Friday, March 20, 2015

A Second Look at Time Segmentation Strategies

One of the retirement income strategies that I didn't discuss in my last post is Time Segmentation (TS). (See Unraveling Retirement Strategies: Time-Segmentation for a description of the strategy.)  I have a tough time nailing down my feelings on this strategy and here's why.

I looked at a Time Segmentation strategy portfolio next to a Sustainable Withdrawal Rates (SWR) strategy portfolio and I couldn't tell them apart. I pointed a bright light at the two of them and still couldn't see the difference. I put on stronger reading glasses than I normally wear and when that didn't help, I took off the glasses and squinted really hard. They still look a lot alike to me.

The big idea behind Time Segmentation strategies (sometimes referred to as "bucket strategies") is that retirees who hold five years or so of expenses in cash may be less likely to panic-sell in a market downturn. It appears comforting to many retirees to know that, no matter how the market behaves, their living expenses are covered for the next five years.

This is a behavioral strategy, not a financial one, and I will be the first to say that a strategy that lets retirees sleep at night has significant value, even if it isn't financially optimal.

I did a web search and found that I wasn't the first to note that these strategies seem to be "twins separated at birth." Michael Kitces wrote about it back in 2011 in a column entitled, "Research Reveals Cash Reserve Strategies Don't Work… Unless You're A Good Market Timer?"

An SWR portfolio is most often organized into asset classes like stocks and bonds, and perhaps sub-classes like small cap stocks or short term bonds, but it could as easily be organized as a TS portfolio with cash categorized as assets meant to cover immediate spending, bonds categorized as assets meant for intermediate spending, and the remainder listed as stocks for long term spending. The following diagram provides an example of a portfolio organized as a SWR portfolio and the same assets organized as a TS portfolio.

TS strategies also recommend spending first from cash, then from bonds, then from equities, but as the Kitces article explains, that is pretty much what happens when we rebalance a SWR portfolio. Rebalancing results in selling assets that have recently experienced the highest growth. If stock prices have fallen, rebalancing insures that it is other asset classes that will be sold. With rebalancing, stocks are sold after their price goes up, not down.

TS strategies use cash for near-term expenses, bonds for intermediate spending and stocks for growth to cover more distant expenses in a form of duration-matching, though less exact. But, so does an SWR strategy, although the common view of an SWR portfolio doesn't typically categorize its assets in that way.

In a paper entitled, "Sustainable Withdrawal Rates: The Historical Evidence on Buffer Zone Strategies", authors Walter Woerheide and David Nanigan showed that the drag on portfolio returns from holding large amounts of cash can be significant. In other words, the comfort of a large cash bucket can come with a heavy cost. According to the authors, the performance drag imposed by a large cash bucket actually leaves the typical portfolio less sustainable. Large cash holdings mean lower expected portfolio returns, and lower expected returns mean a higher probability of ruin.

As Kitces points out, a retiree with a 4% spending rate would need to hold nearly a sixth of her portfolio in cash to cover four years of spending and that has to be a drag on portfolio returns.

Having lots of money, as usual, helps with this problem. With a low withdrawal rate in the 1.5% to 2% range, a retiree can set aside four years of spending and still have a reasonable cash allocation. The performance penalty only comes when withdrawal rates exceed these.

It has also been argued that TS strategies reduce sequence of returns risk, but Moshe Milevsky showed in "Can Buckets Bail Out a Poor Sequence of Investment Returns?" that this strategy cannot always avoid sequence risk. When a retiree spends all his cash in a market downturn he can be left with an extremely risky all-equity portfolio, possibly before the bear market ends.

(The Milevsky paper is sometimes interpreted as saying that cash buckets cannot avoid a poor sequence of returns. Milevsky, however, simply offers a counterexample argument that shows the strategy doesn't always work.)

You could turn most SWR portfolios into a Time Segmentation portfolio simply by over-allocating cash. But, Woerheide and Nanigan tell us that we would actually hurt portfolio sustainability, not improve it, due to the performance drag of a large cash bucket. Milevsky showed that we can't depend on cash buckets to avoid a poor sequence of returns. This leaves only the behavioral benefit of a Time Segmentation strategy to distinguish it, so it becomes more of a different perspective on a SWR strategy than a unique approach, to my thinking.

