NC Unofficial Audit Part 2: Inside the Black Box

[ignatzmouse]

In part one of my analysis of the North Carolina General Election, I identified a discrepancy in voting patterns between the first third of the election via absentee/early vote and the other two-thirds of the election at the November 2nd polls. In the follow-up, I look deep into the county data to try to locate the source of the discrepancy and ask whether or not a solid case can be made for a modified election. The conclusions I've reached are as follows:

I. Most of the discrepancy can be attributed to optical scan and punch card counties. A large question remains as to how much of the discrepancy is the result of central tabulating computers, "spoiled ballots," or higher Republican turnout. The evidence shows that punch cards and certain optical scan vendors had a much lower percentage of their votes cast in the early vote than were cast by other voting systems. Why?

II. Older, manual systems (hand counted paper ballots & levers) showed a precise correlation in the early and election day votes AND with the exit polls.

III. While the DRE electronic voting systems had only a two percent disparity between the early vote and election day, there is stunning evidence to suggest they corrupted BOTH sides of the election.

*************************************************************************

The problem with side by side comparison of North Carolina county data is that it is damn near impossible to profile. Not only are there widely different population sizes to deal with, but the state has a plethora of voting devices and vendors ranging from touchscreen DRE's to levers to punch cards to optical scanners to hand counted paper ballots. (I think there is even one small, sparsely populated county that votes by carrier pigeon. They're very accurate, but by law the poll workers must be hosed down before leaving the building.) I thought a great deal over a solution to get around the complexity, and then I realized that the diversity was a blessing in disguise. It allows us to compare different voting systems and different vendors within the same state during the same election. With larger groups the variables of smaller sized counties simply add into the patterns of the whole.

I looked at the data a number of ways, from many different angles, and even compared county results to demographics. Overall, North Carolina has 2,582,462 (47%) registered Democrats to 1,903,119 (34%) registered Republicans. Add to that 1,021,648 unannounced/Independent voters (19%) and 12,754 Libertarians. They are interesting numbers because if we combined ALL of the Republicans and ALL of the unannounced/Independents we would end up with 53% in relation to 47% Democrats, exactly what the absentee/early vote showed as the relation of Bush to Kerry. That also illustrates how difficult it would be to move past that number. That isn't to say there wasn't some Democratic crossover, but the Independent voters were also not 100% for Bush and most likely voted in the same percentages in the early vote as they did on election day.

Question 1: Was the early vote - poll vote disparity the same across different voting methods? It was not. Here are the results from worst to best.

EARLY VOTE/POLL DAY DISPARITY

PUNCH CARDS
Differential -8 : Early Vote 45K/55B : Election Day 37K/63B

OPTICAL SCAN
Differential -6 : Early Vote 49K/51B : Election Day 43/57

DRE/ELECTRONIC
Differential -2 : Early Vote 44K/56B : Election Day 42/58

LEVERS
Differential -2 : Early Vote 53K/47B : Election Day 51/49

PAPER BALLOTS
Differential -0 : Early Vote 39K/61B : Election Day 39/61


PAPER BALLOTS AND LEVERS

It should be noted that the hand counted paper ballots comprise only three tiny, mostly Republican counties. That said, they were the only system that was dead on -- showing the same percentile of vote in the early/absentee and on poll day. That shouldn't come as a shock. Levers, the other manual method, totaled 5 times the number of votes of paper ballots, had a change of just 2%, and had Kerry in the lead in both the early and election day returns. It's a very hard fact to get around -- manual voting methods were by far the most accurate and reflective of any of the various methods used in North Carolina using the criteria outlined.

DRE'S (ELECTRONIC VOTING)

But electronic voting also showed a 2% differential, didn't it? The DRE's hold a huge voting block in North Carolina, serving 42% of the vote with no less than nine different vendors. One reason for the 2% differential is that the vendors are widely divergent in their returns and tend to average out, but there's more to the story, a lot more. For now take the -2% with a grain of salt, and I'll discuss that issue further down.

