Monday, May 05, 2008

The Early Voting Ethical Dilemma For Pollsters

In the past couple weeks we've based a series of posts, in part, on reports from various polls of the results from so-called "early voting" in the North Carolina Democratic primary election.

Early voting is the process, now allowed in a majority of states (but not the Commonwealth of Virginny), by which voters can cast their ballots either by mail, or at designated polling places, up to a month before "election day." In essence, election day becomes the last day one can vote. It is a convenient and popular innovation in those states that have adopted it.

A number of the polls we've seen in advance of various primary elections this season have included a separate tabulation of those in the poll sample who have already voted. In most states, that percentage has been pretty low, but in North Carolina it has been as high as nearly 30 percent.

That's got us to wondering whether there isn't an ethical issue here for pollsters and the media. A poll of early voters is no different than an exit poll. Instead of sampling voters as to who they're LIKELY to vote for, the pollster is asking them who they DID vote for.

It is common practice NOT to report the results of exit polling before the election polls are closed. The reason is fear that those reports will have an impact on voters who haven't yet cast their ballots, either causing them to stay home or perhaps even to change their minds and go with whoever looks to be the winner.

With the November general election approaching and early voting being much more widespread and popular than just four years ago, pollsters and the media need to make a decision about the ethics of reporting their surveys of early voters.

Suppose, for example, we find out one week before the November general election that McCain is faring much worse (or better--take your pick, but we like worse) in early voting in a key swing state than was expected. [Technically, we wouldn't know for sure--the states won't report results before election day. But a pollster with a decent sized representative sample could get a pretty good snapshot.] Should those results be reported?

We could make a decent argument either way. On the one hand, that information would certainly be newsworthy. You can also bet that the campaigns will be doing their own polling to garner the same data, so they'll know--and be acting on that knowledge.

On the other hand, just as with exit polling, you're now reporting, before the official results are out, on how people actually voted. Such reports are bound to have an impact on the last few days of the campaign.

The one difference we see between a true exit poll and a poll of early voters is that we have no way of knowing whether early voters are representative of all voters. Most likely, they're NOT representative, so knowing how they're voting won't tell us how the vote will turn out. In time, however, pollsters will get pretty good at adjusting their samples to overcome that limitation.

A related issue has to do with the data released by state election boards. Should they say how many voters have utilized early voting? North Carolina released results each day during its early voting process, as have some other states. More significantly, NC also released some demographic data on those who'd already cast ballots, including race and gender, from which some broad generalizations could be made even without an early "exit" poll.

We think it's an interesting issue. The debate should start now--not AFTER the November election.

No comments: