On 23 June 2016 most of the pollsters were wrong, and with good reason. Decades of political and economic complexity, cutting edge campaign techniques and unpredictable events conspired to produce one of the biggest shocks in UK electoral history.

But this wasn’t the only bad call. Both the 2015 and 2017 general elections wrong footed most pollsters as did Donald Trump’s election in 2016. So with another EU Referendum looming and the data showing a clear lead for Remain, it’s worth asking if the polls are getting it right this time?

Leading up to 23rd June 2016 there were already signs that something wasn’t right. We knew that telephone polling seemed to favour Remain while online polling favoured Brexit, so was it something about personal contact that made Brexit voters reluctant to say what they thought, or were there other factors in play?

The 2015 general election may give us a clue, as ‘shy voter’ theories gained traction after polling consistently under represented Conservative support. But in 2015 telephone polling, even with its personal contact, actually favoured the supposedly shy conservatives, with on-line polls favouring Labour.

But beyond margin of error and weighting, there may be a good empirical explanation for telephone polls favouring the Tories and being closer to the actual result. Most ground campaigners will tell you that Tory voters are more likely to be older or retired than Labour voters, therefore less likely to be on-line, more likely to be at home, pick up the phone, less likely to be shy and, critically, more likely to vote.

While that doesn’t dismiss the possibility of shy voters, it does mean the question is more nuanced. In the 2012 London Mayoral election Conservative voters in Labour majority London certainly weren’t shy about Boris Johnson, so ‘shy’ voting may be better explained through context rather than party or region.

In London in 2012 the context gave more license to Tory voters. Johnson was basking in Olympic glory, had bags of appeal, wasn’t identified as a traditional Tory and Ken Livingstone looked like yesterday’s man.

The 2017 General election context provided further clues. In this case it was Labour voters and ‘shy Corbynites’ that upset predictions. Again, it was those who’s party or leader suffered from negative portrayals in the media who were more likely to withhold their true voting intention.

So what of the context in which the EU referendum was set? The stakes were arguably higher than a general election, fear was a currency used by both sides, the campaign was more antagonistic and insults flew from the start.

Then it’s a question of who was more shy? With the establishment firmly behind Remain, it is reasonable to assume that some Brexit voters perceived their behaviour as the social risk in this case and chose to withhold their true voting intention. This may have been amplified by the Joe Cox tragedy, which appeared to damage the Leave campaign with Pound Stirling bouncing as the markets priced in a Remain win.

So are there comparisons with today’s polls? We may be months away from a second referendum, if at all, but the reaction to the shocking terror attacks on the Christchurch Mosques remind us that the context, sadly, hasn’t changed all that much.

For most Brexit voters the issues of immigration, border control and national sovereignty are still important. Yet these views have been increasingly described in pejorative terms such as racist, xenophobic or bigoted. So it isn’t a huge leap to postulate that many who sympathise with a ‘hard Brexit’ or ‘no deal’ may fear guilt by association and be at high risk of withholding their views from pollsters again.

The main counter to this is that most of the polling has been done on-line, so even if Brexit voters feel under siege, they would still feel able to express their views. This would of course depend on an individual’s perception that expressing their views on line would be less risky than on the telephone.

But even if the shy Brexiteer could be discounted, other on-line problems may recur. The fact that older voters are more likely to vote for Brexit and less likely to be on line is still an issue. So elderly voters in on-line sample groups may not be representative of elderly voters as a whole, which could leave pollsters with poor data on the largest group most likely to vote.

So while support for Brexit seems to have softened a little, the polls could still have a sting in their tale. Remember, while the final UK result in June 2016 was 51.9% to leave, with 48.1% to Remain, in England the margin was considerably wider at 53.4% to 46.6%.

So even if current polling is only a point or two out, a second referendum could give us a result where Wales joins Scotland and Northern Ireland as majority remain, while a majority in England, albeit by a reduced margin, still wants a ‘hard Brexit’. If that happens pollsters may have a whole new set of questions to worry about.