Contextual analyses have received increasing interest as a way to understand electoral behaviors during the last decade. However, the growing interest in contextual analysis among political scientists is now limited by conceptual and methodological difficulties related to scale. Indeed, we show here that the scale we use might significantly change the results we get. Our central claim is therefore to plead in favor of a more careful reflection on scale in contextual analysis. This reflection on the impact of scale is made by applying two different methods to explain extreme right voting and ethnocentric attitudes. In the first, we define the context in which individuals are embedded through circles around their home while in the second, we try to capture the real environment of individuals by defining their proximity area, living pool, and employment pool. When applying contextual variables (share of migrants and economic context) to explain extreme right voting and ethnocentric attitudes, the intermediate scale – defined here as the municipality or radius of 4 to 8 km around one’s residence – appears to be the most influential in both approaches compared either to closer neighborhood or larger levels. Our analysis also shows an interesting differential for the effect of contextual variables on ethnocentric attitudes and on extreme right voting, especially concerning the impact of urban versus rural context.