When creating a model to predict who will win the election we generally base our predictions about who will win by creating a survey and asking the public that “if a vote would happen today who would they vote for?” This is known as voter intentions however studies have shown that instead we should ask questions such as “who do you think will win the election?” This is known as voter expectations.
The theory for why this works is that if you are carrying out a survey on a small part of the population and ask them who they will vote for you restrict your sample size to those people. However if you were to ask who they think will win then the person answering will take into account their view and perhaps 20 of their friends views as well so you have managed to increase your sample size by a significant amount. Through the theory of large numbers, the larger our sample size the lower the chances of biases and we’re more likely to have an accurate estimate.
There is more information about the article attached below: