Ebook Hedge fund contagion, liquidity spirals and flight to quality
The hedge fund industry has grown dramatically over the past two decades with estimated assets under management of $33 billion in 1990 and nearly $2 trillion in 2008. Part of the reason for this growth is that clients were promised ’absolute’(or positive) returns. Further, hedge funds supposedly offer portfolio diversification to investors because their investment style allows their returns to be uncorrelated with both equity market returns and with returns of hedge funds with different investment styles. However, in 2008 on average hedge funds lost 18% and the industry shrunk by a quarter reflecting both losses and client redemptions.
It is well known that correlations among equity markets are larger during large downward moves than during upward moves in equity prices. This phenomenon is generally referred to as contagion (Bekaert et al. (2005)). Can the same be said for hedge funds with different styles? Low correlation among different types of hedge funds motivate funds of hedge funds. The popularity of funds-of-funds is based on the notion that they offer the benefits of portfolio diversification by investing in hedge funds with different styles. The benefits of diversification are most important to investors during bad states. However, several studies including Boyson et al. (2008) have documented that hedge fund indices with different styles suffer from contagion. For instance in October 2008 the fifteen hedge fund style indices that are followed by CISDM all had negative returns ranging from -0.2% to -10%. More importantly the equity market neutral style index lost 2%!
In this study we examine the economic determinants of contagion as defined by asymmetric correlation among hedge funds, where the correlation conditional on negative returns of different hedge fund styles is much higher than the correlation conditional on positive returns. Our contribution is twofold. We develop a novel estimation procedure that captures co-dependency in the left tails of two return distributions. We also investigate the causal channels of contagion and show that the returns on highly levered portfolios also vary with funding liquidity, which we measure with changes in futures margins. Our findings are potentially important for empirical and theoretical asset pricing and for portfolio allocations. An example where the asymmetric correlation could be important is where agents have loss aversion preferences Barberis et al. (2001). These agents would treat gains and losses differently and are very averse to downside risk.
The first part of our analysis investigates whether the correlation of returns among hedge funds with different styles is higher conditional on low returns. We measure correlations conditional on being above or below a given quantile. We find that these exceedance correlations are much higher in the left tail of the distribution of hedge fund returns than in the right tail. Correlations are the most asymmetric for the distressed fund index, the merger arbitrage fund index, the relative value fund and the convertible debt fund. The neutral and macro index fund have the most symmetric correlations. We document that high correlations in the left tail occur during periods of high market volatility, as measured by the VIX (implied volatility on the S&P500 index options), during periods of low consumer confidence, and during periods of both high market and funding illiquidity.
We further explore the determinants of correlation between hedge fund returns by estimating the dependence structure between an index of hedge fund returns and the average return of the 7 other indices of hedge fund returns under consideration. Our dependence function (based on the bivariate normal distribution) is similar in nature to that employed in Longin and Solnik (1995) for international equity markets and Patton (2006) for currency markets. This type of dependence structure has the advantage that it yields a simple expression for correlation between two return series. It also allows time-varying correlation. We model the time-varying correlation as a function of the following exogenous variables: implied option volatility (measured by VIX), funding liquidity (measured by the margins on S&P500 futures contracts and USD/JY currency futures), market liquidity (Amihud measure), the Boyson et al. (2008) prime broker index, the TED spread (Libor minus T-Bill rate), corporate bond spreads and the Michigan consumer sentiment index.
We empirically examine the relative importance of three economic channels that may explain contagion or co-movement among hedge funds indices. The effect of the first two channels remains significant after we control for risk factors known to explain expected hedge fund returns. The first channel involves funding liquidity, which we measure by changes in futures margins for members of the Chicago Mercantile Exchange. Brunnermeier and Pedersen (2009) argue that margins are set according to fundamental value volatility and liquidity-based volatility. They show that under certain conditions, margins are destabilizing and funding liquidity and market liquidity are mutually reinforcing, leading to liquidity spirals. In figure 4 of Brunnermeier (2009), “funding problems force leveraged investors to unwind their positions causing 1) more losses and 2) higher margins and haircuts, which in turn exacerbate the funding problem and so on.”
