influence party evaluations in the context of the 2011 Swiss parliamentary ... 2012). In this paper, we provide an alternative account of issue voting by ...... on them in their past election and referendum campaigns and which largely âownâ these.
Swiss Political Science Review 20(4): 697–726
doi:10.1111/spsr.12135
The Psychology of Quick and Slow Answers: Issue Importance in the 2011 Swiss Parliamentary Elections1 LIONEL MARQUIS University of Lausanne Abstract: Attitude importance is a key variable in most models seeking to explain evaluations of political parties and leaders. Unlike attitudes on less central issues, attitudes about aspects of politics which citizens deem personally important are expected to influence their political judgments. In this article, we propose to conceptualize the importance of attitudes as their accessibility in memory and measure it by the time survey respondents take to answer an attitude question. We apply this framework to the way in which issue attitudes (opinions on Europe, immigration, etc.) influence party evaluations in the context of the 2011 Swiss parliamentary elections. As expected, quicker responses tend to be associated with stronger relationships between issue opinions and party evaluations. Promises and limits of this approach to conceptualizing attitude importance are discussed. KEYWORDS: latency, attitude importance, issue opinions, elections, Switzerland
Introduction This article deals with the role of issue importance in the formation of party preferences in the context of the 2011 Swiss national elections. In keeping with recent developments in the political behavior and political psychology literature, it is argued that voters tend to like parties which share their positions on political issues that they perceive as salient and worth their consideration. The other way round, voters feel alienated from parties and leaders who disagree with them on issues they deem personally important. Accordingly, the personal importance of issue attitudes moderates their impact on voters’ evaluations of political parties. Our key argument in this article is that the importance of an issue can be measured by the speed with which it comes to mind. The basic idea that issue attitudes can influence party and leader evaluations is anything but new. A wealth of literature and research on issue voting addresses precisely this possibility and largely confirms that, at least under some circumstances, issue attitudes are significant ingredients of voting choices. In this regard, one of the earliest predictions and most robust findings in the issue voting literature has been that voting decisions mostly depend on issues that are personally salient to voters (e.g. Fournier et al. 2003). More recently, this outcome has been combined with issue ownership theory (e.g. Petrocik et al. 1
I would like to thank participants in the workshop “The 2011 Swiss elections” in Solothurn, 25–26 October 2013, as well as three anonymous reviewers for their helpful comments on earlier versions of this paper.
© 2014 Swiss Political Science Association
698
Lionel Marquis
2003), suggesting that the salience of issues enhances their impact on voting, but only (or more strongly) for those parties which are perceived as competent for dealing with the issues at hand and/or are spontaneously associated with them (Walgrave et al. 2012). In this paper, we provide an alternative account of issue voting by analyzing the electoral impact of issue attitudes in terms of their accessibility rather than salience. Although these two attitude dimensions are sometimes used interchangeably, they are conceptually distinct (see below). More importantly for our present purpose, they are assessed by different measurement methods. While the salience dimension is usually tapped by survey questions about the personal importance or priority of social issues, attitude accessibility is measured by the amount of time respondents take to answer an issue opinion question. When measured in the response collection process, these so-called “response latencies” are often considered as yet another type of “paradata” (or more bluntly as “garbage data”) and are simply ignored by most political scientists. In contrast, psychologists and survey methodologists have long pointed out their usefulness in measuring the accessibility or importance of the issues underlying survey responses. Our aim in this contribution is to foster interest in response latencies as a way to improve our understanding of issue voting mechanisms.
