Wednesday, August 27, 2008

Gaines' perspective

Editor’s note: This week’s Democratic National Convention in Denver and next week’s Republican National Convention in Minneapolis-St. Paul are perfect reasons to stop and gain some perspective. Below is a conversation between Brian Gaines, a political science professor, and Craig Chamberlain, the University of Illinois news bureau social sciences editor. Gaines, a frequent contributor to our Illinois Issues magazine, has appointments in the University of Illinois at Urbana-Champaign’s political science department and the Institute of Government and Public Affairs. He also is a research fellow at the Cline Center for Democracy. Here is an edited version of that conversation:

Questions by Craig Chamberlain of the University of Illinois news bureau:

Many commentators in recent weeks have questioned why Barack Obama is not far ahead of John McCain in the polls, given that issues and circumstances are said to dramatically favor the Democrats this year. Do they have a point or are they missing something? Do you expect a close election in November?

I do. I’ve never bought the argument that this year should be an easy win for the Democrat. The country remains pretty evenly divided, and unpopular incumbents don’t always cast long shadows. John McCain is widely viewed as an independent-minded maverick, so he might be the ideal Republican candidate to withstand a slump in his party’s popularity. Also, retrospective unhappiness with the situation in Iraq doesn’t map neatly into an Obama vote. A lot of voters who wish the U.S. had never invaded still haven’t decided whether Obama or McCain is better qualified to make decisions about Iraq policy in the future.

After every convention, there’s talk about whether the candidate got the expected post-convention “bounce” in the polls. Does this have any significance?

The bounces are usually about the same size, so they’re irrelevant by mid-September. Political scientists are not even sure why they occur. It could be that the undecided get swept up in the enthusiasm of well-scripted love-fests. But it might also be the case that strong Republicans tune out the news (and refuse to talk to pollsters) while the headlines are full of Democratic convention news, and vice versa.

The news coverage of campaigns focuses constant and daily attention on polls, and most often national polls. How and in what ways is this misleading? What should we understand about polls and don’t?

I don’t think the margin of error is well understood. If a poll finds 47 percent saying they plan to vote for Obama, and it is described as accurate plus or minus 3 percent, that is suggesting a level of support anywhere in the range of 44 to 50 percent. And while it is probably true that the actual value (known only by God, not by any pollster) is in that interval, that calculation will fail, just by bad luck, about five out of every 100 times. Moreover, most pollsters do not calculate these intervals quite right. Polls usually need to be weighted to reflect demographic differences — such as in age, gender, or race —between the respondents and the target population.

Such weighting makes the margins of error larger, and so many polls report incorrect margins. There’s even more bad news: Response rates to polls have fallen over the years, and it seems likely that people who are willing to talk to pollsters, or even computers doing “robo-calls,” are different from those who won’t be polled, and in ways much harder to measure than age, gender, or race. The reported margin-of-error assumes that this is not true.

The extremely tight 2000 election, and resulting dispute over the Florida recount, raised some uncomfortable questions about the U.S. voting system. Have we adequately addressed those concerns? Are there other potential issues or controversies waiting in the wings in the event of another close contest?

Unfortunately, there’s no such thing as a foolproof electoral system. Blunders and fraud can creep into many different stages, from ballot design, to eligibility screening, to tabulation. Recounts often reveal serious problems. New Mexico’s handling of the 2000 presidential election was a shambles, but the state was spared scrutiny because all eyes were on Florida. Washington state had an orderly, uncontroversial recount in its U.S. Senate race that year. The secretary of state crowed that his state managed recounts properly, so watching them was “like watching grass grow.” Four years later, his successor oversaw a tumultuous triple recount in which new, previously overlooked ballots emerged late in the process, reversing the outcome. I’ll hope for a controversy-free election, but if it is as close as I expect, there will probably be serious problems somewhere. Personally, I worry about the huge growth of absentee voting. Hardly anyone ever points out that absentee ballots defy modern practice by not being secret. Secret ballots emerged in the 19th century as the main device to prevent vote buying and intimidation of voters. We’ve quietly rolled back that reform in the interest of boosting turnout, on the assumption that decentralized, non-secret ballots are secure. I’m not confident that’s right, and I expect a blowup over systematic abuse of absentee ballots by some campaign one of these days.

No comments: