by Stephen Fisher and Alan Renwick.

We have published two previous forecasts of the result of the UK’s referendum on EU membership, which are based on current vote intention opinion polls for this referendum and on the historical experience of referendum polls and referendum outcomes in the UK and on the EU elsewhere. The main ideas behind this approach are set out here. This post offers an update.

There is little change in our forecast since this time last month. After allowing for some difference between telephone and online polls, the relative levels of support for Leave and Remain have been remarkably steady. The online polls are much more common than telephone polls and they have been suggesting the race is neck and neck. The telephone polls are still showing greater levels of support for Remain than online polls. Our estimate of the current gap between the polls from these two modes is not much diminished from last month. After excluding Don’t Knows, telephone polls tend to put Remain 4.5 points higher than do online polls. Adjusting Remain up by half this amount in online polls, and correspondingly down in telephone polls, leads to a current polling average of 52% Remain, 48% Leave.

Although this is the same as last month after rounding, the underlying estimate is slightly more favourable for Remain. However, because we are closer to referendum day we are also projecting a smaller rise between now and 23rd June in support for Remain in the polls. These changes roughly cancel out and so our forecast for the referendum outcome is still 54% Remain and 46% Leave. Again this is the same as last month after rounding.

The probability of a Remain win is down from 73% last month to 72% now because the forecast share for Remain is slightly lower before rounding than it was last month.

Given we are a whole month closer to the vote it is perhaps surprising that the probability has not increased for a Remain win. The continued high level of uncertainty is due to the historical experience of referendum polls suggesting that there is still a lot of room for change in the level of support for each side. The prediction interval for the share of the vote is still plus or minus 13 percentage points.