Friday, August 3, 2012

The Twit-Dex: The Twitter Political Index

Who looks more smug ... in THIS picture it's suddenly somehow a tossup!
What you are looking at is something new. What you are looking at is something game changing. This is the Twitter Political Index. In 2008, on the day of the election, Twitter got 1.8 million tweets. Today? That happens in six minutes. Twitter describes its full feed as a firehose of data. It's understating itself: Tsunami? Perfect Storm? All You Can Eat Buffet?

Whatever you call it, Twitter is big data at some of its biggest--and now they're mining it. The Twitter Political Index has experts from two polling outfits (Mellman Group and North Star Opinion Research--a left and right set of firms) using their best-thinking algorithms to analyze the sentiment of each tweet. In other words: are the 400million people who tweet per day saying good things or bad things about our candidates? And then they give that a score. At 50% it would be even--one good per bad. Above that? More good. Below that? More bad.

This is the wisdom of crowds with the biggest crowd on the planet.

So What?
Take a look at this:
You eat one foot and they call you a cannibal for the rest of your life!

What you are looking at is Rick Perry's InTrade "Flash Crash." In case you don't know what those words mean, it's this: is a website where you can go to "bet" on real-world events like Rick Perry being elected president or the US finding and killing Osama Bin Laden. For the Perry-Is-President one, people had "bought shares" (bet on it happening) and then, during his disastrous "oops" debate moment, in real-time, those people were frantically hitting their "sell" buttons. The minute-by-minute track at the bottom shows where all those people were realizing they better get out fast to recoup what they could of their 'investment.'

If you knew all of this, consider yourself pretty well politically informed. There's a chance, however, that you didn't. But you did know what Twitter is. And that's true, even if if you're my mother. Everyone knows what Twitter is.

The ability of the InTrade market to react in "real time" to an event (Perry screwing up catastrophically) was kinda 'new.' The ability for Twitter to react "in real time" (the index is updated at 8 AM daily) is unprecedented because it, simply, put, has so much more data. It really is a world-wide / nation-wide poll.

So the first question you should be asking is: So What: Is It Meaningful? Who cares if Romney is getting 10% less positive tweets than Obama.

So What? Is It Meaningful?
6.1.12 Twitter Index Compared to Gallup
I Conclude From This That People Tweet The Way They Do BECAUSE of Gallup. It's The ONLY Explanation!

What you are looking at is the comparison of Gallup to the Twitter Index. As you can see: it tracks. How good is Gallup polling? Decide for yourself (they are within 2% of election results for most cases). If what Twitter is saying matches what Gallup is saying and what Gallup is saying matches what the election results are saying then: yes, it's meaningful ... and it's a lot closer to real-time.

So What's The Big Question?
If we can see daily (and, if you have access to the data, probably more than that) meaningful fluctuations the question will be: what are those people responding to? What caused that shift in the data. The deal is this: it isn't necessarily something that happened on that day. Events take time to move into the public consciousness. Look at May 11th on the first chart--that's the biggest gap. What was going on then? An examination of Google News suggests one thing as a top story in that range: Romney's Bullying Incident. The article linked, from Forbes, is Romney-friendly but says this:
The tone of Romney’s reaction today does not look good on the empathy front. Referring to an assault on a classmate as “hijinks and pranks” is pretty tone-deaf. Let’s say you were told about an incident in your teenage years that you had forgotten, where you behaved cruelly and caused a lot of distress to other people. Wouldn’t you, as Dan Foster describes, feel a little bit ashamed? That’s not at all evident in Romney’s reaction to this story.
I think this is meaningful--and interesting. I assumed the bullying stuff was fairly meaningless--but what if that's not the case. The story was in the news then--breaking (the article is from the 11th--but it was broken before that). If that, indeed, was what the biggest gap was in reference to, maybe that did negatively influence people. Maybe it had some staying power. Consider this article from yesterday (headline: For Undecided Voters, It's The Devil-You-Know Election):
And it's clear the Obama campaign's efforts to define their opponent are working. Several women brought up Romney's time at Bain Capital; an incident in which he forcibly cut a classmate's hair while in school; Seamus, Romney's dog; and Romney's taxes.
Could that Twitter-gap in favorability mean that something about the decades-old hair-cutting incident stuck? Stayed with us to today? It could mean that--and you can bet that the campaigns data-mining experts are looking into this as deeply as they can to figure out what it means.

What Do I Think?
I think this is new and it's important. I believe that political narratives have a lock-and-key dynamic where a specific message will "fit into" and "resonate with" a set (demographic) of voters. Hillary's Red-Telephone-Ad played to a specific set of voters (security-moms) and "fit into" a weakness of Obama's and a strength of Hillary's. It moved the dial in a key swing-state during that primary. Hits to Gingrich's character "fit into" a demographic (women) and moved the dial in Florida. Learning how this works--refining it--will have a significant impact on how campaigns conduct their messaging in the future and I think this is the kind of tool that will bring it to us.

The real question is: behind the raw numbers what does Twitter know about the demographics of the tweeters that make up the message ... and what are they doing with that data? Selling it, I presume ...

No comments:

Post a Comment