A notable feature of constituency polling thusfar is that with the exception of the Gedling poll, they have all shown significantly larger absolute or relative declines (i.e. where the drops have been lower in percentage-point terms, they have been in constituencies where the vote was lower to start with; often a higher proportion of the vote in such seats is shown as lost) in Labour support than suggested by national polling at the time they were conducted. Given the diversity of constituency polled, this is, how shall we say, something of a red flashing light as to their likely accuracy.
How do they generate the constituency polls anyway? Landline calls, then weight them for age?