Theoretical frame Basic assumptions of our approach to accessibility effects Our model of attitude accessibility effects elaborates on specific assumptions about the nature and organization of attitudes, which have to be made explicit. The first assumption explains why accessibility matters. The second and third assumptions propose a general framework to predict when and how attitude accessibility matters for political judgment and behavior. Assumption 1: Constructivist approach to attitudes Like other scholars, we view attitudes not as fixed entities, but rather as transient, “temporary constructions” (e.g. Wilson and Hodges 1992; Zaller and Feldman 1992; W€ anke et al. 1996; Bassili and Roy 1998). These constructions are derived from whatever thoughts happen to be salient in people’s mind at the moment they are asked for their opinion on some topic. This saliency aspect is important because individuals are “cognitive misers” (Fiske and Taylor 1991) or “satisficers” (Simon 1965). Especially when they have to answer two hundred survey questions in a row, people do not base their judgments on all considerations they can retrieve about an object, but only on those considerations that most easily come to mind. In the present case, this “constructivist” view of attitudes implies that evaluations of political leaders, parties, or issues are often generated on the spot rather than directly retrieved from memory as ready-to-use judgments. In this context, which mental objects (attitudes, beliefs, values, stereotypes, etc.) will become ingredients of evaluations is determined by their relative accessibility, i.e., by the ease and speed with which they can be recovered from long-term memory. Among attitudes with equal relevance to evaluating an object, those with greater accessibility in memory are more likely to be retrieved and integrated into working memory with other currently active thoughts to yield an overall assessment of the object. This view also implies that the construction of an evaluation can be easily affected by situational cues, including preceding items in a questionnaire or the © 2014 Swiss Political Science Association
Swiss Political Science Review (2014) Vol. 20(4): 697–726
The Psychology of Quick and Slow Answers
699
order of response options (e.g. Tourangeau and Rasinski 1988; Wittenbrink et al. 2001), and even by the subjective experience of accessibility (e.g. Tversky and Kahneman 1974; Schwarz 1998). Assumption 2: Spreading activation framework The construction of evaluative responses, as described above, occurs because of the assumed associative structure of human memory (e.g. Anderson 1983; Carlston and Smith 1996). According to this view, knowledge or feeling units (“nodes”) are interwoven by a series of relations (“links”) of a semantic, analogical, affective, contextual, or other nature. When a stimulus (e.g. a survey question) triggers some node, an activation “wave” spreads in the network of associative pathways that connect it to other nodes. Importantly, each episode of such “spreading activation” has the effect of reinforcing utilized pathways, while neglected links and memories continue to lie dormant and decay over time. As a consequence, nodes that happen to possess numerous and strong links with top-of-thehead constructs have a higher likelihood to be retrieved than less accessible nodes.2 Besides, although spreading activation can be partly controlled under some circumstances, most of the time the process is inescapable and unconscious. Assumption 3: The accessibility / availability continuum The above description of attitudes as temporary constructions is overly simplistic. Attitude reports can be roughly classified on a continuum spanning from responses made in the absence of any pre-existing attitude about the question topic (i.e., unavailable attitudes) to responses made on the basis of available and highly accessible attitudes (Higgins 1996; Tourangeau et al. 2000). Unlike in Converse’s (1970) attitude – non-attitude continuum, however, we do not consider attitude reports tending toward the “unavailable attitude” pole as nearly random choices. Rather, such attitude reports are based on the “summing” or “blending” of the first few considerations that are retrieved from memory (e.g. Ajzen and Fishbein 1980; Zaller and Feldman 1992). These considerations include accessible beliefs and attitudes about related objects. In our view, party or leader evaluations of the sort measured in election surveys mostly fall in the category of construed attitude statements for which no readily available attitude exists. For example, when asked “how likely they will ever vote for party X”, many people have never actually pondered this question before, and they have never truly considered X as an option. In such cases, spreading activation may bring to mind a number of salient reasons to like or dislike party X. Among these important reasons are issue attitudes. Attitudes about prominent social problems clearly tend toward the other pole of the continuum, that of available and highly accessible attitudes. To be sure, some issue attitudes are poorly accessible, or have to be constructed altogether in response to stimulus questions. But the common point of many political issues is that people do have available attitudes about them — about Europe, about environment, about immigration, and so forth. Admittedly, reports of issue attitudes and other accessible attitudes are sometimes “edited” to fit the response categories provided by the interviewer or to comply with normative pressures of the response context. But accessible attitudes themselves can be activated automatically by relevant stimuli (Fazio 1990a). Therefore, they are privileged 2
To keep things simple, we refer to memory representations of a party or an issue as “a node”, even though it may be more appropriate to speak of a “node assembly”, i.e., a configuration of several tightly associated beliefs and feelings related to a central concept (e.g. a given party).
