Blog Post

Portfolio Construction - Part 1 | Rebalancing: Active Decision or Housekeeping?

Related Articles
You're reading:

Portfolio Construction - Part 1 | Rebalancing: Active Decision or Housekeeping?

Alex Serman • June 06, 2024


Simple Assumptions with Significant Outcomes


Evaluating active investment performance depends on comparing our results to a representative market benchmark. This is “easier said than done” and yet the usual assumption is that the benchmarks we use are a good fit for our active portfolios in measuring their results. Why the disconnect between the level of confidence investors place in these benchmarks and the actual “goodness of fit” that we observe when comparing them to the portfolios?

Some of this “mismatch” is due to poor definition and understanding of the benchmark’s holdings and its structural characteristics. For example, the S&P 500 is often used to evaluate large company stocks, even though the index only holds 85% in large company stocks. An even greater mismatch is between a large company equity portfolio and the Russell 1000 index, which is a blend of its true large company index (the “Top 200”) and the 800 stocks in its mid-sized index. The difference in the patterns of returns in an active large-cap portfolio and either of these indices will generate a surprisingly-high amount of “tracking error” (the volatility of the difference in these “excess returns” that are caused by this structural mismatch.)

Does this mean that we simply need to “tighten up” our benchmark definitions, and use indices that are more closely aligned to the assets of the portfolio? That is certainly the first correction we should make, but we still face a source of structural mismatch between the benchmark and the active portfolio. The second source of return mismatch is caused by the difference in the rebalancing frequency of the benchmark and the portfolio. This source of tracking error is potentially large, and yet it is “off the radar screen” when it comes to the analysis of performance and risk.

In this paper, we begin our series of “portfolio construction” articles by examining the role that rebalancing plays in the market exposure – and the active risk - of the typical actively-managed portfolio.


Our study of the effects of rebalancing starts with a simple benchmark that we can observe every month for nearly a century. Using the S&P 500 and long Treasury bonds, we created a 50-50 allocation that we rebalanced each month. This is the critical assumption that applies to all benchmarks: benchmarks are rebalanced at the frequency of the underlying return data.

If monthly returns are available for the benchmark’s component sectors, then these segment returns are rebalanced to their strategic weights each month. This affects the relationship between risk and return for longer periods of time. If we examine this equally-weighted portfolio of stocks and bonds over the 98 years (1926 – 2023) we see a difference in results, depending on whether we use monthly or annual returns, as noted below. Does rebalancing frequency matter? Clearly, it does!


There is a familiar saying when it comes to rebalancing a portfolio:

“The trend is your friend.”

It is widely acknowledged that markets display momentum, where an asset that is earning above-expectation returns can be expected to continue to do so (for some unknown amount of time.) The same is true for assets that are losing value. How does rebalancing affect the portfolio’s returns in trending markets? We observe that rebalancing in trending markets hurts results, while not rebalancing enhances returns. After all, it makes sense to “let your winners run” which results in higher allocations (i.e., more money) earning stronger returns, producing a compounding effect in terms of monetary outcomes. At the same time, segments experiencing a trend of losses become smaller allocations in the portfolio; this protects portfolio capital from significant and sustained losses. The winning strategy is to refrain from regular rebalancing in trending markets.

We recognize that refraining from rebalancing is an active decision - it is essentially a tactical allocation effect.

Rebalancing in a flat-but-volatile market is another way to enhance performance through allocation. Since asset returns are “mean-reverting” in a flat market, automatic rebalancing will tend to “sell high” and “buy low” at each market movement. This decision to return to regular rebalancing is another tactical allocation decision, hoping to capture upside market movements via asset sales, while adding to segments expected to reverse their prior losses with higher-than-average returns.

Adopting any “trend following” rebalancing methodology is the equivalent of active tactical allocation - one that depends on “timing the market” in terms of its momentum - or lack of it. This also acknowledges the significant effect that rebalancing has on portfolio performance. This paper is not focused on active management; we mention this active rebalancing approach only to demonstrate the significance that rebalancing (and rebalancing drift) can have on portfolio results – especially regarding the mismatch of returns relative to the performance benchmark.


As noted, rebalancing methods have been the subject of investment research for years, with the benefits of trend-following rebalancing being widely understood. We have also seen research that concludes that there is a “rebalancing benefit” from following a disciplined rebalancing approach. This effect is closely tied to the underlying premise of efficient asset allocation, where maintaining diversification requires a reasonably constant allocation to each segment of the asset strategy.

Research on rebalancing has focused only on total return, and sadly, it is silent when it comes to the effect that rebalancing has on volatility risk. We will examine the effects of rebalancing on both total return and active return, while also examining these effects in the context of market risk and active risk.