Our study demonstrates the importance of funding liquidity in explaining financial contagion. The impact of futures margins is both statistically and economically significant after controlling for other measures of funding liquidity. We find that changes in margins required for members of the CME on the S&P500 futures contracts and USD/JY currency futures are a significant determinant of contagion among hedge funds. The return on a portfolio of prime brokers and the TED spread (Libor minus T-Bill) are neither statistically nor economically significant in our specifications. The importance of the futures contract margin suggests that the liquidity spirals of the type described in Brunnermeier and Pedersen (2009) are a potential channel for hedge fund contagion.
The funding liquidity channel is related to the “cash-in-the-market” pricing described in Allen and Gale (1994). In their model, asset volatility increases when liquidity shocks occur and prices are determined by the amount of cash in the market rather than future returns. Limited participation of the type described in that paper fits in nicely with the liquidity spiral hypothesis. Liquidity shocks lead to deviations from fundamental value, causing some funds to incur losses on existing positions which in turn may increase margins requirements, creating greater deviations from fundamental value as funds unwind some of their positions.
The second channel we study is the collateral channel. In addition to trading based on margins, funds can also purchase assets by borrowing money and using assets as collateral. Fostel and Geanakoplos (2008) argue that disagreement between different types of investors over asset values lead to what they call a liquidity wedge. This wedge is the difference between what optimists are willing to pay to borrow funds and what pessimists are willing to demand to lend funds. The greater the wedge, the lower the present value of cash flows but the greater the value of an asset as collateral. Fostel and Geanakoplos (2008) argue that flight to quality arises when this liquidity wedge is high and the dispersion of collateral values between assets is high. Therefore we expect greater left tail co-dependence among funds that trade in securities with low collateral values, such as in low-rated corporate bonds or in convertible bonds.
As the collateral value of these types of securities deteriorates, speculators must post more collateral or unwind their positions, thereby leading to lower asset values among funds with positions in these types of securities. Our measure of the collateral channel is the monthly change in the spread between BAA rated corporate debt and 10-year Treasuries. This variable is a significant determinant of expected returns for hedge fund indices. Moreover, we find that this variable explains the correlation between hedge fund indices (in particular the distressed fund index), even after controlling for its effect on expected returns.
The third channel is the consumption channel described in Fostel and Geanakoplos (2008). This channel is related with the risk premium required by investors. Fostel and Geanakoplos (2008) suggest that contagion is also caused by a portfolio effect in which bad news about one asset class increases the marginal utility of consumption today, thereby increasing the rate of return required to invest in other assets classes (which depresses all asset prices). Our empirical results suggest that the consumption effect is not an important channel for contagion among hedge fund returns.
Our paper is closely related to Boyson et al. (2008) who also document contagion among hedge fund indices. Their results suggest that contagion is caused by funding liquidity proxied by the returns on an index of stock prices of prime brokers. This paper is different from their study in several ways. First, we model asymmetric dependence between funds in a way that explicitly takes into account the left tail co-dependency of hedge fund returns. Second, we identify three potential channels that explain contagion, and third, we use a different measure to proxy for funding liquidity. We also show that contagion is greater among funds that invest in assets with lower collateral values. These are the assets which are the most vulnerable to flights to quality of the type observed in the fall of 2008.
In a related paper, Billio et al. (2008) examine whether hedge fund exposure to risk factors varies across different volatility regimes. Ang and Bekaert (2002) document that correlations between international equity market returns increase during periods of high volatility and negative returns. Pastor and Stambaugh (2003) show that marketwide liquidity is a significant risk factor. Acharya and Pedersen (2005) show that security returns vary with the covariance of its market liquidity and the market return. In contrast our paper emphasizes how funding liquidity affects contagion among funds that employ leverage, such as hedge funds.
Download
PDF Ebook Hedge fund contagion, liquidity spirals and flight to quality
Posted in :
