Now that the Iowa Caucus is complete, the US Presidential Election has now begun in earnest. Over the coming months, words like “momentum” will be omnipresent as candidates from both sides of the aisle jockey for position.
After results are counted in New Hampshire tonight, we’ll likely start to see a number of the Republican candidates drop out of the race due to a combination of a lack of fundraising, low poll numbers, and poor showings in these early primaries (including the Iowa Caucus).
It’s important because a number of these candidates, including Gov. Chris Christie, Gov. John Kasich, and former Gov. Jeb Bush, have put a tremendous amount of time, money, and energy into tonight’s primary. A poor showing for many on the Republican side could mean it is time for them to bow out of the race.
As this author has written previously here and here, I will be using Twitter data and Facebook content to see if these data streams can outperform traditional polling methodologies. Since predictive analytics is the ultimate aim of social media research and measurement, here is my attempt at predicting what will happen at the New Hampshire primary for both parties.
Methodology by Source
Polling Data
For polling data, delegates counts were assigned based on the percentage of support for each candidate as of February 7th, 2016.
Republican Candidates | Trump | Cruz | Rubio | Bush | Kasich |
Poll Numbers as of 2/7/16 | 30.7% | 12.4% | 14.4% | 11.3% | 13.0% |
# of Delegates Expected | 7.061 | 2.852 | 3.312 | 2.599 | 2.99 |
Democratic Candidates | Sanders | Clinton |
Poll Numbers as of 2/7/16 | 53.9% | 40.7% |
# of Expected Delegates | 12.397 | 9.361 |
Twitter Data
Looking at overall Twitter volumes for users that describe themselves as being from New Hampshire from January 7, 2016 to February 7th, 2016, delegates are assigned for each candidate a based on their overall share of voice.
Based on volume of tweets during the past 32 days:
Republican Candidates | Trump | Cruz | Rubio | Bush | Kasich |
Volume of Tweets | 5,517 | 5,708 | 1,820 | 987 | 668 |
# of Delegates Expected | 8.632 | 8.931 | 2.848 | 1.544 | 1.045 |
Democratic Candidates | Sanders | Clinton |
Volume of Tweets | 4,407 | 4,209 |
# of Expected Delegates | 11.764 | 11.236 |
Facebook Data
Because of the rich demographic data that Facebook has at its disposal, Sysomos Scout is well positioned to capture conversations by age. Since I was able to find both historical voter turnout (2012 voter turnout numbers were used for this analysis) and the demographics of Facebook users themselves, I correlated these numbers to find an expected delegate count.
That said, in doing so, I’ve made a significant assumption: because Facebook users tend to be younger, and older Americans are more likely to vote, volumes from older Facebook users are worth ‘more’ than younger users’ mentions.
John Kasich is also absent from this analysis because his Facebook mentions were below the threshold to render results.
Republican Candidates | Trump | Cruz | Rubio | Bush |
Adjusted Facebook Volume –
18-24 Year Olds |
1,166 | 201 | 100 | 80 |
Adjusted Facebook Volume –
25-34 Year Olds |
1,608 | 487 | 219 | 171 |
Adjusted Facebook Volume –
35-44 Year Olds |
1,628 | 662 | 331 | 276 |
Adjusted Facebook Volume –
45-54 Year Olds |
2,050 | 891 | 475 | 416 |
Adjusted Facebook Volume –
55-64 Year Olds |
1,719 | 844 | 406 | 375 |
Adjusted Facebook Volume –
65+ |
1,331 | 665 | 333 | 333 |
Total Adjusted Volume | 9,501 | 3,751 | 1,865 | 1,650 |
# of Expected Delegates | 13.032 | 5.145 | 2.558 | 2.263 |
Democratic Candidates | Sanders | Clinton |
Adjusted Facebook Volume –
18-24 Year Olds |
864 | 482 |
Adjusted Facebook Volume –
25-34 Year Olds |
1,340 | 901 |
Adjusted Facebook Volume –
35-44 Year Olds |
1,352 | 993 |
Adjusted Facebook Volume –
45-54 Year Olds |
1,545 | 1,337 |
Adjusted Facebook Volume –
55-64 Year Olds |
1,500 | 1,375 |
Adjusted Facebook Volume –
65+ |
1,064 | 1,231 |
Total Adjusted Volume | 7,666 | 6,320 |
# of Expected Delegates | 12.607 | 10.393 |
My Prediction:
Based on the methodologies used in this analysis, my assumption is that the Facebook-data approach is more indicative of how the primary will ultimately unfold. Why? A combination of the rich demographic insights found within Facebook data and the correlation of this information with past election results; it leads me to believe this will likely be most predictive.
Keep following this blog as we move through the election cycle. As we continue to analyze more datasets across the country (and as more candidates drop out of the race), we’ll continue to hone in on and study which metrics work best.
(Image credit: Flickr user jamiedfw)