Secession

[pulque]

Originally Posted by Airbornelawyer

The greater variances are somewhat clustered in the direction of the op-scan counties, but when you take into account the other factors (rural voters, military voters, percentage of independent voters, etc.) suddenly there is no statistically significant trend that can honestly be correlated with the type of device used to count votes.

very impressive. I have too much "science" to do today to look closely, but will later on. thanks.

[Airbornelawyer]

Originally Posted by BMT

If I'm not mistaken the way Fl. election laws are written,in the primary elections you can only vote for a candidate from the party you are a member of.

Yes. Florida Statutes, Title IX, section 101.021

Originally Posted by BMT

Canidates in most local elections are Democrats and most of the people in a given county are registered Democrats.

Not entirely, but for the most part. 40 counties have more registered Democrats than registered Republicans. 27 have more registered Republicans than registered Democrats.

Statewide, registered Democrats exceed registered Republicans by 41.5% to
37.8% of all registered voters. 17.5% are registered without a party.

This represents a downward trend against the Democrats in Florida. In 1970, Democrats were 72.4% of registered voters. In 1980, 64.2%. In 1990, this had dipped to 52.2%. By 2000, it was 43.4%. Republicans have gone from 25.4% in 1970 to a peak of 41.9% in 1994. It has slipped to 37.8% today, but as the number has increased and the Democratic percentage has continued to drop, this percentage decline is apparently due to more non-party registered voters.

As for local elections, it depends on the region. Baker County, for example, is in the northeast of the state, just west of Duval County (Jacksonville). It is in the 4th Congressional District. No Democrat even ran against Rep. Ander Crenshaw (R-FL), who received 99.5% of the vote (he received 99.7% in 2002). It is in the 3rd State Senate District, where a Republican won 66% to 34% in 2004 and 55% to 46% in 2002. In 2002, the county voted for a Republican governor, lt. governor, attorney general and commissioner of agriculture, and in State House District 12, the Democrats fielded no candidate, and Rep. Aaron Bean (R-12) defeated a Libertarian candidate 84% to 16%.

[pulque]

Originally Posted by Airbornelawyer

The greater variances are somewhat clustered in the direction of the op-scan counties, but when you take into account the other factors (rural voters, military voters, percentage of independent voters, etc.) suddenly there is no statistically significant trend that can honestly be correlated with the type of device used to count votes.

I don't want to cherry pick, so I'm looking at the 20 counties you selected based on voting population size. I don't discount the idea that there may be specific reasons for the trend regarding the locality and demographics (eg military votes). But I just don't yet see a statistically significant correlation between percentage of otherpartyvoters and variance.

table of those counties ordered by %GOP variance:

[Airbornelawyer]

Originally Posted by pulque

I don't want to cherry pick, so I'm looking at the 20 counties you selected based on voting population size. I don't discount the idea that there may be specific reasons for the trend regarding the locality and demographics (eg military votes). But I just don't yet see a statistically significant correlation between percentage of otherpartyvoters and variance.

That's what you get for cherry picking the data. The drama of this woman's work is the huge variances - Republicans supposedly getting 2-7 times as many votes as they "should" and Democrats getting half as many or less. As I did by reordering the counties by population on an earlier page, showing that the dramatically large variances were clustered in small-population counties, so I do by reordering all counties by the share of Democrats and Republicans of all registered voters. See attachment.

Note that, as I already stated, all significant Republican variances are clustered at the high end, as are all significant Democratic variances. Also note that the e-touch voting counties are heavily clustered at the lower end.

I have no great idea why this is so. At first glance, it seems counterintuitive. One might expect that the greater variances would be in places where there is greater room to vary, i.e., in counties with larger percentages of independent voters. Instead, outwardly at least, the opposite appears to be the case. One explanation lies outside the realm of statistics. The counties which have the highest party ID are not evenly split between the 2 major parties. The top 15 counties by total 2-party ID - Liberty, Lafayette, Taylor, Madison, Calhoun, Gadsden, Holmes, Hamilton, Union, Gulf, Baker, Jackson, Franklin, Jefferson and Dixie - happen to occupy the 48th, 53rd and 55th through 67th places in the rankings of GOP percentages of registered voters. They also rank 1-15 in Democratic Party registration. In other words, these are not just rural counties with small populations, they are also yellow dog Democrat country. But now, they'll vote for the gray elephant, they just don't join his party.

They are also small counties; they occupy 42nd, 43rd, 52nd-55th, 57th-59th, 61st-64th and 66th-67th places in total registered voters.

In other words, given low GOP registration in these counties, the Republicans really had nowhere to go but up. Given these counties' conservatism, the GOP, especially when pitted against the most liberal member of the Senate, had every expectation of going way up, while the Democrats had an equal expectation to go down. And given the overall small population, any big absolute increase or decrease is going to look even bigger when expressed as a percentage.

