Sammendrag
The impact of alcohol consumption on the social and economic environment is a topic of fierce debate in many countries. As usually the only legally distributed leisure drug, production, sales and even consumption of alcohol is often heavily regulated. Advocates of regulations seemingly have a strong case, as alcohol consumption has a number of adverse effects, both to the consumer and to society. Direct health risks to the consumer include cardiovascular disease, cirrhosis, cancer, depression, and even increased risk of suicide (Storvoll et al., 2010, Chapter 3). Alcohol consumption during pregnancy can have inhibiting effects on fetal development as well as adult outcomes for the child (Nilsson, 2008). Furthermore, alcohol consumption is related to increased risk of traffic accidents, both to the consumer and others, as well as criminal activity and the spread of sexually transmitted diseases. In Norway, alcohol is strictly regulated. Limiting alcohol consumption is explicitly stated as a purpose of the law regulating alcohol, as a means of limiting the harm caused by alcohol consumption. The main policy tool is restricting alcohol access. Access is restricted in many ways, both through price, time, and location restrictions. The adverse effects of alcohol do not automatically justify the regulations observed in Norway and many other countries. In this thesis, I first look at why regulations on alcohol could be justified. I argue that restrictions on alcohol could be welfare improving if there are negative externalities from alcohol consumption or if people have time inconsistent preferences. The external costs from alcohol consumption are not incorporated into the decisions on how much alcohol to consume exactly because they are external to the consumers making the decisions. An unregulated market therefore leads to a too high alcohol consumption relative to the social optimum. If people have time inconsistent preferences, they are unable to follow their optimal consumption paths, and restrictions on alcohol could be welfare improving. I distinguish between naive and sophisticated consumers. While the naive consumers are blissfully unaware of their time inconsistencies, the sophisticated agents are able to accurately predict how their future preferences diverges from their current preferences, and they are eager to restrict their future selves in order to follow their preferred consumption paths from the present point of view. There is scope for improving welfare for both types, but while the naive can only be forced towards his optimal consumption path, the sophisticated agent can deliberately use policies as commitment devices in order to (at least partially) overcome his self control problems. I then look at how different alcohol regulations could be welfare improving. Specifically, I look at restrictions on price, time, and location. If alcohol consumption entails a negative consumption externality, Pigouvian taxes improve welfare by forcing the consumers to internalize the external costs resulting from their alcohol consumption. To the naive time inconsistent consumer, restrictions on alcohol could be welfare improving by forcing him closer to his optimal consumption path than what he is able to achieve by the means of his insufficient self control. To the sophisticated time inconsistent agent, a policy forming a commitment device could increase his long term utility by allowing him to restrict the choice set of his future selves. I argue that time and location restrictions could form such commitment devices for the sophisticated time inconsistent consumers, and that even a short extra travel time to the nearest alcohol outlet could decrease alcohol consumption considerably. Next, I turn to testing empirically the prediction that travel times affect alcohol consumption negatively using unique data from Norway. Norway is a large country divided by long fjords and rugged mountains. Off premise sales of beverages with a higher alcohol content than 4.7 percent by volume is only legal through a state monopoly (Vinmonopolet), and a license is required for all other alcohol sales, both off and on premise. Due to a limited number of Vinmonopolet outlets, there are large variations in travel times to the nearest outlet. In order to investigate the effect of travel times to the nearest Vinmonopolet outlet on alcohol consumption, it is not meaning full to look at cross sections. Alcohol consumption may vary between cities and the country side, as does travel times to the nearest alcohol outlet. I therefore exploit variations in travel times to the nearest Vinmonopolet outlet due to the openings of a number of outlets between 2000 and 2012. I use data on travel times between municipalities and sales data from both Vinmonopolet and other enterprises where alcohol sales are possible. During the period 138 new state monopoly outlets opened. Of these 118 opened in municipalities without a previous outlet. Mean travel times decreased by almost 10 minutes. I exploit the variation in travel times between Vinmonopolet outlets and their customer bases due to the increased number of outlets, to investigate the effect of travel times on per person alcohol sales from Vinmonopolet. I use a fixed effects model to