ESRA 2019 Draft Programme at a Glance

Assessing Implicit Attitudes Using General Population Surveys

Session Organisers Professor Elmar Schlueter (Justus-Liebig-University of Giessen (Germany))
Professor Jochen Mayerl (Technische Universität Kaiserslautern)
TimeFriday 19th July, 13:00 - 14:00
Room D31

Survey researchers typically assess attitudes, broadly defined here as associations between concepts and evaluations, using self-report measures. For example, respondents might be asked to explicitly self-assess their views or feelings toward different immigrant groups on a response scale from 1 (dislike this group very much) to 7 (like this group very much). Alternative to such a direct measurement approach, respondents’ attitudes might also be assessed indirectly using implicit measures. Here, one line of research originating in psychology uses respondents’ performance (e.g. response latencies or categorization errors) on behavioral tasks designed to infer the construct of interest rather to rely on self-reports. Recent years have witnessed an unprecedented proliferation of methods seeking to assess such implicit attitudes, including but not limited to measurement approaches like the implicit association test (Nosek, 2007) or the affective misattribution procedure (Payne et al., 2005). Relying on relatively small, non-representative samples, such methods are routinely employed in psychological research. However, the feasibility and utility of using implicit measures in large-scale survey research based on representative population samples is less clear. This session invites papers that focus on methodological as well as substantive issues arising from coupling measures of implicit attitudes with large-scale population surveys. Contributions may cover but are not limited to the following research topics:

• Innovative measurement approaches of implicit attitudes
• Effects of survey mode and/or sampling methodology on outcomes
• Applications with a substantial focus (e.g. research on intergroup relations or sensitive topics more generally)

Please send your paper proposals (no more than 500 words in length) to:
Prof. Dr. Elmar Schlüter,
Prof. Dr. Jochen Mayerl,

Keywords: Implicit Attitude measurement, Population Surveys, Innovations, survey mode, sensitive topics

Success or Failure? Researchers' Implicit Associations with Academic Misconduct and with Questionable Research Practices

Dr David Johann (University of Zurich)
Mr Justus Rathmann (University of Zurich) - Presenting Author
Professor Heiko Rauhut (University of Zurich)
Dr Colin T. Smith (University of Florida)

Research deploying Implicit Associations Tests (IAT) has become increasingly popular in the Social Sciences. This is not surprising as such indirect measures promise to circumvent measurement problems caused by social desirability bias, among other things. However, to date, there has not been an application of IAT in Science Studies, although this subdomain of the Social Sciences deals frequently with norm violations, such as academic misconduct and other questionable research practices, that are highly sensitive and susceptible to biased reporting.
Drawing on a survey among researchers at German universities which includes two Single-Category Implicit Association Tests (SC-IAT) we investigate if researchers link implicit associations with academic misconduct (such as data falsification, data fabrication and plagiarism) and questionable research practices (such as gift authorship and self-plagiarism) to academic success or failure. To be more precise, using the SC-IAT we examine whether researchers rather associate academic and professional success than academic and professional failure with "academic misconduct" or "questionable research practices", respectively. The aim of the study is twofold: (a) We aim to learn more about researchers' propensity to engage in questionable behaviour and (b) we intend to identify factors that are related to academic misconduct and questionable research practices.

Application of the Affective Misattribution Procedure in a Large-Scale Survey to Assess Prejudiced Attitudes

Professor Elmar Schlüter (Justus-Liebig-University of Giessen) - Presenting Author
Professor Jochen Mayerl (Chemnitz University of Technology)
Mr Henrik Andersen (Chemnitz University of Technology)

