Introducing the ExtractAlpha Risk Model

July 20th of this year was the 50th anniversary of NASA's Apollo 11 moon landing. The mission was obviously a technological triumph. But it was also a triumph of risk management! The astronauts and mission control had prepared for a wide variety of adverse events in several different ways:

• Risk control: Extensive preparation for a variety of potential errors and unforeseen conditions, such as overshooting the landing target
• Attribution: If something unusual happened, the crew needed to rely on well-built instrumentation such as the Lunar Module Guidance Computer so they could identify issues and know whether and how to address them
• Tracking: Mission data was collected both for real-time decision making, and so that future missions could be accomplished more effectively

We can draw a parallel to the way many fundamental and, especially, systematic investors rely on risk models in various parts of their processes:

• Controlling portfolios' and signals' exposures to risk factors, for better risk and drawdown management (an example is described below)
• Attributing portfolio, strategy, or stock returns to common versus idiosyncratic components, for both internal analysis and client communications
• Tracking style returns over time, so that funds and their investors understand evolving style trends - as in the following chart which shows common factor returns for the U.S. market over the last 18 months, including the recent uptick in momentum strategies:

Unfortunately, commercial risk models are often cost prohibitive or overly expensive for all but the largest asset managers, and building one internally can be time consuming. ExtractAlpha has been using an internal style-based risk model for our research into alternative data for many years now, and we are now opening up this risk model to our clients as a low-cost alternative to the popular commercial risk models on the market.

As just one example of a risk model's many use cases: in a recent research note, we used the Risk Model to control a growth-exposed factor (the Digital Revenue Signal) to risk factors. We found that the risk control consistently reduced drawdowns across every year and capitalization range. The below table shows dollar-neutral returns, Sharpe ratios, and maximum drawdowns for DRS when unconstrained, and then when risk-controlled for Momentum and Growth factors:

In this case, returns were dampened slightly by the risk controls, but the Sharpe ratios were maintained and the drawdowns mitigated - indicating that, as one would hope, the risk control reduced the risk!

The risk model also has uses in alpha generation. For example, ExtractAlpha's Tactical Model uses a reversal strategy which leverages residual return reversal - that is, the greater likelihood of stock returns to reverse if those returns are stock-specific rather than explained by risk returns. Residual returns are calculated with a simple regression which strips out the effects of the Risk Model exposures from stock returns.

Not every asset manager uses a formal risk management system, but as with the lunar mission, the benefits are clear.

The new ExtractAlpha Risk Model covers all global developed market equities and includes fundamental, technical, and industry factor exposures and factor returns along with supplementary data. Historical data is available for testing. Reach out if you'd like to learn more!