Tracking error is an ideal measure to assess the active risk the manager is taking. Many risk measures (e.g., standard deviation, downside deviation) don't directly take the benchmark into consideration: tracking error does. As Grinold & Kahn put it, "the client bears the benchmark risk, and the manager bears the active risk of deviating from the benchmark." Tracking error measures the active risk: the risk the manager bears.
The math is simple: standard deviation(average portfolio return - average benchmark return):
Goldman Sach's "Green Zone" is an example of a risk management approach that uses tracking error to monitor their managers. Their article (see below for details) received the Dietz Award as a testament to its value. Tracking error ranges (the "green zone") are established for each client and as long as the portfolio stays within the range, everything is fine. But, if it drifts into the "yellow" or "red" zones, then an adjustment is required.
Granted, tracking error does not give an indication of the potential loss or inability to meet objectives (two commonly used definitions for risk), but it's an effective tool in monitoring the risk being taken by the manager. It should be included in your arsenal of risk measures.
Litterman, Robert B., Jacques Longerstaey, Jacob D. Rosengarten, Kurt Winkelmann, & Paul R. Laubscher. "The Green Zone...Assessing the Quality of Returns." The Journal of Performance Measurement. Spring 2001.
.png)
Folks need to be careful with how they use/interpret Tracking Error (TE). The manager can deviate wildly from the index and so long as the deviation is consistently "wild," the TE will be very low. TE measures the deviation of the spreads between the index and the fund of each observation; hence, if all the deviations are the same magnitude, very low TE.
ReplyDeleteInteresting and excellent point: since tracking error measures volatility, as long as the return difference between the portfolio and benchmark remain rather consistent, you'll see a low tracking error. Of course, pulling off major deviations in makeup from the benchmark and yet to be consistent over time might be a challenge, it's still possible.
ReplyDelete