Sunday, June 22, 2008

Democratic nomination voter totals by race, gender

With the Democratic nomination process essentially over, it might be of interest to see basic demographic breakdowns of who voted for whom. The number of voters are in thousands. Green shows the percentage of the candidate's total base the category represents (without respect to Asians and "others", as explained below). Blue shows what percentage of the category each candidate garnered relative to the other.

Hillary's malesObama's malesHillary's femalesObama's females
6,859 38.1% 43.6%8,857 45.7% 56.4%11,166 61.9% 51.5%10,520 54.3% 48.5%

Hillary's WhitesBlacksHispanics
13,617 79.1% 56.0%1,0025.8% 14.9% 2,589 15.1% 63.7%

Obama's WhitesBlacksHispanics
10,68859.7%44.0%5,74132.1% 85.1%1,472 8.2% 36.3%

Notes on the methodology come after my blather. The beginning of that section is indicated by brackets.

The first thing that jumps out is how little things changed since I attempted a similar analysis following the "Super Tuesday" contests that included all primaries that had taken place up to that point. The percentages are nearly identical (Hillary's numbers are very slightly worse across the board as I did not attempt to include caucusing states the first time around but did so this time), even though this time around there are 2.5 times more votes to contend with. Recall how much of a harpy's nest the Democratic party has become--60% of primary and caucus voters this year were women. The evidence that female suffrage drove the US sharply leftward keeps mounting.

The white and black splits nearly mirror how the two racial groups voted in the '04 Presidential election. Hillary received 56% of the white vote; Bush got 58%. Obama received 85% of the black vote; Kerry took 88%.

So if Hillary's support looked like Bush's among whites, and she also cleaned up among Hispanics, why didn't she win? Bush did so comfortably even without Hispanics. Well, this is the Democratic party. It's a lot blacker than the national electorate is. Whites comprised 69.2% of voters in the Democratic contests (not including Asians and others) compared to 81.9% of the '04 general. Blacks made up 19.2% of the Democratic contests compared to 11.7% of the general.

When black support is so overwhelmingly one-sided, that 7.5 point jump makes an enormous difference. It clearly put Obama over the top. Without blacks, Hillary crushes Obama, 57.1%-42.9%; a victory more lopsided than George HW Bush's win over Michael Dukakis in '88.

Since the Hispanic population is growing while the black population is basically holding steady, might this mean Republicans catering to Hispanics will be able to use the Hillary coalition to win in the general? No, for several reasons.

First, polls show Hispanics strongly backing Obama against McCain, 62%-29%, despite the Arizona Senator being the most recognizable champion of open borders in the entire Republican party. It's hardly surprising that a group that is 'eligible' for affirmative action benefits and experiences low levels of educational attainment, low incomes, high poverty rates, and high welfare use rates is going to find a home on the Democratic side.

Second, even if Hispanics are reluctant to vote for blacks, there are more potential Hispanic Presidential candidates on the Democratic side than there are black ones. Obama, the only (half) black Senator, is a unique case. As the Hispanic population continues to grow as a share of the total population, Hispanic representation in the Democratic party will continue to increase.

Third, there is no reason to presume blacks will ever vote for a white Republican over a Hispanic Democrat as Hispanics might do in the case of a black Democrat facing a white Republican--the only way GOP peeling-off could conceivably work is an ever less likely scenario going forward.

Fourth, the Hispanic establishment (in Congress, La Raza, etc) are even more Democratic than the Hispanic population as a whole is. More than nine-of-ten Congressional Hispanics are Democrats.

Fifth, we're looking at Hillary Clinton's support. Unless the Republican party wants to get in step with her politics, the comparison is flawed from the beginning.

More likely, the Democratic contests portend the future of American politics, where white electoral power dwindles and black and especially Hispanic electoral power correspondingly increases. John Savage has argued that an Obama nomination is desirable in that it allows white Democrats to see what their party is becoming--a party whose leadership has made a career of benefitting non-whites at the expense of whites. The white vote is still the 800 pound gorilla of US electoral politics, but it won't rule the jungle by itself for too much longer.

