back to Home Page
Client-Only Articles
Email:
Password:
 

 

Updated February 2021

REBALANCING

Abstract:  Research done by EAM and by DFA indicates that rebalancing may not augment returns due to the high statistical variance in historical financial data.  These make certain rebalancing strategies look good for certain time frames but still leave it unclear whether rebalancing will help or hurt returns longer term.  Various rebalancing strategies are discussed and EAM's preferred rebalancing strategy is presented in this article.


When EAM plans portfolio allocations in passive and index portfolios specific asset class targets for allocations are quantified.  Based on historical financial data these target allocations provide a rough estimate of probable forward risks and returns for a given mix of asset classes in a portfolio.  Most advisors endorse the notion that relatively frequent rebalancing to the original target allocations is beneficial and smooths and augments portfolio returns.  Often monthly rebalancing is suggested.  Rebalancing is often touted as a value-added strategy, a "sell high, buy low" discipline which involves selling those asset classes which have gained the most and buying those that have gained the least.  Two approaches are typically used for rebalancing, calendar based, say every six months, or percentage based rebalancing whenever the deviation from the target allocations exceeds a certain percent.

In 1999 EAM (Evanson Asset Management®) compared the effects of changing portfolio allocation weights on portfolio returns and volatility.  For 1973 through 1998 we constructed a hypothetical portfolio allocating 10% each to U.S. Large Growth, U.S. Small Growth, U.S. Large Value, U.S. Small Value, International Large Growth, and International Small Growth. The remaining 40% was split equally between a one-year fixed bond portfolio and a five year fixed bond portfolio resulting in a traditional "normal" or 60 equity/40 fixed income portfolio found in pension plans.  These asset classes were chosen because DFA data was available back to 1973 and included the 1973 and 1974 bear market as well as the 1982 through 1999 bull market.  They provide some perspective on how shifting asset allocation percentages in portfolios affects portfolio returns and volatility for a 25 year time frame.

Let's see what happens to returns for 1973 through 1998 as we allow first the S & P (U.S. Large) to become overweighted, then multiple growth sectors, then value stocks, then small stocks.  A well allocated, balanced baseline portfolio provided a compound annual return of 13.28% with a standard deviation of 13.81%.  As we let the S & P 500 become overweighted, ramping it up in 3% increments from 10% to 25% of total portfolio value, returns decline slightly to 12.83 but so does risk with the standard deviation dropping from 13.81% to 13.29%.  Overweighting the three growth asset classes produces a return of 12.73% and standard deviation of 13.26%. Overweighting the two value asset classes produces a return of 13.63% and standard deviation of 14.16%.  Overweighting the small cap asset classes produces returns of 13.65% and a standard deviation of 14.26%.

As these portfolios became increasingly out of balance little significant change occurs in returns or volatility.  The range of returns fell between 12.73% and 13.65% with standard deviations (volatility) increasing slightly as returns increased slightly.  Given that the standard error of the mean return, a measure of sampling error, is greater than this range of returns this analysis indicates that within fairly broad ranges shifting equity asset class weights doesn't have a statistically significant effect upon returns and volatilities.  And yet 25 years is a relatively long investment time frame and a typical "investment lifetime" for retired investors.

Another way of conceptualizing rebalancing is to compare the effect of different rebalancing frequencies on compound total returns.  Let's take a portfolio with 70% in equities and 30% in fixed income across six different equity asset classes including growth and value, large and small, Reits, and U.S. and international equity asset classes.  For January 1975 through December 2000 monthly rebalancing produced a compound total return of 3923%, quarterly rebalancing, 3959%, yearly rebalancing, 3971%, and bi-year rebalancing, 4233%.  Less frequent rebalancing resulted in higher returns.  Thus, for this 26 year time-frame monthly rebalancing actually reduced compound longer returns by 7.3% compared to bi-yearly rebalancing.

We then reran the original 60/40 portfolio data using January 1972 through December 2000 since is that is as far back as the data for this asset class mix was available.  It includes the dramatic speculative markets  of 1999 and 2000 in internet/tech stocks.  Once again portfolio returns and volatility differed little as a function of rebalancing frequency.  In this case frequent rebalancing increased volatility (standard deviation) slightly for that 29 year period.  Quarterly rebalancing produced a standard deviation of 10.71% compared to 10.31% for bi-yearly rebalancing.

