(Huge) delegate vote anomaly in Alabama verified

[parocks]

From a pure "compare to 2008" I'd want a

Gingrich 74%. (Giuliani was 4% more than Thompson - go give Giuliani 4% for being first on the ballot. Thompson and Giuliani were out of the race by that time, so presumably only hard core supporters, people who might know the actual delegates, would be voting for them, That "passion factor" drove up Giulianis numbers and Thompsons numbers in 2008. Gingrich would be a 70% candidate, like McCain and Romney in 08. Santorum would be more like Huckabee who was at 72%

Paul was at 75 last time, make it 75 again. Gingrich at 74 (70 + 4) - but there's that question about what happens when you plan to vote for everyone but quit. But the numbers do work with Gingrich at 74%. Romney 70 Santorum 70.

the numbers I have posted somewhere will give you an idea if the wrong votes behave in a reasonable manner.


Gingrich 74% of pres did dele - 1979 wrong voters
Paul 75% - 1791 wrong voters
Romney 70% - 1409 wrong voters
Santorum 70% - 1303 wrong voters

 

[affa]

The 2008 data had undervote rates for the major candidates (except Paul) of around 30%, so that's why I started out in the ballpark I started out in.

Again, looking at all precincts >100 voters.

If you independently correct all candidates to your proposed 70% conversion rate (aka, undervote of 30%), you'd need to correct each candidate by:

Gingrich 47333 votes
Paul 48867 votes
Romney 26960 votes
Santorum 34053 votes

As a percent of their own voters, that's 16.2% for Romney vs. 17.5% for Santorum.

Once again, we have Santorum gaining more 'bad voters' than Romney, or, possibly, more real Santorum voters voting for his delegates than Romney voters voting for Romney's, even though that flies in the face of the 'fatigue/turn ballot over' theories. Why?

Not to mention, you need to correct Paul by even more than Gingrich, even though Gingrich not only should get more 'Down the line' voters, but also more 'first on ballot' voters. It's close, though, but even if we ignore that Romney's still off.

Your solution solves that by... simply modifying each candidate by various variables ("lazy" rate, etc) that is fine tuned for each candidate to smooth the numbers that don't work into numbers that do work. That's rigging it. I can assign weights to those numbers that make it show anything too.

Maybe I'm just denser than usual today but Santorum *was* hurt the worst, in the sense that he had the fewest precincts with overvotes for delegates."

The vote count is far more important than precinct count. Santorum consistently does better than Romney at converting voters to delegate voters. This flies in the face of both the 'down the line' theory and the 'Santorum voters didn't flip the paper over' theory.

 

[dsw]

For me, it keeps coming back to this. This is so different than anything observed in the 2008 data. Why? What is different (besides the year)?

Same ballot style, same equipment, same state, same party, (mostly, because in 4 years some die, turn 18, move, etc.) same people.

Hopefully someone can check my work on this, but I wrote a program to check all the delegate races. In 2008 in Baldwin I found 5 races each with one excess delegate vote, for a total of 5 excess delegate votes. In 2012 in Baldwin I found 1,113 delegate races with excess votes and a total of 62,966 excess votes in all.

The total excess can't be explained by flipping of votes for candidates. You could start looking for an explanation of 2008 vs 2012 in terms of stolen votes. (It would would take fewer than 62,966 stolen votes, because that's the sum over multiple delegate races.) Or with 5 votes unexplained (an anomaly that could be fixed with one more vote for Paul in one precinct and one more vote for Thompson in another) maybe for some reason the people who configured the machines for 2012 neglected to make sure that the rule was enforce.

 

[dsw]

Are there any precincts in which other candidates did not have delegates running? That would certainly explain it -- for example, if only paul had a delegate slate in a particular district, obviously his first delegate would get far more votes than he would as a candidate.

As far as I know the delegate races were the same in all precincts. The delegate races also aren't between delegates of different candidates, they're races between two or more potential Newt delegates, or two or more potential Paul delegates, etc.

