The Art and Science of Polling

Election Context and Primary Dynamics

• Democratic primary polls indicate a majority of voters find the ongoing contest energizing rather than harmful to party unity.
• Hillary Clinton holds an insurmountable lead over Bernie Sanders in pledged delegates prior to counting superdelegates.
• Republican primary polls reveal high voter engagement driven by fear and horror, creating a more divisive environment compared to the Democratic side.
• Polling suggests the Republican debate over issues like border walls and religious bans is failing to resonate with establishment Republicans who wish to avoid such topics.

Role of Polling in Campaign Strategy

• Polling in multi-candidate primaries requires non-traditional data and specific questioning to distinguish between candidates within the same ideological lane (e.g., establishment vs. insurgent).
• Donald Trump's explicit focus on his robust polling numbers is an effective strategy that leverages the "viability effect," where voters prefer candidates with a higher perceived probability of winning.
• SurveyMonkey's daily polling data showed Trump gaining 11 points in the two weeks leading up to the panel, reaching 56% support, breaking previous perceived ceilings.
• Campaigns utilize polling to test specific messages and scenarios to adapt strategy, rather than solely to predict election outcomes, which represents only 5% of their polling goals.

Impact of Media and "Black Swan" Events

• Trump's rise was predicted earlier by online metrics (Google search volume and social media share of voice) than by traditional polling, which initially struggled to model such a non-traditional candidate.
• The correlation between Trump's news coverage and polling numbers is extremely high, exceeding a 0.9 correlation coefficient.
• Traditional predictive models relying on historical data failed to anticipate Trump because he represented a "black swan" event with no prior analogue.
• Media outlets often accelerate the impact of candidate statements by testing policy positions (e.g., banning Muslims) that would not have otherwise been included in standard polling menus.

Crisis vs. Revolution in Polling Methodology

• Industry leaders describe the current state as a "crisis" due to the collapse of the 40-year consensus on high-quality survey standards, now rendered a luxury item due to rising costs.
• Michael Demick characterizes the shift as a "revolution," noting increased transparency and the ability to aggregate diverse data sources to better capture public voice.
• Phone response rates have declined significantly, though historical data suggests this has not yet drastically impacted result accuracy.
• Online polling volume has increased by 250% since the previous cycle for the Republican primary, while phone polling numbers remain flat.

Data Science and Modeling Advances

• Modern modeling has shifted from human-determined likely voter screens to machine-driven models using historical voter file data to predict turnout probabilities.
• Campaigns are integrating unstructured data (e.g., social media sentiment analysis) alongside traditional polling to identify individuals and target communications.
• Aggregators face a "garbage in, garbage out" risk when mixing high-quality and low-quality polls, potentially compromising the reliability of consensus numbers.
• The "art" of polling remains essential for determining priors, framing the right questions, and understanding the "why" behind voter preferences through qualitative methods.

Public Trust and Media Interaction

• The panel rejects the feasibility of banning polls before elections, arguing it would leave the public with unverified punditry rather than data.
• Media pressure for immediate, daily results creates a "fever" environment that leads to over-interpretation of minor statistical fluctuations.
• Qualitative focus groups reveal that voters often lack detailed knowledge of political mechanics (e.g., Senate balance), challenging assumptions held by insiders and pundits.
• Polling serves as a critical democratic check to validate a winner's mandate against the broader, non-voting body politic.

Specific Case Study: Jeb Bush Campaign

• Early polling for the Jeb Bush campaign indicated that Trump was not a viable threat based on historical metrics, failing to capture the organic momentum shown in digital engagement.
• By October 2015, traditional polling finally caught up, confirming Trump's status as the frontrunner and highlighting the campaign's loss of viability.
• Campaigns historically rely on logic-based data and evidence, which proved insufficient against a candidate operating primarily on emotional resonance and visual media.