Relative value (RV) investing has grown and evolved greatly over the past 20 years to become an extremely competitive endeavor. Yet, by pursuing a less common approach, we think it remains a compelling source of alpha.
In the mid-1990s, Long-Term Capital Management (LTCM) famously recruited Ph.D.s to create quantitative models that could exploit price inefficiencies in global fixed income markets (and produced some eye-popping returns until its demise via the Russian financial crisis). However, the subsequent flood of academics into Wall Street, hired by banks and hedge funds to mimic the LTCM model, quickly reduced the edge, or alpha, once available in relative value trades; technology, the proliferation of statisticians in high finance and automated trading have all served to further compress alpha. In short, there is not much low-hanging fruit today.
The most common approach to RV trading relies on the expectation that assets trading at levels dislocated from their historical relationship will mean-revert quickly – inefficiencies that are observable through robust time-series analyses and often captured by “big data.” However, we focus on a second approach: taking advantage of structural dislocations that are likely to persist and not converge to historical means. Rather, the edge is captured in terminal payout space and realized over time as the trades expire.
Identifying relative value opportunities
If two assets exhibit the same profile across all payout spaces, they should trade at roughly the same level. In other words, there should never be a significant valuation discrepancy between an asset and its replicating portfolio. We aim to find assets that contradict this thesis – we look for inefficiencies generated by buyers or sellers who act rationally but with an objective other than maximizing expected value. Often these participants are motivated by regulation or by limited access to the full scope of financial products. We like ideas that are not well hidden or feel like a secret: We are seldom as clever as we think we are, and if we did uncover value that the market overlooked, it would not stay so for long. We are much more assured in taking positions in well-known market imbalances and being thoughtful in implementation and position management.
In rare cases, we are able to find mispriced instruments that allow us to capture near-arbitrage opportunities by buying or selling the assets and hedging with a fairly valued near-replicating portfolio. More often, we are not presented with such closed-form RV opportunities. However, we take the same approach and look for structural forces that push asset prices out of fair value, and we try to harvest that deviation by pairing opportunities within or across asset classes with highly correlated terminal payouts.
One example that can illustrate our approach is the persistent risk premium of long-dated equity index options compared to long-dated interest rate options. Long-dated options on the S&P 500 typically trade at a considerable premium both to short-dated options and to fair value based on long-term realized volatility. This is because market participants with beta exposure to risk assets buy long-dated equity options for portfolio hedging and are often insensitive to the valuation. For example, implied volatility of three-year options has averaged 15% above realized volatility over the past 20 years, while implied volatility of one-year options has averaged only 6% above realized volatility.
In contrast to the asymmetric demand that drives up long-term equity option valuations, long-dated interest rate options are typically depressed by price-insensitive supply, often in the form of structured notes, which are unsecured debt instruments popular with retail investors. In one popular structure, the note holder gives the issuer the option to buy back the note at par periodically prior to maturity. In exchange, the investor receives an above-market coupon, sourced from the option premium synthetically embedded in the callable bond. Investors in structured notes don’t have direct access to over-the-counter (OTC) interest rate options to replicate this exposure and will often purchase the debt as long as it trades with a coupon higher than its non-callable counterpart, providing an ongoing supply of optionality. The implied volatility of three-year expiry options on the 10-year rate has been, on average, 1.5% lower than implied volatility of one-year expiry options.
The supply/demand dynamics driving valuations of long-term equity and interest rate options are unlikely to change, but we are not looking for prices to normalize. In RV trading focused on terminal payout space, our approach lies in capturing realized gains over time.
Evaluating a potential opportunity
Having identified two similar assets (equity volatility versus interest rate volatility) that exhibit opposing valuation deviations, we must decide whether an RV opportunity exists. True relative value trades should be non-directional, with exposure concentrated in normalization of asset valuations either in price or payout space. Thus, it is important in a two-legged trade that the assets be highly correlated. This ensures that the trade is orthogonal to market beta and has equal expectation of profitability in any environment.
To evaluate the potential in pairing long-dated equity volatility with interest rate volatility, we first examine the historical correlation of realized volatility of the S&P 500 and the 10-year swap rate (see Figure 1).
Remarkably, the correlation of three-year realized S&P 500 volatility and three-year realized volatility on the 10-year swap rate is quite high – roughly 0.9 over the past 20 years. (For context, other correlations include: JP Morgan versus Morgan Stanley at 0.18; Wal-Mart versus Target at 0.72; the Nasdaq Composite Index versus the Russell 2000 Index at 0.87.) Equally important, the correlation has been stable through that time series. It seems there is an attractive trade for the patient investor willing to hold the long-dated options to maturity.
Next, we back test our pair trade to see what our hypothetical profit and loss would have been. Here, it is important to be objective about the evaluation process and avoid the frequent mistake of data mining. Calibrating weightings to maximize historical performance and treating incomparable units of risk as if they were interchangeable are common errors that produce mis-weightings and introduce undesirable market beta into the trade.
