Social media based forecasting techniques are very similar to traditional polls and surveys in many ways, except that they leverage much larger sample sizes in real time and with more limited data distortion. Using social media and alternative on-line datasets to forecast elections worked well in the 2016 US and 2017 French Elections as ZettaCap’s Social Media Influence (SMI) forecasts beat polls, betting markets, and traditional analysts in predicting election outcomes.
For the US 2018 Midterms, SMI generally followed consensus forecasts during July and August, giving Democrats a significant advantage. During September, Republicans benefited from the Kavanaugh Hearing which tended to stir the Republican base more than the Democratic base. SMI showed Republicans gaining but not taking the lead. The most significant impact occurred when the debate over immigration returned. The emergence of the Caravan and all of the related news stories greatly influenced SMI. However, these did not have sticking power and the Democrats regained almost everything they had lost.
It looked as though the Midterms would go to the Democrats until late October when the immigration debate hit an exceptional level. A second Caravan formed, US troops were sent to the border, and President Trump floated the idea of getting rid of birthright citizenship. SMI inverted to show the Republicans in the lead for the first time. ZettaCap’s current SMI forecast is for Republicans to win 221 seats in the House, a majority of +3 seats. An average of nine other forecasters shows Republicans with 209 seats, which translates into a Democratic majority of +8 seats. SMI is the only quantitative method showing Republicans maintaining control of the House.
One of the curiosities of the differences between SMI and other forecasts is the changes during the last month. Of those offering historical forecast data, the predominant trend is for Democrats to have gained whereas SMI shows Republican not only gaining but taking the lead.
An important point to highlight is that this election has a considerable unknown, or X factor. There has been a tremendous surge of interest in social media and search trends for items expressing interest in the Midterms and in voting. We assume an unusually high number of occasional and first time voters will turnout, creating fairly large differences from prior elections. If this is the case, many or most approaches based off of prior elections, such as polls, will be off. In this Midterms, expect the unexpected.