On "spoiled ballots" with Range and Approval (and old-style plurality) voting

(Executive Summary)

The issue of "spoiled ballots" – ballots that do not obey the rules of the election system and which hence cannot be used – depends on a lot of factors – not just the election system's mathematical definition, but also how the election officials behave, what kind of voting machines they employ, and what settings they set their machines to.

For example in the critical Bush-versus-Gore presidential contest in Florida 2000, it was known that Blacks were much more pro-Gore than Whites. The Florida state governor was Bush's brother Jeb. The Florida state elections commissioner Katherine Harris was simultaneously Bush's Florida campaign co-chair. Sample voting machines of the types used in each Florida county were on display outside governor Bush's office before the election so the higher-ups could be sure they had been adjusted right. In heavily black counties and districts throughout Florida the machines were intentionally adjusted to silently reject invalid ballots. In contrast, in White counties, and districts expected to be pro-Bush, the machines were adjusted to complain to the voter about invalid ballots, and offer the voter the chance to try again.

Result: Sure enough, the rates of non-counting were considerably higher in the 4 blackest Florida counties than in the 4 whitest (and the electronic voting machines in Gadsden were known to have been adjusted to silently reject invalid ballots without any indication to the voter that it was happening):

Comparison of uncounted "spoiled" ballot rates in the 4 blackest and 4 whitest Florida counties [from Greg Palast's book]:
County Black pop. Uncounted County Black pop. Uncounted
Gadsden 52% 12% Citrus 2% 1/2 %
Madison 42% 7% Pasco 2% 3%
Hamilton 39% 9% Santa Rasa 4% 1%
Jackson 26% 7% Sarasota 4% 2%

Although only 11% of Floridians are black, it was claimed in a speech in Congress on 21 Sept. 2004 that over half the state's invalid ballots appear to have been cast by blacks, and indeed a report by the US Civil Rights commission estimated that over half the invalid ballots USA-wide in 2000 appear to have been cast by blacks, even though blacks are 12% of the USA population. In 2004 in New Mexico, non-Anglo (Black, Mexican-American, and Native American) voters cast at least 89% of the no-choice-counted-for-President ballots despite being only about 55% of the population, i.e. an over 4× higher rate. (Supposedly 21084 NM ballots "did not indicate any choice" for US president; Bush won the state by 5988 over Kerry.) And nationwide data gathered by Harvard Law School Civil Rights Project again indicate that, of the 2 million ballots spoiled in a typical presidential election, about half are cast by minority voters. Cleveland State University Professor Mark Salling analyzed ballots thrown into Ohio's electoral garbage can in 2004. Salling found that, "overwhelmingly," the voided votes come from African-American precincts. Check these maps (pdf) and with caption. Cornell University professor Walter Mebane Jr. analysed ballot-level data from the NORC Florida ballots project and ballot-image files, and concluded that "overvoted ballots in the 2000 presidential election in Florida included more than 50,000 votes that were intended to go to either Bush or Gore but instead were discarded. The primary reason for this was defective election administration in the state, especially the failure to use systems that warn the voter when there are too many marks on the ballot and allow the voter to make corrections. If the best type of vote tabulation system used in the state in 2000 – precinct-tabulated optical scan ballots – had been used statewide then [due to inequities in the distribution of voting machines & settings of those machines] Gore would have won by more than 30,000 votes." [ W.Mebane Jr.: The Wrong Man is President! Overvotes in the 2000 Presidential Election in Florida, Perspectives on Politics 2,3 (September 2004) 525-535]

This also jibes with the Washington Post's analysis of the 2.7 million votes cast in Florida's eight largest counties. It found that Gore's name was punched on 46,000 of their over-vote ballots, while Bush's was marked on only 17,000. (All over-voted ballots were discarded as illegal, so this discrepancy hurt Gore far more than it hurt Bush.) In the eight counties examined, people who overvoted for president but cast a valid vote in the U.S. Senate race favored the Democrat 70 to 24 percent.

