How did we falsely predict the Brexit referendum result?

The outcome of the "Brexit" referendum surprises the "elite" class, especially many investors. Most of the previous polls turned to be in favor of remaining in the EU. Based on these polls, the market had been very positive before the referendum and risen sharply on the day before the 24th of June. As the result of the referendum showed that the "out" won by around 3 percent points, there must have been some biases about the poll surveys. Facebook and some other social networks showed that the "in" would dominate the "out". This is a very bias survey. Most on the social networks are relatively young and have not experienced the age before the UK joined the EU, this group of people is definitely in favor of remaining in the EU and is not a good sample to represent the entire population. Moreover, some Internet surveys have similar problems, some senior people use the Internet to receive information rather than to share their opinions. There is another problematic survey: doing a survey in London. From the referendum result, we can see that almost 60% of the Londoners support to remain in the EU while most other regions in England oppose to this idea. There are many other problematic methods, such as asking for company opinions, expert opinions and etc. To conclude, the failure of making an accurate prediction of the referendum is due to selecting a biased sample to represent the entire population. In addition, the process of running a survey (for example, running an Internet survey) could eventually lead to a biased selection of the sample.