© 2014 Swiss Political Science Association
Swiss Political Science Review (2014) Vol. 20(4): 697–726
700
Lionel Marquis
targets of spreading activation from less accessible constructs such as party or leader evaluations. Overall significance of Assumptions 1–3 Taken together, our three assumptions provide a nuanced account of how citizens in modern democracies form political opinions and judgments. It starts from the basic fact that many citizens, having other things to do than inform themselves and think about political issues, do not have pre-existing attitudes about many aspects of politics and are not prepared to answer all sorts of survey questions about their political preferences. This does not mean, however, that citizens have “non-attitudes” on most political matters. Instead, they construct attitudes from whatever considerations happen to be currently salient to them. At the intrapsychic level, such constructive processes are enacted through spreading activation from a concept (say, the concept steered by a survey question) to related information in memory, which is then retrieved and integrated into a summative judgment. Importantly, activation tends to flow toward information which is highly accessible — either temporarily (e.g., because it was prompted by a situational cue) or chronically (e.g., because it is frequently encountered in everyday discussions and media exposure). Therefore, when a person lacks a true attitude about some object, the accessibility of related attitudes is a key dimension to predict which one(s) of these “competing” attitudes will be relied upon to evaluate that object.
Attitude accessibility and judgments of political parties According to a prominent view in attitude research, the accessibility of an attitude can be conceived as the strength of the link between the attitude object and its evaluation in memory (Fazio 1990a, 2007). Highly accessible attitudes are automatically activated upon mere exposure to (or evocation of) the relevant attitude object. In turn, automatically activated attitudes are particularly likely to influence judgments and behaviors in a relatively direct manner, especially when individuals lack the motivation or opportunity to engage in more careful consideration of alternative courses of action (as should be the case in a typical survey situation). However, the effects of automatic attitude activation are very short-lived and thus unable to explain how a currently active evaluation may influence responses to another question asked several minutes later. Therefore, the concept of attitude importance — “the extent to which an individual cares deeply about and is personally invested in an attitude” (Krosnick et al. 1993: 1132) — has been proposed to explain why some attitudes may be influential regardless of the immediate response context. As it turns out, measures of attitude importance and attitude accessibility are highly correlated (Lavine et al. 1996), and attitudes that are both important and accessible have most influence on candidate preferences (e.g. Krosnick 1988, 1989). Although the strict theoretical and empirical equivalence of attitude accessibility and attitude importance has been disputed (e.g. Miller and Peterson 2004), both constructs may be viewed as subcomponents of a general “attitude strength” dimension (Krosnick et al. 1993; Petty and Krosnick 1995; Bizer and Krosnick 2001). In most accounts, attitude importance is thought to be a major cause of attitude accessibility (Boninger et al. 1995a, 1995b; Bizer and Krosnick 2001). Therefore, accessibility is at least an acceptable proxy for attitude importance, and it is a good predictor of the attitude’s association with related attitude objects. To sum up, highly accessible attitudes (1) tend to be automatically activated upon encounter with their object, (2) are personally important, and (3) are likely to guide © 2014 Swiss Political Science Association
Swiss Political Science Review (2014) Vol. 20(4): 697–726
© 2014 Swiss Political Science Association
(‡)
S.E.
Sig.
3.51 (p < .0001)
B
2.90 (p < .001)
Sig. .048 3705 7.50 (p < .0001)
S.E.
CVP
.188 3694 30.53 (p < .0001)
B
FDP
.486 3726
S.E.
Sig.
122.39 (p < .0001) 4.60 (p < .0001)
B
SVP
S.E.
Sig.
2.55 (p < .01)
.456 3716 108.55 (p < .0001)
B
SPS B
S.E.
GPS Sig.
2.78 (p < .001)
.398 3725 85.82 (p < .0001)
Note: ***p < .001; **p < .01; *p < .05; †p < .10. aReference categories are Male, Rural, and French-/Italian-Speaking. bZ-score.