We begin with market return and risk for our two asset classes (stocks and bonds) as well as cash, for the period 1928 – 2023. Using monthly returns, we created these annualized results:

The “Sharpe Ratio” line connects cash with each asset grouping, and the steepness of each line indicates the efficiency of total return relative to its volatility risk. Individually, bonds are the least efficient asset class, with a relatively small return premium over cash for its higher volatility. Equity shows a strong risk premium relative to its additional volatility risk. And the diversification benefit from combining these assets creates an equally-weighted benchmark with the highest payoff for volatility risk.


We propose five regular rebalancing frequencies: a) quarterly, b) semi-annually, c) annually and d) every 18-months. We also considered a fifth method - an “institutional” approach to rebalancing that is found in many “Investment Policy Statements” (i.e., “IPS.”) The IPS document typically specifies a target allocation to each segment, with a minimum and maximum allowable exposure. This range of exposures for each asset segment requires rebalancing only when the allocation exceeds these limits. While the segment exposures remain within these limits, no rebalancing is required - although the portfolio may be rebalanced at any time at the discretion of the portfolio manager.

While appearing to be a passive, this “rules-based” approach to rebalancing is really an active, tactical strategy with allowable minimum and maximum exposure limits by asset segment. Rebalancing is a matter of portfolio manager discretion, with asset allocation adjustments happening at any time and for any motivation. This does not imply that this approach is undisciplined; rather it indicates that the portfolio is not necessarily rebalanced as a matter of “housekeeping.” Instead, it is driven by the manager’s discretion and perspective on market opportunities.

In our analysis, we set the IPS “guardrails” at +/- 20% relative exposure, or a minimum of 40% and a maximum of 60% in each of the two segments. Rebalancing only occurred when exposures exceeded those limits.


Our baseline is the monthly-rebalanced benchmark, and our rebalancing approaches are compared in terms of total return and total volatility risk. Our routine rebalancing approaches of less than one year (quarterly and semi-annual) are linked in the graph of results, as are the rebalancing methods of one year or longer. The “rules based” rebalancing stands alone.

It becomes clear that the performance benchmark is rebalanced at a somewhat unrealistic frequency, since it is hardly practical to rebalance a portfolio every month, even when using liquid mutual funds for investments. If separate accounts are used for investing, it becomes unreasonable to issue paperwork to direct purchases and sales across all the products – especially when the amounts in question may be immaterial in size. For that reason, a manager may choose to rebalance regularly, but less frequently. Doing so requires understanding the implications in terms of relative risk and return.

We begin with the expected increase in return from quarterly rebalancing. In general, any momentum with each quarter was captured as an enhancement to return, which we estimate at about 15 basis points, with an increase in volatility of only 5 basis points. It is curious that the longer rebalancing approaches produced lower returns, with substantially lower volatility risk. One possible explanation is that momentum tended to reverse within the rebalancing period.

At this point, there is no conclusive “winner” in terms of an optimal rebalancing strategy. We simply confirm our thesis that there is an inherent mismatch between the portfolio and its benchmark, depending on its rebalancing strategy.


We strongly recommend that evaluating the effects of rebalancing requires looking at excess return and tracking error. After all, we began with the concern that rebalancing at a different frequency from the benchmark results in a different pattern of returns in the portfolio. This is an exercise in minimizing unnecessary tracking error relative to the benchmark.

We must also improve our calculation of excess return, so that we consider any additional return in the context of additional volatility risk. Therefore, our analysis will consider volatility-adjusted excess return, or “alpha.” This alpha is then evaluated in the context of the tracking error risk that it generates. Essentially, we are creating a “Rebalancing Information Ratio.”

What a difference we see as we shift our focus appropriately to risk-adjusted excess return (“alpha”) and tracking error!

There are now clear “winners and losers” in terms of the performance results driven by the rebalancing methods. The longest rebalancing method (18 months) is not only the least practical (and least-defensible to clients, who expect their portfolio to remain aligned with the long-term strategy) but it is also the least efficient.

The “rules-based” approach may appear reasonably efficient, but one must question whether nearly 90 bps of tracking error is acceptable to any client, especially when this is the result of what appears to be a “lack of good housekeeping.” Earning less than 20 bps for this degree of “mis-tracking risk” seems like rather skimpy compensation.

Even the lowest tracking error from quarterly rebalancing is at a surprising level of about 50 basis points. That is a truly startling result for what seems like a minimum amount of rebalancing “slippage” relative to the monthly-rebalanced benchmark.