Conducting surveys to assess prejudiced attitudes are typically faced with the problem of social desirability (SD) bias. For decades, survey researchers have tried to assess the ”true” attitudes underlying explicit survey statements which are potentially consciously biased towards social desirability. Different survey methods like scales intended to measure an individual’s need for social approval or trait desirability as well as experimental approaches including Randomized Response Technique or Item Count Technique were developed and tested in survey research. Further, survey researchers have looked at the response latencies of the explicit statements in surveys to get a deeper understanding of the answering process. So far, however, the problem of SD bias still persists. Another stream of research was established for lab experiments in the field of psychology to assess implicit attitudes, i.e. attitude statements that are not under conscious control of the respondents. Two prominent approaches in this field of research are the Affective Misattribution Procedure (AMP) and the Implicit Association Test.
In our study, we adopt the AMP in a large-scale web survey to assess prejudiced attitudes towards various minority- and outgroups (e.g. women, Muslims, homosexuals, Jews, foreigners, refugees, disabled persons and overweight persons). The AMP presents respondents with a prime stimulus (the actual attitudinal object of interest) followed by an ambiguous ‘target’ item (e.g. a Chinese pictograph or abstract painting) that the respondent is then asked to evaluate. Respondents are expressly instructed to ignore the prime. The AMP works because respondents often cannot disentangle affective responses towards prime and target thereby allowing insight into implicit attitudes.
We compare the findings of the AMP with results of more classic survey approaches, i.e. explicit attitude statements after correction for SD scales, response latency analysis and item count technique.

Are immigrants perceived as a security risk?

Professor Alexandra Nonnenmacher (University of siegen) - Presenting Author
Miss Sarah Gierse (University of Siegen)

The general idea of experimental procedures like the IAT or AMP is to reveal attitudes that are unconscious or even unwanted in the sense that they are inconsistent with social desirability, such as negative or discriminating attitudes towards disabled persons or immigrants. In our research, we apply this idea by using portrait photos of representatives of different ethnic groups to study effects on affective and cognitive components of attitudes towards personal security.
In a pilot survey (n=416), we conducted a web-survey which included a tableau with 30 portrait photos, 20 percent of which were representatives of ethnic groups which are typically perceived as immigrants to Germany. In three different versions of the tableau, the „immigrants“ were arranged as a discernible group in the centre, in a corner, or spread unsystematically over the whole page. Following the photographs, we included questions measuring fear of crime (as an affective component of attitudes towards immigrants) or the approval of political measures intended to enhance personal security, for example expanding video surveillance in public places (the cognitive component). We expected the centre version to have the strongest effects, but contrary to that respondents who were shown immigrants spread unsystematically over the page showed significantly more fear of crime; approval of political measures was not affected differently by the three versions. These results indicate that a) implicit measures are better suited to capture the affective component of attitudes, rather than the cognitive component, and b) immigrants may be perceived more as a threat if they are „hidden“ in the crowd rather than standing out as a discernible group. In the following, we intend to applicate the online-questionnaire on a larger population sample to replicate our results and test for differential effects, for example differences between men and women as well as between respondents from different ethnic groups.

Measuring the European political competition through machine learning techniques in Voting Advice Applications.

Mr Javier Padilla Moreno-Torres (LSE)
Mr Guillermo Romero Moreno (University of Southampton )
Mr Jesus Enrique Chueca Montuenga (King's College London) - Presenting Author

In this paper, we discuss how Voting Advice Applications (VAAs) can use machine learning techniques to capture the conceptual political space of the users and explore the ideological distances between voters and political parties in high and low dimensional political landscapes. We present a learning method which can predict the voting behavior of the users and situate them into the European political space thanks to the use of a learning algorithm. Moreover, we test our method in two VAAs launched for the European Parliamentary Election, the EU-Vox 2014 and the EU-Vox 2019, and discuss the methodological and normative assumptions that we face. To capture the political landscape, we combine political affiliation with the positions of the users and the political parties in the most prominent policy issues. Thanks to this approach, it can be studied whether the populist parties which became prominent in these years are situating themselves ideologically in the traditional political dimensions (mainly Economic and Social dimension) or are changing the very nature of the political landscape by creating new dimensions and cleavages of political competition. Our cross-national and cross-age approach allows us to measure whether there is a common pattern of political dimension in different countries and segment of population. Finally, we discuss several methodological and normative challenges of our proposal, and how Voting Advice Applications using machine learning techniques could be improved to capture the political landscape by combining deductive and inductive approaches.