Adj. R2 N F test (whole model) F test for latency block
Variable
Table 2: Continued
The Psychology of Quick and Slow Answers 711
Swiss Political Science Review (2014) Vol. 20(4): 697–726
702
Lionel Marquis
choices is a long-term affair. Attitudes about issues that already enjoy wide public recognition at the outset of a political campaign are likely to remain accessible. Conversely, it may take a good deal of information to build up important and accessible attitudes “from scratch”. Beyond their effects on attitude accessibility, political campaigns bring about “cognitive bonding”, especially among citizens who do not care much about politics (Sch€ onbach and Weaver 1985; R€ ossler and Schenk 2000). In Converse’s (1964) terms, campaigns enable citizens to ascertain “what goes with what”, by allowing them to learn where parties and candidates stand on the issues (Lenz 2009). Campaigns thus create the necessary (but not sufficient) conceptual links for some people to judge parties in line with their political preferences and to vote for them accordingly. Without these links in memory, spreading activation sent out by evaluation questions would not reach issue attitudes at all, and attitude accessibility would not make any difference. To sum up, our basic hypothesis predicts an “Accessibility Effect”, whereby quicker answers to issue opinion questions are reflective of more accessible issue attitudes and have a stronger influence on party evaluations. As a corollary, we hypothesize an “Information Effect”, whereby (i) an issue that has not been linked to a specific party in public discourse does not matter for the evaluation of this party, and (ii) issues that have been a long time on the political and media agenda wield greater influence on party evaluations than issues commanding only temporary attention and accessibility.
Measurements Party evaluations Based on the 2011 Selects Post-Electoral Survey, party evaluations are measured by the socalled party “voting propensities” tapped by the following question: “Could you please tell me how likely it is that you will ever vote for this party?” The 11-point scale ranges from 0 (“very low probability”) to 10 (“very high probability”). Voting propensities for the five main Swiss parties are included in our analysis: the Liberal-Radical Party (FDP), the Christian Democratic Party (CVP), the Swiss People’s Party (SVP), the Social Democratic Party (SPS) and the Green Party (GPS). These are the largest and best known Swiss parties, with which most voters have a very long experience. Hence, even though many voters probably do not know the minutiae of each party’s positions on every specific issue, they probably do have some basis for inferring the parties’ stands on the most important issue dimensions. It then makes sense to assume that voting propensities are momentary constructions based on the most accessible facts about the respective parties, including their positions on issues.
Issue opinions As shown in Table 1, seven issue opinion questions with a clear semantic differential were available from the Selects survey.3 Three issue opinions (social expenses, nuclear energy,
3
In the case of the immigration issue, opinions were measured on the basis of six specific items (e.g. “Migrants exacerbate job market situation”, “Swiss culture vanishes due to immigration”). A reliability analysis of these items indicated a satisfactory internal consistency of the scale (a > 0.8; see Appendix).
© 2014 Swiss Political Science Association
Swiss Political Science Review (2014) Vol. 20(4): 697–726
722
Lionel Marquis
Grant, J.T., S. Mockabee and J. Monson (2010). Campaign Effects on the Accessibility of Party Identification. Political Research Quarterly 63(4): 811–821. Hastie, R. and B. Park (1986). The Relationship Between Memory and Judgment Depends on Whether the Judgment Task is Memory-Based or On-Line. Psychological Review 93(3): 258–268. Higgins, E.T. (1996). Knowledge Activation: Accessibility, Applicability, and Salience. In Higgins, E.T. and A. Kruglanski (eds.), Social Psychology. Handbook of Basic Principles. New York: The Guilford Press (133–168). Holbrook, A. and J. Krosnick (2010). Operative and Meta-Attitudinal Manifestations of Attitude Accessibility: Two Different Constructs, Not Two Measures of the Same Construct. In Forgas, J., et al. (eds.), The