Inevitably when you have a 1,000 sample poll and then adjust for turnout after taking out the don’t knows the numbers you are left with can get very small. But I love the way ICM, in its general election datasets, is now breaking down constituencies by specific types.
The group where there’s the biggest variation across the different seat types are those saying they’ll vote Lib Dem. Notice how in the final column, the CON-LAB marginals, we get the smallest figures. Clearly this voting group has been squeezed. In the LAB facing CON held seats the figure drops to just 2%.
Elsewhere in the data, on a table not shown, we see how turnout looks set to be significantly lower in LAB heartland seats than CON ones. This, of course, is one of the main reasons why the electoral system appears to be biased to the party.
Of course you would love to see this sort of split in a much bigger sample poll but what we’ve got does provide a few pointers to what might happen.