A Review of Public Issue Salience: Concepts,Determinants and Effects on Voting

Labels, Definitions and Properties

As Miller et al. (2017: 125) state, ‘despite its frequent use in the literature, policy issue salience has more often been a vague metaphor than a precisely defined scientific concept with an accepted operationalization’. Political scientists have used multiple terms to refer to the concept that this article calls ‘issue salience’, both on the supply side (i.e. political actors) and the demand side (i.e. the public). These include ‘policy attitude importance’ (Krosnick, 1990), ‘issue importance’ (Fournier et al., 2003), ‘policy issue salience’ (Miller et al., 2017) and, in earlier works, alluded to with terms like ‘importance’, ‘centrality’, ‘ego involvement’ and, indeed, ‘salience’ (for overview of early works, see Krosnick, 1988).

The various definitions of issue salience are similar, though they reveal at least one consequential divergence: those that have defined issue salience in psychological terms of how important an individual believes an issue to be (or how much thought they give to it, or similar) and those that defined salience in purely behavioural terms as the weight an individual gives to an issue when making behavioural choices, typically electoral. ²

In line with the first strand of definitions, and in his seminal series of earlier works, Krosnick (1990: 60) defined salience as ‘the degree to which a person is passionately concerned about and personally invested in an attitude’. Boninger et al. (1995: 62) continued this line of reasoning, defining ‘attitude importance’ as ‘an individual’s subjective sense of the concern, caring, and significance he or she attaches to an attitude’. More recently, Opperman (2010: 3) defined issue salience as ‘the relative importance and significance that an actor ascribes to a given issue on the political agenda’. In this sense, salience itself could be considered a type of attitude.

In line with the second ‘behavioural’ strand of definitions, Bernstein (1995: 488) states ‘salience is the weight that each person applies to the perceived distance from each candidate on each issue’ when making electoral choices. Ansolabehere and Puy (2018: 104, see also Butler and Stokes, 1969) defined issue salience likewise as ‘the relative weight that voters place on each issue, holding constant the positions of parties on those issues’. Similarly, Adams (1997; see also Feld et al., 2014) model voting with a ‘policy salience coefficient’ or ‘weight’. According to this definition, issue salience is inextricably linked to behaviour. In a similar vein, Bartle and Laycock (2012: 680) define issue importance in terms of the effect on one’s vote choice and argue that most important issue (MII) question picks up three components, only one of which is genuinely salience. They label these three components, respectively, as ‘familiarity’, ‘intensity’ and ‘salience’ (with the latter term potentially provoking some confusion). Bartle and Laycock (2012: 681) argue that ‘salience’ ought not be used as a synonym for ‘importance’, which is a function of salience along with the two other components. This strand fails to account for non-voters reporting variations in importance. It also fails to consider that ‘most important issue/problem’ questions refer specifically to the respondent’s life, community or country as the object of the issue, all of which are primarily unrelated to elections.

As the definitions of issue salience have varied, so have the resulting assumptions regarding its properties. For those that have defined issue salience as psychologically exogenous to behaviour, a number of effects have been proposed. These start from the supposition that, rather than individuals being largely unaware of the concept, ‘There is nothing subtle about attitude importance, particularly at its highest levels: People know very well when they are deeply concerned about an attitude, and they know just as well when they have no special concern about one’ (Boninger et al., 1995: 62).

The primary causal mechanism explaining salience’s effects on behaviour is that ‘attaching importance to an issue may activate and engage a person’s emotion systems’ (Miller et al., 2017: 131). Aside from party choice at the ballot box (discussed below), observed resulting behaviours have included additional and selective exposure to information about an issue (e.g. Boninger et al., 1995) and greater consideration of information about that issue and related party positions (Ciuk and Yost, 2016; Fournier et al., 2003; Krosnick, 1990; Walgrave and Lefevere, 2013) because of greater attitudinal accessibility, certainty, stability and influence on less salient attitudes (Boninger et al., 1995; Krosnick, 1988, 1991; Lavine et al., 1996; Miller et al., 2017; Visser et al., 2003). Notably, Boninger et al. (1995) argue that these effects are only observable at the highest level of salience. Comprehensively, Weaver (1991: 68) shows that increased salience of an issue is ‘accompanied by increased knowledge of its possible causes and solutions, stronger opinions, less likelihood of taking a neutral position, and more likelihood of participating in politics through such behavior as signing petitions, voting, attending meetings, and writing letters’. Fox and Schofield (1989) made similar findings regarding protesting. Political attitudes affected by voter issue salience have been shown to be approval and evaluation of policymakers (Edwards et al., 1995; though Grynaviski and Corrigan, 2006, find more mixed evidence) and party support (Neundorf and Adams, 2018). Non-politically, Pedraza et al. (2017) show that Latino US citizens who see immigration as salient are less likely to access healthcare.

