The world is new - here is how, but not why
Equity markets are complex. That makes it difficult to draw simple conclusions.
If we use math to reduce the complexity, we can check if it “is different this time”.
It IS different this time
In DNB Systematic Active Equity Team, we use regression analysis with more than 60 variables to find attractive stocks. Some insights from that model are easily relatable. Since the model is designed to find what stock characteristics drive prices in equity markets NOW, you will see results that correlate well with the general discussion in media.
This allows us to rank stocks according to the current market environment. It is a dynamic approach that surprisingly few people try to follow. Most quantitative models rely on persistent risk premiums rather than dynamic premiums. That to side, our model is ideal for checking if “something is different”.
The 60 factors are a problem. 60 factor payoffs are too complex to evaluate, but this is where Principal Component Analysis (PCA) comes to the rescue.
The set of 60 factor payoffs tells us what ‘view’, what ‘preferences’ that influence market pricing in the period we focus on. We have this analysis since 1973 for the US. PCA analysis can take these data series and tell us what ‘set of preferences’ are most common in this period.
We do PCA analysis on data since 1973 to find out what two general sets of preferences dominate the pricing the most. By design, these ‘views’ will be the most opposing views you can find. They are orthogonal.
It is like reducing the very complex market pricing into two camps. Think of it as one camp of investors that favor this view and another camp that favors that view. –Like the value/growth battle, only much more complex. Remember that we combine more than 60 factor payoffs here.
With the PCA analysis done, we can plot how the battle goes. We plot what camp has the most influence on stock prices. The following graph places the general market performance of every month since 1973 in these two dimensions.
The obvious takeaway here is that there are two very distinct clusters. The two left hand dots mark two market states that seem to occur very often, while some observations are ‘far out’. The observation at the top left is the peak of the IT bubble. The observations out to the right are all observations between 2000 and 2001. All equity managers remember those two years, and these dots are far from the mean for good reasons.
