Figure 1

Figure 2

Figure 3

The impact of correlations

The first asset class characteristic we explored was correlations between portfolio allocations. Conceptually, lower correlations should lead to larger divergences between the portfolio allocations over time, as the allocations move less in line with each other, and higher correlations should lead to smaller divergences, as the allocations move more in line with each other.

To test these assumptions, we started with the baseline results from our simulation analysis but assumed two correlation scenarios: one in which the correlation between equities and bonds is 0.5 lower and one in which the correlation is 0.5 higher. The results made clear that correlation has an important but differing impact for calendar- and deviation-based approaches. Specific data is available on request, but here we highlight the key takeaways:

Calendar-based approaches — Lower correlations led to larger deviations for calendar-based approaches, as the assets tended to move less in line with one another over time. The larger deviations also increased turnover, given that the amount needing to be rebalanced was larger. Higher correlations led to smaller deviations and less turnover as the asset classes moved more closely together.

Deviation-based approaches — Lower correlations led to more turnover as the rebalancing bands were reached more frequently, while higher correlations led to less turnover. In both cases, there was minimal difference in deviation given the constraint of the rebalancing bands.

Risk/return relationship — When correlations declined, rebalancing resulted in increased risk-adjusted returns for all rebalancing approaches, as negative correlations imply a greater rate of mean reversion — so buying a recent underperformer and selling a recent outperformer would be beneficial. Higher correlations, on the other hand, led to lower risk-adjusted returns.

In summary, portfolios with more correlated assets may be easier to manage from a rebalancing perspective, given lower deviation and turnover, than portfolios with less correlated assets, regardless of the rebalancing approach used.

The impact of volatility

The second asset class characteristic we explored was the volatility of portfolio allocations. Conceptually, we can think of lower volatility leading to lower divergences between the portfolio allocations over time as the magnitude of movements is smaller, and higher volatility creating greater divergences as the magnitude of movements is larger.

Again, we started with the baseline scenario from our simulation analysis, but this time we assumed two scenarios: one in which volatility of both equities and bonds was 25% lower and one in which volatility was 25% higher. Like correlation, volatility had an important but differing impact for the calendar- and deviation-based approaches. Specific data is available on request, but the key takeaways include the following:

Calendar-based approaches — Lower volatility led to smaller deviations from target weights, as it resulted in smaller swings in asset values over time (i.e., the magnitude of their relative returns was smaller). The smaller deviations also helped to reduce turnover in the calendar-based approaches given that the amount needing to be rebalanced would be lower. On the other hand, higher volatility led to larger deviations and higher turnover over time.

Deviation-based approaches — Lower volatility led to less turnover as the rebalancing bands were not reached as frequently, while higher volatility resulted in more turnover. In both cases, there was minimal difference in deviation given the constraint of the rebalancing bands.

Risk/return relationship — Unsurprisingly, lower volatility improved risk-adjusted returns, holding all else equal, and higher volatility led to lower risk-adjusted returns.

In summary, portfolios with less volatile assets may be easier to manage from a rebalancing perspective, given lower deviation and turnover, than portfolios with more volatile assets, regardless of the rebalancing approach used.

Additional thoughts on liquidity, costs, and other complexities

Illiquid assets — Illiquid assets introduce an important constraint to the rebalancing process as they tend to be, by definition, difficult if not impossible to resize (a challenge many allocators faced in 2022 with the denominator effect). This can be particularly problematic when allocating to more volatile and less correlated illiquid asset classes. With this in mind, we think allocators should consider sizing allocations to illiquid assets in an “acceptable range” and not at a particular weight. The width of this range would be a function of the volatility and correlation of the liquid assets relative to those of the illiquid assets. The midpoint of the range can be viewed as the target weight, but allocators must be comfortable deviating from it over time.

A further complication arises when the illiquid assets are close to (or above) the top of the acceptable range, making it difficult to maintain the desired absolute and relative exposure to liquid assets as they are “crowded out” by illiquid assets. We think planning ahead for how the portfolio will adapt if the size of the liquid exposures shrinks can help mitigate the risk. One method is to develop a liquid beta equivalent exposure for each private allocation and rebalance to a set of total plan (liquid plus illiquid) beta exposures, with the liquid portfolio carrying the rebalancing burden. The downside of this approach is that the liquid proxy may shrink or even disappear at extremes. A stress-test approach that considers the rebalancing impact of a significant beta sell-off on the liquid portfolio (i.e., how much selling of liquid assets into market weakness is tolerable, what rebalancing tolerances should apply, etc.) can also be an effective way to determine an appropriate illiquid allocation size.

Higher transaction costs — Rebalancing is not free, as each transaction incurs a cost, directly reducing the value (i.e., the performance) of the portfolio. Higher costs should directly bias the trade-off between deviation and turnover toward accepting larger deviations. In turn, this may bias high-transaction-cost portfolios toward deviation-based rebalancing approaches, perhaps with wider rebalancing bands. If there are substantial variations in transaction costs for different allocations (i.e., some exposures incur higher costs to rebalance than others), then allocators can consider a similar approach to the one discussed above for illiquid assets, with an “acceptable range” as opposed to a particular weight.

