[Update: A commenter points out that the 16% difference actually amounts to 50% more Democrats than Republicans interviewed in the referenced poll]
Clinton leads Trump by 8-points– in poll that interviewed 16% more Democrats
The following is a media news analysis
Because this feature is proving popular, here’s another installment of my look at the media’s reporting on presidential polling. This article reports “Clinton leads Trump by eight points” in the latest online Reuters/lpsos poll.
Like the other polls I’ve recently highlighted, the Reuters/lpsos poll appears to have interviewed significantly more Democrats than Republicans. This is not disclosed in the news articles.
According to data on p. 14 of the poll, 46% of the respondents identify as Democrats (strong, moderate or leaning) and 30% identify as Republicans (strong, moderate or leaning).
Assuming 16% more Democrats than Republicans turn out to vote in the November general election, the results may correlate with the poll. But what if an equal number from each party turns out? What if more Republicans than Democrats vote? Obviously, that could change the results, the same way this poll would likely read differently had it interviewed 16% more Republicans. There’s no indication that the poll has adjusted for this factor (though it has adjusted for factors such as gender, age and education).
[quote]Clinton leads Trump by 8-points in poll that interviewed 16% more Democrats[/quote]
Further, what effect might it have for the voting population to hear one poll after another declare a candidate far ahead–without the disclosure that the sample was heavy on one party? Do the polls and reporting risk influencing the electorate by providing an incomplete picture, rather than accurately measuring the trends?
As continuing evidence that it’s possible to make poll results look more or less favorable to a candidate, this poll was universally portrayed as favorable to Hillary Clinton. But Donald Trump actually receives more of the independent vote among both likely and registered voters. That could be significant. Assuming an equal number of Democrats and Republicans were to turn out to vote, it’s the “independents” who could matter most. But this finding went unreported in news articles.
There’s no perfect way to conduct a poll, and no way to know ahead of time what the makeup of the voting public will be on election day. And I still believe trends in polls typically prove accurate over time. But it seems important, for context, to disclose when significantly more people from one party are interviewed.