Minimum Legal Drinking Age and The Social Gradient in Binge Drinking
Minimum Legal Drinking Age and The Social Gradient in Binge Drinking
Minimum Legal Drinking Age and The Social Gradient in Binge Drinking
3 Teenage binge drinking is more prevalent in Europe than in the US, where alcohol is forbidden to people under the age of
4 21. This column looks at the relationship between minimum legal drinking ages and alcohol abuse. Using administrative
5 health and survey data from Austria, it finds significantly increased alcohol consumption – particularly among boys and those
6 from underprivileged backgrounds – when drinking becomes legal. Raising the minimum legal drinking age in Europe could
7 reduce alcohol poisonings and the early socioeconomic gradient in teenage binge drinking.
8 Europe leads the world in alcohol consumption. According to WHO, more than 10.3 million
9 disability-adjusted life years were lost due to alcohol abuse in 2016, and 10% of all deaths are
10 attributable to alcohol (World Health Organization 2018). Teenage binge drinking is much more
11 prevalent in Europe than in the US, a disparity often attributed to the large difference in minimum
12 legal drinking ages (MLDAs). While US teenagers are prohibited from drinking alcohol before
13 turning 21, the MLDA in most European countries is 18. Some countries even allow drinking for
14 teenagers as young as 16. Proponents of low MLDAs suggest that they foster more responsible
15 alcohol consumption later in life, while critics argue that youth drinking may have detrimental effects
16 on brain development.
17 There is a large literature on the effects of the MLDA on a variety of outcomes, such as alcohol and
18 drug use (e.g. Carpenter Dobkin Warman 2016, Crost and Rees 2013), mortality (e.g. Carpenter
19 and Dobkin 2009), crime (e.g. Carpenter and Dobkin 2015), drinking and driving (e.g. Miron and
20 Tetelbaum 2007), and schooling (e.g. Lindo et al. 2013). Most of these studies are based on US or
21 Canadian data, where the MLDA is high. We complement this literature by studying teenage binge
22 drinking in Austria, which has an MLDA of 16 years (Ahammer et al. 2021). Austria is particularly
23 interesting because it is among the countries with the highest levels of alcohol consumption
24 worldwide (see Figure 1). Moreover, we can draw from an excellent data pool in Austria. We use
25 both high-quality survey data and administrative health registers, which allow us to take a closer
26 look at the socioeconomic gradient in binge drinking and the mechanisms behind MLDA legislation.
27 First, we want to know how drinking behaviour changes as teenagers age across the MLDA
28 threshold. To this end, we pull data on self-reported alcohol consumption in Austria from the
29 European School Survey Project on Alcohol and Other Drugs (ESPAD). We compare teenagers
30 who were just above and just below 16 years of age when taking the survey and find a clear
31 discontinuous increase in past-week drinking at age 16, which – according to our regression
32 discontinuity (RD) models – amounts to roughly 0.3 days (see Figure 2). The probability of drinking
33 at least once in the past week increases by 12 percentage points at age 16. We also find interesting
34 heterogeneities by sex and socioeconomic status. Male and low socioeconomic status (SES)
35 teenagers tend to increase heavy drinking more than young women when alcohol becomes legally
36 available.
1
Minimum legal drinking age and the social gradient in binge drinking | VOX, CEPR Policy Portal (voxeu.org)
37 Figure 1 Alcohol consumption: International comparison
38
39 Figure 2 Effect on number of days drinking during the last seven days
40
41 The ESPAD also has information on the quantity of alcohol consumed. Again, there is a noticeable
42 jump at age 16, amounting to about 50 more grammes of pure alcohol consumed in the past week
43 (Figure 3). This is almost double the pre-16 level. These positive effects appear along the full
44 drinking distribution. The probability that teenagers consume at least 180–240 grammes more –
45 equivalent to an extra 9–12 pints of beer – increases by 10 percentage points too. As before, we
46 find that male and low-SES teenagers are more likely to drink excessive amounts.
47 Figure 3 Effects on grammes of pure alcohol consumed during the last seven days
48
49 In a next step, we want to know how these changes in drinking behaviour translate to morbidity.
50 This is important, because the societal cost of drinking alcohol likely skyrockets once it leads to
51 hospitalisation, especially with teenagers. Using high-quality administrative health records from the
52 Upper Austrian Health Insurance Fund (UAHIF), we pull information on hospitalisations due to
53 alcohol poisoning for the full population of children born between 1991 and 1995 in Upper Austria.
54 Alcohol poisonings are a rather rare outcome (one in a thousand children in our data suffer from
55 alcohol poisoning), but the advantage of these data is that we do not have to worry about reporting
56 problems or recall bias.
57 When we look at the probability that a teenager is admitted to hospital with alcohol poisoning by
58 age, we find a very similar jump at the MLDA threshold. The discontinuity estimate suggests a 42%
59 increase at age 16 (Figure 4). Splitting by SES reveals an interesting pattern. The age trends in
60 poisonings are almost overlapping before age 16, but low SES teenagers seem to respond more
61 strongly at the MLDA cutoff. We then see that poisonings are a concave function of age for low
62 SES, but a convex function for high SES teenagers. The resulting socioeconomic gap is largest
63 between ages 17 and 18, after which it converges. Even at age 22, there is still a statistically
64 significant difference in poisonings between low- and high-SES teenagers.