Scholars have also found that public issue salience affects policy outcomes (e.g. Dannevig and Hovelsrud, 2016; Franklin and Wlezien, 1997; Lazarus, 2013, though Wlezien, 2005, found no evidence for this regarding defence). Givens and Luedtke (2005) show that the restrictiveness of immigration policy is positively affected by the salience of immigration. Oppermann and Henrike (2008) showed that higher salience of ‘Europe’ in the United Kingdom than in Germany restricted British governmental manoeuvre in European institutions. Finally, Spendzharova and Versluis (2013) find that salience has a positive effect on the transposition of European Union (EU) directives, while Weissert (1991) showed that legislators who introduce bills related to more salient issues see their personal ratings increase.

Until recently one of the key assumed properties of salience was its supposed stability within individuals over time. To Krosnick (1990: 85–86), the distribution of the salience of various issues was expected to be deep-seated within individuals, resistant to change and, possibly, distal in the causal process that leads to behaviour and even spatial attitudes – ‘Reductions in the importance of a policy attitude are probably unusual’.

However, Miller et al. (2017: 150) argue that issue salience with regard to one’s personal life is likely to be more stable than issue salience with regard to polities. Finally, Hatton (2016: 1) finds that ‘at the national level these two dimensions [policy attitudes and salience] of public opinion move differently over time and in response to different macro-level variables’. Both of these latter arguments seem to be well borne out in recent years in Europe – with political attitudes being highly stable, personal salience being fairly stable and salience regarding one’s country and Europe showing high volatility (see Eurobarometer, 2018).

Measurement

Issue salience is most often measured in surveys by asking individuals what they consider to be the MIIs or most important problem(s) (MIP) affecting themselves or their country (or less commonly some other social unit such as community, family or Europe). The ordinal nature of the question – asking for the respondent’s one, two or three MIIs – necessarily means that the salience of issues are typically reported in relative terms to other issues (see Wlezien, 2005). Individuals are less commonly asked to measure their perceived salience of specific issues (e.g. Arceneaux and Kolodny, 2009; Yang, 2016) according to some metric. In spite of the obvious arbitrariness of any such metric, this approach has the advantage of avoiding a cut off by giving a response on the salience of all issues.

Furthermore, Miller et al. (2017: 127) show that the referent object of the survey question – for example, personal or national salience – has significant effects on how salience predicts evaluations of politicians, voting, policy preferences and social participation; ‘different measures produce very different results’. Some surveys, such as the Eurobarometer, have to some extent resolved this issue by including multiple questions that ask respondents what they believe are the MIIs affecting, respectively, (1) themselves personally, (2) their country and (3) Europe. The results of this survey alone show the very wide discrepancy between the issues that individuals report as important to the various referent units. Less significant seems to be the wording choice between ‘issue’ and ‘problem’, with the two co-varying at the aggregate level (Jennings and Wlezien, 2011).

Survey measures of salience were, until recently, less uniform. Boninger et al. (1995: 162) note that they had variously asked ‘people how important an attitude object is to them personally, how deeply they care about it, and how concerned they are about it’. Furthermore, as Bartle and Laycock (2012: 682) point out, the MII question is sometimes worded directly with respect to the election (e.g. ‘As far you’re concerned, what is the single most important issue in the general election?’). This and other differences in ‘MII’ questions, such as the phrasing of ‘issues’ and ‘problems’, adjoining phrases like ‘as far as you’re concerned’, and, especially, the number of issues that are requested by the survey should be considered in comparative studies.

