Sunday, March 3, 2019

Negotiating The Fog Of Retirement Uncertainty

Households who want to pay retirement expenses from an investment portfolio turn to spending rules like the 4% Rule, fixed percentage rule, or IRS Required Minimum Distribution (RMD) rules, to estimate how much they can spend each year. Retirees hope these rules will offer both a high probability of paying their future bills and a low probability of outliving their savings.

Many retirees and retirement planners are heavily invested in spending rules, with the 4% Rule most widely known. Spending rules attempt to protect us from outliving our savings but don't promise to pay our future bills. Retirees need both. Whether a retired household will actually outlive its savings will be determined by:
  • the length of retirement,
  • realized sequence risk (not market return expectations),
  • portfolio spending needs (actual needs, not spending-rule estimates), and
  • portfolio value.
If we could know all of these future values today, we could precisely determine how to fund retirement, so these are the factors we should include in a model to estimate how to fund it with portfolio spending.  

A household's length of retirement is the most important factor in determining sustainable portfolio spending. (A one-year retirement is easy to fund; a 30-year retirement, not so much.)

The length of retirement depends on longevity at the age of retirement. If two households have the same joint life expectancy but one retires five years sooner, the latter household should expect a 5-year longer retirement.

I simulated household finances for a sample of retired households from the Health and Retirement Survey (HRS). The average retirement age for this sample was 64 for men and 60.4 for women. Life expectancies were randomized using Society of Actuaries actuarial tables. About two-thirds of single-household retirements in this sample lasted from 14 to 23 years.

Spending rules attempt to protect us from outliving our savings but don't promise to pay our future bills. Retirees need both.
[Tweet this]

About two-thirds of two-person household retirements lasted from 22 to 32 years. The length of your retirement is quite uncertain and, since your household is a sample of one, could actually range from less than a year to 40 or more. Life expectancy simply provides an average for many people who are a lot like you and there is no reason to believe yours will be average.

Sequence of portfolio returns is the second largest determinant of retirement success after retirement length (for retirees who spend from a portfolio, of course) and is even less predictable than retirement length. The sequence of your future portfolio returns is unknowable and, as Karsten at[1] explains, the sequence of returns is much more important than average returns. "Precisely what I mean by SRR [sequence of returns risk] matters more than average returns: 31% of the fit is explained by the average return, an additional 64% is explained by the sequence of returns!"  

Spending needs (expenses) are the third most important factor of successfully funding retirement when we spend from a portfolio. (They're second when we don't.) Most spending rules ignore how much you will actually spend or even probably spend and instead make the dubious assumption that whatever amount of spending that will not likely deplete your savings portfolio will also be enough to pay your bills.

In a recently published paper entitled, "LDI Misapplied", David Blanchett and Thomas Idzorek explain how liability-driven investing, when used appropriately, is an improvement over asset-driven portfolio optimizations like Modern Portfolio Theory's mean-variance portfolio optimization (MVO)[4]. The primary difference is that LDI optimization also considers future spending requirements while MVO only considers portfolio assets.

Similarly, spending rules that also consider future liabilities are an improvement over spending rules based on assets alone. In both contexts, adding liabilities creates a more realistic model of future household finances.

Blanchett has published several studies on spending and the cost of retirement. Estimating the True Cost of Retirement finds that on average spending tends to decline as retirement progresses but not for all households. In fact, the study says that "households that are overfunded and not spending optimally (the “low spend, high net worth” group) actually tend to increase consumption." 

In the most recent study, Blanchett and Idzorek find that:
50% of the households experienced relatively small changes between [biannual HRS survey] waves. However, the other 50% of the 288 households experienced larger changes in spending, with the 5th and 95th percentiles indicating large changes in wave-over-wave spending. Focusing on the 5th and 95th percentiles and the five distributions in Exhibit 14, for approximately 90% of the households the wave-over-wave change was less than plus or minus 30%.
Blanchett notes that actual spending variability is probably even substantially higher than measured in this sample, as some outliers were rejected.

These two studies suggest that both the long-term trend of retirement spending and year-over-year spending can vary substantially for individual households. In other words, our household's future retirement spending is relatively unpredictable.

Even when we do include future spending in the spending rule estimate, we do so with substantial uncertainty. If we don't include it, we ignore a lot of risk. 

The value of the portfolio over time is also a key factor in determining sustainable portfolio spending. The future value of an individual household's portfolio is uncertain because the three previous factors are uncertain. Assuming sustainable spending will equal some pre-determined spending rule percentage of an unknowable future portfolio value is equally uncertain. 4% of an unknowable number is another unknowable number.

All four major determinants of sustainable portfolio spending are uncertain individually. Combining the distributions of random variables increases the uncertainty but ignoring one or more of them is worse.

