Much has been written about the role of demographics in the 2012 election, specifically, the decreasing percentage of votes cast by non-Hispanic Whites, and what it portends for the future of American politics. To illustrate the importance of these trends, and the crippling disadvantage they are likely to inflict on the Republican Party in the coming decades, I decided to see whether the result of the only (relatively) undisputed Republican presidential win since 1988 would have been changed if the ethnic characteristics of the 2012 electorate had been in place in 2004.
What I have done is to take the 2004 election results in four states, and add to them the increase in votes cast by each of five ethnic groups in 2012 over their 2004 votes in each state, divided among the parties in the same way as they were divided in 2004. So this assumes that Kerry gets exactly the same percentage of votes in each group as in 2004, and gets more votes only if that group cast more votes in 2012.
I have used the 2004 and 2012 exit poll breakdowns for each of the ethnic groups.
The best way to illustrate the methodology is to work through the easiest of the four states, Nevada.
In 2004, there was a total of 815,880 votes cast for Bush or Kerry. Bush got 418,690 to Kerry's 397,190, for a margin of 21,500. That year, the ethnic breakdown of voters in Nevada, according to the exit poll, was 77% White, 7% African-American, 10% Latino, 3% Asian, 3% Other. Whites voted 56.5-43.5 for Bush (throughout this I'm using only the two party votes, not the total including other candidates); African-Americans 86.9% for Kerry, Latinos 60.6% for Kerry, Asians 51.5% for Bush, and other 56% for Kerry.
In 2012 according to the last numbers listed, there was a total of 531,373 votes for Obama and 463,567 for Romney, a total of 994,940 votes. That's an increase of a whopping 179,060 votes, nearly 22% above the 2004 level. The 2012 exit polls tell us that the 2012 electorate was 64% White, 9% African-American, 18% Latino, 5% Asian, and 4% other.
Applying those percentages, we get the following numbers for approximate totals of votes cast by each ethnic group in the two years.
African-
White American Latino Asian Other Total
2008 628,228 (77%) 57,112 (7%) 81,588 (10%) 24,476 (3%) 24,476 (3%) 815,880
2012 636,762 (64%) 89,545 (9%) 179,089 (18%) 49,747 (5%) 39,798 (4%) 994,940
Diff. 8,534 32,433 97,501 25,271 15,322 179,060
What we're going to do here is take those extra 179,060 votes and distribute them among the ethnic groups. First we find the totals for all ethnic groups in 2012, and then we apply to those figures the 2004 exit poll figures for each group, so that our final figure reflects what the result would most likely have been in 2004 if every group and every person who voted then had voted exactly the same, but the extra voters in each ethnic category had been added.
Increase 2004 Breakdown Kerry Add Bush Add
Whites 8,534 .435-.565 3,712 4,822
African Americans 32,433 .869-.131 28,184 4,249
Latinos 97,501 .606-.394 59,086 38,415
Asians 25,271 .485-.515 12,256 13,015
Other 15,322 .560-.440 8,580 6,742
Total New 179,061 111,818 67,243
2004 Totals 397,190 418,690
Adjusted 2004 509,008 485,933
And that's just common sense, when you think about it. Bush won the state in 2004 by a little over 20,000 votes, but since then 179,000 new voters have been added, and only about 8,500 of them are Whites. Even at 2004 levels where Bush was getting nearly 40% of Hispanics and almost 14% of African-Americans, that large increase among non-Whites would have been sufficient to change the final result by about 45,000 votes, from a Bush win by 21,500 to a Kerry win by 23,000.
Next, Ohio:
In 2004 in Ohio, Bush received 2,859,768 votes, and Kerry received 2,741,167, for a two party total of 5,600,935, and a Bush margin of 118,601. That year, the ethnic breakdown of voters in Ohio, according to the exit poll, was 86% White, 10% African-American, 3% Latino, 1% Asian, and 1% other. African-Americans broke 84-16 for Kerry. No breakdowns by candidate are available for Latinos, Asians, and others.
In 2012, according to the latest numbers available, Obama received 2,819,454 votes to Romney's 2,655,117, for a total of 5,474,571, so that there were actually 126,364 fewer votes cast in Ohio in 2012. The 2012 exit polls tell us that the 2012 electorate was 79% White, 15% African-American, 3% Latino, and 3% other.
Those numbers seem to me adequate to prove the point without a great deal of further analysis. The White share of the vote dropped 6.15 points, the African-American share rose by five. With a total vote of about 5,600,000, those changes would indicate a decrease of 344,000 White voters (due to death, migration, or apathy) and an increase of 280,000 African-Americans. In 2004 African-Americans in Ohio broke 84 to 16 for Kerry, so those 280,000 votes would have been expected to break 235,200 for Kerry and 44,800 for Bush. In addition to that, the reduction of White votes by 344,000 would have trimmed another 41,280 votes off Bush's margin, as he led that group by 12 points, 56 to 44.
So we can conclude that if Ohio voters in 2004 had been 79% White and 15% African-American, instead of 86.15% White and 10% African-American, it would have been Kerry, rather than Bush, who carried the state by about 115,000 votes.
That's two states which cast 25 electoral votes in 2004. As you may recall, Kerry received 252 electoral votes (well, 251, as one Minnesota elector voted for John Edwards), and the addition of Ohio's 20 would have elected him.
Since this one was rather lengthy, I'll post the figures for Colorado and New Mexico in the next couple days, after the final 2012 results are certified.