Abstract
Mixed worded scales contain both positively and negatively worded items. These are introduced in some questionnaires to reduce response style, which is a tendency to answer items regardless of their content. When a respondent fails to switch their response on facing a negatively worded item, the response is inconsistent. By analyzing data of 15-year-old students from 76 education systems that participated in PISA 2018, I investigate whether reading literacy or language spoken at home are associated with the probability of being an inconsistent responder. I expected to find that a respondent is more likely to respond inconsistently if they speak a different language at home rather than the language of the assessment. I also expected that students with higher reading literacy are less likely to respond inconsistently. I found that the proportion of inconsistent responders ranges from 9% to 33% across the education systems with an average of 17%. Using separate logistic regressions for each of the education systems with both the predictors in the model, I found that in 45 out of 76 education systems, as expected, reading literacy showed a negative association with inconsistent response. In 5 education systems, as expected, an inconsistent response is found to be more probable when a different language is used at home. However, contrary to expectation, in 7 education systems the opposite was found. I conclude by discussing limitations regarding the generalizability of the study and discussing how future research can go deeper into understanding how language spoken at home and reading literacy associate with inconsistent responses.