mortality rate for people aged 15-24 across the US in the same time period was around 9 per 10,000 during the same time period (averaging 150 for the males plus 30 for the females)

That should be 15 and 3. The figures on that page are given per 100,000, and I copied without dividing by 10.

However:

From http://www.baltimorehealth.org/info/neighborhood/47%20Sandtown.pdf, we see that in Sandtown, a community near me with an economy based mainly on crack that is one of the most crime-ridden areas in the nation, the mortality rate for people aged 15-24 was 30 per 10,000 in 2002-2006.

From http://www.cdc.gov/nchs/data/nvsr/nvsr57/nvsr57_14.pdf, we see the mortality rate for people aged 15-24 across the US in the same time period was around 9 per 10,000 during the same time period (averaging 150 for the males plus 30 for the females)

That gives us a risk factor increase of 3.33 from the bottom income to the average, which is over half our risk factor of 8.3 (because you multiply, not add, them; we only need another risk factor of 2.5 decrease going from the average to the top quintile to get to 8.3).

If you look on the next page of the Sandtown report, you see that homicide accounts for 7.4 deaths per 10,000 people. The fact that it has a percentage (years life lost / deaths) ratio > 3 indicates that almost all of these deaths are among the young. Also note that deaths from AIDS and drug overdoses are concentrated among the young, though not as heavily.

The rate of death, disease, and crime in our inner cities is mind-blowing to people who don’t live there. For example, 5% of people in DC have HIV. Remember that approximately 0% of people not in the “inner city” have HIV. There are areas where most people don’t have jobs, most people have STDs, AIDS is common, and most men will go to prison at some point in their life.

It would be interesting to compare the next higher quintile against the top, for both the absolute difference and for the trend in time. (Does the paper have a graph that gives death rates for all quintiles?)

Regarding the empirical implementation, lifetime earnings measures are
constructed using long averages of past earnings. For individuals older than 53, earnings
from age 41 to 50 are used to capture years when the person was most closely attached to
the labor market. For younger individuals, averages ranging from 5 to 10 years were
computed without including the immediately preceding 3 years (e.g., for individuals aged
43, earnings from age 31 to 40 are used).

The variable is badly named, but it is not as stupid as its name might lead you to believe. He is try to get at something like Milton Friedman’s “permanent income.” The author makes an attempt to deal with the reverse causation problem by carving out windows (the years right before death do not count), but I find it hard to believe that he is successful.

Death is an unusual thing for youngish men, and this is the group in which you get the eye-popping mortality ratio. How many deaths in that age range are not due to accidents, suicide, and homicide? Of those deaths, how many are not due to chronic, debilitating diseases (known and unknown)? Of the remaining, how many are due to diseases which have known behavioral risk-factors? Causal inference is really problematic in those three cases. I don’t see any discussion of cause of death on a preliminary read.

The idea is that, by getting rid of recent past and all future earnings, you are getting rid of simple reverse-causation stories for the results. But you are also getting rid of simple forward-causation stories. Income five years ago affects mortality today through what causal mechanism? Less health care consumed five years ago has a big effect on health today (despite the modest income elasticity of healthcare demand and the modest effect of healthcare consumption on heath)?

What the paper shows is something sort-of like “There is something permanent about people which is correlated negatively with mortality and positively with income.” The author seems to want that thing to be permanent income (which, I guess, for him, is distributed randomly by Nature), but it might just as well be IQ or conscientiousness or “drive” or whatever.

The conclusion also partakes deeply of the goof-ball normative pose that any inequality in health outcomes which happens to be correlated with identifiable demographic/social/economic groups is bad and any increase in the strength of such correlations is worse. Why these things are bad is usually left unstated. What the Whitehall studies and similar studies in Sweden tell us is that these differences will not go away or get smaller with universal coverage, though.

Do you think the author is against universal coverage?

I suspect there are many other relevant medical conditions producing rather a lot of medical details to track down. Cystic fibrosis life expectancy is now up to 36.8 years, so that is another group dying poor aged 35-49 that didn’t used to live even that long. The article is expensive, priced at $31.50, so I will not be reading it to find out how comprehensive the authors have been.

However, this reasoning should apply to virtually every factor that impacts health. If eating meat is even slightly better or worse for you than eating grains, there should be a big difference in death rates among meat-eaters versus grain-eaters, for relatively young people. I’m not aware of large differentials being the rule among this age group.

2: Cause of death is missing. Violence would have to be a big part of this.

3: High medical expenditures could give .75 years of low quality life. By my crude math that would be equal to 1% difference

4: Hold on a minute… Are these lifetime earnings, or expected lifetime earnings? Anybody who dies at 25 is going to have a lower lifetime earning that someone who lives to 75.