Portfolio Size Matters
First, the lead article by Gordon Irlam is about dynamic asset allocation over the lifecycle. I think this is a fascinating article and is well worth reading. In a recent post, I mentioned that there are three general ways to approach dynamic asset allocation: mechanical glidepaths based on age, valuation-based allocation, or the funded ratio. Gordon's research works at the intersection of mechanical glidepaths and the funded ratio, as he finds that the optimal asset allocation does depend not only on age, but it also very much depends on the ratio of one's portfolio wealth to their desired spending amount in retirement (the Relative Portfolio Size [RPS] which is 1 / withdrawal rate).
In other words, target date funds are inadequate because they base asset allocation only on age, when the funded status of the individual (the RPS) is just as important to determining optimal asset allocation. Of course, the point of target date funds is to move people in the right direction when they don't care about investing and have no idea what their RPS is, but more sophisticated investors should be able to do better than just using a mechanical glidepath.
Calculations are made using dynamic programming, which works backward to determine the optimal asset allocation at a particular age after accounting for what will be optimal at subsequent ages. He analyzes cases with a fixed life expectancies and variable life expectancies, and also for cases with and without a motive to leave a bequest. A summary of what his figures show is:
Figure 1: Success rates are naturally higher when the RPS is higher (implying the ability to use a lower withdrawal rate to meet one's goal). The highest RPS is needed in the years around the retirement date and after. After withdrawals begin, there is less opportunity for the portfolio to grow.
Figure 2: Optimal stock allocations decrease as the RPS gets larger at any particular age. Those able to use quite low withdrawal rates to meet their goals and who have no bequest motive have already won the game (in the language of William Bernstein), and so they can make due with a low stock allocation. Conversely, those with a low RPS will maximize the chances to meet their spending goals with a more aggressive stock allocation. Taking more risk is the Hail Mary pass to try and make the plan work.
Figure 3: This figure moves away from a fixed age of death to a variable age of death. It increases the role for balanced portfolios later in life, since uncertainty remains for how long one can be expected to remain alive.
Figure 4: This figure is really interesting, because it shows the optimal lifetime asset allocations for various individual Monte Carlo simulations. Note that there is a general tendency for a U-shaped lifetime asset allocation path. Stocks allocations are highest when young, lowest near the retirement date, and then increase again at higher ages. This is where my research with Michael Kitces about the rising equity glidepaths fits in. It's not that the rising glidepath is always optimal, but we think it is the best approximation that can be made for someone if we are not otherwise able to incorporate information about their funded status or RPS. The figure shows that in some simulations, the stock allocation does continue to decrease at higher and higher ages. Those would be simulations where things went quite well and the RPS continues to grow throughout retirement, so it is not necessary to have any stocks. Remember, at this stage in the research we are just looking at the optimal asset allocations to meet a fixed spending goal. There is no need for further upside potential because spending will not increase and we don't care about leaving a bequest.
Figure 5: Now he adds a bequest motive. The retiree also cares about leaving a bequest. This is a very interesting figure because it introduces higher stock allocations at low and high ages for people with very high RPS levels. As such, if Figure 4 was re-done with the bequest motive, I'm pretty sure that Figure 5 implies that a U-shaped lifetime asset allocation will apply to even more simulations (i.e. have the lowest stock allocation at retirement, and have higher stock allocations when young or old).
He finishes the article with some sensitivity analysis about how changing assumptions would change the optimal asset allocations, and he also shows how a more optimal asset allocation strategy that includes the RPS will reduce the amount of wealth needed at retirement relative to various rules of thumb or target date fund glidepaths.
Gordon is doing great work, and he has developed www.aacalc.com to allow users the opportunity to test their approach for different circumstances.
The Actuarial Approach
In the next article, Ken Steiner proposes an actuarial approach to planning for taking withdrawals from savings to support retirement. Ken is a retired fellow at the Society of Actuaries, and he hosts the blog, How Much Can I Afford to Spend in Retirement? The five step actuarial process he outlines includes:
1. Gather data
2. Make relevant assumptions about future market returns, future inflation, and remaining time horizon
3. Calculate the preliminary spendable amount, which is a mathematical calculation of the sustainable spending amount that would lead precisely to portfolio depletion (or the desired bequest amount) and the end of the planning horizon
4. Apply a smoothing technique for spending so that annual spending doesn't fluctuate too much based on what is calculated in step 3.
5. Store the results for next year's analysis.
He finishes the article with a comparison for how his approach performs against an RMD strategy, the 4% rule, and a strategy of withdrawing 4% of the remaining blaance each year.
This article is highly worthwhile as well.
And now for the journal announcement:
|
||||
|
||||
|
||||
No comments:
Post a Comment