66
68
69 Another advantage of our register data is that we can investigate whether teenagers that are
70 exposed to alcohol abuse in the family react differently to MLDA regulation. To this end, we identify
71 teens whose parents were diagnosed with alcohol-induced liver cirrhosis. Interestingly, it seems that
72 these teenagers do not change their behaviour at age 16. One might be tempted to interpret this as
73 a deterrence effect, but given that the incidence of alcohol poisonings is much higher in families
74 with a history of alcohol abuse, we interpret this as a sign that teenagers imitate their parents’
75 behaviour and engage more often in excessive drinking, regardless of whether they are legally
76 allowed to drink. In high-risk groups, therefore, MLDA may not be effective.
77 What drives this sharp increase in alcohol abuse at age 16? We have two mechanisms in mind:
78 access and risk perceptions. To gauge the importance of access, we first use new data from a
79 mystery shopping study in Upper Austria, where underage test buyers visit retail shops and attempt
80 to purchase liquor. Out of 4,269 purchase attempts, around 23% were successful. Hence, only
81 three-quarters of retailers comply with MLDA regulation, which may point to an enforcement
82 problem. If we aggregate these attempts at the municipal level and merge socioeconomic
83 information, we find that the socioeconomic composition does not correlate with retailer compliance.
84 This is in line with the observation that, prior to age 16, there is no difference in binge drinking
85 between children from different socioeconomic backgrounds. Also in the ESPAD survey, 84% of 15
86 year-olds perceive access to alcohol as ‘easy’ or ‘rather easy’. Taken together, these findings
87 suggest that a lack of access to alcohol can hardly explain the effectiveness of MLDA legislation.
88 A plausible complementary mechanism is that MLDA legislation has established a normative value,
89 in the sense that some teenagers simply feel obliged to obey and abstain from drinking alcohol
90 before age 16, despite its availability. In addition, parents may become more lenient when their
91 child reaches age 16 and drinking becomes not just legally allowed, but also socially accepted. This
92 normative mechanism is difficult to test empirically. The ESPAD survey, however, includes a
93 question on risk perceptions about alcohol, which we can use as a surrogate outcome. Absent
94 MLDA legislation, we expect risk perceptions to be a continuous function of age, but this is not what
95 we see in the data. In fact, the perceived risk of heavy drinking during weekends decreases
96 significantly at the MLDA cutoff (Figure 5). We interpret this as suggestive evidence for a normative
97 impact of the legislation.
98 To conclude, we find that teenagers increase both the frequency and the intensity of drinking at the
99 MLDA cutoff, and these effects tend to be stronger for boys and low SES teenagers. We show that
100 these effects persist for some years and cannot be explained by birthday effects. We also find some
101 evidence that a normative effect of the legislation, and not access to alcohol, might explain the
102 MLDA effect.
103 Unfortunately, we cannot perform a comprehensive welfare analysis. However, if we are worried
104 about SES disparities, our results suggest that a (stepwise) increase of the MLDA would reduce the
105 number of alcohol poisonings and the early socioeconomic gradient in teenage binge drinking. As
106 an alternative to raising the MLDA for everybody, it might be worth considering measures that
107 particularly target teenagers with a low socioeconomic background to avoid an early socioeconomic
108 gradient in harmful binge drinking. This might also be the preferred option for teenagers from
109 families with a history of severe alcohol abuse, since MLDA regulations are not effective for this
110 high-risk group. Moreover, targeted measures in high-risk geographical areas may be useful
111 (Hinnosaar and Liu 2020).
112 Figure 5 Risk perception of daily drinking and heavy drinking at weekends
113
114 References
115 Ahammer, A, S Bauernschuster, M Halla and H Lachenmaier (2021), “Minimum Legal Drinking Age and the Social Gradient
116 in Binge Drinking”. IZA Discussion Paper No. 13987.
117 Carpenter, C and C Dobkin (2009), “The Effect of Alcohol Consumption on Mortality: Regression Discontinuity Evidence from
118 the Minimum Drinking Age”, American Economic Journal: Applied Economics 1(1): 164–182.
119 Carpenter, C and C Dobkin (2015), “The Minimum Legal Drinking Age and Crime”, Review of Economics and
120 Statistics 98(2): 254–267.
121 Carpenter, C, C Dobkin and C Warman (2016), “The Mechanisms of Alcohol Control”, Journal of Human Resources 51(2):
122 328–356.
123 Crost, B and D I Rees (2013), “The Minimum Legal Drinking Age and Marijuana Use: New Estimates from the
124 NLSY97”, Journal of Health Economics 32(2): 474–476.
125 Hinnosaar, M and E Liu (2020), “Malleability of alcohol consumption”, VoxEU.org, 18 October.
126 Lindo, J M, I D Swensen and G R Waddell (2013), “Alcohol and Student Performance: Estimating the Effect of Legal
127 Access”, Journal of Health Economics 32(1): 22–32.
128 Miron, J and E Tetelbaum (2007), “What Europe can learn from US policies on drinking and driving”, VoxEU.org, 5
129 September.
130 World Health Organization (2018), Adolescent Alcohol-Related Behaviours: Trends and Inequalities in the WHO European
131 Region, 2002-2014.
132 World Health Organization (2019), Status Report on Alcohol Consumption, Harm and Policy Responses in 30 European
133 Countries 2019.