In Experimental Economics, coordination games are used to elicit social norms as incentivized beliefs about others’ beliefs. Conversely, representative surveys like the World Values Survey elicit social norms as personal attitudes and values that are independent of others’ beliefs. Using a representative survey of the Italian population (N = 1, 501), we compare the two ways of measuring social norms with gender roles as a working example and find the following results. At the aggregated level, appropriateness ratings obtained under the two elicitation methods follow the same pattern but differ significantly in magnitude, with the incentivized social norm elicitation depicting a more conservative view on gender roles than the unincentivized one. The analysis carried out at the individual level allows us to explain the previous result. Most respondents report personal values as more progressive than the perceived norm, which may be consistent with a desirability and/or a self-image bias. This occurs irrespectively of whether respondents correctly perceive the social norm or not. We conclude that analyses based on personal values lead to a proxy of gender norms significantly more progressive than the norms elicited in coordination games.