Evidence Against Algorithmic Vote Flipping (no fraud)

[sailingaway]

Sailing, i was being serious. The threads shouldn't spam each other.

If affa continues to post things i've already answered. Isn't that spam? Isn't that derailing the thread?

He is just using fancier words to mock and distort.

I've literally answered most of his points in my OP and FAQs. Not to mention in my prior threads.

the point of the mod comment was as much intended to stop the really unpleasant fighting in the other thread as anything. These views were linked into the other thread. Having a running fight side by side isn't a lot better than the other way, and since your rebuttal is linked in the other thread, appropriate people can find it. This thread is picking a fight, in some ways, and it may go over the line. So I flagged that, because if it does it will be moved.

 

[sailingaway]

However, you can't expect the anti flippers to expel the same energy as the pro flippers. By definition the anti-flippers would like to move on to more important things (to them).

that is the sort of mocking tone I was referring to, and no one is stopping you from moving on. You seem disinclined to do that.

 

[da32130]

Go ahead, continue to 'weight' your charts based on data hand picked by yourself to show what you want, even though the numbers you're choosing to 'weight' by have been proven to be statistically insignificant, unreliable, and highly volatile from election to election. Have fun with your thread, sir.

Affa, you have proved no such thing. I have already responded to the flaws in your analysis. But will try to repost it here.

 

[da32130]

that is the sort of mocking tone I was referring to, and no one is stopping you from moving on. You seem disinclined to do that.

What tone is that? It is a simple fact. I would like to move on. But I also want my side heard.

From my perspective (and you can agree to disagree), if the vote flipping argument is wrong, then it hurts us if it gains widespread acceptance.

I just want the other side heard.

 

[soulcyon]

However, you can't expect the anti flippers to expel the same energy as the pro flippers. By definition the anti-flippers would like to move on to more important things (to them).

The continue and move on, why are you wasting time? Now you're just sounding like an attention whore.

Fine, I'll play your game - why don't you go show your side of the story to a professor or statistics analyst?

 

[da32130]

the point of the mod comment was as much intended to stop the really unpleasant fighting in the other thread as anything. These views were linked into the other thread. Having a running fight side by side isn't a lot better than the other way, and since your rebuttal is linked in the other thread, appropriate people can find it. This thread is picking a fight, in some ways, and it may go over the line. So I flagged that, because if it does it will be moved.

My main point is I want both sides to be heard. I see the link you have. That may be enough. Thank you.

 

[da32130]

The continue and move on, why are you wasting time? Now you're just sounding like an attention whore.

Fine, I'll play your game - why don't you go show your side of the story to a professor or statistics analyst?

1) Because I want the other side to be heard. (if the vote flippers are wrong it could damage Paul's reputation)

2) Because I am one. We just need more demographic data.

 

[sailingaway]

What tone is that? It is a simple fact. I would like to move on. But I also want my side heard.

From my perspective (and you can agree to disagree), if the vote flipping argument is wrong, then it hurts us if it gains widespread acceptance.

I just want the other side heard.

I am going to discuss that with other mods. It is linked in one of the threads. The others are trying to work theirs out. Regardless presenting your view is one thing, fighting them side by side point for point and just not doing it in the same thread, and dismissing 'flippers' is attacking. MY view is that the fights make us look worse than either particular opinion.

 

[sailingaway]

The continue and move on, why are you wasting time? Now you're just sounding like an attention whore.

Fine, I'll play your game - why don't you go show your side of the story to a professor or statistics analyst?

don't you come and attack either. If he does, his thread won't last long here.

 

[soulcyon]

1) Because I want the other side to be heard. (if the vote flippers are wrong it could damage Paul's reputation)

2) Because I am one. We just need more demographic data.

We can never get the demographic data unless you change voting procedures. And Paul's reputation is in Paul's hands, not ours - we're not a cult.

 

[jjockers]

I appreciate da32130's research. Actually, his analysis is better than just about everyone else's work in the other threads. He's at least approaching the phenomenon from an (initially) unbiased, analytic problem-solving perspective (scientific method was emotion). I wish there was more of this perspective in these threads.

