*There seems to be a growing level of agreement that 130/30 is different than simply adding together a 100 portfolio (e.g. an ETF) and a 130/30 portfolio (e.g. a market neutral fund). Some practitioners have pointed to the untrimness of being long and short some of the same stocks (e.g. Jacobs & Levy – see related posting). But others such as First Quadrant’s Jia Ye have argued that adding a short-extension will not always be optimal even for the alpha-producing manager due to the potential volatility of the information coefficient (see posting).*

*Today, guest contributor Srikanth Iyer, Senior VP and Senior Portfolio Manager, Global Systematic Strategies at Guardian Capital LP puts these two ideas together by exploring whether a so-called integrated 130/30 portfolio is always optimal.*

**130/30 “Combined” vs. “Integrated”: The Tail Wagging the Dog**

**Special to AllAboutAlpha.com by:** Srikanth Iyer, SVP, Guardian Capital LP

The rapidly evolving landscape of 130/30 has seen many investment concepts used in interchangeable and often inappropriate ways. As more players enter this space, it’s likely that we will see a further dilution of these core concepts. The debate between a combined and integrated approach to active extension strategies is a classic example of how important concepts relating to return and risk are being bypassed to placate existing investment approaches. The demands of business development add further confusion to the discussion about 130/30 strategies.

An integrated 130/30 portfolio is created using a mean-variance optimizer that uses the correlations between individual long and short securities to achieve an optimal mix for a given risk budget or ex ante tracking error. In contrast, a combined 130/30 portfolio combines an existing mean-variance optimized long only portfolio with an integrated 30 long/30 short portfolio – effectively, combining a long only beta adjusted return with a zero-beta market neutral return.

The chart below from a recent report by Credit Suisse illustrates the subtle differences between these two approaches. (left=integrated, right=combined)

The Credit Suisse report found that ex-ante tracking error for an S&P500 integrated 130/30 mandate was only two-thirds of that for the combined 130/30 portfolio (1.69 vs. 2.68). The reason for this is that when we allow the shorts to interact with a 130 long component and not just the 30 long component, the optimality of negative correlations is better utilized.

While one cannot dispute this fact (theoretically or intuitively), one should not ignore the change in the makeup of the portfolio resulting from the integrated approach.

When one allows an optimized 30-long/30 short portfolio to interact with a 100 long portfolio, it will result in a changes in weights for each stock based purely on risk reduction parameters. This could prove counter-productive if the alpha models used to pick stocks are not based on the same factors as the risk models.

The range of portfolio returns (or cross sectional dispersion of a portfolios return) is the sum product of 2 components: 1) the non-systemic dispersion and 2) the systemic dispersion. Focusing on the non-systemic part will address the nuances of risk and return within the 130/30 portfolios that are often ignored.

To re-cap, the range of non-systemic returns for a 130/30 portfolio are a result of the product of:

- The level of dispersion of active weights,
- The cross-sectional return dispersion, and,
- 1 plus the correlation between active weights and realized returns, also known as IC*TC (IC=Information Coefficient TC= Transfer Coefficient)

The information coefficient (IC) of a strategy is defined as the correlation between projected (i.e. model-based) and realized returns, while the transfer coefficient (TC) measures its implementation (the correlation between the stocks’ projected returns and their active weights). (ed: see previous posting on IC and TC)

The TC improves significantly when one removes the long-only constraint since short-selling allows the manager to underweight a stock beyond just not holding the stock. As a result, this can increase the correlation between stocks’ projected returns and their active weights.

Understanding the role of IC and TC in a 130/30 portfolio is critical to understanding risk. However, mean-variance risk models that are used to integrate 130/30 portfolios invariably address components *1 and 2 only*.

Take, for example, 2 out of 50 stocks in a hypothetical portfolio. One has twice the weight of the other in the index, but half the alpha potential based on a specific stock ranking model. A mean-variance optimization would overweight the stock that has the larger index weight purely to minimize its active weight dispersion.

A mean-variance optimization would also overweight stocks that have the lowest co-variance with other stocks. These co-variances are a function of cross-sectional dispersion and are based on regressions of stocks’ returns with factors such as style, currency etc. But these factors might be very different from the ones used in the stock ranking model. This leads to weight changes that could actually reduce the IC despite a reduction in ex-ante tracking error.

The improvements in an integrated ex-ante tracking error found in the Credit Suisse paper cited above is more than offset by the benefit realized from the stability of a zero-beta market neutral component in a combined portfolio. In turn, this makes alpha attribution easier and can better explain hidden risks in 130/30 strategies.

The relative advantages of an integrated 130/30 portfolio also depends on one’s definition of risk. In an environment of low cross sectional return dispersion, the tendency to make more concentrated active bets further aggravates the hidden risks of IC volatility. In this environment, larger bets are needed to achieve the same level of alpha for a given level of IC. (i.e. We eventually come to a point that what we do not own has a greater impact on the portfolio than what we do own.)

In addition, a large degree of IC volatility comes from the short side since security returns are not symmetrical. Basically, the likelihood of positive returns exceeds that of large negative returns. As a result, the long portfolio will always have a better information ratio than the short portfolio and a mean-variance optimizer will almost always have active risk concentration in the long portfolio.

Which approach, the integrated or the combined, is superior? It depends on your point of view. Risk management suggests an integrated approach is better. But this has serious ex-post risks wherein ex-post Tracking Error will be higher than ex-ante tracking error in a 130/30 strategy. While not optimal from a risk management perspective, the combined approach alleviates these problems, though greater clarity at the zero beta end.

Portfolio managers should be aware of the pros and cons of both approaches rather than let the optics and marketability of these concepts dictate their process. In other words, their first priority should be to engineer the stability of the steering wheel system (lower IC volatility) before pressing the accelerator (TC).

*– Srikanth Iyer, May 25, 2008*

*The opinions expressed in this guest posting are those of the author and not necessarily those of AllAboutAlpha.com. *

## 3 Comments

June 18, 2008 at 6:35 am## Sri Iyer