When Denmark reduced its spirits tax by 45%, on 1 October 2003, and Sweden increased its travellers’ quotas, on 1 January 2004, to levels that made it possible to bring back almost unlimited amounts of alcohol to Sweden from elsewhere in the European Union (EU), it was hypothesized that alcohol consumption in Sweden would peak and that alcohol-related problems would follow accordingly. This was in accordance with the economic literature that links price to alcohol consumption and its consequences (^{Chaloupka et al., 2002}). Although most studies have looked at total consumption or register data (^{Andreasson et al., 2006}; ^{Bloomfield et al., 2009}; ^{Cook 1981}; ^{Cook and Tauchen 1982}; ^{Cook et al., 2005}; ^{Herttua et al., 2008a},^b, ²⁰⁰⁹; ^{Koski et al., 2007}; ^{Wagenaar et al., 2009}), a few studies have also been able to link the decreased taxation on alcoholic beverage to an increased self-reported consumption of that beverage (^{Heeb et al., 2003}) and more reported social problems (^{Mohler-Kuo et al., 2004}).

As southern part of Sweden is closer to Denmark and Germany than the rest of Sweden, southern Sweden had a much higher share of imports and consumption of illegal alcohol compared with the rest of Sweden, prior to the changes (^{Gustafsson and Trolldal, 2004}). It was therefore hypothesized that the changes would have greater effects on the levels of consumption and problems there than in the rest of Sweden. Although northern part of Sweden is assumed to have the same drinking culture as in the southern parts, the effects by the increased traveller's quotas and decreases in prices in Denmark are less due to distance; thus, northern Sweden was used as control. This hypothesis was tested in the Nordic tax study, which also covered similar changes in travellers’ allowances in Denmark and Finland as well as alcohol tax decreases in Finland in March 2004 (an average tax decrease of 33% on all alcoholic beverages or 44% decrease on spirits corresponding to a 36% lower off-premise retail price on spirits).

However, the first analysis of changes in alcohol consumption as reported in population surveys in these countries found that the changes had not had any general effects in any of the study sites (Denmark, Finland and southern Sweden), although there were changes in some subgroups of the population, for example, among women in southern Sweden and men in Denmark (^{Mäkelä et al., 2007}). Follow-up analysis on Swedish data through 2006, 3 years after the changes, confirmed that there had not been any significant overall increase in consumption in southern Sweden (^{Gustafsson, 2010}). These results suggested that there had been a decrease in volumes of alcohol consumption among younger and middle aged men, but an increase among younger and middle aged women in the southern area, whereas roughly the opposite pattern was observed in the north. People in high income groups in the south were showing decreases, whereas in the north only women with a high monthly income had decreased their consumption. Among heavy consumers (the highest 10%), in southern Sweden, women as well as persons older than 50 years had shown increased alcohol consumption, whereas women reported lower consumption in the north (older persons increased in this site as well).

Skog's theory of collectivity of drinking cultures (^{Skog, 1985}) states, in brief, that the whole population moves upwards on the consumption scale when the mean consumption increases (and likewise for decreases). On the basis of a model in which society is conceived of as a large network with many ties between individuals, each actor living in the same society/culture is influenced by and influences other members of society—directly or indirectly through other ties linked to personal social networks. Therefore, Skog argues, each person in a culture will tend to adjust his/her alcohol consumption to other persons within the same culture, and alcohol consumption will fluctuate up and down in concert. In response to ^{Gmel and Rehm (2000)}, ^{Skog (2001)} elaborates on his theory and more explicitly states that collectivity refers to 'cultural differences (between populations) and changes over time (in a specific population)'; thus, the collectivity concept is thus 2-fold, predicting differences between populations and an uniformity of responses within a population. According to this model, consumption in southern and northern Sweden, as part of the same drinking culture, would be expected to behave in the same way. However, with differences in relevant determinants, such as the tax and quota change, the site less affected should be expected to change in the same direction but to a lesser degree. The findings on the changes in reported alcohol consumption (^{Gustafsson, 2010}), however, imply that there was probably not an overall uniform change in amounts of drinking in Sweden at this time; nor did the various subgroups seem to move in concert.