The primary flaws in da's analysis are:

(1) the analysis is done at the county level and therefore tries to explain a county-size::Romney vote % correlation. However, this correlation doesn't exist, so using other factors such as demographics to explain a non-existent correlation is a flawed premise

(2) the analysis weighs each of the other factors together as if they were mutually independent. A proper approach would be to perform a multiple regression with (and without) precinct size and the other factors as independent variables on Romney non-cumulative vote %. This analysis was done for SC and the results indicated that other factors can explain 2/3 of the phenomenon, but precinct size is necessary to explain the remaining 1/3. da has postulated that the remaining 1/3 can likely be explained by improving our demographic data. This may be true, but it's just conjecture until we can show otherwise.

da performs county-size analysis because we don't have demographics at the precinct level. This would be the ideal way if the phenomenon held at the county level (if county-size and Romney vote % were correlated). These variables are, however, not significantly correlated.

My analysis from .. a month ago? .. assumes that precinct demographics are similar to their respective county demographics. This is an approximation and certainly less than ideal, but it's the only method we have available due to data limitation. When da says we agree to disagree, he's referring to our assumption differences in this and the last paragraph.

Conclusion: Demographics can explain 2/3 of the precinct size::Romney vote % correlation in SC, but precinct size is currently necessary to explain the remaining 1/3.

 

[da32130]

MY view is that the fights make us look worse than either particular opinion.

I agree. The fighting isn't good.

However, as the daily paul front page post show and the responses in facebook to Ben Swann's inquiry show, the theory could grow substantially based on the dramatic results.

So I am weighing the fighting vs that future potential of widespread belief.

 

[sailingaway]

I agree. The fighting isn't good.

However, as the daily paul front page post show and the responses in facebook to Ben Swann's inquiry show, the theory could grow substantially based on the dramatic results.

So I am weighing the fighting vs that future potential of widespread belief.

regardless, you are not the last word on weighting that.

 

[da32130]

I appreciate da32130's research. Actually, his analysis is better than just about everyone else's work in the other threads. He's at least approaching the phenomenon from an (initially) unbiased, analytic problem-solving perspective (scientific method was emotion). I wish there was more of this perspective in these threads.

The primary flaws in da's analysis are:

(1) the analysis is done at the county level and therefore tries to explain a county-size::Romney vote % correlation. However, this correlation doesn't exist, so using other factors such as demographics to explain a non-existent correlation is a flawed premise

You are just talking about SC.

Read my part 5 in the OP. It explains this:

The county level analysis for SC and NH shows no vote flipping.

We can agree to disagree whether that analysis transfers to the precinct level.

I believe your precinct level regression on SC doesn't work because you only have county level demographics.

However, we have precinct level data in other states using Obama% and Libertarian%. Those work. We just need the income, age, data for the hard core republican areas.

My analysis predicts those will work based on the prior work at the precinct and state level.

(2) the analysis weighs each of the other factors together as if they were mutually independent. A proper approach would be to perform a multiple regression with (and without) precinct size and the other factors as independent variables on Romney non-cumulative vote %. This analysis was done for SC and the results indicated that other factors can explain 2/3 of the phenomenon, but precinct size is necessary to explain the remaining 1/3. da has postulated that the remaining 1/3 can likely be explained by improving our demographic data. This may be true, but it's just conjecture until we can show otherwise.

I have done this regression. And many others.

Your precinct level regression on SC doesn't work because you only have county level demographics. No reason to expect those to explain precinct level differentials.

da performs county-size analysis because we don't have demographics at the precinct level. This would be the ideal way if the phenomenon held at the county level (if county-size and Romney vote % were correlated). These variables are, however, not significantly correlated.

I have done the analysis at the precinct level in other states.

Also, the correlation does exist at the county level. So you could believe flipping exist at that level. However, demographics explains it.

My analysis from .. a month ago? .. assumes that precinct demographics are similar to their respective county demographics. This is an approximation and certainly less than ideal, but it's the only method we have available due to data limitation. When da says we agree to disagree, he's referring to our assumption differences in this and the last paragraph.

That assumption is way off to me. We can agree to disagree. Your analysis (to me) should not be expected to explain precinct level data. So it is no surprise it doesn't.

My assumption is that the summation of precincts should show the same slopes at the county level. It does. And the demographics explain it.

Conclusion: Demographics can explain 2/3 of the precinct size::Romney vote % correlation in SC, but precinct size is currently necessary to explain the remaining 1/3.

This is based on 1 state (SC) using an analysis that couldn't even be expected to work at the precinct level (only capture county level differences).

My analysis is based on 3 states, 2 at the county and 2 at the pricinct level. One (NH) a full state precinct level analysis.

 

[KingNothing]

I agree. The fighting isn't good.

However, as the daily paul front page post show and the responses in facebook to Ben Swann's inquiry show, the theory could grow substantially based on the dramatic results.

So I am weighing the fighting vs that future potential of widespread belief.

And thank you for that. We have a hard enough time battling the establishment's insults when we go after ACTUAL deceit and corruption. We don't need any more poison pills.

 

[da32130]

regardless, you are not the last word on weighting that.

understood.

 

[jjockers]

You are just talking about SC.

Correct. That's the state where we have the most demographic information (at least among early states)

We can agree to disagree whether that analysis transfers to the precinct level. Also, the correlation does exist at the county level. So you could believe flipping exist at that level. However, demographics explains it. My assumption is that the summation of precincts should show the same slopes at the county level. It does. And the demographics explain it.

No, there is no county-size Romney vote correlation. The graphs below show Romney vote % vs increasing precinct size and county size, respectively.

Note that the x-scale corresponds to bucketing percentage (on precinct size, 1 --> bottom 10%, 2 --> bottom 20%, etc | on county size, 1 --> bottom 20%, 2 --> bottom 40%, etc)

I should have known better than to ask you to 'take my word for it' . Look at those graphs. There is only 1 conclusion to draw: this phenomenon does not exist at the county level, so your primary assumption is completely incorrect (as opposed to a rough approximation).

I believe your precinct level regression on SC doesn't work because you only have county level demographics. Your precinct level regression on SC doesn't work because you only have county level demographics. No reason to expect those to explain precinct level differentials. That assumption is way off to me. We can agree to disagree. Your analysis (to me) should not be expected to explain precinct level data. So it is no surprise it doesn't.

All conjecture, though I agree to an extent. This assumption is unfortunately necessary to perform any demographics analysis on SC. While demographics for small precincts within a county may differ from the county overall, larger precincts are likely to reflect overall county demographics. The assumption is not ideal. It's a necessary approximation.

However, we have precinct level data in other states using Obama% and Libertarian%. Those work. We just need the income, age, data for the hard core republican areas.

I included this factor (McCain vs Obama 08) as an explanation item and it did not significantly reduce the precinct-size 1/3 explanation power. The file linked by Affa on page 2 has everything you need for SC analysis.

*Edited to add that even precinct demographics would be an approximation, because what we'd really need is actual voter demographics*

 

[affa]

The county level analysis for SC and NH shows no vote flipping.

You do realize that nobody is disputing this, correct? The anomaly only is present when evaluating at the precinct level. That's the point. You're examining the wrong thing, and have been from the start.

However, we have precinct level data in other states using Obama% and Libertarian%. Those work.

And I have shown you that libertarian data is statistically insignificant, highly volatile from election to election, and as a result completely unreliable. Generally averaging around .03 to .04 percent of total votes, precincts show wild fluctuation from election to election.

Additionally, I have also shown you that even Obama/McCain precinct results are not particularly reliable. Republican strongholds in 2004 were won handily by Obama in 2008. This means you'd be weighting them as 'democratic' precincts, when looking back in history this may not be so. This may be attributed to a number of factors, including but not limited to: redistricting, Obamamania, and Bush fatigue. Regardless of what caused it, relying soley on 2008 election data for demographic information is flawed.

I wish we had current, 2012 precinct level demographics so you could better do your study. I really do. But your insistence on using data like libertarian results from 2008 to weight your charts when they have been shown to be completely unreliable from election to election invalidates your work.

I have provided you with many examples.

