ABSTRACT
This study explores the acquiescent response style (ARS) among respondents in the Czech Republic. To analyse ARS, confirmatory factor analysis (CFA) was employed and the response style (RS) was modelled as a latent variable. The RS factor in the CFA model must be validated by its relationship to education and age, i.e. proxies of cognitive ability. The two studies presented in this article use large amount of data as all available balanced batteries of items were analysed. In Study 1, the RS factor showed no correlation with age/education in most of the models. Study 2 employed additional measures of cognitive ability and education and confirmed the overall absence of the relationship. Author concludes that the modelled RS factor then cannot be considered the ARS and deduces that the most likely explanation of the identified systematic error variance is the respondents’ carelessness manifested by automatically choosing the extreme agreement, or choosing the first presented answer as a result of memory effect.
Disclosure statement
No potential conflict of interest was reported by the author.
Supplementary material
Supplemental data for this article can be accessed here.
Notes
1. B. Weijters et al. (Citation2013) present an overview of CFA models designed to capture method bias in a balanced battery of Likert type items. They regard J. Billiet and McClendon’s (Citation2000) CFA model to be a special case of the random intercept model proposed by Maydeu-Olivares and Coffman (Citation2006).
2. The non-standardised variance of latent variables is compared.
3. The sum of a respondent’s agreeing responses to all the items in the battery, resp. the items that were entered into the model.
4. The CF-RS model with education as a latent variable and the sumagree index as a latent variable is specified accordingly.
5. The list of these batteries is presented in Appendix 2.
6. Our Society is a continuous sample survey fielded by the Public Opinion Research Centre and conducted 10 times a year, N = approx. 1,000 respondents, face-to-face interviews, the population of the Czech Republic over the age of 15, using quota sampling (characteristics: region (NUTS 3), size of the place of residence, sex, age, education).
7. Data were also sought from the database of the European Social Survey (ESS), but the questionnaires used in these surveys did not contain suitable batteries; the ESS usually measures one-dimensional constructs with fewer than four items and/or the batteries are not balanced.
8. Public Opinion Research Centre, Institute of Sociology, Czech Academy of Sciences; https://cvvm.soc.cas.cz/en/.
9. Czech Social Science Data Archive, Institute of Sociology, Czech Academy of Sciences; http://archiv.soc.cas.cz/en.
10. In some cases, we allowed this correlation as it was a solution in situations when the model did not converge.
11. Heywood cases are negative estimates of variances or correlation estimates greater than ±1 (Revilla & Saris, Citation2013).
12. The first number after the ‘OS’ represents the year and the second number in brackets is the month in which the survey was fielded; for example, survey OS 2009 (3) was conducted in March 2009.
13. In Study 1, the RS factor in the Political Efficacy battery was not found to correlate with education, but a correlation was observed with age.
14. The Death Penalty battery was used in Study 1, it was measured on a four-point scale and it didn’t show any correlation with education.
Additional information
Funding
Notes on contributors
Johana Chylíková
Johana Chylíková is a post-doc researcher at the Czech Social Science Data Archive, a department of the Institute of Sociology of the Czech Academy of Sciences. Her research and academic interests are survey methodology and measurement errors, she is also interested in popularization of social sciences and humanities. She teaches survey methodology and statistics at the Faculty of Social Sciences, Charles University in Prague.