This is a short post in my constitutional design series covering prime minister selection and post-election procedure (including snap elections) in parliamentary democracy.
Prime Minister Selection
Under simple assumptions, game theory says that the median party in the legislature ideologically should get the prime minister spot and all the power in a parliamentary democracy - they can side with either the left or the right, so each side has to give them all they want or they will side with the other.
This may not hold under more sophisticated assumptions. A small party may face repercussions in the next election if they’re being too much of a stickler about the PM spot, for instance, which discourages them from trying too hard for it. Alternatively, the left and right might make a sort of pact to not let the median party be prime minister, perhaps if they prefer the other being in power half the time to the median party being in power all of the time.
In practice, the median party rarely gets the prime minister spot. I asked ChatGPT Deep Research for some estimates, and it said:
The median party gets the prime minister spot about 20% of the time. 70% of the time it is instead the largest party in the coalition, and another 10% of the time it’s a different smaller party. Obviously take those numbers with a grain of salt.
The main factor determining whether the largest party gets the PM spot is how many seats they hold: in a more fragmented system where even the largest party doesn’t hold that many seats, they’re much less likely to get the Prime Minister spot.
Over time, we’ve been seeing the median party in the legislature get the prime minister spot more often.
It sure seems to me like we’re looking at a democratic norm, that the largest party in the coalition has some “legitimacy” and thus gets the prime minister spot.
The choice of whether or not to have this norm is a lot like the choice of which single-member electoral system to have in presidential democracy, one that encourages polarization or one that selects the Condorcet winner (the candidate that would win a 1-v-1 election against any other candidate). The main difference is that when you implement the norm for the legislature, it messes up the legislature representation, because it incentivizes voters to coalesce around large parties to go after the prime minister spot - and similarly for small parties to form alliances. If you have proportional representation with low thresholds, this defeats some of the point: if you want this behavior, you can always just adopt high thresholds.
You might still like this norm, and might even consider making it a forcing mechanism instead:
A prime minister selected must be a member of the largest party (party label given at election time, no shenanigans where you switch party afterwards) for which 90+% of its members selected said prime minister.
Probably I’d only make this mechanism kick in if the party in question got 25+% of the vote, so it doesn’t apply in very divided landscapes.
If on the other hand you don’t like this norm and would rather have the median party take the prime minister spot, here is what you can do. You can make it so that if in a few months after the election there is no coalition with majority support, you have a vote on who the prime minister is among the legislature using one of the Condorcet ranked methods. This encourages the legislature to select that Condorcet candidate.
Controlling the Agenda
I’d like to discuss some of the considerations around the choice of whether to have this norm that the largest party in the coalition gets the prime minister spot. A lot of what I’m saying also applies to the choice of electoral system in presidential democracy, I just did not get to in my electoral systems post.
I think to fully understand the choice, you have to think about the point of the leader to begin with. As I laid out in my post on the leader, my vision is that the leader controls the agenda of a country, determining pieces of legislation the legislature spends their time on. Maybe it actually makes sense that the faction that gets to control the agenda is the largest one in the coalition. A small centrist party that happens to be the median in the legislature is going to have veto power over all laws anyway - why would you want them to control the agenda?
But then again, how much does controlling the agenda really matter when you need the median party to sign off on any action? I argued in my piece on the leader that it’s a bad state to be in when the leader does not command a majority in the legislature. They can’t get anything done, so they end up being blamed/credited for things they can’t control and can’t do things that might make them more popular. Is being the prime minister from a left-wing party that has to rely on a centrist party for a majority that different?
So idk. This is an area where there’s been enough evidence that there is probably an empirical argument you could make, but that’s not for me to make, so I’ll just leave it as uncertain.
Post-Election Procedure
Within a few months of an election, a majority coalition should usually be formed. If not, there are a few options.
One is to simply let the previous prime minister stay interim prime minister indefinitely, alongside the other previous ministers. If the prime minister has the power to fire ministers, they can replace them; if not the other ministers are likely locked in as well.
Another option is to have an election in the legislature using a Condorcet ranked method, as I stated above. You can hold it using a different method if you’d like, but strategic voting means that you’ll likely end up at the Condorcet option anyway. If the prime minister does not have the power to fire ministers, either those other ministers will stay there indefinitely, or you can hold elections for the ministers using a Condorcet method as well. The latter approach has the effect of incentivizing the median party ideologically - the one who likely wins those Condorcet elections - not to agree to any coalition where they don’t get all the power; I dislike this, but I also don’t think it’s a crazy way of doing things.
A third is to do what I suggested in my post on non-partisan entities and make all legislators (or some fraction of them chosen at random) ineligible for re-election if they don’t choose a leader following the election. This adds really high stakes so I don’t love it, but it’s an option.
Finally, and the option almost always chosen in practice, you can let the prime minister stay there in the interim but hold a snap election.
Snap Elections
As I discussed in my post on the leader, the main reason times of no majority are bad is that a lot fewer laws are passed, often including no new budgets, which usually means the budget from the previous year is used. The upside of snap elections is that if it produces a majority coalition, you get to avoid that.
There are a few downsides.
One is that campaign season tends to be a bad time when it comes to governance. People tend to prioritize taking popular positions over good governance, and often they’re too busy campaigning to really do major governance.
A related downside is that because everyone knows there’s an election upcoming, people don’t feel the need to try to compromise and govern. It could be that if you didn’t have snap elections, politicians would actually work together more and pass legislation even if there’s no majority coalition.
Another downside is that not holding elections at regular intervals makes it impossible to hold local elections concurrently, which leads to lower and less representative turnout for local elections. And if you hold snap elections at the local level, politicians may have the incentive to call snap elections so that they’re not concurrent with national elections if they want this unrepresentative turnout.
Lastly, a byproduct of snap elections is that the majority can call a snap election whenever they want, so they can choose to do so when they sense a political opportunity. Because poll numbers naturally fluctuate over time, controlling when the next election is held gives the majority a big leg up in the next election. Granted, it’s pretty common for parties to call new elections only to see their major poll leads evaporate - and you should (IMO) force there to be eg 6 months between an election being called and held to make these more common. Still, controlling the election timing is bound to give you some advantage. I think this is just an unfair advantage that makes things overall less democratic, but I could see a worldview where you like incumbency advantage and so see it as a good thing.
If you don’t want to give up on snap elections, there are a few mechanisms to avoid their downsides:
You can limit the number of snap elections that can be held without successful government formation
You can force elections to be held at regular intervals in addition to any snap elections that may occur in between.
This greatly reduce the incentive of politicians to call an early election for a leg up
It also allows you to schedule concurrent local elections
You can limit the number of elections you can call during a certain period. For example, you can prevent one from being held immediately following a previous election
You can require supermajority (maybe 2/3) to call a snap election instead of simple majority.
This ensures that the majority can’t simply call them to gain a political edge; you need bipartisan consensus that a snap election would be beneficial.
I’m pretty open to all options. I think the option I lean toward is the following: implement 1 - one regularly scheduled election every 4 years - and have a maximum of 1 snap election held between these regularly scheduled elections that can be called by simple majority. This is good in avoiding election fatigue, which serves little purpose, discourages good governance, and can make people jaded about parliamentary democracy in general.