A key debate in the literature during the twentieth century was over the relative merits of close- and open-ended MII questions, with the former supposed by some to be more accurate by ‘properly’ offering parameters of what is meant by salience, while the latter was supposed by others to offer more authentic answers. Geer (1991) offered an overview of this debate before using an experiment to show that open-ended questions do not tap superficial concerns but indeed lead to a genuine expression of salience. Bartle and Laycock (2012; see also Johns, 2010; Wlezien, 2005) argue that individuals are sufficiently bad at knowing themselves that they cannot be relied upon to accurately report salience so that, when responding to MII questions, individuals are actually reporting what they believe the rest of the electorate believe to be important, making MII responses useful only at the aggregate level (for methodological and empirical criticism, see Kurella, 2014).

Overall, though the literature has made significant strides in measuring issue salience via survey data, a number of issues remain. Aside from the consequence of the object of salience, the number of issues that surveys ask for has yet to be properly theoretically considered (particularly considering that communications studies have argued that individuals hold six to nine salient issues at a time; Cronen, 1973). Furthermore, the inherently relative nature of an ordinal approach to measurement necessarily makes isolating the causes and effects of the salience of specific issues difficult. To better understand these dynamics, comparisons between studies using ordinal and absolute metrics of the salience of issues should be carried out in future.

Aside from survey responses, there are two other noteworthy approaches to measuring salience. The second strand of definitions of issue salience – that of a weight on an issue in the voting calculus – sees salience as far simpler to measure, simply using post facto modelling (Adams, 1997; Alvarez and Nagler, 1998; Bartle and Laycock, 2012; Bernstein, 1995). Similarly, issue salience could also be inferred as a revealed preference from behavioural choices in experimental designs (Samuelson, 1938), potentially overcoming some of the shortcomings of survey designs; for example, by asking participants to make vote choices between hypothetical candidates with varying policy preferences or between candidates having been primed with information about recent events that are reliably associated with certain issues’ salience. Future research should consider these approaches. A novel recent contribution to the distinction between salience as a purely psychological or purely behavioural concept, as outlined above, is that of Kiousis et al. (2015: 3351) who state, ‘An important but seldom recognized distinction, though, separates issue salience in general and issue decision salience. Theoretically, an issue might be considered salient without necessarily being important for how people vote or make other decisions’. They accordingly employ survey data that directly ask individuals what the MIIs were in affecting their vote choice.

Finally, Internet search data – notably in the form of Google’s N-Gramme or other search engine results – have been intermittently used, though Mellon (2013, 2014; see also Hayes and Bishin, 2012; Scharkox and Vogelgasang, 2011) compared these findings to those of Gallup’s MIP question, finding congruence in the United States only on some issues, and lower congruence across the board elsewhere. An early work in this vain (Roberts et al., 2002) used online message board mentions as a metric of salience and, later, Vergeer and Franses (2016) measured salience with twitter mentions.

Salience as a Predictor of Electoral Behaviour

As recently as 2003, Fournier et al. (2003: 51) could state that issue salience ‘has received little attention since the publication of Krosnick’s studies’ and that it ‘is rarely integrated into models of political decision’. I proceed to identify four areas in which issue salience is used in electoral studies.

Causes of Variation in Issue Salience

Despite its multiple uses in the election studies literature, the causes of variation in issue salience remain fairly undertheorised and are rarely tested. Krosnick (1990: 74) argued that salience varies between individuals according to three factors: self-interest, social identification and values, whereas it varies across time according to ‘prominent events or problems’ that ‘focus national attention’. Evidence of the predictive power of each of the three between-individual factors was found in five separate studies by Boninger et al. (1995). Neundorf and Adams (2018) also showed that salience was a function of party support.

