The Gallup organization’s long awaited switch to the more accurate “Likely Voter” screen from the “Registered Voter” screen created a lot of fanfare but for many of the wrong reasons. President Obama whose job approval had been mired below 50% (awful for an incumbent this late in the game) received a 5-point bump to 53% in Gallup’s latest survey. This is hugely important because “a president usually pulls in a vote share roughly equal to his job approval rating.” This led to laudatory headlines like: “Obama’s First Term Approval Ratings Now Equal Clinton and Reagan.” You can just feel that Reagan ’84 landslide coming for Obama can’t you? Unfortunately for President Obama and his supporters, Alan Abramowitz at the Huffington Post dug into the numbers and found some unusual changes to the racial make-up of the poll the occurred during this switch:
Evidence from Gallup’s weekly presidential approval results indicates that the racial makeup of its tracking poll changed dramatically between the final week of September and first week of October — a change that coincides with the beginning of Gallup’s reporting of likely voter results in the presidential election. Although Gallup does not report the racial composition of its tracking poll sample in its weekly presidential approval results, we can estimate the racial makeup of the sample by extrapolating from the reported approval rating of the president among whites, nonwhites and all adults. The estimated nonwhite percentage of the sample for the past five weeks was as follows:
Gallup: % of Non-White Voters Surveyed
Changes in poll re-weighting like the above are what drive sharp criticisms from Republicans who, absent reasonable justifications by the polling organizations for such moves, allege bias when incompetence or unseriousness may be the more accurate aspersion. The race alterations specifically to Gallup’s polling assumptions have two fatal flaws: first, they fail to account for the missing White vote from 2008 and second, the steep increase in non-White participation maps out to a chart only a late-90s “dot com” CEO could love.
The Missing White Voter
The near-constant focus on the rise in non-White percentage of the electoral make-up ignores the fact that 1.7 million White voters stayed home in 2008. I addressed this issue at length in “The Reality of 2012 Voter Turnout: The White Voter” where we saw that in the 2008 election a -1.1% election-over-election drop in White participation led to an electorate make-up comprising record high levels of non-Whites totaling 23.7%. This drop in White participation was due various reasons — apathy towards Obama’s opponent, disinterest in expected loss, bad campaign message etc. The thrust of the argument was that while Obama campaign officials argue aggressively for polls with a racial composition at meaningfully greater minority levels than the 2008 historic turnout, there is another side of that coin and it works heavily against the turnout models of both the Obama campaign and of the vast majority of polls being published.
Gallup today, however, proceeds to publish a survey the Obama campaign could wish for only in their wildest dreams. In their survey, the non-White percentage of the electorate comprised 30.6% of the entire poll — a 6.6% increase over the record 2008 level (and 4.9% jump week-over-week). This increase comes in the face of steep drops in enthusiasm among Hispanics and Asians as well as a reduced (albeit small) drop in enthusiasm/support among Blacks. The decreased enthusiasm invariably translates into a decreased propensity to vote. In 2008 Obama had countless advantages — historic candidate, bad economy/financial crises, Bush fatigue, and an ill-financed opponent — that not only do not exist today but actively work against him — bad economy, chronic unemployment even worse among minorities, well-funded opponent and energized voters antagonistic towards Obama. Yet Gallup finds minority interest, voter registration and enthusiasm dramatically superior to 2008 when every measurable aspect is worse? Impossible, unrealistic and unworthy of a serious organization.
The Non-White “hockey stick”
But ignoring the missing White vote is not the only problem with Gallup’s implausibly high 30.6% non-White voter participation. How does Gallup’s 30.6% non-White participation fit in with recent historical trends?
Below we have the non-White percentage of vote in each Presidential election since 1988 according to the Census Bureau, Current Population Survey, May 2010. From left-to-right there is a steady increase in minority participation especially since 1992.
The change in Non-White participation election-over-election beginning in 1992 has averaged a 1.7 percentage point (ppt) increase over those five elections which includes 2008′s historic jump of 2.9ppt over the 2004 level.
Gallup’s assumption today, however, is that in 2012 the non-White percentage will jump 6.9ppt over 2008′s historic level despite all of the disadvantages previously outlined.
1998 “Dot Com” CEOs would be proud of whichever salesman sold that “hockey-stick” 2012 change in non-White voter participation to the Gallup organization. The absurdity of the dramatic increase in non-White participation in the Likely Voter screen is compounded because of this demographic’s historic reticence to participate anywhere near near such levels. Gallup better get that “Obama.com” IPO sold before November 6 because underneath this absurd racial make-up is a dwindling support for Obama among White voters likely to doom his re-election chances. And if Obama has any more debates like his last, he’ll make the Facebook IPO look like the homerun of the century.
Although the sharp increase in Non-White participation helps goose up Obama’s job approval today to a stellar 53% (52.5% actually) it only achieves this because the non-White demographic approves of President Obama at a 77% level in the Gallup survey, largely consistent with historical trends and independent surveys. The White demographic approves of President Obama only at a 41% level — bad, and possibly fatal. Because the above outlined changes are wholly unjustifiable, all they serve to do is mask the reality that President Obama’s job approval remains below 50% which imperils any incumbent’s re-election. If we adjust the non-White participation to more acceptable levels (76/24 White/non-white split used in nearly every national poll), Obama’s job approval drops back below the 50% threshold his campaign is desperate to avoid.
Gallup does a great disservice to polling with changes like those outlined above. Although the racial re-weightings are nearly impossible to justify I’m sure Gallup has their reasons. If nothing else they should buoy the Obama supporters because otherwise Obama supporters may become even less enthusiastic come election day and not even show up.