Measurement errors in survey questions: Explaining response variability