions

DOI: 10.1177/1065912918760729

journals.sagepub.com/home/prq

Article

Introduction

Learning why citizens choose one candidate over another

is key to understanding how representative democracies

function. We cannot deduce the meaning of election out-

comes if we do not understand the reasons underlying

citizens’ choices. What drives citizens’ decision-making

processes when they make their vote choices, and what

role do candidates’ spatial positions on issues play in citi-

zens’ decisions?

According to the proximity theory of voting (e.g.,

Downs 1957), a citizen prefers a candidate whose posi-

tions exactly match her own. Given a choice between two

candidates, she should prefer the closer of the two, even

if the closer candidate is on the opposite side of the posi-

tional midpoint. A slightly left-of-center voter should

thus choose to vote for a slightly right-of-center candi-

date over a more distant far-left candidate. Under the

directional theory of voting (e.g., Rabinowitz and

Macdonald 1989), the citizen is not primarily concerned

with the candidates’ positions, but rather with the sides

that they occupy in issue space. The citizen should prefer

any reasonable candidate on her side relative to any can-

didate on the other side, even if the candidate on the other

side is closer to her own position.1 In this case, a slightly

left-of-center voter should choose the far-left candidate

over the closer slightly right-of-center candidate.2

These models represent fundamentally different ways

to conceive of voter behavior, and they have important

implications for candidate strategy. According to the

proximity model, a voter at the ideological midpoint is

principled and prefers the candidate who is closest to the

midpoint. Candidates should maximize their support by

locating themselves at the median issue position to com-

pete for the support of the median voter. Under the direc-

tional model, however, voters at the midpoint are

indifferent or ambivalent toward candidates at any posi-

tion and candidates maximize support by taking more

extreme positions on issues to better distinguish them-

selves from one another when competing for votes. We

should observe moderate candidates in a proximity

world3 and polarized candidates in a directional world.

This research joins a longstanding literature that com-

pares the proximity and directional models. Like other

scholars, we assume that an issue’s positions can be

graphed on a continuum from liberal to conservative. We

760729PRQXXX10.1177/1065912918760729Political Research QuarterlyKropko and Banda

research-article2018

1University of Virginia, Charlottesville, USA

2Texas Tech University, Lubbock, USA

Corresponding Author:

Jonathan Kropko, The Woodrow Wilson Department of Politics,

University of Virginia, 1540 Jefferson Park Avenue (JPA), S183 Gibson

Hall, Charlottesville, VA 22904.

Email: jkropko@virginia.edu

Issue Scales, Information Cues, and the

Proximity and Directional Models of

Voter Choice

Jonathan Kropko1 and Kevin K. Banda2

Abstract

One of the most important questions in the study of democratic politics centers on how citizens consider issues and

candidate positions when choosing whom to support in an election. The proximity and directional theories make

fundamentally different predictions about voter behavior and imply different optimal strategies for candidates, but

a longstanding literature to empirically adjudicate between the theories has yielded mixed results. We use a survey

experiment to show that the way that candidates’ issue positions are described can cue citizens to choose a candidate

that is preferred under the expectations of either the proximity or the directional theory. We find that directional

voting is more likely when the issue scale is understood to represent degrees of intensity with which either the liberal

or the conservative side of the issue is expressed and that proximity voting is more likely when an issue scale is

understood to be a range of policies.

Keywords

spatial voting, proximity theory, directional theory

Kropko and Banda 773

prompt participants to place themselves on this scale and

to choose between two candidates with positions on this

scale. We place the candidates such that one is the pre-

dicted choice of the proximity model and the other is the

predicted choice of the directional model. Unlike other

studies, which aim to choose the best single model for the

entire sample, we argue that not only are both proximity

and directional voters present in the electorate, but that

citizens can be cued to think about the issue in either way

depending on how the candidates’ positions are commu-

nicated. Proximity voting is more likely when an issue

scale is understood to be a range of policies, whereas

directional voting is more likely when the issue scale is

understood to represent degrees of intensity with which

either the liberal or the conservative side of the issue is

expressed. We use an experiment embedded in a nation-

ally representative survey to demonstrate these patterns.

