NYCU: Research Quality in Surveys
Please log-in using the log-in icon 
on the upper right of the web site or create an account to view this page.
on the upper right of the web site or create an account to view this page.
sampling
ARF Event 11/29/16 – Predicting Election 2016: What Worked, What Didn’t and the Implications for Marketing & Insights
The ARF partnered with GreenBook to assemble a forum of this highly charged topic. Ten industry experts were on hand, with ARF EVP Chris Bacon serving as moderator for the panel discussion. Here are excerpts from the event:
Gary Langer, Langer Research and formerly ABC News:
- There is no real “sample” of voters, only estimates of likely voters. This adds uncertainty, especially in predicting the electoral college vote
- Polling is not just about the horse race, as we use it to collect important information on what voters were thinking
Raghavan Mayur, TechnoMetrica Market Intelligence:
- You can never compensate for bad data
Cliff Young, Ipsos Public Affairs:
- We are in a political era of uncertainty worldwide. It is anti-establishment. This means we also need new methods (for polling)
- This was a “disruption” election, which globally accounts for only 15%. And in these elections, past behavior may not be a useful guide
Matthew Oczkowski, Cambridge Analytics:
- As with marketing research, it is all about finding the right consumer with the right message at the right time
- His role is to help clients win elections. Micro-targeting is a key component. He worked as a consultant for Trump team
Rick Bruner, Viant Inc:
- We need more random control trials (to improve polling). We also need more behavioral inputs
- Underlying values are important, like we do for marketing. For potential voters, this begins with “do I vote?”
Melanie Courtright, Research Now:
- Everything was different in 2016!
- We need samples representing the real population (of voters)
Jared Schreiber, Infoscout:
- The undecided and the indifferent voters matter a lot (swung to Trump)
Dr. Aaron Reid, Sentient:
- Traditional methods are not accurate enough
- Need to measuring the unconscious – people may have had conscious access to answer the researcher’s questions
Tom Anderson, Odin Text:
- Collected 3,000 Americans via Google Surveys one week before the election (inexpensive and predictive). Goal was positioning, finding out what candidates stand for via simple text
- Three issues: Non response bias; “Shy Trump” voter; voter identification
Taylor Schreiner, Tube Mogul:
- How will this (failure) impact CMO’s opinion of research
- We need more experimentation
- Let’s use this election as a teachable and learnable moment