It is essential for the performance measurement staff to communicate these results to the portfolio managers, so that reasonable expectations can be incorporated into the portfolio management process. Without the insights from this analysis, portfolio managers would assume that this tracking error was caused by the underlying funds in the portfolio.


We extend the idea of active return relative to active risk into the context of rebalancing, with our “Rebalancing Information Ratio.” This allows us to evaluate and rank the results of each rebalancing method, bringing together total return and risk, and also active return and risk. Our ratio is rebalancing alpha divided by rebalancing tracking error.

While the tracking error from quarterly rebalancing was the lowest of all methods, it was still a surprisingly-high value at 50 basis points. It is difficult to imagine clients accepting a higher degree of benchmark mis-tracking risk when this can be controlled through a routine rebalancing process. Longer rebalancing periods produce inferior results.


This test case illustrates one type of structural mismatch between a portfolio’s market exposures and the exposures of its performance benchmark; this was caused by a different rebalancing schedule. While this is a common situation, the tracking error that results may be an unwelcome surprise to many investment managers. Although this exposure mismatch risk seems to produce a positive return, we should still seek to minimize this effect. This means that we should focus on aligning the benchmark and the portfolio, rather than seeking excess return from a random factor such as rebalancing misalignment.

But how do we decide on the “best” rebalancing approach?

We suggest combining the routine tasks of withdrawing funds from the portfolio and rebalancing. Most institutional clients make quarterly withdrawals of funds to support their mission. The client’s “liquidity schedule” is a perfect time to “reap the profits” of the stronger returning sectors, turning investment gains into cash. Essentially, the “portfolio-value-minus-spending” amount is rebalanced to the target allocation. This approach addresses liquidity management, concentration risk and rebalancing risk in a single procedure.


Every case study has its limits, and it is wise to gather empirical evidence by testing several strategies over various timeframes. This helps to prevent the common error of “time series dependency” where results may be unique to only a single scenario. We are also trying to avoid an “inductive reasoning error” where we draw a general conclusion from a single, unique case.

The main challenge to testing our rebalancing thesis is the availability of data over a very long time horizon that captures many economic cycles. An additional challenge is that some of the asset classes and investment sectors that are used today did not exist - or were not readily available - even 20 years ago.

We developed a scenario using a more diversified strategy over twenty years, while maintaining the same 50-50 allocation of equity and bonds. The chart below shows the market returns for the 2 bond asset segments and 9 equity segments. Each of these 11 segments was rebalanced at the end of each quarter.

Our second scenario provided a similar quarterly rebalancing return enhancement, but with only half the tracking error that we experienced in our 98-year scenario. Our complete results are listed below.

Can we draw any broad conclusions from a comparison of these two studies? We think that would be unwise. These scenarios are different in terms of diversification, time horizon and economic context. At best, we can see that the “quarterly rebalancing tracking error” has a range that is significant – in this case between 25 bps and 50 bps. This is meaningful, and it warrants attention on an ongoing basis, so that this “otherwise-unmanaged” factor does not cause unexplained (and possibly detrimental) performance effects.


Portfolio rebalancing is an important decision that warrants more attention than it typically receives. We see that the various rebalancing approaches produce a surprising range of performance outcomes, and we believe that greater rigor is required in deciding which rebalancing approach is most appropriate for each client.

There are always trade-offs between the risk required to enhance return, but in the case of rebalancing, we believe that this is a decision that is passive in nature, with the goal of minimizing unwanted and unmanaged tracking error relative to the performance benchmark. We suggested a few approaches, and we provided some empirical evidence for investment managers to “get their arms around” this decision. The goal is to have a reasonable set of expectations around the set of portfolio management tasks. Our analysis and results shed significant light on this underserved topic. And we see that the most practical method for eliminating the “noise” of rebalancing mismatch may be quite simple: rebalance the performance benchmark at the frequency that matches the portfolio rebalancing.


As we continue in this segment, we will be tackling several provocative and compelling topics around the decisions that turn our asset allocation plan into an actual portfolio. The task of investment selection is typically implemented using pooled vehicles such as mutual funds. We will look at the benefits of assembling a “team of funds” that are evaluated holistically, in the context of their contribution to the portfolio. This is in stark contrast to the traditional approach of simply evaluating funds in isolation, relative to their market benchmarks. Our innovative approach provides insight into managing the active process in the context of client goals around earning enough to “pay the bills” to finance monetary goals through a combination of market return and an active return enhancement. We will also look at some challenging questions, such as whether our fund selection could be eroding our asset allocation, instead of delivering an enhanced market return. We will also examine the critical question on role of passive investments within an active, goals-based client mandate.

Written in partnership with Stephen Campisi