Accounting for ceiling effects in gender equality endorsement. A zero inflated modeling approach.

Gender equality endorsement is an intergroup measure present in various survey-based studies and is a prominent indicator among the Sustainable Developmental Goals (SDG) (Sandoval-Hernández et al., 2020). To this end, countries can rely on the gender equality endorsement scale included in the International Civic and Citizenship Study (ICCS), which provides probabilistic samples of 8th-grade students from different countries and assesses gender equality endorsement between men and women. Traditional methods for generating scores with this scale rely on the partial credit model (PCM), a response model that utilizes a normally distributed latent variable to represent students' propensity to respond to the various items included in the instrument. Moreover, researchers rely on regression models to address research questions about related factors and the effects of program evaluation. However, the scale scores of this instrument are highly skewed. This skewness is desirable. It means a noticeable portion of students endorse gender equality at the scale ceiling. Nevertheless, traditional regression models may produce distorted estimates in the presence of ceiling effects on the total scores. We propose a method that relies on the monotonicity property of the PCM scores and creates a reverse sum score. We use zero-inflated models to separate ceiling cases from the rest of the scores, allowing us to make inferences on both sides: the students at the ceiling and those in the remainder of the distribution. This method is a helpful tool for program evaluations dealing with ceiling effects in their attribute of interest.