Can Twitter predict political election results?

I have compared the 2010 Belgian general election, 2011 Singapore general election, and 2011 Irish general election results with Twitter messages. It appears that Twitter users and resulting tweets might in fact play a role in the political elections; although, I will argue that Twitter cannot predict political election results for the following reasons:

• ·         The number of Twitter users is only a small fraction of the voting population, according to the sampling percentages; Belgium 0.11%, Germany 0.24%, and Singapore 4.45%, this prevents an adequate representative sampling.
• ·         When using tweets to conduct scientific research, data miners need to identify the variables of the sample such as the various strata of users based on age, gender, income, political affiliation, technical knowledge, etc. The profiles of Twitter users should play a significant role in sampling research.
• ·         The validity of tweets is questionable due to the fact users can use a pseudo name; therefore, it is possible that tweets come from the same person.
• ·         Even though Twitter users post their viewpoint in favor of one political party, they do not necessarily have a negative sentiment against other parties. How can researchers distinguish or classify the negative and positive roles that tweets play on election results.
• ·         If retweets were counted as separate data for the study, this data probably would skew the analysis.
• ·         A political party might be the center of the discussion among Twitter users, thus the party largely received the number of positive and negative tweets. However, it is not necessary an indicator for the election result; such as the People's Action Party in Singapore.
• ·         We have to bear in mind that popularity is not necessary equal to prediction. The prediction of future events needs comprehensive and in-depth scientific research, not just numbers or contents of the tweets received.
• ·         This research dominantly used quantitative analysis and statistical inference methods. Utilizing combined qualitative methods for these studies might reveal different results.
•  

In my view, the study of tweets and the current events is an interesting research area. From these included studies, we can see the relevance of tweets in the political arena. One cannot draw a conclusion that tweets enable prediction of future events, such as election results; that could require detailed studies.