Sunday, January 18, 2015


Thesis: Social networking technology can be harnessed to create engines of social decision making which can improve, augment and ultimately  replace traditional governments.

Currently in a US presidential election the average citizen contributes 1/200-million'th of a bit of information to a binary hiring decision between two candidates who lied on their job applications. Is this the best we can do? The US constitution attempted, within the confines of 18th century technology, to strike a balance between democratic input and political stability. A key design choice was representative rather than direct democracy, in which citizens delegate their political decision making authority to elected officials and representative bodies. In the era when it took days to travel by horse to the locations of the deliberative bodies there was little alternative. But now?

Direct Budgeting

Consider that one of the main tasks of government is to set annual budgets. Could this be done by an electronic consensus of the electorate? A budget is a tree of allocations, with a pie-chart at each node: this much for defense, this much for roads and bridges, and so on. What if every citizen got to set their preferred allocation, and the budget emerged as some fair aggregation of the individual inputs? Of course not every citizen will have the stamina to wade through the entirety of a federal budget tree with tens of thousands of nested pie charts. The representatives bodies could propose default allocations that would apply unless the citizens overrode them. Citizens could browse as shallow or deep in the tree as they wish to try to influence the allocations -- most would probably look at only the top few layers. Citizens with a particular issue interest -- potholes on their street, rights or the disabled, gun owners, defense contractors -- could drill down just on the subtrees they care about.

What would constitute a fair aggregation method? One possibility is to take the average (or mean) of the allocations of the individual voters, but this encourages single-issue extremists to take the most extreme positions possible in order to maximize their influence on the outcome. For example, a defense contractor trying to maximize defense allocation might choose to zero out everything in the budget except defense. While it is possible that the average of many extremist allocations pointing in different directions might be a reasonable policy, it is also likely to be unstable. Statisticians have a solution to the problem of averages being dominated by extreme outliers, called robust statistics. The simplest robust statistic is the median, which is similar to the mean in that it measures a central tendency, but unlike the mean it is relatively unaffected by extremes. Taking the median of participating voters allocations at every level of the tree would produce a robust consensus budget.

Median budgeting may suffer from unfairness in some circumstances. Suppose 51% of voters want to allocate 100% of the budget to guns, while 49% want to allocate 100% to butter. The median will side with the majority and allocate 100% to guns, effectively disenfranchising the minority, however large. A mean-based system (or equivalently, one in which everyone decides how to spend their own fraction of the total budget) would allocate 51% to guns and 49% to butter, which seems more fair.

However a mean based system encourages a type of extremism, in which one overallocates to one's pet cause(s), in an attempt to drag the mean in a preferred direction. Such extremists advocate a budget which is more extreme than they want, in the expectation that dilution with the votes of others will achieve something closer to what they really want. It is hard to design a system which avoids disenfranchising large minorities and yet is robust to extremists. One possibility is to use a trimmed or truncated mean. Judged sporting events often use this approach by throwing out the highest and lowest scores.

Any democratic budget-setting process is subject to attempts by interested parties to game the system. The current American democratic process has been to a large extent captured by special interests who have learned that the investment of comparatively small amounts of money to purchase political influence pays large dividends, often at the expense of the interests of the majority of the citizens. In the proposed democratic median budget, opportunities for influence peddling and purchasing are reduced but not eliminated. By limiting the role of representatives to setting the default budget, the influence of backroom deals in smoke-filled rooms is reduced -- the public can override those deals, not just by the relatively coarse-grained method of electing different representatives in the next election cycle, but at the level of the line-item. (The public, in effect, gets the line-item veto that Ronald Reagan lobbied for, unsuccessfully, as a presidential power.) Since most citizens will not be interested in second-guessing every detail of a complex budget, the ability to set budgetary defaults would remain a significant type of power, worth purchasing in many cases. Representatives, for better or worse, would still play a significant role, as will campaign contributions / bribes.

Another opportunity to game this particular system resides in the structure of the budgetary tree itself. Assuming that most citizens will only peruse the top few levels -- expressing their views on the relative balance between defense and social spending, say, or on environmental, educational or infrastructural priorities -- the best way for a special interest, working through their captive representatives, to encode their priorities is to bury the funding for, say, the F-37 fighter deep within a budget subtree with an innocuous, public-friendly top-level description, such as "military readiness". It is much more challenging to design a democratic algorithm that creates budget trees than for allocating resources within an existing tree (perhaps a subsequent topic).  The fact that opportunities remain for representatives to exert (and sell) their influence may be a feature of this system rather than a bug, otherwise they would relentlessly oppose its implementation.

Direct democratic budgeting could perhaps reduce the effect of horse-trading. Currently budgets are created by a series of compromises ("you don't want to know how sausage is made") in which representatives sell their consent on some issues they care less about in order to secure a coalition in favor of issues that they or their patrons care more about. The resulting "camel" budget is a pastiche of special-influence interests. If the public's "line-item veto" can override these compromises, representatives may be less able to trust the exchange of favors. Would the sausage making process break down? Will representatives become more prone to "vote their consciences" (whatever that would mean), or perhaps even, as a last resort, the interests of their constitutents? A presidential line-item-veto would have had a similar effect of undermining horse-trading, which was probably one reason it never received serious consideration. However a president can telegraph his intent on particular items, and the representatives could take that information into account, discounting the value of a veto-likely item in budget negotiations. It may be that public-opinion research, or transparent early voting, could play a similar role in providing advance warning of the likely effect of democratic choices on budget priorities, thereby allowing more effective market-pricing of vote trades and campaign contributions.

