December 13, 2004
Much Smoke In Ohio. But Is There A Fire?
Much to-ing and fro-ing in Ohio:- The Ohio Supreme Court was asked to step in
- The state's electors, however, have already voted
- Once again, a Republican stepped in to prevent ballots from being counted.
- Ohio Recount Affidavit - Richard Hayes Phillips is an affidavit (by a hydrologist who claims statistical expertise) suggesting a pattern of vote fraud in Ohio. Whether it would be enough to alter the result is unclear, but if true it would get close. (As the site is hard to reach, I've quoted it in full below).
OHIO RECOUNT
AFFIDAVIT
December 10, 2004
From: http://web.northnet.org/minstrel/supreme.htm
I, RICHARD HAYES PHILLIPS, do swear and affirm the following:
1. I hold a Ph.D. in geomorphology from the University of Oregon. I am a professional hydrologist and am well versed in standard techniques of statistical analysis, with special expertise in spotting anomalous data. A copy of my curriculum vita is attached to this Affidavit as Exhibit A.
2. I have analyzed unofficial precinct level results from the November 2, 2004 general election in nine Ohio counties, including Cuyahoga, Franklin, Warren, Butler, Clermont, Miami, Montgomery, Hamilton, and Lucas. In have compared these results with those from the November 7, 2000 general election where such data is available. I have examined the unofficial and official results for the November 2, 2004 election, county by county. I have examined, in Franklin County, data on the number of voting machines deployed in each precinct. I have also examined United States census data for 2000 and 2003.
3. There are numerous examples of incorrect presidential vote tallies in certain precincts in Cleveland, Cuyahoga County. These irregularities include at least 16 precincts where votes intended to be cast for Kerry were shifted to other candidates’ columns, and at least 30 precincts with inexplicably low voter turnout, including 7.10%, 13.05%, 19.60%, 21.01%, 21.80%, 24.72%, 28.83%, 28.97%, and 29.25%, and seven entire wards where voter turnout was reportedly below 50%, even as low as 39.35%. Kerry won Cleveland with 83.27% of the vote to 15.88% for Bush. If voter turnout was really 60% of registered voters, as seems likely based upon turnout in other major cities of Ohio, rather than 49.89% as reported, Kerry’s margin of victory in Cleveland has been wrongly reduced by 22,000 votes.
4. The systematic withholding of voting machines from predominantly Democratic wards in Columbus, many of them with high black populations, severely restricted voter turnout in these wards and cost John Kerry 17,000 votes. I have meticulously compared election results with the number of registered voters per voting machine for each precinct in Columbus, and for each ward in Franklin County. In Columbus, the median Bush precinct had a 60.56% turnout, while the median Kerry precinct had only a 50.78% turnout. County wide, the 73 wards with fewer than 300 registered voters per machine had a 62.33% turnout; 58 were in the suburbs, and 54 were won by Bush. The 73 wards with 300 or more registered voters per machine had a 51.99% turnout; 59 were in Columbus, and 58 were won by Kerry. In addition, there were 68 machines not provided to anyone, according to data provided by the Board of Elections.
5. It has been widely reported that in Warren County, the administrative building was locked down on election night and no independent persons were allowed to observe the vote count. Based upon the official Board of Elections reports, there has been a 15.51% increase in voter registration in eight months time, and voter turnout was reportedly above 80% in 55 precincts. Since the 2000 election, voter registration was reportedly up by 79.0%, 38.3%, 32.4%, 31.0%, 29.7%, and 28.4% in six townships that provided 68.75% of Bush’s margin of victory in Warren County. While the county population has increased by 14.75% since the 2000 census, 87 of 157 precincts had shown declines in voter registration at other times since the 2000 election, and yet every single precinct, 157 of 157, showed increases in voter registration since March 2, 2004. In Butler County, there are nine precincts and two entire townships where Kerry received fewer votes than Gore despite a sharp increase in voter turnout; and there are precincts with reported increases in voter registration, since November 7, 2000, of 177.9%, 143.5%, 69.3%, 65.5%, 64.5%, 48.2%, 43.3%, 38.8%, 36.9%, 34.3%, 34.0%, and 33.8%, compared to an increase in population of only 3.12% county wide. In Clermont County, where the population has grown by 4.39% since the 2000 census, voter registration was reportedly up by 85.4% and 67.6% in two precincts, and down by 49.4% in another precinct, all in the same township; there were 23 precincts where turnout was up, but Kerry got fewer votes than Gore. All these data are indications that votes may have been shifted from Kerry to Bush. According to the official results certified by the Ohio Secretary of State, these three counties combined provided Bush with a plurality of 132,685 votes, which is 13,910 votes more than his statewide plurality of 118,775 votes. Given that George Bush carried these counties by 95,575 votes in 2000, the net loss for John Kerry could be as high as 37,000 votes.
