Polls: What They Should Tell You (But Usually Don’t)

By February 4, 2016Uncategorized

By Hovannes Abramyan, Data Scientist, TPC

During the run up to a presidential election, you’ll likely hear about the release of a new poll every day. Typically, some organization in conjunction with some other organization poll a sample of likely voters on their current pick, be it for the primaries or the general election, and these numbers are blasted out for media outlets to report and discuss. But lost in the discussion is an understanding of what the polls are (and aren’t) telling us about the dynamics of the contest. Here’s a look at the most common polling approach, its relative value and deficiencies and an explanation of why TargetPoint Consulting’s (TPC) different approach – our Presidential Candidate Consideration (PCC) study – has so much value.

Current Hypothetical Vote Choice

The most commonly reported results are aggregated hypothetical vote choices. Here’s an example from a recent Washington Post-ABC News poll: “If the primary or caucus in your state were today, for whom would you vote?” Sometimes survey respondents are also asked to choose in a hypothetical head-to-head matchup (e.g., between candidate pairs from opposing parties to simulate a general election). Percentages based on these responses form the basis for much of the commentary regarding current political dynamics and election forecasting.

The problem is, five months – or even one week – before a general election or primary, these numbers may show little resemblance to the true dynamics of what is going on inside voters’ heads. They can only hope to reflect voter preferences during the narrow time period the poll was conducted. This is especially problematic in primary elections, where generally lower levels of attention and the absence of a party cue combine to create an electorate with many undecided voters, continuous vote shifting and large-scale instability in responses. The numbers have value, but are not as helpful as they appear; they don’t disclose how firm supposed preferences are, or whether they’re likely to stay intact on election day. Admittedly, these issues lessen the closer you get to the election, but by then you’ve already endured months of (probably) volatile polls.

There is value in gleaning early inclinations and watching for trends – donors need to know if their investments are having a measurable impact, and candidates will want to know if their methods and messages are working – but it’s unclear that directly asking survey respondents to make an early choice in a hypothetical election is the best way to go about it.

Candidate Consideration and Derived Ballots

TPC’s PCC study takes a different approach by polling candidate considerations, and using these survey responses to generate a “derived ballot.” Just last week, we published results from our most recent candidate consideration poll – an in-depth look at the considerations and preferences of likely Republican primary voters.

Rather than use a forced-choice ballot that asked which candidate each respondent preferred, we asked likely voters to rate how strongly they are considering each candidate in the race on a 7-point scale. From this set of ratings, we can determine the share of respondents who have a clear preference, those who are still deciding between a couple of candidates, and those who remain undecided. These numbers are leveraged to form a derived ballot, which gives a more complete overview of the state of preferences than does posing a question about a hypothetical election held today.

Now on our fifth GOP nomination consideration poll, we have been able to show at several points during the campaign which candidates have led, which have had potential to grow in support and which have largely been dismissed by the primary electorate. In addition, our ability to run statistical analyses on consideration reports means we can determine what types of voters are attracted to each candidate, and where support is likely to shift if any one candidate drops out of the race. These are important insights for media commentators and campaigns, alike.

Our consideration polls provide a reflection of current political dynamics, but they are much more forward-looking and useful than basic measures of preference.

Leave a Reply


Contact Us

WeWork C/O TargetPoint Consulting
1201 Wilson Boulevard
Arlington, VA 22209

P – 703.535.8505

Sign up for email updates