Angus deaton

Age 50–54

500

510

480

470

490

2015 dollars

2015 dollars

Age 65–69

1,500

1,600

1,700

1,800

28,000

30,000

26,000

24,000

38,000

40,000

36,000

34,000

1,900

2,000

2000

2005

2010

Year

1995

2000

2005

2010

Year

1995

Mortality 

(left axis) 

Median household income 

per member (right axis) 

---

422

 

Brookings Papers on Economic Activity, Spring 2017

women age 50–54 together (dashed lines). Mortality and income match 

closely. The bottom panel shows mortality for the age group 65–69, and 

median real income per member in households headed by someone in this 

age band. This older group has done well since 1990 in part because, for 

those who qualify, initial Social Security payments are indexed to mean 

wages and are subsequently tied to the Consumer Price Index; mean 

wages have done better than median wages. Real incomes for those 

age 65–69 increased by a third between 1990 and 2015, while incomes 

for all middle-aged groups show an initial increase followed by subsequent 

decline, though the timing and magnitudes are different across age groups. 

Online appendix figure 8 shows that while the matching of mortality and 

household income is strongest for the 50–54 age group, it also appears at 

other ages, albeit less clearly. This looks like good evidence for the effects 

of income on mortality, not at an annual frequency, which the graphs 

clearly show is not the case, but because of the (approximate) matching of 

the timing of the turnarounds across age groups.

When we disaggregate by educational attainment in figure 15, there is 

less support for an income-based explanation for mortality. The left panel 

10.6

10.4

10.2

10

665

400

250

150

Log 2015 dollars

Deaths per 100,000

a