Government 2: Oversight

December 15, 2010, last update
 

Oversight

 

The need to guard the guardians has been a slow concept to develop in human history. Rulers had God on their side, or were gods, and it’s only in the last few centuries that there’s been even a concept of checks and balances. But, as usual, the powerful find workarounds and we’re back to the point where “you can’t fight City Hall” is considered a cliché.

In a sustainable system, that is, in a fair one, that feeling can’t exist. If it does, the system won’t be sustainable. Without effective oversight of the government, a sustainable system won’t happen.

Real oversight depends on the independence and intelligence of the overseers, with the first factor apparently much more important than the second. The financial meltdown, the false evidence for the US to start a war in Iraq, the astonishing loss of consumer control over purchased digital products: these all depend(ed) on cozy relationships among the powers-that-be. Those powerful people are, if anything, smarter than average. So the first requirement of oversight is that it has to be distributed, not concentrated among an elite. It should, furthermore, stay distributed and avoid creeping concentration from re-emerging. This is where current methods fall down. Elections happen too rarely and voters can be manipulated for the short periods necessary. Lawsuits take too much time and money to be more than a rearguard action. Our current feedback loops are far too long to work consistently.

There are a few pointers showing what will work. Although no current government has effective oversight unless officials cooperate (which suggests the system relies more on the hope of honesty than a requirement for it) many governments do employ elements of oversight. Some things clearly work better than others.

First is that transparency prevents the worst behavior, or at least enables citizens to mobilize against it. Full scale looting of public money and dreadful decisions taken for self-serving ends require secrecy.

Second is that elections do cause politicians to sit up and take notice. Unfortunately, when the elections occur on a schedule, the politicians only pay attention on that schedule. One essential change is a framework to call elections — in the system described here that would be recall elections — at any time.

Third, and now we’re already in largely untested territory, is the need for finer-grained feedback. It’s essential to have ways of communicating dissatisfaction and implementing the necessary corrections which are both effortless and scalable. An escalating system of complaints might be a solution.

Fourth, in the case of plain old criminal wrongdoing, there need to be criminal proceedings, as there officially are now.

Fifth is the need for oversight to have its own checks and balances. Government is a tensegrity structure, not a pyramid. Officials need to have ways of responding to complaints. And the system of oversight itself must be overseen.

Who does the overseeing is as important as how. Distributed — in other words, citizen — oversight is the essential part, but citizens have other things to do besides track officials. There also need to be professional auditors whose task is the day-to-day minutia of oversight. They keep tabs on all the data provided by transparency, take appropriate actions as needed, and alert the public to major developing problems that are overlooked. This is another area where professionals are needed for the boring or technical details and the general public is needed to ensure independence and functional feedback.

So, starting with transparency, the first thing to note is what it isn’t. It is not a data dump. It is not an exercise in hiding some data under a mountain of other data. It does not mean keeping every record for the sake of theater. It doesn’t mean stifling interactions by perching the Recording Angel on everyone’s shoulders.

Transparency means the optimum data for the purpose of promoting honesty in government. Information not relevant to that purpose is just noise that masks the signal. However, as always when an optimum is required, the difficulty is finding the right balance. If the right to know is defined too narrowly, there’s the risk of oversight failure. If it’s defined too broadly, official interactions may be stifled for nothing or privacy may be recklessly invaded. Organizations such as Transparency International have considerable data on how to approach the optimal level. The more governments that have a commitment to providing clear and honest data, the more we’d learn about how to achieve the most useful transparency with the least intrusion and effort.

What we know so far is that financial data and records of contacts and meetings provide two of the clearest windows on what officials (also non-governmental ones) are up to. Financial data in the hands of amateurs, however, can be hard to understand, or boring, or vulnerable to attitudes that have nothing to do with the matter at hand. For instance, accountants have the concept that tracking amounts smaller than what it costs to track them are not worth following. Amateurs, on the other hand, will notice a trail of gratuitous donuts but feel too bored to follow statements of current income as present value of net future cash flows. The fact that the statement is Enron’s means nothing beforehand. It is important to ignore the easy distractions of small stuff and to notice the big stuff, especially when it’s so large it seems to be nothing but the background.

There were plenty of auditors involved in the Enron fiasco who knew better, of course. That wasn’t a simple matter of amateurs distracted by ignorance. However, the auditors received vast sums of money from the company, which simply underlines the need for true independence in oversight, including independence from membership in similar networks. Real transparency would have made the necessary information available to anyone, and would have prevented both the disaster and the pre-disaster profits, some of which flowed to the very people supposedly in oversight … which is why oversight wasn’t applied in time.

