According to the 2016 election forecast, the chance of winning for Hillary Clinton is 71.4%. However, we already know the final result that DT is in the house. What makes the difference between the forecast and the reality? I asked some students; one of them told me that she abstained from voting because she doesn't like any of the candidates. And another possible explanation is that publishing the election prediction may influence the final result. There is a story of psychology about this kind of situation. If there is a car accident happened and only one witness saw that, there is a high chance that the witness would call the police for help. However, if there are a lot of witnesses, maybe none of them will call the police because they think other people may have already called for help. The same situation happened in the 2016 election as well. The voters might believe that the candidate they like got enough votes, so it doesn't matter if they vote someone else.
And the methods of doing research for the forecast is essential as well since different classes have a different lifestyle. Due to the survivorship bias, people tend to think that other people are similar to themselves. But the gap between classes is more significant than we think. Some people may never use the internet in their life, how could the result be accurate if the conclusion is drawn by internet research.