Wednesday, October 26, 2011

The Fungible Nature of Retirement Spending Goals

Today’s blog post is inspired by 3 sources

1) Duncan Williams and Michael Finke’s article about how retirees may be willing to accept higher failure probabilities for the opportunity to spend more in early retirement. I described their article in a column at Financial Advisor.

2) Robert Curtis’s “Monte Carlo Mania” article which I reviewed here the other day.

3) A series of email exchanges I’ve had with Michael Kitces, whose ideas as expressed on his blog and his Kitces Reports have been a guiding inspiration for my research this year. His take on these matters helped to prompt this analysis.

Step 1, I think the main idea of the Williams and Finke article can be expressed in a paragraph I wrote at Financial Advisor:

Figure 2 shows the case for a risk-tolerant male retiring at age 65, who has guaranteed inflation-adjusted income sources of $20,000 (Social Security, for instance), and a $1 million nest-egg.  Far from the 4 percent rule, the figure shows how this retiree can maximize his utility using a 7 percent withdrawal rate with a 70 percent stock allocation. Rather shockingly, based on the traditional shortfall risk approach, this strategy actually would lead to a 57 percent chance of running out of wealth within 30 years. This, indeed, is the most surprising and thought provoking insight coming from Mr. Williams and Dr. Finke’s research. Acceptable failure rates might be much higher than we ever imagined.  Calls for low failure rates may not have properly accounted for the risk aversion or the other sources of income available for retirees who may be willing to risk higher failure for the opportunity to spend more earlier on in their retirements.

Step 2, in “Monte Carlo Mania,” Robert Curtis argues that prospective retirees do not have just one overall spending goal for retirement. They have different goals, and these goals can be prioritized by their order of importance.  Here is an adaption I am making from his article about some annual spending goals in order of priority:

1. Minimum annual living expenses: $50,000 (shelter, food, basic needs, basic travel)

2. Re-occurring expenses on ownership of nice family car: $5,000

3. International Travel: $10,000 

4. Gifts: $5,000 

5. Extra living expenses: $10,000

(Total Annual spending: $80,000)

Step 3, what I want to argue here is that these goals don’t just have an order of priority, but they also have a weighting based on their importance.
Williams and Finke discuss income sources from outside the savings portfolio which serve as a guaranteed income floor.  Suppose, for instance, that the prospective retiree can expect $20,000 from Social Security. Should the retiree exhaust all of their savings, the retiree will be forced to live on only Social Security.

In our email exchange, Michael Kitces argued that this would be a devastating outcome for most retirees, as it implies a drastic reduction to one’s standard of living. 

But as economists always want to say, life is about balancing tradeoffs.  Someone who really wants to keep their portfolio failure rate very low, will need to use a rather low withdrawal rate, which may mean they can only expect to meet their Goal #1, and must forgo the opportunity to achieve Goals #2 - #5.  

In facing this tradeoff, others may make a conscious decision to increase their withdrawal rate and make more of their goals achievable early in retirement. Finding how to balance the tradeoff is the whole purpose of the Williams and Finke article. But the answer depends on one’s weightings given to all these various goals, which in their article is defined as the coefficient of risk aversion.  I already provided a quotation describing an outcome for a more aggressive-style retiree with a risk aversion coefficient of 1. But in my review article, I also explain the case for a more conservative retiree with risk aversion of 4 who has the same underlying guaranteed income floor:

This new research does not always call for such high withdrawal rates. The scenario shown in Figure 3 is the same as in Figure 2, except that now we are investigating the case for a more conservative and risk-averse 65-year old male. For this retiree, a 4 percent withdrawal rate with a 40 percent stock allocation is utility maximizing. This particular result is close to what is generally recommended in traditional shortfall risk studies.

What I want to do here is re-cast the risk aversion coefficient instead as a “coefficient of goal flexibility.” Low numerical values mean that the retiree is flexible about their spending levels, while high numerical values mean that the retiree is averse to flexible spending, or in particular is averse to a drop in their spending levels.

To see this, I want to give a utility weighting to the goals from Robert Curtis’s framework.  I’ve illustrated below the utility provided by different spending levels for both the “flexibility tolerant” retiree and the “flexibility averse” retiree. Both naturally get more utility as their spending increases, but the flexibility averse retiree suffers from a big drop in utility when spending falls below the $50,000 minimum expense floor.  This retiree is really unhappy to subsist only on Social Security.  And so this retiree would want to accept only a very low failure rate, as is assumed in traditional safe withdrawal rate studies.

Meanwhile, the flexibility tolerant retiree does not suffer so much if spending must fall to $20,000, and also enjoys relatively higher marginal utility from increasing spending beyond $50,000.  This is the retiree who is more like the first case described above… they will be willing to increase their spending higher to even potentially $80,000, with the clear understanding that this leaves them with a rather strong possibility that they will be spending some of their retirement with much lower spending. 

The utility maximization approach provides a systematic way to analyze the shapes of these curves along with the probabilities of exhausting wealth with various withdrawal rates, putting it all together, and giving an answer about what strategy will afford the retiree with the most overall expected lifetime satisfaction over their whole retirement period.

But to see this, it must be clear that there is not just an all-or-nothing retirement spending goal.  Goals are fungible.  Different people will have different takes on it. And everyone can find a strategy that will provide them with the most expected lifetime utility given the constraints they face about their guaranteed income sources and their level of retirement savings.