Post by northbruin40 on Jun 30, 2020 19:29:52 GMT -8
Was it just the case that Biden is increasing his lead in deep blue states, thus it has little impact on the electoral college? Was Trump losing support in deep red states, so that he will win by less there, but still little impact on the EC? Are the swing states more elastic than the national average so that the impact is magnified there?
Well I looked at the state polls over the last 3+ weeks that has been the time of a Biden peak and a Trump trough.
For the "deep blue states" I found just 3 state polls. Small sample size, but for that sample Biden averages 5.1% better than the 538 prediction for Clinton and 0.2% better than the actual 2016 results for Clinton. So the "making the rubble bounce" effect does not appear to be overly large. It should be noted that the 538 prediction tended to underpredict the real margin in 2016 in blowout wins, so sometimes it's better to compare polls vs. prediction rather than polls vs. actual result.
For the deep red states, there were 10 polls. Biden averages 5.6% better than the old 538 prediction, and 11.9% better than the actual result. So depending how you look on it, Trump is either proportionately losing support in the states he is overwhelming likely to win anyway or is especially losing support in states where it's probably not gonna matter anyway given the his lead there was so large.
For a few selected competitive states:
Michigan (11 polls) Biden averages 7.4% better than the 538 2016 prediction and 11.9% better than the actual 2016 result
.
Pennsylvania (4 polls) Biden averages 3.6% better than the 538 prediction and 8.0% better than the actual result.
Florida (7 polls) Biden averages 6.5% better than the 538 prediction and 8.3% better than the actual result.
Wisconsin (6 polls) Biden averages 2.7% better than the 538 prediction and 8.8% better than the actual result. The 538 prediction had a larger 2016 error here, and there was a rare Trump poll with him leading (1%) for this state.
Arizona (4 polls) Biden averages 6.2% better than the 538 prediction and 7.5% better than the actual result
N. Carolina (8 polls) Biden averages 2.2% better than the 538 prediction and 6.5% better than the actual result
Texas (3 polls) Biden averages 8.2% better than the 538 prediction and 8.7% better than the actual result
If Trump is looking for "good news" a case could be made that Wisc. and Penn. are less elastic than other states, and he may be losing less ground here.
I'm very curious as to how well the polls have been corrected for errors in the electorate in the Northern strip from Maine to the Rocky Mountains. Trump was routinely outperforming the polls by about 4% in that strip (I believe that region made for a net difference of about 1.6 million votes in the national popular vote compared to polls). The net error was considerably less elsewhere.
Well I looked at the state polls over the last 3+ weeks that has been the time of a Biden peak and a Trump trough.
For the "deep blue states" I found just 3 state polls. Small sample size, but for that sample Biden averages 5.1% better than the 538 prediction for Clinton and 0.2% better than the actual 2016 results for Clinton. So the "making the rubble bounce" effect does not appear to be overly large. It should be noted that the 538 prediction tended to underpredict the real margin in 2016 in blowout wins, so sometimes it's better to compare polls vs. prediction rather than polls vs. actual result.
For the deep red states, there were 10 polls. Biden averages 5.6% better than the old 538 prediction, and 11.9% better than the actual result. So depending how you look on it, Trump is either proportionately losing support in the states he is overwhelming likely to win anyway or is especially losing support in states where it's probably not gonna matter anyway given the his lead there was so large.
For a few selected competitive states:
Michigan (11 polls) Biden averages 7.4% better than the 538 2016 prediction and 11.9% better than the actual 2016 result
.
Pennsylvania (4 polls) Biden averages 3.6% better than the 538 prediction and 8.0% better than the actual result.
Florida (7 polls) Biden averages 6.5% better than the 538 prediction and 8.3% better than the actual result.
Wisconsin (6 polls) Biden averages 2.7% better than the 538 prediction and 8.8% better than the actual result. The 538 prediction had a larger 2016 error here, and there was a rare Trump poll with him leading (1%) for this state.
Arizona (4 polls) Biden averages 6.2% better than the 538 prediction and 7.5% better than the actual result
N. Carolina (8 polls) Biden averages 2.2% better than the 538 prediction and 6.5% better than the actual result
Texas (3 polls) Biden averages 8.2% better than the 538 prediction and 8.7% better than the actual result
If Trump is looking for "good news" a case could be made that Wisc. and Penn. are less elastic than other states, and he may be losing less ground here.
I'm very curious as to how well the polls have been corrected for errors in the electorate in the Northern strip from Maine to the Rocky Mountains. Trump was routinely outperforming the polls by about 4% in that strip (I believe that region made for a net difference of about 1.6 million votes in the national popular vote compared to polls). The net error was considerably less elsewhere.