Every Tweet Counts: Examining Spatial Variability of Twitter Data Representativeness

The growing global Twitter population has prompted social scientists to examine the potential of Twitter-generated sentiments to serve as an alternative to public opinion polls. This thesis intends to study this potential by evaluating the variability of sampled data representativeness that is voluntarily submitted through Twitter. This research examines a case study: President Barack Obama's public approval as viewed by the United States population. The sentiments generated from Twitter were compared to the sentiments from public opinion polls in order to measure the degree of representativeness at both national and state level. The results show that Twitter data are not representative at the U.S. national level. At the state level, Twitter data representativeness is highly varied and such variability can be linked to individual state’s total population