Morningstar - Q4 2020 - 62

Strategies

demonstrated in E X H IB IT 1 , which includes
estimated retirement periods (in years) at various
survival probabilities based on different health
status levels (poor, average, or excellent)
and smoker status for the base couple (John and
Jane, both age 65). Again, all estimates
are obtained from the Longevity Illustrator.

probability periods are nine, 12, 15, and six
years for John, Jane, either, or both, respectively.
In contrast, if the household is assumed to
be in excellent health and residents don't smoke,
the median survival probabilities are 23, 25, 29,
and 18 years for John, Jane, either, or both,
respectively. There is a roughly 15-year differential
in these different periods, which is relatively
constant across survival probabilities.

The median (50%) survival probability period, as
well as other probability estimates, changes
materially based on the assumed attributes. If the
household is assumed to be in poor health
and the residents smoke, the median survival

speaks to the importance of ensuring
that mortality estimates are personalized to
the respective household.
The Accuracy of Subjective Mortality Estimates
The Health and Retirement Study is a longitudinal
household survey conducted by the Institute
for Social Research at the University of Michigan
that has been administered biennially since 1992.
In the first wave of the study, researchers
asked, "What do you think are the chances that
you will live to be 75 or more?" With an
average respondent age of 58, it is possible to
observe the accuracy of these estimates.

Changing the assumed length of retirement
by 15 years can materially affect required
savings or recommended spending levels and

EXHIBIT 1

Retirement Periods Based on Household Types
Poor Health, Nonsmoker
Probability (%)

Poor Health, Smoker

John

Jane

Either

Both

90

5

7

14

3

75

10

13

19

50

17

20

25

23

10

28

John

Jane

Either

Both

90

1

2

6

1

7

75

4

6

10

3

24

13

50

9

12

15

6

26

28

18

25

14

18

20

10

31

33

23

10

19

23

25

14

John

Jane

Either

Both

Average Health, Nonsmoker
Probability (%)

Probability (%)

Average Health, Smoker

John

Jane

Either

Both

90

7

9

17

4

90

2

4

9

1

75

13

16

22

10

75

6

8

14

4

50

20

23

27

16

50

12

15

19

8

25

26

29

31

22

25

18

22

24

13

10

31

34

35

26

10

24

28

30

18

John

Jane

Either

Both

Excellent Health, Nonsmoker
Probability (%)

Probability (%)

Excellent Health, Smoker

John

Jane

Either

Both

Probability (%)

90

8

11

19

6

90

3

5

11

2

75

16

18

24

12

75

8

10

16

5

50

23

25

29

18

50

14

18

22

10

25

29

31

33

24

25

21

25

28

16

10

33

36

38

28

10

27

31

33

21

Source: American Academy of Actuaries and Society of Actuaries, Actuaries Longevity Illustrator.

62

Morningstar Q4 2020

EXHIB IT 2 includes information about the projected
versus actual probabilities of survival, based on
the 11 increments available to respondents, where
the zero to 10 responses are converted into
probabilities, from 0% to 100%, in 10% increments
There is clearly some predictive information in
subjective mortality estimates, given the relatively
linear relation between self-reported probability
of living to age 75 and the observed (actual)
mortality experience. However, there are
also notable errors at both extremes, even when
focusing on the averages for each group.
For example, among respondents who provided
a 0% probability, the actual survival rate was
approximately 50%, and among respondents who
provided a 100% probability, the actual survival
rate was approximately 80%.

Additional analysis was performed, and it
appears that once self-reported health status is
controlled for, the predictive ability of
individuals effectively vanishes. In other words,
people do a relatively good job incorporating
their health into mortality expectations,
but not much beyond that. Things like gender,
income, and smoking status, which should
matter, do not appear to be correctly considered.
These findings were reinforced when
reviewing values from the 2016 Survey of
Consumer Finances.
Do Financial Planners Customize Assumptions?
The information in EXHIB IT 1 suggests that
retirement periods can vary significantly based on
household attributes. An important question,
then, is to what extent financial planners consider
Morningstar - Q4 2020

Table of Contents for the Digital Edition of Morningstar - Q4 2020