As mentioned above, when designing the study, the southern and northern regions of Sweden were selected on the assumption that they were part of the same society and culture, thus part of the same drinking culture, the difference being that one region (south) would possibly be affected by the tax decrease in Denmark and the increased Swedish traveller's import quotas, while the other region (north) would not. The results found in the study of self-reported alcohol consumption (^{Gustafsson, 2010}) raise questions about this model. In fact, although the differences in overall trends were not significant between the two sites, they seemed to be moving in opposite directions from each other; where there was an increase in average consumption in the northern area between 2003 and 2006, and persons in the southern area (if anything) rather decreased their consumption. Additionally, there seemed to be differences in patterns of change between different population subgroups, within sites and between sites. For example, men in the north increased their consumption, whereas the trend among men in the south seemed to be negative; among women the opposite pattern was observed. Moreover, different age groups did not behave in a similar manner: in groups where consumption seemed to increase in the north, it decreased in the south. Additionally, the different age groups (by sex) behaved differently within the same region. The same was true for income groups and for the subgroup of high consumers. Likewise, ^{Romelsjö (1989)} has shown that various population groups change differently when the average consumption level changes. These studies can be seen as examples of a weakened theory of the collectivity of drinking. ^{Skog (2001)} in some way allows for this in his response to ^{Gmel and Rehm (2000)}, as he cites examples of ‘exceptions from the overall pattern’. However, in elaborating on his theory from 1985 in this way, there is not too much left of the hypothesis of collectivity at all.

Alcohol-related problems do usually follow consumption trends. When consumption and alcohol-related problems do not follow each other, there is often a lagged effect for problems (^{Skog, 1987}). This, however, is usually related to certain types of harms, for example, liver cirrhosis, which develops over a longer period of time than do acute harms, but even for these types of harms an effect may be observed in a shorter perspective on an aggregated level (^{Herttua et al., 2008a}). There are also some Nordic studies, showing that the effect of policy changes on the rates of alcohol-related problems (at least as reported in registers) may sometimes be greater than the effect on the (total) alcohol consumption level; sometimes, an effect on alcohol-related problems is also observed even when no measurable effect on overall drinking level has been found (^{Room et al., 2002}). It has additionally been suggested that when there is a change in alcohol-related problems (liver cirrhosis rates) without a measurable change in alcohol consumption, this can be explained by the changes in treatment of alcohol abuse and AA membership (^{Mann et al., 1988}; ^{Smart and Mann, 1991}, ¹⁹⁹³; ^{Smart et al., 1998}). The changes in the rates of self-reported problems, however, can reflect the changes in the perception and in the threshold for what is considered as a problem, as well as changes in more objective events. Thus, when ^{Midanik and Clark (1995)} found that self-reported alcohol consumption decreased while the rates of self-reported problems did not change (except for in some groups where it increased), two potential explanations were offered. One was that alcohol consumption decreased particularly among the general public, where those who had problems generally experienced minor problems, whereas problem drinkers did not change their drinking pattern. The second potential explanation was that there is a threshold for when problems are counted as problems, and that this may vary; when consumption decreases, what is counted as a problem may also change, especially as observed in surveys.

^{Bloomfield et al. (2010)} looked into some reported alcohol-related problems, reported in a comparable manner for all three countries in the Nordic tax study, by sex and age. They did not find increases in problems in Denmark or southern Sweden, although in Finland and northern Sweden increases could be found among some subgroups, and separate analyses of register data on problems from very heavy drinking showed increasing rates in Finland (^{Mäkelä and Österberg, 2009}). Post hoc explanations have suggested that Denmark and southern Sweden have more ‘continental’ drinking habits, and that they follow the general decrease found in many other European countries, whereas Finland and northern Sweden are lagging behind this development, thus retaining a typical Nordic drinking style (^{Norström, 2001}). Additional problem items were included in Sweden, which would give us the opportunity to look more closely at alcohol-related problems as two indexes and by additional population groups.

From this perspective, it is also of interest to study the patterns of alcohol-related problems in Sweden after changes in price and availability in light of Skog's theory. Although Skog's theory concerns consumption rather than alcohol-related problems, it can be argued that alcohol-related problems should be expected to follow accordingly, due to the strong relation between the two variables, and that we should be able to obtain a rough approximation of whether there was any collectivity in problems. On the basis of the theory of collectivity of drinking cultures, we would thus expect the changes in alcohol-related problems to change (relatively) and to rise and decline together in the southern and northern parts of Sweden. On top of this, in comparing rates in the two problem indexes in 2003 with rates in 2004–2006, we should find a greater increase in southern Sweden because alcohol prices, broadly considered, decreased and purchases abroad were already more common in the southern site. Thus, people in the north are expected to follow, but initially to a lesser extent than in the south. According to the 1985 version of the collectivity hypothesis, all subgroups in a population should behave in a similar way, so that both in the north and in the south men and women, different age and income groups, and the highest 10% of consumers versus others, should also move in concert. For the analyses, besides comparing the overall changes in alcohol-related problems to overall changes in consumption, it is also logical to make comparisons of problem rates by population groups.

