“How did you resolve which variables to incorporate in your mannequin, and which did you intentionally exclude?”
The worth of the query lies in what it reveals. You aren’t asking for a listing of variables. You’re asking whether or not the inclusion and exclusion selections have been grounded in financial reasoning reasonably than statistical match alone.
In my conversations with each allocators and managers, the responses fall into three distinct classes.
A robust reply: The supervisor explains the financial mechanism behind every variable’s inclusion. Crucially, they focus on variables they excluded and why, displaying that specification was a deliberate design alternative. They distinguish between variables that drive their goal issue and variables that end result from it. The strongest managers hint a sequence of financial causality: how macro forces undertaking onto stock-level indicators, and why the mannequin displays these causal chains reasonably than mining for correlations.
An ordinary reply: The supervisor cites statistical standards: info ratio, R-squared enchancment, significance assessments. That is present trade apply. It’s not flawed, however it’s incomplete. Statistical match alone can’t distinguish between a variable that belongs within the mannequin and one which introduces distortion whereas enhancing match metrics. That is precisely the lure within the opening story.
A regarding reply takes one among two varieties: “We use all accessible variables and let the mannequin choose” indicators structural vulnerability to issue mirages. Then again, “Our variable choice course of is proprietary” could replicate authentic IP safety. However a supervisor who can’t clarify the reasoning behind their specification, even with out disclosing particular variables, can’t show that the reasoning exists.
