We Put the Bucket System Through Additional Stress Tests

My recent stress test of a bucketed retirement portfolio yielded an encouraging result--and many terrific reader comments.

Although I introduced the portfolios in 2012, I assessed how they would have performed had they launched at the beginning of 2007, the year before the financial crisis began in earnest. The model portfolio, combined with the bucket system, acquitted itself well during this admittedly arbitrary time period. (You can read more about the bucket system here.) Assets in the portfolio were higher at the end of 2012 than they were at the outset, despite the fact that the portfolio had supplied living expenses for a hypothetical retired couple during that time and withstood the biggest market drop since the Great Depression.

Readers said they liked being able to see the mechanics of portfolio rebalancing and withdrawals on a year-by-year basis, but they noted that different variations would be helpful. Some posters said my model portfolios--and in turn my stress test--were too complicated; wouldn't a good two- or three-fund portfolio do just as well? Others noted that because my simulation allowed for withdrawals at the end of each year, it didn't factor in living expenses at the very beginning of retirement.

To help address some of these issues, I conducted some additional stress tests, using different investment mixes and different assumptions. (Readers also said they'd like to see the portfolio stress-tested during longer time periods. All of my portfolio holdings didn't exist during the dot-com crash, but I'll be tackling a simulation from 2000 through 2012 in a future article to incorporate two bear markets.)

First Distribution at Start of 2007
Starting Portfolio Balance: $1,500,000
Ending Portfolio Balance: $1,489,816
Amount Withdrawn: $449,906

In my original simulation, I didn't factor in withdrawals for living expenses at the beginning of year 1 of retirement. What would have happened if a retiree turned to bucket 1--the cash component of the portfolio--to fund those year 1 expenses? For this new simulation I left the portfolio holdings--and the rebalancing and withdrawal assumptions--from my previous simulation intact. You can view the simulation in this spreadsheet (Microsoft Excel required) or this PDF. 

Perhaps not surprisingly, given that this new simulation includes an additional $60,000 year 1 withdrawal that wasn't there in my earlier stress test, the results of this simulation aren't quite as strong. The portfolio would have ended 2012 with $1,489,816 in assets, modestly under its starting value of $1.5 million. That's still decent, however, when you consider that it provided $60,000 in income, adjusted upward annually for inflation, in year 1 of retirement while also withstanding an epic bear market.

The simulation does demonstrate the value of getting a running start in year 1 of retirement--that is, not depleting a year's worth of living expenses from bucket 1 right off the bat. The portfolio system assumes that a retiree maintains a baseline of two years' worth of living expenses in bucket 1 on an ongoing basis, so raiding bucket 1 at the outset of retirement meant that the cash bucket was chronically underfunded relative to the two-year target. (In years like 2011, for example, cash dropped to just $14,362--a cushion that many retirees would deem uncomfortably small.)

Moreover, rebalancing proceeds from long-term assets were constantly being plowed into cash rather than being used to top up other long-term assets. Following 2012's equity market rally, for example, all rebalancing proceeds not used to fund living expenses were moved into cash, whereas in my previous simulation some of those rebalancing proceeds were moved into international stocks and commodities.

The takeaway from this simulation is that unless the long-term portion of the portfolio dramatically increases in value in year 1 of retirement and the proceeds can be used to top off the cash bucket, you're better off not raiding your liquidity bucket at the beginning of year 1 of retirement. An alternative to the assumptions in my simulation would be to come into retirement with three years' worth of living expenses in cash--with the understanding that ongoing liquid reserves would be two years' worth of living expenses. Alternatively, a retiree could fund living expenses in year 1 by using rebalancing proceeds from the last working year.

A Simple Portfolio
Starting Portfolio Balance: $1,500,000
Ending Portfolio Balance: $1,507,653
Amount Withdrawn: $378,548

What about a bucket system that's radically simpler than the above-mentioned portfolio composed of nine funds plus cash? To help assess the viability of a simple portfolio in retirement, I started this stress test with a fund that is the essence of simplicity-- Vanguard Balanced Index (VBINX)--plus two years' worth of liquid reserves. The fund consists of a 60% position in a total stock market index portfolio and a 40% weighting in a total bond market index portfolio, regularly rebalanced back to those targets. I used the same rebalancing and cash flow rules as in my initial stress test. The first withdrawal, from the cash bucket, came at the end of 2007, not the beginning. I rebalanced positions when they exceeded 110% of the starting value. You can view the simulation in this spreadsheet (Microsoft Excel required) or this PDF. 

The results of this exercise were solid. With $1,507,653 in assets at the beginning of 2013, the ultrasimple portfolio finished slightly above its starting value and supported our $60,000 annual withdrawals, adjusted for inflation. The simple portfolio's ending balance was, however, lower than the $1,637,996 final total for the multifund portfolio in my original stress test.

Asset allocation played a role in the balanced portfolio's underperformance versus my original simulation with the multifund portfolio. The balanced-index-plus-cash portfolio had a slightly higher equity weighting (about 4 percentage points higher) than did my nine-fund portfolio; that hindered its results during the period examined because bonds outperformed stocks. (That also explains one reader's observation that a position in  Vanguard Wellesley Income (VWINX), which features a roughly 60% bond/40% stock asset allocation, would've beaten my nine-fund portfolio during the time frame studied.)

The simple balanced portfolio also ran out of cash early. Because the balanced fund didn't run up enough in 2007 to warrant scaling back the position (it didn't exceed 110% of its starting value), we were forced to raid the cash bucket in both 2007 and 2008, leaving our portfolio with little to no cushion during the depths of the bear market. Cash ran dry again at the end of 2011, forcing us to liquidate part of the balanced fund to help meet living expenses even though it hadn't yet exceeded our 110% rebalancing threshold.

The multifund portfolio, by contrast, had various holdings hit the 110% target at various points in time; rebalancing out of those positions provided funding for living expenses and kept us from having to raid the cash bucket for those expenses. However, it's worth noting that a different rebalancing protocol might help address this shortcoming of the simple balanced portfolio. For example, one might scale back the stake in the balanced fund once it exceeded its original position size rather than using the 110% threshold, or employ a 105% threshold for rebalancing.

But even with a different rebalancing framework, a retiree wouldn't have the latitude to pick and choose which long-term subasset class(es) to sell and which to add to at various points in time. Instead, the retiree selling a single fund spanning multiple asset classes would sell proportionate shares of stocks and bonds at once, even though one of those asset classes may have enjoyed a big runup and the other had not. (The fact that the balanced index is, itself, rebalanced regularly helps mitigate this problem somewhat.)

The takeaway from the balanced-portfolio simulation is that even though simplification is a worthy goal, holding at least a single stock and a single bond fund gives a retiree more discretion over where to go for cash than is possible via multiasset funds. 

See More Articles by Christine Benz