Unhealthful News 162 - Total cost of smoking calculation, revisited

A couple of days ago I wrote some general overview points about putting a dollar measure on the gain or loss of something "invaluable" like a life-year or a bit of nature.  This is needed for doing cost-benefit analysis calculations that allow us to compare apples to oranges when making decisions, something that we absolutely have to do.  As part of that, I commented on some economists' calculation of the total cost of smoking, measured per pack of cigarettes, adding up the health costs, the production costs, and so on.  This was based on a 2004 book, but one of them was blogging about it this week.  I suggested there were some serious problems with their approach, but I wanted to wait for the last in the series of posts about it to be sure.

In my previous post I mentioned that this whole exercise is inappropriate, and can only create results that are misleading, since smoking is a case of a consumer decision where the consumer is paying the costs.  Those calculations where we try to assign values, as best we can (which is often not very good), to goods that are not traded in markets and perhaps are "invaluable" are a necessary evil for policy making where someone needs to make tradeoffs on behalf of society because individual market decisions are not possible.  But it make no sense for private market decisions.  If I calculated that the costs of riding motorcycles exceed the benefits, considering the health risks and assigning a number to the non-market goods that are someone's life and health, and told you that therefore you do not want to ride a motorcycle even though you think otherwise, I trust you would laugh at me.  Why would you pay any attention to my calculations of the average costs and benefits, and in particular what value I think you should place on your own probabilistic death or injury, when you can make your own decision based on how you actually value the benefits and costs.

Smokers themselves pay the biggest single contributor to the cost of smoking, which is between 90% and 100% of the total health costs (depending on whether you believe anti-tobacco's claims about the effects of second hand smoke, or the claims of those who say there is no proof of any risk, or something in between).  They also pay for the production and distribution of the product via the purchase price.  Moreover, in most rich countries they pay taxes that more than cover the external costs – i.e., those that are foisted on someone other than the decision maker.  The external cost is primarily the much talked about, but actually fairly small (and negative by some estimates) net additional health care costs.  So, smokers are basically paying the total costs, and we have a functioning market; so a calculation aimed at making decisions in the absence of a market simply makes no sense.  It makes no more sense than calculating the net costs and benefits of making a new computer game.  This is something else that could be very roughly estimated if needed, but it is not needed – the market and consumers balancing their options in life can figure that out just fine. 

There are some other glaring flaws in the calculation too.  The costs that they claim are suffered by the smoker's family, predominantly the spouse's risk of earlier death, are estimated to be well over 20% of the health costs to the smokers themselves.  But even the clearly overestimated party line numbers for the impact of second-hand smoke puts the total impact at only a little over 10% of that experienced by the smokers (e.g., the oft repeated numbers for the US claim about 450K people die from smoking annually, with about 50K of that being from the second-hand smoke).  They also make claims about lost productivity from sick-leave and what is presumably smoke breaks.  But they say nothing about the increased productivity that smoking provides for people whose brains work better on nicotine.  Also, there is an important ethical question about whether someone "owes" society their maximum possible productivity, since employers choose to pay for only the actual productivity.  Finally, as a commenter on my previous post pointed out, the cost of the smoke break is caused not just by the smoking but by the rules that force the smoker to travel to a distant place to smoke; however much those rules might be justified, they do cause this productivity loss. 

But the point I wanted to expand upon a bit today is the double counting that comes from the estimate of $100,000 per life year lost prematurely.  This is obviously not an actual price tag.  In fact, despite the apparent confusion of people who use numbers like this a lot, it has no inherent meaning whatsoever.  It is just a cutpoint for deciding when a policy is worthwhile and when it is not.  If saving a statistical life year costs more than X then we can more efficiently devote our resources elsewhere instead, and $100,000 is an estimate for the right value of X.  The math is such that this number can be used in cost-benefit analysis calculations as if it were a real price, but it is not.  It does not matter that we use observations like how much people are willing to spend on safety or demand as a premium to work risky jobs – how they value their own statistical life, as it were – to reach a decision about the number.  The number still exists simply because it is what we have decided to be the right way to make decisions, and while we call it "the value of a life year" it really does not have a natural inherent meaning.

This is a very good thing for those of us who want to use numbers like this, because figuring out everything an extra life year entails, and pricing out each bit, would be impossibly complicated.  A life year is a complicated bundle of enjoying life, providing enjoyment to one's companions (these first two are by far the most important), as well as production that benefits society and consumption that takes away from society, further complicated by lower order costs and benefits (taking away a job from an unemployed young person, contributing to crowding and traffic accidents, imparting the wisdom of age, being a selfish old person with antiquated views who votes against public school funding and progressive policies).  Obviously these vary dramatically by person.  So instead of trying to figure all of these out, we take an aggregate number and just go with it.  At the very least, using the same number across policy decisions (e.g., for both air quality rules and medical expenditures) makes our decisions more efficient, even if we might disagree about what number to use.

What the authors of the book and blog posts about smoking did, however, was take the number $100,000, and declare it to be just the value of the life year for its own sake, the enjoyment of life (presumably also including other people's enjoyment of that life).  So they claim this justifies adding back in earnings, which increases this by about 25%.  (Note: Earnings per se are not a benefit or a cost, just a transfer of money, but they are a proxy for productivity which is a benefit, and thus represent a loss when someone dies earlier.)  But this really does not work.  It is impossible to sort out the different components of the value of a life year, so you have to use an aggregate.  If you are adding in salaried productivity, how can we not justify adding in the value of teaching your wisdom to your grandkids?  If they wan to pick a larger number for the value of a lost life-year, fine, but they should own what they are picking and not try to sneak in more.

Furthermore, since the number is somewhat arbitrary it is necessarily extremely rough, so adding $100,000 to a calculation of productivity down to the last dollar makes no sense.  If they want to just declare that they are using a value for a lost life of $125,000, that would be fine (the $100,000 is at the low end of what most experts consider reasonable).  But what they did reminds me of the old joke, where a pupil asks his teacher, when they are visiting a natural history museum, how old those bones are and she replies "five million and three years old".  When the kid asks how she could possible know that, she replies "I asked a museum guide that question when I first visited and she said 'five million years', and since that was three years ago…." 

If the authors of the smoking cost study do not want to take the most practical course, to pick a single number, because they want to divide the figure between the smoker's own costs and what is externalized to others, they could do that.  The biggest factors in each part would require a rough quantification of the "invaluable" – of loss of one's own life's enjoyment and others' loss of family/friend – so both would be extremely rough.  But what they are doing by including the big number and then adding in others is (a) making a muddle of the fact that the big number properly includes those other values because it is impossible to separate them all, (b) making the "five million and three" mistake of calculating some values down to the last dollar when the biggest number is arbitrary with a factor of tens of thousands of dollars, and (c) trying to hide their important assumptions and value judgments behind a lot of seemingly impressive numerical details.

I guess this calls for another one of my "never trust anyone who…."

But you know what?  Neither my previous post nor this one has explored the biggest problem with the whole exercise.  I will come back to that sometime next week.