The presidential campaign is causing dismay down ballot. When WikiLeaks unloads another damaging batch of Hillary Clinton’s emails, or a tape of Donald Trump’s “locker room talk” gets released, congressional, gubernatorial, and state legislative candidates collectively cringe.
It’s an understandable reaction, wondering if coattails, positive or negative, are going to filter down from the top to impact turnout or sentiment in a given state or district.
Good consultants got to where they are because of solid gut instincts. But the wild swings of this cycle require more than educated guesses.
If you’re not polling, and polling regularly, you’re at the whim of conventional thought, or chattering-class assumptions. Both are dangerous places to be considering the unconventional presidential election, candidates, and coverage.
Create Your Own Polling Average
The New York Times recently gave four pollsters the same raw data of a national survey. Two university polling operations, one Republican polling firm, and one Democrat polling firm all looked at the Upshot/NYT/Siena College sample of of 867 likely Florida voters. The results ranged from Trump up by 1 point to Clinton up by 4. Same sample. Same questions. Different polling methodologies.
Why does this matter to your state senate or state treasurer race? Because this is a single poll interpreted five different ways. If you’re making all your strategic campaign decisions from a single poll, you’re building your campaign table on a single leg.
Always look at multiple polls over time as you close in on Election Day. If you’re lucky, there will be public polling to build your own polling average, like 538 does for the presidential race. Otherwise, work with your current pollster to conduct affordable brush fire surveys, or find a specialized polling firm that focuses on low-cost (but not IVR-only) surveys. You should have at least three polls in the final stretch of the campaign.
Test Hypotheses
In these surveys, get creative about testing hypotheses. We have been inundated lately with Republican clients that want to see what effect Trump is having on them. They’re testing messages that look at the electorate’s view of a candidate who is Republican, but publicly denounces the party nominee. Or they’re looking at how a voter’s opinion of Trump or Clinton affects how he or she views other candidates of the same party.
Increasing the number of questions does up the cost of a survey, but going from anecdotal assumption to empirical answer is well worth a couple hundred dollars. Questions like those mentioned above help fill in the holes as to why your numbers are moving as the presidential race shifts.
Pay Attention to the Details
As the aforementioned Times article proves, how your polls are conducted matters — a lot. No pollster can perfectly project turnout, but some have it pegged better than others.
This election is not anything like 2012 or 2008. It’s even different compared to 2000, the last race with an heir-apparent nominee (Vice President Gore then, Clinton now). Voter trends are changing. Ohio is not the bastion of purple it used to be. The list goes on and on.
So as you build your polling average, closely watch these factors:
- Sample Approach: RDD (random digit dial), RBS (registration-based sampling), or Modified-RBS
- Collection Methodology: Live, Web, IVR, or Hybrid (IVR to landlines, live to cells)
- Cell Phone percentage: How does it compare to the district census and voter file?
- Voter History: Comparison to past turnouts and individual voter participation frequencies
- Demographic: Race, Age, Gender, and Geographical distributions
Hopefully the mere fact of looking at your movement between polls will smooth out the slight variations in polling methodology. However, you can make your average more accurate with consistent approaches to your surveys.
Prep for Future Polling Approaches
At one point in the past, pollsters were anguished about moving to telephone polling. How could it be accurate compared to tried-and-true door-to-door survey collection?
Now we find ourselves at the juncture of low telephone answer rates and high access to online survey tools. Both have their faults. One is past its peak of maturity and the other in infancy.
Start testing new polling methods now. The complete craziness of the Trump-Clinton races gives you the perfect opportunity to “compare notes” between survey collection methodologies. Don’t limit yourself to phones and Google Consumer Surveys.
Think text messages, emails, Facebook, in-app push notifications, old-school canvassing, Google search trends, et cetera. Patrick Ruffini recently sat down with the Harvard Business Review to discuss the future of polling, and he makes some great points about where we are and where we are most likely going. Read the article if you haven’t yet.
We recommend sticking with accurately conducted phone surveys as your baseline, at least for the next four weeks. But do step out there as see what other survey methodologies tell you in comparison.
At some point in the near future, one of those ways you test will be the main polling technique or part of an important data acquisition blend. So you better start getting comfortable with it now.
The bottom line is that you shouldn’t just rely on your gut. If you haven’t polled in the last two weeks, get in the field now and book at least two more surveys for the coming weeks with your pollster.
Use the sheer “pandalerium,” as Jeff Foxworthy once said of a trailer park tornado victim describing the event, of 2016 to test some alternatives to phone polls.
And whatever you do: don’t put anything confidential in emails and always check your mic before speaking behind closed doors.
Brent Buchanan is a managing partner at Cygnal, a GOP communication, digital, and data/research firm.