I have a fairly conventional list of things I want from government.
I want it limited, accountable and transparent. I want it close enough to the people being governed that they can affect it. I want officials to have enough authority to do the jobs we've given them without acquiring enough authority to become casually dangerous. I want to know who made a decision, why they made it, what it cost and how we get rid of it if it turns out to have been a terrible idea.
I would also prefer not to spend Saturday night reading the municipal budget.
This may be a problem.
Because self-government isn't administratively free.
We tend to talk about citizen participation as though the only alternatives are civic virtue and apathy. Good citizens pay attention. Bad citizens can't name their state senator and are watching Netflix while democracy dies.
But that puts nearly all the responsibility on the citizen and very little on the design of the system they're being asked to supervise.
Maybe the useful question isn't simply whether citizens are paying attention. Maybe it is: How much attention does this particular form of government require them to pay?
Fifty experiments
I started wondering about this while trying to compare the governments of all fifty states.
I started with rankings. There are plenty of them, and many sound useful until you try to use them. A state can be ranked more free, better governed, more business-friendly or more fiscally responsible, but the ranking has already decided how much the things being measured should matter relative to one another.
That wasn't the question I wanted answered. I wanted to know what each system was buying and what it was paying for it. Then I could decide which bargains I preferred.
So for each state I began asking: what hurt badly enough that people changed the rules? What did they build in response? What did that solution buy? And what bill did it send afterward?
Explore the companion 50-State Government Design Review.
The results are rather wonderful. Government turns out to be full of scar tissue.
California's direct democracy grew partly from distrust of representative institutions influenced by powerful interests. Giving voters initiative and referendum power created a route around those institutions. That bought something real. It also meant voters could create complicated policy that somebody else would later have to implement, maintain and perhaps repair.
Texas reacted against concentrated power by distributing authority and embedding limits in a detailed constitution. That bought checks against executive concentration. It also made coordination harder. Then Texas built a Sunset process that periodically asks whether many state agencies should continue to exist: a patch for the maintenance problems created partly by earlier patches.
Florida gives unusually strong legal protection to public access to records and meetings. Excellent. Now find the record. Then find the other record. Then figure out which public body made the decision and connect its decision to the contract, expenditure and result. Transparency turns out not to be the same thing as legibility.
I kept finding versions of this. A protection solves a real problem, then the protection creates a new cost. Another institution appears to manage that cost. Eventually an ordinary person trying to understand why the thing at the end of her street hasn't been fixed may need an organizational chart, three browser windows and the disposition of a minor administrative appeal from 1987.
This isn't necessarily evidence of stupidity or corruption. Sometimes it is evidence that tradeoffs accumulate.
Civic load
I've started calling the result civic load: the amount of informed attention and participation a governmental system assumes ordinary citizens will supply.
There are at least two parts. One is attention: how much must I know, monitor and understand? The other is participation: how much must enough of us actually do? Those aren't the same.
A government can publish every document it possesses and still require an unreasonable amount of attention to understand what matters. A town can hold a perfectly accessible public meeting at two o'clock on Tuesday afternoon and still make participation rather expensive for the person who has a job. A ballot can give me direct control over a complicated fiscal question without giving me the expertise required to understand what I'm controlling.
None of those things makes transparency, public meetings or direct voting bad. That's the trap I want to avoid. The right to supervise government is valuable, but the ability to supervise it depends partly on what supervision costs.
And when that cost becomes too high, the power doesn't simply sit there patiently waiting for citizens to return.
Somebody will pay attention
Government doesn't stop operating because I'm making dinner. Budgets continue. Rules are written. Contracts are awarded. Hearings occur. Ballot language gets drafted.
And plenty of people are paying attention. Lobbyists, agency employees and political staff are paid to pay attention. So are many lawyers, contractors, unions, trade associations, advocacy organizations, journalists, campaign workers and professional activists.
There is nothing inherently sinister about this. Most complicated systems need people who understand them professionally. But the asymmetry matters. The organized participant can send someone to every Tuesday-afternoon hearing. The ordinary citizen has Tuesday.
I wonder whether some of what we describe as capture, disengagement or voter ignorance begins one step earlier. The system may simply require more sustained attention than ordinary lives can supply. If so, telling everyone to become more civically engaged isn't much of a solution. We can't all become unpaid policy analysts.
Delegation isn't surrender
There is an obvious answer to this problem: delegate. Humans have been doing that for a very long time.
We elect representatives. Hire civil servants. Use courts. Appoint boards. Employ experts. Read newspapers. Join organizations. Ask the person who actually understands municipal bonds what on earth the paragraph means.
Delegation isn't a failure of self-government. It may be one of its prerequisites. The problem comes when delegation makes authority too expensive to check.
I don't need to personally inspect every bridge. I do need some reasonably inexpensive way to discover who was responsible for inspecting it, whether they did, what they found and what happened next.
That suggests a standard I like much better than either constant suspicion or blind trust: earned trust with cheap verification. I shouldn't have to reproduce the government's work to determine whether government did its work.
