Conventional wisdom is you should never mistake a poll for a prediction. Instead, a poll should be treated as a measure of opinion at a point in time. Whatever about in the case of general elections, the predictive power of polls for referendums has been called into question by the Seanad result.

With a week to go, the Yes side was comfortably ahead. The Irish Times/Ipsos MRBI poll conducted the previous weekend indicated a 24-point margin in favour of abolishing the Seanad. Then it vanished.

The gradual erosion of the Yes vote has always been a feature of referendum campaigns, dating back to the divorce referendum in the 1980s. But until recently the decline was slow and somewhat predictable. Nowadays it can, and often is, dramatic.

Driving this volatility is a preference for leaving any serious campaigning until the last minute, itself a reflection of a growing realisation that voters are simply slower to engage on some topics.

Notwithstanding some punches being traded over which savings estimates were more accurate, the media debate over abolishing the Seanad was slow to ignite. The televised debates came late in the day, long after the pollsters had put away their clipboards or put down their phones.

A lack of voter engagement is not just down to the media and the politicians. Lately, there has also been a lack of relevance to or understanding of the arguments voters are being asked to consider.

Prior to the final week of campaigning in the children's rights referendum, most voters claimed to be only vaguely or not at all familiar with the issues. Add to the mix the holding of referendums on legal matters such as Oireachtas inquiries and courts of appeal and most voters switch off. And when they eventually tune in, they are easily swayed because their minds are open and the arguments few, as illustrated by the successful intervention of the attorneys general in the Oireachtas inquiries referendum.

The more referendums the Government chooses to hold, the more difficult it will be to inject the enthusiasm and energy needed to sustain a debate for long enough for all arguments to be rehearsed.

It feels now like there is referendum every year, with sometimes two referendums on the same day. With so many changes to the Constitution proposed, and in the pipeline via the Constitutional Convention, referendum fatigue must now be a concern. The volatility that accompanies voter apathy looks like being with us for some years to come. The job of polling in referendums will not get any easier.

If we accept that the Seanad debate was slow to ignite, the question becomes why was the swing away from abolishing the Seanad? Previous polls seem to indicate that the swing tends to go one way, in favour of the No side and against government.

We can speculate as to the other reasons why the Yes campaign lost its advantage in the final week. The argument that the undecideds all went to the No side is too convenient and statistically improbable, which means some voters must have changed their minds. The televised debates may have been influential in this regard.

Confusion may also have played a part – anecdotally the two polling cards and a No vote meaning Yes to the Seanad were disorientating for many. This is not something to be ignored.

Looking ahead, the challenges are significant, for Government and for pollsters. Referendums that do not excite the public and the media because they are too irrelevant or too esoteric will increasingly hinge on a mood or a moment that arrives too late for polling companies to capture or for Government to control.