Predicting election results with social media

Ciarán Mac Mathghamhna

Active member
Joined
Jun 7, 2007
Messages
185
As some of you may have read elsewhere online, a certain level of success in predicting the recent US midterm elections was observed using data from Facebook and Twitter.

In 71% of races, the candidate with the most Twitter followers was ahead in the polls.
New Data: Can Twitter Predict Elections? | Dan Zarrella

Number of Facebook fans correctly predicted of 74 percent of the House races and 81 percent of the Senate races:
Facebook, Twitter analytics successfully predict 2010 election winners | VentureBeat


Do you think that these methods would be successful in this jurisdiction? Currently, in the Donegal Southwest race, we have Facebook friends/fans:

Doherty (Sinn Féin) 1746
McBrearty, 909 (Labour)
Ó Domhnaill, 909 (Fianna Fáil)
O Neill, 1005 (Fine Gael)


Only Doherty is on Twitter, as far as I know (221 followers). This would suggest that he will run away with the seat.
 


neiphin

Well-known member
Joined
Aug 23, 2009
Messages
5,511
would it be possible for dave to set up a poll
where you could only vote once and in one constituency ?
 

Ciarán Mac Mathghamhna

Active member
Joined
Jun 7, 2007
Messages
185
I've been thinking about something that! Wouldn't it be very interesting to see how well an online poll would do in comparison to TNS/MRBI/Landsdowne/Red C/Quantum Research etc?

Actually, I have a very good idea of what the build would look like and how a revenue stream could be generated.

It would be very interesting, but would require a significant amount of work if it was to be done right.

Any unemployed web developers out there with a social conscience?
 

pandora

Well-known member
Joined
May 3, 2009
Messages
346
As some of you may have read elsewhere online, a certain level of success in predicting the recent US midterm elections was observed using data from Facebook and Twitter.


Do you think that these methods would be successful in this jurisdiction? .
Interesting idea but you have to remember that Donegal isn't the US in terms of intrastructure.
Many of us up here don't have fast broadband connections and those of us who do pay a lot more for it than Americans would.
 

Ciarán Mac Mathghamhna

Active member
Joined
Jun 7, 2007
Messages
185
Interesting idea but you have to remember that Donegal isn't the US in terms of intrastructure.
Many of us up here don't have fast broadband connections and those of us who do pay a lot more for it than Americans would.
That's exactly why I'm interested in it - if it works in Donegal, it'll work for the general election.

However, I'm not entirely sure that Donegal would have poorer connectivity than many parts of the US - in fact, as a nation, on many metrics, we score above them.

For example, this report states:"for our average of 3.9Mbps, we pay about $40 per month."

US broadband's average speed: 3.9Mbps

How does that compare with what you're paying in Donegal?


However, the penetration is considerable higher - Donegal I believe is about 45%, whereas the US is about 77%


As I say, it will be interesting to see how the internet data relates to the ballot box.
 

Goldencircle

Well-known member
Joined
Mar 2, 2010
Messages
1,201
Fianna fail

1,378 people like this

-

lets get rid of FIANNA FAIL controversial corrupt ministers anTaoiseach.

443 people like this
Well Fianna Fail, You have finally done it. The country is finally ruined

2,148 people like this.
Get Brian Cowen and Fianna Fail out of government

437 people like this.
Remove Fianna Fail From Office

238 people like this.
GET FIANNA FAIL OUT

1,034 members

BAN FIANNA FAIL FROM IRISH ELECTIONS

33 people like this.

Fianna Fail Out!

30 people like this.

Punching Fianna Fail TD's in the face when they come looking for votes
Government


4 people like this.

Wanted for Treason! The entire Fianna Fail Parliamentary Party

18 people like this.

Fianna Fáil should be renamed to Fianna FAIL
6 people like this.

Vote Fianna Fáil, Vote FÁILure

30 members

(It goes on and on)




Fianna fail gets a grand total of -3010
 

QuizMaster

Well-known member
Joined
May 26, 2004
Messages
3,193
Website
www.quizmatic.com
I live in rural Donegal and have a fairly good connection with 3 mobile.
Most people who want broadband have it.
 

jmcc

Well-known member
Joined
Jun 12, 2004
Messages
42,359
So what do those data predict then?
Not having access to them, I would expect them to provide a more accurate prediction than the social media data. The main reasoning is that they would provide a far larger and demographically wider data set. The Social Media data would be from a self selecting group and would be dependent on the number of local voters interested in the subject. Now whether they would translate to real votes is open to question. The difference between social media is that it is largely responsive (people responding to a follow, friend or pack activity) whereas the search engine traffic is aquisitive (people who want to find out about a candidate). Sorry about my brief reply yesterday but I was in the middle of working on a pile of internet statistics.

