There is more than one way to do a poll.

The US  Presidential elections… according to the Gallups and the Yougovs are swings and roundabouts or neck and neck… but link domain stats suggest a landslide win for…

Well read on…

Some time back, Majestic SEO found a great correlation between links to a political candidate’s official page and the result. Well something really interesting has happened between Romney and Obama:

The last month has seen a MASSIVE swing to Mitt Romney. While the polls are suggesting this is a close fought race, the closeness may well be dependent on whether the incumbent’s past “branding” will keep him going past the winning post, because cumulatively, Obama has more sites linking to him than Romney does, but the rate of acquisition suddenly changed last month.

So what has happened?

If we look at links seen during the last 60 days, rather than all time, we can look at the links seen day by day:

As you can see, the whole game seems to have changed immediately before the first Presidential debate. Ever since that point, Mitt Romney seems to have kept up the online brand messaging. Here is the same chart on a cumulative basis.

During the London Mayor election, we used back link history (the way in which website link to each other) as a proxy for opinion between the two main candidates. Boris Johnson – now known around the world after the Olympics – won reasonably convincingly and our graph mapped the opinions of the public well… Spot the similarity?

Boris finally won the elections in London by about 6 points. Right now, I have to predict that unless Obama changes the dynamics a bit, Romney will do rather better than the “pundits” are saying on election day.

…But read on…

Is Crowd-sourcing a better predictor than polling?

Using Majestic SEO as a prediction tool may have some advantages over other more traditional methods.

  • The data contains a massive sample. Majestic SEO sees about 3 billion URLs every day
  • The data is international in nature, but since all URLs are checked against a Geo-targeting database, detailed analysis could break down data by country. (In this case, the data is US centric mainly because of the subject matter and the two sites compared.
  • The data is FAST! We update our index constantly although these charts are currently being updated every day, other data updates most hours.
  • The polling does not JUST have to be looking at electoral Maths. Many other concepts can be visualized through link analysis.

…but read on…

 Some Provisos

This data was collected at a top level. At the top level it is only a good predictor if the data was not manipulated. In a democratic election, manipulation is infact the whole name of the game! Knowing where all these links came from may dramatically change one’s opinion on what this data means. Luckily, we can easily see where Mitt Romney acquired all his links over a three day period through the New links tab on our Site Explorer. OK – this requires a $49 subscription… but that’s GOT to be cheaper than phoning 3,000 people to try and get a sample.

You are welcome to register with Majestic (it only takes a minute) to see where these thousands of links came from. If you are just here for the read, here are a few:

However – closer inspection reveals that the vast majority of links came and went in the same day (or over a short period). This suggests a specific form of marketing – a heavy PR campaign – which resulted in a spike in the results which rapidly left the home page of websites, but may still be on these news and blog sites somewhere:

So I take my hat off to a PR/SEO team that must have spent years preparing that push. I am sure it had an effect, but because the links arrived SO quickly, they stand out and in this instance may not represent the change in fortunes Mitt Romney may have hoped for and suggested at the top level analysis. Even so – the idea of using link analysis to predict an election still holds true. Just beware of wild spikes.

If you are interested in other ways in which Majestic SEO’s data might be useful within your own predictive modelling, either consider a web based account or contact Majestic’s support for more information about our Enterprise level APIs.

Comments

  • Eider Vasconcelos

    Very detailed and I take my hat off to you guys for this analysis. Very interesting. I specialize in hispanic markets and your analysis has triggered my thought to follow your example in Latin American countries.

    October 22, 2012 at 3:30 pm
    • Dixon

      Thanks. I think anyone could be MUCH more detailed if they really wanted to investigate who is saying what. If you analysed the sentiment of the URLs linking, for example, you would get even more useful data. The spike, though, masks the true “lie of the land” I think.

      I appreciate the feedback though, Eider!

      October 22, 2012 at 3:34 pm
  • Cody Sharp

    This is a great insight. However, remember that only a few key states will actually matter when it comes down to it. Ohio, PA, Florida, Colorado, etc. Unlike the London race where it was a straight majority vote to win. Still, an unorthodox analytics idea. I enjoyed it.

