Business

Hedge Funds, Expensive Beta, Low-Cost Alpha – Replication Is Better

Hedge Funds, Expensive Beta, Low-Cost Alpha – Replication Is Better

May 23, 2016

by Maneesh Shanbhag

Advisor Perspectives welcomes guest contributions. The views presented here do not necessarily represent those of Advisor Perspectives.

Hedge fund returns, like most strategies, are a combination of market risk (beta) and manager specific risk (alpha). Depending on an investor’s goals with a hedge fund investment, high risk-adjusted return or diversification, replicating the hedge fund in order to avoid the detrimental effects of high fees is better than a direct investment.

For investors seeking high risk-adjusted returns, or alpha, many top managers can be copied sufficiently well. This is because one only needs to know their largest positions with some timeliness (most managers provide more than enough of both with monthly and quarterly reporting). If one has high confidence that a top performing hedge fund manager will repeat their outperformance in the future, replicating the manager’s portfolio should outperform their net-of-fee return. We review a formula that can be used to determine whether one can successfully replicate a manager.

Alternative beta exposure, for diversification, is the other common reason for investing in hedge funds. Investors hope the beta from strategies such as trend-following and event-driven trading diversifies their stock and bond allocations even if their returns may be in line with these traditional asset classes. We show how to replicate many alternative betas using only a few publicly traded asset classes. This begs the question of whether alternative betas should be diversifying to stocks and bonds in the first place. We would argue not and, therefore, the only potential benefit of hedge fund investing in the first place is high alpha potential, which is extremely difficult to achieve.

Hedge Funds – Mind your alphas and betas

Following the dot-com collapse in equity prices, hedge funds were pitched as the solution to market volatility, sold as having the ability to generate high returns in any market environment and thus worthy of their “2 & 20” fee structure. After failing to deliver on this promise of high returns, the industry has pivoted, claiming hedge funds are a diversifier to traditional assets. This too has proven false. The chart below shows how a passive portfolio of stocks and bonds tracked the broad index of hedge funds with more than 80% correlation over the last decade. The hedge fund industry, with over $3 trillion under management, has become the market.

Hedge Funds

Hedge funds is the HFRI. Replication is 50% MSCI World, 20% Barclays US Treasury Index, 30% T-Bills. Source: Bloomberg. Data from Jan. 2005 to Dec. 2015.

Hedge fund returns, like all strategies, can be separated into their component drivers of return:

Total Return = Risk-Free Rate + Beta (buy and hold market risk) + Alpha (skill-based return)

From the chart above we can see that a broad index of hedge funds delivers no skill-based return – both portfolios earned 2.9% annualized over cash since January 2006. When we dissect how much of the hedge fund return is driven by beta risk versus alpha risk (tracking error), it is mostly beta, as illustrated in the pie chart below. While the 21% alpha risk may seem low to some since hedge funds are supposed to be all alpha, this number actually looks high to us.

Hedge Funds

Tracking error is HFRI minus Greenline’s beta replication portfolio. Alpha is assumed to have 0 correlation to the beta replication in calculating the alpha risk share.

Alpha risk on its own does not equate to outperformance as we can see; rather, it is often only the tracking error versus a selected passive benchmark. This serves as a reminder that beating markets is very difficult and few will achieve it over the long run. Those who over-diversify while paying high fees certainly are not likely to earn alpha and will, in turn, lag passive benchmarks by approximately their total fee. But this does not mean hedge funds should be completely disregarded. We separately discuss their beta and potential for alpha, and how we would go about replicating each piece more cost effectively. First a review of the characteristics of betas and alphas.

There are two primary reasons to select an investment other than a passive index fund:

  1. Higher risk-adjusted returns
  2. Diversification potential

Market risk, beta, like buying a passive equity index fund, should offer moderate risk-adjusted returns to compensate investors for taking risk, but no higher given the ease with which all investors can earn this return. If everyone could easily earn high returns, then we would all be rich. Most betas are also linked to similar risk factors, namely shifts in economic growth and inflation. With common risk factors, there is limited ability for diversification among betas. This logic should apply to alternative betas, as we will show.

A single investment with both desirable characteristics, high return and diversification, can only be driven by skill, which we call true alpha. True alpha is determined by the skill, insight and disciplined process of a manager. By definition, skill is rare because markets are a zero sum game (and negative sum after fees). True alpha can have much higher risk-adjusted returns than beta because of its rarity and also more diversification potential if it is driven by unique manager skill. Below we analyze how to separately replicate the alphas and betas embedded in hedge fund.