OPTICAL SCANS

The second worst system for reflecting parity in the first 1/3 and final 2/3 of the vote was Optical Scan. Optical scan ballots hold the biggest chunk of North Carolina's votes. At 47%, they are nearly HALF of all votes cast and swung wildly from a nearly EVEN race in the early vote 49/51 to a plummet of 43/57. They went from a 2% differential to a 14% differential over the weekend. Behind the large differential, Optical Scans counties, however, tend to break into three camps, those where Kerry maintained a lead or actually gained, those that slipped a little, those that slipped way out of proportion to the others. Of the 49 primarily Optical Scan counties, 16 had Kerry in the early lead. Twelve counties also had him ahead on November 2nd. He *gained* or maintained his percentage of votes in 17 counties. The great optical scan disparity, then, is a product of a handful of counties. Red flags that might warrant further study are the Republican strongholds of Ashe, Mitchell, Avery, Johnston, and Randolph who bottomed out even past normal Republican county standards. These counties are all fairly small, though, so there is probably more merit in examining the Democratic county drops particularly in the population center of Wake County which reversed a 53/47 Kerry early vote lead into a 47/53 win for Bush on election day. Chatham, Cleveland, and Orange Counties also need scrutiny. Orange County posted the state's biggest drop at 14%. It is unusual in that they may have possibly had the most successful GOTV early voter lead I have seen. 57% of all Democratic votes in Orange were cast by absentee/early vote. In comparison, 61% of the Republican votes were cast at the polls. It may have reasonable explanation, but it is bizarre.

Most of the 48 optical scan counties are controlled by either ES&S (24) or Diebold (19). ES&S showed the largest drop between early vote and election day at -6% after beginning with an even 50/50 split of the vote. Diebold on the other hand started low, a 44/56 split of the early vote, and dropped even lower to a 42/58 16% differential on election day. It's also of note that ES&S's Republican dominated counties all bunched toward the mean in absentee/poll differential because heavily Republican counties tend to be monolithic and turn out roughly the same numbers in early votes and on election day. Diebold's Republican counties, however, showed an unexplained further drop of the Democratic vote on election day as noted above. Airmac's two counties are so small, it is hard to make much of a statement on them. Fidlar Doubleday, however, fared well in its three counties. Despite the overall numbers being dominated by the much larger Surry County with a heavy Republican turnout, Democratic Hertford and Bertie posted sizable Kerry leads in both early voting and on election day . Of ES&S and Diebold's FORTY-THREE counties, ONLY THREE have comparable numbers to just two of Fidlar Doubleday's three counties.

PUNCH CARDS

Conclusively, Punch Cards showed the worst correlation of all systems. Their sample was much smaller than DRE's or Optical Scan Ballots, but they were also a sizable 9% of the vote. Not much can be said, except punch card counties careened into the abyss, going from a 45/55 10% early differential to a 37/63 26% election day differential! The six punch card counties are all controlled by ES&S with the biggest fallouts occurring in two counties, Watauga and Cabarrus.

Watauga County is dominated by ES&S punch cards but also includes a mix of systems including a few touchscreens. In the early vote Kerry took a comfortable 52/48 lead (5351 votes to Bush's 4865). But on Nov. 2nd the bottom dropped with Bush prevailing 43/57 -- that's a 19 point turnaround over the weekend. So bad was the turnabout, if one subtracted the early vote numbers from election day, we would be left with 2929 votes for Bush against 530 votes for Kerry. Maybe a lot of rural Dixiecrats wandered in, but that's a little ridiculous. A question remains over how much of the disparity can be attributed to an early GOTV for Democrats, fueled by ballot misreads or perhaps a toxic cocktail of mixed voting systems. Overall, Watauga's drop was one of the day's biggest discrepancies and yet was exceeded by another ES&S punch card county, Cabarrus. As the county had already established an excessive 32% lead for Bush in the early voting (34/66), it is absurd to see it plunge another 12% points to 22/78, within 2 points of the largest margin in the state. The county is tilting red, breaking 46/54 Republican but it also has 11,730 registered black voters. Kerry garnered just 19,803 total votes to Bush's 60,824. The 22% Kerry election day vote was the lowest percentage in the state, exceeding even the most staunchly Republican counties. They're not just tilting red in Cabbarus, something or someone went tilt. The question is how many "spoiled" ballots turned up in both counties? One very odd thing about the 6 ES&S punch card counties is that their early vote accounted for only 23% of their overall votes. The state early vote average is 30% while the other systems averaged 34% DRE's, 32% Levers, and 29% Optical Scans. It is an open question as to whether or not the Punch Cards unexplainable 7% differential from the mean (and the smaller differential of optical scan ballots) is a red flag for "spoiled ballots".

