The Living Consequences

There are polls this week suggesting that McCain may be gaining a bit of ground nationally, and even that the presidential race may be fairly close. This is true, but also misleading, at least in part.

First, the national polls have tightened slightly, with McCain seeming to make inroads with independent voters. However, it’s worth noting that this effect is quite minor compared to the natural “noise” of fluctuating polls. While this tightening may be genuine, it may be merely an artifact of recent polling.

The more important issue is how Obama and McCain are doing at the state-by-state level, particularly in battleground states, and not whether McCain may be winning over voters in California, Texas, the northeast or the traditional south.

In the battleground states, the polls show no tightening, at least in favor of McCain. In the last few days, several battleground states have shifted towards Obama, or have seen polls which confirm that an Obama advantage is holding. These include Nevada and Ohio, toss-up states which are now leaning Obama; Montana, a red state which is now a toss-up; and Arizona, which is now only leaning towards McCain, its senior senator.

There is another issue, however, which some of the polls have put on the table again. This is the question of whether all of these polls can be trusted. What many in the public do not appreciate is that political polls are not simply the result of surveying voters and then tallying the responses. Pollsters make a host of critical choices in their poll designs, involving such issues as how to choose respondents and how to weight their responses to correct, rightly or wrongly, for various factors perceive to affect the results.

This reality explains why even the national polls, for instance, show such a wide variation in support for the candidates (with Obama’s lead currently anywhere from 2% to 15%), and why the McCain campaign’s internal polling shows Obama doing nowhere near as well as the independent polls do.

Each of these polls makes assumptions about the turnout among various demographic groups, such as black voters and younger voters. Both of these groups historically have turned out in relatively low numbers, but each is expected to be more motivated to go to the polls this year. Just how many of these voters, many of whom have registered for the first time in recent months, will turn out on Tuesday? Which of the campaigns will do a better job overall at turning out their base, and at converting independent support into voters who show up at the polling place?

The design of each poll also considers what assumptions to make about likely voters who can’t be reached (who use only cell phones, for instance), or who refuse to respond to questions (as many who are surveyed do). Might voters be more reluctant this year to indicate support for the Republican candidate, for instance, because the Democratic candidate is perceived as highly popular or because he is black?

To give a concrete example, the Gallup daily tracking poll currently shows Obama with a seven-point lead using a model which assumes a higher turnout of first-time voters. That lead shrinks to two points, however, using a model which provides for turnout similar to past elections.

Most political scientists and professional polling experts doubt that the polls this year will be wildly off from reality, given the attention paid to these questions in designing and analyzing polls. This conclusion is enhanced by the fact that so many polls show a consistent Obama lead, despite reflecting different decisions about the issues above.

However, these same experts agree that this year poses unique challenges, in part because Barack Obama is the first black candidate to run as a major-party nominee and because he draws his support so heavily from newly-registered voters in groups which traditionally fail to turn out to vote.

Worse yet, consistent results across different polls mean nothing if those polls are grounded on similar assumptions which turn out to be wrong. Likewise, similar results across a number of battleground states could prove to be wrong across the board if the polling assumptions used in each state are similar and prove to be incorrect.

None of this is to suggest that the current Obama lead is likely to evaporate. But there is ample reason to be cautious. If, for instance, the voters who turn out to cast ballots on Tuesday turn out to look very much like the voters we saw four years ago, we can be certain that the national polls will prove to have been quite misleading, and the state-by-state polls we’re now relying on will be systematically wrong.

In other words, a commanding Obama victory is still the most likely possibility. But there’s plenty of reason not to assume anything, and to watch the results carefully on election day.