Monday, February 10, 2014

Unraveling Retirement Strategies: Time-Segmentation

If you've been following my posts on retirement funding strategies, beginning with Unraveling Retirement Strategies: Systematic Withdrawals, you know we have come down to the last of what I consider four major categories of strategies, time-segmentation, or as they are sometimes called, “bucket” strategies.

About a fourth of financial advisers prefer breaking down long retirements into more manageable time periods using a time-segmentation strategy. (Systematic withdrawal strategies are the most popular, though I'm not fond of them.) One benefit is that we can build a plan from multiple smaller plans, in 5-year segments perhaps, instead of a single 30-year financial plan.

Another benefit is the ability to manage our assets for various future time periods, or “buckets”, based on investment horizon, investing in the best asset class for how far into the future our goal is.

In his classic book, Stocks for the Long Run, Wharton professor Jeremy Siegel noted that stocks historically outperform bonds and cash about 65% of the time for one to three year periods, but about 99% of the time for 30-year periods. The table below was created using data from his book. If we are financing a bucket twenty or more years into the future, we are likely to see better results from stocks than from bonds.
Likewise, we are likely to see better results from bonds than cash if our investment horizon is greater than a couple of years or so.

Some have interpreted this data as proving that stocks are safer than bonds if held for twenty years or more, but this data doesn't show stocks are safe. It just shows that they usually provide a higher return than bonds when held a long time. Stocks are risky no matter how long you hold them.

In a retirement investment portfolio, the higher the volatility (risk) of the investment, the greater the chance that the retiree might have to sell that investment after a price decline. Over short periods of time, there is a greater probability of having to sell stocks at a loss than bonds, and bonds at a loss than cash.

Put these risks and returns together and cash appears to be a better investment for liabilities up to three years in the future and stocks appear to be the best bet for funding retirement years roughly 15 or more years into the future. Bonds seem to be best suited to the periods in between.

Time segmentation exploits these characteristics by recommending that we hold enough cash to pay our living expenses for a couple of years or so and then fund the next 8 to 10 years with bonds. Any remaining savings are invested in stocks.

Time-segmentation is a much less-granular approach to liability matching than floor-and-upside. Floor-and-upside matches each future year of liabilities individually, while time-segmentation matches buckets of years. It doesn't match resources to future liabilities so much as it matches asset classes to future liabilities.

Another important difference is that floor-and-upside demands that all years of retirement be financed with the safest possible investments, while time-segmentation would risk the most distant buckets with stock investments in the hope of generating higher returns. Both strategies would invest short term in cash and intermediate term in bonds.

Perhaps the biggest difference between time-segmentation and safe withdrawal strategies is that SW determines a stock/bond allocation using MPT portfolio allocation, optimizing portfolio return at the desired level of risk (volatility). In other words, SW recommends your portfolio allocation based on how much risk you believe you can tolerate.

Time-segmentation calculates the cash and bond allocation based on the amount of desired spending over the next 10 years or so of retirement and invests the remainder in stocks. The two allocations can be meaningfully different.

As an example, assume a retiree saves $500,000 and expects intermediate bonds to return 5% and cash 3%. With a SW strategy, he decides he could live with no more than a 25% portfolio loss in a bear market, so he allocates 60% of his portfolio to stocks and 40% to bonds.

With a time-segmentation strategy and 4% annual withdrawals, he would purchase bonds and cash to provide ten years of spending $10,000 a year. That allocation would be about $155,516 to cash and bonds (31%) and the remaining 69% to stocks.

Given these expected returns, bucket sizes and withdrawal rate, a time-segmentation portfolio will hold about 31% bonds. The same retiree might choose a larger or smaller portion of bonds for a SW portfolio depending on her risk tolerance, providing larger or smaller amounts of upside and downside risk.

The typical spending strategy for time-segmentation is the same percentage-of-remaining-balance method employed by systematic withdrawals. Some advisers, however, choose a desired income, instead. Also like systematic withdrawals, there is no secured floor of spending with time-segmentation and there is upside potential for the retiree's standard of living.

The spending range for a time-segmentation strategy will look similar to that of a systematic withdrawals strategy, but perhaps with more or less less upside and downside risk. The cash and bond allocations needed for time-segmentation may result in a larger or smaller stock allocation than systematic withdrawals would dictate, as described in the example above, resulting in a different expected portfolio risk-return than with SW. (Although this chart indicates less risk than the SW chart shown in my previous blog, in some cases there will be more.)
Time-segmentation strategies are more often advocated by financial planners than economists, who prefer a secure income floor.

Longevity risk management is not as powerful with time-segmentation as life annuities or floor-and-upside. It will be similar to systematic withdrawals, depending on the bond and cash allocation you end up with.

Moreover, studies have shown that the high levels of cash using this strategy are a significant drain on portfolio return, as are the transaction costs of constant selling of securities and bonds to re-balance the buckets.

Many authors tout the behavioral finance benefits of time-segmentation. They say that this approach focuses the retiree on smaller pieces of the larger 30-year funding problem and makes him regularly re-evaluate his finances. In my opinion, if you aren't going to regularly evaluate whatever strategy you choose, you just just buy a life annuity. It's as close to a set-and-forget strategy as you will find and has less potential to get you into trouble if you ignore it.

Cash management strategies are a form of time-segmentation that advocate holding several years of spending in cash, typically at least five. The theory is that having this cash will make a retiree more comfortable in a market crash because she knows she won't have to sell stocks at their bottom.

I don't know about other retirees, but that logic doesn't work for me. When the market tanked in late 2007 and went on to fall well over 50%, I never worried that I couldn't pay my bills for the next five years. I worried that I wouldn't be able to pay them for the 25 years after that. I drew little comfort from my cash on hand as I watched my portfolio fall 15%.

Who would find time-segmentation strategies attractive?

Retirees who want to manage their future liabilities with greater granularity than the systematic withdrawals “large pile of wealth” method might prefer time-segmentation, but I'm not sure it's a lot less work to maintain than floor-and-upside, which manages liabilities on an annual basis.

I like investing money I won't need for decades, as time-segmentation mandates, in stocks more than I like the long bonds of floor-and-upside after a minimum secure floor is established. I hate long bonds and small cap growth stocks. Their returns don't historically justify their risk.

Yes, this introduces sequence of return risk. Long bonds add risk, too. No strategy is perfect. We have to choose our risk (see TIPS and Risk).

Retirees with smaller portfolios may be able to invest more in stocks for greater upside potential with time-segmentation than with floor-and-upside. Time-segmentation strategies will require perhaps half the bond allocation of a floor-and-upside strategy, since the retiree will be funding maybe ten years with bonds compared to 30 years with bonds for floor-and-upside. On the other hand, that makes time-segmentation riskier.

Some retirees will like focusing on the next five years or so of their retirement instead of the bigger picture, but I prefer to start with the bigger picture and work my way down.

In my next post, Build a Floor, Place a Bet,  I'll summarize and compare the four major strategies for funding retirement.


  1. Thanks again for another helpful post, Dirk.
    In your post "How many Rungs" (of TIPS) you presented a "My approach". I really liked that approach. It seemed like it allowed for more of the upside potential of a Time-Segmentation strategy with the increased safety (and minimizing sequence of return risks) of a Floor and Upside approach. A full "Floor" approach is, as you pointed out, extremely expensive at present. Your "my approach" would seem to be a much less expensive approach with not a lot more risk, especially if the TIPS ladder was kept extending to 7 to 10 years out.. If someone was able to afford it, they could always have a TIPS ladder out to 15 years or so and reduce the risk even more. It would seem to be a good compromise between cost and risk (and still allows the option of a fixed annuity in ones 70s instead of extending the TIPS ladder). Maybe in your next post (or future post) it would be interesting to hear if you think a hybrid of the different strategies is worthwhile for a lot of people to consider.

    1. Thanks, Derek. I think hybrid strategies are probably not uncommon. The important point I made about my strategy is that my standard of living is not at risk even if my stock portfolio fails.

      There are many things you might do, like use an annuity instead of a TIPS ladder, or intermediate bonds instead of the long portion of a TIPS ladder. If you understand how the pure strategies work, you can see how to pick parts of each to meet your individual needs.

      If you have other sources of flooring, like a pension or a large SS benefit, it opens up more options.

      I will be talking about this a little in future posts.

      Thanks for reading.

  2. I found your blog from reading your comments at Wade Pfau's blog. I have enjoyed the articles at your blog.