The number of Hispanics eligible to vote in the selected, displayed swing states is greater than the margin of victory in each of those states was in 2008. Consequently, the Hispanic vote will determine who wins in all of them! Of course, the WSJ could have just as easily done the same using the number of white voters instead of Hispanic voters with the result being that instead of readers seeing that there are twice as many Hispanics in Colorado as there were more voters who voted for Obama than for McCain, they'd see that there is something like ten times as many white voters as there were more Obama voters than McCain voters in the same election. That would give the impression that the white vote is a lot more important than the Hispanic vote, though, so that's not what we get.
Inductivist has already pointed out that the swing states are whiter and less Hispanic than the settled states are. Deja vu? We saw something very similar in 2008.
Let's systematically compare the demographics of the toss up, leaning, and settled states, and configure an estimate of the ethnic/racial vote shares in each of these three groups of states. To do so, I use categorizations from the NYT's interactive electoral map to classify each state as either toss up, leaning, or settled. Using 2010 US Census information, I then figure the shares of whites, blacks, Hispanics, and Asians (all groups other than Hispanics are comprised of non-Hispanics) in each of the three categories, with the "other" racial category being excluded from all calculations. I multiply these shares by a particpation factor for each racial grouping that is computed by taking the share of total votes cast in the 2008 presidential election and dividing it by the share of the total population the corresponding racial grouping represents. I proportionally correct for rounding errors and the demographic change that has occurred in the last three years so that the estimated percentages equal 100% (instead of 102%, 104%, and 96%, respectively). I don't make an attempt to correct for differential voting rates by race across states (if you're aware of an easy way to do this, feel free to alert me to it), electing instead to use the national rates and apply them equally to each individual state. While the country has continued inexorably towards becoming less white and more Hispanic and Asian in the intervening three years, exit polls consistently understate the white share of the vote and overstate the minority shares of the vote so the small discrepancies are more-or-less a wash*.
That said, the predicted racial total vote shares for the toss up, leaning, and settled states:
How inconvenient that homogenous states like Wisconsin and New Hampshire are the most hotly disputed while vibrant states like California and Texas are settled contests. And in Florida, the most racially diverse of the competitive states, the significantly Cuban Hispanic population is unrepresentative of Hispanics in much of the rest of the country on issues like taxation and immigration. The competitive states are more white and less black, Hispanic, and Asian than the states that have essentially already been decided are. More than four out of five of the 'determining' votes cast will be cast by non-Hispanic whites. It looks like appealing especially hard to the concerns of (mostly middle American) whites is the key to victory in November. Even this nation's electoral map is racist!
* That methodological description isn't the easiest thing to follow, so if you're interested in the raw data and how I used them, email me at firstname.lastname@example.org and I'll send the excel file right over.