6. It is my professional opinion that there is compelling evidence of fraud in Miami County. Early on election night, when 31,620 votes had been counted, and later, when 50,235 votes had been counted, John Kerry had exactly the same percentage, 33.92%, and the percentage for George Bush was almost exactly the same, dropping by 0.03%, from 65.80% to 65.77%. The second set of returns gave Bush a margin of exactly 16,000 votes, giving cause to question the integrity of the central counting device for the optical scanning machines. Compared to 2000, voter turnout increased by 20.86%, while the population increased by only 1.38%. Voter turnout was reported at 98.55% and 94.27% in two precincts in Concord, numbers nearly impossible to achieve. Voter turnout was reported to have increased by 194.58% and 152.78% in two precincts in Troy compared to the 2000 election, and by more than 30.0% in ten other precincts. There are no data for voter registration in 2000, so the ballots cast offer the only meaningful comparison. Comparing the results of the 2004 election to the results of the 2000 election, there is one precinct where the reduction in turnout exactly matched the reduction in votes counted for the Democratic presidential candidate. It is my professional opinion that these numbers are fraudulent, in that the true election results have been altered. Given that Bush officially carried Miami County in 2004 by 16,394 votes, and that Bush carried Miami County in 2000 by 10,453 votes, the net loss to John Kerry could be as high as 6,000 votes.
7. In Toledo, Lucas County, there were 50 precincts with less than 60% reported turnout. All of them were won overwhelmingly by John Kerry, by a margin of better than 5 to 1 in the aggregate. There were 45 precincts with more than 80% reported turnout; 12 were won by Bush, 33 were won by Kerry, and most were competitive. When the precinct numbers are combined into totals for each ward, data not provided by the Board of Elections, a clear and unmistakable pattern emerges. The 14 wards with the highest reported turnout were won by John Kerry by a margin of 11 to 7 in the aggregate. The 10 wards with the lowest reported turnout were won by John Kerry by a margin of 6 to 1 in the aggregate. It is my professional opinion that the election in Lucas County was rigged, most likely by altering the vote totals in each ward by a percentage chosen for that ward, plus or minus, based upon voting patterns in past elections. If turnout in Toledo had been as high as that reported elsewhere in the county, John Kerry’s plurality would have been 7,000 votes larger.
8. There are still 92,672 uncounted votes in Ohio, exclusive of any uncounted provisional ballots. According to unofficial results provided by the Ohio Secretary of State, there were 5,574,476 ballots cast, and 5,481,804 votes counted, which leaves 92,672 regular ballots (1.66%) still uncounted. The official results, now certified, do not include these ballots, but differ from the unofficial results only in the addition of provisional ballots and some absentee ballots to the tally. In Montgomery and Hamilton counties, these uncounted votes come disproportionately from precincts that voted overwhelmingly for John Kerry. In Montgomery County there are 47 precincts, all of them in Dayton, where the percentage of uncounted ballots is 4% or more. Kerry won all 47 of these precincts, by a margin of 7 to 1 in the aggregate. County wide in Montgomery County, the percentage of uncounted ballots was 1.70%. In Hamilton County there are 26 precincts, 22 of them in Cincinnati, where the percentage of uncounted ballots is 8% or more. Kerry won all 26 of these precincts, by a margin of 10 to 1 in the aggregate. Altogether there are 86 precincts in Cincinnati where the percentage of uncounted ballots is 4% or more. Kerry won 85 of these precincts, by a margin of 5 to 1 in the aggregate. County wide in Hamilton County, the percentage of uncounted ballots was 2.34%. Although I have not yet had time to examine similar data for Cleveland, Columbus, Toledo, Akron, Youngstown, Canton, or elsewhere, it is possible that the same pattern will emerge in these cities as well. If these 92,672 uncounted votes were cast for Kerry by a 5 to 1 margin, this would reduce the statewide margin between the candidates by another 61,781 votes.
9. There are still provisional ballots uncounted in Ohio. On election night the Ohio Secretary of State reported that 5,481,804 ballots had been counted, and 155,428 provisional ballots had been issued. According to the official results, now certified, 5,625,621 votes have now been counted, an increase of 143,817, which represents the number of newly counted ballots. Some of these were absentee ballots. The reported count of provisional ballots was 79,482 for Kerry, and 61,505 for Bush. This would leave 14,441 provisional ballots uncounted.
