Monday, September 08, 2008

McCain's Life Expectancy

John McCain has a 20% chance of dying before his first term would be complete, and a 44% chance of dying before the end of his second term.

At least, that's my best calculation according to actuarial tables. I'm not an expert at these kinds of calculations. And it should certainly be acknowledged that no one individual life is is actuarially accurate -- there are too many variables relating to genes, life experiences, diet, medical care, etc. In fact, I highly suspect these variables overwhelm any error I might have made in my calculation above.

I actually asked the American Academy of Actuaries this question, but they played it safe and wouldn't answer, saying

Mortality tables are used to project the life expectancy potential of large groups of people of the same age and gender and are not meant to predict the life expectancy of one particular individual. There are hundreds of additional questions involving family history, medical history, lifestyle decisions, etc., to which we cannot possibility know all the answers. And even if we did know all of the risk factors for a particular individual, some aspects of mortality are simply random. Therefore it does not make sense for us to try to look at a table and predict one person's life expectancy.

Fair enough. But, of course, many companies do exactly this -- apply actuarial tables to individuals. Isn't that the whole point?

5 comments:

crf said...

Could you give an example of what you mean in saying companies apply actuarial tables to individuals?

When a company uses these tables, I'm sure it only makes sense if they are large company, applying this use to a large number of workers. For example, they might look at the average age of their population of workers, and make certain decisions on managing their pension fund. (If I'm being cynical, they might use it to justify preconceptions: for example, to avoid hiring an employee who is well qualified, but is a smoker, or overweight. But this is likely a statistically unjustified misuse.)

A calculated average risk of someone in a category (like McCain's category) dying within so many years reasonably applies only to a sample from a large population. A sample of one random individual will have high variance. So it isn't about McCain's individual risk at all, and improper to say it is McCain's risk. It is about the average risk in whatever population McCain is in.

Also, let us remember: he is in lots of different possible populations: including the one you mention in the actuarial table. One could invent all sorts of populations McCain belongs to, and calculate their risks of dying. What if those risks are different? Why should one calculation be seen as better than the other?

Trying to calculate McCain's particular risk of death is, I think, a misuse of statistics.

(BTW, I'm not American, I am quite uneducated, and I don't wish for McCain to win :D )

David Appell said...

An example: if I apply for life insurance, my rate is partly determined by actuarial data. That is, they are applying group data about the averages of people "like m e" to me as an individual.

Anonymous said...

Some factors to include: McCain will be 72.5 years old in January when the swearing-in of the next President takes place.
McCain's father died from a sudden heart attack at age 70.
McCain's grandfather died unexpectedly from a sudden heart attack at age 61.
He has survived four skin cancers (melanomas), including one in 2000 that was classified as Stage IIa.
ex-smoker.
Tortured in Vietnam.
Multiple broken bones.
Is able on good days to perform light exercise.
Premature graying.
Inability to remember recent events.

crf said...

[quote]An example: if I apply for life insurance, my rate is partly determined by actuarial data. That is, they are applying group data about the averages of people "like m e" to me as an individual.[/quote]

No company sells insurance to just one person. The company is making an expectation not only on your life, but also an expectation on all the other people in your cohort it sells to at the same time. The expectation made on your life cannot be separated from that made on all the other people in your cohort, neither by you, or the insurance company.

Anonymous said...

Obviously, one cant predict an individuals life expectancy.What I would like to know is if you took 10,000 American men 72 years of age, without a known fatal illness,within 4 years , how many would die and how many would die in 8 years. This would give an estimate of the chances of the vice president becoming president.Do the same for a 44 year old.