Information from the epidemiological literature shows that the number of diseases where low alcohol consumption has a detrimental effect far exceeds the number of diseases with a protective effect. In addition does this study indicate that there is a net detrimental effect of low alcohol consumption, especially for working-aged individuals, measured as medical cost and number of care episodes. It thus seems that the health-link argument in the alcohol-wage literature fails to account for those studies where alcohol consumption is shown to increase the risk of disease. It is therefore doubtful if the common explanation of health as the link between alcohol consumption and increased wages is valid in its existing form. There is a beneficial effect of low alcohol consumption on medical cost for the age group 80 years and above, most likely following the increase in prevalence with older age in those diseases with a beneficial effect of alcohol. However, this age group is normally outside the workforce and thus receives no wages. The net detrimental effect of low alcohol consumption on working aged individuals is expected to have a negative effect on wages which is contradictory to the hypothesised health link in the alcohol-wage literature. We see no obvious reasons why the results of this study could not be applied to other (comparable) countries. On the contrary, as Sweden has a high prevalence of coronary heart disease as a fraction of total disease burden compared to other countries , we expect that the net cost and care episodes due to low alcohol consumption will be lower in Sweden than other countries. That is, since the study still shows non-positive effects on health care cost and episodes, this indicates that the results and conclusions can be applied to other countries as well. A note for concern might be the rather hazardous drinking pattern in Sweden . However, the use of relative risks from international studies, due to lack of country specific information, and given that the hazardous pattern should reduce the beneficial effects, this should not hinder the application of the results from this study to other comparable countries.
Earlier work studying the effect of alcohol consumption on health care utilisation (compared to the alcohol-related health care utilisation in this study) have normally shown that alcohol use decreases utilisation compared to abstainers but also a negative relationship between different levels of consumption and utilisation, e.g. Zarkin et al. , Rodriguez Artalejo et al.  and Rice et al. . Other studies have failed to find significant differences between abstainers and low-risk drinkers,, after controlling for rudimentary and health-related confounders [25, 26]. Although this study employs a different outcome measure, i.e. alcohol-related diseases, the result seem to give some support to those studies that fail to find a protective effect on health care utilisation from low alcohol consumption compared to abstainers. Based on the epidemiological literature, one would expect an increase in health care utilisation from low alcohol consumption.
As discussed above, the result of the current study seems to sever the health link commonly used in the alcohol-wage literature, and the beneficial effects are found most pronounced for age-groups outside the labour force. The question that remains is how the 'missing link' between low alcohol consumption and increased wage should be explained. A number of other factors have been suggested such as family background and genetic endowment . Human capital accumulation  has also been suggested, for example if alcohol consumption affects educational attainment. A third discussed issue is social networking  where it is often suggested that low consumers, as opposed to abstainers, join in after-work activities to a larger extent and thus creates a stronger social network that, for example, facilitates promotion. Another type of argument why low alcohol consumers tend to have the highest wages is that low alcohol consumption might increase subjective as opposed to objective (e.g. as measured in the current study) health. This possible advantage in subjective health might manifest itself in different manners such as improved quality-of-life, and/or reducing the number of days absent from work. Especially the latter is interesting as sick leave should affect wages both in short and long term. However, performed studies on the effect of alcohol consumption on sickness absenteeism show differentiated results [e.g. [28–30]]. Further studies are needed focusing on the link between low alcohol consumption and wages as there are many possible explanations and the literature in many related fields (e.g. alcohol and subjective health and sickness absence) is inconclusive [28–33].
It should be noted, however, that if diseases where low alcohol consumption has a protective effect are more important in terms of sickness absence (and other labour market outcomes) compared to those diseases with a detrimental effect of low alcohol consumption, the health link might still be valid. This would be in despite of the fact that most of the protective effects are for the retired population. This should be investigated in future research in addition to studies investigating if there are differences in effects between working and non-working individuals.
As is evident from the division of cost on inpatient-, outpatient and primary care, the net detrimental effect of low consumption is driven by the two later as inpatient care shows a protective effect. It has to be acknowledged that the data material for outpatient- and primary care are of a lower quality than inpatient care. If there is a serious bias in the reporting of disease codes in outpatient and primary care, for example if alcohol related diseases are given a disease code to a lower extent than non-alcohol related diseases, this could have a major effect on the results of this study. However, we expect such bias in coding to be more associated with diseases that are socially stigmatised, in this case fully alcohol related diseases such as alcohol abuse and alcohol dependence syndrome. Therefore are possible bias in outpatient and primary care of lesser concern as this type of diseases are excluded from the study due to the difficulties of dividing these on consumption groups. However, exclusion of fully alcohol related diseases and accidents have lead to an underestimation of the negative effect of low consumption. The assumption made in this paper was that the excluded diseases and accidents does not burden low alcohol consumption. The underestimation caused by this is perhaps most evident for injuries resulting from accidents, for example falls and fires caused by alcohol consumption, but also the risk of being victim of violence and accidents caused by others' consumption, where it seems probable that this could happen also to low consumers. This built-in underestimation can be expected to further strengthen the effect found in the study of increased medical care cost and episodes resulting from low alcohol consumption.
If the result of the current study holds and along with it the contradiction of the common link in the alcohol-wage literature, it implies that somewhere has a false step been taken. For example could the health link be invalid as an explanation of the wage premium from low alcohol consumption or, alternatively, could the underlying epidemiological literature be in error. The latter is by no means controversial as when research advances, new diseases are shown to have a medical relationship to alcohol consumption and current relationships are adjusted. The possibility of this, however, does not 'save' the health link as it should be based on the existing level of knowledge. Further studies are required, also using econometric methods, before any final conclusions can be made.