There’s a lot of chatter this morning about a group of election lawyers and cybersecurity experts who believe they’ve found evidence of or at least statistical anomalies suggesting that the voting results may have been hacked in the three critical states of Wisconsin, Michigan and Pennsylvania. Note that these are the three states where Clinton lost the election. Ohio, North Carolina and Florida were huge too. But Trump was always likely to win Ohio and North Carolina and Florida were always toss-ups, although Florida seemed pretty promising.
So here’s my take.
1. Always be extremely skeptical of your theories which purport to explain why something you really did not want to happen in fact did not happen.
2. Something being possible in theory does not constitute any evidence that it happened. This is the downfall of so many ‘vote fraud’ conspiracy theories.
3. The ground zero of election tampering stories goes back to the 2004 presidential election where there were various arguments about election anomalies, final results which substantially diverged from election night exit polls, etc. The main questions centered on Ohio where the race was won and lost. None of those theories ever panned out as far as I could tell. Elections are messy. Things that you think are anomalies usually aren’t. Perhaps most critically, exit polls are not remotely reliable enough to stand up as some standard against which you judge actual results.
4. The one thing that makes things somewhat different in 2016 is that we know to a pretty high degree of certainty that a hostile foreign power was actively tampering with the electoral system. This is needless to say a pretty big deal in terms of evaluating these claims. The biggest doubt I have about these claims in general (i.e., not just the 2016 cycle) is thinking through the logistics. You need a lot of people to be involved. The criminal penalties are really high. When lots of people are involved things tend to leak out and people get caught. All of these conditions make widespread tampering not only difficult but highly risky. But of course none of those things apply to manipulation by a foreign state. They have high levels of technical knowledge and they don’t need to worry about getting prosecuted. That doesn’t mean this happened. I just put this out there because it’s the one thing that makes 2016 different – and different in a significant way – from other cycles. We don’t need to speculate that someone was trying to influence/tamper with the electoral process. They were.
5. But even for a foreign power, this would be very difficult to pull off. Here’s one reason. The big story from this election is that Donald Trump over-performed Mitt Romney in a series of more rural, more working class white regions. That was particularly important in the three above mentioned states because they’re whiter than the country at large and have more people without college degrees. At the same time, Clinton slightly under-performed in her areas. The theory would have to be that an outside tamperer goosed those numbers in those rural, conservative white areas. But remember, we didn’t know this was going to happen going into election night. In fact, the vast majority of experts thought something different would happen.
So this sort of broad electoral manipulation requires you to react on the fly with manipulation that makes sense or is plausible in the context of the election trends. Critically, you have to do this when there are many different voting technologies, some much more hackable than others. So for instance, let’s say that the working class white voters who seemed to have delivered the election for Trump actually hadn’t. But in the limited number of regions that you were able to hack you goosed those numbers up beyond their historical highs. That really would stick out like a sore thumb. I admit this all goes a bit down the rabbit hole. But my point is that to seem plausible election manipulation has to be done with a solid grasp both of electoral history and an ability to come up with a plausible manipulation based on election night trends you can’t know until the election reporting is underway. That is really, really challenging. This might be easier if there were one national system. You could just create the national trends yourself. But there’s not. Someone systems are much more hackable than others. Many probably aren’t hackable at all. My point is that even if machines are hackable and even if someone was trying to hack them, doing it plausibly would be considerably more difficult than I think most people realize.
6: Finally, the report that people are talking about comes from people who are supposed to be knowledgable and serious people. I’m not familiar with them myself. But that doesn’t mean anything. People I do know and respect are familiar with at least one of them. They take their expertise seriously. We’ve heard that this evidence has been presented to the Clinton campaign. But they seem not to have acted on it. That doesn’t prove anything. But it suggests they don’t think it adds up. Given the seriousness of such a charge and the huge consequences if such a thing had happened, the researchers should simply release their findings. It’s hard even to be a skeptic if we don’t even know what they say they found. That’s the obvious thing to me: release what you found and let experts from various fields scrutinize the data. If time is running short, release the findings today. I’m sure there are lots of statisticians, election data experts, security experts and more who are happy to sit down with the datasheets all through the holiday weekend.
7: Put me down as highly, highly skeptical that anything like this happened. Highly skeptical. But again, maybe the evidence they found would change that. So release the findings.
8: I will say that going into election night I had serious concerns that there would be election tampering, but of a different sort. I mean just sabotage. Making votes disappear, crashing systems. This is far easier to do because you don’t have to hide what you’re doing. You can do it in plain sight. It just disrupts the process, makes people doubt the results. Maybe it makes it impossible to count the results if a bunch of results have simply been lost, etc. etc. But again, this is entirely different – in terms of the challenges of pulling it off – than the kind of manipulation that is being alleged.
9: I’m highly, highly skeptical. But again, release the findings. There are lots of statisticians and computer security experts who could say yea or nea pretty quickly.