ONE MORE THING

Comparing the disparity of the different technologies, I decided to combine results from a mix of older voting systems (minus the hanging chad, machine counted punch cards). The systems included hand counted paper ballots, levers, and a very old type electronic voting system made by an extinct company called Shoup in 1987 that still uses the original software. The combined results of these 8 counties (3 Kerry, 5 Bush) were:

EARLY VOTE
Kerry: 10864 48%
Bush: 11718 52%

POLL VOTE
Kerry: 23398 48%
Bush: 25287 52%

Amazingly enough, they are both perfectly correlated between the early vote and election day returns and are also dead-on the final state polls and election day exit polls. Though the numbers are relatively small, it includes eight counties, five of whom went for Bush. You can mix and match results all day from any of the other voting methods and nowhere, no when, no how will you come close to the parity of these results and their agreement with poll data.



Question 2: Was the early vote - poll vote disparity the same across different vendors? It was not.

VENDORS (State Avg: Early 46/54 Polls 42/58)

SHOUP
(+4) : Early 44K/56B : Polls 48K/52B : Pct. early 34% : 1/30,470 (Counties:Votes)

AVM
(+2) : Early 54K/46B : Polls 56K/44B : Pct. early 37% : 1/11,527 (Counties:Votes)

DANAHER CTRLS
(+0) : Early 38K/62B : Polls 38K/62B : Pct. early 30% : 7/152,724 (Counties:Votes)

MICROVOTE GNL
(-2) : Early 48K/52B : Polls 46K/54B : Pct. early 32% : 8/485,543 (Counties:Votes)

DIEBOLD
(-2) : Early 43K/57B : Polls 41K/59B : Pct. early 23% : 21/469,727 (Counties:Votes)

AIRMAC
(-2) : Early 35K/65B : Polls 33K/67B : Pct. early 22% : 2/9,994 (Counties:Votes)

FIDLAR DBLDY
(-3) : Early 46K/54B : Polls 43K/57B : Pct. early 27% : 4/46,724 (Counties:Votes)

SEQUOIA
(-3) : Early 46K/54B : Polls 43K/57B : Pct. early 36% : 6/240,103 (Counties:Votes)

HART INTERCIVIC
(-3) : Early 34K/66B : Polls 31K/69B : Pct. early 38% : 1/58,460 (Counties:Votes)

ES&S
(-5) : Early 48K/52B : Polls 43K/57B : Pct. early 32% : 43/1,920,303 (Counties:Votes)

UNILECT
(-7) : Early 40K/60B : Polls 33K/67B : Pct. early 24% : 2/57,840 (Counties:Votes)


That's an incredibly diverse group, and it isn't really fair to judge a small sample size of a vendor with a single county against a vendor with half the state, so I will comment mostly on the larger vendors.

ES&S with the largest number of contracts in the state also has a broad spectrum of voting devices from the touchscreen DRE's to optical scanners to punch cards. They also had the second worst differential (-5%) of all vendors between the early vote (32% of their totals) and election day. Likewise, Unilect Corporation posted a series of severe numbers. They had the worst differential at -7%; they posted the greatest disparity between candidates on November 2nd (Bush by 34%); and they had an abysmal early vote percentage (just 24%). Unilect is the company that oversaw the fiasco in Carteret County where 4,500 votes totally vanished. With that record, one wonders how many other votes between their two counties disappeared.

