1992: States where Perot cost Bush the electoral votes

[Nym90]

There's no evidence that Perot's voters would have disproportionately backed Bush over Clinton.

[Nym90]

I ran a regression analysis a few months ago and there was actually a negative correlation between Perot's vote percentage and the Democratic "trend" (in Atlas terminology) in a state, while if Perot was taking votes disproportionately from Bush, you'd expect a positive correlation. Someone else did mention they had seen studies that there was a positive correlation at the county level, however.

So I would say the above, combined with exit polls showing that Perot's voters would've split evenly, makes it impossible to say with any certainty that Perot affected the outcome one way or the other.

A more persuasive argument might be that Perot tended to attack Bush more than Clinton during the campaign and thus drove up Bush's negatives higher than they would have been if he hadn't run, though I'd be interested in seeing data on that as opposed to just anecdotes.

[Nym90]

There's no evidence that Perot's voters would have disproportionately backed Bush over Clinton.

That's right, but it is generally believed that in some states Perot took more votes from Bush, and in others he took more from Clinton, right? As the OP said,

Clinton still would have won the election without Perot, but it would have been closer.

Ok, but I'm disputing it would have been closer overall. Sure, some states he took from one more than another.