Can we predict the NH Gubernatorial primary outcomes by watching campaign page statistics on Facebook?
Here we compare the ‘Likes’ and activity ‘people talking about this’ for each page – one month before the primary (8/8 and one day before the primary (9/10):
First the Republican race: Ovide had and has tremendously more likes than Smith. Smith has a better slope on Like count (1.3 vs. 1.1) but the difference is so large as to not matter. Similarly, the activity slopes favor Smith (2.4 vs. 1.7) but the magnitude is again so much different that it won’t matter. Smith is very popular among some, but not enough.
The Democratic race gets less attention on Facebook in general and is a bit closer (Hassan has roughly half the lead over Cilley as Ovide does over Smith). Hassan has a better slope (1.4 vs. 1.2) on Likes but Cilley has a better slope on activity (2.1 vs. 1.6). Cilley will pose a stronger challenge to Hassan than Smith will to Ovide, but it will not be enough to win it for her.
Using the perhaps-bogus method of multiplying the current likes by the activity slope, and doing percentages (should probably use a standard normal table instead) the outcomes will be:
Republican: Ovide: 73%, Smith 27%
Democrat: Hassan: 67%, Cilley 33%
We’ll find out tomorrow to what degree Facebook pages reflect the voting population at large and just how bad that calculation is. New Hampshire has such a small percentage of party voters that it’s probably not useful to extend this to the general election at this time.
Update: actual results:
Ovide: 67.7%, Smith: 29.8%
Hassan: 53.1%, Cilley 38.9%
Not bad considering how “WILDLY WRONG” some prominent politicos declared the predictions of this method!