Openly worded prejudice items measure bias well only while prejudice stays socially acceptable.

Evidence

One good starting point for further discussion and testing is the Islamophobia Scale, developed by Lee et al. (2009). These researchers used a 5-point Likert-type scale from “strongly disagree” to “strongly agree” on which 223 undergraduate students responded to questions such as “If I could, I would avoid contact with Muslims,” and “Islam is a dangerous religion.” These types of questions may have diminished utility if expressing open Islamophobia becomes less socially acceptable, but in the current historical context, they are direct measures with a high probability of accurately revealing Islamophobia. — Erik Bleich, Defining and Researching Islamophobia, 53

Categorization

• SEED: Article angle: how the social desirability ceiling silently retires survey instruments, making prejudice look like it vanished when it only went underground.
• XREF: Connects to the broader measurement problem in studying racism and antisemitism after WWII, where overt-attitude scales lost validity.