Perhaps the most bizarre electoral news since the midterms is the allegation that Republican operatives and outside groups used clandestine Twitter accounts to share polling data.
The Twitter accounts, which had no followers, tweeted polling results for House races in what might have been an effort to circumvent laws that preclude coordination between parties and outside groups. Their existence was first reported by Chris Moody of CNN, and the accounts were deleted moments after CNN contacted the National Republican Congressional Committee.
The mere existence of surreptitious Twitter accounts is not enough to prove a violation of the law. It is not illegal for campaigns or outside groups to post information, even if another group might use it to supplement, validate or even replace its own polling data.
But what's intriguing about the tweets is that they appear to be 'coded,' as Mr. Moody put it, though 'unlabeled' might be a better word. If the tweets require knowledge to interpret, then the case for illegal coordination becomes more credible.
One tweet, for instance, looked like this:
CA-40/43-44/49-44/44-50/36-44/49-10/16/14-52->49/476-10s
CNN posted screenshots of more than 100 tweets last week, and it's possible to figure out a lot of what the data means if there are enough tweets for context.
Yet at the same time, it's hard to be certain about the data or its assumptions. Without additional information, it's a stretch to imagine that the tweets were useful to anyone.
Decoding most of the meaning of the tweets is possible by identifying races with distinctive characteristics, like New York's 19th District, in the mid-Hudson Valley and Catskills, where the Republican incumbent Chris Gibson won easily over Sean Eldridge. The result was all but assured, but it was a district that voted for President Obama. Two West Virginia districts were among the few competitive races where Mr. Obama's approval rating was sure to be in the 20s. These characteristics let you identify which results from the tweet seemed to match the expected yet distinctive finding.
30pts.
Republicans Outperform Poll →
← Republicans Underperform Poll
0pts.
40pts.
← Republicans Underperform Poll
Republicans Outperform Poll →
Using this method, the tweets appear to offer sets of data, each separated by a hyphen. First is the state, indicated by its postal code. Then come two ballot tests, the first being of an unknown character (perhaps the generic ballot, a generic re-elect question or perhaps the incumbent's approval rating) and the second being the head-to-head result. What follows seems to be the favorability ratings for the two candidates, the president's approval rating and the congressional district.
If you prefer Rosetta Stone form:
Tweet: NY-48/36-56/33-59/19-25/27-39/57-10/23/14-19
New York
Republican incumbent 48, Democratic challenger 36 (or perhaps another generic polling result)
Gibson 56, Eldridge 33
Gibson Favorable 59, Unfavorable 19
Eldridge Favorable 25, Unfavorable 27
Obama Approval 39, Disapproval 57
10/23/2014
19th Congressional District
If the findings after the first set of numbers are indeed the results, they were more accurate than the public polls, though not by much.
On average, the findings from October differed from the results by 5.9 percentage points, compared with 6.6 points from 58 House telephone polls recorded on The Huffington Post's Pollster page for top House races. The polls were biased toward Democrats by an average of 1.85 points, but there were three polls (all in Massachusetts) where Democratic chances were understated by at least 10 points.
The polls were tweeted by two accounts, @Brunogianelli44 and @TruthTrain14. (Bruno Gianelli was a character on 'The West Wing.') Of the two, the results tweeted by the @Brunogianelli44 account tended to be modestly more accurate, differing by just 5.5 points. The results from the @TruthTrain14 account had a larger average error of 6.3 points.
If anyone was using these tweets, my guess is that they wouldn't have done so without more information about the tweets or the polls. Without that information, one would be taking a huge risk; there's no way to be sure about the findings.
But this poses a big challenge to regulators. Anything on the Internet is public; Twitter or obscure, unlinked websites could easily be a medium to transmit private messages through 'public' channels. It's hard to see what regulators can do about it.
Tidak ada komentar :
Posting Komentar