In a recent paper published in Addiction, we documented that drinking among Swedish 15–16-year-olds halved during the period 2000–12 [1]. Our findings further suggested that this reduction was observed in all drinking groups, from light to heavy consumers, supporting the collectivity of drinking theory [2]. As noted by Hallgren [3], this outcome counters the polarization hypothesis that received support in the study he and his co-workers published in 2012 [4]. According to their results (focusing Stockholm youth between 2000 and 2010), the top 5% increased their consumption, whereas drinking decreased in the remaining part of the distribution. In our discussion, we attributed this discrepancy in findings to the difference in analytical approach. We focused on the consumption trends in various drinking groups over the whole study period, whereas Hallgren et al. looked at the change between two selected years, 2000 and 2010. In our data, too, there was a consumption increase among the top 5% between selected years simultaneous with a decrease in overall consumption, but these selected observations were rather exceptions from the global pattern of collectivity. However, in his letter, Hallgren argues that although the collectivity of drinking trajectories that we found may be valid at national level, the growing availability of alcohol in Stockholm makes this city an exception, and hence polarization a plausible phenomenon. Although we do not wholly agree with the analytical approach applied by Hallgren et al., we find their notion of the polarization hypothesis novel, well argued and important, and thus worthy of further testing. More specifically: does the discrepancy in our findings depend upon methodological differences, or to differences between Stockholm and the remaining part of the country?

We do not have the data necessary to provide a conclusive answer to this question, but we can glean some hints by analysing the subsample of our data living in the county of Stockholm. Information on place of residence was missing for 2000, so our data span the period 2001–12, comprising repeated cross-sections of about 1000 15–16-year-olds (with an approximately equal number of boys and girls) in each wave. We applied the same scheme as in our analysis of the national data [1]; that is, the respondents were divided into seven drinking groups based on their relative ranking on the consumption variable. The outcome for the county of Stockholm echoed the findings for the whole of Sweden; there was a marked consumption decrease which was synchronized across all drinking groups. This resulted in a marked positive relation between overall consumption and mean consumption in each of the seven drinking groups (Fig. 1). Thus, the findings show no indication of polarization in drinking, but rather supports the collectivity of drinking theory. It should be noted, however, that the city of Stockholm (which was the focus of the Hallgren et al. study) comprises about a third of the county of Stockholm (which our analysis is based upon). There is thus scope for more conclusive testing of the polarization hypothesis by the time additional waves of the ongoing drinking surveys for the city of Stockholm are obtained.