Thursday, October 30, 2014

Revisiting the Recommended Smoothing Algorithm

I first proposed a recommended smoothing algorithm for The Actuarial Approach in my post of October 11, 2013.  The purpose of the recommended smoothing algorithm was to balance the desire to have relatively constant real dollar budgets from year to year with the need to keep spending budgets reasonably on track with the "actuarial value" determined using the applicable spreadsheet from this website (generally Excluding Social Security V 2.0), recommended assumptions and the retiree's actual data as of the beginning of each year.

In general terms the recommended smoothing algorithm involves taking the budget amount from the previous year, increasing that amount by inflation (the increase in CPI) for the previous year and making sure that the resulting value is not more than 110% of and not less than 90% of that year's actuarial value.  If the resulting value falls within the corridor, the inflation adjusted value is used to develop the budget for the year.  If the value falls above or below the corridor limit, the applicable corridor value is used in the budget calculation.

Recently, in a conversation with Ian McGugan, reporter for the Globe and Mail in Toronto, Mr. McGugan pointed out that I wasn't terribly specific about what budget amount was being used in this smoothing calculation.  Mr. McGugan (a bright guy) correctly pointed out that in the various examples contained in my website, I had inconsistently used several different amounts for the previous year's budget amount for this calculation.

I will readily admit that there was more art than science involved in developing this smoothing algorithm.  For example, I don't know whether some or all retirees who may use the Actuarial Approach would be better served with a 12% or 8% corridor.  Perhaps someone like Wade Pfau or some other retirement academician can utilize Monte Carlo modeling to develop an optimal corridor.  Similar arguments can be made with respect to the budget (or portion of the budget) that should be subject to the corridor.

If the retiree has an immediate defined benefit pension benefit or life insurance annuity as a retirement income source, she has at least three choices with respect to how the recommended algorithm can be applied:

  1. to the portion of the budget attributable to accumulated savings only
  2. to the portion of the budget attributable to accumulated savings plus the pension/annuity, or
  3. to the total budget, including income from Social Security
If the retiree has no immediate pension or annuity income sources, then items 1 and 2 above will be the same.

Personally, I would prefer that the spending budget stay reasonably close to the actuarial value, so I will recommend that the 90% to 100% corridor be determined using the portion of the budget attributable to accumulated savings plus the pension/annuity rather than using the entire budget including Social Security, as it will produce a somewhat smaller corridor range, all other things being equal.

As a result of Mr. McGuggan's excellent suggestion to try not to confuse readers when possible, I have revised the examples in the June, 2014 article "Using the Actuarial Approach to Determine Your Annual Spending Budget in Retirement" to reflect the more specific recommended smoothing algorithm above.

I thank Mr. McGuggan for his feedback and encourage other readers to submit their suggestions for improvements to this website.

Friday, October 24, 2014

You Can Spend it Now or You (or Your Heirs) Can Spend it Later

The title of this post is a re-working of the saying used by the mechanic in the old Fram Oil Filter TV commercials.  His famous line was, "You can pay me now, or you can pay me later."  A similar principle applies to spending your accumulated savings--assets you spend today will reduce the amounts available to you (or your heirs) tomorrow and visa versa.  The trick, of course, is to figure out the best way to meet your spending objectives, including making your money last your lifetime.  This difficult "balancing act" is what this website is all about. 

It is also important to remember that most systematic withdrawal strategies produce a spending budget.  As with any other budget, you are free to deviate from the budget and spend what you feel is an appropriate amount.  Some systematic withdrawal strategies (like the Actuarial Approach recommended in this website) automatically adjust for such deviations.  Other approaches may not.   

As promised in my post of October 9, this post will discuss the systematic withdrawal approach suggested by David Zolt in his recent Journal of Financial Planning article.  Readers can also find more details about his approach in his website

In summary, Mr. Zolt uses a Hybrid safe withdrawal/actuarial approach.  He incorporates Monte Carlo modeling using historical returns from 1926-2013 and then uses a deterministic approach similar to the Excluding Social Security V 2.0 spreadsheet in this website (which he calls the Target Percentage Test) to monitor whether spending remains on track during retirement.   In general, if the proposed withdrawal in any year, measured as a percentage of remaining accumulated assets, falls below the Target Percentage, Mr. Zolt's approach would require the retiree forgo that year's cost of living increase.  His most recent article proposes another alternative to forgoing the entire cost of living increase if the test is failed. 

