Election 2016 Intro: Summary of Two Possible Scenarios
There is a significant divergence between two groups of indicators in calling the 2016 US Presidential Election. Only one of these groups of indicators can be correct as the difference between their projections is so vast. The outcome could have a dramatic impact on future election analysis and could change much of the political analysis industry.
The first group uses traditional methods and metrics and is focused on political pundits, analysts, and betting markets analyzing news and polls to forecast the election. This group has for over the past year called the general election heavily in favor of Clinton with some calculating the probability of a Clinton win at 80% and above with others declaring a coming landslide victory for Clinton. However, traditional methods have had a terrible track record this election season having completely misdiagnosed both the Democratic and Republican nomination races by discounting Trump and Sanders.
The second group uses newer data and metrics and is focused on social media and on-line activity to forecast expected election results leveraging the idea that voters’ unfiltered actions are better indicators of their voting intentions than polls which are more open to error and manipulation. This group clearly sees Trump having the dominant position and forecasts a fairly easy win for him in the general election. These newer data and methods successfully forecasted a Trump victory in the Republican primary before voting began and warned of a very close race between Clinton and Sanders, with Clinton winning. This appears to be the only quantitative method to correctly forecast the main parties’ primaries outcomes prior to the start of voting.
Two main scenarios emerge for the outcome of the general election, either:
1. Polls and Pundits are Correct – Clinton will win the election, perhaps in a landslide. Social media and on-line activity showed that Trump received much more interest and traffic on-line. However, it turned out that this was mostly due to the fact that Trump was a curiosity and people wanted to know more for entertainment not because they intended to vote for him. The polls were rock-solid. Turnout more or less replicated 2008 and 2012 levels. There was no Social Desirability Bias as people expressed their support for Trump in polls without regard to what others might think. The negative attacks on Trump were not exaggerated, just politics as usual. This election was just like those before it and the traditional methods for analysis proved extremely useful once again. Moral of the story => social media and on-line activity are sideshows and people don’t change a poll or survey response just because they might feel uncomfortable, leave politics and political analysis to the professionals.
2. Social Media and On-Line Activity are Correct – Trump will win the election. People showed consistently throughout the primaries and the general election what their true opinions were by how they spent their time on-line and how they interacted with social media. These unfiltered indicators consistently favored Trump. The polls were inherently biased in favor of Clinton as they were partially based on the results of the last two elections which saw very enthusiastic turnout for Obama which was not replicated in 2016 for Clinton. Also, polls inherently favored Clinton as many Trump supporters became shamed against publicly supporting him in polls due to an unusually strong Social Desirability Bias. The media, the sitting president, and many popular celebrities perhaps unknowingly created the environment for such a bias as they strongly attacked Trump like no other candidate in recent memory labeling him as a ‘Hitler’, an ally of the KKK, a misogynist, a racist, a tax cheat, a conman, a bombast, and someone who should not have access to nuclear codes. This Social Desirability Bias caused many Trump supporters to change their responses in polls throwing them off by more than the margin of error. This created a self-reinforcing cycle where negative attacks lowered live interviewer based poll results for Trump which just emboldened more negative attacks. All the while, betting markets incorporated the data showing that Clinton would win easily and Clinton ‘stock’ skyrocketed. Moral of the story => it is more important to see what people are doing and how they are spending their time in an unfiltered environment than to ask them what they intend to do in a public environment, especially during a very emotional election.
Only one of these can be correct.
The first scenario is the status quo. The pundits keep their jobs and the polling-industrial complex remains intact with no changes.
The second scenario is more or less where I place myself. The power of social media and on-line activity to forecast future actions and of emotion and social pressure to influence decision making are significant and should increase in an interconnected world.
A potential counter argument to the social media and on-line activity analysis is that much can be explained away by more curiosity surrounding less traditional candidates. However, if this were the case you would expect such activity to be very spotty with spikes around curious events (like a Trump gaffe). But what we have seen is consistently more interest for all things Trump over long stretches of time. Additionally, there is considerable proof that polls have significant amount of bias involved which go a long way in explaining why polls can be so pro-Clinton and on-line activity data can be so pro-Trump.
With this as a backdrop, please enjoy the posts.
With this as a backdrop, please enjoy the posts. For comments or specific questions regarding the social media and on-line activity analysis, please use kevin@zettacap.com. For all other inquiries, feel free to contact me via kjcoogan@yahoo.com or on twitter at @kjcoogan.