January 20, 2005


Study Cites Human Failings in Election Day Poll System: A study concluded that the polling system erroneously showed John Kerry to be leading the presidential race
because of more human variables. (JACQUES STEINBERG, 1/20/05, NY Times)

The new system was constructed after a precursor, the Voter News Service, contributed to the networks' calling of Florida in the 2000 election first for Al Gore, then for George W. Bush, then for neither.

While the new system avoided those pitfalls, the researchers found that their projections overstated the votes won by Mr. Kerry in 26 states, while overstating the votes won by Mr. Bush in four.

The system was not without its technical glitches on Election Day. For example, the research firms found that at least in the early afternoon, the system "overstated the proportion of women in the electorate," a problem apparently caused by a programming error.

In positing why the overstatements of Mr. Kerry's performance were so pronounced, the researchers said they were convinced that the technical foundation on which their work was based was sound.

Instead, the report concluded that at least some of the breakdown was rooted in the "interactions" between some pollsters and some voters. In general, the surveys appeared to overrepresent younger voters, who tended to vote for Mr. Kerry, and to underrepresent older voters, who tended to vote for Mr. Bush.

In analyzing the results, the researchers were careful to indicate that there was no evidence that the surveyors had embarked on any conscious effort to skew the vote.

The researchers focused instead on the median age of the surveyors, 34, and they hypothesized that perhaps younger voters felt more comfortable than older voters submitting questionnaires to younger surveyors.


Posted by Orrin Judd at January 20, 2005 7:38 AM

Bu[nk]. They lied. They tried to fix the election the way they tried to fix the 2000 election in Florida.

Posted by: Lou Gots at January 20, 2005 7:43 AM

It would be interesting if the polling company would offer up the names of their "young surveryors" so that anyone wanting to could run a Google search on those names to see if any affiliations with other political organizations pop up. But since this is a CYA report those names will be guarded by the company about as tightly as the reviewing stand at the U.S. Capitol will be guarded today.

Posted by: John at January 20, 2005 8:12 AM

Even in today's media controlled world, wishing still doesn't make it so. Give them a bit more time though, and they may do it yet.

Posted by: erp at January 20, 2005 8:41 AM

There is a preliminary analysis of the report at Daily Thoughts. The hypothesis that "... perhaps younger voters felt more comfortable than older voters submitting questionnaires to younger surveyors" is contradicted by their own data. There is a table on p.37 of the report showing completion, refusal, and miss rates, and there is no significant difference with respect to precinct partisanship. Nor were the problems due to bad precinct selection -- the selected precincts accurately account for the results when the precinct vote totals are used instead of the survey results.

It seems more likely that the problem was due to poor recruitment and training of canvasers, improper (i.e., non-random) selection of interviewees, and unconscious or intentional misreporting.

Posted by: jd watson at January 20, 2005 11:48 AM

Interesting how they waited until after there was any possibility of an effective challenge in Ohio before releasing the study.

Posted by: Bret at January 20, 2005 12:13 PM

Face it. Polling is GIGO. And the sooner it is dicredited and dropped as a subject of news reporting, the better.

Posted by: Robert Schwartz at January 20, 2005 3:14 PM