Range of the Reverse Regression Slope

Reported in interviews at The Voleon Group Walleye Capital

Problem statement

You fit a simple linear regression of $y$ on $x$ by ordinary least squares and get slope $\beta_1$, so the fitted line is $y = \beta_1 x + a$.

Now you flip the roles and regress $x$ on $y$, getting slope $\beta_2$ in $x = \beta_2 y + c$.

What is the possible range of the product $\beta_1 \beta_2$? And given $\beta_1$, what can you say about $\beta_2$?

Show hint

How can each least-squares slope be written in terms of sample covariance and variances?

Relevant roles: Data Scientist, Applied Scientist and ML Engineer.

Related free problems