top of page
  • Julean Albidone

Why Macro Factors Don't Predict Equities, and How You Can Change That

If you go to any conventional discretionary or quantitative equities shop, you will find most analysts don't incorporate macro-economic variables into their long / short stock picking process. Talk to any quantitative analyst and they will loathe macro-economic factors. They will caution you against using them, saying macro factors are not for the stock picking.

At first glance, those analysts are correct. If you test any common macro factor - be it credit spreads, currency movements, overall stock market returns, etc. - you will find they have very little predictive power in forecasting relative stock returns.

But let's pause for a minute and think; does macro factors truly add no value to equities? What about market regimes, or business cycles? Wouldn't that be an important indicator of whether a stock is going to beat it's benchmark?

To properly answer this question, I first need to break down the way most equity models are constructed.

Most models are created by ranking various factors (e.g. growth, momentum, dividend yield, value, etc.), and going long the stocks that have the most positive collection of factors, and shorting the stocks with the poorest collection of factors. These factors are tested independently for statistical significance, and added to the collection of factors in the model once they pass various robustness tests.

Let's illustrate this with a simple example; if you know nothing about a stock other than it is highly valuable, then on balance you would expect that stock to go up. Here's another example; if you know nothing about a stock other than it has a great dividend yield, then on balance you would expect the stock to beat the benchmark. This simple logic is how most models are constructed, simply stitching together a bunch of individual factors.

Let me take this same premise, but flip the script slightly. You know nothing about a stock, other than credit spreads in the market are very high - is the stock going to beat the benchmark or not? Here's another example; you know nothing about a stock, other than oil is at an all time low - is the stock going to beat the benchmark or not?

The answer to both of those questions is simple; you have no idea because I haven't told you anything about the stock in question.

Therein lies the crux of the problem. It's only when macro factors are coupled with stock specific factors that they become useful. Macro-economic factors act as a conditional on stock specific factors.

Let's internalize this with some additional examples. If we were looking at a highly levered stock, AND we're at the top of the cycle, then you could guess that this stock would underperform the benchmark going into a sell-off. If we were looking at a high value stock AND we are at the bottom of the cycle, then you could guess this stock would outperform as the market recovers.

So how do we actually incorporate this into a model? Well there are two ways.

The first way - which is an adaptation of the ranked long / short model above - is to simply create a separate factor that combines a macro factor and stock specific factor into one. For example, multiplying credit spreads by the value of the company can be a rudimentary way to create a new credit spread dependent value factor that you can test.

Another way to do it is to use a non-linear machine learning model which captures these interactions between variables. For example, you could include a credit spread factor and a value factor into the same model, and let the ML model decide the exact interaction between these factors.

In conclusion, while macro-economic factors on their own may seem irrelevant in stock picking, their true value emerges via their interactions with stock-specific factors. By acknowledging and integrating these conditional relationships, you can be better positioned to capture these business cycles and market dynamics.


bottom of page