How close was the 2012 US Presidential election? Not as close as you we thought, apparently. The reason: Most polls and campaigns neglected social media. In doing so they missed a significant audience that was otherwise unreachable: The many young, Latino, African American and other voters who own cell phones but not land lines, or just prefer to talk online.
Light Rail in Canberra: Too much, too little, too late: Is the price worth th...
The social candidacy
1. THE SOCIAL CANDIDACY
Poll results, Facebook, and the 2012 Presidential campaign
Ian Lurie, CEO, Portent
www.portent.com
@portentint
ian@portent.com
2. Overview
With the election over, it’s very clear: Candidates, pollsters and
the media must factor in social media.
Political candidates have brands, just like companies do. They’re
influenced by many of the same market forces, and social media
is one of those forces.
And pollsters can make their models more accurate, sooner, if
they factor in social media.
The team at Portent put together this evaluation of the Romney
and Obama presidential campaigns based on Facebook data
collected since April 2012.
3. Our analysis
Two months ago, we analyzed Facebook and Twitter data. We
reached two conclusions:
1. The Obama campaign was overly cautious, behaving like a classic
big brand, and therefore letting the Romney campaign stay
closer than they otherwise might.
2. The Romney campaign suffered from a ‘universe problem’. While
he had a fiercely dedicated audience, he was unable to effectively
expand it.
You can read the full analysis in our blog post and slide deck:
portent.com/blog/internet-marketing/social-media-election2012.htm
4. Our hypothesis
Based on those conclusions, we hypothesized that President
Obama would:
1. Win the electoral vote by a large margin.
2. Win the popular vote by a small but measurable one.
3. Pollsters missed a huge pro-Obama audience who no longer
have land lines and don’t reply to calls.
5. Conclusion
Two key election players – pollsters and the GOP – suffer from a
universe problem.
The pollsters need to expand their analysis to include the social
media universe. If they don’t, they’ll be excluding a key part of
the electorate. Doing so will give them more accurate data.
The GOP must find a way to appeal to the social media
mainstream. This will give them the ability to access the
audience that cost them this election.
And, the Democratic Party should learn from this election:
Study the types of media and messages that best performed,
and refine their demographic targeting.
6. The outcome
President Obama appears to have won the electoral college
332 to 206.
He won the popular vote by 2.2-2.4%.
Compare that to the 2004 Bush/Kerry election:
Popular Vote Margin Electoral Vote Margin
2.40% 140
120
100
80
60
2.20%
40
20
0
Bush/Kerry Obama/Romney
Bush/Kerry Obama/Romney
7. The outcome
Historically, Gallup has correctly predicted presidential election
outcomes 9 of the last 10 elections. Their only miss was Carter’s
defeat of Ford in 1976.*
This year, they missed their prediction by 3-4%. Not much. But
it’s a big 3-4%.
CNN’s “Poll of Polls” had the national race too close to call
November 5th. They showed Obama ahead less than the margin
of error in Ohio.
Other ‘tossup’ states included Virginia, Colorado, Florida and
Nevada.
*They predicted a popular vote tie in 2004.
8. The outcome
Polling was unable to predict the outcome. We wouldn’t go so
far as to say they were wrong. But their data had the outcome
in doubt right up to election day.
Maybe it was just close. Maybe a lot of people made up their
minds at the last minute. Maybe the Obama campaign had a late
surge.
Or, maybe social media, combined with poll data, provided a
clear preview the entire time.
9. The outcome
Candidates and polling companies should supplement their
polling data:
1. With social media demographic data in crucial swing states,
where campaigns might be investing heavily in their
Facebook and other campaigns.
2. On a national level, to look at trends and clarify projections.
10. Swing states
Facebook data may help predict the outcome in crucial swing
states.
We looked at 4 swing states where the Huffington Post’s Poll of
Polls showed the outcome as a tossup: Florida, Colorado,
Virginia and North Carolina.