If you have a lot of money and a low spending rate, you can hold a cash buffer that covers four or five years of spending without much damage to expected portfolio returns. The benefits will be largely psychological, but will have little or no financial cost. Retirees with a withdrawal rate of 3% or more may find the psychological benefits of this mental accounting worth the financial cost, but need to understand that it comes at a price.

I'm crossing Time Segmentation off my list of sound retirement income strategies, not because it is flawed or dominated, but because I don't believe it is distinct enough from SWR strategies to warrant separate consideration. In the end, it is largely a SWR strategy with perhaps too large an allocation to cash for its own good.

I believe Time Segmentation will provide a useful way for many retirees to view their finances, so just look at your portfolio from both perspectives. And, again, I don't want to minimize the value of making retirees comfortable, even if there are more efficient financial strategies. Over-withholding one's taxes, for example, isn't an efficient way to save money, but some people have trouble saving any other way.

I think a sub-optimal strategy is better than no strategy. Or, as my friend Peter is fond of saying, bad breath is better than no breath at all.

Please note the sidebar has been updated with a link to a recent New York Times column by Jeff Sommer on how often mutual fund managers beat the market.

Tuesday, March 17, 2015

Dominated Strategies, Illogical Strategies, Problematic Strategies and Strategies That Just Make Me Queasy

In Pure and Mixed Strategies, I noted that we can make life simpler for retirees by winnowing out dominated strategies and strategies that are logically unsound. I showed in Dominated Strategies and Dynamic Spending that dynamic updating of sustainable spending amounts dominates the SWR-Fixed and SWR-Variable strategies, so I cross those off my personal list of reasonable alternatives for retirement income plans.

I believe that there are strategies that are flawed, some logically sound strategies with problems, and others that just give me a queasy feeling. So, in this post, I will share my feelings on those three categories of retirement income strategies.

Dominated Strategies: There are strategies that appear to be dominated by better strategies and, according to game theory at least, should not be played. In previous posts, I noted that SWR-Fixed and SWR-Variable strategies are dominated by a dynamic updating strategy. Are there other dominated strategies that are often proposed? 

The RMD Strategy, in which the retiree bases her spending on the required minimum distribution amounts the IRS mandates for IRA’s is also a dominated strategy. (You can read about RMDs here.) Dynamically updating all important variables of sustainable spending will always perform better than a strategy of updating only the retiree’s age and new portfolio balance, as the RMD strategy does.

Logically Unsound Strategies: There are also retirement income strategies that simply aren't logically sound, like the "Spend Dividends Only" strategy. This strategy seems to be based on a misunderstanding of how dividends work.

I won’t devote a lot of explanation here, because the topic has been thoroughly vetted by others. I recently recommended this explanation by Canadian Couch Potato.  But, in short, when a company pays you a dividend on a stock worth $10.00, they pay you perhaps 30 cents in cash and the market immediately reduces your stock’s value to $9.70. You are no better off and no worse off. (You may actually be a tad worse off if you hold the stock in a taxable account because dividend payments can be a taxable event.) Furthermore, investing primarily in dividend-paying stocks will limit your portfolio’s diversification.

I cross Spend Dividends Only off the list because it is illogical and, as Wade Pfau recently put it, simply isn’t a valuable strategy. 

SWR is another logically unsound strategy when applied literally. It ignores conditional probabilities of failure. A common counter-argument is that no one implements SWR literally. I hope that is correct, but I doubt it.

Problematic Strategies: Now, let’s look at an example of a strategy that has not been shown to be dominated and that is logically sound, yet still problematic. The Bond Ladder and Longevity Insurance (BLLI) strategy proposed by Professor S. Gowri Shankar, for example, is logically sound but suffers from two problems.

(If the BLLI strategy is unfamiliar, there is a nice summary of it and several other strategies in this paper by Wade Pfau and Jeremy Cooper.)

The BLLI strategy proposes building a 20-year TIPS Bond Ladder and funding later years (beginning age 85 for someone retiring at 65) with Deferred Life Annuities (DLAs). The idea is that the retiree won’t have to give up control of his capital for the first 20 years – a common complaint with annuities – and that deferred life annuities are cheaper than immediate life annuities. (See Wade Pfau's Why Retirees Should Choose DIAs Over SPIAs.)