Regards...jmcc
 

Ciarán Mac Mathghamhna

Active member
Joined
Jun 7, 2007
Messages
185
I agree that having access to that kind of data would give us a broader picture, but I don't necessarily think that it would be more accurate. I mean, if you become friends or 'Like' a candidate, I think that's a clearer, stronger and more public measure of preference than googling them. It's one thing to 'like' someone, quite another to 'google' them. I mean, you might google all the candidates, but you will probably only 'like' the one you're actually a fan of.

At any rate, we don't have access to the data. The beauty of the Facebook numbers is that anyone can access them - no need to get a company of pollsters or access search engine records.
 

Tea Party Patriot

Well-known member
Joined
Oct 31, 2010
Messages
11,468
I've been thinking about something that! Wouldn't it be very interesting to see how well an online poll would do in comparison to TNS/MRBI/Landsdowne/Red C/Quantum Research etc?

Actually, I have a very good idea of what the build would look like and how a revenue stream could be generated.

It would be very interesting, but would require a significant amount of work if it was to be done right.

Any unemployed web developers out there with a social conscience?
Very hard to prevent manipulation of online polls, especially with dynamic IP addresses. Remember A Nation Once Again being the song of the century on BBC :D
 

Ciarán Mac Mathghamhna

Active member
Joined
Jun 7, 2007
Messages
185
Very hard to prevent manipulation of online polls, especially with dynamic IP addresses. Remember A Nation Once Again being the song of the century on BBC :D
Very true. Although I wonder which party would benefit most from that type of carry-on :)


I would envisage getting around that issue by making users register with the site, and only allowing voting once a week.


All I need is about ten grand and two or three web developers ....
 

jmcc

Well-known member
Joined
Jun 12, 2004
Messages
42,359
I agree that having access to that kind of data would give us a broader picture, but I don't necessarily think that it would be more accurate.
I would think that it would be far more accurate given that it is a bigger dataset and that it is not self-selecting.

I mean, if you become friends or 'Like' a candidate, I think that's a clearer, stronger and more public measure of preference than googling them. It's one thing to 'like' someone, quite another to 'google' them. I mean, you might google all the candidates, but you will probably only 'like' the one you're actually a fan of.
Becoming a friend or following a candidate is a weak link and is no real guarantee of intent. What the search data reveals is the raw intentions and interests of voters. They are in the aquisitive mode and are looking for information. This is often information that would not be gained by pollsters. Just on instinct, the communications flow between these popular politicians and their followers/and friends might be almost completely from the politicians to the followers with minimal interaction. This is the push model. The search engine and web data would be the pull model in that the searcher is looking for information. The more searches being made on a topic is a general indication of the popularity of the topic.

Google Trends actually provides some indication of search volume by region:
Google Trends
Google Insights for Search

Regards...jmcc
 

pandora

Well-known member
Joined
May 3, 2009
Messages
346
I live in rural Donegal and have a fairly good connection with 3 mobile.
Most people who want broadband have it.
Most people can get mobile broadband in Donegal but many still can't get fixed line. It seems pretty hit and miss in this area, sometimes one house can get it but the person directly across the road can't. You have to have a direct line of sight to the mast so if there is a hill in the way it's tough luck, and there are a lot of hills in Donegal.

In Donegal, and I suspect in other parts of Ireland, people like to see the candidate, shake hands and just size them up in person before they vote. Maybe they will settle for checking them out on the internet and maybe they won't.
 

Ciarán Mac Mathghamhna

Active member
Joined
Jun 7, 2007
Messages
185
I think we can safely say that election results can be predicted with social media.

I called them the night before as follows (Twitter)

Doherty, 38%
McBrearty, 19%
O'Domhnaill, 21%
O'Neill, 18%
Pringle, 4%


These percentages were simply based on the number of Facebook Fans or Friends each had (e.g. of the total of about 5k, Doherty had about 2200)

Obviously, Pringle is underestimated as he didn't have much of a Facebook campaign.

It also seems that McBrearty had built up a similar level of support as O'Domhnaill and O'Neill. His comments about the by-election being pointless may well have convinced about half of that to stay at home.

