"Asked for their views in an opinion poll prior to or at an early stage of the campaign, the pragmatists might well have shown sympathy for change. However, these views would not have been particularly strongly held and would have been likely to be susceptible to change. In short, the hypothesis would be that the centrists or pragmatists swung from YES to NO over the course of the campaign because they were open to persuasion. (...) It is essential to take the intensity and salience of voters' (and non-voters) opinions into account and (...) these two features are likely to vary systematically depending on voters' proximity to the centre of the distribution of opinion."
Richard Sinnott, in Cleavages, parties and referendums: Relationships between representative and direct democracy in the Republic of Ireland.
"Another explanation for the no-side's success is that they were able to set the agenda and claim the middle ground. In this light, the Danish referendum seems to fit the patterm know as "the opinion reversal referendum". Canadian psephologist Larry LeDuc has found that the yes-side in a referendum often loses if the no-side captures the centre ground. (...) Therefore, while the government is often ahead when the referendum is called, the no-side often closes the gap - if it succeeds in appealing to the median voter."
Mads H. Qvortrup, in How to Lose a Referendum: the Danish Plebiscite on the Euro.