Mei Preliminary 2018 Election Polls Reveal How Young Voters Plan To Vote In Senate and Governor Races

To pilot our new polling platform, we recently sent 10,000 short midterm election polls for Senate and Governor races ac­­­­­­­­ross the US to our Mei app users.  We hoped to demonstrate the potential for polling to share intelligence and opinions on a large scale through Mei.  Given the upcoming midterm elections, we thought this was a perfect opportunity to pilot our anonymous polling platform with our users.  Soon, Mei users will be able to conduct polls similar to this.

Polls were sent on various dates between October 26 and November 5.  Of the 10,000 polls sent, we received responses from nearly half (4,600), the majority of which were completed within 30 minutes of being sent. 

Many election sources are reporting an increase in early voting among young voters so far. Given our user base is significantly younger on average than the voting age population (26 years old vs. median age of voters in 2016 elections of 52), our polling results provide a unique perspective on how young voters can affect the outcome at many of the hotly contested elections across the country. 

Not surprisingly, our predominantly younger demographic tends to show a significant lean towards Democratic candidates.  However, how much young voters tend to lean varies by candidate.  The contrast is apparent in the Senate and Governor races in Texas, where young voters (age 18-29) have shown up in force in early voting and appear to heavily favor Democrat Beto O’Rourke over Republican Ted Cruz as Senator, by a factor of 2 to 1 according to our poll.  In the Governor’s race however, young voters don’t appear to show any meaningful preference between Republican Greg Abbott and Democrat Lupe Valdez. 

We summarize our results and compare them to the latest major publicly available poll conducted on October 28-30 by Emerson College, as reference.  In Texas, we received polling results from nearly 200 respondents, of which approximately 75% were in the age range of 18-29.

Our results contrast significantly from Emerson’s mainly because of age.  Whereas participants under the age of 45 made up 97% of our poll, respondents in that age range make up less than half in Emerson’s (as well as most other major polls).  So, despite our polling sample being smaller overall, our count of young voters is larger than most polls, giving us the potential to better understand this under-represented group (which we may further explore in later posts). 

The following table details the age mix for the Texas Senate, along with how each group favors each candidate.  For example, the 18-29 group in our poll favors O’Rourke over Cruz 1.9-to-1.

We published these preliminary results as an example of the intelligence and analysis made possible through our new polling platform.  In the coming days, we will be providing other election polling results, so visit our website for the latest updated polling information, including an interactive map.

Our users will soon be able to conduct polls similar to this with the credits they earn in Mei. 

METHODOLOGY

Our poll was conducted within our Android mobile app and therefore reflects a demographic distribution that correlates with overall smart phone usage.  Polls were sent in several clusters over the course of October 28 to November 5, in part to test our polling system.  In the future, these polls can be conducted at a single point in time.

Participants were sent respective state polls based on user inputs or according to the IP address of their phone network, which proved to be approximately 90% accurate.  Users were asked up to 5 questions depending on their responses and their state of residence.  The first two questions established whether they lived in the state and whether they were registered to vote.  A “no” response to either immediately ended the poll. 

Q1.  Do you live in [state]?

Q2.  Are you registered to vote?

We had user-disclosed ages for about 90% of respondents, many collected prior to the poll.  Because our app is able to make machine-learned predictions of age based on text messages, we used those predictions to estimate ages when they were not provided.  Questionable polls where users responded, but did not appear to be of voting age were minimal (3%) and excluded.  As a deterrent to false responses, users who opted into polling were informed that false responses may negatively affect their ability to earn rewards for future polls. 

Depending on the state, where there is or isn’t a senatorial or gubernatorial election, users were asked one or both of Q3/Q4 (Q4 twice where there is a special election) accordingly. 

Q3. Who do you plan to vote for as Governor?

Q4.  Who do you plan to vote for in the Senate?

Users had three options in each, the top two candidates and ‘undecided’.  We limited to the top 2 to minimize noise.  The candidate favored to win according to recent polls was usually the left option and the second candidate was the middle option.  The ‘undecided’ button was always on the right. to earn credits via polling in the future may be affected.