Of the companies with substantial shares of North Carolina's election, MicroVote General posted the most consistent numbers. Among their counties is Mecklenburg, a large population center that compares favorably to Wake County run by ES&S, and it gives us a unique opportunity to compare results. Let's see what we find.

WAKE
201035 registered Democrats 54%
167995 registered Republicans 46%
106912 unannounced/Independents
366378 White 80%
89186 Black 20%

MECKLENBURG
219507 registered Democrats 56%
172022 registered Republicans 44%
108577 unannounced/Independents
342162 White 71%
137842 Black 29%

Here are the absentee/early vote statistics.

EARLY VOTE:
WAKE (ES&S Optical Scan)
Kerry 52411 53%
Bush 46311 47%

MECKLENBURG (MicroVote E-Voting)
Kerry 55405 54%
Bush 47209 46%

Extremely similar and right in line with their demographics. Here's what happened on election day.

ELECTION DAY VOTE:
WAKE (ES&S Optical Scan)
Kerry 117498 47%
Bush 131013 53%

MECKLENBURG (MicroVote E-Voting)
Kerry 111423 51%
Bush 107875 49%

FINAL RESULTS:
WAKE (ES&S Optical Scan)
Kerry 169909 49%
Bush 177324 51%

MECKLENBURG (MicroVote General E-Voting)
Kerry 166828 52%
Bush 155084 48%

Nearly identical counties with nearly identical early vote counts that reflected their demographics, but on election day they diverged. Although Kerry's vote shifted by 3% in Mecklenburg, he maintained a lead and won the county. In Wake, however, there was DOUBLE the shift to a 6% decline in Kerry support. Counties do not always follow the same pattern, but a doubling of differential in counties so similar makes one look twice at the two companies running the elections in question. Why were the ES&S results so counter to the MicroVote results?

Diebold is second on the county list, consisting of 20 optical scan counties and 1 e-voting county, and yet they had the second lowest percentage of early voting (23%, 7% under the state average) to Airmac (22%) another optical scan company. Was the early vote suppressed, was there a preponderance of "spoiled ballots," or were the half a million people in its counties just not interested in voting early?

Lastly, Danaher Controls provoked interest with their seven counties of e-voting. Their differential was spot on, but the disparity of the vote was enormous. Looking more closely at their counties, Lenoir is open to big questions. The county is 66% registered Democrats, just 24% Republican, and 40% Black. Bush won 56/44! Likewise their results for Bladen county were...uh...interesting. Bladen is 73% Democratic, 14% Republican, and 36% Black. Bush won with a shave over 50% of the vote by winning the early vote 56/44. It's dumbfounding to see. And Danaher's bottom feeding 24% cumulative spread for Bush in both the early vote and at the polls led me to an interesting realization about e-voting which I think is the real story here...

Question 3: What role did electronic voting play?

While there is a huge correlation discrepancy that must be reconciled with the other major voting systems, Optical Scanners and Punch Cards, the paperless computerized voting systems had only a 2% deviation as a group. It ran counter my assumptions. But when I looked more closely, I saw why. Even though, the DRE's had the largest voting block and highest percentage of early voting (which has been established as a mark of the Democrats' extensive GOTV campaign), the DRE's had the WORST margin of disparity between the candidates, a 10% LEAD for Bush at 56% to 44%. It suddenly dawned on me that the DRE's were suppressing BOTH sides of the election, and it was the crazy mix of voting systems in this election that was revealing it. The starkest comparison is that DRE's and Optical Scanners each hold huge chunks of North Carolina's electorate. They are nearly mirrored in percentage and size for North Carolina's votes -- 47% and 1,655,779 total votes for Optical Scanners and 42% and 1,480,040 total votes for DRE's, enough to marginalize any discrepancies, and yet there that one glaring discrepancy -- the big difference in the early vote percentages each system posted for Bush and Kerry.