The BLLI strategy, however, has an inflation problem. Because DIAs will provide income well into the future, insurance companies typically won’t offer them with inflation protection and, of course, "well into the future" is when inflation takes its biggest toll. When inflation protection is offered, it does not cover the period from purchase to the first payout, in this example 20 years.

This is different than funding half of your retirement income with a TIPS bond ladder and the other half of the income with annuities. BLLI suggests exclusively funding the first half (or so) of retirement with a TIPS bond ladder and the second half exclusively with annuities.

The idea of maintaining control of capital and liquidity is also somewhat problematic, since the retiree will need the money in her bond ladder for living expenses to age 85 and can’t really spend it in an emergency. Funding the entire 30 years with a TIPS bond ladder and prematurely spending the most distant rungs is problematic enough, but it is possible that she won't live the full 30 years and may not need to spend that money ever. Diverting funds meant to meet living expenses during the first 20 years of retirement is significantly riskier.

Nonetheless, it is neither flawed nor dominated and stays on the list.

Strategies that Make Me Queasy: Some strategies can be logical and not dominated and perhaps not problematic to some, but still make me feel uncomfortable. To wit, the Floor-leverage Rule and Zvi Bodie's floor-and-upside strategy of 90% TIPS bonds and call options (LEAPS). Both are known as "barbell strategies" because they invest in extremely safe and extremely risky assets with nothing in between.

As I suggested in Hope and Your Retirement Portfolio, most retirees won't be comfortable with the possibility of losing all or most of their upside "hope" even with a comfortable floor in place. Retirees who purchase call options will often see those options expire "out of the money", in other words, worthless. The financial argument will be that the options served their intended purpose, even the ones that expired worthless, and that is correct. Most of the retirees I know, however, will find that cold comfort when they see a few ten thousand-dollar call positions disappear in a poof of smoke. Actually, it isn't even that dramatic. It's just not there, anymore.

Zvi Bodie's friend, Jeremy Siegel, seems to agree (download PDF). On multiple occasions Siegel has said, "You know, I find it a little strange — Zvi says he’s giving conservative investment advice, and then advising all your clients to buy call options."

The Floor-leverage Rule makes the LEAPS approach look tame in comparison. The idea for its upside portfolio is to employ triple leverage at the equity end of the barbell. The typical investor can't purchase a stock portfolio with triple leverage. It's illegal. But, as Sharp and Watson point out, they can purchase shares of a 3x leveraged ETF like UPRO.

I'm crossing Floor-leverage off the list of reasonable strategies not because it makes me queasy to apply huge leverage to my entire upside portfolio (it does), but because the upside strategy doesn't work. The problem is that these ETFs are not the same as a triple-leveraged stock index fund. In fact, they aren't portfolios of stocks, at all. They're portfolios of derivatives that only track stock indexes for short periods of time. They're best suited to short term investments, which shouldn't be part of a retirement income plan.

Bodie's options strategy stays on the list. It's riskier than I can accept and I don't think it will fit the temperament of most retirees, but it isn't flawed. It will outperform other strategies in some scenarios, so it isn't dominated.

Some of the strategies I've crossed off the "sound" list may fit your individual financial situation and may be bets you're willing to take. Some of them will even be recommended by advisers. I would pare down the list of reasonable retirement income strategies by at least a third, as shown in the following table. This is my personal perspective and not everyone would agree.

Friday, March 6, 2015

Glide Paths

Retirement researchers Michael Kitces and Wade Pfau published papers on glide paths in the Journal of Financial Planning in 2014 and 2015, including Retirement Risk, Rising Equity Glide Paths, and Valuation-Based Asset Allocation, suggesting a rising glide path for retirees beginning with a low equity allocation early in retirement and rising throughout. Researcher David Blanchett has published contradicting studies that show better performance from declining equity glide paths throughout retirement, which has been the traditional recommendation, in Revisiting the Optimal Distribution Glide Path.

As you can imagine, these conflicting recommendations have caused more than a little anxiety among advisers and do-it-yourselfers and experts like William Bernstein have even entered the discussion.

Ultimately, the research has left some retirees asking whether they should follow a rising glide path or a declining glide path.

The correct answer is "none of the above" – they should follow their personal financial situation as retirement progresses. (See Dominated Strategies and Dynamic Spending.)

