In turns, Republicans and conservatives have been infuriated and worried about the media polls of supposed “likely voters” who are between 5% and 11% more Democratic than Republican. Citizen Pamphleteer and other blogs like Unskewed Polls have been busy correcting these polls as fast as they come out to reflect the reality of a tied electorate as in 2004 and 2010.
This still leaves the question of why the media polling is so biased towards the Democrats. The answer is probably less a grand conspiracy to screw with the minds of Republican poll consumers and more a weighting of their polls to conform to their preconceptions of the racial makeup of the electorate.
In a recent article, Dick Morris explains:
[Telephone surveys have] too few blacks, Latinos, and young people and too many elderly in its sample…because some don’t have landlines or are rarely at home or don’t speak English well enough to be interviewed or don’t have time to talk. Elderly are overstated because they tend to be home and to have time. So you need to increase the weight given to interviews with young people, blacks and Latinos and count those with seniors a bit less.
Normally, this task is not difficult. Over the years, the black, Latino, young, and elderly proportion of the electorate has been fairly constant from election to election, except for a gradual increase in the Hispanic vote. You just need to look back at the last election to weight your polling numbers for this one.
But 2008 was no ordinary election. Blacks, for example, usually cast only 11% of the vote, but, in 2008, they made up 14% of the vote. Latinos increased their share of the vote by 1.5% and college kids almost doubled their vote share. Almost all pollsters are using the 2008 turnout models in weighting their samples…
If you adjust virtually any of the published polls to reflect the 2004 vote, not the 2008 vote, they show the race either tied or Romney ahead, a view much closer to reality.
Actually, the poll weighting methodology is often more skewed than Morris believes. Many media polls like Gallup are weighting their polls using the racial and ethnic makeup of all United States residents in the 2010 census, which includes millions of disproportionately minority residents who should not vote because they are not citizens or have not bothered to register to vote because they have largely checked out of the political system. This thumb on the polling scale creates an electorate even more favorable to Democrats than in 2008.
Weighting according to the census creates a heavily minority and Democrat-leaning electorate of which Democrats (and apparently media pollsters) have often dreamed, but which does not actually exist. Apart from the 2008 election where many white voters stayed home and minority voters turned out in record numbers, every other election including the last cycle in 2010 have substantially higher white and lower minority turnout. For example, the 2008 exit polling showed an electorate of 74% white, 13% African American and 9% Hispanic. Two years later, the 2010 exit polls showed a far more standard electorate of 78% white, 10% African American and 8% Hispanic. In contrast, this week’s Reason/Rupe poll projects a 2012 electorate even more skewed towards minority voters than the 2008 outlier election – 72% white, 11% African American and 12% Hispanic. The Rupe poll’s projected hispanic turnout is 50% higher than just two years ago!
Like most fantasies, this media poll created electorate will not survive contact with reality in November. My Republican friends need not be concerned and my Democrat friends need to treat the media polls like Lucy offering to hold the football for Charlie Brown.