Ministry's figures need analysis (updated)

so says Eric Crampton in this morning's Press here in Christchurch. And he is right. The figures in question are the Ministry of Health's estimates on the health costs of smoking - $1.9 billion. Eric writes,

You could be forgiven for thinking that the health system could save $1.9 billion if tobacco had never existed. That's what the Ministry of Health says smoking costs the public health system.

But, you'd be wrong.

Not that the ministry would correct your error. Eric goes on to make the real point about the ministry's numbers,

The ministry's latest estimate of the cost of smoking has nothing to do with the costs that smokers impose on taxpayers or the costs that could be avoided if smoking were to disappear.

Rather, it's a politically convenient number whose promotion has much to do with gaining voter support for anti-tobacco initiatives and nothing to do with real economic costs.

Interestingly the ministry's number

[...] was much higher than the previous estimate of $350 million dollars - a figure produced not by the Big Tobacco lobby but rather by Des O'Dea in a report commissioned by anti-tobacco crusaders Action on Smoking and Health.

So, you may ask, how did the ministry get their figure?

After sorting the population by age, gender, income, ethnicity and smoking status, they then compared the costs of providing health services to smokers as compared to nonsmokers for each group.

The excess costs of the smoking group were tallied up to produce the $1.9b figure.

But there are problems with such an approach. Eric deal with two of the big ones.

It's easiest to think of smoking as bringing forward a whole lot of end-of-life costs.

Smokers die earlier than nonsmokers.

We know that.

And the costs to the health budget of somebody who is dying are rather higher than the costs of somebody who is healthy.

But everybody dies sometime and most of us will incur end-of-life costs that will be paid for by the public health system.

Suppose that a smoker will die at age 65 and a nonsmoker will die at 75. Comparing 65-year-old smokers to 65-year-old nonsmokers and calling the difference the cost of smoking then rather biases upwards the measured costs of smoking.

We ought to be comparing the health costs of a smoker dying at age 65 with the health costs of a nonsmoker dying at age 75.

And, perversely, the deadlier cigarettes are, the greater will be this bias. The younger smokers are when they die of smoking-related illnesses, the greater will be the measured cost difference between smokers and non- smokers because a smaller proportion of comparable nonsmokers would be incurring end-of-life costs.

The figures assume that in the absence of smoking, smokers would never have imposed end-of-life costs on the health system. But for their smoking, all smokers in this scenario would have died of a sudden, and cheap, heart attack and would only have had average health costs up to that point. That's clearly nonsense, but the $1.9b figure only makes sense if it's true.

Further, we might well expect that there are differences between smokers and nonsmokers beyond those accounted for by income, gender, ethnicity and age.

Imagine a 45-year-old white female of average income who happens also to be an active jogger, moderate drinker and health food enthusiast. Is she more or less likely to be a smoker than a 45-year-old white female of average income who happens also to avoid the gym, drink too much and never touch a vegetable?

On average, we'd expect that folks who are more health conscious on other margins are also less likely to smoke.

But the ministry's method, which doesn't correct for those other health-related behaviours, necessarily lumps all of the differences between smokers and non- smokers into the cost of smoking rather than into the cost of having a generally unhealthy lifestyle that includes smoking.

It's only if smokers and nonsmokers are otherwise identical, on average, in their health-related behaviours - after correcting for income, gender and ethnicity - that the ministry's figure holds up. But that's also pretty clearly nonsense.

So the ministry may not be telling us the truth, the whole truth and nothing but the turth. Anyone surprised? As Eric notes we should

[...] be as sceptical of numbers coming from the Ministry of Health as you would be of numbers produced by the tobacco industry. Neither is a disinterested party.

Well said Crampo.

Update: Eric writes more on the matter here.