Unlike in the US, where proper exit poll data is released to the public after each election so we have a good sense of how different groups in society actually voted, in the UK there is no official data beyond the overall election result. As my colleague Joe Twyman explains, predicting turnout in a referendum is a tricky business. This interactive represents our best estimate at the overall shape of turnout, based on different groups' self-reported likelihood to vote. But we'll leave the prediction to you: adjust the settings to what you think seems a realistic turnout chart and see how it affects the overall result.
The four dials on the left are:
1) Underlying poll number
This is the result before any turnout adjustment has been applied. It defaults to the latest YouGov polling (our last three polls combined) but you can adjust this to your best guess of the state of the race on the day.
2) Overall turnout
This is the overall percentage of the country who you think will vote in the referendum. Hint: 85% of people voted in the Scottish referendum but only 41% of people voted in the AV referendum in 2011 so it's likely to be somewhere between these two!
3) Age Skew
This is the degree to which older people vote more than younger people. Obviously at 100% turnout there is no skew because everybody votes, so it only comes into play as you bring turnout down. The basic shape is taken from different groups' self-reported likelihood to vote, at an overall turnout of 66% which was the level at the General Election in 2015, but you can double the age skew or reduce it to zero.
4) Class Skew
This is the degree to people in higher 'social grades' of AB and C1 vote more than people in C2 D and E. This is marketing speak for how educated and affluent people are. Notice how at the estimated level the skew cancels out the age skew, but only partially, meaning that the net effect is still to benefit the Leave campaign by a couple of percentage points.
The data for the levels of support within each group comes from the past three YouGov polls combined (sample size 5016).