Here's one:

London Bridge, VBC:
1984 - Reagan won with 76% of the vote.
1988 - Bush Sr. won with 70% of the vote.
1992 - Bush Sr. won with 49% of the vote (Clinton 30, Perot 20)
1996 - Dole won with 49% of the vote (Perot took 9%)
2000 - Bush won with 56%
2004 - Bush won with 59%

and...
2008 - OBAMA won with 52% of the vote.


Republican voters absolutely abandoned McCain - around 500 less votes than normal (a loss of about 1/3). Other precincts show either similar Republican abandonment, or, in some cases, massive Obamamania.

In 2009, by the way, the Republican running for Governor (McDonnell) won with 63%, so it's still 'Republican'.
In 2010, Rigell (Republican) won with 55% for the House of Representives.

There are many, many precincts like this.

I realize you're stuck because 2008 precinct level election data seems like it should be indicative of the voting population. However, it just isn't. 2008 was an odd election, and turnout was not normal. Using 2008 data to identify 'Democratic' precincts for use as demographics is flawed. 2008 is not necessarily a good indicator.

 

[da32130]

Correct. That's the state where we have the most demographic information (at least among early states)



No, there is no county-size Romney vote correlation. The graphs below show Romney vote % vs increasing precinct size and county size, respectively.

Note that the x-scale corresponds to bucketing percentage (on precinct size, 1 --> bottom 10%, 2 --> bottom 20%, etc | on county size, 1 --> bottom 20%, 2 --> bottom 40%, etc)

I should have known better than to ask you to 'take my word for it' . Look at those graphs. There is only 1 conclusion to draw: this phenomenon does not exist at the county level, so your primary assumption is completely incorrect (as opposed to a rough approximation).



All conjecture, though I agree to an extent. This assumption is unfortunately necessary to perform any demographics analysis on SC. While demographics for small precincts within a county may differ from the county overall, larger precincts are likely to reflect overall county demographics. The assumption is not ideal. It's a necessary approximation.




I included this factor (McCain vs Obama 08) as an explanation item and it did not significantly reduce the precinct-size 1/3 explanation power. The file linked by Affa on page 2 has everything you need for SC analysis.

*Edited to add that even precinct demographics would be an approximation, because what we'd really need is actual voter demographics*

Here is the standard chart per my 2nd post in this thread, analysis part 4.

standard analysis (using county data because demographic data is county data)
http://img706.imageshack.us/img706/6950/totalvote.jpg

What you notice is that Newt falls and Romney rises in the 2nd half of that graph. Newt falls in the whole graph.

My question is this (to Affa as well): why isn't that evidence of flipping?

In SC (hard core repub state in non democrat areas (last 2/3-1/2 of graph) and NH, VA (hard core repub in last 10-15% of graph) you need demographics. The indie and libertarian vote only matter in the other areas.

While using county income and precinct obama% demographics are all that is available in SC, I wouldn't expect either to add much value to the above problem in SC. So a precinct level regression doesn't move the ball forward for me. In that case the assumption is fatal for me.

Also, for Affa, I believe Obama% and Lib% do capture what they are suppose to capture. The data is enough. I'll try to provide that analysis later.

 

[jjockers]

Here is the standard chart per my 2nd post in this thread, analysis part 4.

standard analysis (using county data because demographic data is county data)
http://img706.imageshack.us/img706/6950/totalvote.jpg

What you notice is that Newt falls and Romney rises in the 2nd half of that graph. Newt falls in the whole graph.

My question is this (to Affa as well): why isn't that evidence of flipping?

Newt falls and the other three rise slightly. I can't say why conclusively or if that's a recurring phenomenon, as we haven't researched any county-size correlations. It is interesting, and should give the flippers pause when assuming the precinct-size phenomenon is fraud. Your analysis suggests the county-size graph can be explained via demographics. I agree. Romney's blip up at the end is due to Beaufort, Richland, and Charleston (higher median income), and Gingrich does better in the counties with low % white. This can be seen easily with the file Affa linked on page 2.

That said, it's clear from both our graphs that the precinct-size:romney vote % correlation is not reproduced by county-size, so any county-size graph explanations don't explain the precinct-size phenomenon.