You are on page 1of 1

Date: September 6, 2016

To:
Senate Majority PAC
From: Civis Analytics
Subject: Analysis from Florida Senate research
Summary
Civis Analytics surveyed 1,436 likely voters in Florida between August 9th and August 15th , 2016. The major
takeaways from this research are:

This is an extremely competitive race and our survey finds that its a statistical tie. Murphy leads Rubio
45% to 44% with 9% undecided. Civis modeling expects undecided voters in the survey to break toward
Murphy, and allocating these voters grows Murphys lead to 51.6 versus 48.4 for Rubio.
There is substantial opportunity to grow Murphys vote share with groups where we expect support to
coalesce after the primary and as voters become more familiar with the candidates nearing Election Day.
We suspect that there is not yet Democratic consolidation due to this survey being conducted before the
primary.
Rubios favorability is on a steady declinedropping 9 points in favorability since announcing his
candidacy.

The strength of the Murphy vote


Patrick Murphy is tied with Rubio despite running behind
Democratic benchmarks among several key groups
(Democrats, non-white voters and Millennials). Murphys
support is buoyed by running ahead of Democratic
benchmarks among older, whiter, and more conservative
voters. In short, Murphy has a lot of room for growth.
The weak Rubio vote
Rubio is running behind expectation among traditional parts
of the Republican coalition in the state. Weve seen drops in
his favorability over the summer, with a 5-point decrease in
favorability among Republicans over the last month.
Rubios support is also inflated by his temporary advantage
in name ID, as he is winning voters who havent heard of
Murphy by 2-to-1.
Conclusion
Rubio is already a well-known candidate in the state and is likely at his ceiling. In contrast, Murphy has room to
grow among several Democratic leaning groups that should be responsive to communications. He is well
positioned to win this seat in the fall.

Civis Analytics surveyed 1463 individuals in a sample representative of voters from August 09, 2016 to August 15, 2016 with a
landline/cell phone mix. Civis stratified the sample based on demographics and modeled variables including Age, Modeled
Partisanship, Race, and Sex to be representative of the full population of voters. Individuals were sampled with probability inversely
proportional to phone contact rates in order to ensure a representative sample. Civis weighted returns back to full population
estimates in order to further reduce bias. The Margin of Error (+/ 2.8) accounts for sampling error from a weighted survey, b ut does
not account for potential nonresponse, coverage and measurement error.

You might also like