Polling: Unskewed Polls and comments on polling (moved from Harris Thread)

My “comment on polling” is that the margin of error concept often cited makes little to no sense.

First, it is based on the idea that there could be up to a five percent chance an individual poll is outside the margin. The number five is arbitrary. If a post of mine, or any of us, had less than a fifteen percent chance of being wrong, IMHO that would be excellent by SDMB standards.

And most discussion incudes multiple similar or averaged polls. Add the sample sizes together, and the traditional margin of error gets really small.

ON THE OTHER HAND, there are bigger reasons polls can be wrong other than sheer sampling bad luck. In the U.S., the response rate is as low as one percent. This creates big to impossible possibilities for error compared to door to door surveys in nations where people are very friendly, commonly home, and most cooperate (Example: Bengladesh).

Polls are great for showing trends. Harris is doing better than Biden did. Comparisons, like that Harris is doing better in Pennsylvania than Florida – probably, but less sure. Who will win? Less sure yet. And pegging a firm number to those levels of uncertainty is impossible.
.
.