Can someone answer some questions for me regarding Dr. Shiva's analysis? So in Michigan, you can vote in one of two ways, making a straight party selection or voting individually by race.
And their precincts report not only how many votes Trump (and Biden got), but what form those votes took, i.e., how many votes Trump received via voters making a "straight party Republican selection" vs. voters making an individual selection for Trump in the presidential race? Dr. Shiva's graphs looked at the total Trump vote percentage minus the straight Republican party percentage and graphed that against the Republican percentage of straight-party votes in that precinct? So if there were 200 votes in a precinct, 150 of which were via straight party voting, and Republicans (and thus Trump) got 100 of those votes and Democrats got 50 -- and then an additional 50 votes which were via individual voting and of which Trump and Biden each received 25, that would show up on the graph as a point at 66.7% on the x-axis (i.e., Trump's percentage of straight party voting, i.e., 100*150/200) and negative 4.2% on the y-axis (i.e., 62.5% - 66.7%)? Do I have that right?
The percentage of straight-party ticket votes for Trump is being used as a proxy for the "Republican-ness" of a particular precinct. Presumably you would get a similar result if you used different proxies for that quantity, e.g., the percentage of votes that were cast by voters who are registered as Republicans (or the percentage of total registered voters who are registered as Republicans)?
How well-established is the baseline as far as what this kind of graph "should" look like? They showed that Wayne county didn't have the pattern? Was that true of any other Michigan counties? What about counties in other states? Or the same or different counties in past presidential elections? That seems like a pretty key piece of information in terms of proving that there isn't an innocent / natural explanation for the phenomenon, no?
The problem is mostly in the fact that the effect is so linear. That is totally unnatural. I hope they put out their data or at least give us their linear regression results, otherwise this is all for nothing. We need to see their actual analysis in a data table so we can confirm.
They will probably give updates and I hope they are working with the Trump team.
You actually answered my question with your question!
Michigan I think is an edge case (although there may be other similar states) that break out straight-party vote versus individual ones. Therefore the slope indicating weighting and when it kicks in is fairly pronounced in Dr. Shiva's examples.
I imagine that we could use the percentage of total registered Republicans as a proxy for this, although it wouldn't be as exact. We know that Trump had high approval ratings in the Republican party, but of course it's not 100% and certainly not uniform from county to county. However, I think we could use a "fudge" number of 90% or so and if applied uniformly we could see similar adjustments being made by the voting system. As the number of registered Republicans in a county increased, if this weighting algorithm holds true, we should see that same downward slope kick in at a certain percentage -- again, just not as perfect a line as Michigan.
Can confirm. This is beyond the most disturbing thing I've ever seen.
Can someone answer some questions for me regarding Dr. Shiva's analysis? So in Michigan, you can vote in one of two ways, making a straight party selection or voting individually by race.
And their precincts report not only how many votes Trump (and Biden got), but what form those votes took, i.e., how many votes Trump received via voters making a "straight party Republican selection" vs. voters making an individual selection for Trump in the presidential race? Dr. Shiva's graphs looked at the total Trump vote percentage minus the straight Republican party percentage and graphed that against the Republican percentage of straight-party votes in that precinct? So if there were 200 votes in a precinct, 150 of which were via straight party voting, and Republicans (and thus Trump) got 100 of those votes and Democrats got 50 -- and then an additional 50 votes which were via individual voting and of which Trump and Biden each received 25, that would show up on the graph as a point at 66.7% on the x-axis (i.e., Trump's percentage of straight party voting, i.e., 100*150/200) and negative 4.2% on the y-axis (i.e., 62.5% - 66.7%)? Do I have that right?
The percentage of straight-party ticket votes for Trump is being used as a proxy for the "Republican-ness" of a particular precinct. Presumably you would get a similar result if you used different proxies for that quantity, e.g., the percentage of votes that were cast by voters who are registered as Republicans (or the percentage of total registered voters who are registered as Republicans)?
How well-established is the baseline as far as what this kind of graph "should" look like? They showed that Wayne county didn't have the pattern? Was that true of any other Michigan counties? What about counties in other states? Or the same or different counties in past presidential elections? That seems like a pretty key piece of information in terms of proving that there isn't an innocent / natural explanation for the phenomenon, no?
The problem is mostly in the fact that the effect is so linear. That is totally unnatural. I hope they put out their data or at least give us their linear regression results, otherwise this is all for nothing. We need to see their actual analysis in a data table so we can confirm.
They will probably give updates and I hope they are working with the Trump team.
You actually answered my question with your question!
Michigan I think is an edge case (although there may be other similar states) that break out straight-party vote versus individual ones. Therefore the slope indicating weighting and when it kicks in is fairly pronounced in Dr. Shiva's examples.
I imagine that we could use the percentage of total registered Republicans as a proxy for this, although it wouldn't be as exact. We know that Trump had high approval ratings in the Republican party, but of course it's not 100% and certainly not uniform from county to county. However, I think we could use a "fudge" number of 90% or so and if applied uniformly we could see similar adjustments being made by the voting system. As the number of registered Republicans in a county increased, if this weighting algorithm holds true, we should see that same downward slope kick in at a certain percentage -- again, just not as perfect a line as Michigan.