Ditch the President, Vote on the Cabinet
I often talk about other peoples' ideas about government and policy which I like, and feel a certain discomfort in discussing original ambitious ideas in this area. In large part this is because I am not an expert in political economy. I know there are going to be important considerations I am missing if a view hasn’t been hashed out by others already (well, more egregious gaps than usual).
That said, there is a political idea of mine (which I’m sure other people have come up with, but I haven’t seen spelled out elsewhere) which has gotten a warm reception from everyone I’ve discussed it with, despite it entailing a radical change to current systems of governance. This seems like a good place to share it, some strengths it has, and possible objections to it. My main comparisons here will be to a presidential system since it is the one I am most familiar with, but I believe the strengths I discuss can be applied to other forms of liberal democracy just as well.
The idea in question is something like this: replace the presidential election with a cabinet election. Not necessarily something exactly like a presidential executive branch, but some sort of powerful branch of government that elects representatives issue to issue. The basic justification for this is that electing one generalist dilutes both the democratic and technocratic strengths of a governing body. Currently, a president can be elected on the basis of their stance on one issue and has equal power over every other issue they have some opinion on. In an ideally functioning democracy, winning candidates would tend to approach being those with the most majority-favored opinions on a variety of issues, but this force does not seem very strong in practice. Our current election system in particular tends to funnel candidates into two camps, neither of which, even when they win, represents the majority view on each of their positions.
The primaries in particular almost guarantee that plausible general election candidates first need to give a credible signal of their partisan leanings based on their platforms. Primary voters will tend to vote for the candidate who is pleasing to wider society to the minimum degree necessary for them to win the race. In an ideally competitive world, this would result in a race to the bottom between the two parties resulting in candidates that are more and more pleasing to the majority. It seems that at most this occasionally gives more centrist candidates an edge, but by no means works ideally, or else general election candidates would tend to be closer to each other than to their original party platforms 1. It also doesn’t help that only some elections involve simultaneous primaries by both parties. For those that don’t, the primarying party has no incentive to approach the most popular positions any closer than necessary, even in theory. After all, they already know how electable their opposition will be, and their choice won’t affect that 2. Given this stagnation, it is unlikely either party will ever approach the majority position on each issue fast enough to get anywhere near it before the majority positions change. The result is an unresponsive, binary system of government.
Even if our voting system were changed, and significant competition between multiple parties took place, politics would approach majority positions only gradually at best, and still would be likely to hit a suboptimal equilibrium point determined by how fast the majority is approached versus how fast that majority changes. Electing an official for each issue (or some natural clump of issues) would be the far more direct way of achieving this ideal than hoping the electoral incentives line up and the implicit trades are made in time to happen upon the ideal point for each issue. Not only does an alternate voting system fail to solve this issue as well as this more direct method, my proposed system could address counter-productive polarization even within a FPTP, two party system, as each specific issue will be split into two options, rather than the entire slate of issues. An alternate voting system would still improve the available options within this issue by issue vote, but would arguably be a much more minor step.
While people often worry that greater direct democracy like this can erode expertise, this issue by issue election could actually improve expertise overall. Perhaps your average individual candidate who believes every popular position will not be as good as your average candidate who believes something close to the most popular views in each area. In particular there seems to be little reason to assume that the set of the most popular positions at a given time will all be correlated with a coherent, well-researched worldview that a single expert might arrive at. More likely the winner will be selected from the pool of people who don’t have coherent worldviews that connect different issues, or who only run on each position because they are popular. However, it seems clear to me that putting every one of these issues in the hands of a single candidate is trusting too much in our ability to find exceptional generalists.
Electing more domain-specific experts will still involve some populist pressures, but it is also more possible to find institutions and accolades that will reliably signal strong domain-specific expertise. If someone has worked on one problem their whole twenty year career, they probably understand its nuances reasonably well. In terms of the information synthesized through an election, an issue by issue vote seems likely to be both a democratic and technocratic improvement over the pick-the-most-popular-general-candidate system.
A remaining objection might be that splitting politics into independent, issue by issue branches could lead to an incoherent vision, or even a conflicting one. As an example, let’s say that most voters in an electorate want more welfare, and most voters want more immigration, but, because of fiscal concerns, they do not want the combination of the two. The problem is that people conditionalize their preferences about a given issue on the status quo elsewhere, since they don’t know until the election what the other winners will look like 3. Concerns of this sort are the ones I take most seriously, they seem likely to be a genuine problem.