To see that this kind of thing was happening not only at the county level, but also at the level of individual districts, check this histogram [from the Netrinsics.com study] of the Bush-Gore 2000 results in all the 268 precincts in Duval County Florida and notice the amazing anti-correlation between the percentage of "missing" (i.e. spoiled, hence left uncounted) ballots and the pro-Bush fraction in those districts, with the most pro-Gore districts having an incredible spoilage rate of over 20% of the ballots, while the most pro-Bush districts had 3%. A fluke result due to random fluctuations? I don't think so. The chance the two orderings – by missing-fraction and Bush-fraction – came out the same by luck, was 1/10!=0.000028%. This would be expected to happen just once every 4618 years among the 3143 counties in the USA, assuming one presidential election every 4 years. With a more sophisticated statistical analysis the amazingness would grow by at least a further factor of 10. Was this Duval-County-only effect large enough to swing the statewide Bush-Gore election result in Florida (which was decided by an official margin of only 537 votes)? Easily.

Duval spoiled ballots histogram

Heck, even the votes for Buchanan accidentally cast by more than 2000 Democratic voters in Palm Beach County (due to the famous butterfly ballot foul-up) would, if corrected, have been far more than sufficient to swing the election from Bush to Gore. [Jonathan N. A. Wand, Kenneth W. Shotts, Jasjeet S. Sekhon, Walter R. Mebane, Jr, Michael C. Herron, Henry E. Brady: The Butterfly Did It: The Aberrant Vote for Buchanan in Palm Beach County, Florida, American Political Science Review, 95,4 (2001) 793-810.]

Although generally speaking overvotes (e.g, votes for both Bush and Gore) are useless due to ambiguity, that is not true of so-called "write-in overvotes." One fact which has received near-zero media attention is that many of the "overvotes" in Florida 2000 were of this kind, where e.g. voters both voted Gore and wrote-in "Gore." Professor Lance deHaven-Smith writes (in Florida State University's Research in Review Magazine, concerning his book)

If you look at where those votes occurred, they were in predominantly black precincts. And (when you look at) the history of black voting in Florida, these are people that have been disenfranchised, intimidated. In the history of the early 20th century, black votes would be thrown out on technicalities, like they would use an X instead of a check mark. So you can understand why African Americans would be so careful, checking off Gore's name on the list of candidates and also writing Gore's name in the space for write-in votes. But because of the way the vote-counting machines work, this had the opposite effect: the machines threw out their ballots.

In an email to daily kos, deHaven-Smith summarizes:

A number of studies have been conducted of the "spoiled ballots" in Florida. The most thorough was by the National Opinion Research Center (NORC) at the University of Chicago. It examined all of the 175,010 spoiled ballots that were excluded from the vote tabulations. Ballots were examined by multiple coders to assure reliability, and a dataset (available on NORC's website) was constructed that included each coder's judgment on each ballot. NORC also provided a program for applying different criteria and calculating the results. If the most restrictive criteria are used, Gore's margin of victory is 158 votes. By "most restrictive," I mean ballots are counted only if the voter's intention is totally unambiguous (no partially punched chads, no unclear write-ins, etc.). [With less restrictive criteria the margin would be many thousands.]
Today, this would be widely recognized if NORC had published a report on its findings, but instead it chose simply to release the data and let everyone make sense of it on their own. My assumption is that NORC and the research sponsors did not issue a report because they did not want to embarrass President Bush. The study was concluded just after 9-11-01.

And we haven't even mentioned the famous "Felons" list, and see this about the Colorado felon list.

The same kind of amazing correlations also happened throughout the critical state of Ohio in the 2004 Bush-Kerry election, where again, amazingly, the Ohio Commissioner of Elections J.Kenneth Blackwell just "coincidentally" happened simultaneously to be Bush's campaign chair. [Many of these figures are extracted from the immense data collections documenting Ohio election manipulation collected by Ph.D. statistician Richard Hayes Phillips, as well as from Ohio newspapers.] E.g.:

Summit County:
There were exactly 71 precincts with more than 3.0% of the ballots spoiled hence uncounted. Kerry won 70 of these 71, by an aggregate ratio of about 4 to 1. Meanwhile the countywide totals were Kerry 156587, Bush 118558.
Hamilton county:
(Which includes Cincinnati) Bush won with 222616 votes to Kerry's 199679. There were exactly 26 precincts with more than 8% of the votes left uncounted. Kerry overwhelmingly won all 26 of them, by an aggregate margin of 10 to 1.
Montgomery County:
Had exactly 588 precincts, and exactly 47 of them had more than 4.0% of the regular ballots uncounted. Kerry won every one of those 47 precincts, by an aggregate margin of 7 to 1. The countywide official vote totals, meanwhile, were 142997 for Kerry and 138371 for Bush. In these 47 precincts the rate of ballot "spoilage" was 5.16%, compared to 1.31% for the rest of the county.
Allegheny County:
More voter complaints than any other county in the entire USA. (In second place, not to be daunted: Broward County FL.) Each polling place furnished with 12 (yes, twelve) provisional ballots.
Stark County:
There were exactly 28 precincts with more than 3.33% of the regular ballots uncounted. Kerry won all 28 precincts, by an aggregate margin of 2.7 to 1. Meanwhile the countywide totals were Kerry 95337, Bush 92215.
Lucas County:
The 88 precincts with the lowest turnout all were won by Kerry by an aggregate ratio of over 4.5 to 1. All 88 were in Toledo (which contains over 60% of the county population). Meanwhile 21 of the 25 highest-turnout precincts in Toledo were won by Bush.
Chairing the Lucas County election Board was Bernadette Noe, who was simultaneously Bush's campaign chair in that county. She was married to Tom Noe, who later pled guilty to money laundering charges and continues to face massive fraud charges in "coingate" (in which Noe invested $50 million of Illinois public worker's comp money – he was the sole fund manager – in rare coins sold by Noe's personal numismatics business, then some coins mysteriously vanished... resulting in $13 million in losses for the state fund. Noe did this by paying multi-thousand-$ kickbacks to others who also pled guilty, such as Terrence Gasper...). In October 2004 the window in the Democratic Party Lucas office was smashed and computers with sensitive election information were burglarized; other offices in the same building were not entered, nor were less-sensitive Dem-party computers or any of their petty cash taken, and everything was left neat and tidy, suggesting run of the mill criminals were not behind this. ("They knew what they wanted" according to one witness quoted in the Toledo Blade.) The thieves would then (especially if they were in communication with Bernadette Noe) have been in a perfect position to try to suppress votes in the areas the Dems liked the most, for example by moving polling places to unexpected new high-crime areas, "forgetting" to designate voting machines in order to lengthen lines, using defective machines, "accidentally" sending voters to wrong locations, etc (all of these are complaints that were filed by voters in these 88 precincts).

Obviously, most or all of these were not "flukes." They were "pre-arranged" just like in Florida.

And on and on. I have plenty more Ohio 04 data of this kind. More Ohio stories. Think those Ohio biases were just a sad, unintentional side effect of the fact that poorer, blacker, too-stupid-to-vote-validly, and more urban voters tend to vote Democratic? Cornell Professor Walter Mebane Jr. doesn't agree. He compared the distribution of voting machines in the Ohio 04 presidential election and the primary election held the previous spring. They differed. Amazingly enough, in the primary "there was no sign of racial bias in the distribution of the machines" but when it came to the general election in November, lo and behold, "there was substantial bias with fewer voting machines per voter in areas that were heavily African-American." Mebane concluded his study of Franklin County with "The allocation of voting machines in Franklin County was clearly biased against voters in precincts with high proportions of African Americans when measured using the standard of the November, 2004, electorate. The most conservative estimate... [is that this] reduced voter turnout by slightly more than 4% in precincts in which high proportions of the voters were African American, while shortages in precincts where very few voters were African American reduced voter turnout by slightly less than 1.5%." Essentially the same conclusions were reached in an independent study using different methodology by Benjamin Highton at UC-Davis. Ohio statewide, African-American voters reported waiting an average of 52 minutes in line to vote, compared to white voters' 18 minutes; but in Franklin County's Columbus 55th ward, on the ground election protection volunteers clocked an average wait of 3 hours 15 minutes. The Franklin County Board of Elections, under the control of former Franklin County Republican Party Chair Matt Damschroder, failed to employ 76 voting machines (by his own admittance; the Democrats claimed 122) – all 76 from the Democrat-rich city of Columbus. Damschroder and Blackwell met with Bush and Karl Rove on election day in Columbus Ohio [reported sources to the Free Press]. The official story, e.g, reported in the Chicago Tribune, was that Bush had flown to Columbus that morning from his home in Crawford Texas, in order that he could join phone bank members by making a single phone call with them to a voter; then after that inspiring call, Bush went back to the airport and flew on to Washington DC.

