Scaring seniors with ... living too long?

I enjoyed reading this NPR article discussing retirement saving for some insights into how individuals and their advisors view and plan for spending needs in retirement. The president of EBRI, Dallas Salisbury, describes his "Rule of 33," which is to have assets (including Social Security) equal to 33 times your annual spending before retiring.  That sounds adequate for roughly 33 years of additional life.  But if average asset returns are perhaps 3% in real terms and if typical consumption growth is 2%, a quick calculation reveals that 33-fold buffer to be good for more like 40 additional years of life.

Needless to say, there's plenty that can cost a bunch in old age, namely out-of-pocket medical or nursing care. But I think the factor of 33 may be a little on the high side.  For the average American male born in 1950, average years remaining at age 65 (in 2015) are 17.6, according to Social Security cohort forecasts, while they are 20.3 for women.  The median remaining years are about 20 and 21, which is another way of saying that such individuals have a 50% chance of living to 85 and 86.

To be sure, education, income, genetics, smoking, and many other things can either increase or decrease one's survivorship probabilities.  But I was really shocked to see what must have been a flat-out error in Mr. Salisbury's quote about his own parents' survivorship:

"They were in their mid-80s, as old as they ever expected to live. But he pulled out his computer to show they had a 50 percent chance of living until age 99."

This seems exaggerated to me. In the figure below, I've plotted survivorship probabilities for men and women in the 1920 and 1930 birth cohorts, and in the 2000 period life table.  All are from the SSA data, and I grant that there are subgroups that will live longer and shorter than those shown here. There are some differences between these groups, but you can see the rapid reduction in survival starting at age 85.  The median age at death for these groups is around 91, and only about 10% live to age 99. I'm not sure what Mr. Salisbury had on his computer, but I can't imagine that any life table would predict such a huge difference in survivorship probabilities --- something like 40 extra percentage points --- for even the most healthy seniors at these ages.

UPDATE:  Dallas Salisbury clarifies --- he meant that starting from age 65, the probability of at least one spouse surviving to age 92 is 50% in actuarial life tables he had obtained from an insurance company.  Surprising, right?  Perhaps that estimate is a little optimistic (or pessimistic, depending on your point of view); compared to Social Security forecasts it is.  But it's a sobering look at the very real risks of living too long.  There is wide variance in length of life.