CommunityFDL Main Blog

The Poisoned Landscape VI: Race And The Gender Gap Over Time

landscape.jpg[Editor’s note: This is the last of a six-part series by Paul Lukasiak on what polling reveals about how Americans will vote in the coming election. For more details, see parts 1, 2, 3, 4, and 5.— DN]

PART VI: When Race Becomes an Issue — The Gender Gap Over Time

The SUSA 50 state poll provides an excellent opportunity to describe the contours of the political landscape at a specific point in time, but cannot tell us anything about voter behavior over time within that landscape. The image is static; it’s not a movie, but a snapshot (or, perhaps more appropriately, a multi-dimension holographic image), and where voters are “pooled” like a liquids in 50 different areas of that landscape.

As the contours of the landscape change, the voters wind up “pooling” in different places, and a political campaign is the equivalent of candidates and other people trying to shift the contours of the landscape to get the voters to “pool” where they want them. While the actions of the candidates and others can change the contours of the landscape by raising issues and providing new information, the landscape can also change “naturally”. But ultimately, it is “the media” that controls the landscape by controlling the access, and more crucially, the nature of the access, to the levers and pulleys that change the contours of that landscape.

It is “the media” that decides whether “illegal immigration” or “universal health coverage” or “Hillary Clinton’s tax returns” is actually the force determining the contours of the landscape, and decides who has what kind of access to the levers of power as the landscape changes. The SUSA 50 state poll provides a still picture of what looks like a “political landscape”, but when that picture comes to life and the contours begin to shift, it is because what we are really looking at is a “media landscape”.

The media manifestly shape voter behavior, and that couldn’t be clearer than when we examine how their handling of issues of race (and accusations of racism) over the course of the campaign alters voting patterns in ways that leave the gender gap relatively unaffected.

So as the landscape changes, and the voters “pool” in different areas, the gender gap itself usually changes. South Carolina and Ohio provide two examples of how the gender gap changes as the “media landscape” changes — but how, when the media decides to make “race” an issue, it can shift large numbers of voters without substantially altering the gender gap.


When, subsequent to her win in New Hampshire, false accusations that the Clinton campaign was deliberately injecting race into the campaign, it created significant changes in the established trends that defined the changes in the gender gap in South Carolina. And contrary to popular opinion, the efforts of the Obama camp (as documented by Sean Wilentz) to cast the Clinton campaign as racist backfired – while Clinton lost 3% of her Black support, and Obama gained 5 points among Blacks, the real impact was on Obama’s loss of a quarter of his White support while the controversy raged, and the shift of 12% of the White vote to Clinton.

Once it ebbed, when the voters went to the polls, Clinton had lost the White support she gained during the controversy, but only a small fraction of the White votes (1/7) that wound up “in play” found their way to Obama. Ironically, all these changes had little visible impact on Clinton’s gender gap – and considering the major shift in candidate preference, the impact of these changes in gender distribution for Edwards and Obama was negligible in the short term. If the “racial controversy” in South Carolina had any impact on the gender gap, it was in altering the trends that would have continued had the controversy not taken place.

Thanks in no small part to overwhelming positive media coverage (and a media that all but ignored John Edward) for Barack Obama. By early November 2007 Obama, a candidate with less than three years of national political experience, had managed to establish himself solidly in the No. 2 position for the Democratic nomination. Hillary Clinton was still getting twice as much support in national polls as Obama, but Obama was getting twice the support of third-place Edwards. Obama was rapidly becoming the candidate of choice for the “anyone but Hillary” crowd, and had even managed to overtake Edwards in Iowa.

But after starting strong in the national polls in the winter of 2007, by early spring Obama had stalled, and started fading over the summer as Clinton increased her own support. Obama began picking up steam again in early October, apparently because Al Gore had made it clear that he was not running. (By mid-October, most pollsters were no longer including Gore in its questions on the Democratic nomination.)

And thus the stage is set for the first Democratic primary poll taken by Survey USA in South Carolina, conducted on Nov. 9-11. This poll showed Clinton with a solid 14-point lead over Obama, with Edwards mired in third place. And while Obama was already ahead of Clinton among males by 19 points, her even more substantial, 33-point lead among women (who were projected to make up 62% of the electorate) gave Clinton a nice cushion. The gender gap was at its peak in November in South Carolina, with a 52-point spread between Obama (LIGHT BLUE line) and Clinton’s (PINK line) gender gap numbers.

