Inside the new social media lab monitoring South Carolina's feelings about 2020 politics

[Uncle Joe]

Kait Park trains faculty members on using the lab’s software to enhance their research as part of a series of training sessions she offered for USC’s College of Information and Communications faculty. Kim Truett/Provided

(snip)

The Social Media Insights Lab at the University of South Carolina has been monitoring what social media users in the Palmetto State are tweeting, posting and sharing about the field of Democratic presidential candidates.

It isn’t a poll, but the findings can be just as telling, lab manager Kait Park said.

“I call it the canary in the coal mine,” Park said of social media conversations. “It’s a barometer. Social can show where excitement, pressure, and fear lies.”

It can also show indicators that polls don’t always capture.

(snip)

https://www.postandcourier.com/politics/inside-the-new-social-media-lab-monitoring-south-carolina-s/article_2d4fd09c-2023-11ea-b2d0-4ba4661c8fc9.html

[tiredtoo]

1. Every day someone posts a new poll here

Showing their candidate to be leading the pack. This most likely pleases the poster and the candidates supporters. However, if i remember correctly the polls had Hillary winning. So my friends, get your kicks from the latest polls but beware. Some trump supporters of my acquaintance have told me when they get polled they lie like trump

[karynnj]

6. The 2016 polls were not that wrong

First of all they did not poll up to the day of the election. The ones done within a week of the election - on average had her about 3% ahead nationally. The trend was her lead was going down slowly. The election results WERE HRC winning the popular vote by over 2%. One thing the exit polls showed was that people who disliked both, broke for Trump. Here is a summary of polls from a right leaning site - https://www.realclearpolitics.com/epolls/2016/president/us/general_election_trump_vs_clinton_vs_johnson_vs_stein-5952.html )

Very very few state polls were done in the last week. Here is the RCP electoral politics projections at the end. In 2016, almost any state that was close went to Trump. (As Nate Silver had pointed out before the election, it was likely that shifts in various states would be in the same direction. In the 2006 Senate elections that happened where we won in every even remotely likely state except Tennessee.)
As you can see, the states at the end were not as uniformly pro Clinton as the national result. (Note that Wisconsin's polls really were suggesting a Clinton win.)

[PhoenixDem]

2. Guessing overall trends from social media

is like getting an opinion on alcohol only from bar patrons.

Statistically useless.

[tiredtoo]

3. No more useless then the polls being posted daily. eom

Actually it is more important than the polls, that's how Russia helped steal our last election.

[PhoenixDem]

5. Polls have the force of statistics behind them

with proper sampling, they can be much more accurate.

Social media have a built-in sampling error because social media users are vastly younger and opinionated activists.

[tiredtoo]

8. Social media users are vastly younger and OPINIONATED activists

Wow, is that your activated opinion or do you have some data to back that up?

[Algernon Moncrieff]

9. It provides insight

Much like a focus group. It's not looking at the electorate at large in a statistically meaningful way but provides ideas on what social media consumers are thinking. Even that needs to be drilled down further. Facebook, for example, skews much older and to less tech-savvy users.

[ritapria]

4. Cinventional polling is much more accurate than social media

Social media is a barometer of voter enthusiasm for a particular candidate, not breadth of support …...The social media buzz generated by Trump was an indicator he would overperform the results he generated in conventional polling …...He drew much larger crowds than Hillary in the fall of 2016 ...That was also a lead indicator that his base was going to turn out in larger numbers than the Dem base - that Hillary was in trouble ….. Enthusiasm matters - at the margins - in close races .

[karynnj]

7. Interesting to see, but there are a lot of flaws here too

1) There is a correlation between age and the likelihood a person is on social media that would undercount the oldest voters, who are also very likely to vote.

2) If they were looking at 2016 -- did they attempt to eliminate the many Russian bots?