The 2-folded aim of the present study will be to:

1. study whether self-reported alcohol-related social problems, mainly presented by two indexes describing (a) impaired self-control/dependent behaviour and (b) extrinsic problems, followed the same pattern of changes in southern and northern Sweden as did self-reported alcohol consumption as reported by ^{Gustafsson (2010)};
2. study whether self-reported alcohol-related problems according to these two indexes changed collectively, that is, in the same direction, in the two sites and among various population groups after the Danish tax decrease and the Swedish abolition of import quotas on alcohol; this will be analysed on the basis of Skog's strong version of the collectivity of drinking cultures theory (^{Skog, 1985}).

Data from the Nordic tax study were used; a telephone survey with a sample from the adult Swedish population, aged 16–80 years old, was carried out during the third quarter (July–September) of 2003, prior to the Danish tax change and the increase in quotas, by the national survey institute Synovate Sweden AB (former Temo). The samples were generated by random digit dialling, and interviews were carried out using a computer-assisted telephone interviewing system. New cross-sectional samples were drawn in each site the following years until 2006, following the same procedure as in 2003. From these samples, a southern and northern sample was included in the design. The southern part, including Scania and Gothenburg, was defined by its proximity to Denmark via bridge or ferry (for definition and map, see ^{Mäkelä et al., 2008}). The northern part, which was believed to be relatively unaffected by the changes and served as a control area, included an area 150 km west of Stockholm up to 150 km from the Finnish border in northern of Sweden.

As initial analysis of the problem indexes for each year separately did not differ significantly between the 3 years (2004, 2005 and 2006) following the changes, the results will therefore be presented as a comparison between before (2003) and after (2004, 2005, 2006 together) the Danish tax decrease and the abolishment of private import quotas had taken place; this should also result in more stable results. The samples analysed included 1425 persons in the southern sample and 1336 in the northern sample in 2003, and between 972 and 1406 in the southern sample and between 994 and 1343 in the northern sample in 2004, 2005 and 2006. The cooperation rates were 50.5% in 2003 (before the changes) and 45.3% in 2004, 2005 and 2006 (after the changes) in the southern parts of Sweden, and the corresponding figures for the northern parts were 48.4 and 42.3%. These response rates are similar to those in other similar Swedish surveys (^{Hradilova Selin, 2004}; ^{Leifman and Gustafsson 2003}), but the implications of these rates are further discussed in the last section of the paper.

Initial analysis by single alcohol-related problem item did not show a great deal of variation in changes of prevalence pre- and post-intervention for the questions separately (Table 1). At the level of the two scores, there was only a significant change for extrinsic problems: a decrease in the south. For an overall problems score, whether the division was at one or more, two or more, or three or more problems, the results went in the same direction, with significant decreases in problems in the south, but generally non-significant change in the north. In the southern site, the few specific problem items showing significant change moved in a negative direction; they were less common after the interventions. These were ‘having had guilty feelings after consuming alcohol’, ‘having been a passenger in a car where the driver had been drinking alcohol’, whether ‘alcohol had harmed work or studies’ or whether ‘alcohol had harmed relationships with family or the relationship with a partner’. The change in answers to the questions about harming family or partner relationships was also shown to be significantly different from the change in the north (small non-significant increase). In the northern site, individuals had experienced memory loss more often after the interventions and this difference in change was also significant between the sites, although the decrease in the south was not in itself significant.

Looking at the direction of change at the item level and ignoring significance, all seven impaired self-control items in the south decreased, while five of them increased in the north. For extrinsic problems, all nine decreased in the south, while seven of them increased in the north.