explain changes in per person alcohol sales by changes in travel times. I find a negative and highly significant relationship between mean travel time from Vinmonopolet outlets to its customer bases and the per person alcohol sales from Vinmonopolet. In particular, I find that a ten minute decrease in mean travel time in the customer base is related to an increase in per person alcohol sales from Vinmonopolet of 0.36 liters of pure alcohol per year. This is a 7.4 percent increase relative to the mean per person total alcohol sales between 2000 and 2010. The effects of changes in travel time on alcohol sales are likely to differ between people, depending on their initial travel time to the nearest outlet. I therefore estimate some other specifications to allow for a non-linear relationship. First, I include travel time squared in my regression equation. This gives the expected signs and highly significant estimates for both the linear and squared coefficient. Using this model, I predict the effect of a ten minute decrease in mean travel time from the year 2000 mean travel time. A ten minute decrease in travel time is related to an increase in per person alcohol sales of 0.54 liters of pure alcohol per year. I then estimate the model using the log of travel time. Again, the estimates are highly significant. The predicted increase in per person yearly alcohol sales from a ten minute decrease in travel time from the year 2000 mean travel time, is then 0.43 liters of pure alcohol. Finally, I estimate a model where I allow for a linear relationship that differs between outlets for which the customer bases have different mean travel times. For the outlets with mean travel time below or equal to 30 minutes, I find that a ten minute decrease in mean travel time is related to an increase in per person alcohol sales from Vinmonopolet of 0.66 liters of pure alcohol per year. This is a 13.6 percent increase relative to the 2000-2010 mean of total per person annual alcohol sales. The estimate is highly significant. For the outlets with mean travel time between 30 and 60 minutes, a ten minute decrease in mean travel time is related to a highly significant increase in per person sales from Vinmonopolet by 0.36 liters of pure alcohol per year. For the last group of outlets, those with mean travel time for the customers above 60 minutes, the coefficient is not significant, and it is quite small. In all the specifications, travel times to the nearest Vinmonopolet outlet are thus negatively related to per person alcohol sales from Vinmonopolet. The effect of changes in travel time is smaller the higher are initial travel times. These findings do not necessarily mean that shorter travel times are related to higher total per person alcohol consumption. Rather, it could be that the travel time to the nearest Vinmonopolet outlet only affects the composition of different forms of alcohol consumption. To investigate the possibility of substitution, I look at the effect of changes in travel time to the nearest Vinmonopolet outlets on per person alcohol sales through other channels than Vinmonopolet. The types of enterprises with potential to obtain licenses for alcohol sales are pubs and restaurants (on premise licenses) and grocery stores (off premise licenses). If there is substitution between alcohol from these sources and alcohol from the state monopoly, we would expect the relationship between the travel time to the nearest Vinmonopolet outlet and sales from pubs, restaurants, and grocery stores to be positive. For restaurants, the relationship is in fact positive. A ten minute decrease in travel time to the nearest Vinmonopolet outlet is related to a 22.2 NOK yearly decrease in per person sales from restaurants and cafes. However, this only to a very little extent compensates for the estimated increases in sales from Vinmonopolet, and the estimate is not significant. For grocery stores and bars, there is a negative relationship between per person sales and the mean travel time to the nearest Vinmonopolet outlet in the customer bases. For bars, the estimate is very small and insignificant. A ten minute decrease in mean travel time is related to a 2.46 NOK increase in per person yearly sales. For grocery stores, the estimate is higher and highly significant. A ten minute decrease in mean travel time for the customer bases of grocery stores is related to a 209 NOK increase in per person sales per year. This in turn leads to a negative relationship between total per person sales and travel time, as sales from grocery stores constitute most of the total per person sales. A ten minute decrease in travel time is related to a 171 NOK increase in total yearly per person sales. This could be due to a complementarity between alcohol acquired through different channels. However, it is also possible that sales in grocery stores are positively affected by increased proximity to a Vinmonopolet outlet, because people shop at grocery stores close to Vinmonopolet outlets while they are in the neighborhood. Decreases in mean travel time thus seem to be related to not only an increase in alcohol sales from Vinmonopolet, but to an increase in actual alcohol consumption, as there is little evidence of substitution from alcohol from other sources.