The tables above only consider white, black, and Hispanic votes in the 50 states and DC. Including all primary votes (and caucus estimates) in addition to the four participating US territories, Obama comes out ahead, 19,555,861 to 18,344,087 (51.6% to 48.4%). As discussed below, that artificially inflates Obama's numbers by assuming states holding primaries and states holding caucuses saw similar voter turnouts, but I think for the purpose of distilling the 'will of the people', it makes sense to go about it this way.

Even though Hillary outdid him for the last three months of the campaign, her 'momentum' (a word that vanished from media discourse as soon as it shifted from Obama to Hillary) came too little, too late. I remember recently reading commentary on how Hillary's campaign ran on leftist assumptions (she staffed her campaign with women and beta males, believed whites in flyover country were too bigoted to vote for a black man, etc) while Obama's took a more hard-nosed realist approach. I can't recall where I read it, though (please let me know if you're aware of what I'm referring to). Adding to that, she clearly should've made Wright and Obama's racialist obsessions an issue instead of waiting for it to finally spill out 'on its own' after being made public by Steve Sailer more than a year ago.

['Discussion' of methodology used to compute totals follows]

States where no exit polling was conducted (primarily small states that caucused rather than holding primaries) do not yield vote totals for the candidates, only state delegates assigned to each. To estimate total turnout, I looked to a nearby state with a similar demographic and political profile (ie, for Wyoming I referred to Montana, for North Dakota I referred to South Dakota). I took the percentage of voters who backed Kerry in '04 from both states and multiplied the respective percentages by each state's total population. Then I divided that figure for the state without exit polling data by the figure for the referential state to come up with a multiplier (sometimes less than 1). I then took the referential state's total votes for this year's Democratic primary and multiplied it by the, uh, multiplier, to come up with an estimate for the total voter turnout in the caucusing state without an exit poll.*

This presumes caucus turnout is similar to primary turnout, which isn't the case. Caucuses bring smaller numbers of voters than primaries do, but I see little reason to presume that Obama's support would've suffered substantially in these states had they held primaries instead of caucuses (it would have required a greater investment of campaign resources though). The intention is to get an apples-to-apples figures by state on the way to voter totals at the national level. Still, this methodology favors Obama, primarily because older folks (who tended to support Hillary) are less likely than younger ones (who tended to support Obama) are to participate in caucuses, which involve greater raucous than primaries do.

Racial/ethnic totals are estimated for whites, blacks, and Hispanics. The exit polling data for Asians and others are to lacking to be computed nationally**. Consequently, the total for the three categories included falls short of the total estimated votes for each candidate.

For states with exit polling data but for which black and/or Hispanic sample size is insufficient to report voting patterns, votes are assigned in accordance with the national trend***. This is reasonable as neither candidate enjoyed a clear advantage among states where the method is necessitated.

For states without exit polling numbers, I simply assume those voting in the Democratic caucus matched the state's demographic profile according to the US Census bureau. Black votes are assigned at a ratio equivalent to how they voted nationally in all states where exit polling was conducted.

Hispanic votes are split equally between the two candidates. Initially, Hispanic votes were allocated in the same way black votes are, but that quickly struck me as improbable. As Obama won all of these states at margins between 2-to-1 and 3-to-1, giving Hillary two-thirds of the Hispanic vote surely led to underestimates of her white support and also of Obama's Hispanic support.

Reductions to the total are made to account for Asian and other voters with the assumption that these groups voted in the same way as the state on the whole, and then the remaining votes are assigned to each candidate as white voters in accordance with each of their total estimated votes in the state.

This method is used because whites show the most variability in their voting patterns, while blacks voted in a predictable way irrespective of which candidate actually won the state. It likely undercounts black votes and overcounts whites a bit. But all states where exit polls were lacking are overwhelmingly white (with the exception of Hawaii and Colorado) and small, so the effects on totals are very marginal. Anyway, exit polls tend to, if anything, overrepresent minorities to ensure sample size adequacy. In the '04 Presidential election, for example, exit polls showed 8% of the electorate to be Hispanic, while Pew later discovered the actual figure to be 6%.