Obama was also well ahead among African American voters, getting the support of a clear majority (DOTTED BLUE line) of 52% to Clinton’s 39%. But perhaps most significant, in terms of this discussion, is Edwards’ incredibly poor showing among Black voters: after winning a plurality (37%) of the Black vote in the 2004 primary, and spending four years focusing on issues of important to the Black community (like poverty and economic justice), Edwards was only getting 3% of the Black vote.



But Obama was lagging well behind in the White vote, with Clinton at 55% (SOLID RED line) to Obama’s 15% (SOLID BLUE line.). This data suggests that Obama had established himself as a viable enough candidate for African Americans to support as a gesture of positive social identification, in early November White voters on the whole remain unconvinced. The data also strongly suggests that in early November, there was a considerable gender gap among Black voters — Clinton’s overwhelming support among women, Obama’s strong showing among men and African Americans, and weakness with White voters (and subsequent gender and racial polling trends) seem to show that a significantly large percentage of Black women than Black men were supporting Clinton at that time.

Two polls taken over the subsequent five weeks (December 7-9 and Dec 17-18) show that both Clinton and Obama maintain about the same percentage of the male vote, with Edwards gaining 11 point among men. Edwards also picks up six points among women, while Obama gains 8 points while Clinton loses 8 points among women. As a result, Clinton’s lead over Obama was cut to 2 points. The loss of women for Clinton and the increase in female support for Obama results in the shrinking of their respective gender gaps.

However, during the same period, there was little change in Clinton’s Black support (-2%) or Obama’s White support (+3). Obama also picks up an additional 5% of the Black vote. But the big change is in the White vote loss of 11 points for Clinton, and Edward’s gaining 15% more of the White voters. Clinton was looking increasingly vulnerable, but White voters shifted to Edwards, and not Obama.


It is during the next three weeks that the most dramatic changes in the gender gaps took place – a period during which Obama overtook Clinton in Iowa polls and subsequently won the Iowa caucuses. The poll taken on Jan 4-6 shows that while Clinton lost only 2% of the male vote during the period, one third of her female support fell away (going from 50% to 33%) while Obama picked up 14% of women voters. Obama also increased his male support by 6%, and Edwards lost the same amount of male support.

As a result of these shifts, major changes in the gender gaps for the candidates too place. Clinton’s gap was reduced by nearly 70%, going from negative 22 to negative 7 (a negative gender gap number means that a candidate has more support among woman than men, a positive number means male support is greater). Edwards’s gap also dropped substantially (by 7 points), to where there was almost no difference in how the genders voted for Edwards. And Obama’s gender gap was reduced by more than half, from 15% to 7%.

It also during this “pre-controversy” period that Clinton loses the biggest chunk of her Black support, going from 37% of the Black vote to 23%, while Obama picks up an additional 12% of Blacks (and Edwards loses 1%). Clinton and Edwards also loses 8% and 4% of their White vote respectively, while Obama picks up 11%. Because Edwards gender gap drops substantially, and he lost 4% of Whites and 1% of Blacks, it is safe to say that most of the gender gap drop was due to the loss of White males. This fact, and the fact that Clinton lost only 2% of her male support, tells us that the Black voters that shifted to Obama were significantly disproportionately female.


It is at this point that the “racial code-word campaigning” accusations against Clinton started to fly, and they reached their peak right before the next polling was done on Jan. 16-17. (Obama finally admitted that there was nothing “racial” in the comments of Bill or Hillary Clinton right before the Jan. 15th debate in Las Vegas, but never acknowledged that not only his supporters, but his own campaign, were responsible for the phony controversy.) And while the biggest impact was on racial voting patterns, it also had an impact on the gender gap.

The changes that occurred can best be accounted for by simply assuming that South Carolinians, both Black and White, know the difference between real “dog whistle” racism, and false accusations of the intentional use of “racial code.” Black voters recognized that the accusations were false, and during the controversy Clinton lost only 3% of her Black support, while Obama’s Black support rose by only 5 points. (Indeed, given that Obama wound up with 78% of the Black vote in SC, when he got at least 85% of the Black vote in subsequent southern primaries in Louisiana, Georgia, Mississippi, and Alabama, the controversy may have dampened African American enthusiasm for his candidacy in SC.)

