# Econstudentlog

## Alcohol consumption, a few numbers

Two great graphs, from gnxp (there’s more at the link):

The ‘wordsum’ variable above is a vocabulary test; a higher score indicates that an individual has a more extensive vocabulary.

This report (pdf) concludes that in Denmark, “there’s no (/significant? /clear? /obvious?) connection between average alcohol-consumption and length of education.”

(“Der er ingen tydelig sammenhæng mellem gennemsnitligt alkoholforbrug og
uddannelseslængde, men umiddelbart har skoleelever det højeste gennemsnitlige forbrug.”)

I’m not sure that’s the right conclusion to draw and they make no attempt to justify it, they just state it as if it was a fact without any further comments. Here’s a graph of the results:

I have excluded from the graph the two variables ‘skoleelev’ and ‘anden skoleuddannelse’ because I have no clue precisely what those varibles mean and who they include (both relate to people somewhere in the educational system, but that’s all I know). Does a high school student belong to the ‘skoleelev’ or the ‘anden skoleuddannelse’ segment? When is something a ‘skoleuddannelse’? I have no idea. Stuff like that can just really make me mad – the communication here is so damn poor that I’d rather throw away the results than post them here. Anyway, I was surprised that people with the shortest education drink (much) less than average and less surprised that people with a long education drink more than average. I’d assumed that the relationsship between education and alcohol consumption was v-shaped. Maybe selection bias is at work here, I don’t know, but these data certainly do not support the hypothesis of a v-shaped relationsship. The reason why I believe their conclusion (‘no connection btw…’) is incorrect is that there do seem to be a relationsship between education and alcohol consumption; it looks like people without much schooling drink much less than average and that people with a relatively long education drink more than average. No, you can’t draw a straight line, but who says the ‘relationsship’ should necessarily be a straight line? The fact that alcohol consumption varies widely across education levels is a clear sign that the two variables are in fact connected, even if we don’t know precisely how they are connected. Note that these numbers are not age-adjusted and that I’ve taken out people from the educational system. There are a lot more 20-30 year old plumbers who’re no longer in the educational system than there are people with phd’s. That’s another way to say that the group of people with long educations should actually be drinking less than the other groups, all else equal, because they’re on average a few years older (and ‘older’ here means ‘drink less’ for the relevant age demographic).

The report has a lot more data. Some of the results are quite surprising. Here’s one bit that took me by surprise: ‘41,9 % of the participants think that it is ok to drink alcohol in order to become inebriated. 58,1 % do not think it is ok.’ (41,9% af deltagerne i denne undersøgelse mener, at det er i orden at drikke alkohol for at blive beruset. Tilsvarende mener 58,1% at det ikke er i orden.). Ok – so more than half don’t think it’s ok to drink alcohol in order to get drunk? Seriously?

May 2, 2010 - Posted by | alcohol, Data, Demographics

1. Undersøgelsen virker ikke ret repræsentativ. Kun 57% har svaret tilbage, og allerede her undersøger man ikke befolkningen generelt, men den del af befolkningen der frivilligt bruger tid på at udfylde et tilsendt spørgeskema.

En lidt far fetched teori kunne være, at de lavt uddannede der svarer på et sådant skema, er de der ikke drikker så meget. Højere uddannede er måske bedre i stand til at have et vist alkohol forbrug, og samtidig interesserer sig for deltagelse i sådanne ting, mens det hos de lavtuddannede er en mere specifik type mennesker der gør dette? Eller sagt kort, de lavtuddannede med et højere forbrug, vil ikke have “ressourcer” til at deltage i denne undersøgelse, mens det samme ikke i så høj grad er tilfældet for højere uddannede?

Comment by polyb | May 2, 2010 | Reply

2. @Polyb:

Jeg skriver faktisk specifikt i relation til sammenhængen at: “Maybe selection bias is at work here…” – den bemærkning dækker netop over det forhold at den adspurgte sample sandsynligvis ikke er repræsentativ. At jeg ikke brugte flere kræfter på det forbehold i posten skyldtes vel nok primært, at jeg var mere interesseret i hvad de tilgængelige tal nu engang fortalte, end jeg var i at spekulere over, hvorfor tallene måske er misvisende. Jeg ved meget lidt om dansk alkolholkultur.

Når det er sagt tror jeg ikke, der kan være stor tvivl om, at den andel af de adspurgte, der ikke svarer, sandsynligvis har et højere alkoholforbrug end resten, og samplen er sandsynligvis også trunkeret, i og med at jeg er ret sikker på, at der ikke findes nogle individer med et dagligt indtag på, eks., 12 genstande, som har deltaget i undersøgelsen. Jeg ville så i øvrigt endvidere også have en formodning i udgangspunktet om, at de forskellige uddannelsesgruppers angivne antal genstande til en hvis grad afspejler det, der kunne betegnes den pågældende gruppes sociale accept af alkoholindtagelse; det er mere acceptabelt for folk med højere uddannelse at indtage alkohol, bl.a. fordi det er finere at drikke cognac end papvin eller billige øl og fordi de højere socialgrupper i uddannelsessammenhæng tilsyneladende etablerer det, der vist roligt kan kaldes et relativt afslappet forhold til alkohol, relativt tidligt.

