We have found that stocks with a high degree of valuation uncertainty are more sensitive to investor sentiment and quality metrics, while stocks with a low degree of valuation uncertainty are more sensitive to valuation ratios and discounted cash-flow models. Our dynamic factor-weighting model assigns different factor weights to different groups of stocks, depending on their profile. The model also takes into account the evolution of a stock’s attributes over time. For example, as a stock moves from early growth to maturity or through market environments (from low to high uncertainty), we adjust factor weights accordingly. We believe this approach diversifies our factor exposure, improves our alpha models, and has the potential to reduce drawdown risk.

Our model assigns risk attributes to each stock and predicts how combinations of stocks may contribute to overall portfolio risk. But our model also goes several steps further, seeking:

• Improved risk-prediction accuracy through the use of forward-looking risk indicators
• Faster response to changes in risk regimes
• Integration with our alpha models, allowing for efficient implementation of stock ideas
• Crowd avoidance through differentiated holdings
• Application of geographic risk factors and dynamics

We aim to distinguish intentional risks (arising from our alpha models) from unintentional risks (arising from style exposure or residual risk).

We integrate long-term return forecasts with our proprietary risk models, while also assessing the likelihood of short-term price movements. This multi-horizon alpha-modeling process seeks to ensure that each portfolio reflects an attractive combination of return potential, risk, and cost. Complete transparency in our research and portfolio management infrastructure enables this level of detail.