Morningstar - Q2 2020 - 63

Strategies

A Better Retirement Spending Rule
for Everyone
Finally, a model that accounts for
an investor's preferences, needs, and
circumstances.

Q UA N T U

Paul D. Kaplan
After saving up for many years, an investor faces
a new set of decisions upon retirement. Two
of the most important are how much to spend each
year in retirement and how to invest the money
to fund retirement spending.
In the literature, there are two general approaches
to modeling retirement spending: static and
dynamic. In static models, the retiree selects an
amount to spend in the first year of retirement,
and that amount is increased annually by inflation.
In dynamic models, the amount spent each
year varies with the value of the retiree's portfolio.

out of money before dying. (The even better
models incorporate survival probabilities
into the calculations.) This yields a trade-off
between spending level and success probability,
taking as given the investment strategy (which
is usually expressed as an asset allocation).
The retiree can pick a desired success probability
and select the asset allocation that maximizes
spending. I have developed a model that does this
using Monte Carlo simulation (Kaplan 2006).
Milevsky and Robinson (2005) have developed
a version that uses formulas rather than
Monte Carlo simulation, which I described in
the February/March 2016 issue of Quant U
(Kaplan 2016).

A problem with the static approach is that the
retiree risks running out of money before dying
because the real withdrawal amount remains
constant. This is the case with the best-known
static rule, the "4% rule," originally suggested
by Bengen (1994). Under the 4% rule, the retiree
withdraws, and presumably spends, 4% of
the value of retirement funds in the first year of
retirement. This rule is not personalized to
the retiree and is a one-size-fits-all approach.

Dynamic rules avoid running out of money by
varying spending with portfolio value. A fairly
simple approach that has gained a lot of attention
was proposed by Waring and Siegel (2015).1
In their model, the retiree picks a date far enough
into the future to be nearly certain of not surviving
until then. Call this date T. At each year t ≤ T,
the retiree calculates (and then recalculates the
following year) the price of a sequence of $1
payments from year t through year T. The amount
of spending each year is wealth divided by
this price.

A better static approach is to use a success
probability model. In this approach,
for each spending level considered, the model
calculates the probability of not running

The only parameter in the Waring-Siegel model
that is specific to the retiree is T. In every other
way, it is a one-size-fits-all approach. This is why
I consider their model to be imperfect and too

simple for use in practice. In my view, a retirement
spending model should account for the
preferences, needs, and circumstances of each
retiree. David Blanchett and I (Kaplan and
Blanchett 2019) have developed such a model.
In this issue of Quant U, I present this model
with some simplifications to ease exposition.
While the model is somewhat complex, it
can be implemented in an Excel spreadsheet.
The model I present here is based on the same
principles that I discussed in the three-part Quant
U series "Investing Over the Lifecycle" (October/
November 2017, December/January 2018, and
February/March 2018). The main differences are
that here I model only the retirement period
(although the model could be extended to include
the working period) and I assume that annuities
are not available to reflect the fact that many
retirees do not use annuities, even though they
could benefit from them.
Preferences, Needs, and Circumstances
As I indicated, retirees differ from one another in
their preferences, needs, and circumstances.

In the Kaplan-Blanchett model, there are three
types of preferences:
gRisk tolerance-the retiree's attitude toward risk.
A higher risk tolerance leads to a more aggressive
investment portfolio.
gFlexibility-how willing the retiree is to shift
consumption between periods.
gImpatience-how much the retiree discounts
future consumption because consumption today is
better than consumption in the future.
By needs, I mean nondiscretionary consumption
for necessities like food and housing.
Any spending strategy must cover this.
The circumstances of a retiree include:
gLongevity-the probability of surviving to each
possible age once retired. This depends on age,
gender, and health. Also, when planning for a
couple, a spending strategy should consider the

1 The authors won the prestigious Graham and Dodd Award of Excellence from the Financial Analysts Journal for this article.

morningstar.com/lp/magazine

63
http://www.morningstar.com/lp/magazine
Morningstar - Q2 2020

Table of Contents for the Digital Edition of Morningstar - Q2 2020
