These are some very enjoyable charts! I was expecting something gloomy along the lines of How We Die.
It is also a great opportunity to practice rationality. Some these charts show surprising patterns, with captions like "one line in this chart looks unusual!" It is easy to click to the next slide and see the explanation, which then makes perfect sense. Hindsight bias is quite alluring. I tested myself to see if I could explain the trends before they were explained to me, and did so-so.
To keep myself honest, I took notes as I paged throught he charts, and have pasted them (spoilers obscured) below.
"But one line in this chart looks unusual!"
Between 1983 and 1995 the 25-44 year-old cohort saw an increase in mortality, from 65 to 79 based on an index of 100 in 1968. The chart is not baselined at zero, but the 20% increase still seems a significant departure from the overall downward trend.
First theory: The crack epidemic. Violence and overdose. I would expect the below-25-year-old cohort would reflect this as well, but that group enjoys a pretty steady improvement from 1970 to 2010.
Second theory: War? But that also (at least in the Vietnam era) affects mostly the below-25 group. And what war? Seems a stretch.
Third: Disease? Only AIDS stands out during those years, and the age group seems a good match. But did AIDS really kill so many people to upset the prevailing trends? This seems like a stretch.
Fourth: Accidents? Can't think of any pattern that makes sense during those years.
Fifth: Statistical anomaly? Something like how cancer deaths are increasing, but this is really the flip side of good news: that people are getting better at avoiding other illnesses and accidents, and something is going to get you in the end. But for this relatively young cohort, it is hard to see how something was not killing them more during those years.
Best guess: Crack caused 2/3 of the effect, AIDS 1/4, and 1/12 is due to random variation.
Another oddity is that Americans aged 45-54 have made surprisingly little progress in mortality since the late-1990s.
Analysis: Does this have to be explained? No change can be the result of an absence of cause. But there is a happy downward trend from 1970 to 1998 that abruptly ceases.
From 1970 to 1980 the line went from 97 to 77, dropping 2 points per year. From 1980 to 1990 it dropped to 63, 1.4 points per year, so the improvement was already waning. In 2000 it was 56.6, dropping 0.6 points per year (when perhaps we would have predicted 0.8 if we projected a similar change in the rate of change).
If we continue to adjust our expected rate of change, we would apply the average of the observed declines in the rates: 2.0, 1.4, and 0.6, which is 0.7. So in the 2000s we would expect a "drop" of -0.1 per year, or an increase of 1 point of mortality by 2010, arriving at 57.6. In fact the number is 54, so the picture might be better than expected, or at least within expectation.
The decline was not linear from the start, the rate of improvement has been decaying all along. A little curve-smoothing around 1997 would make this more apparent.
The next slide adds to the confusion. It shows that cancer and heart disease have become "much less deadly" over the years for this cohort (though the cancer band seems to have pretty consistent thickness to my eye) (mousing over shows the raw numbers, which do indicate significant change). But on this slide there is a noticeable increase in death after 1990. This chart is "number of deaths" while the previous slide showed mortality per 100K people. So the size of the cohort is important. Did a lot of people with good hearts and low cancer risks get killed by AIDS just before, leaving them vulnerable? No -- we need more individuals around to explain growth in the raw number of deaths. Sliding back and forth, I see almost every cause of death (maybe not AIDS) took more lives by 2010 than it did in 1990. There simply must have been more people around. This must be the baby boom.
The next slide implies that suicide and drug deaths are an important factor, because these rates have increased, but the previous rainbow slide showing all causes makes it obvious that these were minor contributors to the big picture.
Motor vehicle deaths were up and down since the 70's, but trending slightly down. Are seat belt and air bag laws visible in the chart? Seat belts came first, probably in the 70's. There's a big improvement after 1973, but mostly reversed by 1979. Air bags appeared in the 80s. 1988 is the start of an improvement, but it's really just a statistical local high.
The real question is what happened since 2005? The total death count dropped from about 46,000, close to the historical low, to about 36,000 in 2010. Did this trend continue? Does it reflect fewer passenger miles, more air travel, more biking and public transportation? More expensive gas?
Suicide is tricky. Number of deaths is misleading, we should consider number of deaths per capita. And there may be a balancing effect here too. "Cancer is up" sounds bad, until we realize that accidents and violence are down, leaving more people around to get cancer. Some number of suicides occur among people who would have died in earlier years, but have lately been preserved from accidents and disease-caused death, and perhaps are being maintained in life but very ill.
If we assume that an eventual death is mandatory, and ask ourselves what form we would prefer that death to take, a self-directed exit after deliberation seems hard to criticize. This is more palatable if we name "depression" or something similar the cause of death for those who take their lives while still possessed of a healthy body.
Cause of death among those aged 75+ is probably not so simple to categorize, as is done here. If an Alzheimer's patient suffers a fall and dies from a heart attack, where do they fit?