White Democrats, meanwhile, seem to focus not merely on the fact that the accusations were false, but where those accusations were coming from, and as a result of the controversy Obama lost 7%, or nearly a quarter, of his White support. White undecideds fell from 5% to 2% in this period, and Edwards also lost 2% of his White support, resulting in Clinton going from 38% of the White vote to 50%. This strong White shift back to Clinton can best be explained by the idea that White Democrats began to identify with Clinton, based on their own fears (or experiences) of being falsely accused of racist intent.



Ironically, these changes had no impact on Clinton’s gender gap – she picked up 6% of both the male and female vote. But Obama lost 7% of his male support while losing only 2% of his female support, resulting in a 5% reduction in his gender gap. And Edward’s male support went up by 2% during this period, while his female support went down by 2%, resulting in a net increase of 4% in his gender gap.

The subsequent SUSA poll, taken on Jan 23-24 after the controversy had died down and right before the Jan. 26 primary, shows little change in the Obama’s percentages of Black and White voters, but his gender gap numbers decline anyway by four points – this can be explained by noting that the change in the racial distribution of the sample occurred. (see Note 4) And while there was a 12 point shift in the White vote from Clinton to Edwards, given the distorting effects (see note 4) of the change in demographic distribution, it is difficult to say that any real change is reflected in the slight changes in Clinton’s and Edward’s gender gap numbers.

What is most surprising about the final gender gap numbers from the South Carolina exit poll is how small they are, and how the injection of race appears to have altered the trends that were established. Less than two months before the primary, there was a 51% difference between the Clinton and Obama gender gap numbers, the exit poll numbers show only a 7 point gap. The “racial code” controversy seems to have completely erased the gender differences with regard to Obama, leaving only the 14 point gap between the White male and White female candidates to testify to the existence of such a gap, and gender gap numbers that were inconsistent with the trends established prior to the controversy erupting.

In other words, South Carolina showed on a temporal basis what was seen in the SUSA 50 state poll snapshot – racial considerations tend to “flatten out” in elections.


As a crucial swing state for November, Ohio has been among the most polled of all states when it comes to prospective presidential matchups. Thus it provides an excellent setting in which to examine the changes in the gender gap over time – and what happens to the gender gap when “race” suddenly becomes an issue.

Survey USA’s first “match-up” polling that included both Clinton and Obama against McCain was taken on Dec 3-5, 2007. It was a time when few Ohioans were focusing on the presidential campaign (after all, the general election was almost a year away, and the nominations would be settled well before the Ohio primary, right?) It was a time when McCain was mired in the teens in GOP primary polling, a newcomer named Barack Obama was getting a lot of good press, and while Clinton was maintaining her lead in the national polls, she was not looking quite as inevitable as a few months prior before.

Thus this poll, and the next two SUSA polls, taken Dec 13-15 and Jan 1-4, should be given little emphasis in terms of how McCain, Obama, and Clinton were doing relative to each other. But it they do provide the first clues of how the gender gap changes as voter attitude changes. (Note that gender gap number is be based on Male percentages MINUS female percentages, and is not directly related to changes in male and female margins. A positive gender gap number means greater support from men, a negative gender gap number means higher support from women.)



As Clinton’s margins against McCain in these polls went up by six points (from –6% to –2%, her gender gap was reduced by seven points (from –17% to –10%), entirely due to a relative increase in male support (from a 26% deficit to a 12% deficit among males.) But while Obama also improved against McCain (from -11% to –7%) his gender gap expanded, going from –13% to –16%. This occurred because Obama did better among both men and women, but he improved the most among women, resulting in a lower negative number.

The next two SUSA polls (Feb 15-17 and Feb 26-28) both took place after “Super Tuesday” and “Potomac Tuesday”, when the race for the Democratic nomination was focused on Ohio (and Texas). Both Clinton’s and Obama’s margins jumped to 10 points over McCain, a 12-point gain for Clinton, and a 17-point gain for Obama between the early January and late February polls.