It is extremely unlikely that our actual spending path throughout retirement will even remotely mirror sustainable spending predictions. The 4% Rule suggests larger percentages of spending as remaining life expectancy declines. RMD requires percentage withdrawals from tax-deferred portfolios that increase with age. Fixed percentage rules suggest a constant withdrawal percentage at all ages. All three are percentages of an unknowable future portfolio value.

When actual spending exceeds the spending rule estimates, the household is exposed to greater risk of underfunding retirement than the spending rule previously suggested. When estimated spending rates exceed actual needs, the household becomes more likely to underspend.

This isn't to say that spending rules have no value but they're at best a ballpark estimate from within an enormous ballpark. On the other hand, as my friend, Peter frequently reminds me, bad breath is better than no breath at all. The errors of the estimates are reduced as we age and we experience diminishing uncertainty about the future.

Spending rules that consider all four factors provide a better model and should provide a better estimate. Most rules consider three or fewer.

The key is to recognize that a spending rule estimate is good for perhaps a year. They should be recalculated at least annually. Retirement plans based heavily on spending rules have a one-year planning horizon.

Managing with a one-year retirement planning horizon is like driving while looking only at the road immediately in front of your car. When we can't see clearly what lies ahead, on foggy days perhaps, most of us respond by becoming less confident and driving more conservatively.

The important question is how confident we should be in spending rule estimates and the answer is not very.

Why is this important? As I mentioned in Honey, What's Our Retirement Plan?, the most important decision you will make in retirement planning is how much of your resources to allocate to the upside and floor portfolios. The less confident we are in our upside portfolio's ability to deliver on its promises, the more we should allocate to the safe floor portfolio.

Many retirees and even some planners seem to be massively overconfident in upside-portfolio spending rules.

Perhaps they haven't noticed the fog.


[1] blog.

[2] LDI Misapplied, David Blanchett and Thomas Idzorek.

[3] Estimating the True Cost of Retirement, David Blanchett.

[4] Liability-Driven Investment.


  1. Dirk, one of your best writings I have read on Retirement Cafe. I am reasonably confident in my upside portfolio at 50/50 in retirement with a solid floor from pension and Social Security. Do you think I am too optimistic with 15 to 20 years left. I couldn't agree with you more "Many retirees and even some planners seem to be massively overconfident in upside-portfolio spending rules."

    1. With a "solid" floor, 50% equity allocation, vigilance, and assuming you are not underestimating longevity, I don't think you are overly optmistic. That all sounds quite reasonable.

  2. Can you explain more about SRR: how is a poor sequence defined? One could assume that it means negative returns, but could you give a more detailed explanation / examples?

    1. Sequence risk is complicated. If you search my blog for "sequence" you will find many posts on the topic. But here are a couple of important points.

      Savings are only exposed to Sequence of Returns Risk to the extent that we save to or spend from an investment portfolio.

      A poor sequence of returns doesn't have to be losses. To cause problems, it merely has to be 5-7 years of so of returns that are significantly lower than your withdrawal rate. Several consecutive years of even mediocre returns can lower your portfolio balance enough that it will be difficult to recover without draconian spending cuts. (Actual losses are obviously worse.)

      Second, average market returns are far less a problem than the sequence of those returns. An average portfolio return of 2% with a fortunate order (sequence) of those returns will probably succeed while a 6% average return with most of the losses early in retirement ( a "poor" sequence") will probably fail. The point being that assuming in your retirement plan that market returns will be very conservative doesn't lower your risk of prematurely depleting your savings much. It's the sequence of those returns that will likely determine the outcome.

      Thanks for writing!

  3. There's a lot of uncertainty in projecting spending in retirement and it's not a one-size-fits-all spending rule. That's why I urge going into retirement with a margin of safety. If you think you'll need $1.2, try to make it $1.3 and have additional back up plans.

    1. Correct, a spending rule is simply an estimate of an amount that is supposedly "safe" to spend this year. In reality, there is no way to estimate that with any degree of accuracy. It's simply a budgetary amount.

      Regarding your last sentence, I will say from having run a gazillion retirement plans, actual, historical and simulated, that the probability of success given $1.3M of savings versus $1.2M will be indistiguishable.

      Thanks for writing!

  4. Well said. The sequence of returns and withdrawal rate are complicated by another variable, the volatility of your portfolio. Limit the volatility, and your chances of long term success generally increase. However, a 75/25 portfolio is also more likely to succeed than a 25/75 portfolios. Unfortunately those last two sentences are in opposition to one another.

    Personally - I have a decent slug in Vanguard's minimum volatility portfolio (not low volatility stocks, different thing.) I think ACWV does something similar, but I won't swear that it's the same. In any case, I expect to increase this allocation nearer retirement. It won't solve the problem, but it's performance so far suggests the possibility that I'm likely to get a skoosh of breathing room. On the other hand, if it were a sure thing, they wouldn't call it risk.