BTW, of those 15 e-touch voting counties, 11 have more registered Republicans than Democrats. Consider that, corollary to the analysis above, one of the factors keeping the GOP variance lower in these counties was the fact that with stronger GOP ID already, there wasn't as much room to do better.

[Airbornelawyer]

In other words, BS in, BS out. A flawed methodology that assumes rather than shows that actual voting patterns should follow voter registrations. An inability or unwillingness to recognize the numerous alternative explanations for the data.

I originally was agnostic on whether Ms. Dopp was ignorant or mendacious. I have since learned that she purports to have an advanced degree in mathematics. Therefore, I am now disinclined to give her the benefit of Hanlon's Razor ("Never attribute to malice that which can be adequately explained by stupidity").

BTW, for an even more dramatic illustration of the disconnect between party registration and voting results, and of the power of smaller absolute numbers to make for really dramatic percentage changes, consider Ralph Nader.

In 2000, the Green Party of Florida had 2,728 registered voters, or 0.031% of all registered voters. 5,963,110 votes were cast in 2000. Applying Ms. Dopp's methodology, the Green candidate should have expected to receive 1,859 votes. Yet Nader received 97,488 votes. That is a statewide variance, again applying her methodology, of +5,144.1%.

In 2004, the Reform Party of Florida had 3,872 registered voters, or 0.0376% of all registered voters. As of the first set of unofficial returns, 7,609,810 votes were cast in 2004. Again applying Ms. Dopp's methodology, the Reform Party candidate should have expected to receive 2,860 votes. Yet Mr. Nader, running as the Reform Party candidate in '04, received 32,971 votes. That is a statewide variance, again applying her methodology, of +1,052.8%.

So either (1) Nader's people engaged in massive voter fraud or (2) people voted for Nader who were not registered with the parties whose nomination he carried. Do you suppose it's possible the same might have happened with the President?

[Airbornelawyer]

BTW, if this all really was part of the Dark Lord Karl of Rove's nefarious designs, he is not very good at this sort of thing.

The formula for the variance is v=(A-E)/E. Therefore, A=Ev+E.

The average GOP variance in the e-touch voting counties was +28.6%. Applying that variable to the total "expected" GOP vote for the op-scan precinct counties gives you 1,719,666 votes. The actual GOP vote for these counties was 1,950,213. The difference, then, is 230,547. Bush's margin of victory in Florida was 381,000 votes. If the secret cabal rigged those machines to win, the secret cabal needs a managerial shake-up, because this difference was not the margin of victory.

BTW, look at these numbers another way. Dopp's "surprising result" is the difference in the variance between e-touch voting counties and op-scan precinct counties and is the result of the DU-fed myth that op-scan technology was produced by companies (principally Diebold) headed by Republicans and therefore must have some secret election-theft chip.

So let us assume the +28.6% variance is the "proper" variance. That means that 1,719,666 votes should be the "expected" vote, i.e., we should expect the GOP to receive 28.6% more than the product of the GOP percentage of registered voters and the total votes cast.

Plugging 1,719,666 into the formula gives you:

(1,950,213 - 1,719,666)
1,719,666

The result is +13.4%.

So, in 52 counties, most of which are rural and more Republican-friendly than the state as a whole, and 8 of which - Escambia, Santa Rosa, Okaloosa, Walton, Bay, Duval, Brevard and Polk - host military installations, Bush did about 13.4% better than what might be considered the statewide standard. Suddenly the result does not appear so "surprising."

[Shark Bait]

Originally Posted by Roguish Lawyer

The Dems should get red and we should get blue. For obvious reasons.

Lets give the Dems yellow, for obvious reasons.

[Airbornelawyer]

Add another problem with Ms. Dopp's mendacious piece of pseudoscience - results from other states:

Massachusetts:

Total registered voters (as of 10/13/04): 4,098,634
Republican: 532,319 (12.99%)
Democrat: 1,526,711 (37.25%)

Total votes cast: 2,888,083
Bush: 1,067,163
Kerry: 1,793,916

"Expected" votes for Bush=(.1299)(2,888,083) =375,096
"Expected" votes for Kerry=(.3725)(2,888,083)=1,075,790

Variance for Bush=(A-E)/E=(1067163-375096)/375096= +184.5%
Variance for Kerry=(A-E)/E=(1793916-1075790)/1075790= +66.75%

Rhode Island:

Total registered voters (as of 3/2/04 Presidential primary): 639,459
Republican: 64,153 (10.0324%)
Democrat: 240,512 (37.6118%)

Total votes cast: 437,126
Bush: 169,045
Kerry: 259,753

"Expected" votes for Bush=(.100324%)(437126)= 43,854
"Expected" votes for Kerry=(.376118%)(437126)= 164,411

Variance for Bush=(A-E)/E=(169045-43854)/43854= +285.47%
Variance for Kerry=(A-E)/E=(259753-164411)/164411= +57.99%

More up-to-date statistics for voter registration in Rhode Island would probably lower that number somewhat, though perhaps not by much. Using the numbers above gives you a turnout of 68%, which is close to the 70% turnout estimated by Providence media.

