Adjusting Response Style in Work Engagement Scores in Japan ―Using the Anchoring Vignette Method and a Multidimensional IRT Model―

This study aimed to quantitatively examine the extent to which response style (RS) — a bias arising at the time of responding and previously discussed as a possible source of cultural bias — contributes to the internationally low work engagement (WE) scores of Japanese workers. A questionnaire survey was conducted with 442 regular employees recruited through an internet survey company. The survey included three items from the Ultra-Short Measure of the Utrecht Work Engagement Scale (UWES) and six anchoring vignette items. Bayesian analysis using a multidimensional item response theory model combined with the anchoring vignette method was employed to estimate a model controlling for respondents' RS. The results showed that, although model comparison suggested the presence of RS effects, no RS of meaningful magnitude was identified for nearly all respondents, and RS did not substantively affect their WE scores. Furthermore, the correction effect from controlling for RS was extremely small at 0.013 points, representing only 1.3% of the approximately one-point difference reported in previous international comparisons. The effect size (Cohen's d = 0.010) was also negligible. These findings suggest that the low WE scores in Japan are difficult to explain by RS, and that other potential influences — including biases that depend on item content, as well as job characteristics and the work environment — warrant further examination.