 

[affa]

As far as I know the delegate races were the same in all precincts. The delegate races also aren't between delegates of different candidates, they're races between two or more potential Newt delegates, or two or more potential Paul delegates, etc.

Correct. Look at the sample ballots posted throughout this thread.

 

[dsw]

You're responding to my comment about 2008, but are you talking about 2008 or 2012? The undervote percentages were posted earlier for 2008. They weren't identical. But for the ones (other than Paul) who had the largest percentages of votes, 30% is a good ballpark number.

Sorry, I still don't get what you mean by more bad voters for Romney, when Romney was the one with more overvotes (assuming we're talking about 2012). Santorum has almost no overvotes. So in what sense do you mean that he has more bad voters?

And finally, when you object to different "laziness" rates are you assuming that supporters for the different candidates all have comparable degrees of motivation and in such a way that ballot position should have no effect? Because that's the only way you would *not* expect to have different rates for the four candidates.


EDIT: You wrote: " Santorum consistently does better than Romney at converting voters to delegate voters. "

How are you calculating that? Santorum almost *never* had a delegate total equivalent to getting all of his supporters to vote in the delegate race. Why isn't that a low "conversion rate"? And where Newt and Romney had *more* delegate votes than candidate votes, why isn't that a *high* conversion rate?

EDIT: Looking at the first delegate for each, Newt had 27k too many votes, or a conversion rate over 100%. Mitt had 702 too many votes, or a conversion rate of almost exactly 100%. Santorum fell 18,796 votes short of a 100% conversion rate.

Again, looking at all precincts >100 voters.

If you independently correct all candidates to your proposed 70% conversion rate (aka, undervote of 30%), you'd need to correct each candidate by:

Gingrich 47333 votes
Paul 48867 votes
Romney 26960 votes
Santorum 34053 votes

As a percent of their own voters, that's 16.2% for Romney vs. 17.5% for Santorum.

Once again, we have Santorum gaining more 'bad voters' than Romney, or, possibly, more real Santorum voters voting for his delegates than Romney voters voting for Romney's, even though that flies in the face of the 'fatigue/turn ballot over' theories. Why?

Not to mention, you need to correct Paul by even more than Gingrich, even though Gingrich not only should get more 'Down the line' voters, but also more 'first on ballot' voters. It's close, though, but even if we ignore that Romney's still off.

Your solution solves that by... simply giving each candidate modifying each candidate by various variables (lazy rate, etc) that is fine tuned for each candidate to smooth the numbers that don't work into numbers that do work. That's rigging it.

 

[drummergirl]

Can you be more specific about what you mean by broken?

No, I think the machines were broken in 2012.

 

[drummergirl]

Hopefully someone can check my work on this, but I wrote a program to check all the delegate races. In 2008 in Baldwin I found 5 races each with one excess delegate vote, for a total of 5 excess delegate votes. In 2012 in Baldwin I found 1,113 delegate races with excess votes and a total of 62,966 excess votes in all.

The total excess can't be explained by flipping of votes for candidates. You could start looking for an explanation of 2008 vs 2012 in terms of stolen votes. (It would would take fewer than 62,966 stolen votes, because that's the sum over multiple delegate races.) Or with 5 votes unexplained (an anomaly that could be fixed with one more vote for Paul in one precinct and one more vote for Thompson in another) maybe for some reason the people who configured the machines for 2012 neglected to make sure that the rule was enforce.

Exactly. If you look at the Alabama presidential preference vote data only (no delegate contests) the Paul line is flat as a pancake. You do see some flipping, but from Gingrich and possibly Santorum to Romney.

But the delegate votes! holy cow! 5 excess votes in 2008, nearly 63,000 in 2012 in the same place?!? From what I understand from previous posts by Alabamans is that the rule was the same in 2008, but not enforced in 2008 either. There were just so few problems in 08 that it didn't matter.

