portfolio rebalancing when it works
In the previous blog we saw how volatility drains your compounding gains and how diversification is an effective tool to limit the impact of volatility drain.
We worked with a toy example of 2 assets – Nifty 50 and Gold and saw how a quarterly rebalanced portfolio yielded a better CAGR than both the assets.
What happens if we don’t rebalance our initial allocation?
We will return to this, but before that, let’s quickly understand rebalancing and how it differs from buy-and-hold approach.
Contrary to popular perception, rebalancing to a fixed weight is NOT a passive strategy.
Suppose you invest in a 60:40 equity-bond portfolio. Rebalancing back to 60:40 monthly (or any other frequency) is an active strategy. If you do not rebalance and just buy and hold, you allow the weights to drift away from the initial 60:40 allocation. This is a passive strategy.
Rebalancing can come in many flavours depending on frequency of rebalancing, rebalancing thresholds (bands) and so on. In this blog, we will assume that we are working with a fixed allocation portfolio that is rebalanced annually.
Optimal rebalancing frequency and optimal rebalancing bands is a topic for another blog.
Let’s now come back to our toy example. We will consider 4 portfolios:
The chart below shows the growth of Rs 1,000 invested on 1st Jan 2007 till 31st Dec 2021.
Table below provides summary stats for the 4 portfolios:
We can clearly see that a rebalanced 70:30 portfolio (Portfolio 4) delivers the best CAGR while the performance of buy-and-hold 70:30 portfolio (Portfolio 3) is very similar to Nifty 50 and worse than Gold.
A closer look reveals that average annual returns of Portfolio 4 is same as weighted average of Nifty 50 and Gold returns – 0.7 x 15% + 0.3 x 12%. Average annual return of Portfolio 3 is closer to Gold.
Why is this happening? Let’s look at the chart below that shows how weights changes over time in Portfolio 3 and 4:
As is evident, for a buy-and-hold portfolio, the weights are all over the place. We started with 70:30 in 2007. By 2012, the weights had drifted to roughly 50:50. On the other hand, for a rebalanced portfolio, weights remained stable at 70:30.
In this example, weights revert to 70:30 in buy and hold portfolio because we see a period of outperformance by Gold followed by a period of underperformance. In general, and over a longer horizon, portfolio starts getting dominated by outperforming assets and hence the benefit of diversification is lost.
And that’s why rebalancing works. In the long run, it prevents the portfolio from being dominated by fewer assets and hence “preserves” the intended diversification. This will become clearer in the next section where we conduct some monte carlo experiments.
We will now design some monte carlo experiments to compare the performance of buy-and-hold strategy vis-à-vis a periodically rebalanced strategy.
We consider 2 assets (A and B), both with expected one-period (annual) return of 15% and volatility of 30% (closer to sample mean and volatility of Nifty 50). Expected CAGR of both these assets is roughly 10.5% (refer to the blog on Variance Drain).
Consider 2 equal-weighted portfolios (50% in each) held for 30 years. Portfolio 1 is buy-and-hold while portfolio 2 is rebalanced quarterly. We vary correlations between A and B from 0 to 1 (perfectly correlated). For each correlation, we generate 1000 scenarios of prices of A and B and calculate the 30-year CAGR for each scenario. We then report the average CAGR for both the strategies.
Chart below shows the results of this experiment:
As we can see, rebalancing clearly improves CAGR and the improvement is more pronounced when assets are uncorrelated. As the correlation increases, performance difference between between buy-and-hold and rebalanced strategy decreases.
This is expected because rebalancing adds value by “preserving” diversification. When you hold 2 highly correlated assets, there isn’t much diversification to preserve.
In our previous experiment, we considered 2 assets and changed correlation while keeping the weights fixed at 50%. In this experiment, we will use a set up that closely resembles classic stock-bond portfolios. We will keep correlation fixed (sample) and change the weights.
We consider 2 assets (stocks and bonds). Returns of Nifty 50 Index and Nifty 5Y GSEC index are used to estimate expected return, volatility, and correlation. Since correlation is fixed, in this experiment, we consider different weighting schemes with stock weight going from 0% to 100%.
Rest of the details are same as the first experiment.
Chart below shows the results of this experiment:
We can see that when Nifty 50 weight is below 40%, buy-and-hold strategy fairs better. Plausible reason could be that for portfolios with equity weight less than 40% have very low volatility and hence impact of variance drain itself is limited. Therefore, allowing higher expected return asset (stocks) to drift (and dominate) yields better performance than rebalancing.
On the other hand, above 40%, we clearly see the value added by rebalancing. Controlling volatility through rebalancing when allocation to riskier asset is high adds a CAGR premium over the long run.
So far, in all our experiments, the horizon was 30 years. What happens if the horizon is very short?
We conduct a third experiment. The setup is exactly the same as the first experiment except that the holding period is 1 year instead of 30 years. For rebalanced strategy, we rebalance monthly.
This is yet another interesting result. It shows that for shorter horizons, rebalancing doesn’t really add any value. Performance of buy-and-hold strategy is similar to or better than rebalanced strategy. And if we take transaction costs associated with rebalancing, a short-horizon investor is better off buying and holding rather than rebalancing.
However, this result directly stems from assumption that asset returns are independent and identically distributed (IID). If we instead assume some form of mean reversion, it can be shown that rebalancing adds significant premium even for shorter holding periods. Interested readers can refer this paper – Dubikovsky, Vladislav and Susinno, Gabriele, Demystifying Rebalancing Premium and Extending Portfolio Theory in the Process (May 20, 2015).
Results presented here are based on estimates and certain distributional assumption (normal). In reality, we do not know the true distribution (and the parameters of the distribution). Results of these simulations convey a broader point and should be construed accordingly.
We also note that if we use a heavier-tailed distribution such as student-t, the impact of rebalancing premium is further magnified
We have seen conditions when rebalancing adds value and when it doesn’t.
Rebalancing is likely to add significant value for a (1) long-term investors who (2) diversifies across relatively uncorrelated assets and (3) takes decent exposure to riskier assets.
Most of the investors satisfy (or should satisfy) the above 3 conditions. In that case, one should rebalance rather than buy and hold.