Alcohol Control Policies And Alcohol Consumption
A great deal of economics research on alcohol use has focused on estimating the effects of alcohol control policies on alcohol consumption, both because this type of policy evaluation is independently interesting and because many policy-induced changes in alcohol consumption can be used to identify causal effects of alcohol use on outcomes (e.g., mortality and morbidity). Research on alcohol control policies is particularly appealing to economists because of the fundamental tenet in economics that demand curves slope downward. That is, the price of a commodity and the quantity demanded of that commodity are inversely related. In the context of alcohol consumption, this means that policies and practices that raise the full price of drinking either directly (e.g., through alcohol taxes, which are passed through to consumers in the form of higher alcohol prices) or indirectly (e.g., through other types of availability restrictions) should reduce the quantity of alcohol consumed. Although alcohol taxes are probably the most widely studied alcohol control policies in the economics literature (and have been summarized in multiple recent meta-analyses), many others have also received scholarly attention, including the presence of government liquor monopolies; age-based alcohol availability restrictions (e.g., minimum legal drinking ages (MLDAs)); drunk driving laws (e.g., BAC limits, driver license suspensions, random breath tests, and sanctions/penalties for driving under the influence); spatial restrictions on alcohol availability (e.g., liquor license restrictions); temporal restrictions on alcohol availability (e.g., Sunday alcohol sales bans or bar/pub closing hours); advertising and sponsorship restrictions (including health warnings); other ‘circumstance’ regulations such as prohibitions on alcohol sales at sporting events; and legal liability for bartenders and bar owners for serving intoxicated persons, among others.
The most common approach taken in this literature to test whether the demand curve for alcohol slopes downward has been to relate drinking rates (using either individual-level survey data on alcohol consumption or aggregate data on alcohol sales) to variation across places in the alcohol control environment at a point in time. This approach is made possible by the fact that places (e.g., localities, states, provinces, countries, etc.) vary substantially in their chosen menu of alcohol control policies. For example, some places have higher alcohol taxes and/or severe penalties for alcohol-involved driving compared to other places. A finding that drinking rates are lower in places where alcohol is more difficult to obtain (i.e., where individuals face higher full prices to drink) is usually taken as evidence that demand curves slope downward, or that drinking is negatively related to price.
One weakness of the type of approach described in the previous paragraph is that the types of designs that rely on variation across places at any one point in time may suffer from the omitted variables biases. For example, if places that are very religious are the ones that are more likely to have high alcohol taxes and strict availability restrictions, then the inverse relationship might be observed between the full price of obtaining alcohol and drinking rates that is due to the religious attitudes of people in that area, not due to the policies and prices themselves. This criticism has in the past decade led economists to incorporate other types of research designs commonly found in other applied microeconomics disciplines (most notably labor, public, and development economics). As such, the more commonly accepted standard for evaluation research on alcohol control policies and alcohol consumption is to compare changes in drinking rates coincident with changes in alcohol control policies (e.g., alcohol excise tax increases or tightening of availability rules). The advantage of this ‘changes on changes’ or ‘DID’ type of specification is that, because areas usually adopt different alcohol control policies at different times, researchers can use the staggered timing of adoption to rule out the possibility that permanent unobserved differences about individual places are driving the observed relationships between alcohol prices (broadly defined) and alcohol consumption. In practice, this amounts to including dummy variables, or fixed effects, for each area in multivariate regression models of drinking that include controls for area-specific alcohol policies.
Results from this and other types of quasi-experimental approaches have been somewhat less conclusive about the role of alcohol taxes in determining alcohol consumption behaviors, in that they have not uniformly returned evidence of significant relationships between alcohol excise tax increases and alcohol consumption decreases, particularly for research on youths and for research focusing on the US. Part of the lack of clarity around the effects of alcohol taxes on consumption is that in the US there have been relatively few large alcohol tax increases in the past three decades; by construction, this makes estimating difference or change-based models more difficult. (The lack of alcohol tax change variation is notably different from the case of tobacco.) Similarly, studies of spatial, temporal, and other ‘circumstance’-type regulations of alcohol availability have not produced overwhelming evidence that these policies seriously affect overall alcohol consumption, which is perhaps not surprising because it is not particularly costly to undo the effects of these types of restrictions (e.g., purchasing alcohol on Saturday can undo the effects of a Sunday alcohol sales prohibition). There is, however, ample evidence from these types of stronger designs that age-based alcohol restrictions (such as MLDAs) causally reduce alcohol use. For example, research in the US has shown that state experimentation with lower drinking ages in the 1970s and early 1980s led to higher drinking rates among youths who were newly legal to drink, and similarly state increases in drinking ages back to age 21 (the current MLDA in the US) led to lower drinking rates among youths who were no longer legally allowed to drink. Moreover, research has also shown that alcohol use increases sharply and discretely exactly at a country’s MLDA, even when other policies do not change discontinuously at the same threshold. This further bolsters the idea that minimum drinking age policies causally reduce alcohol consumption. Because drinking ages affect the total price of obtaining alcohol through time and convenience costs for youths who are too young to legally consume alcohol, studies of minimum drinking ages have played an important role in confirming that demand curves do, indeed, slope downward for alcohol.
Finally, it can be noted that research exploiting changes in place-specific alcohol control regulations to identify the effects of higher effective prices for obtaining alcohol on drinking rates – with improvements over comparisons of drinking rates across areas at a point in time – are not a panacea. Specifically, these studies must also contend with the fact that alcohol policy changes may themselves likely be the result of unobserved population preferences , because in democratic societies voters elect officials who make or change policy. If sharp changes in attitudes toward alcohol underlie the changes in alcohol control policies, then studies using DID can still be biased. In this situation, other strategies that are less prone to these criticisms, such as RD or IV, can be useful alternatives.