Great idea. After learning about ancient Athenian democracy, and the multiple instances of inherent randomness they employed, it got better to wonder why modern democracies seem to incorporate no deliberate element of randomness at all.
Switching up the incentives might cause even larger problems that are unexpected though. Maybe, rather than focusing on policy early in their terms they just do a constant slow burn of focusing on reelection. Just because the expected date is 5 years away, doesn’t mean they won’t get hit after 1. It could also lead to a lot of popular unrest, where a suddenly unpopular official, out of sheer luck, remains in their position for years.
The ‘constant slow burn’ of election prep is part of the intention. Ultimately you can’t stop politicians from trying to get re-elected and you want them to be somewhat disciplined by expected public reactions, the point of the memoryless process is that you can precisely tune the tradeoff politicians face and make that tradeoff consistent over time. If officials are spending too much time on election prep, then tune the probability of election down until they are making the tradeoff you want.
On the ‘unpopular official’ point, I would note that this is already a problem for existing terms, though they certainly have the advantage of bounding negative outcomes. It might be possible to mitigate this with maximum term limits as mentioned by @plasmabloggin but I think doing so might also break the memoryless property of the incentives.
Very fun idea, though the incentives would shift as soon as they find out they're up for reelection, and, if something like this were ever implemented, it would probably be a good idea to still have some maximum term length just to keep someone from getting really lucky and staying in power for a long time against the will of the voters. Though perhaps you could pair it with some process for recall elections to keep that from happening. The change in incentives once someone has been selected for reelection could also be mitigated by making that period short, which would probably be good anyway.
In Denmark it is actually the prime minister who gets to decide the time of elections, within a maximum window of some 4 years. This adds a new layer of planning to politics, as they usually try to keep their term for as long as possible while at the same time looking for a moment where their own party is doing particularly well, and hopefully the opposition is down on some reason or other.
I think your random system would work better. One problem though is how to make sure nobody is cheating with the random generator, since a lot could depend on exactly when it comes down. I wonder if we have some lottery method that is completely hack proof.
Yeah, unfortunately, it does seem difficult to find a source of randomness that people could trust. Ideally, you would have a completely deterministic process that everyone could verify the outcome of, but which people are uncertain about beforehand, and you could use that as your source of randomness. For example, you could use the nth digit of pi as your source of randomness. The difficulty is in making the outcome uncertain enough beforehand, since if the barrier is merely computational, then people may just be able to invest in more computation to learn the outcomes before they are meant to.
Also some offices should be selected randomly, by lottery, not elections (sortition). While it does not result in better candidates than the average people, it does not result in worse ones either. They would call out corruption, probably. Also the American judge+jury system shows that random people informed by one expert can make good decisions. One half of the house would be the expert but crooked career politicians, and one half random people, non-experts but presumably honest (because it is their tax money being wasted).
Doesn't this risk legislators always acting as if they're up for reelection in the next year, because they could be? Under the current system, a Senator is guaranteed not to face reelection for six years from the start of his term in office. Do we really want to weaken that guarantee?
In that case you would want to lower the probability of election (and thereby raise the expected future term length). The coin flip mentioned is not meant to have a strictly 50/50 probability. If an expected time remaining of 6 years is optimal, then you could tune the probability to accomplish that. This should still work even if officials are risk averse, since that just increases the optimal expected time remaining.
Great idea. After learning about ancient Athenian democracy, and the multiple instances of inherent randomness they employed, it got better to wonder why modern democracies seem to incorporate no deliberate element of randomness at all.
Switching up the incentives might cause even larger problems that are unexpected though. Maybe, rather than focusing on policy early in their terms they just do a constant slow burn of focusing on reelection. Just because the expected date is 5 years away, doesn’t mean they won’t get hit after 1. It could also lead to a lot of popular unrest, where a suddenly unpopular official, out of sheer luck, remains in their position for years.
The ‘constant slow burn’ of election prep is part of the intention. Ultimately you can’t stop politicians from trying to get re-elected and you want them to be somewhat disciplined by expected public reactions, the point of the memoryless process is that you can precisely tune the tradeoff politicians face and make that tradeoff consistent over time. If officials are spending too much time on election prep, then tune the probability of election down until they are making the tradeoff you want.
On the ‘unpopular official’ point, I would note that this is already a problem for existing terms, though they certainly have the advantage of bounding negative outcomes. It might be possible to mitigate this with maximum term limits as mentioned by @plasmabloggin but I think doing so might also break the memoryless property of the incentives.
Very fun idea, though the incentives would shift as soon as they find out they're up for reelection, and, if something like this were ever implemented, it would probably be a good idea to still have some maximum term length just to keep someone from getting really lucky and staying in power for a long time against the will of the voters. Though perhaps you could pair it with some process for recall elections to keep that from happening. The change in incentives once someone has been selected for reelection could also be mitigated by making that period short, which would probably be good anyway.
In Denmark it is actually the prime minister who gets to decide the time of elections, within a maximum window of some 4 years. This adds a new layer of planning to politics, as they usually try to keep their term for as long as possible while at the same time looking for a moment where their own party is doing particularly well, and hopefully the opposition is down on some reason or other.
I think your random system would work better. One problem though is how to make sure nobody is cheating with the random generator, since a lot could depend on exactly when it comes down. I wonder if we have some lottery method that is completely hack proof.
Yeah, unfortunately, it does seem difficult to find a source of randomness that people could trust. Ideally, you would have a completely deterministic process that everyone could verify the outcome of, but which people are uncertain about beforehand, and you could use that as your source of randomness. For example, you could use the nth digit of pi as your source of randomness. The difficulty is in making the outcome uncertain enough beforehand, since if the barrier is merely computational, then people may just be able to invest in more computation to learn the outcomes before they are meant to.
Also some offices should be selected randomly, by lottery, not elections (sortition). While it does not result in better candidates than the average people, it does not result in worse ones either. They would call out corruption, probably. Also the American judge+jury system shows that random people informed by one expert can make good decisions. One half of the house would be the expert but crooked career politicians, and one half random people, non-experts but presumably honest (because it is their tax money being wasted).
Doesn't this risk legislators always acting as if they're up for reelection in the next year, because they could be? Under the current system, a Senator is guaranteed not to face reelection for six years from the start of his term in office. Do we really want to weaken that guarantee?
In that case you would want to lower the probability of election (and thereby raise the expected future term length). The coin flip mentioned is not meant to have a strictly 50/50 probability. If an expected time remaining of 6 years is optimal, then you could tune the probability to accomplish that. This should still work even if officials are risk averse, since that just increases the optimal expected time remaining.