It’s the self-projection, stupid.

What I mean by that is everything Trump accuses someone else of doing, he’s doing himself. Sexual assault by Bill Clinton? Turns out Trump is a serial groper–and boasts about it. The Clinton Foundation is thoroughly corrupt? The Trump Foundation is nothing more than a tax dodge. It goes on and on.

So when Trump is claiming HITLERY KLINTOON is going to steal the election, that it will be “rigged”, it would absolutely be par for the course to assume Trump (and his campaign) are actively working on rigging and stealing the election. It would be out of character if they weren’t.

Some of what Republicans are doing involves their ongoing effort to disenfranchise likely Democratic voters. While that’s not getting the media coverage it deserves, that is out in the open. But consider this: imagine if the polls for Pennsylvania hold up–Clinton has a large lead–but on election day, the results are a lot closer. Any Democrat who claims she wouldn’t think something odd has happened is lying. While I think Pennsylvania’s margin is tamper-proof, what would that mean for down ballot races?

I use Pennsylvania as an example for this reason:

Because this should scare the shit out of every single Democrat (boldface mine):

But most identified Pennsylvania as the greatest concern. There, according to Verified Voting 47 counties of 67 vote on digital voting machines without a written backup record if something were to go awry—a reality that is very much on the minds of state officials (legislation is working its way through the House to examine the issue of voting modernization.) In Pittsburgh and Philadelphia—two Democratic strongholds whose turnout typically decide the fate of the state’s outcome—around 900,000 voters will cast ballots entirely on paperless touchscreens DREs, if previous elections are any guide. Then, at least from the voters’ perspective, they will disappear into a sea of ones and zeroes.

Montgomery County, a crucial Democratic redoubt in the suburbs of Philadelphia—an area sometimes seen as having the potential to swing the entire state—is one such locality that uses a paperless electronic machine, and only one machine, for all 425 precincts: Appel’s Sequoia AVC Advantage.

With all the talk about how Trump is wrong about the elections being rigged–against him–I worry that we’re setting ourselves for a real problem in the other direction.

Just use optical scans [scanning machines] and publicly announce precinct tallies at the precincts. This is too important not to get right.

Online, I’ve seen several people arguing that it would be impossible to ‘steal’ an election. But this isn’t a legal issue, this is an information transfer issue: can people’s votes be accurately turned into reliable and representative vote totals? One of my day jobs is to try to break a complex computational system, to find all of the ways things can go wrong, and to make the system generate gibberish (it’s kinda fun actually). So when people make extravagant claims about the integrity of our voting system, that just tells me they haven’t been paying attention to the actual system.

Moreover, our voting system has already been hacked:

In Fairfax County, Virginia, an investigation into a 2003 school board race found that a vote was subtracted for every 100 votes cast for one of the candidates on 10 machines. With margin sizes small enough to be noticed, local elections were vaulted into the forefront of these debates…

But the tipping point came in 2006, when a major congressional race between Vern Buchanan and Christine Jennings in Florida’s 13th District imploded over the vote counts in Sarasota County—where 18,000 votes from paperless machines essentially went missing (technically deemed an “undervote”) in a race decided by less than 400 votes….

The auditing techniques of most states aren’t sophisticated enough to detect a subtle attack—every 100th vote switched from Trump to Hillary Clinton, for instance. “The whole point of a Risk Limiting Audit is not to find the tally down to the last digit,” explains Wallach. “The problem you’re trying to figure out is if the error rate is big enough that I could change who won.”

(and that’s just a taste. The article has much, much more). And I’m not the only one who thinks this could happen, security expert Bruce Scheier does too.

We haven’t even broached the subject of altering vote totals at county or state registrars (well, I guess I just did). Many of these ‘systems’, according to Politico, are nothing more than Excel on an unsecured computer. Not exactly impregnable.

At this point, someone will argue that the system is so decentralized, it would be impossible to hack national or state elections. Well, the 2008 and 2012 elections might suggest otherwise:

Here’s what they found in the GOP Iowa caucus in 2012:

What the authors did is look at the percentage of votes cast (the vertical axis on the left) according to the number of votes cast, where voting precincts where sequentially added from largest to smallest (the horizontal axis on the bottom). In other words, if the smallest precinct had 100 votes, with the next largest having 110, then the 210 mark on the graph would give the results of those 210 votes. Admittedly, one graph is tinfoil helmet territory, but look at New Hampshire in 2012:

[more states at the original post]

…Oddly enough, this pattern occurs in seven other states [eleven total]. And it always favors Mitt Romney. This doesn’t seem to be an urban/rural divide either: the authors took a geographically randomized subset of the data and found the same patterns. In addition, these patterns also hold within counties. What it looks like is that someone is electronically putting his or her thumb on the scale in large precincts to tilt things towards Romney (the one state that doesn’t appear to have this problem is Utah where Romney won in a blowout). This also isn’t opponent specific: in some cases Paul took the hit, while, in others, Santorum did.

