With the 2014 Scottish Referendum [1] and with the 2015 Greek Referendum [2], opinion polls indicated that results might be "close" (say around 4%), but the former ended with a 10.3% lead, and the latter with a 22.6% lead. But why? Does this reveal a systematic problem with polling? Is it a problem with the sample (e.g. size, distribution, etc.)? Is it the nature of the questions in a referendum that make choosing a sample harder than in traditional elections?[1] https://en.wikipedia.org/wiki/Opinion_polling_for_the_Scottish_independence_referendum,_2014#2014
[2] https://en.wikipedia.org/wiki/Greek_bailout_referendum,_2015#Opinion_polls