    October 22, 2012 at 3:42 pm
    • Dixon

      Thanks Cody. It would be strightforward to look at all these links by State… well – OK – 35,000 of them appearing that quick may need a bit of looking at I guess… but doable.

      October 22, 2012 at 4:18 pm
  • Joe Henners

    Very interesting. But there might be some factors wich make it difficult for very precise Internet statistics. I Think older people prefer Romney, but they are not so active on the Internet than the Obama fans wich might be of younger age. Such Demographic factors (Digital Divide) could make it difficult to get the full information. On the other hand I think such methods could be very useful for analysts to track the performance of Web comapanies, Social Networks and so on. Such Data might be useful to make a prognosis of future development which would be indeed very interesting for the Investment Guys in the Big Banks, Fonds and Hedgefonds.

    October 23, 2012 at 1:29 pm
    • Dixon

      You make a good point that the data would be influenced by other demographics such as age. Again, however, anyone who wanted to build on this modelling seriously could do so – for example by overlaying our data with Microsoft Passport API data – which includes broad age range demographics for users to any website.

      October 23, 2012 at 2:46 pm
  • Jason Hawkins

    I agree strongly regarding age, as most people over the age of 60 (just in my opinion) would likely not know how to “link” to something. Sure, there is exceptions to this rule, however feel it is an important point to consider. Joe Henners, love the points you outlined regarding Obama supporters being younger and hence may be a slight clue as to why Obama is more heavily linked. (Not to mention the favorable media coverage, which yields links).

    With heavy commercials being ran, obtaining a great amount of links in a relatively short amount of time would not raise a red flag in my book. Additionally, social media has been heating up between Obama & Romney, which may cause sharing (Google +1’s now count as links), etc.

    October 23, 2012 at 8:05 pm
    • Dixon

      Thanks for the add , Jason.

      If you think about it, the oldies don’t NEED to know how to link… map the link data to the user age demograohics (Microsoft Passport) and THEN map the URL onto traffic data (Hitwise, Compete.com, Comscore or Alexa) and then you have the full mix. Of course… Oldies are less likely to use the Internet at all. I don’t have THAT layer of te puzzle sorted yet.>

      October 23, 2012 at 8:30 pm
  • Oscar Gonzalez

    Awesome analysis. And when you have such big data available I do think you can predict a lot of stuff with a lot of accuracy. As much as we think that we are “random” or we try to change how people and machines see us, we are highly predictable given enough data.

    October 24, 2012 at 2:57 am
  • Jamie

    Many factors could come into play here with out seeing more of the digital plan.

    I read recently that Obama’s team was spending far more on SEO/PPC compared to Romney, who favored print and TV advertising.

    I believe over the last 2 weeks they have been pushing TV very heavily so this may have had some correlation towards people linking after they see the commercial on TV possibly?

    but overall the news is limited in Australia for this type of research, thank you for sharing it with us.

    October 24, 2012 at 6:41 am
    • Dixon

      The rise in links to the Obama site don’t look “natural”… but then again, neither presidential campaign looks very “natural” in any way shape or form anyway.

      October 24, 2012 at 9:54 am
  • Cathie

    So are you saying this was “SEO campaign” for making Romney look more popular than Obama and the marketers were counting on someone to notice, but someone less sophisticated than Majestic Seo? They were counting on a less astute observer, is that what you mean? I’m confused, because I’m not sure how the ranking higher for “mitt romney” would help him. So not sure what to take from this post, although it is very interesting.

    By the way, the anchor text, did you see some of them? “Dirty cat litter,” was one of the keywords. Lol.

    October 27, 2012 at 1:40 am
    • Dixon

      I am not saying it was for an SEO push. After all – one person would type in “Obama” and the other “Romney”. But the online campaign put in a huge spike which is far from “natural’ link acquisition.

      October 27, 2012 at 7:05 pm
  • Dixon Jones

    So the elections are done and dusted. Obama landslide. I’m putting that down as a win for using link analysis as an opinion prediction tool.

    November 18, 2012 at 12:51 pm

Comments are closed.