Government needs maintenance
This may be the part of the fifty-state project that surprised me most. We talk constantly about what government should do. We talk much less about who is responsible for noticing when something it already does no longer makes sense.
Who owns maintenance?
Texas has formal sunset review for many agencies. Virginia maintains a catalog intended to make mandates on local governments more visible. Ohio requires implementation plans following certain audits. Other states have auditors and evaluators who can identify failures but cannot themselves require anyone to fix them.
Which raises an almost comically basic question: who owns making sure the whole thing still works? Not who made the original decision, not who audits it, and not who currently administers one piece. Who owns the system?
Software people would find some of this familiar. A reasonably maintained software project has version history, named owners, dependencies, issues, tests, review, reasons for changes and a way to deprecate things that no longer belong.
Government often has all of those things somewhere. The citizen's problem is discovering where. A law authorized something. An agency interpreted it. A board approved something else. A contractor implemented part of it. An auditor later found a problem. The legislature changed the funding. The original official left office twelve years ago. Good luck.
Perhaps every consequential governmental decision ought to leave a visible trail: authority → evidence → decision → implementation → expenditure → result → owner → review date → correction route.
Not because every citizen should follow every trail, but because a citizen should be able to follow one when it matters.
How much liberty can we maintain?
This leaves me with an uncomfortable thought. We often want substantial control over government while resisting the work that control requires. I understand the impulse. I share it.
But some forms of liberty require maintenance. Local control means somebody local has to pay attention. Direct democracy means voters have to understand more than candidates. Transparency means somebody has to look at what was disclosed. Checks and balances mean more than one institution may need watching. Limits on concentrated authority can make responsibility harder to locate.
None of those is an argument against the thing being protected. It is the bill, and there is no reason to assume the bill is always worth paying.
Sometimes an old protection may be guarding against a danger we've merely forgotten. Sometimes it may have outlived the problem that justified it. Sometimes the protection still matters but could be maintained more cheaply.
That's why I increasingly want to ask of institutions: What pain created you? What did you fix? What did the fix cost? Does the pain still exist? And what would happen if we changed the fix?
The last question matters. Every reformer should have to answer it.
A staff for the citizen
Technology may change this bargain. Large institutions already have staff. Ordinary citizens increasingly can too.
A personal AI could read meeting agendas, budgets, audits, proposed regulations and legislation. It could trace a decision back to its source, notice that an audit finding has remained unresolved for three years, or tell me that something I care about is actually being decided next Tuesday rather than demanding that I monitor 900 pages every week in case something relevant appears.
That could reduce one of the oldest asymmetries in government: organized interests have paid attention; citizens have lives.
But there is an important line here. I don't want AI to become the citizen. I want it to become the citizen's staff: research it, monitor it, explain the competing arguments, show me the source, tell me what changed, and tell me when my judgment is actually required. Then give it back.
The vote is mine. The decision is mine. The letter goes out because I authorized it. Otherwise we've solved the maintenance cost of self-government by quietly eliminating the self.
Keep the questions in circulation
None of this is a new way of thinking about government. People have been arguing about concentrated power, local control, representation, administrative competence, citizen participation and the costs of each for a very long time. Many of the institutions I was examining exist because earlier generations had exactly those arguments.
But governmental ideas don't remain useful merely because somebody once thought them through. The questions have to remain in circulation.
Circumstances change. Institutions accumulate. Old protections acquire new costs. New technology changes what citizens can realistically supervise. A bargain that made sense under one set of conditions may still be worth making, or it may need maintenance.
The danger isn't that we haven't invented the perfect form of government. It's that we stop asking what our existing forms are buying, what they're costing, and whether we still accept the trade.
The bargain
After looking across fifty state governments, I don't think there is a frictionless institutional design hiding somewhere that everyone else has simply failed to discover.
Every structure buys something and sends a bill. More localism can mean more control and more fragmentation. More centralization can make responsibility easier to locate while concentrating power. More permanence can protect promises and entrench mistakes. More turnover can prevent entrenchment while transferring institutional memory to the people who don't leave. More direct democracy can increase popular control while asking citizens to make increasingly technical decisions.
The question isn't which one of those is always correct. It's which costs we're willing to accept for which protections.
That may be the maintenance cost of liberty: not endless participation, not reading every budget, and not transforming the population into a nation of amateur administrative-law specialists. Enough attention to know what we've delegated; enough visibility to see what happens afterward; enough maintenance to notice when yesterday's protection has become today's obstacle; and a cheap enough route back into the machinery that ordinary people can still use it when something matters.
Maybe the question isn't, What is the best form of government? Maybe it's: What pain are we willing to tolerate in order to preserve the things we value most?
And before answering, it would be useful to know what bargain we're already making.
Follow the consequences. Choices and Everyday Consequences adds a five-state pilot to the Government Design Review: household costs, housing decisions, migration, and questions for the offices responsible. The observations, sources and remaining uncertainties are available together.