So flipping doesn't explain that, a sudden surge in voter stupidity doesn't explain it, so what does? That is the $64 question.

 

[dsw]

From what I understand from previous posts by Alabamans is that the rule was the same in 2008, but not enforced in 2008 either.

Except for an error so small that it could be corrected with just two additional candidate votes, enforcing the rule in 2008 and not in 2012 would explain why the 2008 data looks the way it does. It may be the same machines but it's not the same configuration files. Simple human error.

 

[affa]

How are you calculating that? Santorum almost *never* had a delegate total equivalent to getting all of his supporters to vote in the delegate race. Why isn't that a low "conversion rate"? And where Newt and Romney had *more* delegate votes than candidate votes, why isn't that a *high* conversion rate?

Are you intentionally driving this conversation in circles?

You don't need to have 100% conversion rate to have a high conversion rate, heck, based on your own number, anything above a 70% conversion rate is a 'high' conversion rate.
I've never said Newt had a low conversion rate, in fact, I said his was 98.53%.

Looking at all precincts >100 voters, Santorum has
194794 votes and an average delegate vote count of 170408.67 for a conversion rate of 87.48%
.
Romney has
166316 votes and an average delegate vote count of 143381.27 for a conversion rate of 86.21%

This does not compute with either the 'down the line' voter theory, which, with fatigue, should give Romney more 'mistake' votes than Santorum, nor does it compute with the 'Santorum voters are too stupid to turn the ballot over', which should subtract 'correct' votes from Santorum. Both of those theories in combination should jack up Romney's conversion rate, and hurt Santorums, but that is not what is seen.

And correcting all candidates to 70% conversion rate shows the same thing - Santorum converts more voters than Romney, which makes no sense based on the ballot, nor on any of the theories put forth so far.

 

[affa]

Except for an error so small that it could be corrected with just two additional candidate votes, enforcing the rule in 2008 and not in 2012 would explain why the 2008 data looks the way it does.

Completely disagree that ballot mistakes on 2012 explain 2012. Yes, it would bring some numbers 'more' in line, I agree with that. But there are still significant oddities present.

 

[dsw]

We are *totally* not communicating here. FWIW what you're describing as "correcting all candidates to a 70% conversion rate" makes no sense relative to what I was saying, and I don't think the way you're calculating what you call "conversion rate" is meaningful in the first place. Let's just agree to disagree.

Are you intentionally driving this conversation in circles?

You don't need to have 100% conversion rate to have a high conversion rate, heck, based on your own number, anything above a 70% conversion rate is a 'high' conversion rate.
I've never said Newt had a low conversion rate, in fact, I said his was 98.53%.

Looking at all precincts >100 voters, Santorum has
194794 votes and an average delegate vote count of 170408.67 for a conversion rate of 87.48%
.
Romney has
166316 votes and an average delegate vote count of 143381.27 for a conversion rate of 86.21%

This does not compute with either the 'down the line' voter theory, which, with fatigue, should give Romney more 'mistake' votes than Santorum, nor does it compute with the 'Santorum voters are too stupid to turn the ballot over', which should subtract 'correct' votes from Santorum. Both of those theories in combination should jack up Romney's conversion rate, and hurt Santorums, but that is not what is seen.

And correcting all candidates to 70% conversion rate shows the same thing - Santorum converts more voters than Romney, which makes no sense based on the ballot, nor on any of the theories put forth so far.

 

[The Man]

To date, all of the “vote flipping” theory is based on the comparison between low and high vote total vote percentages. Whether you are a believer or not, many of us have devoted hundreds of hours to proving and disproving vote fraud.

