An analyst for RealClearPolitics takes a look at possible issues with the racial breakdown in the samples of presidential election polls. A few of the issues:
First, as Chait repeatedly concedes, we don’t know what the ultimate electorate will look like this November. That really should be the end of the argument — if we don’t know what the racial breakdown is going to be, it’s hard to criticize the pollsters for under-sampling minorities. After all, almost all pollsters weight their base sample of adults to CPS (current population survey) estimates to ensure the base sample reflects the actual population; after that, the data simply are what they are.
It’s true that the minority share of the electorate increased every year from 1996 through 2008. But there’s a reason that 1996 is always used as a start date: After declining every election from 1980 through 1988, the white share of the vote suddenly ticked up two points in 1992. In other words, these things aren’t one-way ratchets (and while there is no H. Ross Perot this year, the underlying white working-class angst that propelled his candidacy is very much present, as writers on the left repeatedly have observed)…
“The U.S. Census Bureau allows for multiple responses when it asks respondents what race they are, and Gallup attempts to replicate the Census in that respect. While most pollsters ask two separate questions about race and Hispanic ancestry, Gallup goes a step further, asking five separate questions about race. They ask respondents to answer whether or not they consider themselves White; Black or African American; Asian; Native American or Alaska Native; and Native Hawaiian or Pacific Islander.”
In other words, how you ask the question could impact how people self-identify with regard to race and ethnicity, which could in turn affect how your weighted data look. This is a polling issue that will likely become more significant as the nation grows more diverse, and more multi-racial.
Trying to figure out who exactly is going to vote is a tricky proposition and it is little surprise that different polling organizations have slightly different figures.
I hope people don’t see stories like this and conclude that polls can’t be trusted after all. Polling is not an exact science; all polls contain small margins of error. However, polling is so widely used because it is incredibly difficult to capture information about whole populations. Even one of the most comprehensive surveys we have, the US Census, was only able to get about 70-75% cooperation and that was with a large amount of money and workers. Websites like RealClearPolitics are helpful here because you can see averages of the major polls which can help smooth out some of their differences.
A final note: this is another reminder that measuring race and ethnicity is difficult. As noted above, the Census Bureau and some of these polling organizations use different measures and therefore get different results. Of course, because race and ethnicity are fluid, the measures have to change over time.