> Vote buying is not a serious problem

It's not a problem because there is too much "friction". Vote buying is already not impossible but way too hard to establish itself. When it becomes gradually easier it still won't be a problem, because it is still not yet established. So you allow even more changes that make it easier, because it's not a problem yet. Call it a slippery slope argument all you like, but there is a crazy amount of inertia in the absence of vote buying that is protecting democracy now, but that will turn against us once it is overcome. When barriers are lowered so far that the inertia is overcome, the same inertia will make it incredibly hard to get rid of vote buying again. Keeping honest people honest is orders of magnitude easier than making them if they are not. I would not want to risk it without a promises of truly significant gains and I just don't see those with e-voting.

But vote buying is not even the problem I would focus on. Much more pressing is the form of soft coercion that is enabled by allowing voting in what I would call "unchecked privacy": imagine you are part of a group where everybody assumes that all would vote the same. There is a documented tendency (proudly showing off your ballot on Twitter) to scrap vote secrecy in favor of virtue signaling for "the cause", whatever cause that might be. As soon as there is a group with supposedly aligned opinions, the true believers will tend to erode secrecy and establish an expectation that the others follow. Maybe your spouse won't beat you, maybe your friends won't shun you for insisting on voting in secret, but the easy path is to just go with the flow and play along. "What difference does a single vote make?" Optional secrecy is a serious weakness to the democratic process.

All of those things are possible now, yet they don't happen. I see no reason to think electronic voting would make it more likely. And in fact, we have a test case: Estonia. Is there any evidence of any of this happening there?

By definition, electronic voting is easier to game. Simple example: you just asked for evidence on a procedure that didnt have a physical paper trail.

I asked for evidence of a specific phenomenon. The specific phenomenon I requested evidence of would be no easier to have evidence for in the electronic or paper case.

It takes much less energy and time to modify something electronically. Once physical items are required to be collected, edited or destroyed then evidence becomes significantly more durable.

You don’t even need that. In many systems, you can vote multiple times, and only your most recent vote counts. As a buyer you have no way of knowing if your agent didn’t vote after they photographed them self voting for the party you picked.

That's definitely a good improvement over a single final commit, but there has to be some cut-off time and it would be perfectly possible to control the voter's ability to make last moment corrections. Buy/coerce the vote close to the deadline, then keep them occupied until it's over.

A countermeasure might be an undisclosed deadline lottery: a guaranteed voting window until some time t, then allow corrections until an individually randomized cutoff moment t+x, with a sufficiently big range for x (up to two days, perhaps?). Don't provide feedback wether a correction went through or not to make it even more opaque to a possible buyer.

> but there has to be some cut-off time and it would be perfectly possible to control the voter's ability to make last moment corrections

There is. In the Estionial election it is something like the day before election day. But even if there wasn’t if you need to spend the time and effort to coerce your agent for anything longer then few hours, buying enough votes to sway any election is going to be unfeasible.

That means my real vote might not get counted, thanks to the random deadline.

But that only happens when for some contrived, malicious reason you felt pressure to cast a vote that did not reflect your real opinion. The random deadline extension weakens that attack.

People free from interference would simply make their first vote their real one, cast safely before the earliest possible deadline.

That's an excellent point, I hadn't thought of that.

we have evidence of widespread irregularities in casting votes via mail for people abroad in italian elections (there are a ton of italian citizens abroad, due to ius sanguinis rules).

That entails basically going to people, buying a voting slip and casting it on their behalf. Or better, buying packs of slips from officials. For e-voting it would be the same.

Sure, you can increase jail time and reward revealing it, but that is not a panacea, as the existence of _any criminal activity_ proves.

Those kinds of penalties and worse don't stop organized crime, for instance.

Politicians also tend to do all sorts of crazy, risky and/or illegal things to get elected or for personal profit. Nixon and Trump spring readily to mind.

If the reward is large enough, someone will risk it. Sometimes the reward doesn't even have to be large at all -- witness rich celebrities shoplifting, for instance.

People can also be compromised and blackmailed in to committing crimes, or otherwise feel desperate and at the end of their ropes, so they'll try anything.

That's to say that such laws shouldn't be made, but I am skeptical that they'll be enough.

None of that makes any kind of sense. You can't risk it when the equilibrium is that bad for you. Remember, to make any appreciable dent in an election outcome, you need to convince thousands of people to vote for you, and not turn you in, despite the 100k reward. There is zero chance of that ever succeeding.

If there is a $100k incentive for framing someone for vote buying, people will do it.

That's what we have investigative agencies for. We already offer these kinds of monetary incentives to criminals all the time. We also offer them for reporting tax evasion. Seems to work fine there.