Although this correlation [does exist at the individual level], you cannot infer it from the Flowing Data map, due to the [Ecological Fallacy].
What the map shows is that counties with higher poverty rates have higher rates of smoking. You cannot then infer that individuals who are poor are generally more likely to smoke. Here's why: suppose that most poor people live in a small number of counties (concentrated urban areas), and have low rates of smoking. Then overall, it is possible that poverty and smoking are inversely related at the individual level, even as they are positively correlated across (unweighted percentage rates) counties.
Poor states are more likely to vote Republican. But poor people are more likely to vote Democratic. These are mutually consistent, because the position and slope of the income-political affiliation line varies by state -- see Andrew Gelman's work on this.