In a simple two-legged trade, we believe the best approach is often to select weightings based on the realized volatility of the time series. If the realized volatility is not stable over long periods, the structure is less compelling as an RV trade. The necessary rebalancing would increase transaction costs, and because it is difficult to forecast the future ratio of volatilities, the back-testing results would carry less weight. In our example, the three-year realized volatility of the 10-year swap rate has been consistently about 3.5 times as volatile as three-year realized volatility on the S&P 500. Thus, in our back test we evaluate the historical profit and loss of a trade with 3.5x the exposure to S&P 500 realized volatility to that of realized volatility on the 10-year rate (see Figure 2).
The back test reveals what we were hoping to find – the two legs, properly beta-weighted, form a compelling relative value trade that appears relatively unexposed to the overall level of market volatility. Importantly, based on the back test, the pair trade would have performed well during bouts of extreme volatility (the global financial crisis, the eurozone crisis) as well as periods of calm.
Implementing the relative value trade
Once we have identified what we think is a capturable dislocation, it’s important to select the instruments that maximize exposure to this risk. The simplest choice would be to buy an at-the-money long-dated straddle on the 10-year rate and sell a matched-expiry straddle on the S&P 500. However, in doing this we would only capture the discrepancy in realized volatility if we delta-hedge and if the forwards remain near the strike prices.
A better alternative, in our opinion, would be to trade a series of options across strikes (strangles) in each asset class. This minimizes slippage but still requires active delta-hedging and dynamic management of the options on the 10-year rate. To gain the exposure we truly desire, we want to use volatility or variance swaps, which are OTC contracts that pay out the realized volatility (or variance) of an asset in exchange for a fixed rate agreed at inception.
Sizing is also an important consideration. Too often investors focus on target profit rather than the capital they are willing to risk when sizing a trade. Because we are focused on capturing value via the discrepancy in realized volatility ratios over time, we need to ensure we are able to withstand mark-to-market volatility as well.
The easiest way to lose money on an RV trade involving dislocated valuations is to be forced to close out of the position when valuations become even more dislocated. Running stress tests on both realized volatility (focused on terminal payout) and implied volatility (which will drive mark-to-market valuations) would be prudent. In addition, we consider the risks in the trade in the context of the entire portfolio; a trade that adds to existing portfolio exposure should be sized accordingly. Ideally, relative value positions trade orthogonally not only to market betas but to one another as well.
Finally, it’s important to note there is no empirical relationship between the realized volatility of the S&P 500 and that of the 10-year rate. While the historical correlation between the two has been stable throughout many regimes and both should be similarly exposed to the macro landscape, it is certainly possible for the relationship to deviate going forward. The most compelling relative value trades exhibit price discrepancies across assets with empirical relationships in payout space, but these opportunities are rarely scalable. Scanning for assets with strong, stable historical correlations and with an intuitive rationale for their connection is a logical extension.
Our active approach in relative value
Investors have access to low-fee exchange-traded funds (ETFs) and other passive instruments to express RV views on emerging market versus G7 sovereign debt, EU banks versus U.S. banks, and just about any combination of pair-wise positions in different markets. However, those types of trades are highly dependent on making correct calls on macro outcomes, the global risk environment, central bank policies, and so forth; they are also subject to directional noise and exogenous factors that can reduce Sharpe ratios. If the underlying assets in a pair trade do not have a correlation close to one, investors are unintentionally exposed to factor risks in addition to the “relative value” risk. Even with correct identification and structuring of an RV trade, the cleanest expression of an RV opportunity – one that maximizes the relative value capture and minimizes variance from other factors – often requires the flexibility of OTC instruments.
We believe that many high quality, non-directional RV opportunities with return profiles that are orthogonal to traditional portfolio exposures are only available through active management. While traditional regression- and statistical arbitrage-based RV opportunities have become increasingly competitive, our emphasis on finding and exploiting structural price dislocations that stem from explainable, natural imbalances between buyers and sellers should enable us to continue uncovering new RV ideas.
Furthermore, passive management is for the most part limited to analyzing securities in one asset class, missing out on cross-asset RV opportunities. Even with the advent of “big data” and ever more sophisticated statistical computing packages, machines are only as good as their creators.
Because of the complexity involved in identifying, evaluating and implementing relative value trades, we think success ultimately comes down to the experience and acumen of the manager. The recent increases in bank capital requirements, the Dodd-Frank Act and other regulations that followed the global financial crisis have also discouraged many financial intermediaries from warehousing positive-expected-value but difficult-to-offset risk. This unintended consequence of financial regulation should continue to result in supply/demand-driven inefficiencies.
One of the factors that differentiates PIMCO from other investors is our long-term approach. In relative value investing, we can aim to extract value from long-term market inefficiencies that others may not be willing or able to pursue, and in this way, we can offer our clients the potential for superior alpha generation.