So here you have two states where registered Republicans are heavily outnumbered by registered Democrats, and furthermore, two states decisively won by the Kerry campaign (though not as decisively as Gore in 2000, given Kerry's local boy status). Party ID on the whole is rather low, however, with half the electorate in each state unaffiliated with a party.

Given the high number of independents, positive variances are to be expected for both parties. Indeed, with half the electorate unaffiliated, if you assume voters turn out in the same ratios that they registered (which is not necessarily a reasonable assumption), and you assume independents break as affiliated voters do (again not necessarily true), then both parties should have a variance of +100%. In reality, registered independents are less likely to vote than affiliated voters, which would push the variances down, but in a Democrat-dominated state, registered independents will often include people who vote Republican but choose not to label themselves as such. This would tend to push the GOP variance up.

The actual results seem to confirm this, albeit with perhaps larger-than-expected variances. Both parties did better than "expected" but the Republicans did much better, with rural Florida county-type variances.

I have no idea what kind of voting machines they use in Massachusetts and Rhode Island, but that's the point. These outwardly dramatic variances are not uncommon and can be explained by many factors. You don't need nefarious conspiracy theories.

[Roguish Lawyer]

Pulque:

I think it is time for you to concede.

[Airbornelawyer]

Not to keep beating this into the ground, but I am not a mathematician, so I didn't notice this until now. Someone with an advanced degree should have noticed this a while ago.

Dopp supplies the following data:

a=% of registered voters who are Republican
b=% of registered voters who are Democrat
c=total number of registered voters
d=votes received by Bush
e=votes received by Kerry
f=total number of votes cast

Dopp applies the following formulas:

x="Expected" REP vote=(f)(a)
y="Expected" DEM vote=(f)(b)

REP variance=(d-x)/x
DEM variance=(e-y)/y

Here's the thing: (d-x)/x = (d/x)-1 and (e-y)/y = (e/y)-1

This stands to reason: if I received 150% more votes than I "should" have, then what I have is 250% of what I "should" have. If I received 25 votes, and should have gotten 10, I received 15 more. 15 is 150% more than 10. I have 25, which is 250% of 10. In math terms, (25-10)/10 = 25/10 - 1 = 1.5, which as a percent is 150%.

Bear with me: for the Republican, you now have (d/x)-1. But x=fa. So the formula is (d/fa)-1, which is the same as (d/f)/a-1 . We know what "a" is: the percentage of registered voters who are Republican. But we also know what (d/f) is: the percentage of votes cast for the Republican candidate.

The point is this: skip all the formulas and tables and "actual minus expected over expected" and all that. For each party, all you need is the percentage of the vote its candidate received and its percentage of registered voters. Divide the former by the latter and subtract 1. Express as a percentage.

A convoluted path (you can see why I'm not a math teacher either), but hopefully an easy example:

Republicans were 12.99% of Massachusetts registered voters. Bush received 36.95% of the vote. 36.95% divided by 12.99% is 2.845. Take away 1 and you have 1.845, or 184.5%. Two variables and 1 formula, instead of 6 variables and 2 formulas.

In my equations, the percentages are rounded, which builds in rounding error, but Dopp's builds it in as well, since she gives you the percentage of registered voters for the GOP and Dems rather than the actual numbers.

So while her tables look impressively full of numbers, to reach her conclusions you do not need most of the information provided, and if you really want to rigorously check the numbers, you need information she does not provide (the actual numbers of registered Democrats and Republicans in each county, so you can get a more accurate percentage).

Ironically, some of the extraneous information was useful for debunking her work. Having the total numbers of registered voters allows you to rank them based on size, and shows the strong correlation between smaller counties and higher variances. But note that this number is not used in any of the calculations. Having these totals and the total votes cast would allow us to calculate turnout, which may or may not be useful information, but she doesn't take that step. But turnout - like size (small vs. large county), voter ID (% of independents among registered voters) and demographics (rural vs. urban, military voters, etc.) - is just one more variable affecting the results.

Ms. Dopp is a poor mathematician - extraneous information and overcomplicated calculations to arrive at a result with built-in error - and a poor scientist - she seems intent on forcing the data to fit her hypothesis, regardless of its weaknesses. In the end, she is a poor sport.

[NousDefionsDoc]

So President Bush won, right?