Psychology of Attitudes and Attitude Change. Florence (KY): Psychology Press (109–124). Holbrook, A., M. Berent, J. Krosnick, P. Visser and D. Boninger (2005). Attitude Importance and the Accumulation of Attitude-Relevant Knowledge in Memory. Journal of Personality and Social Psychology 88(5): 749–769. Holbrook, R.A. and T. Hill (2005). Agenda-Setting and Priming in Prime Time Television: Crime Dramas as Political Cues. Political Communication 22: 277–295. Holtgraves, T. (2004). Social Desirability and Self-Reports: Testing Models of Socially Desirable Responding. Personality and Social Psychology Bulletin 30(2): 161–172. Huckfeldt, R., J. Levine, W. Morgan and J. Sprague (1999). Accessibility and the Political Utility of Partisan and Ideological Orientations. American Journal of Political Science 43(3): 888–911. Johnson, M. (2004). Timepieces: Components of Survey Question Response Latencies. Political Psychology 25(5): 679–702. Karpinski, A., R. Steinman and J. Hilton (2005). Attitude Importance as a Moderator of the Relationship Between Implicit and Explicit Attitude Measures. Personality and Social Psychology Bulletin 31(7): 949–962. Krosnick, J. (1988). The Role of Attitude Importance in Social Evaluation: A Study of Policy Preferences, Presidential Candidate Evaluations, and Voting Behavior. Journal of Personality and Social Psychology 55(2): 196–210. - (1989). Attitude Importance and Attitude Accessibility. Personality and Social Psychology Bulletin 15(3): 297–308. Krosnick, J., D. Boninger, Y. Chuang, M. Berent and C. Carnot (1993). Attitude Strength: One Construct or Many Related Constructs? Journal of Personality and Social Psychology 65(6): 1132– 1151. Lavine, H., J. Sullivan, E. Borgida and C. Thomsen (1996). The Relationship of National and Personal Issue Salience to Attitude Accessibility on Foreign and Domestic Policy Issues. Political Psychology 17(2): 293–316. Lenz, G. (2009). Learning and Opinion Change, Not Priming: Reconsidering the Priming Hypothesis. American Journal of Political Science 53(4): 821–837. Lodge, M., M. Steenbergen and S. Brau (1995). The Responsive Voter: Campaign Information and the Dynamics of Candidate Evaluation. American Political Science Review 89(2): 309–326. Mayerl, J. (2013). Response Latency Measurement in Surveys. Detecting Strong Attitudes and Response Effects. Survey Methods: Insights from the Field. Online: http://surveyinsights.org/? p=1063 [accessed 20.12.2013]. Mayerl, J. and D. Urban (2008). Antwortreaktionszeiten in Survey-Analysen. Messung, Auswertung und Anwendungen. Wiesbaden: VS Verlag. McNamara, T. (2005). Semantic Priming: Perspectives from Memory and Word Recognition. New York: Psychology Press.
© 2014 Swiss Political Science Association
Swiss Political Science Review (2014) Vol. 20(4): 697–726
The Psychology of Quick and Slow Answers
723
Merrill, S. and B. Grofman (1999). A Unified Theory of Voting: Directional and Proximity Spatial Models. Cambridge: Cambridge University Press. Meyer, M. and H. Schoen (2014). Response Latencies and Attitude-Behavior Consistency in a Direct Democratic Setting: Evidence from a Subnational Referendum in Germany. Political Psychology 35(3): 431–440. Miller, J. and D. Peterson (2004). Theoretical and Empirical Implications of Attitude Strength. Journal of Politics 66(3): 847–867. Mulligan, K., J.T. Grant, S. Mockabee and J.Q. Monson (2003). Response Latency Methodology for Survey Research: Measurement and Modeling Strategies. Political Analysis 11(3): 289–301. Nimon, K., M. Lewis, R. Kane and R.M. Haynes (2008). An R package to compute commonality coefficients in the multiple regression case: An introduction to the package and a practical example. Behavior Research Methods 40(2): 457–466. Petrocik, J., W. Benoit and G. Hansen (2003). Issue Ownership and Presidential Campaigning, 19522000. Political Science Quarterly 118(4): 599–626. Petty, R. and J. Krosnick (1995), (eds.). Attitude Strength: Antecedents and Consequences. Mahwah (NJ): Lawrence Erlbaum. Price, V. and D. Tewksbury (1997). News Values and Public Opinion: A Theoretical Account of Media Priming and Framing. In Barnett, G. and F. Boster (eds.), Progress in Communication Sciences, Vol. 13. London: Ablex Publishing (173–212). Roskos-Ewoldsen, D., B. Roskos-Ewoldsen and F.D. Carpentier (2009). Media Priming: An Updated Synthesis. In Bryant, J. and