The first case study in Mr. Zolt's recent article describes how his approach could work for a couple, both age 63 with $650,000 in accumulated retirement assets, total Social Security benefits of $3,000 per month and a fixed dollar $1,000 per month benefit from a pension plan.  Based on his Monte Carlo analysis,  Mr. Zolt determines that this couple has a "superb" chance of making the money last with an initial withdrawal rate of 5.7% if they are willing to forgo cost-of-living increases whenever the proposed withdrawal for the year measured as a percentage of accumulated assets exceeds the Target Percentage for this couple.  By comparison, if we input this couple's data in our Excluding Social Security V 2.0 spreadsheet and use our recommended assumptions , we would get an initial withdrawal rate of 3.57%.  There are two reasons why Mr. Zolt's initial withdrawal rate is much higher than ours for the same couple:  1) We use different assumptions for future experience and 2) The Actuarial Approach attempts to keep total retirement income (including income from the fixed dollar pension) constant from year to year in real dollars. 

Using Mr. Zolt's Target Percentage spreadsheet, we can determine that if payments are to last 32 years and increase by 3% per annum, the couple's assets must earn a minimum of 7.7% per annum.   Tables 1 and 2 below compare spending budgets every five years produced by Mr. Zolt's approach (assuming the target percentage is determined using 7.7% investment return, 3% cost of living increases and payment for 32 years) vs. the Actuarial Approach (using recommended assumptions) assuming future experience is either 5% investment and 3% inflation (Table 1) or 7.7% investment return and 3% inflation (Table 2).  Table 3 compares spending budgets assuming future experience of 7.7% per annum and 3% inflation, but changes the assumptions used under the Actuarial Approach to 7.7% investment return and 3% inflation while Table 4 assumes future experience of 5% per annum and 3% inflation, but the assumptions used under the Actuarial Approach are changed to 5% investment return and 0% inflation.  Under all four tables, the retired couple is assumed to spend exactly the spending budget under either approach each year.

Under the assumptions for future experience in Table 1, spending budgets decline significantly in real terms under the Target Percentage Approach.  The couple's money does, however, last the entire 32 years.  Under the Actuarial Approach and recommended assumptions, spending budgets remain constant from year to year in real dollar terms.  Under the assumptions for future experience in Table 2, spending budgets decline somewhat under the Target Percentage Approach as the approach does not adjust spending from accumulated savings to make up for inflation erosion of the fixed dollar pension benefit.  Under the Actuarial Approach (using recommended assumptions), the more favorable investment experience produces significant increases in the spending budget in later years. 

The purpose of Tables 3 and 4 is to illustrate that similar results can be achieved under the Actuarial Approach by varying assumptions used to determine the annual spending budgets from the recommended assumptions. 

Which approach is right for you?  There is no easy answer to this question.  Both approaches are dynamic in that spending budgets may need to be increased or decreased during retirement since both approaches involve annual checking to see how last year's spending budget increased with inflation compares with an actuarially determined target.  The Actuarial Approach using recommended assumptions assumes future investment returns will be 2% real, while Mr. Zolt's approach uses a real rate of return closer to 4.7%.  His approach also does not consider the effects of inflation on other fixed dollar income (such as pensions or annuity contracts) or amounts desired to be left to heirs.   There is nothing mathematically superior about either approach.  While Mr. Zolt's approach can be made more conservative by using more conservative target percentages, his reliance on historical rates of return makes his approach less conservative (more likely to involve decreasing real dollar spending budgets in future years) than the Actuarial Approach (which tends to tie investment return assumptions more closely to interest rates inherent in current immediate annuity contracts).  This may be ok for some retirees, as long as they understand that higher amounts spent early in retirement mean smaller spending budgets later, all things being equal.  Other retirees, however, may be less concerned if the spending budget increases in later years as a result of more favorable than assumed experience.   Being a retired actuary, I tend to favor the more conservative approach.   But, in the end, the bottom line of this post is the same as its title, and it is up to you to determine how conservative you want to be in determining your spending budget (and your actual spending) from year to year.  




Wednesday, October 22, 2014

Thank You Michael Kitces

In his October 22 blog post, Michael Kitces points out several of the problems associated with Monte Carlo modeling and determination of "safe withdrawal rates" based on assumptions about future experience that will not be realized.  Mr. Kitces suggests that the solution lies in "reframing" the model outcomes.  He says, "consider what happens if we actually call it a probability of adjustment, instead of a probability of failure. When we frame the outcomes as failures, the nature [sic] response from clients is to think up terrible images of what failure might look like, and then seek to avoid it at all costs. But when we frame the outcomes as “adjustments” it leads to very different – and much more productive – conversations instead, such as “How big would the adjustment be? When would I have to make the adjustment? How will I know when it’s time to adjust?”

