Quick. Thirty-two degrees. Hot or cold? It depends. In Fahrenheit, it’s literally freezing; in Celsius it’s sweltering. One hundred dollars. A lot of money or a little? A veritable fortune in Timor, where it’s a quarter of the average annual wage. In Manhattan, not so much. Numbers seem absolute, but they only have meaning in context.
So it is with poll data. Is 80 percent of the African-American vote a lot or a little? It sounds like a lot, but if you’re a Democrat heading into Election Day with just 80 percent support from African-Americans, you ought to be quaking in your boots — it’s rarely enough to win.
{mosads}What gives these numbers meaning is two bits of context — how they compare to numbers in similar situations across time or space and how they compare to what one needs in order to emerge victorious.
Yet rarely do pundits, or even pollsters, provide readers with the context they need to interpret poll numbers meaningfully. Our firm has routinely provided that context for years, but I can’t tell you how many times I’ve listened to folks crow about a candidate garnering 53 percent of the female vote, or 80 percent of Democrats, or leading independents by 5. None of that is good news if you need 58 percent of the female vote or 93 percent of Democrats, or a 10-point margin among independents to actually win the race.
One prominent example of the failure to provide context, and therefore meaning, has been on display in the running commentary declaiming “Obama Struggling With White Voters.” Some analyses are vague, pointing to Obama’s “eroding position among whites.” Another treated the president’s then-44 percent approval among white voters as potentially career-ending, without providing any sense of what constituted a “good” number, based either on history or electoral necessity. Karl Rove saw the president’s declining support among whites as evidence he would lose the election. The numbers behind Rove’s assessment: White support for the president was down 25 points from his inauguration. Rove at least tries to provide context, albeit by making an absurd comparison to the inaugural, when most presidents reach an artificial high.
A recent ABC/Washington Post poll stimulated a rehashing of Obama’s “problem” with white voters under nearly identical headlines. Yet none of these stories, save the Post’s own, put the numbers in any context. In that poll, Obama garnered 39 percent of the white vote. Jon Cohen and Karen Tumulty wisely note that Obama had received only 43 percent in 2008, while winning the general election decisively, and that John Kerry and Al Gore had posted results similar to Obama’s current standing.
Few, if any, of the other myriad articles stimulated by that poll provided that context. None noted that despite his showing among white voters, Obama bested Romney by a 3-point margin in the Post poll or that he needs only about 38 percent of the white vote to win in November.
Moreover, we now know from research by Columbia University’s Andrew Gelman that the bases of partisan support vary considerably by state. Gelman focused on class, but there are differences in racial patterns as well. In 2008, Obama’s vote share among whites ranged from 10 percent in Alabama to 86 percent in Washington, D.C. In swing states there was substantial variation between those extremes, with the president capturing 55 percent of whites in Colorado, 54 percent in Wisconsin, 51 percent in Iowa and Michigan, 45 percent in Nevada, 39 percent in Virginia and 35 percent in North Carolina. With such wide variation by state, the “national white vote” is not a particularly telling indicator in a system where state victories build Electoral College majorities.
It’s just impossible to understand the import of the poll data without context. Reporting poll numbers without the information that renders them meaningful does everyone a disservice.
Mellman is president of The Mellman Group and has worked for Democratic candidates and causes since 1982. Current clients include the majority leader of the Senate and the Democratic whip in the House.