A conversation with Prof. Ben Knoll on public opinion, post-convention bounce and voter IDs
Assistant Professor of Government Ben Knoll is one of several faculty experts who can speak to issues and subjects related to the upcoming Vice Presidential Debate at Centre College on Oct. 11. To see other faculty experts, click here.
Q: Is there really a post-convention bounce? How do public opinion polls generally change after the conventions? What have the public opinion polls looked like after the RNC and DNC this year?
Knoll: When we talk about a post-convention “bounce,” we’re referring to the general pattern that candidates tend to receive a small increase in their poll standings during their convention and for a few days after. Over the last fifty years, candidates have almost always gotten this bounce in the polls. The size of that bounce, however, has varied from election to election, anywhere from no bounce for John Kerry in 2004 to an 11 percent bounce for Bill Clinton in 1996. On average, the size of a post-convention bounce tends to be about four to five percent.
This year, the preliminary polling data seems to indicate that Romney received about a one to two percent bounce after the Tampa convention—which quickly faded—and that Obama enjoyed somewhere around a four to five percent bounce after the Charlotte convention—which is quickly fading. We should remember that they’re called “bounces” for a reason: the polling numbers usually return to their pre-convention levels within a week or two of the conventions. Take a look at the polling numbers a week from now and we’ll see if there is any more long-lasting effect that the conventions had on public opinion. It’s very difficult to tell right now so soon after the conventions have ended.
Q: What is the significance of the post-convention bounce to the overall general election?
Knoll: Post-convention bounces are significant to the final outcome if the “bounce” eventually turns into a “bump.” Most of the time, polling averages return to where they were before the conventions started (in which case there’s no lasting effect), but sometimes they turn into longer-lasting “bumps,” in which case they become more significant to the final outcome.
As far as being able to make predictions about the ultimate outcome of the election based on the size of the bump, history suggests that there’s not a very strong relationship between the two. Ultimately, post-convention bounces are useful in the sense that they tell us how the public responded to the conventions in the short-term, but I wouldn’t use the information to make definitive predictions about the eventual outcome of the election.
Q: Who gets polled in public opinion polls?
Knoll: If polls are done well—and many of them are—they should include a random sample of Americans over the age of 18. Anyone who is eligible to vote should be included in the sample. If not, the poll results won’t be an accurate measure of the attitudes of the entire population. Also, a good poll should include at least a thousand respondents so that the margin of error is sufficiently small—three percent or less—to make reasonable inferences.
At this point in the election, polling companies start limiting their sample to registered voters and/or “likely voters” so as to try to get a more accurate prediction of how the election will turn out. It’s important to look at who the sample population is, how many people were polled and the margin of error of the result before making any firm conclusions.
Q: How much stock can we put in polls?
Knoll: Assuming that the poll is conducted by a reputable, professional polling operation, polling results present a fairly reliable “snapshot” of public opinion at any given time. The important thing, though, is not to put too much stock into any single polling result. It’s better to look at several different polls and see where the average is. The more polls that are out there, the more reliable the “average” polling number is. There are some good poll aggregation websites out there, like pollster.com and realclearpolitics.com.
Q: When reporters show maps and color in different states based on electoral college predictions, how do they arrive at those predictions?
Knoll: Usually, news organizations base those predictions on polling done on registered or likely voters within each of those states. There are some political scientists and professional statisticians out there who combine that state-by-state polling data with economic conditions within each state and create a more fine-tuned projection. A good example of this can be found at www.fivethirtyeight.com.
Q: How many people are truly undecided so close to the actual voting date?
Knoll: In past elections there were usually many people still undecided two months before the election. In the last few elections, though, partisanship has begun to exert a much stronger influence on people’s voting choices than in previous decades. This year, nearly all Democrats had decided to vote for Obama by the middle of the summer, just as most Republicans had already decided at that point to vote for whoever the GOP nominee would be. At this point in the election this year, there’s only a small slice of the population—about five percent—that haven’t made up their minds yet.
Q: Do debates ultimately have an influence on public opinion?
Knoll: Debates definitely influence how the public views the two candidates in terms of their personal characteristics. Debates also serve to educate the voting public about the policy positions of the two candidates. Further, debates have an important effect on the campaign narrative—think “Joe the Plumber,” who was unexpectedly made famous after being mentioned in one of the 2008 debates.
However, presidential debates don’t often sway many voters to change their minds and switch from one candidate to another—although that does happen sometimes. Most viewers are watching to cheer for their preferred candidate rather than to be persuaded one way or another. So debates don’t often exert a slam-dunk, decisive effect on the ultimate outcome of the election, but they definitely influence how the public views the candidates and helps them cast a more well-informed vote, both of which are very important for our democracy.
Q: What do you think voter turnout will be for this election?
Knoll: Over the last sixty years, turnout in presidential elections has averaged between 50 and 60 percent, with the last two elections right around 55 percent of the voting-age population showing up to vote on Election Day. I see no reason so far to suspect that this year will be much different.
Q: What are your thoughts on the voter ID card issue—what it is all about, where it stands currently and how/if it could affect this election?
Knoll: There are two key questions with the voter identification laws recently enacted in Pennsylvania, South Carolina, Wisconsin, Tennessee, Texas, Rhode Island, etc. First: what are the political implications of these laws? Opponents—often Democrats—accuse supporters—often Republicans—of supporting these laws as a way of discouraging minorities and the poor—who tend to vote Democrat—from voting. Republicans respond that these laws are important to safeguard the integrity of the voting process. There’s some empirical support for the partisan argument, but its effect is small: Republicans perhaps get an advantage of up to one percent in the vote, but even the research is mixed on that. So Democrats can breathe a little easier—voter identification laws will very likely not sway the election one way or another. The bottom line is that the people who tend to be affected most by these laws—minorities and the poor—are the same types who tend to vote at lower levels anyway, even in states where the laws are much less restrictive.
The second question is more normative: is it “right” to pass a law that disproportionately affects minorities and the poor. On the flip side, is it unreasonable to show evidence that you are who you say you are when you go to vote? Respectable people have honest differences of opinion on these issues. Personally, I take Republicans at their word when they say that their motive is to prevent voter fraud, but I also sympathize with Democrats who have legitimate concerns about the effect it might have on historically disadvantaged populations.
It will be interesting to see how this all plays out. Just this last week a federal court struck down the voter identification law in Texas, but another court upheld the law in Pennsylvania. This all suggests that there will likely be more litigation and controversy over voter identification laws in the future.
To read more about the upcoming Vice Presidential Debate at Centre College on Oct. 11, visit Centre’s debate website here.