A New Measure for the Tendency to Select Ideologically Congruent Political Information

International Journal of Communication 10(2016), 200–225 1932–8036/20160005

A New Measure for the Tendency to Select Ideologically

Congruent Political Information:

Scale Development and Validation

YARIV TSFATI

University of Haifa, Israel

Selective exposure is a popular research construct, but the strategies used to

operationalize ideologically congruent exposure in contemporary correlational research

are problematic. This article offers a novel approach, asking people directly about their

tendency to seek information that is ideologically congruent with their opinions. A new

measure for the tendency toward congruent selective exposure is proposed and was

tested on three different data sets. In all three studies, confirmatory factor analysis

revealed a two-factor model, with one factor representing the tendency to select

congruent information and the other representing the tendency to avoid incongruent

information.

Keywords: selective exposure, selective avoidance, confirmation bias

The term selective exposure (SE) refers to the fact that “people tend to see and hear

communications that are favorable or congenial to their predispositions” (Berelson & Steiner, 1964, p.

529; for similar definitions, see Childs, 1965; Klapper, 1960; Lipset, Lazarsfeld, Barton, & Lintz, 1954).

Dating back to the seminal Lazarsfeld, Berelson, and Gaudet (1948) study, the notion of SE is probably as

old as communication research itself, and the tendency of individuals to expose themselves to like-minded

communications has been documented in a variety of contexts (for a meta-analysis, see D’Alessio & Allen,

2007). SE is said to be “one of the most widely accepted principles in sociology and social psychology” and

“a basic fact in the thinking of many social scientists about communication effects” (Sears & Freedman,

1967, p. 194).

Despite its popularity, SE has also received severe criticism (Sears & Freedman, 1967). However,

changes in the media landscape in the past two decades, most importantly, the multiplicity of online

Yariv Tsfati: [email protected]

Date submitted: 2015–01–18

1 Studies 1 and 2 were funded by the Israel Science Foundation [Grant #477/11]. Study 3 was funded by

the Israeli Ministry of Science and Technology [Grant #3-8784].

Copyright © 2016 (Yariv Tsfati). Licensed under the Creative Commons Attribution Non-commercial No

Derivatives (by-nc-nd). Available at http://ijoc.org.

International Journal of Communication 10(2016) Scale Development and Validation 201

media channels and cable television and the reemergence of partisan television and online outlets as

popular sources of political information for large parts of the electorate, have restored the relevance of the

concept of SE to communication scholars (Bennett & Iyengar, 2008). Indeed, in the past decade,

communication research has seen a major “comeback” of empirical research and conceptual work on SE.

The improvements in methods and data quality have enabled scholars to examine the dynamics of SE over

the course of time (Stroud, 2008), demonstrate selectivity using behavioral measures (Iyengar, Hahn,

Krosnick, & Walker, 2008), and develop a more nuanced understanding of different types of SE,

differentiating SE from selective avoidance (Garrett, 2009b). Recent studies also demonstrate that social

identity processes explain SE (Knobloch-Westerwick & Hastall, 2010), not just individual factors as

previously demonstrated (e.g., homophily discussed by Wheeless, 1974).

It would not be exaggerating, then, to say that the notion of SE had an immense influence on the

social sciences in the 1950s and 1960s (Festinger, 1957; Hovland, 1959), and that this concept is

reemerging as a central line of inquiry about the effects and consumption of news in the contemporary

media landscape. Nevertheless, despite the centrality of SE for political communication research, previous

studies were severely limited by a lack of standard measures operationalizing the extent to which different

people tend to prefer ideologically congruent political information. In this article, I propose a novel

strategy for measuring the preference for ideologically congruent SE, present a new measurement tool,

and test it in three different contexts. The article also provides some preliminary evidence about the tool’s

psychometric qualities and convergent, discriminant, and construct validities.

Current Approaches to Measuring SE and Their Limitations

The extant literature on SE can be divided into laboratory studies offering participants a choice

between concordant and discordant information (Iyengar et al., 2008; Knobloch-Westerwick, 2012) and

survey studies that typically ask respondents to report on their habitual exposure to various ideologically

slanted and mainstream media channels (Stroud, 2008). Of the four strategies employed to measure SE in

the literature reviewed by Clay, Barber, and Shook (2013), two were survey-based strategies and the

other two were observational. The two survey-based strategies included retrospective reports of whether

people had attended to specific types of information in the past and their behavioral intentions to attend

to specific types of information in the future; the three studies cited as examples for this approach (e.g.,

Garrett, 2009a) presented participants with synopses of articles and asked them about their preference

for and intention to read the full articles. Beyond the studies reviewed by Clay et al., the self-report

retrospective strategy seems to be the most widely used in the extant research on SE, in particular when

it comes to survey research (e.g., Daniller, Silver, & Moehler, 2013; Dvir-Gvirsman, 2014; Garrett et al.,

2014; J. Kim, 2015; Lawrence, Sides, & Farrell, 2010; Tsfati, Stroud, & Chotiner, 2014).

Although observational approaches have substantial face validity, because the experiments offer

the participants a choice between ideologically slanted information sources and measure their subsequent

choice, the survey studies using retrospective strategies are limited in several respects in addition to the

limitations reviewed by Clay et al. (2013, pp. 151–152; i.e., possible confounding with selective attention

or selective retention and possible biases due to self-presentation; see also Knobloch-Westerwick, 2015,

Chapter 3). First, such studies typically assume ideological consonance within a given media outlet. For