In this case the change side usually wins: Scotland 1979 (the measure failed because of a turnout restriction, but it was a Yes victory in raw votes); Scotland 1997 twice, on the establishment of a parliament and its tax-varying powers; Wales 1997 (narrowly); Greater London 1998; and Wales 2011. The no-change side has won only in Wales 1979 and north-east England 2004. That is an anti-status quo bias of 6 to 2. This is even more pronounced if Northern Irish referendums are included.

Consider how referendum campaigns work. Imagine some enthusiasts wish to outlaw lollipops. Normally, anti-lollipop activists will have a spectrum of views and support a variety of parties. However, on the crucial lollipop question, they unite for the sake of change. A referendum is called. This pressure group goes on to be the core of the Yes campaign. So, the change side starts the referendum possessing the formidable advantages of cohesion, motivation and the initiative. But they are almost certainly a minority. If there was widespread public demand to outlaw lollipops, every party would pledge it in its manifesto. There would be no need for a referendum. The task facing the Yes campaign, then, is to snap out of being a pressure group geared towards persuading politicians and become a campaign to persuade a majority of voters.

By and large people do not establish Save the Lollipop groups on the off-chance that someone might one day try to ban them. Anti-change campaigners start behind, forced to react to the Yes group. Before now their political differences have always been more important than their shared love of lollipops. Thus, anti-change campaigners must first learn to sink their differences. That is another source of delay as they catch up with their opponents. Although the anti-change side represents a potential majority, support will still have to be mobilised. Their ability to do so depends on how successful they are in assembling a unified No campaign.

Every referendum can be explained by the dynamic between these two factors: can the Yes campaign break out of its minority position; can a coherent No campaign mobilise its blocking majority? The two extremes were demonstrated in 2011. In the Welsh referendum in March it was not possible to form an effective No campaign. The change side received a free pass and won with 64 per cent. The AV referendum held in May saw a broadly-based No campaign drawn from Conservatives, Labour supporters and even advocates of other rival voting systems. The change side was pinned down on 32 per cent.