For gender totals in the caucus states without exit polling, a similar method is used. After coming up with a state's total turnout as described previously, I allocate total votes for each candidate in accordance with the number of state delegates awarded to each. Hillary's male and female totals are then arrived at by using the her total support by gender from all states with exit polling data. Obama's support is arrived at by taking the state's estimated total votes and presuming that the gender distribution for primaries with exit polling data is the same as that state's distribution. Hillary's male total is subtracted from this overall male total to come up with Obama's male support. The same thing is done for females^.

Michigan is included, with all uncommitted votes assigned to Obama. This provides a little balance to the way total votes in caucus states are estimated. Puerto Rico, American Samoa, the US Virgin Islands, and Guam are not figured in to gender or racial/ethnic voter totals. They are only included in the national totals for each candidate. To estimate turnout for the latter three, I took Puerto Rico's Democratic primary turnout, 9.7%, and assumed the other places experienced the same turnout, except for the Virgin Islands, where I arbitrarily doubled the estimate to account for the reported surge (actual numbers were not recorded either in '04 or this time around, only delegate totals were) in voter turnout relative to '04.

Why the uptick? Blacks comprise 76.2% of the Virgin Islands' population. And they back Obama, big time. He took 89.9% of the vote. It is interesting to note how clannish blacks are in this majority black territory. In no state do whites come remotely close to showing such racial solidarity. It would be fun to see Richard Cohen chew on that!

This seems like a lot of best-guessing on my part, but these methods only apply to a small piece of all votes. The estimated vote totals from caucuses without exit polls, including three American territories, only comprise 10% of the national total. The other 90% comes from hard numbers recorded in the actual primaries.

* In '04, 39% of Montana went for Kerry, while 30% of Idaho did. Montana's population as of '06 was estimated to be 944,632. Idaho's was 1,466,465. Multiply 944,632 by .39 to get 368,406 and 1,466,465 by .30 to get 439,940. Then divide 439,940 by 368,406 to get 1.1942. To come up with Idaho's turnout this year, multiply 181,423 (Montana's total in this year's Democratic primary) by 1.1942 to get 216,655, Idaho's estimated caucus total.

** Hillary took the Asian vote in California, 250,900 to Obama's 88,345 (71%-25%). Hawaii, by contrast, is 40% Asian, and backed Obama by a 3-to-1 margin, so they probably favored him at about the same rate Asians in California opposed him (although, in absolute terms, there were about five times as many Asian voters in California as there were in Hawaii).

*** One percent of South Dakota voters were black, according to exit polls. There were 97,590 votes cast for either Hillary or Obama. It is assumed that 976 (97,590*.01) of these voters are black. Nationally, Hillary received 14.9% of the black vote. So she is assumed to have received 145 (976*.149) of these 976 votes, with the rest going to Obama.

^ In North Dakota, Hillary's total is estimated to be 27,192. From the states with exit polling data, 61.8% of her support came from women. Thus her female support is estimated to total 16,807. Her male support then comes to 10,385. The total number of North Dakota voters is estimated to be 72,023. From the states with exit polling data, it is found that 42% of Democratic primary and caucus voters are male. By multiplying 72,023 by .42 we get 30,193. Subtract 10,385 from this (Hillary's male support) to arrive at 19,808, Obama's estimated male voters. The same process, using 58%, is repeated to come up with Obama's female total.

Data are here.


Dragon Horse said...

The beta male comment I believe was made by Maureen Dowd, but I am certain you got it from here:

"Does Hillary surround herself with girly men?
Does Hillary surround herself with girly men?"

April 9, 2008

By Camille Paglia

Audacious Epigone said...

Hehe. If not, I must've picked it up from someone who got it from her.

Half Sigma said...

Good analysis.

Anonymous said...

Include p-values for significance.