Absentee/Early Vote

Optical Scan:
Bush: 249,385 51%
Kerry: 235,175 49% -2%

Electronic Voting:
Bush: 280,354 56%
Kerry: 221,235 44% -10%

Let me put that in further perspective. The Optical Scan counties had already posted a weak early turnout at 29% of the total vote, mainly propelled by Diebold's ridiculously low 23% early turnout over their twenty counties. The DRE's had the HIGHEST percentage of early votes at 34%. A high early turnout is supposed to favor the Democrat's huge GOTV campaign and yet in the 36 DRE counties with the highest early vote totals, Kerry drops from 2% behind in Optical Scan territory to 10% in touchscreen Mecca. Something is seriously messed up.

Here are the numbers:

Optical Scanners: 48 Counties 1,655,779 total votes (47% of all votes)
DRE's (E-Voting): 36 Counties 1,480,040 total votes (42% of all votes)

EARLY VOTE/ABSENTEES
Optical Scan: 484,560 votes 29%
DRE's: 501,589 votes 34%

But there's more. There are major differences in the e-voting vendors. Let's take a look at how each recorded the election. (Note: these are only a company's e-voting totals.)

DANAHER CONTROLS (7 counties: 3DEM, 2REP, 2 SPLIT)
Early/Absentee
Kerry (pct) Bush (pct)
17592 (38%) 28566 (62%)
Election Day Polls
Kerry (pct) Bush (pct)
40640 (38%) 65926 (62%)

ES&S (13 counties: 8DEM, 1REP, 4SPLIT)
Early/Absentee
Kerry (pct) Bush (pct)
69239 (44%) 89479 (56%)
Election Day Polls
Kerry (pct) Bush (pct)
110368 (44%) 140734 (56%)

MICROVOTE GENERAL (7 counties: 3DEM, 1REP, 3 SPLIT)
Early/Absentee
Kerry (pct) Bush (pct)
74655 (48%) 80093 (52%)
Election Day Polls
Kerry (pct) Bush (pct)
152089 (46%) 178703 (54%)

SEQUOIA (3 counties: 2DEM, 1REP)
Early/Absentee
Kerry (pct) Bush (pct)
36193 (45%) 43420 (55%)
Election Day Polls
Kerry (pct) Bush (pct)
58573 (42%) 80481 (58%)

UNILECT (2 Counties: 2SPLIT)
Early/Absentee
Kerry (pct) Bush (pct)
5447 (40%) 8209 (60%)
Election Day Polls
Kerry (pct) Bush (pct)
14013 (33%) 28429 (67%)

DIEBOLD (1 county: SPLIT)
Early/Absentee
Kerry (pct) Bush (pct)
5172 (36%) 9141 (64%)
Election Day Polls
Kerry (pct) Bush (pct)
15082 (31%) 34111 (69%)

FIDLAR DOUBLEDAY (1 county: DEM)
Early/Absentee
Kerry (pct) Bush (pct)
898 (45%) 1076 (55%)
Election Day Polls
Kerry (pct) Bush (pct)
1024 (37%) 1757 (63%)

HART INTERCIVIC (1 county: REP)
Early/Absentee
Kerry (pct) Bush (pct)
7421 (34%) 14604 (66%)
Election Day Polls
Kerry (pct) Bush (pct)
11437 (31%) 24998 (69%)

As a key to looking into those numbers, let's look at one more study. This is a list comparing how many times Kerry posted a lead within a given county in different voting systems.

KERRY LEADS STUDY
E-voting: 36 Counties; 5 early leads 13.8%; 5 Nov.2 leads 13.8%
Optical Scans: 49 Counties; 16 early leads 32.7%; 12 Nov.2 leads 24.5%
All other: 14 Counties; 5 early leads 35.7%; 3 Nov.2 leads 21.4%
(note: Lee County did not post early vote counts.)