I am not confident that the above distortions were large enough to swing Ohio (and therefore the presidency) from Kerry to Bush. I am confident that they were in the same ballpark as the amount required to do that. This conclusion is also supported by the fact that CNN (based on exit polls of 1963 Ohio voters) predicted slightly after midnight EST that Kerry would defeat Bush in Ohio by a margin of 4.2 percentage points, whereas the official election results were that Bush won Ohio by 2.5 percent. (More seriously, there was a countrywide discrepancy of the same sign as this, which was biggest in the critical states such as Ohio, Florida, and Pennsylvania. CNN's exit polls after midnight predicted a countrywide popular-vote victory for Kerry 51-to-48% based on 11027 respondents, but the official totals gave it to Bush by 50.7-to-49.1%.) In November 2004, Senator Richard Lugar(R), in Kiev, cited the gap between exit polls and official results as solid evidence of "blatant fraud" in the Ukraine election. Hence the Bush administration refused to accept Ukraine's official vote total.

And in the 1960 Kennedy-Nixon presidential contest, most historians agree that Kennedy won the state of Illinois because of electoral fraud in Daley-controlled Chicago, and future US president Lyndon Johnson got his start by winning a massively blatantly fraudulent Senatorial primary election in Texas [e.g. see Caro and Feinberg's books].

For an eye-opening history of all kinds of USA election fraud and manipulation, check out Gumbel's book, and for a powerfully comprehensive and documented report on the Ohio 2004 shenanigans, see Kennedy's Rolling Stone article. (A less-good report about Ohio 2004 is Miller's Harpers Magazine article.)

Reduced spoilage under Range & Approval voting? Our recommendations?

Theoretically, by using computerized machines that simply refuse to accept invalid votes and demand the voter keep trying until his vote obeys the rules, any voting system could be made spoilage-free. Sounds simple to eliminate the problem, right?

But sadly, in reality, evil manipulators could always reprogram such machines to avoid complaining to voters about invalid ballots, or just to commit blatant fraud by discarding or faking ballots.

It has been argued that with approval voting, (especially the version of AV where you are allowed to leave votes blank) ballot spoilage is impossible because every possible combination of approvals and disapprovals is a legitimate vote. (Unlike in the plurality system where, e.g., some combinations – such as voting for more than one candidate – are illegal.)

That argument is however not really true because voters can always find some way to mangle their ballot to render it unreadable by the machine, etc., so there will always be some spoilage rate.

With range voting, every possible combination of 2-digit range votes (including blanks) is legitimate, so again "ballot spoilage is impossible." But again this is not really true because of e.g, voters who wrote an unreadable digit. In range voting with "blanks," which CRV recommends, we could agree to treat illegible digits as blanks in which case the ballot would not be spoiled.

Range and approval voting definitely do not cure the problem of election fraud and manipulation. But they do reduce spoilage.

And the problem of ballots "spoiled" by "accident" is really hardly a problem at all (true accidents would tend to cancel out and lead to only a very tiny noise level) – the real problem is intentional manipulation as in the examples above intended to create biases that do not cancel out.

And a cure (or at least a partial cure) for that is to have uniform national standards for ballot "spoilage" and not to allow individual districts and counties and states to make up their own each election. For example, immediate voter-notification of ballot non-validity, should be required everywhere. It is an outrage that this still has not happened after over 200 years of US history. We strongly recommend that this be done as part of whatever law enacts range voting.

But we think it is fair to say that ballot spoilage would be a smaller problem with range and approval voting because if your vote about candidate A is spoiled, then your votes on all the other candidates still can be used, so only part of your vote gets lost. This is exactly the right thing to do – unreadable scores for a candidate are regarded as blanks and not incorporated into his average. (And of course, this policy should be demanded by the uniform national standard.)

Return to main page