M. Oliver (eds.), Media Effects. Advances in Theory and Research (3rd ed.). New York: Routledge (74–93). R€ ossler, P. and M. Schenk (2000). Cognitive Bonding and the German Reunification: AgendaSetting and Persuasion Effects of Mass Media. International Journal of Public Opinion Research 12(1): 29–47. Schoenbach, K. and D. Weaver (1985). Finding the Unexpected: Cognitive Bonding in a Political Campaign. In Kraus, S. and R. Perloff (eds.), Mass Media and Political Thought. An InformationProcessing Approach. Beverly Hills: Sage (157–176). Schwarz, N. (1998). Accessible Content and Accessibility Experiences: The Interplay of Declarative and Experiential Information in Judgment. Personality and Social Psychology Review 2(2): 87–99. Simon, H. (1965). The Logic of Rational Decision. British Journal for the Philosophy of Science 16 (63): 169–186. Srull, T. and R. Wyer (1980). Category Accessibility and Social Perception: Some Implications for the Study of Person Memory and Interpersonal Judgments. Journal of Personality and Social Psychology 38: 841–856. Tourangeau, R. and K. Rasinski (1988). Cognitive Processes Underlying Context Effects in Attitude Measurement. Psychological Bulletin 103(3): 299–314. Tourangeau, R., L. Rips and K. Rasinski (2000). The Psychology of Survey Response. Cambridge: Cambridge University Press. Tversky, A. and D. Kahneman (1974). Judgment under Uncertainty: Heuristics and Biases. Science 185(4157): 1124–1131. Visser, P., J. Krosnick and J. Simmons (2003). Distinguishing the cognitive and behavioral consequences of attitude importance and certainty: A new approach to testing the common-factor hypothesis. Journal of Experimental Social Psychology 39: 118–141. Walgrave, S., J. Lefevere and A. Tresch (2012). The Associative Dimension of Issue Ownership. Public Opinion Quarterly 76(4): 771–782. W€ anke, M., H. Bless and B. Biller (1996). Subjective Experience Versus Content of Information in the Construction of Attitude Judgments. Personality and Social Psychology Bulletin 22(11): 1105–1113. © 2014 Swiss Political Science Association
Swiss Political Science Review (2014) Vol. 20(4): 697–726
The Psychology of Quick and Slow Answers
705
Mayerl 2013; Meyer and Schoen 2014).6 Unless these two components are distinguished, response latencies may simply reflect a continuum of capacity to provide quick answers and may be little more than surrogate measures of political expertise and other background variables. Hence, attitude accessibility is best assessed by comparing every respondent’s response latencies to her own response time standard. Incidentally, this procedure also allows controlling for differences in latencies that are due to varying interviewers’ average speed for reading the questions (Mulligan et al. 2003). Although the proposed method may not be appropriate in all situations (e.g. Bassili and Krosnick 2000: 113), we believe it does make sense in the present case, since the speed with which respondents indicate their positions on issue opinion scales is expected to be highly dependent on some general “know-how” factor. Accordingly, we simply subtract each individual’s “baseline speed” (BS) from the latencies of her responses to each specific question. BS is computed as the average of (outlier-corrected) latency scores for the seven opinion variables and two similar variables.7 The overall stepwise procedure for measuring “net” response latencies from raw, uncorrected scores is illustrated in Figure 1. At this stage, some readers may wonder whether and why “baseline speed” (BS) really matters for the measurement of attitude accessibility. To show why it is necessary to incorporate a baseline in the computation of latency scores, let us see what this baseline looks like in the first place. In particular, are there systematic differences in BS between certain categories of respondents? And if such were the case, to what extent are these differences not reducible to genuine variations in the underlying accessibility of concepts? The answers to these questions are quite straightforward. First, the linguistic area from which respondents were drawn is clearly related to aggregate variations in BS. Germanspeaking respondents were about 1.5 seconds faster than French-speaking respondents to answer the issue opinion questions, and about 2.5 seconds faster than Italian speakers (both ps