Sunday, January 25, 2015

Forecasting Future Investment Experience--Please Use Monte Carlo Simulations Responsibly

Just after I finished yesterday's post, I received a question from a reader asking if I would consider doing a post on the latest article from Messrs.  Blanchett, Finke and Pfau, "Retiring in a Low-Return Environment", and how the results of their latest research might affect use of the Actuarial Approach.

Since these three gentlemen consistently produce very good retirement research, I was more than happy to read their article.  As it happens,  I was pleased to see that their article generally reinforced the concerns I was expressing  yesterday about the Monte Carlo simulations used in the "Perfect Withdrawal" post.  They said,

"The generous capital market returns of the prior century bolstered a comfortable and long-lasting retirement portfolio. But they will give 21st-century clients a false sense of security and prejudice products and strategies that would do a better job of meeting retirement income goals."

This is not the first time these gentlemen have sounded the alarm about using historical rates of return and Monte Carlo modeling to develop safe withdrawal rates (SWRs) in the current economic environment.  See my post of January 19, 2013 about their January 15, 2013 article, "The 4% Rule is Not Safe in a Low-Yield World."  This new article supplements the older article with information supporting  their assertion that future equity returns may also be lower than historical equity returns.

So, after depressing us all with their view of future investment returns, what do the authors recommend we retirees do other than use lower SWRs?

  1. Rather than plan on living until age 95, plan on living your life expectancy and hedge longevity risk by purchasing a deferred income annuity.
  2. Plan on decreased levels of spending as you age and be more flexible in terms of accepting decreased spending levels if investment experience is not as favorable as expected. 
Item 1 is not so easily accomplished when a retiree is using a SWR approach unless the retiree knows exactly how to adjust the SWR for shorter expected payout periods.   On the other hand, both items can be easily accomplished when using the Actuarial Approach, which just supports my general recommendation to ditch SWRs and use the Actuarial Approach instead. 

In terms of the question of what a low-return environment means if one is using the Actuarial Approach, my response is if you are using the recommended assumptions for 2015 (5% investment return, 3% inflation and a payout period of 95-current age, or life expectancy if greater) then you are already using assumptions that are consistent with a low-return environment and you don't necessarily need to change anything.  The recommended assumptions are also consistent with the 2% real investment return assumption advocated by Dr. Pfau for deterministic projections.   As I have indicated in previous posts, if you desire a more-front loaded spending budget pattern consistent with item 2 suggested above, you can use a lower assumption for inflation.  For more discussion of the 2015 recommended assumptions, see our post of December 3 of last year.

If you are using the Actuarial Approach with assumptions that are significantly more aggressive than the recommended assumptions (or an approach that produces significantly higher current withdrawal rates), you either need to believe that these researchers are being unduly pessimistic about future investment returns or you need to be prepared to reduce your future spending. 

Saturday, January 24, 2015

The Perfect Withdrawal Amount?

Recently two faculty members of the Trinity University Department of Finance and Decision Sciences and an independent financial advisor released a paper entitled, "The Perfect Withdrawal Amount:  A Methodology for Creating Retirement Account Distribution Strategies."  From my perspective, the approach set forth in this paper has many positives and a few potential negatives.  Not surprisingly, I agree with aspects of the approach that are similar to the Actuarial Approach recommended in this website and generally disagree with the aspects of the approach that are dissimilar. 

Items of Agreement.  The authors recommend a (dynamic) process involving annual redetermination of the amount to be withdrawn based on accumulated savings, assumed rates of future return, age, risk preference and amount of bequest motive at time of redetermination.   As I, the authors are critical of safe withdrawal approaches and variations of safe withdrawal approaches that incorporate "adaptive rules."  The Actuarial Approach basically follows the same process recommended by the authors.  In addition, it develops a budget that is coordinated with other annuity/pension income.  In fact, it develops what the authors call a "Perfect Withdrawal Amount" (or perhaps "Better than Perfect" if the retiree has other fixed dollar annuity income) each year.  The only difference is that the Actuarial Approach uses deterministic assumptions where risk preference is varied through the use of more or less optimistic assumptions for longevity, future investment returns and inflation.    In addition, I recommend the use of a smoothing algorithm to keep budgets from being too "jumpy" from year to year. 

