Ask HN: Any research about decomposing sources of uncertainty in polls?

I've got a question for the data-science/stats side of HN that's proving difficult to google.

We're all familiar seeing the results of polls reported along with a confidence interval of +/- a few percent. I'm curious if there have been historical attempts to develop survey methods to decompose the source of that uncertainty.

In particular, it would be great if we could disentangle:

- uncertainty due to the generalizing from a random sample to a full population

- uncertainty due to the voters actually being uncertain what they'll do

i'm guessing that somebody tried this and found that getting people to accurately self-report partial support was too noisy, but i was wondering if there have been other attempts