Odd statistics

I was going through Estadao's neat statistics UI on Sao Paulo's governor's election poll. Not very complicated, but has some odd things (which is more likely to mean that somebody entered data incorrectly than anything else):

If you look at the general election numbers by education level and income level, you will see that there is a positive correlation on both factors with voting for Alckmin. You'd expect that, as there is a positive correlation between education level and income level.

But when you go to the 2nd round is when things get odd. There is now still a positive correlation to education, but not on income.

I'm a little away from Brazilian politics (although I'll be in Brazil this year during the 2nd round - but I won't be able to vote as of now, because my voting registration was canceled, because I haven't voted for so long), but one thing that I can relate to is the rejection rate on Paulo Maluf. I'm somehow in shock that the guy is still running for a public office. He has a huge track record of corruption, but apparently he still thinks that he was able to get more things accomplished while giving money and taking money from "his friends", so he deserves to be put in power again. The scary thing is that he is not completely wrong about it, just a little narrow-sighted.