Items of Disagreement.  The authors model future investment experience using a Monte Carlo approach and the paper refers to historical equity performance from 1957 to 2013.  It is not clear whether this data  has been adjusted for inflation or to reflect the current economic environment.  As I indicated in my post of July 15, 2014, while Monte Carlo modeling appears to be more sophisticated than deterministic projections, it is no better at projecting the future, and in fact could be much worse if the data is not properly adjusted.  The authors use the period 1957 to 2013.  As an example of how that period may not properly reflect the current economic environment,  let's take a look at the prime rate charged by banks on short-term loans to businesses (available in Table H15 of the Federal Reserve System historical interest rates).  Over the 33-year period of 1969 to 2001 (about 58% of the period used by the authors), the average prime rate was in excess of 9% per annum.  By comparison, the rates for every year since 2008 have been 3.25%.  Running a model 20,000 times with what could be questionable data does not give me a good feeling about the reasonableness of the projections.  By comparison, the recommended investment return assumption for the Actuarial Approach is selected to be somewhat consistent with interest rates inherent in current fixed immediate annuities available on the market (with the understanding that higher investment returns generally come with higher levels of risk that should be discounted).

Bottom Line. 
Depending on the data used to model future experience, the author's approach may produce a reasonable annual withdrawal amount.  Care should be taken to make sure that the effects of future inflation have been considered to the retiree's satisfaction.  I would suggest comparing withdrawals under this approach with the withdrawals determined under the Actuarial Approach, with reasons for significant differences in results explained to the retiree. 

Thursday, January 22, 2015

Bud Hebeler's Autopilot Withdrawal Rule

I've been a big fan of Bud Hebeler and his Analyze Now website every since I retired and started this website.  In fact, his website was one the first one I listed in the "Other Calculators/tools" section of this website.

Bud writes many articles about retirement, and is featured as one of the MarketWatch Retirementors.  In his most recent article,  Bud touts the virtues of a fairly simple method of determining annual withdrawals from savings he calls the "Autopilot" approach.  While I don't think it is necessarily better than the Actuarial Approach advocated in this website, subject to the caveats discussed below,  I do think that it can provide you with another point of reference with respect to the amount of your spending budget that may be withdrawn from accumulated savings. 


Bud's autopilot approach involves taking 75% of last year's budgeted withdrawal increased with inflation for the previous year and adding 25% of what Bud calls the "Planning Method" (or Planner Method).  Under the Planning Method, all you need is a calculator (like the HP 12c) or an internet financial planner with financial functions, and you solve for the annual annuity payment (PMT) given the period of payments (n), the interest rate (i) and the amount of accumulated savings (PV).  Bud suggests using IRS Publication 590 life expectancy tables for "n" and an interest rate of 5% and inflation of 3.5%.  Note, if you are using the HP 12c calculator, you will be doing the calculations assuming beginning of year payments and an interest rate of 1.45%.

Let's take a look at Bud's suggested withdrawal rate at age 65 where the IRS 590 life expectancy is 21 years.  Bud's initial withdrawal rate at age 65 using the Planning Method would be about 5.5%.  By comparison, the withdrawal rate under the Actuarial Approach (using recommended assumptions) is about 4.3%.  The primary difference in the two approaches results from using life expectancy of 21 years under Bud's method vs. assuming a 30 year payment period under the Actuarial Approach.  We discussed why you might not want to use life expectancies for your retirement planning period in our post of Wednesday December 3, 2014, but perhaps Bud feels his conservative investment return/inflation assumptions counterbalance this somewhat unconservative longevity assumption. 

The Autopilot approach also uses a different technique to smooth actual experience and spending variations.  If readers like Bud's smoothing approach better than the smoothing algorithm we recommend, you could simply replace the "Planning Method" result with the result from our Excluding Social Security spreadsheet and use a combination of the two approaches. 

Caveats regarding the Autopilot rule:

  1. It assumes no bequest motive 
  2. It does not coordinate with other sources of retirement income such as immediate or deferred pension/annuities
  3. As discussed above, it recommends the use of life expectancy.  As the retiree ages, life expectancy (in years) will decrease but not by one year for each year of advanced age, so this approach will result in longevity experience losses that will decrease future withdrawals unless these losses are offset by other experience gains.