The graphs below make NO camparison between low and high vote total precincts. The analysis is TOTALLY AND INDEPENDENTLY based on the differential between the vote reconstruction graph using reported vote totals AND the delegate reconstruction graph using the average of ALL reported delegates in each precinct for each candidate's delegate "score". Although all of the graphs below utilize this averaging method for computing delegates, I have constructed all of the below graphs using delegate 1, delegate 2. delegate 3, and variations of weighting. Every single method I used demonstrated the same anomaly that is shown below;The rate at which Romney received new precinct votes increases by 4.2% (of the overall votes cast) at vote count 300k, or precinct votes cast = 850. Also from this precinct and greater which includes more than 250 of the largest Alabama precincts, Ron Paul receives less than 100 extra votes total in precincts with more candidate votes than delegate votes. The largest 100 precincts, Paul NEVER has a single precinct where he has a single vote more than delegate votes! In those same precincts, Santorum and Romney consistently receive hundreds of candidate votes versus delegate votes in each precinct.

It's beyond me how to begin to explain this ANY other way besides vote theft. It appears to corroborate the notion that Romney is receiving votes that are not intended for him in larger precincts. Note that graph totals will vary depending on the delegate score methods used, but what DOES NOT change is that ]At 300k total votes with precinct delegate totals arranged in ascending vote total order, the rate at which Romney receives new votes increases by 4.2% of the Total Vote.

Note that at 300k total votes, Romney's new vote receiving percentage instantaneously increases by 4.2 percent in relation to the TOTAL vote, which gains him 12k additional votes.

 

[affa]

We are *totally* not communicating here. FWIW what you're describing as "correcting all candidates to a 70% conversion rate" makes no sense relative to what I was saying, and I don't think the way you're calculating what you call "conversion rate" is meaningful in the first place. Let's just agree to disagree.

It makes complete sense. You tried to apply drastically different 'lazy' rates to each candidate, along with another variable, to smooth data. I took your assumption, and applied it directly (without making up per candidate weights to tailor the data to what i want it to show) to see what we got.

And 'conversion rate' is a very, very simple and direct term: Average Delegate Votes / Votes for Candidate.

No different than, say: Number of People Buying Something / Number of People Entering Store equals your conversion rate of shoppers into buyers.

It directly measures the ratio of voters for, say, Romney, to the voters for his delegates. It doesn't care if it's a correct vote or an 'incorrect' vote, but it lets us see how each candidate compares to each o, ther very easily. Gingrich and Paul fit the 'idiot voter' argument' to differing degrees, but Romney and Santorum don't, which means the theory doesn't fit. Santorum has a better conversion rate than Romney, which makes no sense when the entire 'idiot voter' argument hinges on Romney being 3rd on the ballot, and Santorum being 4th on the ballot + on the flip side of the ballot. Santorum should have the worst conversion rate, by far.

Liberty's shown other various flaws in the 'idiot voter' argument, I'm just approaching it from a different angle and finding more inconsistencies.

 

[drummergirl]

It's beyond me how to begin to explain this ANY other way besides vote theft. It appears to corroborate the notion that Romney is receiving votes that are not intended for him in larger precincts. Note that graph totals will vary depending on the delegate score methods used, but what DOES NOT change s that [/COLOR]At 300k total votes with precinct delegate totals arranged in ascending vote total order, the rate at which Romney receives new votes increases by 4.2% of the Total Vote.

WOW!

 

[affa]

Wow to #2, especially. Just wow.

 

[The Man]

It's beyond me how to begin to explain this ANY other way besides vote theft. It appears to corroborate the notion that Romney is receiving votes that are not intended for him in larger precincts. Note that graph totals will vary depending on the delegate score methods used, but what DOES NOT change s that [/COLOR]At 300k total votes with precinct delegate totals arranged in ascending vote total order, the rate at which Romney receives new votes increases by 4.2% of the Total Vote.
WOW!

