Income comparisons are often performed through the construction of reference groups. These groups are highly dependent on the sociodemographic characteristics collected by survey data. Gender is usually included in these characteristics only when the number of cases is large enough to have separate samples for women and men. So far, there has been no empirical proof on the fact that comparisons are within or between people of the same sex. With the support of specific questions collected in three waves of the pretest of the German Socio-Economic Panel, this study analyses income comparisons within and between gender groups. Results suggest that income comparisons are mainly within people of the same sex. On average, women compare more than men independently from the gender composition of the sector of employment. Despite the predominance of within-gender comparisons, between-gender comparisons exist. Regressions that test the effect of income comparisons and reference groups on subjective well-being are indeed better explained when gender is not included as a characteristic for the construction of reference groups.