From Under the Ivory Domes

Mark Tapley of the BNP Paribas Hedge Fund Centre provides a sneak preview on some upcoming research

Mark Tapley
Originally published in the February/March 2006 issue

Greek tragedies, I vaguely recall, sometimesend with a tragic hero destroying something he loves, often unaware of doing so until in the final act the awful reality dawns on him, and remorse sets in. In parallel but less dramatic fashion, recent research by faculty at London Business School suggests that the investing institutions’ appetite for hedge fund product could bring about its own source of disappointment, with the cherished alpha playing the part of the object of their desires. The difficulty faced by individual funds as they take in more capital, and the difficulties of the industry – or segments of it – namely continuing to add value as it expands, is arecurrent theme in a number of working papers posted on the Centre’s website so far this year. In this article I shall focus on two of them.

Do hedge funds add alpha?

The first paper is jointly authored by Robert Kosowski of INSEAD, Narayan Naik of London Business School and Melvyn Teo of Singapore Management University, and its title contains a simple question, namely “Do Hedge Funds Deliver Alpha? A Bayesian and Bootstrap Analysis.” Simple, that is, to ask; but much harder to answer. As the authors explain, it is not just a matter of poor data sets, with biases introduced by selective self-reporting, incubation periods, and back-filling. More so than long-only mutual funds, hedge fund returns will tend to exhibit fat tails, negative or positive skewness, and be subject to structural shifts. These departures from the normal distribution strew statistical hazards in the path of the researcher. Distinguishing luck from skill itself requires some skill in itself.

The conventional analysis is to do ordinary least squares regressions and study the alphas. Ranked over two years, then observed over the succeeding 12 months, the authors confirm what the typical sceptical academic or scornful journalist has always known; there is no persistence in the alphas, or at least no statistically significant persistence. Backing winners on this basis will not provide excess returns.

Kosowski, Naik, and Teo looked at the problem of alpha persistence afresh. They examined not just alpha, but also the t-stat of alpha, a measure not of the alpha’s size, but of how reliably it differs from zero. Secondly they used two much more robust tests of significance that are less sensitive to departures from normality. Bayesian analysis was one tool, and bootstrapping procedures the other. Both have been around for some time, but as statistical techniques they have themselves been refined. The particular refinements used here are quite recent. Bayesian analysis is particularly relevant to hedge fund studies because it is stronger than OLS on short return histories.

Lo and behold! Measured more exactly there is persistence in top-decile performers. The findingchallenges the classical finance theory view that markets are efficient, and that top hedge funds are simply the lucky ones. In a battery of further tests, the authors divided their sample into six broad categories – directional traders, relative value, long/short equity, security selection funds, multi-process, and finally funds of funds. Perhaps even more surprisingly than their basic finding was that all six categories showed alpha persistence that could not be ascribed to chance. Nor was the finding restricted to, or dependent on, only small funds. In fact, larger funds showed greater persistence. The persistence also stood up to extending the ranking periods to three, four, and five years, and evaluation periods from one to two and three years.

But back to the Greek tragedy. In an article full of surprises, at least for their academiccolleagues, though probably not for practitioners, the one caveat concerns funds with high inflows compared with closed or low-inflow funds. And you’ve guessed it – funds with high inflows had less persistent alphas. Money, it seems,is a great leveller.

Convertible arbitrage

The second paper is an updated study of the convertible arbitrage strategy. The number of convertible arbitrage hedge funds exploded between 1995 and 2002, with several new offerings every month for extended periods. They played a particularly significant role in providing liquidity to the primary and secondary markets during the 2000 to 2002 period, when primary issuance in equity markets was miserably low. In spite of being tiny compared with the main equity market in market cap terms, only 2% of its size, the convertible market matched its big brother in terms of new issuance, at about $300 billion. Fast forward to last year, and of course the tables had turned, with the main market in fine fettle, and the convert market languishing. Therein of course lies one of the attractions for institutions: hedge funds help provide stability of returns, not just because they themselves have less downside risk than the orthodox asset classes, but because their correlations with those asset classes are low.