One possible solution to the incoherence problem is to elect the positions one by one over a long period. This way, for each new role, people will decide which candidate they most want for a given role conditional on who they already know will be in the other roles. The limitation of this possibility is that there is no way to know in advance the ideal order.
Continuing my hypothetical, maybe the pro-welfare candidate is elected first, and so an anti-immigration candidate is elected next, however most people with the stated preferences would prefer more immigration to more welfare. The apparently optimal solution to this is to run every possible combination of issue representatives against each other at once, so that the most popular combination wins. This is something like just electing a normal generalist candidate, but with a more salad-barish granularity. Many-option single-winner elections all have similar theoretical problems, so despite improved granularity, and especially given the sheer number of combinations you might end up with, this could lead to democratic problems that are almost identical to those with electing an individual.
One defense of a system in which each issue’s candidate is just elected independently is that our current general candidates aren’t necessarily measured that well for ideological coherence either. As mentioned, our democracy just isn’t very fine-grained, and the existing coalitions often shift. Perhaps some sort of mixed system, in which people vote for candidates separately, and are also able to vote against particularly undesirable combinations if they wish, would be a decent compromise, but more work would need to be done to figure out the precise balance, and I don’t have well-realized ideas yet myself.
There are three other ways this problem might be mitigated, which offer trade-offs but no ready-made perfect balances. One lever you can move is how you group different issues together. If you try to make each department responsible for many of the things that will be most directly associated with each other, and are most likely to involve frequent trade-offs, you will improve the effectiveness of issue-by-issue votes. Another lever is how many departments there are. If you group the departments so that there are fewer of them, then the slate by slate voting option becomes more viable, since there will be fewer options. (At the extreme, in a two party system with only three departments, you can get the field down to eight options, which seems reasonable to list completely on a single ballot if you use a non-FPTP voting system.)
The final lever you can move is the power of some arbitrator. It seems as though a system like this will inevitably require some sort of higher judge that is, for instance, able to handle disputes between different departments with investment in the same issue. The more power you give this arbitrator over these departments, the more governance becomes like a standard presidential system, with the modification of a voted-on cabinet. Even this extreme seems to me like an improvement on a cabinet selected by a single president. Indeed, the reverse might be a good idea, in which the cabinet is elected by the people, and then an arbitrator is voted on by the cabinet. Regardless, pushing any of these levers too far – decreasing the number of departments, grouping too many diverse issues into a single department, increasing the power of some arbitrator – will correspondingly decrease the previously discussed democratic and technocratic powers that come with greater granularity. That said, it seems unlikely to me that the right answer is what we have right now.
Another problem is deciding just what role a branch like this might play in government. Just transferring the powers of the presidency to a council like this seems to map poorly. Could this council simply take over all legislative functions of government and replace the existing congressional/parliamentary systems? While there is some temptation and elegance in this, congress/parliament serves at least two unique functions that would be harmed by any attempt to collapse them into this branch. The most straightforward is that they pick local representatives, and so represent the interests of localities rather than just the popular view on a given issue. This seems like a worthwhile function to me, especially in large and diverse countries like the US, in which legislation might benefit from a more fine-grained understanding of what local interests exist. Getting information about different local interests for policies is harder when you aggregate everything to the Federal level, in the same way that getting information about the popular views on different issues is harder if you aggregate everything into a single candidate.
The other function is probably even more important, which is that, if the country were split into departments like this with sole authority over their own issues, there would be many many laws that would be decided by a single administration unilaterally. Things would be accomplished efficiently, but probably far too efficiently. It would be an extremely winner-takes-all system, and even a serious hazard for dictatorship. All that would keep the departments in line would be other departments constantly making dubious claims of conflicts of interest to be handled by the arbitrator, but making this too easy seems dangerous as well.
A possible alternative to converting the executive branch to these issue-by-issue councils would be converting the legislative branch to these councils. Each locality can elect representatives for each issue, and then legislation on given issues is debated in as many separate congresses as there are separate departments. If there is some risk of gridlock in both the current congressional system, and in arbitrating departmental conflicts in my suggested system, this possibility seems like a recipe for endless gridlock. Sure the representatives will be more responsive to democratic will, but they will be less able to act on it than ever before. Once again, the delay in accomplishing anything will create a suboptimal equilibrium with the time it takes for majority opinion to change. Perhaps even this isn’t the end of the world. It rewards positions that manage to maintain their popularity over longer periods of time, but it seems like a significant risk for sclerotic governance that can’t respond to new situations or trending values well. I think that the ideal role for a government body of this sort would be as its own branch – or as some alternative to the executive branch, not as either the total seat of legislative power or as an alternative system of congress.
