Last June, there was a critical Presidential election in Iran, An election that fooled all the traditional surveys and predictions, but I predicted the actual result with the highest accuracy using WordPress!
Living outside Iran, I followed the process of Iran’s 2013 election on the internet and other social media. A few hours before the polls opened, I published a statistical graph on Facebook and Twitter, the last in a series of predictions. As it turned out, that graph and others that I published during the weeks preceding the election were the most accurate statistical analysis among a range of predictions by news outlets as well as polling institutes inside and outside Iran.
Measuring public opinion could be widely applied in social studies, public relations, marketing and politics. In this article I try to explain how I did this. I will also try to point out the strengths and weaknesses of my and others’ analysis. It may be worth noting from the outset that my analysis was in no shape or form influenced by political speculation. And at no point did I intend to campaign for a certain candidate. In fact, for reasons out of the scope of this article, I didn’t vote in the 2013 election. My analysis, as elaborated in this article, like any other scientific research, has been solely based on existing data with no room for my personal views.
What I did was an indirect measuring of public view. I looked at the ‘trace’ Iranians left behind on the internet: people’s words, curiosities, questions, hopes and concerns, all of which they would discuss on social networks, search in Google or write about in blogs and new sites. In this method I didn’t have to ask people about their votes. I didn’t have to prepare a questionnaire as is usually done in any poll. Instead of asking individuals in a sample population, in this method we look at all that’s published and shared on social networks. In the case of a major event, such as an election, this public, existing information is called Big Data – which is identified, collected, and assessed in different ways.
The Beauty of This Kind of Survey is to Get the Most Honest Answer From the Public Without Asking a Single Question!
Sources picked, one has to extract their data. To do this I used the ‘feeds’ and API, looking for certain keywords or hashtags. For other data with a smaller size (e.g. Google Trend), I used Excel and manually stored what I needed. For data gathered via feeds and API, however, I needed a database. Here I used WordPress as interface and as the link between content and the database. (Each tweet is turned into a post in WordPress so that all the data can be stored in MySQL database. It would then be possible to search and process the raw data with PHP My Admin.
There are a number of sites as well as online programs for data analysis, some limited, some advanced, some free, some costing a fortune. The one I utilized, however, was one I have developed myself. I needed a cheap yet reliable option – specially with content in Persian.
I love WordPress because of its flexibility, you can use it as a canvas for any type of application, some define WordPress as the best interface between Data and Applications, Some say WordPress is the Web OS, I like to call WordPress an app engine or even a social engine.
It’s crucial in any analysis to distinguish between the various types of collected data in terms of their importance/significance. A tweet about the election, a comment on Facebook, a search in Google, or a blog post, each has a different ‘weight’. I had to estimate, as accurately as possible, the ratio of Iranians on Twitter (or those using Google as their search engine) to the entire population. I then gave ‘weight’ to each ‘source’ of data accordingly.
If you are interested to read the whole story of this survey please refer to the article that I had published in Iran Opinion.