In the past few years, there have been many awesome mechanism design developments (e.g. quadratic voting, quadratic funding, product nash rule). These mechanisms are proven optimal for a variety of things like efficient preference expression, incentive-compatibility, decomposabiltiy and robustness. However, the research often leaves open the question whether the outcome of the mechanism should be.
More concretely, should people vote to choose a representative? Or should they vote directly to implement an actual change in legislation or budgeting?
If direct democracy is to gain more support, it would be great if they can be shown to be superior to some form of indirect democracy (where the public votes for representatives, not policy).
Let's dive into how a potential comparison between direct and indirect democracy can be made.
How to compare voting mechanisms
(1) Principle Agent Problem
A first factor to look at is the Principle Agent Problem. It aims to explain how the introduction of an agent to act on someone's behalf, creates the potential for the agent to act in ways which do not correspond to their preferences.
In some way, some direct democracy schemes might have this problem too, when your outcomes can be influenced by other people. However, in many direct democratic mechanisms, selfishness of all players are assumed anyhow, overcoming this problem. This entire factor is therefore basically in favour of direct democracy.
"The starting assumption instead is that an agency problem exists between voters and their elected representatives because of free rider problems in monitoring and disciplining officeholders, giving elected officials leeway to pursue policies that are not in the interests of their constituents (Kau and Rubin, 1979; Kalt and Zupan, 1984; Peltzman, 1984)" - source
(2) Information problem
A second factor which we can look at when comparing voting mechanisms is the Information problem. There are two types of information to take into account when taking decisions: people's preferences, and states of the world. These types of information may either inform you about current or future events, which gives us the following 2x2 diagram of information which people need to take decisions:
|Current states of the world
|Future states of the world
People themselves are the best judges regarding the current preferences of themselves and those around them (top left quadrant), but it might be hard for them to judge their future preferences (bottom left quadrant), as well as how the world works in general.
There is a lot of evidence point out the lack of knowledge of "the public" and that voters are frequently uninformed about public policy. On the other hand, there has also been lots of evidence that experts in many scenario's don't do better than the average person when it comes to predictions. To complicate matters even more, the only way experts find their way as representatives in most democracies is through voting, making it very context-dependent whether or not direct or indirect democracy would do better:
"One could argue that, if anything, uninformed voters are more likely to make mistakes when voting on candidates than ballot measures because candidates represent bundles of issues and characteristics, while ballot propositions typically involve only a single issue" - source
Are some decisions more suitable for direct democracy?
So it seems that despite the Principal Agent Problem, indirect democracy might still perform better for certain decisions because representatives have better information. Can we deduce when representatives would perform better?
To answer this question, there are a number of empirical findings which might help (though they are limited in their sample size and how well they determine causality - source):
- "initiative states cut the combined spending of state and local governments by about 5 percent and cut state government spending by over 10 percent"
- "iniative states also seem to adopt more conservative social policies"
- "iniative states also keep their governments on a shorter rein" - through shorter term limits
- "A fair amount of evidence supports the overspending theory, particularly its implication that spending increases as the number of representatives increase" - source
The complexity of the subject at hand makes it untenable to find a generic answer to our question. Perhaps the only two moments when we can be certain that direct democracy will be better than indirect democracy is when:
- the issue is mainly about defining a value, such as "whether to use capital punishment or allow physician-assisted suicide"
- the people that will be influenced by a decision can be identified, and the level of knowledge about the issue in the public is well distributed, such as when spending a very local city budget
For all other questions, I'm unfortunately still at a loss for an answer. There just doesn't seem to be enough data! So is there any way to get more data?
How to compare different governance systems in practice
Alright, let's say that we are considering to implement direct democratic mechanisms like quadratic funding or voting on prize winners. Comparing their effectiveness to an indirect democracy is hard.
Doing a deductivate evaluation is hard because indirect democracy involves the decisions of free-thinking political representatives. The way that we overcome this limitation in economics, is by pretending the agents are fully egoistic Homo Economicus - they take incentives seriously. However, politicians are usually shielded from incentives to a large extent. Arguments like the median voter theorem might indicate that representatives are likely to pursue the median preference of its voters in certain scenario's, but every layer of indirection weakens the economic assumptions.
Doing an empirical evaluation is hard because it is unclear what should be regarded as optimal when comparing the two systems. We get into a recursive problem: who is most capable to tell whether a representative or the public is most capable, a representative or the public? This is a deep constitutional question. If you think this is a purely values-based question, then it should be the public who decides on this. On the other hand, if you think that experts might have some knowledge on this... you know the drill. For now, let's say this problem is outside of scope for now, we can choose a welfare scoring rule, and hope the democratic apparatus figures things out for us.
Imagine we see Direct Democratic Approach A achieves better results than Indirect Democratic Approach B, we can start to implement more of Approach A or adjust Approach B.
We can argue that we should be doing a lot more experimentation, to let multiple programs run side-by-side so we can A-B test which program people are actually happiest with. It would be great if we can go about as rigorously as Twitter's A/B testing team. Call it scientific governance. Or - as Robin Hanson does in his seminal work Shall we Vote on Values but Bet on Beliefs? - we can take an engineering approach and encourage experimentation with different processes.
Experimentation is worth it as long as the cost of experimentation is less than the expected value. So above all - bringing the costs of experimenting with scientific governance down itself would be a worthy goal.
Despite some concrete issues (voting on values or local city budgets), I still have a hard time understanding whether direct or indirect democracy would perform better. Perhaps more diversity and analysis of governance systems would allow us to gather more knowledge in this area. The scientific revolution should make its way to governance.