Researchers find a way to spot clusters of stocks that move together safely

A team developed a method to find groups of stocks with stable, predictable relationships using correlation networks that filter out weak statistical noise. This means portfolio managers can now identify which stocks actually hedge each other in real crises instead of relying on assumption.

For decades, stock clustering has been guesswork — pick a correlation threshold, hope it holds. This paper shows a measurable way to find clusters that survive market stress because they're built on genuine structural balance, not arbitrary cutoffs. In the Chinese stock market from 2013 to 2024, these balanced clusters visibly contracted during calm periods and expanded during crashes like 2015, which means they're actually tracking something real about how markets reorganize under pressure rather than academic noise.