Another effect of democratic budgeting would be to provide a more powerful role for media in publicizing where the budget is being captured by special interests. While there is already plenty of coverage of these issues now, it tends to have minimal impact on the sausage-making process, which can shrug off the complaints of an outraged punditocracy as long as the money flow is not affected. In the world of direct budgeting, however, a reporter or blogger who blows the whistle on some piece of budgetary misallocation could spark a popular movement to correct the problem.

Since direct budgeting would reduce the ability of special interests to capture the budgetary process by purchasing representatives, it will force those interests to make their cases to the public directly. In effect, they will have to bribe the electorate rather than their representatives, which is likely to be more expensive. The cost of political influence will increase, but the benefits will flow directly to the voters. In effect the voters can sell themselves out rather than electing representatives to do it for them. The current system, by contrast, may be seen as a system where representatives collude with special interests to sell out the public at a discount and split the savings.

The notion of a single default budget created by elected representatives may be unrealistic, since nothing would stop political parties, or indeed any other organizations or individuals, from proposing their own budgets, and enabling voters to endorse them in whole or in part. More or less formalized trust networks can delegate budgeting authority among citizens without the need for elected representatives. This contradicts what I said above about preserving a role for representatives in the brave new world of direct budgeting. Do I contradict myself? Very well, then I contradict myself...


Could the direct budgeting idea could be extended to the revenue side, to set taxation policy? Here the pie chart is collected rather than disbursed, an even more contentious subject. A tax policy that was strongly democratic would likely turn out to be a populist, soak-the-rich, Robin Hood policy, since the poor are more numerous. Arguably one of the main roles the American government has assumed over the years has been to deflect this natural democratic tendency away from its patrons. On the other hand, as Piketty has argued, the Matthew effect makes our current economic system unstable in the long term unless a redistributive counterforce exists to stabilize the wealth and income distributions. Perhaps a taxation policy that was more responsive to majority interests would provide that counterforce. The risk is that it could overshoot into Soviet-style redistributive dystopia. Whether a democratic taxation system could avoid such an overshoot is unclear. As long as an individual voter perceives a net benefit by increasing the tax rate on those with higher income (or wealth) than himself, he might be expected to favor an increase. Would a reasonable equilibrium exist? If everyone gets to vote on everyone else's tax bracket, they might all vote to zero out their own bracket and maximize everyone else's. In this case, the most numerous bracket wins, i.e., the poor, and you get runaway redistribution with a completely egalitarian equilibrium.

One possibility is to allow democratic input on the parameters of a constrained taxation model. Consider a simple flat tax system with two parameters: the individual deduction I, and the proportional rate R. An individual's income taxes would be determined by T=max(0, R*income - I).
What would we expect if median voting were used to determine I and R? Since the poor are most numerous, they zero out their taxes by having I > income, making all of their income tax deductible. Conversely, R can be arbitrarily large, since it only affects the rich, i.e., those with incomes greater than I. In the equilibrium limit, you could have confiscatorily high R applied to a vanishingly small fraction of the superrich defined by a very high I.  Not even the most ardent modern leftist would endorse that outcome, unless they were completely oblivious to the redistributionist excesses of the 20th century.

If we assume that the taxation policy is designed to yield a certain total expected revenue which is determined independently of taxation parameters, then the policy would have only one degree of freedom. It would suffice to allow the voters to choose the individual deduction I, with the marginal rate R following from the revenue target and the income distribution. The policy could also be designed to encompass nonflat (progressive/regressive) options by adding additional parameters to govern the curvature of the tax vs. income function.

If a democratic flat tax does not have a reasonable equilibrium, are there other parameterized models which might? One possibility is weighted voting, where one's parameter choices are weighted according the proposed taxation level that they would imply for the voter. Here the rich would have greater influence, if and only if they pay more taxes. This would generate conflicts between the liberal rich, the Warren Buffet types, and their more conservative "keep what you got" counterparts, with an expected equilibrium shifted somewhat towards higher taxation by those Buffet-like "class traitors" willing to impose higher taxes on their wealthy but less generous brethren. Moreover the rich would have far more control over the parameters impacting the poor than the poor, who would be at the mercy of the equilibrium between the compassionate rich and the let-them-eat-cake conservatives, who would try to drive the individual deduction downward. This might drive the poor to an antitax rebellion.

Weighting by expected taxation would be difficult to implement, since it would be hard to assess expected impact. Perhaps weighting by last year's taxes would suffice.

Should there be a link between taxes paid and influence on budgetary allocations? The limit of this linkage would be a purely philanthropic budget, government by charity. This arguably has been tried at various points in history ("gilded ages") and been found wanting.

Another candidate for democratic parameter setting is monetary policy. Who needs the federal reserve? Let the whim of the electorate determine the rate of printing money. What equilibrium would we expect? This depends in part on what is done with the money that is printed. The inflationary effects of increased money supply probably fall hardest on the poor, so they might be expected to oppose loosening, unless they can persuaded to see contrary benefits, such as increased employment. The Federal Reserve was created out of the realization that elected officials inherently lack the incentives to use the power of the money presses wisely, i.e., in a manner consistent with a stable equilibrium rather than a runaway. It is not clear that the public would be any less prone to instability.

Any such electoral machine constitutes a kind of steering mechanism for public policy, and creates a risk that the electorate will oversteer. Oversteering arises when there are time lags in a feedback system: the steering process overcompensates for the slow response of the underlying system, and then overshoots the target. The result is oscillation around the target. Something of the sort may be happening with our current electoral system, where one party causes damage, the other cleans it up as best it can before it gets thrown out for not fixing things fast enough, allowing the other party to resume its damaging ways. (Either party can probably agree with this story while disagreeing on who is who.) Will the public collectively learn to drive such a device over time? Most of the electorate is relatively innocent of policy considerations. Would they learn? Or delegate? Or oversteer into such violent oscillations that the system breaks apart?


No comments:

Post a Comment