There are related studies in Baseball Between the Numbers and The Book, two awesome books that broke a lot of OG sabermetric studies that form the basis of much that's still used today. Unfortunately, the articles from both books aren't available for free online. I own and have read both of these --they're fascinating reads to this day, even though many of the concepts have been improved on since their release.
In short, the basis for the use of walks/strikeouts/homeruns came from studies performed by Voros McCracken that showed those rates as stable predictors of their own future values in later years, whereas ERA (and most other commonly used stats) were not. I don't remember off the top of my head how they accurately scaled FIP to ERA, but the end result was the FIP predicts future ERA better than ERA predicts future ERA.
Now, the important distinction here is that this is only going to hold true if the relevant rates (walks/strikeouts/homeruns) hold steady. So the statement "his performance suggests he will regress" is less accurate than the statement "if he continues to pitch as he has been pitching, his performance suggests he will regress." In other words, if you think that Reynaldo Lopez will improve/refine his command and thus strike more guys out/walk fewer guys/continue to repress homeruns going forward, you may not believe that will regress. But it IS accurate to suggest that he has not shown the characteristics of a sub-3 ERA pitcher so far.