The tracking error predictions of risk models are swayed by recent market conditions. These predictions change significantly depending on the time period of measurement and do not properly capture the absolute level of a portfolio’s active risk.
What’s Inside?
To control risk, institutional investors often require their managers to invest within a tracking error range. A high tracking error denotes that active return is volatile and that the portfolio strategy is thus riskier. As a result, managers usually use an ex ante tracking error estimate produced by an equity risk model. The model calculates an estimate of future tracking error based on how the holdings have moved in the past and the exposure to common factors.
The author’s research suggests that because the models are based on historical data, they produce changing estimates of risk. Thus, portfolio bets relative to the index change in significance depending on recent market movements, and the portfolio’s riskiness relative to a benchmark cannot be effectively monitored.
How Is This Research Useful to Practitioners?
This research could benefit board members, consultants, and risk and compliance managers who need to evaluate investment managers. Problems arise when practitioners use ex ante (predicted) tracking error as an absolute risk measure. For example, an analyst is likely making an incorrect conclusion if he assumes that because the forecasted tracking error has doubled from 2% to 4%, the portfolio has become more active. Such a change is often the result of movement in the model’s view of risk within the market.
This research is significant for investment adviser contracts that require that a portfolio’s ex ante tracking error remain within a specific range. This type of constraint could cause the portfolio to trade or rebalance simply to adjust for a changing estimate based on recent market conditions.
How Did the Author Conduct This Research?
Ex ante tracking error is calculated on a portfolio, rather than a target, with two different risk models for quarter-end periods. Both risk models are global fundamental risk models with such common factors as valuation, momentum, growth, and so forth.
The author creates the portfolio and target portfolios using stocks with a market capitalization of $2 billion or more. Each stock is alternatively assigned to the portfolio or the target, with the resulting portfolios containing 449 securities each. No changes to securities are made from period to period. The author graphically compares quarterly predicted tracking error versus time for each model from 2007 to 2013, along with tracking error for common factors and specific factors. The graphic display of changes of predicted tracking error over time is dramatic, ranging from a low of around 2% to a high of more than 5%. Although the portfolio remains the same, the increased tracking error might cause a practitioner to assume that the portfolio had taken on a lot more bets than the target.
Abstractor’s Viewpoint
This well-written, short article is ideal for an analyst or consultant who may need a refresher on the use of tracking error. The author identifies the problems that occur when professionals do not appreciate that estimates of risk can change based on recent market experience. These types of risk models should not be used on their own but should be included with other tools to understand portfolio structure.