Tweets Can Foretell Votes, Study Finds

National Journal
Alex Roarty
Aug. 12, 2013, 2:29 p.m.

Who needs polls? A study pub­lished Monday re­ports that cam­paigns could use Twit­ter to suc­cess­fully pre­dict the win­ner of most races, find­ings that might bol­ster the so­cial me­dia ser­vice’s already ro­bust polit­ic­al pres­ence.

The key meas­ure, re­search­ers from In­di­ana Uni­versity found, was a can­did­ate’s “tweet share,” the per­cent­age of total tweets about a race that men­tion them. The more of­ten a can­did­ate is men­tioned on Twit­ter re­l­at­ive to their op­pon­ent, the study re­por­ted, the great­er their chance for vic­tory.

The find­ings were com­pre­hens­ive: An ana­lys­is of tweets from the 2010 midterm elec­tions found the data cor­rectly pre­dicted the win­ner in 404 of the 406 House races.

“We plot­ted it and thought, ‘Holy moly, it was a very strong cor­rel­a­tion,’ ” said Fa­bio Ro­jas, a so­ci­ology pro­fess­or at In­di­ana and one of the study’s coau­thors. He ad­ded that pre­lim­in­ary ana­lys­is of last year’s con­gres­sion­al elec­tions show sim­il­ar res­ults.

The find­ings rest on two im­port­ant points: The raw num­ber of tweets about a can­did­ate doesn’t mat­ter, and neither does wheth­er the tweets are pos­it­ive or neg­at­ive. Ro­jas and his col­leagues, who col­lec­ted hun­dreds of thou­sands of tweets from the 2010 race, ini­tially meas­ured the total num­ber of times the can­did­ate was men­tioned, but the find­ings failed to cor­rel­ate with which can­did­ate won. Well-known can­did­ates, like Rep. Michele Bach­mann, R-Minn., or can­did­ates from big­ger and wealth­i­er dis­tricts would in­her­ently re­ceive more at­ten­tion.

It was then, Ro­jas said, they real­ized that what mattered was the Twit­ter “horse race,” or the num­ber of tweets a can­did­ate earns vis-à-vis his or her op­pon­ent. Just as the can­did­ates would com­pete for a lim­ited per­cent­age of the vote, they would also com­pete for a lim­ited per­cent­age of the total Twit­ter traffic.

Per­haps most in­ter­est­ing, wheth­er the tweet praised or cri­ti­cized the re­cip­i­ent was ir­rel­ev­ant. When it comes to Twit­ter and politi­cians, ap­par­ently all pub­li­city really is good pub­li­city.

“Are you go­ing to talk about the guy who loses or the guy who wins?” Ro­jas asked. “You’re go­ing to talk about the win­ner, even if you hate the win­ner.”

He ad­ded that al­though cam­paigns could seem­ingly skew the res­ults by pay­ing so­cial me­dia dir­ect­ors to tweet or by ask­ing vo­lun­teers to pitch in, such a prob­lem has not yet aris­en. Such ef­forts are usu­ally can­celed out by sim­il­ar ac­tion taken by their rival, he said.

In Ro­jas’s view, the find­ings should re­vo­lu­tion­ize how cam­paigns con­duct them­selves. Rather than spend­ing hun­dreds of thou­sands, or even mil­lions, of dol­lars on sur­veys, cam­paigns could simply gauge their status on Twit­ter. That should help cam­paigns with few­er re­sources com­pete with well-heeled in­cum­bents, he said.

“The point is, it’s cheap,” he said. “Once you start up soft­ware for col­lect­ing tweets, it’s very cheap. It took one of my Ph.D. stu­dents a couple of weeks to set it up.”

Of course, pro­fes­sion­al polling isn’t likely to dis­ap­pear from polit­ics any time soon. For one, it’s used for more than just the horse race — cam­paigns test a vari­ety of things with polls, in­clud­ing their mes­sage. Twit­ter doesn’t of­fer help that way. And few politi­cians would be will­ing to switch off from a battle-tested poll­ster in fa­vor of a tech­no­logy un­proven in the heat of a crit­ic­al race.

But Twit­ter is also of­ten de­rided as a hangout for polit­ic­al and me­dia elites, pro­du­cing a de­bate that bears little re­semb­lance to the thoughts and opin­ions of most voters. This study sug­gests that the so­cial me­dia out­let does provide an ac­cur­ate re­flec­tion of the elect­or­ate.

Guy Har­ris­on, a former ex­ec­ut­ive dir­ect­or of the Na­tion­al Re­pub­lic­an Con­gres­sion­al Com­mit­tee, said he’d have to see a Twit­ter ana­lys­is pro­duce ac­cur­ate res­ults in a com­pet­it­ive race be­fore put­ting his faith in it. But the ser­vice’s im­port­ance to the polit­ic­al de­bate, he said, is bey­ond ques­tion at this point.

“So­cial me­dia and di­git­al me­dia across the board is here to stay, and it’s go­ing to be here a long time,” he said.

{{ BIZOBJ (video: 4382) }}

What We're Following See More »
ON GUN RIGHTS
Trump Jr. Meeting with GOP Members
11 hours ago
THE LATEST
FLOPPY DISKS
US Nukes Rely on Decades-Old Tech
11 hours ago
THE DETAILS
CONTRARY TO REPORTS
Ryan Not Endorsing Trump Just Yet
16 hours ago
THE LATEST
OTHER SECRETARIES AT FAULT, TOO
State Dept. Review Faults Clinton Email Management
18 hours ago
THE LATEST

"A State Department audit has faulted Hillary Clinton and previous secretaries of state for poorly managing email and other computer information and slowly responding to new cybersecurity risks. ... It cites 'longstanding, systemic weaknesses' related to communications. These started before Clinton's appointment as secretary of state, but her failures were singled out as more serious."

Source:
CRUZ STILL TOOK DELEGATES AT THE CONVENTION
Trump Rolls in Washington Primary
18 hours ago
THE LATEST

Donald Trump "was on course to win more than three-quarters of the vote in Washington's primary" last night. Ted Cruz's defunct candidacy still pulled about 10 percent. "Cruz dropped out of the race on May 3, but won 40 of the state's 41 delegates up for grabs at last weekend's state GOP convention."

Source:
×