"One reason I am dubious about what this Institute says is that I don't see how Bush can have won as he did in the actual count if the 45% figure is not accurate."
I think what we are seeing here is an unanalyzed result of hte interaction of Latino immigrants. Cubans were famous for their business acumen before Castro's dictatorship, but they have multiplied that acumen manyfold here in the U.S. Other Latinos look around at what Cuba has done here and it looks pretty good. Why listen to the demagogues who for centuries have said that they were going to improve things down south and have completely and repeatedly failed. Latin American people are not stupid and they are figuring it out. Cuba in the U.S. is the most succesful Latino country.
The answer to that question is the fact that Bush boosted his share of the white vote from 54% to 58%. It was Bush's poor showing with whites (for a winning Republican) that nearly did him in last time, and it was his improvement with them this time that was most responsible for his winning 52% of the popular vote.
Dick Morris says latinos made up 12% of the electorate this time, but the exit polls I've seen puts it at 8%. So lets' just split that and say its 10. Whites made up about 77%. Simple math says a 4 point boost with nearly 80% of the electorate is a bigger overall deal than a ten or eleven point boost with 10% of the electorate.
Now, you are not going to hear this mentioned in the MSM. Apparently the GOP is not supposed to try and boost its share of the white vote, or brag about it when they do as they can with any other ethnic group, and as Democrats can do with all groups including whites.
As to the specific states:
Texas: Bush won over 70% of the white vote. That was the primary reason for his and the GOP's dominance in that state. Now if he did in fact win the Hispanic vote there then that's great, but it wasn't key to him winning there.
Arizona, New Mexico, Florida: I don't have the demographic breakdowns of those states. I'm sure improved performance with Hispanics helped in each state, but again any increase in the white vote in those states would be significantly more important than an eqivalent gain with Hispanics.
Now of course as the nation becomes less white, and more Hispanic (though that itself is dependent upon current demographic trends and public policies continuing, so its not like its some irresistible force of nature), then winning more of the Hispanic vote will become more and more important, but the fact is that Bush's improved performance with whites was a bigger deal this time than his boost among latinos.
And of course we are basing this all on exit polls, and who knows how reliable they really were.
It's possible that the media exit polls are right about him getting 45% of Latinos, but the validity of those polls is highly questionable given their inability to predict the election. I also agree that this poll is suspect, but there's nothing I see that gives me any reason to doubt this poll more than the media polls.