A 2002 post of mine from soc.history.what-if (with an URL updated):
FWIW, exit polls found that Perot voters were split fairly equally between
Bush and Clinton as their second choice, with a large percentage saying
that they would not have voted at all if Perot were not in the race.
(Indeed, 1992 represented a temporary reversal in the long-term decline in
turnout for presidential elections, and Perot was probably a major reason
for that, attracting some voters, especially young ones, who had never voted
before.)
It has to be remembered that for no-Perot to have enabled Bush to win,
the exit polls had to have been not just wrong but drastically wrong.
Clinton defeated Bush by 43.3 to 36.7 percent of the vote.
http://psephos.adam-carr.net/countries/u/usa/pres/1992.txt To have Bush ahead in the
popular vote the Perot vote would have to go for him by almost 2-1. And if
a substantial number of Perot voters had abstained, Bush would have had to
get *more* than two-thirds of the remaining ones to win. (Yes, I know
that the electoral vote is what counts, but generally one cannot get an
electoral majority unless he at least comes very *close* in the popular
vote, as Harrison did in 1888 and GW Bush did in 2000. And GHW Bush would
have had to win well over 60 percent of the Perot vote to even come
close.) I see no reason to assume Perot voters would be that
overwhelmingly in favor of Bush, not only becuase of the exit polls but
because of the following facts:
(1) Although Perot did well in "conservative" states (e.g., in the
Rockies) he also did well in "liberal" ones like Massachusetts and Rhode
Island. OTOH, he did poorly in the most socially conservative region of
the country, the South; the only southern state where he exceeded 20
percent was his own state of Texas.
(2) Perot was largely running on dissatisfaction with the economy, and a
belief that the parties were neglecting it in favor of social issues.
People who believed this were not particularly likely to vote for Bush,
and people who were satisfied with the economy or who believed it was
imprtant to outlaw abortions were unlikely to vote for Perot (who was
pro-choice).
(3) In 1996, when Perot did much worse, much of his old support in New
England and the Northeast seems to have gone to Clinton; in the South and
the Rockies it went more to Bush. This is consistent with the finding
that the Perot vote was about evenly split in 1992.
(4) Again, if the Perot vote had been overwhelmingly Republican, one would
have expected the GOP to do much better in Congress than it did. Yes, the
Democrats did lose a few seats compared with 1990. My World Almanac indicates
a 267-167 Democratic edge in the House after 1990, compared to 258-176 after
1992. So that's a grand loss of nine seats, still leaving them exactly as
many seats as they had after the 1986 midterm election. And this loss, small
as it was, can easily be explained by reference to three facts: (1) The
House had been reapportioned after the 1990 census, and some of the
Democrats' strongest states in the Northeast had lost seats. (2) The House
bank scandal, though it hurt Congressmen of both parties, probably hurt the
Democrats more because there were more of them involved (simply because there
were more Democrats in the House, period). (3) The Democrats' House margin
had long been artifically inflated by the fact that many Southern districts
that hadn't voted for a Democratic presidential candidate in many years still
continued to re-elect their old conservative Democratic Congressmen who had
first been elected in another era. As these Southerners gradually retired,
the Democrats' margin was bound to erode.
In spite of these facts, the Democrats still won the Congressional vote
easily--which would not have been possible if the Perotistas were as
heavily Republican as some people imagine.
On the whole, I agree with the analyis of the 1992 elections in Michael
Barone's and Grant Ujifusa's *Almanac of American Politics 1994*:
"But when the Perot vote is allocated as exit polls suggested it should
be, split evenly between the two candidates (with perhaps a few more
votes for Bush in the South, the Mountain States and Ohio, but that gets
us quickly into the realm of spurious precision), the result is a 53%-47%
Clinton victory, almost precisely equal to the 52%-46% by which
Democratic House candidates beat Republican candidates in preliminary
figures." p. xxix
This seems to me about right--I think that had there been no Perot it
*might* have cost Clinton the electoral votes of Georgia, Colorado,
Montana (all of which Clinton carried narrowly in 1992, and lost in 1996
when Perot's vote declined) Nevada (which Clinton carried narrowly in both
elections) and maybe even Ohio (which Clinton carried narrowly in 1992 and
by less than his national margin in 1996). But if he lost all those
states he would still have gotten 321 electoral votes--well over the 270
necessary to win.