Thursday 14 June 2018

Gotta adjust for population

I know it's tempting if you want to run a scare story on how bad something is to just run totals over time and to ignore population growth.

But it's pretty poor practice.

Here's a piece over in Stuff. It's got the headline "New Zealand: Where alcohol is normalised - and that means more drinking". It doesn't have a by-line, presumably because the author is too ashamed to admit to writing it.

If you want to say more drinking, you either need to compare it to other places, or over time.

The piece has this chart on alcohol availability from Figure NZ. It gives total alcohol consumption, by quarter, from 2012 Q4 through 2017Q4.

It looks like this.

Now I don't know why they didn't go through March quarter 2018. That data's now up. Maybe Figure had an easy-to-find one that had that time range. Who knows. But that's the more minor problem. The more major one is that we've had substantial population growth since 2012.

So you'd find an upward track in total consumption of most stuff in New Zealand. More people means more consumption. Except this total track is pretty flat. I bet that means decreasing per capita consumption. Well, it's a sucker-bet because I know the data. 

Here's what the per capita data looks like, out of Infoshare, using population aged 18+ as denominator. 

Whatever hypothesis they're trying to run about normalisation leading to more drinking has to contend with the substantial decline in per capita alcohol availability since liberalisation with the 1989 Sale of Liquor Act. There was a mild rise through 2010, then it came back down again. I suppose you could base predictions of another increase on mean reversion or something 

And we're low to middling in the international per-cap consumption stats too. 9-ish litres per capita is around the middle of the OECD tables. If you want to look at overall global tables, maybe don't rank us against countries in the mid-east that will cut your head off for drinking. 

And I don't think this is just the web-person at Stuff chucking in some pretty(ish) charts. From the article: 
And alcohol sales continue to rise. In the 12 months to March, they reached $1.6 billion - a $200 million increase from last year, Statistics New Zealand figures showed.
Sales, in dollar terms, can go up for all kinds of reasons. You can spend more while drinking less by shifting up-market. You can have an increasing total dollar spend just with more people being here - and more tourists rolling through and drinking as they go. Or you could have actual increases in per capita alcohol consumption accompanied by increasing expenditure. You need to check the volume per capita stats, not the total expenditure stats. 

The article takes the latest SHORE work by Huckle as hook.
Boozing has become normalised in New Zealand, and that means it's likely we'll drink more - and at higher risk levels, new research says.

One of the study's authors, Massey University's Docter Taisia Huckle, said: "What does normalisation look like? It looks like New Zealand.

"We have a situation where alcohol is completely normalised in society, through advertising, marketing and availability, alcohol is reasonably priced."

School children could walk past three liquor outlets on the way to school or see advertising on social media, she said.

We're a high-income country - and that means our drinking frequency is higher than middle-income nations, according to the researchers, who studied drinking patterns across 10 nations in a report published on Thursday in the journal Drug and Alcohol Review.
Ok. SHORE and Huckle have a couple pieces in Drug & Alcohol Review.

The first one is survey work asking drinkers where they purchase their alcohol, at what time they purchase it, how long it would take them to obtain alcohol and, if under-aged, how often they're asked for ID and get served. It doesn't say anything about normalisation. It does show that New Zealand kids are the most likely to report being asked for ID out of the set of places surveyed, and are third least-likely to report being served alcohol. 

I don't get the link to normalisation in any of this. The article's Table 1 reports the "percentages of drinkers purchasing alcohol at an on-premise or take-away outlets across countries at least once in the last 6 months", but we have no clue about baseline proportion of drinkers - the table just says that Kiwi drinkers are a bit less likely than Oz drinkers to buy at pubs, more likely to buy at duty-free shops and at the cinema, and more likely to buy at clubs and restaurants. For a normalisation argument, wouldn't you need to show increasing proportions of people at different venues who consume alcohol there, rather than the venue choices among current drinkers?

The second one is more survey work comparing Oz, England, Scotland, Thailand, Peru, Vietnam and NZ. 

Table 1 shows that NZ has the smallest proportion of heavier drinkers, the second-highest proportion of low-risk drinkers (Peru has 74% low risk, NZ has 62% low risk) and the second-lowest proportion of higher-risk drinkers (Peru has 2% higher risk, NZ has 15%). I also don't see anything in there about normalisation. 

Upshot:
  • Always run a population correction for this kind of thing. It's absurd to point to total spending on something or total consumption of something as being a bad thing, unless that thing is bad in a total way. Total carbon dioxide emissions can go up despite drops in per capita emissions (say) and that would be bad because the harm comes from the total. Alcohol isn't like that.
  • The research forming the hook doesn't really say anything about normalisation, and instead shows NZ to have less high risk drinking and more low risk drinking than Australia, England, Scotland, Thailand, and Vietnam - but not Peru. 
  • I expect the journalist didn't sign the article out of embarrassment, and is right to feel shame. 

No comments:

Post a Comment