Inddragelsen af de danske tal i posten var mest tænkt som et sjusket forsøg på at lave en grov sammenligning med de amerikanske tal, også selvom det ikke er helt det samme, der måles på, med fokus på uddannelsesvariablen. Eftersom tilsvarende selektionsproblemer kunne forventes at eksistere i de amerikanske tal mente jeg ikke, at det var så uproblematisk endda bare at bruge de tilgængelige tal uden at inddrage de mange nødvendige forbehold eksplicit.

Comment by US | May 2, 2010 | Reply

3. @US
Jeps, det skulle ikke forståes som et angreb på din metode – blot et lille bidrag til hvad der måske kunne ligge bag tallene. Jeg er ret enig i dine betragtninger 🙂

Comment by polyb | May 3, 2010 | Reply

4. Jeg er heller ikke så overrasket. Dels har jeg set folk i hjemmeværnet med typisk korte uddannelser drikke så godt som ingenting, mens jeg kender masser af akademikere, der drikker en masse vin. Jeg har før skrevet om disse iagttagelser på min blog.
Jeg skyder på to sammenhænge:
1) hvis du forlader folkeskolen og begynder at arbejde på en fabrik eller et andet sted, hvor du skal møde klokken seks om morgenen, så kan du ikke have den samme drikkekultur som folk, der har første forelæsning klokken 10.
2) Den anden ting er, at det også er en kultur. Arbejderklassen er måske lidt bange for alkohol, og forbinder det med social deroute, mens der er en udbredt romantisering af vin og kunst, noget livsnyder-iscenesættelse og overklassekultur over at kunne have et forbrug af ‘god rødvin’ og spiritus, specialøl o.lign.

Comment by Ulla Lauridsen | May 5, 2010 | Reply

5. Jeg kan ikke finde kilde lige nu, men tror man skal tjekke for andre rusmidler, før man drager for vidtgående konklusioner om ruskulturen. Fx har jeg set undersøgelser, der peger på, at ting som amfetamin findes som misbrug blandt folk med hårdt fysisk arbejde, fx en byggeplads, efter alkohol i senere år i stor stil er blevet bandlyst fra danske arbejdspladser. Men funktionen var den samme: at dæmpe smerterne.
Hvis man møder klokken seks kan man i øvrigt nå at drikke sig fuld inden aften og sove den ud.

Comment by info | May 5, 2010 | Reply

6. @Ulla:

Info er også inde på det: De to overlapper. Hvorfor er det forhold, at man møder tidligt automatisk et udtryk for, at man skulle drikke mindre? Med mindre man arbejder flere timer er der nøjagtigt lige mange timer at drikke sig fuld i, i løbet af et døgn. Og lavt-uddannede arbejder på det aggregerede niveau færre timer end højt-uddannede, ikke flere – de har flere timer at drikke sig fulde i, men færre penge at gøre det for. Dine bemærkninger i 2 overlapper med mine i kommentaren til Polyb. Bemærk især her, at hvis der er mere skam forbundet med druk i de lavere sociale lag, hvad jeg godt kunne forestille mig, vil disse individer kunne forventes i højere grad at lyve om deres forbrug, hvis de indtager relativt store mængder alkohol, end deres mere veluddannede sammenligningsobjekter – det var bl.a. derfor, jeg i min kommentar til Polyb var inde på at antallet af genstande også fungerer som et estimat for den sociale accept af alkoholindtagelse i de forskellige sociale lag. Der er mange skjulte effekter af potentiel interesse her.

@info:

Det er en god pointe at få med at der findes substitutter til alkohol, og at disse ligeledes bør inddrages i en meningsfuld analyse. Jeg skal ærligt indrømme, at jeg ikke har nogen anelse om, hvilke karakteristika forbrugere af illegale rusmidler besidder, og hvordan det marked er bygget op.

Comment by US | May 5, 2010 | Reply

7. Probably tangentially related, since I cannot read the comments in Danish, but… have you seen this map? I know from personal experience (hangovers and all) that Bulgaria is not a wine country – I’d estimate it’s 45% distilled liquor (traditional “rakiya” – plum brandy), 45% beer, and 10% wine… But it also puts Denmark in the beer belt. I am a fan of Carlsberg (Tuborg is hard to get around here), but still… really? I always pictured (perhaps wildly inaccurately – apologies) Danes as a vodka people.