Records of meetings and contacts are another important source of data about what officials are doing, but that raises other problematic issues. People need to feel able to speak candidly. After all, the whole point is to promote honesty, not to generate ever newer layers of speaking in code. A soft focus is needed initially, followed by full disclosure at some point. Maybe that could be achieved by a combination of contemporary summaries of salient points and some years’ delay on the full record. The interval should be short enough to ensure that if it turns out the truth was shaded in the summaries, corrective action and eventual retribution will still be relevant. The interval would probably need to vary for different types of activity — very short for the Office of Elections, longer perhaps for sensitive diplomatic negotiations between countries.

Presentation of information is a very important aspect of transparency, although often overlooked. This may be due to the difficulty of getting any accountability at all from current governments. We’re so grateful to get anything, we don’t protest the indigestible lumps. Governments that actually served the people would present information in a way that allows easy comprehension of the overall picture. Data provided to the public need to be organized, easily searchable, and easy to use for rank amateurs as well as providing deeper layers with denser data for professionals. If these ideas were being applied in a society without many computers, public libraries and educational institutions would be repositories for copies of the materials.

A related point is that useful transparency has a great deal to do with plain old simplicity. The recordkeeping burden on officials should be made as simple and automatic as possible. In an electronic world, for instance, any financial transactions could be routed to the audit arm of government automatically and receive a first pass computer audit. Some simplicity should flow naturally from requiring only information relevant to doing the job honestly, rather than all information.

Information relevant to other purposes, such as satisfying curiosity, has nothing to do with transparency. Or, to put it another way, the public’s right to know has limits. An official’s sex life, for instance, unless it affects his or her job, is not something that needs to be detailed for the sake of transparency, no matter how much fun it is. Officials, like everyone else, do have a right to privacy on any matter that is not public business.

The professional audit arm of the government would have the job of examining all departments. The General Accounting Office fulfills some of that function in the US now, but it serves Congress primarily. In the system I’m thinking about, it would serve the citizenry with publication of data and its own summaries and reviews. It would also have the power to initiate recalls or criminal proceedings, as needed. (More on enforcement in a moment.)

A professional audit agency doesn’t provide enough oversight, by itself, because of people’s tendency to grow too comfortable with colleagues. Citizens’ ability to take action would counteract the professionals’ desire to avoid friction and ignore problems. But citizen oversight, by itself, is also insufficient. Its main function is as a backstop in case the professionals aren’t doing their jobs. The day to day aspects, the “boring” aspects of oversight need to be in the hands of people paid to pay attention.

Feedback is the next big component of oversight, after transparency. The sternest form of feedback is recall elections, but in the interests of preventing problems while they are still small, a system that aims for stability should have many and very easy ways to provide feedback.

The easiest way to give feedback is to complain. That route should be developed into an effective instrument regarding any aspect of government, whether it’s specific people, paperwork, or practices. Complaints could address any annoyance, whether it’s a form asking for redundant information or it’s the head of the global Department of Transportation not doing his or her job coordinating intercontinental flights. Anyone who worked for a tax-funded salary, from janitors to judges, could be the subject of a complaint.

It’s simple enough to set up a complaints box, whether physical or electronic, but effectiveness depends on three more factors. The first two are responsibility and an absence of repercussions. People have to be able to complain without fear of retribution. On the other hand, complaints have to be genuine in all senses of the word. They have to come from a specific individual, refer to a specific situation, and be the only reference by that person to that problem. Anonymity is a good way to prevent fear of retribution. Responsibility is necessary to make sure complaints are genuine. Somehow, those two conflicting factors must be balanced at an optimum that allows the most genuine complaints to get through. Possibly a two-tiered system could offer a solution: one tier would be anonymous with some safeguards against frivolous action, and one not anonymous (but with bulletproof confidentiality) that could be more heavily weighted.

The third factor is that the complaints must actually be acted upon. That means sufficient staffing and funding at the audit office to process them and enforce changes. Since effective feedback is crucial to good government, staffing of that office is no more optional than an adequate police force or court system. It needs to be very near the front of the government funding line to get the money determined by a budget office as its due. (The budget office itself would have its funding level determined by another entity, of course.)

It’s to be expected that the thankless jobs of officials will generate some background level of complaints. Complaints below that level, which can be estimated from studies of client satisfaction across comparable institutions, wouldn’t lead to action by others, such as officials in the audit agency. An intelligent official would note any patterns in those complaints and make adjustments as needed. If they felt the complaints were off the mark, they might even file a note to that effect with the audit agency. The fact that they were paying attention would count in their favor if the level of complaints rose. But until they rose, the complaints wouldn’t receive outside action.

