In software engineering, an anti-pattern is a pattern that may be commonly used, but is
ineffective and/or counterproductive in practice (wikipedia).
Retirement planning, as any other activity with thousands of participants, has its own
patterns. Not all of them are good. Here we will discuss five some of the most prominent anti-patterns
in retirement planning.
Assigning value to portfolio return
Most retirement calculators as well as many professional retirement planning software products
require the user to input the exact portfolio return value (or yearly returns). Some variations
include assigning mean and standard deviation to the entire portfolio. Asset classes are not
factored in the analysis at all in this approach. Thus, the burden of the most difficult and
sophisticated calculations gets put on the user. The quality of decisions based on such approach
is very questionable.
Using quantitative rules of thumb
A typical quantitative rule of thumb related to retirement portfolio asset allocation sounds as
following: the percentage of your portfolio invested in bonds should equal your age.
A more recent modification of this rule is: subtract your age from 120 to get the percentage
of your portfolio invested in stock.
The 4% withdrawal rule recommends that you withdraw 4 percent from your retirement portfolio
annually and adjust it for inflation.
You may have noticed that some of these rules contradict each other and hence are questionable.
In addition, these rules of thumb do not take into account individual financial and family situation
and the individual risk tolerance. You can find more information on quantitative rules of thumb
and corresponding references here.
Static asset allocation
The vast majority of professional retirement software products that provide the ability
to construct retirement plan portfolio from different asset classes do not allow changing asset
allocation of this portfolio during the entire retirement phase. The same is true for online
retirement calculators with a very few exceptions.
This sounds terrible from both practical and theoretical points of view,
but this is the sad reality. Target date funds do a much better job providing a glide path
(an annually changing asset allocation), but they have another problem: the only personal
information used by a target date fund from the participant is her age. Target date fund
retirement plans are maintained uniformly for very large groups of people with different
wealth and risk tolerance. This is not enough for an effective individual or family
retirement plan.
Applying Monte Carlo methods where they are not effective
Risk valuation of retirement plans is a good practice in the retirement planning. More and more
software retirement planning products implement this feature. All of them currently use Monte Carlo
approach for this goal, which is by far not the most effective way to achieve it. Monte Carlo
requires significant additional efforts if you want to analyze retirement scenario with a variable
asset allocation. Monte Carlo fails in cases when you want to analyze many such scenarios. Analyzing
various scenarios is natural for decision support systems or retirement plan optimization.
You can find more detail about this anti-pattern here. Multiple retirement risk calculators
on this site demonstrate implementation of a much more effective alternative to the Monte Carlo
approach for retirement plan risk valuation.
Separating asset allocation optimization from the retirement plan risk valuation.
It is natural to use retirement goals as criteria to optimize the retirement plan asset allocation.
This means that retirement plan optimization has to include risk valuation of retirement plans.
Unfortunately, all software products that use Monte Carlo methods for retirement plan
risk valuation fail to provide this common sense requirement.
Current retirement planning software products do not use the final retirement goal(s)
in optimization. They utilize a one-period static mean-variance optimization framework, which is
traditionally used for short-term investing. and apply it for multi-period retirement investment,
where this framework is not really applicable. Such approach produces asset allocation that remains
the same during the entire retirement phase. The final retirement goals are not a part of this framework.
Criteria used in one-period frameworks are very far from retirement goals and
in some cases contradict them. Besides, the Monte Carlo method for valuation of risk to reach retirement
goal(s) is applied only after the calculation of the asset allocation.
Such workaround significantly decreases the quality of the resulting retirement plan because it:
produces non-changeable asset allocation portfolio for the entire phase of the retirement plan
separates the asset allocation optimization from the retirement plan risk valuation
Conclusion:
Today’s retirement planning software products utilize multiple anti-patterns that
significantly decrease the quality of the resulting retirement plans. Their quality is far behind
the 21st century demands. Fortunately, the technology and science together can significantly improve
the situation. The first online publicly available retirement glide path optimizer that uses effective
alternatives to these anti-patterns can be run free from this site.