The paper ‘Risk and Return in Convertible Arbitrage: Evidence from the Convertible Bond Market’ is by Vikas Agarwal and Yee Cheng Loon from Georgia State, and Bill Fung and Narayan Naik from London Business School. The data set included more than 3,000 US and Japanese convertibles. As for the CA hedge fund universe, the authors pooled data from three sources, namely the CISDM, CSFB Tremont (CT), and the Hedge Fund Research (HFR) databases.

While the first article I reviewed above discussed one set of data problems with hedge fund research, this second article reveals another. No single source neatly provides index data and constituent return data. Relying on any one source could easily give biased results, depending on the criteria used by the database providers for inclusion, and the reporting preferences of the individual funds themselves.

A major project at London Business School is nearing completion that will go some way to answering these problems, by matching and cross-checking return data from commercial providers, with a view to eliminating errors and data gaps. At the same time, the lack of overlap in fund coverage among the commercial providers means that the resulting dataset is not just much cleaner than the raw material, it is also much larger. That is shown in the Venn diagram below. I should caution now, however, that under the agreements between the Centre and its suppliers, the pooled dataset is available only for use in academic research.

Distribution of Convertible Arbitrage Hedge Funds by Data Sources

So much for the data and the motivation for the research. Looking at the authors’ analysis, the hypothesis was that CA hedge funds face three ‘hedgeable’ risks: the risk of the underlying equity into which the bond is convertible; the credit or default risk of the issuer; and thirdly, interest rate risk. Typically, the convertible arbitrage hedge funds seek to exploit mis-pricing of one or more of the three risk factors, usually though not always by shorting the underlying equity and going long the convertible bond. Here, the authors attempt to isolate three so-called ‘primitive’ trading strategies, and study the returns to each over a decade-long period. The first of these strategies (labelled CARRY by the authors) builds portfolios with positive carry, i.e. where the yield on the bond plus interest on the proceeds of the short sale (delta-adjusted) of the underlying equity exceeds the cost of financing the long position in the bond. The second primitive strategy (CREDITARB) builds portfolios of bonds which have relatively little chance of conversion. Long/short portfolios of bonds are constructed with the equity and (government) bond market risk hedged out. The third ‘primitive’ strategy (VOLARB in the paper) is akin to the second, but instead looks only at bonds with a conversion parity close to 100% (i.e. the bond has a similar value when converted as when not). Furthermore as well as shorting the matching underlying equity, the strategy in this case also shorts the BAA corporate bond market, not the government bond market as in the second strategy.

The results support the hypothesis that these three primitive strategies account for a substantial proportion of the returns of hedge funds in the convertible arbitrage category both in- and out-of-sample. In two extensions to the analysis the authors divide their samples into new issues of less than 90 days since issue, and seasoned issues. The motive is to test whether convertible arbitrage hedge funds appear to benefit from the apparent under-pricing of initial public offerings by issuers. The studyconfirms that the hedge funds do indeed appear to benefit from their activities in the new issue market. In the second extension, the study looked at the CA hedge fund industry pre- and post-LTCM, finding as one might expect, a structural break, with the funds gaining excess returns in the latter period from the expansion of credit spreads that took place then – only of course to see them shrink again by 2005.

These two studies shine very different lights on the hedge fund industry. The first is wide-ranging and provides encouraging evidence of sustained and identifiable managerial skill. It does so across a broad range of hedge fund categories. The second looks in depth at one fairly narrow category of fund, and is less convinced about managerial skill. But the latter study does show how funds can harvest an illiquidity premium in years when other, more familiar forms of systematic risk are not being rewarded. Whether that’s just a dressed-up beta masquerading as alpha or not, as the cynics might argue, matters less. There was a harvest to be gathered in, and we needed some skilled hands to help us do it.

And the Greek tragedy? What happened in 2005 wasn’t simply that the new issuance market dried up. It was also that momentum-chasing investors continued to pile into CA funds, not looking at the rapidly changing fundamentals of the market, but simply extrapolating on past success. The resulting severe supply/demand imbalance caused a six-standard deviation event that many are now regretting. And it is this that is the common theme between these two papers and others authored at the Centre. The weight of money can be crushing. Or, when seeking allocations, be careful what you wish for. You might get it.