Moreover, the Census figures do not translate well into voter breakdowns, because the Census also counts non-citizens who can't vote. So, the drop-off in the Latino proportion of the vote is exponential to the drop-off in the Latino percentage of the population.
Whatever the case, here's how the top 20 Latino counties in the U.S. voted in the 1996 presidential election:
1 Starr County TX: 86.94% Clinton; 10.41% Dole
2 Maverick County TX: 80.58% Clinton; 15.94% Dole
3 Webb County TX: 76.67% Clinton; 19.02% Dole
4 Brooks County TX: 84.43% Clinton; 11.84% Dole
5 Zavala County TX: 82.32% Clinton; 14.50% Dole
6 Jim Hogg County TX: 79.30% Clinton; 16.94% Dole
7 Hidalgo County TX: 66.49% Clinton; 28.84% Dole
8 Duval County TX: 84.94% Clinton; 11.65% Dole
9 Zapata County TX: 72.90% Clinton; 21.27% Dole
10 Willacy County TX: 70.02% Clinton; 24.62% Dole
11 Cameron County TX: 61.76% Clinton; 32.63% Dole
12 Presidio County TX: 70.22% Clinton; 22.32% Dole
13 Dimmit County TX: 74.96% Clinton; 20.19% Dole
14 Santa Cruz County AZ: 64.17% Clinton; 27.62% Dole
15 El Paso County TX: 62.33% Clinton; 32.11% Dole
16 Mora County NM: 68.76% Clinton; 23.43% Dole
17 Guadalupe County NM: 69.59% Clinton; 25.12% Dole
18 La Salle County TX: 69.69% Clinton; 26.10% Dole
19 Val Verde County TX: 53.05% Clinton; 41.10% Dole
20 San Miguel County NM: 72.51% Clinton; 20.09% Dole
You'll likely dispute my use of 1996 instead of 2000 or 2004, but most of these are in Texas and GWB had a strong home-state effect. If it makes you feel better to go with 2004 though be my guest..
Cameron was more than 84% Latino in the 2000 census. Also, Latinos who don't live in Hispanic ghettos, tend I suspect to be more GOP.
I think the drop off is straight line, but i could be wrong. Having said that, Bush ran very well among rural Latinos in Texas.