Multi-cycle forecasting of congressional elections with social media

Analyzing Political Communication with Digital Trace Data

“…Fink et al 2013;O'Connor et al 2010;Shi et al 2012), the share data in their relationship to various metrics of electoral success (e.g. Diaz et al 2014;Huberty 2013;Jungherr et al 2012). This has led various researchers to emphasize that case studies seemingly successful in finding a statistical relationship between some metrics in Twitter data and some metrics of electoral success should not be taken as indicators for the generalizability of their findings (e.g.…”

“…Diaz et al 2014;Gayo-Avello et al 2011;Huberty 2013;Jungherr 2013;Jungherr et al 2012;Metaxas et al 2011).…”

Can we vote with our tweet? On the perennial difficulty of election forecasting with social media

“…DiGrazia et al (2013) suggested that metrics based on message volumes alone could contribute value to forward-looking election predictions in US House races. However, Huberty (2013) shows that such volume metrics do not contribute any predictive value above and beyond a candidate's prior vote share when they are employed in true forward-looking electoral forecasts. The complexity of political speechparticularly on Twitter, where the 140 character limit leads to unique grammar -should throw doubt on the ability of such simple metrics to tease out real voter sentiment or intent.…”

Twitter in the Analysis of Social Phenomena: An Interpretative Framework

“…This necessarily leads to a rather limited number of observations which can serve as the basis for establishing a statistical relationship between Twitter messages and election results. This limits the potential for establishing a valid statistical relationship between the two metrics (Diaz et al 2014;Huberty 2013).…”