The main exception to the lack of attention given to the causes of variation in issue salience is that of the agenda-setting literature. As Fournier et al. (2003: 52) state, ‘agenda setting concerns the influence of factors – most notably, but not exclusively, media coverage – on the issues that individuals consider important and unimportant’ (Behr and Iyengar, 1985; Salwen, 1988). A rich literature has gone on to consider when and why agenda setting works, with first memory accessibility and later type of news content shown to arouse levels of emotions that determine salience (Miller, 2007). Arguably in support of the agenda-setting literature is the finding of Einsiedel et al. (1984) that exposure to crime news is a better predictor of salience of crime than is personal experience with crime. Aside from the media as an agenda-setter, Dellis (2009; see also Glazer and Lohmann, 1999) shows that salience is a function of policy-makers’ decisions, which are strategically made to influence salience for electoral reasons, though Hayes (2008) shows that politician cues only have an effect via an interaction with media cues. Finally, Bonnie Meguid’s (2010) work on niche parties argues that their support is subject to the manipulation of the salience of issues on which they campaign (e.g. environmentalism) by the approach of mainstream parties to those issues.

However, attitudinally, panel data studies have shown that media preferences tend to follow consumer preferences, rather than vice versa (Gentzkow and Shapiro, 2010) while experimental studies have shown that the effects of messaging and framing on political attitudes tend to be non-durable (Chong and Druckman, 2007). Moreover, the prevalence of the agenda-setting literature has long led scholars to criticise as exaggerating the role of media (and politicians) in their ability to affect salience (Erbring et al., 1980; Feld et al., 2014). Ansolabehere and Puy (2018: 105) criticise the assumption that ‘the weight given to an issue is entirely a function of media attention to the issue, and not of some underlying preferences of voters’ by academics who ‘use campaign strategies to estimate the emphasis on each political issue’.

Perhaps the most consistent finding regarding the causes of cross-time variation is that issue salience responds to actual events and their gravity. Behr and Iyengar (1985; see also Iyengar, 1979) show that ‘real-world cues’ affect issue salience both via the positive effect they have on news coverage and directly, with Erbring et al. (1980) reporting similar findings for ‘real world conditions’ (see also Wlezien, 2005). For example, Carey et al. (2014) find that social protests increased the salience of immigration among Latinos. Related is the finding of Dannevig and Hovelsrud (2016), that salience is positively affected by the perceived existence of feasible policy solutions regarding the issue.

Conclusion

Despite issue salience’s increasing visibility and some consistently supported effects, it remains curiously tangential in the political science literature. There are important conceptual divergences, an underutilisation of its predictive power and a lack of understanding of its antecedents. From the findings of this review, a number of paths forward become clear.

First, issue salience should be treated as a political attitude like any other, rather than a behavioural weight. Treating it as a weight simply defers the concept (and effects) of attitudinal importance elsewhere and is likely to result in theoretically underspecified models, particularly because, when salience has been assumed to exist exogenously to the electoral process, it has shown to be an effective predictor of a number of attitudes – both political and otherwise. As such, main effects should accompany any interaction effects with policy congruence in spatial models, as already argued by Krosnick.

Second, and contrary to Krosnick’s argument, the assumption that issue salience is stable within individuals must be rethought following substantial volatility in all European countries in the past decade. Third, there should be reconsideration of the theoretical justifications for the number of ‘MIIs’ for which surveys ask, as well as of how the ordinal nature of the question makes isolating the causes and effects of the absolute salience of specific issues more difficult and may have biased results in relevant studies.

Fourth, there should be further research into ‘real world’ effects on salience, which have thus far been strongly empirically supported. Moreover, how to theoretically isolate these effects from less well-supported agenda-setting effects remains unclear. Fifth, the possibility that issue salience at the individual-level has a direct effect on voting behaviour has received some empirical support, yet a causal mechanism remains elusive and may be theoretically implausible.

Finally, the link between the success of certain party families and the salience of certain issues at the aggregate-level should be further investigated. Is the salience of immigration a universal predictor of radical right success? What about Europe, terrorism, crime, unemployment or the economy? What about other party families such as the social democrats and the – at the time of writing – notably low salience of social issues such as health and education, which had been so prevalent across Europe in the late 1990s during the centre-left’s most recent hey-day? By resolving these issues, there may be considerable potential to improve our understanding of political attitudes and political behaviour moreover.