10. In summary, it is my professional opinion that John Kerry’s margins of victory were wrongly reduced by 22,000 votes in Cleveland, by 17,000 votes in Columbus, and by as many as 7,000 votes in Toledo. It is my further professional opinion that John Kerry’s margins of defeat in Warren, Butler, and Clermont counties were inflated by as many as 37,000 votes in the aggregate, and in Miami County by as many as 6,000 votes. There are still 92,672 uncounted regular ballots that, based upon the analysis set forth above of the election results from Dayton and Cincinnati, may be expected to break for John Kerry by an overwhelming margin. And there are 14,441 uncounted provisional ballots.
11. My research into the topics discussed in this affidavit is continuing, and I reserve the right to modify my conclusions as new information becomes available.
TO THIS I SWEAR AND AFFIRM,
___________________________________
Richard Hayes Phillips, Ph.D.
Posted by Michael : December 13, 2004 06:32 PM | Politics: US: 2004 Election | TechnoLinks
Slashdot
It!
I firmly believe that a full and open investigation of all results in Ohio should occur. (I personally think it unlikely that there was any state-wide conspiracy to steal votes, although there were probably localized instances of misconduct, generally to the benefit of the party controlling election monitoring, as is the case in most elections -- but I am certainly willing to change my mind on this issue if adequate evidence is presented.)
But the affidavit of Dr. Phillips is very far from offering any significant evidence whatsoever on the subject; indeed, it is the kind of amateurish product that makes it easy to tar those who want an investigation as kooks and crackpots. Leaving aside the issue of Dr Phillips' qualifications and focusing only on the substantive claims: [3] claims that the voter turnout in various counties is "inexplicably low"; yet there is no a priori reason to expect a uniform distribution of turnout. An analysis of standard deviation of turnout by county in previous elections, or in other states, would be a good start; from there one could move on to a detailed analysis of variance taking into account population density and other relevant factors, in order to see whether these Cleveland precincts were actually all that anomalous. None of this is done here. [4] is a real point, although methodologically very poorly done; why not use a standard correlation analysis to try to get a better picture of the relation between the inequities in voting machine distribution and other variables? But in any case unequal distribution of voting machines almost certainly does not provide legal grounds for overturning an election result. [5] again provides no point of comparison: what is the expected level of swings/divergences in registration over time and in different areas? There's no real way to know whether this is anomalous without a lot of statistical work. (And is there really no 2000 data)? [6] is just silly: in any large quasi-random sample, some data will hit "nice" numbers like 16 000; in itself, this isn't "compelling evidence" of anything. [7] is again badly designed and, even if it holds up, provides only one correlation that could be explained by any number of factors. As for [8] and [9], was any attempt even made to contact the relevant offices to ask about the uncounted ballots? Are they just spoiled ballots? If so, what level of spoilage is normal?
Dr Phillips claims to be "well versed in standard techniques of statistical analysis"; yet the affidavit employs none of them; nor does it employ the kind of methodological care, caution about seemingly "striking" results, and sensitivity to possible confounding variables one would expect of any scientist doing any sort of experiment. Indeed, it seems to employ very little except addition and multiplication in the service of a very selective presentation of a small number of data points with no useful context. Those who are serious about getting to the bottom of the election results in Ohio should realize that this sort of slipshod work can only hurt their cause.
Posted by: Evelyn Blaine at December 14, 2004 08:19 AM[3] claims that the voter turnout in various counties is "inexplicably low"; yet there is no a priori reason to expect a uniform distribution of turnout.
a uniform distribution of turnout would predict that some counties would have higher turnout and some lower. However, considering that Ohio was known (1) to be significant battleground state that (2) could swing either way, county-level turnout well below the national norm (7%-28% vs. 60%(?)) seems mighty suspicious.
[6] is just silly: in any large quasi-random sample, some data will hit "nice" numbers like 16 000; in itself, this isn't "compelling evidence" of anything.
[6] was a paragraph and yet you highlight one number from it that 'in itself' isn't evidence of anything.
The "Voter turnout was reported at 98.55% and 94.27% in two precincts in Concord, numbers nearly impossible to achieve." is much more compelling as evidence of possible fraud.
[7] is again badly designed and, even if it holds up, provides only one correlation that could be explained by any number of factors.
take a look at This for a fuller explanation of his reasoning. His reasoning seem legit enough to verify the toledo vote with an independent recount.
Those who are serious about getting to the bottom of the election results in Ohio should realize that this sort of slipshod work can only hurt their cause.
I dunno, I think we need to throw up what we can and see what sticks. If everything alleged in the affidavit sticks, Kerry would have won with a margin of over 30 thousand votes if I tallied the professor's numbers correctly.