11. Swing states
According to data gathered by Walker Sands PR, Facebook
audience size by candidate correctly predicted the outcome in 3
of 4 states:
Facebook audience size by candidate and state
1,800,000
1,600,000
✔*
1,400,000
1,200,000
1,000,000
Obama
800,000
✔ Romney
600,000 ✔
400,000
200,000
0
Colorado Florida North Carolina Virginia
*Florida hasn’t been called, but Obama leads by 50,000 votes with 100% of precincts reporting.
12. Swing states
Looking across all swing states, Facebook audience size was
correct 80% of the time:
✔*
1,800,000
1,600,000
1,400,000
1,200,000
1,000,000 ✔
800,000
✔
600,000 ✔ ✔
400,000
✔
200,000 ✔
0
*Florida hasn’t been called, but Obama leads by 50,000 votes with 100% of precincts reporting.
13. Swing states
In non-swing states, Facebook did a far poorer job of predicting
the outcome: Audience size only correctly predicted the
outcome 27 of 41 times.
Why the difference?
Effort.
The campaigns likely focused their online efforts on swing
states, just as they did offline. In those states, their efforts
influenced audience sizes. In non-swing states, they didn’t
influence audience size, and lower participation meant a less
accurate predictor.
14. Social media trends
We updated our Facebook growth rate data to reflect the last
month of the election.
Each debate had a short-lived effect. The trend remained the
same: After the Democractic National Convention, President
Obama maintained or increased his fan growth rate; After the
Republican convention, Governor Romney saw an inconsistent
growth rate. In fact, he lost momentum after the convention.
15. Social media trends
President Obama's Facebook growth rate: Rolling avg
0.250%
0.200%
0.150%
0.100%
Pre-convention growth rate
0.050%
0.000%
16. Social media trends
Governor Romney's Facebook growth rate: Rolling avg
4.000%
3.500%
3.000%
2.500%
2.000% Average pre-convention growth rate
1.500%
1.000%
0.500%
0.000%
17. Social media trends
President Obama started with a massive advantage on
Facebook: He had over 25 million fans in April. By November 1,
he had over 31 million. His campaign showed all the attributes of
a big brand: Huge followership, and an over-abundance of
caution. This came through in his first debate performance. He
needed to go on the attack and take risks to secure the
campaign, and did so.
Governor Romney started with 1.6 million fans in April. He did a
fantastic job growing his Facebook base thru August, reaching
almost 8 million. But after that, his growth rate plateaued and
began to fall. The Romney campaign was a classic small brand
with a limited universe. They were able to grow within that
universe, but couldn’t get beyond it.
18. Demographic gap
Facebook reflects the thoughts of a demographic that polling
companies can’t access. Facebook users belong to the
demographic group most likely to have a cell phone as their
only telephone.
Pollsters couldn’t contact that group, so they missed
President Obama’s winning margin. (From Democracy Corps).
Facebook taps that audience. And freely-available, anonymous
data that Facebook offers via their API provides a fantastic way
to observe and measure their voting intent.
19. Conclusion: Polling
If polling companies want to remain accurate, they need to add
social media data to their models. At a minimum, they should
use:
1. Facebook fan growth rate;
2. Facebook comment sentiment;
3. Facebook audience response – comments per post, likes per
post;
4. Twitter audience growth rate;
5. Facebook and Twitter audience quality.
20. Conclusion: Campaigns
Campaigns must have a strong ‘ground game’ in social media.
Like pollsters, they can access an audience that’s mobile and
hard to reach. This audience is only going to grow in the next
four years.
To succeed, they will need to:
1. Monitor social sentiment.
2. Post, measure and refine their content strategy continuously.
3. Invest heavily in video for social distribution.
4. Learn to use demographic targeting, if they aren’t already.
5. Use tools like re-targeting to segment between supporters
and potential supporters.
21. Conclusion: The social candidacy
We’ve entered the time of the social candidacy. Internet
marketing and social media provide access to a mobile, young,
diverse audience that can swing elections.
Campaigns and polling organizations take note: The 2012
election was a sea change in campaign media channels. Adjust
strategies accordingly.
22. Q u e s t i o n s?
ian@portent.com
@portentint
http://plus.google.com/+IanLurie
www.portent.com