Future Regression

by Ben Perreira

Oftentimes economists, through their training in statistics, use multiple regression to determine correlations between various variables. This has some utility in mature, established fields, but little utility elsewhere. Financial analysts use regression (technically it’s cousins variance and co-variance) to determine beta, or firm’s market risk. Beta has been consistently proven close to useless in predicting a firm’s risk relative to the market, in part because it ignores the investor’s true desires – increase upside risk and decrease downside risk. Beta treats all risk the same. So does regression.

But what if there was a type of regression that allowed us to look forward?

Daniel Kahneman, to whom I’ve referred for his book Thinking, Fast and Slow, has a great technique. I don’t remember Dr. Kahneman assigning a name for it, so I’ll give it the paradoxical name “future regression.”

Future regression works as follows:

Your team is looking to make a move in the next few months – an investment, acquisition, sale, whatever. Gather the top 5-10 decision makers with different backgrounds and tell them to put themselves in the same seat in precisely twelve months. Next, they must imagine that the project has failed and they need to write down all the possible reasons why. This technique predicts that you will see a wide variety of possibilities that remain latent when the team has already committed to move forward, offering a more accurate glimpse into the future.

Why is this superior to multiple regression? It may not always be, but it can be quite difficult to predict the next bends of a country road by staring into the rear-view mirror. Besides, who better to predict the future than those who know the area best?