Retirement Portfolios - When to Rebalance?



The question of how often to rebalance a portfolio in retirement is a complex one.  For one thing, one must consider whether the rebalancing is done in a taxable or nontaxable account or a combination of the two.  Clearly, there are negative implications for those who must rebalance within taxable accounts.  Triggering taxable events often negates the advantages of periodic rebalancing.  The advice from most investment professionals seems to be that one should consider rebalancing their portfolio every year, if not every quarter.  This of course can trigger quite a few taxable events and transaction costs, particularly with quarterly rebalancing. With this in mind, I wanted to look at retirement portfolio performance as a function of rebalancing less frequently than what most investment advisers recommend.


I put together a small study showing historical 100% SWRs (inflation adjusted portfolio safe withdrawal rates) for 1927-2002 using different asset classes. I looked not only at the conventional year-end rebalancing but also at rebalancing at various thresholds, either by percentage change in the components of the asset mix or by time multiples of the base one year period.

The data comes mostly from the Global Financial Data sample series (

The pre-1994 small cap value (ScV) data comes from Kenneth French's data library: French's Library

The post-1993 data is the actual S&P SmallCap 600/Barra Value index data:

The pre-1994 ScV data was extracted from the deciles of French's "25 Portfolios Formed on Size and Book-to-Market (5 x 5)" file which most closely match the S&P SmallCap 600/Barra Value index data.  I chose this approach since currently the best widely available ScV index fund is, in my opinion, the S&P SmallCap 600/Barra Value index ETF (ticker: IJS).

As for the mechanics of the study, I looked at withdrawal periods up to 55 years starting in 1927 through 2002. This means only about 21 were full 55 year periods and the rest (about 55) were partial.  I used expense ratios of 0.25% for the S&P500, 0.4% for ScV, 0.35% for EAFE, and 0% for Treasury Inflation-Indexed Securities (TIPS) and T-bills. Expenses were taken out at the start of each year.

TIPS have only been available since the late 1990's so their performance history is short.  I simply modeled them at the historical CPI-U inflation rate plus a fixed coupon which I arbitrarily chose at 2.5%. In reality this approach more closely models TIPS' cousin, inflation indexed US savings bonds (I bonds), but the differences are not great and it is easier to implement in software code.




In the table below under rebalancing, values for '0%' and '1yr' are the same since at 0% (or greater) change the portfolio is rebalanced every year.  'X%' means that the portfolio is rebalanced only if at least one of the components changes by X percent or more from its base value at the end of the year (5% means, for example, that T-bills now comprise at least 30% or less than 20% of the portfolio rather than a targeted 25%).  'X yr' means rebalancing every X years regardless of percentage changes in the portfolio.  The last column, 'None', means no rebalancing at any time.


S&P500 ScV EAFE TIPS T-Bills   1 yr 2 yr 3 yr 5 yr   0.0% 5% 10% None
75%       25%   3.8% 3.6% 3.7% 3.7%   3.8% 3.7% 3.8% 3.7%
25% 25% 25%   25%   4.5% 4.4% 4.5% 4.5%   4.5% 4.7% 4.6% 4.3%
75%     25%   4.5% 4.9% 4.9% 4.3%   4.5% 4.5% 4.6% 4.1%
25% 25% 25% 25%     4.9% 5.1% 5.1% 4.6%   4.9% 5.0% 4.7% 4.4%
60%   40%     5.1% 5.4% 5.4% 4.5%   5.1% 5.1% 4.9% 4.4%
38%   38%   25%   3.9% 3.7% 3.9% 4.0%   3.8% 3.9% 3.9% 3.8%
75%     25%     3.9% 3.9% 4.0% 3.9%   4.0% 4.0% 4.0% 3.9%
      Average SWR: 4.4% 4.4% 4.5% 4.2%   4.4% 4.4% 4.4% 4.1%







One probably should rebalance periodically.  As you can see, the average SWR for the theoretical portfolios is less for the non-rebalanced portfolios compared to those that were periodically rebalanced.  However, the differences are not that great and if one has to rebalance within taxable accounts then he or she may want to consider rebalancing with considerable restraint, if at all.

Diversifying outside the S&P500 and cash (T-bills) really does help.  This is not really a surprise given what we know about diversification and the "free lunch" it provides.  A simple Coffeehouse-type portfolio (The Coffeehouse Investor) split evenly between S&P500, EAFE, ScV, and T-bills would've increased the SWR by a hefty 0.7% per year (4.5% versus 3.8% for portfolios rebalanced every year) over a 75:25 S&P500:cash portfolio.  I think that this is a great option, particularly if you believe in the efficient market dictum that all asset classes should perform about equally on a risk-adjusted basis going forward.  By the way, being able to withdraw an extra 0.7% per year is huge.  This means that you could withdraw nearly 20% more from your portfolio each year.

ScV was the real star performer during this time period but should be balanced with other equity classes or a sizeable dollop of fixed income to offset volatility. TIPS work especially well with ScV.  At the time of this writing ScV also is currently undervalued relative to the S&P500 and I would expect its outperformance to continue in the future, maybe even by a wider margin.

Even if one is rebalancing within nontaxable accounts the rebalancing really doesn't have to be done every year.  In fact, you might be better off rebalancing only after your portfolio components get seriously out of whack by 5-10% or every 3 years or so.  Even in a nontaxable account there may be some transaction costs involved so decreasing the frequency of rebalancing may be desirable even if taxes are not a consideration.  In fact,  I believe that there may be a reason that rebalancing about every three years seems to work just as well, if not slightly better, than rebalancing more frequently.  Wm. Bernstein  has shown that there is some momentum effect, albeit weak, with regard to short-term domestic and foreign equity returns .  As you can see in the table above, there is a very slight advantage to rebalancing every third year versus the other strategies.  The effect is so slight that I doubt that it is statistically significant.  However, both Bernstein's research and my own tend to indicate that mean reversion does not occur over periods much less than three to four years and that there may be some benefit to following the weak momentum that equity markets seem to exhibit prior to reversion to the mean. 


Last edited: 06/23/2004


Home ]