This shows us rather explicitly that although e-voting dominated the early voting and captured 42% of the statewide votes, electronic results stifled, snuffed, and skewered the results. As bad as we believe the optical scans were, Kerry posted early vote leads in 16 of the 49 optical scan counties (32.7%) and had an even better showing in the full combination of punch card, lever, and paper ballot counties boasting a 35.7% showing by taking 5 of 14 counties. But in the 36 counties with e-voting, he barely made scratch with just 13.8% and 5 of 36 counties. If one looks inside the numbers, though, it gets worse. Three of those leads were all posted for a single vendor, Microvote General, in just seven counties. For the rest of the e-voting vendors, the results were 2 out of 29, 6.9%. If all the other voting systems are posting Kerry leads at 33% and it's running in parallel numbers of e-votes at only 7%, it should shock the hell out of you.

Lest you think this is simply an aberration of demographics, let's look at the election results of the biggest e-voting vendor ES&S in comparison to the state's demographics and another e-voting competitor with a similar share, Microvote General.

DEMOGRAPHIC COMPARISON

GROUP A: Statewide
DEM 2582462 47%
REP 1903119 34%
LIB 12754 0.20%
UNA 1021648 19%
Total 5519983
White 4224098 79%
Black 1112959 21%

Group B: ES&S E-Voting
DEM 295384 47%
REP 219759 34.50%
LIB 1239 0.20%
UNA 117374 18.50%
Total 633756
White 487416 78%
Black 134289 22%

Group C: Microvote General E-Voting
DEM 335529 44%
REP 269203 36%
LIB 1786 0.20%
UNA 149551 20%
Total 756069
White 550657 75%
Black 179733 25%

The figures are nearly identical. ES&S has the same Democratic ratio as the rest of the state. They have a hair-split more Republican to Unannounced ratio. They even have 1% higher Black population. ES&S has 3% more Democrats than Microvote but also 3% less in Black population. In all cases, it is essentially a toss-up. The ES&S e-voting counties are dominated by Democratic stronghold counties (8 of their 13 counties) and with but one small exception not once did Kerry have a lead in the early/absentee voting. In the one county that Kerry tabulated a 234 vote lead in the early vote, it swung wildly from a 52/48 Kerry lead in the early vote to a 45/55 Bush success on election day. In comparison, Microvote General has seven counties (not including Lee) of which only three are Democratic strongholds and yet Kerry posted early leads in three counties by 6, 8, and 40 percent margins.

Now in deference, Pasquotank County was under ES&S e-voting, and they had one of the most dramatic shifts toward Kerry from the early voting. What are we to make of that? A whopping 48% of the votes were cast in the early election, a factor supposedly favoring Kerry. This is a HEAVILY Democratic county with 13904 registered Democrats to 5221 registered Republicans. It also has a large number of Black voters, 38%. But Bush won the early vote 53% to 47%. It raises the question of how Kerry came back with a 56% to 44% win on election day. It's possible that out of the 5221 registered Republicans in the county and the total 6609 votes for Bush, they had simply run out of enough Republican voters to feed the illusion.

We see the same thing with Perquimans County. Kerry appears to gain under ES&S e-voting from 38% in the early/absentee voting to 41% on election day. Perquimans has a Democrat: Republican ratio of 7:3 and a 24% Black population. But here Bush won handily on both sides of the election, garnering 2956 votes to 1952 Republican registrations.

Here is a table of the ES&S E-Voting Counties, the percentage of the vote that was early/absentee, the demographic ratio of Democrats to Republicans, and the early vote percentages for Kerry/Bush. Right down the line, we see the same results, from Guilford a population center with a heavy early turnout and a large black population (30%) giving Bush the early lead to Greene where an 81:19 ratio of Democrat to Republican turned into and early 59/41 lead for Bush to Stanly where a split electorate (47:53) voted at the same gargantuan early percentage for Bush (71%) as heavily Republican Davie County with a 30:70 Democrat:Republican ratio.

It's all very neat but in a messy state like North Carolina, the neatness is its undoing. We can compare methods, vendors, and the different halves of the election and use that messiness to peer into the black box. I hope this is a model for other messy states before we arrive at an imposed uniformity without recourse. The test of a democracy is in its openness and accountability. For it is in secrecy and impunity that power may perceive freedom as a threat and seek to control not only its expression but also its exercise. No matter one how views these findings, it is time for an open and accountable election.