23 October 2015 - 7 minute read

Institutional investors rarely invest in just one hedge fund. Typically, exposure to hedge funds will involve investing with several managers. The value of these holdings will evolve over time – given performance differences in the underlying funds – and the investor or plan sponsor will eventually face the challenge of finding an effective rebalancing methodology.

Here, we attempt to shed some light on potential rebalancing strategies, by conducting a simple experiment to assess their advantages and disadvantages. We find that no single strategy performs best on all measures, and with a number of possible pitfalls considered, we are unable to find a strategy that clearly outperforms a simple reweighting strategy.

Methodology

Our focus is how an institutional investor might reweight their hedge fund portfolio over time based on the performance of the underlying funds. Fund selection or determining the number of funds to include in a portfolio is out of scope.

We examined the following rebalancing strategies, purposely keeping them simple to avoid overfitting and make it more likely that they retain their observed properties in the future [1]:

  • Buy and Hold – Each fund has an initial allocation of 5% and weights evolve over time due to fund performance.
  • Calendar Rebalancing – We reweight each fund to 5% of the portfolio at year end, resulting in various weightings during the year. This is our base case for comparison.
  • Profit Chasing – Each year we increase allocations by 5% to the two funds in each portfolio with the highest cumulative returns over the past 24 months. We decrease allocations by 5% to the two funds with the lowest returns.
  • Contrarian – The opposite of profit chasing: we increase our allocation to the two worst performers by 5% and decrease our allocation to the two best performers by 5%.
  • Volatility Weighted – We weight our funds by the inverse of their trailing 24-month volatility. This is a close approximation of a risk-parity strategy, but does not account for correlation between the funds.

We then simulate the performance, between 1995 and 2015, of each strategy on 2,000 portfolios of 20 funds selected at random from the Hedge Fund Research (HFR) database, with four funds selected from each of the Equity Hedge, Event-Driven, Macro, Relative Value and Fund of Funds groups [2].

In our first experiment, we only use funds with at least $200 million in assets. We remove the first year of returns to mitigate reporting bias, as investment managers report voluntarily and are more likely to report funds that start well [3].

We also take into accounts funds that are withdrawn from the reporting process by assuming a 25% loss occurs in the month following the end of the reported performance history. While some funds close because they are making money or have hit capacity limits, the opposite is much more likely to be true.

We reweight our portfolios yearly, and adjust weightings only after the full advance notice period and lockups of each fund have expired. This allows for operational restrictions that hedge funds may enforce on reweighting frequencies.

The turnover per annum was approximately 10 to 15% for each of the strategies, except buy and hold. Our results also take into account a 35 basis point turnover cost, which would typically only reduce returns by around 5 basis points per year; however, we do not show the results of this model here.

Hedge-fund horse race

To show how the rebalancing strategies compared, we ranked each on four performance measures from 1 (higher return, lower risk) to 5 (lower return, higher risk). Figure 1 shows the proportion of times each strategy achieved each rank.

Figure 1: Five rebalancing strategies ranked from 1 to 5 on four performance metrics

For instance, the top-left chart shows that the volatility-weighting strategy achieved the highest Sharpe ratio 40% of the time, whereas the profit-weighting strategy had the lowest Sharpe ratio 50% of the time.

Figure 1 shows how strategies that attempt to manage volatility or risk gave a higher Sharpe ratio, with a lower drawdown and volatility – but, in many cases, also deliver a lower return. This increase in Sharpe ratio shows that, accounting for the risk-free rate, the portfolio volatility was decreasing faster than return.

As we have discussed previously, we might expect drawdown to be proportional to volatility, given the consistent length of the return series [4]. The pattern for each statistic is roughly the same, but drawdown appears to be controlled in a less consistent manner than volatility. This could indicate that tail risk was more difficult to predict, although a larger confidence interval on the statistic is also to blame.

How great, how small?

The next question is whether the differences between rebalancing strategies are significant, reliable and useful. In Table 1, we report: a) the distribution of each measure over our samples; b) the probability that the strategy improves upon the calendar rebalancing strategy; and c) the distribution of that improvement.

Table 1: Performance statistics for five rebalancing strategies

For each rebalancing strategy, we show the distribution of each measure over our samples, the probability that the strategy improves upon the calendar rebalancing strategy; and the distribution of that improvement.

The buy-and-hold and profit-chasing strategies gave the highest return, but with large increases in volatility and drawdown. The average return for profit chasing was 0.71% higher than calendar rebalancing, with a disproportionate 2.85% increase in average volatility and a 7.83% increase in average drawdown. Volatility weighting reduced the overall portfolio volatility in 99% of cases and gave the highest average Sharpe ratio, although returns were 1.08% lower than calendar rebalancing on average. Calendar rebalancing gave the second highest average Sharpe ratio, with middling returns and a relatively low volatility and drawdown.

