In a previous post on how the "rope-a-dope" analogy is not an apt description for Obama's strategy in Wednesday night's defeat by Mitt Romney, I referenced this piece on how presidential debates have historically had little to no impact on how individual's choose to vote on election day.
There are some sections in the Bloomberg piece that deserve a bit more exploration:
"Where you started the debate season is pretty much where you end the debate season," said Christopher Wlezien, a political science professor at Temple University and co-author of the book "The Timeline of Presidential Elections."
No candidate who was leading in the polls six weeks before the election has lost the popular vote since Thomas Dewey in 1948, according to Wlezien and Robert Erikson, a political science professor at Columbia University. They studied polling data going back to 1952 and computed a running average "poll of polls" for each presidential election...
Wlezien and Erikson found only one campaign with a big movement in opinion polls from the start to finish of the debate series - and then it was the candidate widely judged to have lost the debates who gained in the polls...
What influence debates have had on public opinion historically has stemmed from matters of style rather than substance. A glance at a watch or a distant reaction to an emotionally charged question have been more consequential than clashes over war, taxes or economic policy.
A 2008 Gallup review of polling data surrounding presidential debates concluded the events are "rarely game- changers" yet may have made a difference in 1960 and 2000, both among the closest presidential contests in U.S. history.
Barack Obama is the country's first black president. As such, he is playing a game which is not designed for him. Given that these models of how debates impact voters have been based on white presidents, are they a good fit for assessing the relationship between Obama's debate performance and the vote choice on election day?
Despite what right-wing pundits would have you believe--that being black in America is a net advantage, or that the American people will have "pity" on Obama and give him a do over because of his skin color--serious people suggest that racism cost Obama about 5 percentage points in 2008's election.
Moreover, the politics of white racial resentment and overt racism have been repeatedly used by conservatives to subvert support for the country's first black president, and were the driving force between the white political insurgency known as the Tea Party.
Optics matter: there is a symbolic power to Obama as the country's first African-American Chief Executive that many white folks, especially on the Right, are repulsed by; the stated and unstated burdens of blackness, what Du Bois famously summed up with the question "how does it feel to be a problem?", are the background radiation which colors how many in the public perceive the President. He can't get angry. He can't show emotion. He can't talk about race. And he most certainly cannot remind anyone that he is black.
My instincts would suggest that the cultural politics, the white racial frame, and our country's long history of white racism, must in some way be impacting how members of the public assess his performance in the debates. However compelling, instincts are not a substitute for empirical rigor.
Teach me something if you would.
I do qualitative research. I can read the stats and explain it within reason, but don't ask me to run a simultaneous equation or do the matrix algebra for a regression by hand.
For those of you who are quantitatively trained social scientists, how would you go about working through the puzzle of comparing Obama's performance, and the public's perceptions of it, with that of his predecessors?
For the debates, I would imagine that modifying existing models of the relationship between debate performance and vote choice by simply inserting "dummy" variable where "0" is used for white presidents, and "1" is used for Obama, would be wholly insufficient.
Would you have to construct an index variable of some type that collapses together measures of white racial resentment from other surveys, and then include that into the model for Obama? If so, how would you maintain internal consistency in the model when such information would not be relevant in the same way for Obama's white predecessors?
Once the election occurs, and subsequently there is a full set of data available, i.e. we know who won and have metrics for the debates, would it be a matter of comparing like cases of debate performance (and other standard measures such as likability, the economy, approval polls) from the past that most closely resemble Obama's in the present? In essence, looking for differences in vote choice as the dependent variable?
Quantitative analysis is a great tool to have in the proverbial tool box. I just worry about how well a formal model would do in this case, where it would have to pick up all of the noise in the social ether that is directly tied to how racial animus, stereotypes, and the white racial frame impact the public's perceptions of the country's first black president and his performance.