Handbook of Exchange Rates

17.5 Empirical Evidence: Fund Managers on the DB FX Select Platform

Evaluating hedge fund performance is challenging due to the usual biases affecting hedge fund databases.³⁵ In particular, backfill and survivorship bias can be severe. Malkiel and Saha (2005) report that backfill bias averages 7.3% per year and survivorship bias averages 4.4% per year for hedge funds.

To address backfill and survivorship biases, we make use of the same database as used in Pojarliev and Levich (2010), that is, daily return data for currency managers listed on the DB FXSelect trading platform. The FXSelect data is unique relative to other hedge fund databases, as it provides actual return data; this has been made possible because gains and losses are computed by DB based on real trades processed through DB prime brokerage. The return data are audited by an independent third party. In contrast, hedge fund databases simply collect return data submitted by managers and are affected by numerous biases. This makes the FXSelect dataset especially useful to study the currency management industry.

We rely on data from April 6, 2005, until March 26, 2008, as in Pojarliev and Levich (2010) and then extended through June 30, 2010, or slightly more than 5 years overall. During this sample period, 107 currency funds were active at some point on the platform. Of these, only 67 funds were active as of June 30, 2010. We label these as live funds. Another 40 managers joined the platform and exited before the end ...