He also expresses concerns about surpluses that are ignored in the safe withdrawal rate determination, saying, "it’s not just the probability of failure that’s misnamed. It’s also the probability of success, which is more like a probability of Excess. It’s the likelihood of having excess money left over, and sadly makes no distinction about how much will be left over! A Monte Carlo analysis in traditional retirement planning software treats having $1 left over the same as $1M and the same as $10M – they’re all “successes” – yet clients would react to this very differently. When you call it a probability of “excess” it again raises the question “how much of an excess are we talking about?” and a more productive conversation."

While I agree with Mr. Kitces' assessment of the problems, in my opinion, reframing the outcome is not the answer.  The actuarial approach suggested in this website automatically provides valuable input for Mr. Kitces "productive conversations."  As indicated in my previous post, retirees need to wean themselves off the safe withdrawal rate Kool-Aid and  live with the fact that that sometimes their spending budget may increase or decrease in real terms from year to year, depending on actual investment experience and spending.

Thursday, October 9, 2014

20 Years of Drinking the "4% Rule" Kool-Aid

This month's Journal of Financial Planning celebrates the twentieth anniversary of publication of the article, "Determining Withdrawal Rates Using Historical Data", by William P. Bengen.  This article formed the basis for what is now known as the "4% Rule" or "4% Withdrawal Rule."

In this month's Journal article, Jonathan Guyton, CFP, offers his thoughts on the original research performed by Mr. Bengen, and Mr. Guyton's article contains glowing praise from many others for Mr. Bengen's research.  According to Mr. Guyton, "Bill Bengen framed a deceptively complex question and crafted and elegant simple answer that remains relevant two decades later for hundreds of thousands of financial advisers and easily 100 million retirees."

As someone who doesn't understand the appeal of the 4% Rule, or it's many safe withdrawal rate progeny, please forgive me if I don't enthusiastically join in this celebration.

Based on historical returns provided by Ibbotson Associates' Stocks, Bonds, Bills and Inflation 1992 Yearbook, Mr. Bengen noted that hypothetical individuals who retired in each year from 1926 to 1976 who invested at least 50% of their assets in equities and who withdrew 4% of accumulated assets in the first year of retirement and increased that amount by inflation each year would be expected to have their assets last at least 30 years (Figure 1b of Mr. Bengen's article).  As a result of this analysis of these historical returns, Mr. Bengen concluded that it was "safe" going forward for almost all retirees to withdraw 4% of accumulated assets in the first year of retirement and increase that first year amount by inflation for subsequent years if they invested at least 50% of their assets in equities and rebalance their portfolio each year.  In fact, Mr. Bengen said, "It is appropriate to advise the client to accept a stock allocation as close to 75 percent as possible and in no cases less than 50 percent."

I agree with the concerns expressed about the 4% Rule expressed by researcher Nobel Laureate, William Sharpe in his article, "Staying Flexible on Retirement Spending" (highlighted in my second post in 2010) where he points out the problems inherent in combining a fixed spending strategy with investment of a portfolio with variable returns. 

I list what I believe to be the many shortfalls of the 4% Rule in this website and in my recently published article in the Journal of Personal Finance.  For brevity sake, I will not be re listing them here.  Suffice to say that the 4% Rule requires the retiree to have faith that historical performance is a good indicator of future experience and the retiree should "stay the course" irrespective of actual investment returns or spending. By comparison, the Actuarial Approach advocated in this website suggests looking to see whether the retiree is "on track" on an annual basis and making appropriate adjustments.  

In the twenty years following release of Mr. Bengen's original research, many researchers have suggested modifications to the 4% Rule to deal with its significant shortcomings and to refine the assumptions and methodology (Monte Carlo modeling) used to forecast future experience.  In fact, an example of such refinement is also featured in this month's Journal of Financial Planning entitled, Retirement Planning by Targeting Safe Withdrawal Rates, by David Zolt, CFP, EA, ASA, MAAA.  This is an update of Mr. Zolt's original article which I discussed in my post of May 23rd of last year.  Like many suggested modifications to Mr. Bengen's original work, Mr. Zolt has dealt with some of the problems, but his approach and various tables are much more complicated than Mr. Bengen's "elegant simple answer."  I will be discussing Mr. Zolt's approach in a subsequent post.

Bottom line:  Retirees who invest in risky assets should not expect those assets to produce a fixed level of real income in retirement as implied by the 4% Rule or other safe withdrawal rate approaches.  They need to use a dynamic withdrawal approach (like the Actuarial Approach advocated in this website) and live with the fact that that sometimes their spending budget may increase or decrease in real terms from year to year, depending on actual investment experience and spending.