Hedge fund alpha replication: A formulaic approach

We have already explained why alpha is hard to earn and especially over long periods of time. Therefore, the word “replication,” implying “easy” or “low cost,” should not be consistent with the definition of “alpha.” Take this as fair warning as to the difficulty in outperforming by simply copying others. This being said, we should be able to create a logical framework for what strategies and managers we can copy and expect to increase the odds of earning higher risk-adjusted returns.

The expectation with replicating a manager is simple: if they are good and we can copy them accurately, we should earn even higher returns than a direct investment with the manager by avoiding their high fees. In order to replicate a manager, we need transparency into their positions delivered in a timely manner. If we cannot get full transparency into a manager’s holdings in real time, this does not mean we cannot closely replicate their performance. As long as one can replicate a large enough portion of their holdings so as to overcome the fee hurdle, then replication should be superior to investing directly with the manager.

Through a thought experiment, we develop a logical framework that managers should be able to replicate successfully. We think there should be a trade-off between:

  • Fraction of portfolio that is transparent and fees paid as percentage of expected return

For a manager earning 10% per year and charging a 2% management fee plus 20% on performance, their net of fees expected return is 6.4% (10% – 2% management fee – 20% x 8% of the remaining return = 6.4%). In this case, the investor keeps 64% of the gross return, which will vary with fee structure and expected return.

If we cannot replicate a manager’s entire portfolio, then how much is sufficient to at least match their net of fee alpha? If we assume all positions equally contribute to return, then we should only need to replicate enough of their portfolio to outperform their net of fee returns. If net of fees return is 64% of their gross return as in our example above, then we should have to replicate at least 64% of the portfolio. In practice, all positions are not equally weighted nor contribute equally to portfolio return. Furthermore a manager’s highest conviction ideas get the largest weights and are also expected to earn proportionally more than the smaller positions. Mathematically, using the Herfindahl index, we can show that a 50 stock portfolio with half its weight in the largest 10 names is similar in diversification to a portfolio of 25 equally weighted stocks. Therefore we should be able to replicate less than 64% of the portfolio by at least a factor of 2 if most of our replication is of their highest conviction positions in this 50-stock portfolio example.

Mathematically, we can say that, over time, replication should earn more than investing with the manager if:

% of portfolio not replicated < % of return paid in fees

Furthermore, we can divide the ‘% of portfolio not replicated’ by 2 for many managers given the likely concentration in their top ideas.

  • Time lag of position reporting divided by average holding period and fees paid as a percentage of expected return

Since we cannot get immediately transparency into positions when a manager places a trade, how much lag in reporting positions still allows for successful replication? Said another way, with a given time lag when position data is reported relative to when a manager puts on their position, what is the maximum turnover a manager can have to allow for successful replication? To understand this, we show a thought experiment illustrating the trading of the perfect manager to arrive at an upper bound for turnover.

The chart below shows what happens to the returns of a portfolio before, during and after the perfect manager holds it. The perfect manager initiates a position just as it begins making money and exits at the top. To further simplify, let’s assume the portfolio makes its expected return at a linear rate over its holding period. This way, we can estimate the lost return from entering the same positions at a later time than the perfect manager.

Hedge Funds

In this stylized example, we can see how much return is lost by implementing the manager’s portfolio with a time lag. Assuming, again, the standard 2 & 20 fee for a manager earning 10% per year, as long as the time lag between the manager initiating a position and it being reported (and copied) is less than 36% (% of return paid in fees in the example above) of the manager’s average holding period, then replication should outperform the net of fee returns. Said another way, if this manager had an average holding period of 12 months, as long as we can implement this portfolio less than 131 days of the manager doing so, then replication should outperform. In practice, we get transparency more frequently. With equity managers, they must report their positioning to the SEC within 45 days of each quarter-end. Managers of other strategies often give sufficient transparency into their top holdings in monthly reports and quarterly letters to also facilitate replication.

Mathematically, with lag, we can say that replication should be successful when:

Reporting lag / Holding Period < % of return paid in fees

Similarly as with the fraction of the portfolio we replicate, we can divide by a factor of 2 the ‘Reporting lag / Holding Period’ to remove some conservatism because markets do not behave linearly nor do managers consistently catch the tops and bottoms of trades. We can now combine the fraction of portfolio factor and reporting lag factor to estimate whether replication for a given manager should be successful or not with the following formula:

[ % of portfolio not replicated + Reporting lag / Holding Period ] / 2 < % of return paid in fees

Next, we look at a few examples of actual managers we have replicated and how our formula applies in each case. We do not disclose the individual manager names, but the studies are based on real managers and their actual returns and holdings information.

PDF | Page 2