So here is what I'm talking about. Each of these three graphs plots the difference in every precinct betwen the candidate's vote total in that precinct AND his delegate total (averaged between ALL delegates for that candidate). The precincts are arranged from lowest vote total to highest left to right as usual. There are 1,864 precincts shown on each one of these graphs. Look at Paul's graph and see how many times the red line rises above the X-Axis, or the zero point. Look at Newt's graph, who is unaffected in this particular race by vote fraud. If the delegate "errors" were random, all 4 candidates' graphs would look like Newt's, with a fairly balanced portion of positive and negative points in relation to the Zero point. Then look at Santorum and Romney's graphs. Santorum clearly stays above the zero point in most precincts from start to finish because, as I previously showed, he is siphoning from Paul in small and large precincts. Romney, on the other hand, increases his vote vote total by 4.2% at a vote count of 300,000 and can be seen in the graph below, although not as clearly as in post 493.

All 1864 precincts are plotted in each of these graphs above. Notice how Paul's "votes minus delegates" almost NEVER gives a positive point. Notice how Newt's graph looks like you'd expect from "random" delegate vote error, keeping an average gain/ loss at zero. Then look at how Santorum usually gains votes versus his delegate totals. Romney ascends above the zero axis at 300k total votes when his new vote receiving percentage increases by 4.2%. EACH precinct calculation of 'candidate votes minus delegate votes' is plotted on these graphs.

 

[drummergirl]

All 1864 precincts are plotted in each of these graphs above. Notice how Paul's "votes minus delegates" almost NEVER gives a positive point. Notice how Newt's graph looks like you'd expect from "random" delegate vote error, keeping an average gain/ loss at zero. Then look at how Santorum usually gains votes versus his delegate totals. Romney ascends above the zero axis at 300k total votes when his new vote receiving percentage increases by 4.2%. EACH precinct calculation of 'candidate votes minus delegate votes' is plotted on these graphs.

I don't know how anyone (who understands what that means) can look at those graphs and think that is an accident.

 

[tremendoustie]

It's just as irrational, if not more irrational, to assume fraud is unlikely or impossible when voting is not remotely transparent.

There's a big difference between objecting to people instantly crying "fraud" for every data set that doesn't match their expectations, and saying that no fraud occurs anywhere ever.

If a kid cousin of mine declared that every single footprint she saw was a lion track, I'd object. That doesn't mean I don't think lions exist.

We aren't jumping to conclusions, we're spending tens of hours studying it and making a case for it.

You're spending tens of hours trying to make the facts fit the result you want to see, not tens of hours really trying to understand the reasons for the data.

It's the people stating fraud is impossible or jumping to the 'demographics' caused it argument in the face of all evidence that are leaping to conclusions.

The "flipping" b.s. is absurd. Sorry, it is. If you asked me to draw what I'd guess the plots would look like, sorting precincts by size, I'd draw exactly that -- guaranteed, depending on the region, certain candidates will do better or worse in districts near or in certain municipalities, than they will in the surrounding countryside.

I would bet 20 to 1 that fraud has occurred during this primary, at least once, somewhere. There's a big difference between that and supposing that any plot which doesn't immediately match what you imagine it should be is "mathematical proof" of some giant multi-state fraud conspiracy.

There are so many basic -- really basic logical mistakes in the things people in these threads are saying, it's embarrassing.

It's like people who immediately believe in conspiracy theories, even if those theories don't even come close to fitting or explaining the facts.

We're looking at the data. You aren't. No need to insult us.

The kid saying every footprint came from a lion is "looking at the data" too. Have you ever heard of the "boy who cried wolf"? There are wolves. You folks calling everything with four legs a wolf doesn't help our credibility.

Ok, back to the point at hand: Unlike the other data, this data does seem unexpected to me. I'm not going to jump to declare "fraud", because:

1. I don't even understand it yet -- for example, why did other candidates also sometimes get more votes for delegates?
2. Fraud would most likely affect delegates too. It'd be trivial to add that in.

 

[tremendoustie]

Holy crap, 50 pages?!

Jeez.

Has anyone actually looked into why any candidate would ever get more votes for their delegate than for themselves? Plotting the data 14 ways to sunday is an exercise in time wasting if you don't even understand the data in the first place.

Were there districts where some candidates did not have delegates running?