Big Data Analytics Predict Obama Win

Big Data Analytics Predict Obama Win

Here’s an interesting article from InformationWeek        

which discusses how University of Illinois professor Sheldon Jacobson and American University professor Allan Lichtman have both devised different methods of election forecasting which have each separately predicted an Obama victory. It gives a particularly vivid description of Lichtman’s method, which relies on thirteen keys in order to make a prediction. The keys are:

1. What were results of 2010 mid-term elections?
2. Was there a serious challenge to the incumbent’s party nomination?
3. Is the candidate the sitting president?
4. Is there a third party with a chance of getting 5% or more of the national vote?
5. Is the economy in recession during reelection year?
6. Strength of the long-term economy?
7. Has the current administration undertaken major policy change?
8. Is there major social unrest in the U.S., such as what occurred in the 1960s?
9. Is there a scandal like Watergate or the impeachment of Bill Clinton?.
10. Has the administration avoided a catastrophe abroad, such as losing a war?
11. Has the administration achieved a major foreign policy success?
12. Is the incumbent candidate charismatic?
13. Is the challenger charismatic?

Analysis of this from one election analysis, by nature, raises a few questions. First of all there is the issue of bias. While there does not seem to be any strong evidence that Lichtman is biased, one problem with this form of analysis is that, with some of the more subjective questions, such as questions about the charisma of candidates, the person reading the questions or doing the analysis runs a risk of letting his personal biases influence the answer he gives. For example, while Lichtman chose to assume that neither candidate was particularly charismatic when conducting the analysis, there are doubtlessly some who would disagree with that statement and say that one candidate was vastly more charismatic than the other. This could lead to some problems when using the method as well as accusations of bias from people on either side of the election. Nonetheless, many of the questions themselves seem fairly valid as long as they can be answered fairly objectively and the idea of trying to predict the outcome of elections without relying on polling voters is an interesting one.

This entry was posted in Uncategorized. Bookmark the permalink.

3 Responses to Big Data Analytics Predict Obama Win

  1. alexjking11 says:

    It’s funny you should have a blog post about this, as I wrote about a very similar thing for my term paper: the use of data and modeling in election forecasting. I also brought up Lichtman’s “13 Keys” method as an example of a model that uses subjective, not objective, data points.

    It seems to me that the leader in the use of data in election forecasting is Nate Silver. He has a blog at the New York Times called FiveThirtyEight and his forecasting model mainly relies upon state-by-state polling. What is your opinion of Silver and his model? Obviously, Lichtman’s is much easier to implement – I could probably answer all of his 13 questions off the top of my head. Is there value in Silver’s massive undertaking (for instance, he runs 100,000 simulations per day based on thousands of data points), or are Lichtman’s 13 questions enough to determine who will win the election?

  2. samargolis says:

    Not to cause any panic, but you weren’t alone, Alex; I wrote my paper on a similar thing (More emphasis on polling). There doesn’t seem to be much similarity between these models. Princeton has one which has an Obama win and the University of Colorado had one with Romney winning by a pretty wide margin (I remember the right-wing blogs going ga-ga over it).

    Silver seems to be the most solid, but he only has the 2008 general and 2010 midterms under his belt. In regards to Lichtman, surely each of the 13 questions is weighted in a different manner; I would assume that some of those questions are more valuable than others.

    I got quite a bit angry while writing my paper because many of these number crunchers whose name doesn’t rhyme with Sate Nilver seem to just make up stuff as they go along. I saw this guy cited on the Daily Caller and Drudge Report:

    “What? Obama can’t possibly be in the lead!!! I’ll adjust the numbers in how I see fit.” However, it isn’t just modelers, it’s also pollsters who either don’t really release their cross-tabs along with their toplines (Rassmussen) or pollsters who have a methodology which will make any high school graduate sad (Baydoun-Foster White).

    Don’t get me wrong, polling and good models are generally accurate, but anyone can release a high-visibility poll and considering polling is viewed by many as a legitimate study, that is very depressing.

  3. kristenmayer says:

    I think the system of using this set of keys is an interesting way of predicting the outcome of the election. How were these questions chosen over other potential questions? And should the questions be given equal weight? To me, it seems as some of the questions might be more useful in predicting a win for either candidate (for example, whether or not the economy is in recession seems more important than whether the candidate is charming). Additionally, some of these keys can be rather ambiguous. A win was given to Obama for primary healthcare reform, but people who didn’t support the healthcare reform would likely be swayed not to vote for Obama because of this.

    Personally, I’m not sure I believe that these 13 questions can adequately predict the outcomes of elections – though this method is 7 for 7 so far. It will be inserting to see if it will add another correct prediction to it’s record after the election on Tuesday.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s