Lying with stats - alcohol and violence edition

[andrewducker]

Lots of headlines over the last few days about the fact that violence is down, and that alcohol is more expensive and ARE THESE THINGS CONNECTED?!?!?!?!

Which would be pretty exciting if true. So let's take a look at the stats:


Ok, violence definitely down, and as alcohol is less affordable that could be making a difference, right?



Oh.

So, alcohol is _massively_ more affordable than it was in the 80s, and more affordable than it was even in 2000, but because it's less than it was at the peak, journalists are happy to blame the violence on alcohol? (And yes, the alcohol curve starts in the 80s, while violence is only showing the last decade, but even peering closely at the alcohol affordability for the last decade I can't see anything which looks anything like correlation. Particularly as the alcohol affordability peaked in 2008 and violence...didn't.)

This isn't to say that there is _no_ link, but the sheer eagerness of the newspapers to leap onto this is incredibly annoying.

Both graphs from this Full Fact article on Violence and Alcohol.

Comments

[cheekbones3]

Journalists can't cope with trends that are more complex than, say, a straight line, possibly with one kink in it. As for correlation/causation, ha! Usual behaviour is to parrot back what we tell them has happened, or if it's a big story and they have an agenda (or a lack of one), focus on the one thing that might cause a row, no matter how tiny a nuance.

^Slightly jaundiced view, but if you ever wonder why statistical news releases have tended towards the extremely equivocal/anodyne in recent years, this could be why. However, we're being encouraged to editorialise more in recent times, which may help in some circumstances.

[marahmarie]

Journalists can't cope with trends that are more complex than, say, a straight line

I find that a rather stunning claim. A true journalist, and there are a few, seek the truth and tell it as they see it (whether it happens to draw itself out in straight lines or not). What you're talking about is more like the very thing you say might help: editorialization/clickbait, both which suit themselves to becoming virtually bottomless breadbaskets of nothing but straight lines tailored to the journalist's exact liking, and both which are becoming more common these days as true journalism becomes a dying art.

[cheekbones3]

I was unfair in my generalisation to an extent. Someone like Goldacre for example is brilliant at reporting and using stats, and will tease out the truth where possible. Many bog-standard journos however simply don't have the skills to interpret charts, or to understand any deeper nuances, and there is an ongoing programme to address this problem, to an extent. However, it requires buy-in and is most definitely not a quick process.

Maybe the other side of the coin that we can do little about are the journalists who will peddle the same line no matter what the numbers are saying, and worse, the people/publications who wilfully misrepresent. At least we now have some power to publicly censure such behaviour.

[major_clanger]

In my experience as a barrister, both in criminal and family cases, a very substantial portion of incidents of violence involve alcohol. But - and it's a bit but - that does not mean that violence correlates with alcohol availability.

My view is more that there is a fraction of the population that (a) likes drink, and (b) is more prone to get violent when drunk. Such people will tend to get drink unless we make it prohibitively expensive (and probably will even then) and have always done so, so the general increase in alcohol availability isn't affecting them. I've often seen criminal records showing long histories of alcohol-related offences, or social work reports that indicate a lengthy history of alcohol abuse. So I suspect few people who commit offences while drunk have recently started drinking; they've probably been doing so for years.

As such, it's no surprise at all that there is no upward trend in violence despite the increasing availability of alcohol. Most of the people who will be violent when drunk have been drinkers for years. Rather, the gentle decline is a result of better and more focussed policing and the introduction of ASBOs and/or alcohol treatment orders as a condition of community sentences or suspended sentences.

[redbird]

I think this may also have to do with how people are used to reading graphs. The second graph is labeled "affordability of alcohol," but at first glance it looks like hundreds of "price of XYZ" graphs, in which the curve moving up means the item is more expensive, not more affordable.

This probably isn't a case of deliberate lying with graphs, but journalists aren't trained statisticians either. (Most of us aren't, and I'm certainly not.) So I suspect that what most people are seeing is "the crime rate curve is higher, and the alcohol price curve is higher," before or instead of realizing that the second curve isn't the price of alcohol, but a calculated measure of affordability.

(This feels a little like trying to figure out what percentage of the increase in "billion-dollar weather disasters" is due to changes in the weather, what percentage is due to people building more expensive buildings in vulnerable areas, and what percentage is due to inflation.)