Autocorrelation

We saw no evidence of profit-chasing or contrarian strategies improving upon buy-and-hold or calendar rebalancing; in fact, they gave poorer performance in our experiments.

This could be explained by autocorrelation in returns. In Figure 2, we plot the 1-lag autocorrelation of returns in the HFR database, resampled to various intervals and accounting for fees. We saw strong positive 1-lag autocorrelation in monthly and quarterly returns that may be explained by illiquidity [5]. This effect is allegedly due to a smoothing of returns and it is typically argued that it is not possible to profit from it. We observed little significant correlation between returns over the yearly frequency at which we are rebalancing [6].   

Figure 2: 1-lag autocorrelation of returns in the HFR database

Resampled to monthly, quarterly and yearly intervals, accounting for fees.

Conclusions

We did not see a method that clearly outperformed calendar rebalancing, as this simple method gave some of the best results under all of our measures. In particular, the received wisdom that contrarian strategies outperform profit chasing is not reflected in our results.

We did see some indication that volatility-weighted strategies might increase risk-adjusted returns, but we noted that any gain would depend on the level of unreported losses in our database. The strategy had a lower return that an investor might be able to be use to maximise risk-adjusted returns through leverage, but this will depend on the cost of borrowing, restrictions and judgement of associated risks.

Buy-and-hold strategies may increase return, but are rarely implemented. This could be wise as they led to increased risk in our experiment, even though the differences were small.

Contrarian strategies were found to give returns on par with calendar rebalancing strategies, with an increase in associated risk. Profit-chasing strategies were found to increase return slightly compared to buy and hold strategies, but this came with a large increase in risk.

Overall, these results suggest that there is likely to be little to nothing to be gained by reallocating to managers based on recent performance, be it positive or negative. Given the uncertainty and risk associated with such quantitative strategies, we believe that the best policy is simple calendar rebalancing, as this has the best chance of creating a diversified portfolio without the need for assumptions. We do not believe this method is unbeatable, but we hope to have raised some concerns over how you might choose to deviate from it.


References

[1] The philosophy behind this can be found in Winton Research, Blinded by Optimism, 2013.

[2] When a fund stops reporting, we reinvest in a randomly chosen fund of the same strategy.

[3] A further discussion of this effect can be found in Adam L. Aiken et al., Out of the Dark: Hedge Fund Reporting Biases and Commercial Databases, Review of Financial Studies, 2012.

[4] D. Harding, The Pros and Cons of Drawdown as a Statistical Measure of Risk of or Investments, AIMA Journal, 2003.

[5] Andrew Lo et al., An Econometric Model of Serial Correlation and Illiquidity in Hedge Fund Returns, Journal of Financial Economics, 2003.

[6] Winton Research, Autocorrelation of Trend-Following Returns, 2013 offers a more comprehensive review of autocorrelation for trend-following hedge funds.

This article contains simulated or hypothetical performance results that have certain inherent limitations. Unlike the results shown in an actual performance record, these results do not represent actual trading.  Also, because these trades have not actually been executed, these results may have under- or over-compensated for the impact, if any, of certain market factors, such as lack of liquidity and cannot completely account for the impact of financial risk in actual trading.  There are numerous other factors related to the markets in general or to the implementation of any specific trading program which cannot be fully accounted for in the preparation of hypothetical performance results and all of which can adversely affect actual trading results. Simulated or hypothetical trading programs in general are also subject to the fact that they are designed with the benefit of hindsight. No representation is being made that any investment will or is likely to achieve profits or losses similar to those being shown.

The information included in the HFR products is obtained from various third party sources and is provided on an “AS IS” basis. HFR has not performed any audit on or verified the information provided by the third parties. HFR is not responsible for and does not warrant the correctness, accuracy, or reliability the Limited HFR Data in the RIA Materials. RIA has extracted and produced calculations from data contained in the HFR Products. HFR has not reviewed, verified, approved of or endorsed RIA’s work product. RIA, RIA’s company and HFR are not in any way related or affiliated, and HFR does not endorse any of the information or recommendations contained in the RIA Materials. Any information obtained from the HFR Products does not constitute (a) an offer or a recommendation by HFR to buy or sell any security or financial product whatsoever or (b) any type of tax or investment advice or recommendations. Past performance of investment product is no indication of future results. TO THE MAXIMUM EXTENT PERMISSIBLE BY APPLICABLE LAW, HFR EXPRESSLY DISCLAIMS ANY AND ALL WARRANTIES, EXPRESS OR IMPLIED, OR OTHERWISE, INCLUDING, WITHOUT LIMITATION, ANY WARRANTY OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE WITH RESPECT TO ANY INFORMATION IN THE RIA MATERIALS OBTAINED FROM AN HFR PRODUCT. USER ASSUMES ALL RISK OF DAMAGE OR LOSS RESULTING FROM THE USE AND/OR RELIANCE OF THE INFORMATION IN THE RIA MATERIALS OBTAINED FROM AN HFR PRODUCT.

Portfolio Construction