Imagine two 90-year old retirees. Rising glide paths would suggest that they both hold large equity positions, while declining glide paths would suggest they both hold large bond positions. Now imagine that one still has tons of savings left, more floor income than she needs and is, as William Bernstein puts it, investing for her heirs. The other has barely enough wealth left to cover another five years of expenses. Should both invest with the same asset allocation because they are the same age? A similar argument can be made at any age and the correct answer is that we need to consider more than age or retirement date when determining our asset allocation.

This research aims to isolate one factor, the retiree's planned retirement date (or, roughly speaking, the retiree's age), and to determine its effect on the retirement investment problem. The question it attempts to answer could be framed in a number of ways, like:

  • What would be the best asset allocation pattern for a retiree if we only knew that retiree's planned retirement date and nothing more?

  • What allocation should we recommend to a retiree who wants to do little or no investment planning? (Note that this person is unlikely to be a reader of this blog.)

  • What glide path should be implemented by a target-date mutual fund that serves a broad target market of retirees who happen to share the same planned retirement date?

I feel certain that all of the authors of these papers would agree that a customized, individual retirement plan will always be preferable to one based entirely on the retirees age/retirement date. (One of the authors actually told me this, so it isn't entirely a guess.)

As explained in Wade's post of March 2nd on, To Rise or Not to Rise, the authors are currently comparing notes to understand why their results differ. Initial thoughts are that the differences are explained by different capital market expectations. Both sets of expectations, or assumptions about future markets, are imperfect predictions and either (or neither) might turn out to be right.

As Wade explains, "… The choice of glide path is not always fixed, as it does vary with the stock market valuation level at the time of retirement."

So, which glide path is best if we ignore everything about a retiree except the planned retirement date? We're not sure, but the research does seem to show that it depends on stock market valuations and interest rate expectations at the time of retirement (no surprise).

How does this information help the typical retiree invest her savings if she doesn't really care how the mechanisms of retirement funding work? It doesn't help much, in my opinion, at least not at this point. As Blanchett suggests, more research is needed.

More importantly, this information won't apply to most advisers or do-it-yourselfers because they will be willing to do more work than to simply write down a planned retirement date. They can find a much better fit.

An individual household's optimum asset allocation should be estimated by considering several more factors, including how well their retirement is funded and the maximum loss they could stomach in a bad market crash. By continually managing asset allocation based on current conditions of the retiree's finances, at the end of retirement that retiree's glide path will have followed some curve that will probably look like one of the half dozen or so currently recommended glide paths simply because they pretty much cover the entire range of possibilities. (The 2014 Investor's Guide issue of Money magazine identified six glide path recommendations that had little in common. Pozen's has a tongue sticking out the end. What's up with that???)

We will not know in advance which glide path that retiree will end up following. It will be determined by a random walk through market returns, how long we live, our expenses, the decisions we make and many unpredictable factors.

It's important to understand that research of any kind is new information, or a new argument, that we should consider in light of what we already believed to be true. It doesn't automatically replace what we thought we knew. (See Think Like a Bayesian Pig.)

When this evidence conflicts, it gives us a chance to learn more. This research doesn't tell us that one glide path is better or worse — both sides make good arguments — but it begins to tell us under what conditions one glide path might perform better. That doesn't necessarily mean one is right and the other wrong or even that a well-planned retirement needs one. Please keep that in mind when you read academic papers. They are usually new evidence or new arguments and not universal truths that dispel old beliefs.

What should a retiree do with this information about glide paths? Probably nothing. You should customize your asset allocation instead of following a glide path. There are a couple of interesting points you can use, however. First, it turns out that a 60% equity portfolio throughout retirement isn't a bad "couch potato" allocation.

Second, as Wade points out, you might give more consideration to cash and short bonds than to intermediate bonds. "Another important point in that summary is that we did look at both short-term (6 month or 1 year) bonds as well as 10-year bonds in our analysis, and we found that the shorter-term bonds were of much greater help than longer-term bonds. When the focus is on protecting from downside risks, the additional volatility caused by the 10-year bonds hurt retirement outcomes by more than could be compensated by their higher average yields. This ties into David’s article as well, since his bond allocation was 75% to (I believe) 10-year bonds, and 25% to cash."

This is an important point to remember when you think about taking on more risk to increase the return of the bond portion your retirement portfolio. Bonds and cash are there to protect against downside risk, not to maximize portfolio return.

You can also just follow the ongoing research. I find the discussion quite enjoyable. It's a work in progress.