Similar patterns are found in some of the 2008 primaries and in the 2008 general election. In 2008 in Cuyahoga County, OH, the same type of pattern favored McCain over Obama, even though there is no relationship whatsoever between the ratio of Democrats to Republicans and precinct size (R-squared is 0.01). Interestingly, these patterns also weren’t seen in the 2008 Democratic or Libertarian primaries.

In 2015, a mathematician stumbled across this paper and set out to disprove it, only to end up agreeing with its findings:

A Wichita State University mathematician sued the top Kansas election official Wednesday, seeking paper tapes from electronic voting machines in an effort to explain statistical anomalies favoring Republicans in counts coming from large precincts across the country….

Clarkson, a certified quality engineer with a Ph.D. in statistics, said she has analyzed election returns in Kansas and elsewhere over several elections that indicate “a statistically significant” pattern where the percentage of Republican votes increase the larger the size of the precinct.

While it is well-recognized that smaller, rural precincts tend to lean Republican, statisticians have been unable to explain the consistent pattern favoring Republicans that trends upward as the number of votes cast in a precinct or other voting unit goes up. In primaries, the favored candidate appears to always be the Republican establishment candidate, above a tea party challenger. And the upward trend for Republicans occurs once a voting unit reaches roughly 500 votes.

“This is not just an anomaly that occurred in one place,” Clarkson said. “It is a pattern that has occurred repeatedly in elections across the United States.”

The pattern could be voter fraud or a demographic trend that has not been picked up by extensive polling, she said.

“I do not know why this trend is there, but I know that the pattern is there and one way to establish that it is or is not election fraud is to go and do a physical audit of paper records of voting machines,” she said.

Clarkson wants the hard copies to check the error rate on electronic voting machines that were used in a voting station in Sedgwick County to establish a statistical model…

Clarkson became more interested in the issue after reading a paper written by statisticians Francois Choquette and James Johnson in 2012 of the Republican primary results showing strong statistical evidence of election manipulation in Iowa, New Hampshire, Arizona, Ohio, Oklahoma, Alabama, Louisiana, Wisconsin, West Virginia and Kentucky.

Clarkson said she couldn’t believe their findings, so she checked their math and found it was correct and checked their model selection and found it appropriate. Then she pulled additional data from other elections they hadn’t analyzed and found the same pattern.

…there is something really weird going on here, and it is consistent with–which does not mean it is due to–someone trying to subtly weight the favorites.

Unfortunately, Clarkson lost in court (always get a lawyer). But eleven states across multiple elections with the same trend is not a ‘local’ problem.

Keep in mind, in 2008 and 2012, McCain and Romney very well might not have known about this (personally, I don’t think they did; I do think someone did this intentionally, however). In fact, anyone familiar with complex software with multiple dependencies could very well conclude no one is at fault–that this is a glitch in the code (like I mentioned, when you spend your time trying to break things, you stumble across some crazy shit), though I think that’s unlikely. But something very weird happened.

To be blunt, it’s a lot harder to hack your credit card information than it is to hack your vote. Which inspires confidence, doesn’t it?

That’s why I’m worried about Democrats arguing that the elections can’t be ‘rigged.’ What if they are–against Democrats?

So what should Democrats–and anyone else who cares about accurate voting results–do? First, electronic-only machines need to be eliminated. Second, for this election cycle, those machines that don’t provide a paper record as well should be randomly inspected (i.e., code and hardware review). Third, even if Democrats win the presidency (and I think we will), we need to analyze the precinct-level results and be ready to go to court to impound voting machines and paper trials, as well as gain access to the code and hardware. Fourth, precinct tallies need to be publicly announced at the precinct to identify any downstream tampering. Fifth, random precincts need to be spot-checked to see if they match the paper tally.

While this could be a problem for the presidency (again, I doubt it), we really don’t know if there is anything going on with state and local races–which, again, has happened before. It’s worth investigating.

Note: A pdf of the paper can be found here.