And the gender gap for both candidates also shrank over that period. Clinton’s was reduced 4 points to –6% as she gained 17% (from –12% to +5%) among men, and 8% points (+8% to +16%) among women. Obama’s gender gap was reduced from –16% to –2%, as his male margin against McCain leaped from –12% to +5%, while his female margin increased only 3 points, from +8% to +11%.

Ohio provides an excellent way of understanding how the gender gap operates over time. It grows and shrinks based on the relative increases and decreases of support of male and female voters of both candidates. And Ohio shows how men and women don’t make the same decisions at the same time.

Or that is, they don’t make the same decision at the same time most of the time.



The next SUSA poll was taken on March 14-16, and shows that Clinton’s overall support against McCain (RED line) dropped by 4 points, while her male margin (VIOLET line) dropped by three points, and her female margin (PINK line) dropped by 7%. But her gender gap remained the same. In fact, her male and female support remained unchanged (supported by 47% of men, and 53% of women in both the late February and the March poll), the change in margins was due entirely to undecided voters choosing McCain. The previous poll showed 11% male, and 10% female, undecided voters. This new poll showed that only 8% of men and 3% of women remained undecided.

Some of this should be attributed to a significant change in the voter sample, which went from 26% Republican and 49% Democrat to 34% Republican and 44% Democrat.) and a possible reporting error. The previous poll Party affiliation numbers added up to only 93% (26% GOP, 49% Dem, 18% Independent); in previous polls, and the March poll, those numbers added up to at least 96%. In fact, given these disparities, its not unlikely that there was little or change at all in the actual relative support for Clinton between these two polls

But while Clinton’s support remained relatively stable, the same cannot be said for Obama. Those three sharply decending lines on the Chart A3 represent Obama’s male (LIGHT BLUE), female (LIGHT GREEN), and overall (DARK BLUE) margins against McCain. Obama’s 9 point advantage over McCain among men became a 10 point deficit, his margin among women dropped from +11 to minus four, and he went from leading McCain by 10 points to trailing McCain by 7. However, Obama’s gender gap did not change.

Obama’s went from having the support of 49% of men and 51% of women to only 42% of men and 44% of women, while McCain’s support among men went from 40% to 52%, and his support among women went from 40% to 48%. Even adjusting for the potential glitch described above, McCain’s support would still have risen to 49% among men, and 41% among women.

All this seismic shift in Obama’s support can be explained with two words: Jeremiah Wright.



Obama’s white (SOLID ORANGE line) support had risen steadily, and because Ohio’s electorate is 87% White, his 12 point increase in White support (from 34% of White voters in mid-December to 47% at the end of March) meant an increase of 10.4% of voters overall. And most of Obama’s gain in White support came at McCain’s expense—McCain dropped from 51% to 42% of the White vote during the period that Obama was gaining, contributing another 7.8% of voters to his overall margin shift of +19 points during that period.

But when the Wright controversy hit, Obama’s White support dropped from 47% to 37%, representing 8.7% of voters. Meanwhile, McCain went from having 42% of the White vote against Obama to 56%, an increase of 13.9% overall. Even assuming the potential glitch mentioned above was all White voters who were added to McCain’s support, McCain’s White support still increases by 4%, resulting in an increase of 3.5% of all voters, and a combined net gain for McCain of 12.2% of voters.

And while Obama’s Black support did rise from 78% to 84% during this period, because African Americans are only 10% of the Ohio electorate, that 6 point jump represents on 0.6% of voters, far too few to make a significant difference in the decline of Obama’s margin against McCain.



And, based on the reported data, Obama’s gender gap did not change during this period. And while we can’t be certain that this is the case because of the potential glitch, because the impact of that glitch seems to have been felt only in McCain’s numbers, Obama’s gender gap would have been unaffected.

Overall, this massive change in voter preference represented an end to a trend where Obama’s gender gap was shrinking rapidly, and the Wright controvery brought that to a screeching halt. Unfortunately, Survey USA has not done a poll since the controversy died down, so we can’t determine if/how much longer term damage was done to Obama in Ohio, and how those changes are reflected in the gender distribution of his support.

Previous post

MassResistance hard up for ca$h

Next post

Sourcing Skepticism ...

Paul Lukasiak

Paul Lukasiak