Sure, the Toledo vote should be subjected to an independent recount. But going from that to "[i]t is my professional opinion [!] that the election in Lucas County was rigged" is ridiculous. On the question of voter turnout, I have no idea what maximal turnout levels for blocks of 400-1000 geographically contiguous voters are. Intuitively it seems quite plausible, in an election with overall regional turnout of 70% (that's the statistic for Cleveland as a whole), that some groups of that size could reach 90%+, through the combined effects of simple chance and historically persistent disparities in turnout (probably correlated in large part to economic and demographic factors.) There are lots of ways I might try to back this thesis up or falsify it: one of the most obvious starting places, for instance, would be to find the standard deviation of turnout by ward relative to precinct turnout (or state turnout, etc.) over a variety of sample sets (Cleveland 2004, Cleveland in the past 2 or 4 or 10 elections, all of Ohio for similar timescales, all of the US, all major cities ... ) I could also figure out correlations between turnout variance and a whole slew of relevant variables. At this point, we would have a lot of data to look at and could at least begin posing some interesting questions. But not only does Dr Phillips not do that, it never even seems to have occurred to him that it could or should be done.
"Throw[ing] up what we can and see[ing] what sticks" just means descending to a Rovian level. Those of us who went into the election proclaiming our allegiance to the "reality-based community" have a responsibility to do better; I'd like to think that that phrase is not just an anti-Bush slogan but represents a real intellectual value, which we're not going to compromise simply because it might lead us to conclusions we don't like.
Posted by: Evelyn Blaine at December 14, 2004 09:43 PM[3] claims that the voter turnout in various counties is "inexplicably low"; yet there is no a priori reason to expect a uniform distribution of turnout.
there is, however, an excellent reason to expect high turnout in those precincts that the democrats would have targeted for turnout efforts with the expectation of a large Kerry margin.
Your overall objection seems to be that Phillips did not complete a study that would take an independent researcher at least 18 months to 2 years worth of data collection in slightly over one month. Phillips provided the groundwork upon which a legal case could be built that would make it possible for a team of researchers to demand information from county officials in a very short period of time under a court order.
You haven't proven your case----where is the data that says that what Phillips is claiming did not happen? Why don't you prove that the APPARENT pattern of discrimination and vote suppression has a "reasonable" explanation.
You can start by explaining why the Secretary of State, who took office in 1998, suddenly decided in September 2004 to enforce a law that was on the books for ages which required that all voter registration forms be submitted on post card stock----and ordered ONLY that those people who registered using non-post card stock paper in 2004 be removed from the voter rolls, rather than require that everyone who had ever registered without using the correct thickness of paper be removed....
Posted by: paul lukasiak at December 14, 2004 10:16 PMAs I stated earlier, I firmly believe that a full investigation into the Ohio results should be conducted. My objection to Dr Phillips' analysis is not that his observations are, at best, crude, unsystematic, and preliminary; all science starts from crude, unsystematic, and preliminary observations which are then progressively refined. My objection is that he presents these observations as though they led inexorably to one and only one possible conclusion. The best that can be said for this is that, if it isn't simply intellectual dishonesty or self-delusion, it evinces shockingly poor understanding of the way research should be done.
Among a lot of people, there seems to be a great misunderstanding about what statistical analysis can and cannot demonstrate with respect to the election results. Statistical analysis, however refined, is not magic; it will not prove or disprove the existence of voter fraud. The most that it can possibly tell us is that, for some set(s) of data, some result(s) differ from what would be expected on the basis of a predictive model by a greater amount than can be accounted for by chance. (Obviously, how interesting this result is will depend on how strong and refined the model is: it might be based on a simple synchronic or diachronic superset of the data in question, or it might take into account a very large number of independent variables.) Once we have a result like relative to a suitably strong model, we can look at those data points and try to investigate causes for the divergence. One of those causes, to be sure, might be fraud. But we are still very far away even from that result. And sacrificing our skepticism and our willingness to think critically will not bring us any closer.
Posted by: Evelyn Blaine at December 15, 2004 12:05 AM"Throw[ing] up what we can and see[ing] what sticks" just means descending to a Rovian level. Those of us who went into the election proclaiming our allegiance to the "reality-based community" have a responsibility to do better; I'd like to think that that phrase is not just an anti-Bush slogan but represents a real intellectual value, which we're not going to compromise simply because it might lead us to conclusions we don't like.
Horsepuckey.
it's an acknowledgement of the reality that not everything may pan out...doesn't mean it shouldn't be checked out, as the flimsiest of evidence might actually be all that's visible of very real/widespread vote fraud.
If vote fraud was indeed perpetuated in Ohio, it is unlikely it was actually limited to JUST Ohio...
Posted by: Patrick (G) at December 15, 2004 02:42 PM