“No, you can’t draw a straight line, but who says the ‘relationsship’ should necessarily be a straight line? The fact that alcohol consumption varies widely across education levels is a clear sign that the two variables are in fact connected, even if we don’t know precisely how they are connected.” – Bingo. Someone was too lazy to run a multivariate probit or a bayesian regression, or they did, and did not like the results. An interesting side note: can the results be skewed by poor and non-drinking Muslim immigrants? I admit to near-ignorance about Denmark’s minorities, but it sounds like a plausible explanation…

Comment by Plamus | May 7, 2010 | Reply

8. @Plamus:

Yes, I’ve seen that map before. Back then I thought that Denmark was probably in the right place, it was my impression that Denmark is a ‘beer country’, though there’s probably quite a bit of variation in the consumption patterns of different groups of consumers.

I just decided to take a look at some numbers from the Danish Bureau of Statistics in order to tjeck whether my intuition was right. In 2008, Denmark consumed 520 million liters of beer, 184 million liters of wine and 25 liters of distilled beverages. In terms of pure alcohol, the corresponding numbers are 24, 21 and 8. So there’s no doubt Danes drink a lot more beer than they drink other types of alcoholic beverages, even though in terms of units of alcohol, wine is probably not all that far behind.

“Someone was too lazy to run a multivariate probit or a bayesian regression, or they did, and did not like the results.”

I can tell you without a shred of doubt that the authors of that study had likely no idea how to set up a probit model. Most likely it was neither laziness nor selectivity, it was ineptitude. They probably hadn’t even realized that doing something like that was possible. There’s nothing in there that indicates that the authors even know what a statistical distribution is. They basically just report the data and every now and then draw conclusions from those without ever doing any sort of statistical analysis (besides perhaps calculating the average).

I had completely overlooked the effect you mention in the end and you are right – that might be the explanation. There are very few Danes under the age of, say, 50, who’ve gotten less than 10 years of education, whereas that’s very often the case for first generation immigrants, so even a relatively small amount of uneducated immigrants can bias that stat quite a lot, especially if their consumption patterns diverge a lot from that of their Danish counterparts; it does sound like a plausible explanation. None of the Danish commenters mentioned this: I consider it somewhat interesting that I had to have that likely effect pointed out to me by a foreigner.

Comment by US | May 7, 2010 | Reply

9. I cannot take all the credit you ascribe to me – when working with data about the US of A, one of the first things you learn is to account for effects, biases, etc., that various minorities introduce. For my first ever project in econometrics, I chose to try to explain violent crime in the US based on a simple regression of a bunch of variables, and found out that if you exclude immigration data, income and education were your best predictors. If you accounted for immigration, then ethnicity knocked out income. Colinearity and all 😉 – basically ethnicity and income were too correlated. The lesson stuck…

I also keep another lesson in mind: a story I read back maybe 7-8 years ago, written by an Iranian immigrant to Sweden. She basically described how shocked she and her family were to come to Sweden and be given a luxurious (by their standards) apartment and a generous monthly income – some kind of a “stipend”, if I remember correctly. The part that struck me – some 3 months later a social worker came and told them that since they had caused no trouble, their stipend was going to be increased by 50% or so. I readily admit to ignorance about how most of Europe works, even though I grew up in what’s now a part of Europe; I just try to learn and apply the lessons learned on the other side of the Big Puddle…

Cheers!

Comment by Plamus | May 11, 2010 | Reply

10. Good comments. I’m pretty sure dealing with minorities and how they affect the data is a bigger deal for someone living in the US than it is for someone living in Denmark. It only recently became (somewhat) acceptable to even ask such questions. If you asked if Muslims committed more crimes than Christians in the 90’es and maybe even set out to go have a look at the data, you were either a troublemaker, a racist or worse. So very few people did. Lack of good data is also a problem; for instance, if I remember correctly it is still illegal for the Danish Bureau of Statistics to register people based on religious affiliation.

Sweden is not an all that good country to extrapolate from when it comes to the situation in the rest of Europe(/EU). They have a very liberal immigration policy and their influx of muslims (which is, aside from the occasional debate about Poles and people from Eastern Europe, basically what Europeans talk about when they discuss this subject) is somewhat higher than that of Central-European countries (and ie. Denmark). They also have a lot of problems with the immigrants they have, something that is obvious to anyone who’s taken a closer look at ie. their crime data, although many Swedes seem to refuse to even talk about these problems.

The Danish immigration policy looked a lot like the Swedish until 2001, where the election results made it pretty clear to the politicians that a lot of voters were unhappy with the way things were going. I don’t believe we ever actually rewarded immigrants monetarily for not behaving like criminals though.

Comment by US | May 12, 2010 | Reply