Morningstar - Summer 2019 - 58

Strategies

20.3

19.1 30
19.0
13.4 20

20
10.5
5.9 10
0

EXHIBIT 2

12.3
10
11.5
0

60-65among
80
85 Real consumption
66-70 retirees
71-75 tends
76-80 to81-85
Spending
Smile
decline
Age Group
in the first two decades of retirement before increasing at later ages. The effect
varies by income group.

60
70
The Retirement
Household Age

Expenditure Level:

$25,000

$50,000

Real Spending Level

$100,000

$1.10

This lack of spending is found in other domains.
For example, the value of retiree portfolios remains
constant or increases over time (see Browning,
Huston, and Finke, 2015) even after taking
required minimum distributions, which are
subsequently reinvested in other financial assets
instead of spent (Poterba, Venti, and Wise, 2011).

1.03

1.00

0.88

0.90
0.80

0.75

0.70
0.60
65
Age

100

Source: Morningstar.

Above Threshold
Below Threshold
Diﬀerence
Δ in Nondurable Spending
change. Alternative consumption baskets such
0
as the CPI-E have grown faster than more
traditional consumption baskets such as the CPI-U
used to measure inflation, in large part due
-5
-8.4
to the higher weighting on health expenditures.
-10.1-10
Given that healthcare represents a rising allocation
of expenditures for retirees as they age, and
-15
that healthcare expenditures increase at a rate
-18.6
higher than inflation, it would seem that
-20
5
10 grow faster
15
20 inflation.
0retiree spending
should
than
Threshold
(% of
Total
Spending)
For example,
if we
assume
that Medicare
premiums represent 65% 9 of total healthcare
expenditures, that Medicare premiums increase
4% higher than base inflation, and that the
remainder of health expenditures are regular
medical expenses that increase 2% higher than
inflation, the total medical/health expenses
component of total retiree expenditures would
increase by about 3.3% per year in real terms
during retirement. In the absence of any other
considerations, retiree spending should increase in
real terms through retirement. For example, a
financial planner might assume the retirement
income need would increase by 2.5% throughout
retirement versus a general inflation rate of 2%.

9 See Fronstin and VanDerhei, (2017) and Cubanski et al. (2014).

Morningstar Summer 2019

Above Threshold
Below Threshold
Diﬀerence
Δ in Nondurable Spending
Despite Inflation, Spending Falls in Retirement
0
In reality, however, retirees appear to decrease
spending in real terms during retirement. While
there has been much research exploring changes-7.4 -5
in spending and consumption at retirement
-7.8
(dubbed the "retirement consumption puzzle"),
-10
there is little research exploring how retirees
-15.2
change their consumption throughout retirement.
Wolfe and Brazier (2018) is one example. Another is -15
a study I conducted in 2014. I found that real
-20
0
5
10 initially
15 and then
20
consumption
tends
to decline
Threshold
(%ages,
of Total
Spending)
increase
at later
noting
a "retirement
spending smile" (Blanchett 2014). The effect varies
by income level; those with higher expenditure
levels tend to see larger declines (likely due to
higher levels of discretionary spending). EXHIB IT 2
documents this effect, based on equation 1 in my
2014 paper, for a 65-year-old retiree, assuming
initial spending levels in the first year of retirement
of $25,000, $50,000, and $100,000.
It is not clear whether retirees reduce consumption
throughout retirement in real terms by choice
or necessity. I noted in my study that even wellfunded retirees tend to reduce spending, which
suggests some degree of choice is involved.

The fact that retiree spending does not simply
increase annually by inflation has important
implications for other retirement modeling
assumptions, such as how spending may change
after a health shock. The default assumption might
be that spending is unaffected (other than
having to pay for the additional health expense),
but this may not be the case. Spending might
increase if the retiree incurs higher ongoing
medical expenses, or spending might decrease
if the retiree cuts back on spending due to health
limitations (e.g., less travel and entertainment).
We sought to explore this effect in greater detail.
Our Analysis
We analyzed data from the Health and Retirement
Study, a household survey of individuals over
the age of 50 in the United States. We focused on
actual healthcare expenditures to determine
a potential shock, with a particular focus on outof-pocket expenditures, including those for
drugs, health services, and medical supplies. We
excluded health insurance premiums from
our estimate of healthcare expenditures because
these would not be considered a shock-they
are known with relative certainty in the near-term
(although the longer-term value is less certain
given the unknown inflation rate). We used an event
study framework, comparing health expenditures
over time to base values in 2005. We applied
several filters to the data to ensure data quality.

We looked at four types of spending: nondurable
spending, durable spending, total spending, and
total consumption. Durable spending was
estimated by subtracting nondurable spending
from total spending. We suspected spending
would decline after a health shock, but it may well
have increased, too. For example, De Nardi et
al. (2016) notes that medical spending is persistent

Morningstar - Summer 2019

Table of Contents for the Digital Edition of Morningstar - Summer 2019