It should be emphasized that the overall rates of reported alcohol-related problems in the southern and northern parts of Sweden did not change, and certainly not in the direction expected in the original hypotheses of the Nordic tax study, although changes were observed in some population groups. Instead, as discussed in the paper on changes in consumption in Sweden (^{Gustafsson, 2010}), there seem to have been effects in the opposite direction to what was expected. Thus, even if the development in consumption and alcohol-related problems is not always parallel (^{Skog, 1987}), it does not seem to have changed in accordance with expectations. The overall rates of problems decreased in both the south and north, but (marginally significantly) more in the south, contrary to the hypothesis in the Nordic tax study. For a couple of items (memory loss and alcohol harming the family), the changes were additionally significantly contrary to the hypothesis in the two sites.

In relation to self-reported consumption, the (significant) increases in consumption in the north shown by ^{Gustafsson (2010)} were matched by the alcohol-related problems according to both indices, although here the results were non-significant. The results presented for alcohol-related problems (significant for extrinsic problems) were also in accordance with the (non-significant) decrease found for consumption in the southern site. The increases in consumption found among men in the north seemed to be valid for self-reported alcohol-related problems as well, although when broken down by age group the relationship is not as clear.

Given the nature of the interventions, where the actual decrease in price took place in another country, the results imply that the price changes may have been too small to result in further increases in consumption and changes in rates of self-reported problems. According to ^{Gustafsson (2010)}, spirits consumption in the southern site decreased by 13% (while it increased by 25% in the north). A change within Sweden may well have been more likely to give significant results in the expected direction, as price changes usually are expected to change the demand for alcohol (^{Chaloupka et al., 2002}). It is important to remember that there had already been great increases in alcohol consumption in the general population in the southern site before the tax reduction and quota changes (^{Boman et al., 2006}), and that before the policy reforms started in 1996—after entering into the EU—southern Sweden actually had the lowest consumption levels in Sweden, as well as the lowest share of binge drinkers and risk consumers (^{Kühlhorn et al., 1999}). Concerning the results in the northern site, where problems (as well as spirits consumption) were observed to have increased, the decreases in taxes in Finland may have had more influence across the border than expected, though municipalities closest to Finland were excluded from the sample.

What is shown here are two parallel changes in the opposite directions, but both occurring in Sweden. In southern Sweden, the decrease in reported problems (and consumption) could very well be related to a more ‘continental’ manner of drinking with meals during the working week rather than intoxication during the weekends, as also suggested by ^{Bloomfield et al. (2010)}. In discussing various typologies of the cultural position of drinking, ^{Room and Mäkelä (2000)} ended up with four main approaches—one of them being banalized drinking, where drinking is more woven into daily life, as opposed to binge drinking occasions. The frequency of people who never binge drink has been shown to be increasing in Sweden (^{Ramstedt et al., 2009}); however, banalized drinking has been suggested to be practised especially by those living in the southern parts of Sweden and the amount consumed by binge drinkers (referring to the past year) has also decreased in this site, but not in the north (^{Gustafsson, 2010}). Yet, some alcohol-related harm can still be related to banalized drinking; ^{Room and Mäkelä (2000)} mentioned two examples: traffic casualties and liver damage. In fact, drunk driving was one of the two single problem questions in this study for which no decrease was observed in the southern site. In northern Sweden, on the other hand, problems (and consumption) were rising. It may be that there is a lagged effect in the northern site, such that the north is now catching up on the major increases in consumption that have already taken place in the south (^{Boman et al., 2007}). It has been suggested that alcohol consumption has reached a point of saturation in the south due to the rising consumption levels in this site (^{Gustafsson, 2010}), although as ^{Room et al. (2009)} pointed out, there may be other explanations as well.

There are also differences between and by subgroups. In relation to Skog's theory of collective drinking (^{Skog, 1985}), the present results do not seem to follow the predictions. If the extended version of Skog's original theory of the collectivity of drinking presented here was to hold true for alcohol-related problems, we would expect alcohol-related problems to increase/decrease evenly (relatively) or at least in the same direction by subgroups when policies change, or for that matter when there is a general societal change. When a policy change affects one site in a society but not another, we would expect a change in accordance with the policy. On the contrary, we observed rather that northern and southern Sweden were moving in opposite directions to those expected, and there were substantial variations in patterns of change between different population subgroups. Additionally, the difference in patterns of change between subgroups was not the same in the two sites. For the collectivity hypothesis to be supported, sub-populations too would need to be moving in the same direction, which is not the case here. In fact, how strongly the trends in two subgroups matches each other could be an indication of the extent to which they are part of a single collectivity with respect to drinking. Given the present results, clearly changes can happen in one part of a population and not in another. This has been shown by others as well (^{Midanik and Clark, 1995}; ^{Romelsjö and Diderichsen, 1989}; ^{Room, 1984}; ^{Sulkunen, 1981}). As already mentioned, ^{Skog (2001)}, in his response to comments by ^{Gmel and Rehm (2000)}, allowed for subgroups to change differently when he retreats to a less strong position than in 1985, stating that there are exceptions to the overall pattern, for example, between the sexes and across different social-economic strata.