Since I have dwelt a bit on some of the strongest problems and trade-offs that need working out with this suggestion, I want to end with a look at one of the aspects of it that excites me the most – the strength of this idea from a RadicalXChange perspective. One of the ideas in Radical Markets that most intrigues and frustrates me is Quadratic Voting. The idea is that, instead of votes, we should be given an allowance of “Voice Credits” which we can spend on votes for different issues. The number of votes corresponds to the square root of the number of voice credits spent on it (to account for the marginal cost to others of extremism). This would allow people to vote less for the issues they have a weaker preference about (are either more uncertain on the right answer to, or care less about), and more about the ones they have stronger preferences about.
While very elegant in theory, in practice elections are rarely split by issues someone might care about, but various representatives one might care about. If someone has a given set of views, and they have already decided which party has the stronger fit for these views as its platform, they might care similarly about most candidates from a given party, or at least it will be harder to choose unless the candidates emphasize in advance which trade-offs they are willing and unwilling to make.
One thing that could help this is increasing the number of elections these voice credits can be applied to, so that you really can compare elections where there is a significant difference in how much you care. Weyl/Posner suggest that these voice credits can be saved up, and spent over multiple elections. This seems crazy to me; making rational choices, not even with linear spending, but with weird square root spending, over the course of many elections you don’t simultaneously compare and which you won’t know about in advance, requires a sort of fiscal literacy that would rival the sleekest investment bankers to even approach organized decision-making.
There are canonical algorithmic solutions for maximizing expected value in some decisions over time – given uncertainty – but even these only get so good, and I’m not sure there is any such solution known for something like quadratic spending. If elections really used QV over time, pretty much everyone would make their decisions with more randomness than intention, and many would likely be disillusioned with what they see as an overly-complicated, elitist election system. 4
The best results for QV would be during a single election, when all options are simultaneously laid out (even then a digital visual aid would be important for making the trade-offs, since the alternative is mixing and matching different possible quadratic distributions by hand over multiple options). Elections featuring multiple ballot initiatives for instance are a good place for QV, but ballot initiatives feature all the problems of direct democracy and, in part because of that, are usually the smaller part of how elections shape the overall policy landscape.
Choosing issue representatives seems like an almost ideal fit for QV. It would be a reliable, multi-issue election format, that would help shape nearly all policy. Answer some concerns with such a system also relies on taking different degrees of concern into account anyway. As Nick pointed out to me, expanding the number of elections held at once will almost guarantee a lower turn-out per election. The strongest defense of this is that much of this will come down to single-issue voters getting to satisfy their one issue, or however many issues they really have developed opinions on. The absence of votes on an issue someone cares less about boosts the relative strength of votes from those who care most about an issue. This is just a binary version of quadratic weighting, which does even better by these lights.
Insofar as I am usually writing to the Effective Altruist or Rationalist communities in my posts, I want to target this post, more than anything, at the RadicalXChange community. I believe it is a uniquely promising idea, both for the movement’s goals of a responsive, less polarized government, and the specific policies favored by RXC, and would particularly welcome feedback and interaction from people of this community. Is there an elegant solution to some of the hard balances I mentioned? Some fatal problem with anything like this? A modification that significantly improves it?
Ed. Note: I recall an argument mentioning a “basket-of-policies” pattern. First, people hold some ideas at the same time as other ideas. Like supporting both weed legalization and gay marriage. Then, tribes form based on those shared ideas. And then more ideas are included based on what the tribe wants, rather than as part of a coherent worldview. Some of the beliefs are held for signalling reasons, especially if the other tribe happens to hold the opposite belief. This would lead to more polarization, especially around the powerful single-person office of the presidency. ↩︎
If you expand the scope to the next election, there may still be an incentive based on re-electability. ↩︎
Ed. Note: This especially comes up in more extremist/polarized politics. “If we only get 90% of our vision, it’s still polluted by the other 10%!”. Or, even further, “If only 90% of the system changes to our liking, the other 10% will undo/negate all the progress!” ↩︎