Clinton won the popular vote by the amount that polls predicted. She lost the EC due to razor thin margins in some battleground states, which were well within the margin of error of polls in those states.
Everyone thinks 2016 was a miss for polls, it was not. It was a miss for forecasting models that had Clinton at 90%+ chances of winning while her numbers in many battleground states were really tossups.
There were definitely polls that tried to account for the EC, yet nearly everyone was still flat-footed when Trump won. I don’t see how the attitude should be “people interpreted the polls wrong!” if basically everyone did so.
We don’t tell users they have to type perfectly formed commands in order to open a webpage, and then if they don’t, format their hard drives. We build systems that try to assume users will not understand the underlying concepts and yet we still accommodate them. I see no reason that polls shouldn’t be designed with the same ethos. Until they are, I will assume anything a poll tells me is quite likely to be misleading.
And yes I understand statistics enough to know that polls were never meant to tell me exactly what the results will be, so please don’t lecture me on that…
Clinton won the popular vote by the amount that polls predicted. She lost the EC due to razor thin margins in some battleground states, which were well within the margin of error of polls in those states.
Everyone thinks 2016 was a miss for polls, it was not. It was a miss for forecasting models that had Clinton at 90%+ chances of winning while her numbers in many battleground states were really tossups.
There were definitely polls that tried to account for the EC, yet nearly everyone was still flat-footed when Trump won. I don’t see how the attitude should be “people interpreted the polls wrong!” if basically everyone did so.
We don’t tell users they have to type perfectly formed commands in order to open a webpage, and then if they don’t, format their hard drives. We build systems that try to assume users will not understand the underlying concepts and yet we still accommodate them. I see no reason that polls shouldn’t be designed with the same ethos. Until they are, I will assume anything a poll tells me is quite likely to be misleading.
And yes I understand statistics enough to know that polls were never meant to tell me exactly what the results will be, so please don’t lecture me on that…