^{Norström and Romelsjö (1998)} have suggested that social and cultural boundaries could hamper the overall collective process; the existence of several parallel drinking cultures in one society is not a new phenomenon (^{Ullman, 1958}) and it has even been suggested that some population groups refuse adaptation to the mainstream drinking culture as a way of differentiating themselves (^{Room, 2005}). Perhaps, then, northern and southern Sweden could be perceived as having two separate drinking cultures, which do not necessarily move together, although they are part of the same country. Skog was in fact never explicit about what the boundaries for collectivity and for cultural drinking networks were; although he most often used national data in his analysis, the collectivity he refers to is not necessarily that of a nation. Additionally, further elaboration of the theory would call for an analysis of the extent to which habits in networks strive towards imitating each other, and under what circumstances habits are more likely to develop differentiating patterns.

Naturally, the present study has certain limitations; there is a risk that survey data will exclude some population groups that have been exposed to and affected by the interventions under investigation. Regarding these particular changes, travellers are an especially interesting group; if they travel often, they are more difficult to reach using telephone surveys. Another drawback of the surveys is that they may make it difficult to reach problem drinkers. Although analyses of these data show that those who are harder to reach do not differ a great deal from those who are easier to reach (^{Wennberg et al., forthcoming}), we cannot know whether those not responding to the survey differ from our sample. In terms of change in consumption, a significant increase was observed among heavy consumers and heavy consuming men (^{Gustafsson, 2010}). We already know from earlier studies on alcohol surveys (^{Kühlhorn et al., 1999}) that high consumers are underrepresented in surveys. Thus, we cannot expect that our highest 10% will include everyone in the highest 10% in the Swedish population and accordingly we can assume that our group has a drinking less problematic. This can partially explain why the significant changes in a few subgroups among high consumers (highest 10%) in the southern area most often involved a decrease. A more plausible explanation may be that problems are a less precise measure than is the amount of consumption. Although I would argue that we should be able to get a crude measure of whether problems changed collectively, changes in problems may be harder to detect through surveys than are changes in consumption. Naturally, this is a problem with surveys, which is why register data can be useful. In fact, ^{Gustafsson and Ramstedt (forthcoming)} did note an increase in the number of alcohol poisonings (but not in the number of police-reported cases of drunk driving or violent assaults) in southern Sweden between January 2000 and December 2007 linked to the increased import quotas. Perhaps surprisingly, though, initial analysis in the study by Gustafsson and Ramstedt did not show an increase in alcohol poisonings or any of the other harms in the northern parts of Sweden. In a longer perspective, ^{Ramstedt (2009)} has reported increases in alcohol-related mortality and alcohol-related hospitalization among women, especially in the oldest age groups, and with regard to hospitalization also among the youngest groups (both men and women) in Sweden. Still, the analyses presented here, of relatively common problems in the general population and in subgroups of it, would not be possible with register data.

To conclude, self-reported problems did not increase in the south as was expected given the tax decrease in Denmark and increased quotas in Sweden. Instead, the patterns of change in problems were in accordance with the results already presented for alcohol consumption. It was also established that patterns of change in impaired self-control problems and extrinsic problems from drinking did not change collectively, as would be expected by a version of the collectivity theory of drinking originally brought forward by ^{Skog (1985)}. Rather, there were variations in change between and within subgroups, both within and between regions. A substantial part of the divergence in the ‘wrong’ direction was related to what happened among the heavier drinkers, and particularly among men in the low-income group and younger male drinkers.

This work (the Swedish part of the study) was supported by the Joint Committee for Nordic Research Councils for the Humanities and the Social Sciences (NOS-HS, project 20071), the US National Institute on Alcohol Abuse and Alcoholism (R01 AA014879) and partial funding for the 2003 data collection from Systembolaget (the Swedish alcohol monopoly).

The author would like to thank Robin Room and Karin Helmersson Bergmark for their insightful comments, and Pia Mäkelä and Christoffer Tigerstedt for commenting on earlier drafts.