In other words, both proximity and directional think-

ing are present in the electorate and the way in which the

candidates’ positions are communicated matters because

this information can cue individuals to use one model or

the other. Political discourse that emphasizes binary posi-

tions—the liberal side or the conservative side—and the

strength or purity of candidates’ positions encourages

directional thinking, which may aid extreme candidates.

Discourse that emphasizes the policy content of interme-

diate positions on an issue, on the contrary, encourages

proximity thinking, which in turn may benefit moderate

candidates.

Prior Comparisons of the Proximity

and Directional Models

Many substantive results are predicated on proximity vot-

ing and might not hold if the directional model is instead

correct. The proximity and directional models posit differ-

ent sets of incentives for candidates. The median voter

theorem (e.g., Black 1958; Downs 1957; Hotelling 1929),

for example, only holds under proximity voting. If voters

are directional, then the positions that maximize the

chance of electoral victory are the most extreme accept-

able positions for each issue on the same side as the

median voter rather than the precise positions of the

median voter herself. It follows that the elected represen-

tative in a single-member district, such as in elections for

the U.S. House of Representatives, may be much more

ideologically extreme than the median voter in the district.

Thus, this result may provide an explanation for polariza-

tion in Congress beyond the effects of redistricting and

primary elections. In addition, while proximity voting

implies that centrist parties in multiparty democracies—

where centrist is defined as adopting a platform that takes

moderate positions on most issues—can have stable con-

stituencies of moderate voters, directional theory predicts

that centrist parties are likely limited to unstable support

(Macdonald, Listhaug, and Rabinowitz 1991).

Given their differing implications, it is understandable

that the question of the relative validity of the proximity

and directional models has led to a voluminous literature.

Tomz and Van Houweling (2008) identify more than fifty

independent studies that attempt to adjudicate between the

models. The theories have been directly compared in their

ability to model observational data from American presi-

dential elections, American congressional elections,

Canadian and European elections, and even in how voters

are influenced by their spouses (see, for example, Adams,

Bishin, and Dow 2004; Blais et al. 2001; Cho and

Endersby 2003; Dow 1998; Iversen 1994; Johnston,

Fournier, and Jenkins 2000; Kenny and Jenner 2008;

Krämer and Rattinger 1997; Lewis and King 1999;

Macdonald, Listhaug, and Rabinowitz 1991; Macdonald,

Rabinowitz, and Listhaug 1995, 1998, 2001; Merrill 1994,

1995; Merrill and Grofman 1997; Montgomery and

Nyhan 2010; Platt, Poole, and Rosenthal 1992; Rabinowitz

and Macdonald 1989; Rabinowitz, Macdonald, and

Listhaug 1991; Westholm 1997).

Although the proximity and directional models are

empirically testable and comparable, the studies that com-

pare the models have not yielded a consensus. There are

four difficulties inherent in observational research that

obscure a valid and general comparison. First, a large

branch of this literature shows that the divergent results are

conditioned on different data coding decisions (e.g., Cho

and Endersby 2003; Gilljam 1997a, 1997b; Macdonald,

Rabinowitz, and Listhaug 1997, 1998; Merrill and

Grofman 1997; Pierce 1997). Given these concerns, Lewis

and King (1999, 32) conclude that “there exists essentially

no evidence . . . to distinguish between the two models”

using observational data. Second, none of these studies

allow for the possibility that directional and proximity vot-

ers coexist in the electorate (Morris and Rabinowitz 1997).

If some voters follow the directional model, and some vot-

ers follow the proximity model, then in the population we

expect to see a mixed model emerge as the best overall fit.

Third, few of the observational studies have attempted to

isolate the positions in ideological space for which the

proximity and directional models make different predic-

tions. A voter who reports the most extreme ideological

self-placement, for example, will prefer the most proxi-

mate candidate under both the proximity and directional

models. As a result, a large proportion of the samples used

in these studies contains observations that offer no guid-

ance on the relative validity of the models. Last, none of

these studies have considered how the information context

in which voters learn about candidate positions cues voters

to consider candidates in ways that are congruent with one

model or the other. We address these last two points in

more detail below.