[DunbarFC]

Originally Posted by Airbornelawyer

Add another problem with Ms. Dopp's mendacious piece of pseudoscience - results from other states:

Massachusetts:

Total registered voters (as of 10/13/04): 4,098,634
Republican: 532,319 (12.99%)
Democrat: 1,526,711 (37.25%)

Total votes cast: 2,888,083
Bush: 1,067,163
Kerry: 1,793,916

I have no idea what kind of voting machines they use in Massachusetts and Rhode Island, but that's the point. These outwardly dramatic variances are not uncommon and can be explained by many factors. You don't need nefarious conspiracy theories.

There are only 532,000 Republicans in Mass ? Wow. I knew I was in the minority but wow.........

In Mass it varies from town to town and city to city what kind of voting "machines" we have. In the city I live in now you use a scantron type sheet to fill in the arrow next to the candidate you want to vote for.

In Boston proper I believe they use lever based voting machines

In Little Rhodey - you vote how the godfather tells you to vote

[Airbornelawyer]

As far as I'm concerned, Ms. Dopp's horse is long since beaten to death, but ...

What have we learned? Dopp, having apparently heard the various Internet conspiracy theories about how Diebold and others were conspiring to rig voting machines, crunched the data the State of Florida was nice enough to provide and organized it by the one variable - type of voting machine - she cared about. Lo and behold! She found something that looked suspicious.

Many others, including - immodestly - yours truly , looked at her data and noticed anomalies. Those of us actually familiar with the State of Florida and its history and demographics were especially cognizant of these anomalies (I was born in West Palm Beach, Florida and until the age of 16 spent a grand total of 3 days outside of Florida. My family has lived in Florida for generations, stretching back to the Spanish colonial era and the migrations of Seminoles southward).

It was immediately apparent that Dopp ignored or downplayed a long list of demographic factors that might color the results. As previously stated, these included the size of the counties, the nature and level of voter ID and the nature of the counties' populations. It turns out that the salient factors were not the voting machines, but the fact that most of the counties with anomalous results were inhabited by those God-fearin', gun-totin', cousin-marryin', country-lovin' bogeymen of the New York Times editorial page and MSNBC commentators.

Had Dopp been a true scientist, she might have corrected for these factors or at least accounted for them. She might have did what I did, which was correlate the data by those other factors, or she might have established control groups. Any number of alternatives were there. An amateur like myself had no problem (though I've learned a lot about the "formula" feature of tables in Word in the past week). For someone ostensibly holding a Masters in Mathematics, it should have been a cakewalk.

BTW, as far as control groups go, note again several salient characteristics: southern counties with relatively small populations where registered Democrats outnumber registered Republicans by significant numbers. I looked around and many states did not provide sufficient publicly accessible data on voter registration, but at least one did: Louisiana.

In Louisiana, registered Democrats outnumber registered Republicans 1,515,211 to 651,794, or 56.0% to 24.1% of all registered voters. Louisiana was, however, far from a swing state. Bush won Louisiana with 56.7% to Kerry's 42.2% of the vote. That means the Republican variance was +135.3%, while the Democratic variance was -24.7%.

Louisiana does not have counties, but parishes. Parish results show similar patterns. For example, in tiny Acadia Parish, registered Democrats outnumber registered Republicans 23,875 to 6,715, or 66.1% to 18.6% of all registered voters. This is somewhat stronger Democratic Party ID than the state as a whole. In Acadia Parish, Bush beat Kerry 63.7% to 35.4%, a wider margin than that that by which he took the state. So the GOP variance was +242.7%, while the Democratic variance was -46.4%. Sounds like one of those Florida counties.

The state of Iowa, by contrast, is relatively evenly divided. Among registered voters, 31.06% are Republicans and 30.84% are Democrats. Bush's margin of victory over Kerry was about 10,000 votes (and Gore's margin over Bush was about 4,000), or less than 1%. Bush won Iowa with 50.02% to Kerry's 49.15%. The GOP's +61.0% variance was virtually indistinguishable from the Democrat's +59.4%.

California is not as evenly divided, as Registered Democrats outnumber registered Republicans by about 8.3 percentage points. But Kerry's margin of victory (+9.9) was not significantly different from the voter registration margin, so the variances were close: +28.2% for the GOP and +26.4% for the Democrats.

[Airbornelawyer]

BTW, the difference between the California and Iowa variances is explained by stronger voter ID in the former. Republicans and Democrats together account for only 61.9% of Iowa registered voters, but 77.7% of California registered voters. Thus, while independents in both states split for Bush and Kerry relatively evenly, Iowa had a greater percentage of independents to split.

[The Reaper]

Ouch!!

That is going to leave a nasty mark!

Who'd a thunk it?

AL, Science Guy!

TR