Yet another causality fail

Think about couples where spousal abuse is common. Would you expect that partners in those dysfunctional dyads might just be a bit different from other folks, on average? Perhaps more impulsive and higher rate of time preference? Would it be surprising then to find on survey data that abusive spouses also exhibit other behaviours consistent with impulsiveness?

The Press happily takes the correlation as causal:

Binge-drinking couples are twice as likely to be involved in physical aggression towards each other, University of Otago research has found.

Professor Jennie Connor of the department of preventive and social medicine in Dunedin said it was the first time the link between partner aggression and drinking has been studied in New Zealand.

Researchers studied all forms of physical aggression, as violence in partnerships commonly escalated from less severe aggression, she said.

The study found if one or both partners had a pattern of heavy drinking episodes - or binge drinking - then physical aggression was more common.

...

They were also more likely to report that their partner had been drinking when physically aggressive towards them and this situation was associated with the highest levels of severity, anger and fear.

"Making changes to the price, availability and promotion of alcohol to reduce the amount of heavy drinking across the whole population will be a good start to reducing the frequency and severity of physical aggression in New Zealand homes,'' Connor said.

The full paper's here. And there's absolutely nothing in the study design that allows for causal inference. The correlation that winks more suggestively at causation is that spouses experiencing violence report that such instances are more frequent when the partner has been drinking heavily. And it's likely that some of these reports will involve violence that would not otherwise have taken place. But other instances of violence could easily be shifts in when the incident takes place: a displacement effect rather than newly induced violence.

Professor Connor's press release suggests policy changes:

“Making changes to the price, availability and promotion of alcohol to reduce the amount of heavy drinking across the whole population will be a good start to reducing the frequency and severity of physical aggression in New Zealand homes.”

It would be nice if she had any evidence of the elasticity of heavy drinking with respect to these policy instruments (hint: for price, it's very low).

HT: ed.co.nz