However, it is remarkable how close the figures for Doherty, O Domhnaill and O'Neill are - very close indeed. The days of telephone opinion polling are numbered.
 

KingKane

Well-known member
Joined
Aug 19, 2003
Messages
2,323
Website
www.danielsullivan.ie
Twitter
kingkane
I think we can safely say that election results can be predicted with social media.

I called them the night before as follows (Twitter)

Doherty, 38%
McBrearty, 19%
O'Domhnaill, 21%
O'Neill, 18%
Pringle, 4%


These percentages were simply based on the number of Facebook Fans or Friends each had (e.g. of the total of about 5k, Doherty had about 2200)

Obviously, Pringle is underestimated as he didn't have much of a Facebook campaign.

It also seems that McBrearty had built up a similar level of support as O'Domhnaill and O'Neill. His comments about the by-election being pointless may well have convinced about half of that to stay at home.

However, it is remarkable how close the figures for Doherty, O Domhnaill and O'Neill are - very close indeed. The days of telephone opinion polling are numbered.
Are you off your trolley or can you simply not add? Your %'s are well off the final results. You've got McBreaty on twice what he got, and Pringle on half what it got, it's a joke of a prediction and you're as bad as FF or Labour claiming the result represents a success.
 

Ciarán Mac Mathghamhna

Active member
Joined
Jun 7, 2007
Messages
185
Are you off your trolley or can you simply not add? Your %'s are well off the final results. You've got McBreaty on twice what he got, and Pringle on half what it got, it's a joke of a prediction and you're as bad as FF or Labour claiming the result represents a success.

I think you are the one with mental deficiencies, specifically in the area of reading. Have a look at the title of this thread, the original post and replies. I predicted that Doherty would win. Doherty has won.

If you can read, then have a look at the rest of the preceding posts - it was said that this method of predicting election winners would not work, because Donegal doesn't have great broadband connection, and because the method was flawed. Again, because I did pick the winner with this method, I've shown those criticisms not to hold. Of course the figures aren't perfect, but they're well within the range of both the final result and the polls taken beforehand - that is a good result, proof of concept, and a success in my book.


Red C/Paddy Power:
Doherty, 40%; McBrearty, 14%; O’Domhnaill, 19%; O'Neill, 15%

Exit Poll:
Doherty, 33%; McBrearty, 19%; Ó Domhnaill, 21%; O’Neill, 18%; Pringle, 8%

First Preferences:
Doherty, 39.8%; McBrearty, 9.8%; Ó Domhnaill, 21.3%; O’Neill, 18.6%; Pringle, 9.9%

Facebook Prediction:
Doherty, 38%; McBrearty, 19%; O'Domhnaill, 21%; O'Neill, 18%; Pringle, 4%


If anyone still wants to argue over the figures, fine - go ahead, I have better things to do with my time. However, if you stop and have a little think about this, you might realise that I am actually giving you a very useful resource. You don't have to wait until an opinion poll comes out any longer: you can get a reasonable idea of how your candidate is doing from their Facebook page.
 

jmcc

Well-known member
Joined
Jun 12, 2004
Messages
42,359
If you can read, then have a look at the rest of the preceding posts - it was said that this method of predicting election winners would not work, because Donegal doesn't have great broadband connection, and because the method was flawed.
It was valid criticism.

Again, because I did pick the winner with this method, I've shown those criticisms not to hold.
No you did not. All you did was make a guess on a situation that was already decided.

Of course the figures aren't perfect, but they're well within the range of both the final result and the polls taken beforehand - that is a good result, proof of concept, and a success in my book.
Maybe in social science but not in statistics or real science. You chose a case where the outcome was, to a great extent, already decided and now you are trying to gather "figures" to support your guess.

If anyone still wants to argue over the figures, fine - go ahead, I have better things to do with my time.
That's what distinguishes cargo-cult science from real science and predictions. Real science, and statistics, depend on numbers to make predictions.

However, if you stop and have a little think about this, you might realise that I am actually giving you a very useful resource.
Did it ever occur to you that people here might already know about such methods and may even be more familiar with them as a predictive tool than you appear to be?

You don't have to wait until an opinion poll comes out any longer: you can get a reasonable idea of how your candidate is doing from their Facebook page.
Relying on a single source of data and an extremely limited dataset when many factors are at play is not a reliable method of analysis. Perhaps you should consider astrology as a career. ;)

Read the book that I recommended earlier as it gives a far better insight into using web activity to predict outcomes.

Regards...jmcc
 


New Threads

Popular Threads

Most Replies

Top