Norming studies of gender ratios (e.g., Misersky et al. 2014) are frequently used for stimuli selection in psycholinguistic studies (Kim et al. 2024). We examine the stability of stereotype ratings of activities in German (e.g., Zug fahren 'take the train') with two unipolar binary gender scales (Sato et al. 2025) to test their interdependence. 306 German speakers estimated the proportion of women (version 1) or men (version 2) engaging in each of 96 activities on a 0-100% scale (between-subject). Version 2 was recoded (100−estimate) to quantify measurement inconsistency for each activity using a paired Z-test between the mean proportions of women of the two versions. Crucially, even though estimates were highly correlated (r = 0.92), only 29 activities showed no significant differences between the two versions (|Z| ≤ 1.96), confirming psychometric stability for merely ~30% of the items (r = 0.99). For highly stereotyped activities (e.g., female: Harfe spielen 'play the harp'/male: Pfeife rauchen 'smoke a pipe'), the ratings demonstrate a clear elicitation bias caused by gender group salience: Minority-focused questions enhanced the stereotype, while majority-focused questions led to stereotype attenuation. While the latter might be due to social desirability, the former could reflect a socially motivated confirmation bias, where participants maximize the estimated underrepresentation of the minority group to conform with perceived inequality. For neutral activities, the recoded female proportion was significantly lower than the direct proportion. This study not only provides a stable subset of norms, but also a robust methodological framework for assessing stereotype stability.