Active managers — We view rebalancing as predominantly a beta decision, but recognize that in practice the rebalancing decision often involves managers, not asset classes. While the broad dynamics of manager rebalancing are similar to asset class rebalancing, there are a few nuances worth noting. For example, rebalancing to maintain equilibrium between multiple active managers within the same asset class (e.g., emerging market equity) often means selling the outperformer to add to the underperformer, even though this may feel counterintuitive. However, where managers exhibit “lumpy” profiles (extended periods of outperformance/underperformance) or where the allocator’s confidence in an underperforming manager is waning, rebalancing may not be advised, as the winning manager may keep winning. In general, we think allocators likely underappreciate the benefits of rebalancing and may be too quick to lose confidence in an underperforming manager, but ultimately the key to success is having a process for making these determinations.

Rebalancing with flows — While many portfolios experience continual inflows or outflows, we don’t think this should change the broad dynamics of the rebalancing decision. There is often uncertainty around the timing and size of the flows, so basing rebalancing on the flow dynamic would effectively “outsource” the rebalancing decision — not an ideal approach given the impact rebalancing can have on performance.

That said, predictable flows may be useful as a natural rebalancing mechanism. For deviation-based rebalancing approaches, predictable flows can help reduce the number of times a portfolio needs to be rebalanced. For calendar-based approaches, when the flows occur outside of the set rebalancing schedule, they can reduce the intra-period deviations experienced by the portfolio. In either case, this may effectively reduce transaction costs directly associated with rebalancing.

Timing the rebalancing — Allocators sometimes try to “time” the rebalancing process (e.g., make a discretionary decision to ignore a rebalancing signal), especially when markets are volatile and rebalancing can have a large impact on performance. It may be useful to think of this decision as being a function of 1) the size of the required rebalancing (which drives the magnitude of the impact of the rebalancing decision); 2) the confidence the allocator has in their near-term market views; and 3) the tolerance for getting the decision wrong. In general, when the size of the rebalancing is large, the confidence in the market view is low, and the tolerance for getting the decision wrong is also low, we believe the allocator should adhere to a disciplined rebalancing policy.

A rebalancing framework

In our analysis, the outcomes of all the rebalancing approaches were clearly superior to the Drifting approach (i.e., no rebalancing). But which rebalancing process is the right one? Given allocators’ differing operational, investment, and governance objectives, there is no “one size fits all” answer. Instead, we offer the following framework to help guide this decision.

Step I: Select type of approach: Calendar-based or deviation-based

The first step is to choose between a calendar-based and a deviation-based approach. The deviation versus turnover trade-off can be similar between calendar-based and deviation-based approaches, depending on the frequency of the calendar-based approach and the size of the rebalancing bands for the deviation-based approach. For example, our analysis suggests that a +/-2% deviation-based approach has a similar deviation versus turnover trade-off to a quarterly calendar-based approach.

Therefore, the decision between calendar-based and deviation-based approaches should focus on operational and governance characteristics (Figure 4). For many asset owners, a standardized calendar-based rebalancing schedule will be easiest to implement. The trade-off for implementation simplicity is that it is impossible to know in advance what the divergence between portfolio and target weights will be in any given period. (In addition, those very sensitive to trading frequency and transaction costs may find that even an annual rebalancing schedule is too frequent, which will likely lead them to prefer a deviation-based approach with bands wider than +/-5%.)

On the other hand, those that are most sensitive to deviations from target will likely prefer deviation-based approaches. However, allocators pursuing this approach must be willing to take on more operational complexity in exchange for more certainty on the size of the deviations. In particular, they will need a process to track deviations on a continuous (day-by-day) basis and a process for initiating rebalancing trades whenever the deviation bands are breached.

Figure 4

Step II: Select frequency (calendar-based) or band size (deviation-based)

Once the type of rebalancing approach has been determined, the next step is to select the frequency for calendar-based approaches or the band size for deviation-based approaches (Figure 5). This is where the trade-off comes in. If an allocator has lower transaction costs and/or higher asset liquidity in the portfolio and wishes to put more focus on reducing deviations, then that would indicate a bias toward higher frequency for calendar-based approaches and smaller bands for deviation-based approaches; in other words, approaches that accept higher turnover to seek smaller deviations. On the other hand, if an allocator already faces higher transaction costs and/or has lower asset liquidity in the portfolio and wishes to put more focus on cost reduction (transactional or operational), that would indicate a bias toward lower frequency for calendar-based approaches and larger bands for deviation-based approaches; in other words, approaches that accept larger deviations to seek lower turnover. Ultimately, larger deviations mean portfolios that may run with a volatility or market beta that is well above (or below) the long-term average. Each allocator will have to gauge their own comfort with those types of deviations.

Figure 5

It is also important to consider the effect of correlations and volatility on the trade-off between turnover and deviation. Higher volatility and/or lower correlations in the portfolio imply a larger trade-off, while lower volatility and/or higher correlations imply a smaller trade-off. Therefore, it may be helpful to think about the trade-off as a range of potential outcomes and consider one’s comfort with the deviation versus turnover outcomes across the different scenarios. For example, for a monthly rebalancing approach, if volatility is 25% higher than the baseline scenario, average annual turnover in our analysis would be about 5% higher and average deviation would be about 0.25% higher. So, an allocator focused on limiting turnover may want to choose a less frequent rebalancing interval if they are concerned about entering a higher-volatility regime. 

To read more, please click the download link below.