Skip to main content

A Poll Sample’s Party Composition


A note on party composition in polling samples.

Some commenters have noted that the Democratic advantage in the latest Monmouth University Poll (Aug. 8, 2016) is larger than in our poll taken just prior to the two parties’ conventions . Specifically, voters in the current poll self-identify their party leanings as 35% Democrat, 26% Republican, and 39% independent or other.  In the July poll it was 33% Democrat, 28% Republican, and 39% independent or other.

Contrary to some misperceptions – largely by those unhappy with the overall results of the latest poll – Monmouth did not “choose” the sample to look this way.  Party identification is a self-reported attitude based on where people see themselves fitting in the current political environment.

It is not the same as party registration or partisan voting behavior (e.g. consistently voting in one party’s primaries), which is a more stable metric. I wrote about these differences in more detail a few years ago (Party ID Apples and Oranges).  While the data in that analysis were drawn from New Jersey voter files and poll samples, the underlying message is the same.  Party self-identification can move with the political climate, while party registration is more stable.

Monmouth’s 2016 presidential polling uses a combination of voter lists and random digit dialing. The voter list includes data on voter registration and past primary voting.  According to this metric, 34% of the Monmouth sample are registered or active Democrats, 34% are Republicans, and 32% are independents or something else.

In other words, the Monmouth sample is evenly divided between Democrats and Republicans when it comes to registration and past voting behavior.  Yet when asked how they see themselves politically, these same voters are 9 points more likely to call themselves Democrats rather than Republicans.

The question you should be asking yourself, in light of events over the past few weeks, is why that might be so.

4 Replies to “A Poll Sample’s Party Composition”

  1. Larry says:

    Many years ago, if the phone rang, I picked it up and talked to whoever was on the other end. Now if I don't recognize the number of the caller, I send it straight to voicemail. I don't think I'm alone in this behavior. I wonder how THAT skews polling results?

  2. Anonymous says:

    Same here. We don't know anyone who answers the phone any more. We've been wondering the same thing.

  3. Aj H says:

    There are 250 million voters in the US. They only need about 1,000 of them to answer the phone and respond. It's fairly easy to do that although it probably does take longer than it used to, but nowadays they also have automated computer dialing, so that process is much faster than it used to be. When all is said and done, it's no more difficult now than it used to be to conduct a poll.

  4. Unknown says:

    But do polls reflect disaffected voters, such as Sanders or Cruz voters? The pollster will find 35% democrat respondents and 35% republican respondents-but how many were deemed "unlikely" voters? The questions should be: 1) are you dem, rep or ind 2) are you voting 3) if so, for whom. From that, we can better see a true measure of the electorate.

Comments are closed.