Truman Clark of DFA examined rebalancing in detail in three papers published online on the DFA website in Fall of 2001.  He concluded that, "the proposition that a rebalancing strategy can increase expected return is dubious," and that "rebalancing costs definitely reduce expected returns."  He cautions advisors that rebalancing "simulations with historical returns may provide misleading estimates of the benefits...don't employ naive, mechanical rebalancing rules... and don't rebalance just to appear to be doing something."

Marlena Lee, a Ph.D. research statistician with DFA, reexamined rebalancing strategies in a paper published for DFA advisors in October 2008.  She looked at returns and standard deviations for February 1996 through December 2004 and rebalancing frequencies of 1, 5, 10, 20, 60, 125, and 250 market days for rebalancing when targets were 5%, 10%, 15%, 20%, 25% and 100% from their initial targets.  For example, if one's rebalancing trigger is +/- 20% then an initial allocation asset class allocation of 10% would be rebalanced when it was below 8% or over 12%.  Her target portfolio was 25% S&P 500, 20% Russell 2000, 5% DJ commodities index, 10% DJ Reit index, and 40% in a US government bond index.

Marlena Lee's comments and results are consistent with our earlier analyses and those of Truman Clark.  She also reviews studies examining the relationship between rebalancing and returns and notes the lack of robustness in historical asset class data,"Any analysis must use historical realized returns which are very noisy."  Very noisy.  It takes at least 20 years of returns to statistically distinguish the equity premium from zero and for size and value premiums it takes 20 and 30 years respectively.  "One must be cautious when interpreting reported benefits of rebalancing strategies that are formulated using historical returns because noise (volatility) in realized returns will make certain strategies look good for certain periods."  For the 1996 through 2004 period she concludes "...the noise in historical return data make it impossible to say one (rebalancing) strategy outperforms better with any statistical certainty."  This also means that investors who engage in rebalancing strategies will not know if the practice hurt or helped their portfolio returns for at least a few decades.  Advisors have some interest in persuading clients that rebalancing adds value and justifies their fees but it's unlikely rebalancing can be counted on to add value for investors.

Marlena Lee also looked at rebalancing for a much longer period, July 1926 through June 2008.  For this analysis asset allocation targets were 12% each in large value, large neutral, and large growth, 8% each in small value, small neutral, and small growth, and 10% each in one-month Treasuries, five year Treasuries, long Treasuries, and long corporate bonds, a 60% equity and 40% fixed income "balanced" DFA portfolio.  DFA offers its balanced portfolios as examples, not recommendations.  Dr. Lee performed an analysis of risk and return using Fama-French's five-factor returns model and concluded that, "Aside from some large positive differences in the 1930's....the rebalanced portfolios do not appear to be reliably larger than portfolios rebalanced seldom or never" and "Over the entire 82-year sample it is clear that the rebalanced portfolio had no significant advantage over the portfolios rebalanced less frequently" and "..trying to predict which strategy will have the highest returns going forward will likely lead one down a path of unproductive data mining" and "Aside from avoiding excessive trading there are no optimal rebalancing rules that will yield the highest returns on all portfolios and in every period."

Thus, research by DFA's statisticians clearly indicates that rebalancing has unknown consequences and may even reduce portfolio returns.  Probably the best rule of thumb is to look at the overall stock/bond/alternative asset ratios quarterly since they are the primary determinants of expected returns, volatility about those returns and worst case portfolio drawdowns.  That's what we do at EAM.  EAM recommends rebalancing when asset classes and particularly, the equity/fixed/alternative ratios drift out of balance enough to produce a statistically significant difference in forward expected returns, volatility and drawdowns.  As noted elsewhere on this site in "Portfolio Design" each increment of 10% in equities increases probable forward returns by about 0.4% per year in equities while increasing the risk of potential portfolio declines in equities by about 6.5%.  Thus, a diversified portfolio 100% in equities might be expected to return 4% more per year long-term than one 100% in high quality short term fixed assets with a potential top to bottom decline in a worst case global bear market of around 65% for equities and only 4% for short term fixed income, a huge difference in volatility.  In EAM's experience the need for rebalancing is a function of market and asset class returns with big market moves like 2003 through 2007 more likely to trigger the need for rebalancing.  Due to positive correlations between equity asset classes most of the time they move up and down somewhat together but it is impossible to predict the causes of these moves and when they will occur.

EAM will rebalance portfolios whenever clients request but recommends it only be done when percentage targets in asset classes have drifted sufficiently from initial targets to alter statistically significant probable forward returns and risks.

Site Navigation