In setting that baseline level, the idea would be to err on the low side, since expressed complaints are likely indicative of a larger number of people who were bothered but said nothing. The level also should not be the same for all grades. People are likelier to complain about low level workers than managers whom they don’t see.

Once complaints rise above what might be called the standard background, action should be automatically triggered. Some form of automatic trigger is an important safeguard. One of the weakest points in all feedback systems is that those with the power to ignore them, do so. That weakness needs to be prevented before it can happen. The actions triggered could start with a public warning, thus alerting watchdog groups to the problem as well. The next step could be a deadline for investigation by the audit office (or the auditors of the audit office, when needed). If the volume of complaints was enormous, then there’s a systemic failure somewhere. Therefore, as a last resort, some very high number of genuine complaints should automatically trigger dismissal, unelection, or repeal of a procedure. Transparency means that the volume of complaints and their subject is known to everyone, so when that point was reached would not be a secret.

I know that in the current environment of protected officialdom, the feedback system outlined sounds draconian. The fear might be that nobody would survive. That’s not the intent. If that’s the effect, then some other, more well-tuned system of feedback is needed. But whatever it’s form, an effective system that’s actually capable of changing official behavior and altering policies is essential. In a system that worked, there would be few complaints, so they’d rarely reach the point of triggering automatic action. In the unattainable ideal, the selection process for officials would be so good that the complaints office would have nothing to do.

I’ve already touched on recall elections, but I’ll repeat briefly. The recall process can be initiated either by some form of populist action, such as a petition, or by the government’s audit arm. After a successful recall, an official should be ineligible for public service for some period of time long enough to demonstrate that they’ve learned from their mistakes. They’d be starting at the lowest rung, since the assumption has to be that they need to prove themselves again.

The fourth aspect of oversight, punishment for crimes, corruption, or gross mismanagement by officials, is currently handled correctly in theory, but it needs much stricter application in practice. The greater the power of the people involved, the more it’s become customary, at least in the U.S., to let bygones be bygones after a mere resignation. It’s as if one could rob a grocery store and avoid prosecution by saying, “Oh, I’m sorry. I didn’t mean it.” Worse yet, the response to criminal officials is actually less adequate than it would be to treat shoplifters that way. The state sets the tone for all of society, so crimes committed by representatives of the state do more damage. So there must be real penalties for any kind of malfeasance in office. If there has been financial damage, it should be reimbursed first out of the official’s assets. Nor should there be any statute of limitations on negligence or problems generated by officials. They are hired and paid the big money to pay attention. If they don’t, they must not escape the consequences merely because people are so used to letting government officials get away with everything.

So far, my focus has been on ways of controlling officials, and that is the larger side of the problem. But, that said, officials do need ways of contesting overactive oversight, downright harassment, or unjustified recalls or prosecutions. The point of draconian oversight — draconian from our perspective that teflon officials are normal — is to have public servants who live up to their name. They must serve the public.

On the other hand, they’re an odd sort of servant because a significant part of their job is doing things for the public good which the public usually doesn’t like very much at the time. (If that weren’t true, there’d be no need for officials to do them. They’d get done without help.) So, there are two reasons why officials must be protected from capricious oversight. One is that servants have rights. Two is that they have to be able to take necessary unpopular steps without fear of reprisals. The public, after all, is as capable of abusing power as the individuals it’s composed of.

Complaints are the earliest point where an official could disagree with the public’s assessment. If the complaints are caused by a necessary but unpopular policy, the official needs to make clear why it’s necessary, and why something easier won’t do. Complaints happen over time, so if they’re a rising tide, that’s not something that can come as a surprise to the responsible party. The official’s justifications would become part of the record. When complaints reach a high enough level to be grounds for unelection or dismissal, and if the official is convinced that a decisive proportion of the complaints is unwarranted, they could ask for independent review. (An example of possible specifics is given below.)

The official and the complainant(s) could agree on one or more audit experts, depending on complexity, to review the case. If that didn’t lead to a resolution, the next step would be the legal process, with appeals if needed. Only decisions that went against an official should count as bad marks on their record. Simply feeling the need to defend oneself is not.

However, after repeated cases that go against a complainant, those individuals should no longer have standing to bring future cases of that type. A case that went against an official would result in recall or firing, so repeated cases wouldn’t arise.

The audit arm of the government is another source of oversight that an official could disagree with. The ability to respond to its charges would be built into the system. If the responses passed muster, the case would be closed. If not, there could be appeals to a review board, and eventually the legal system. As in the example box, I’d see this as a process with limits.

Recalls are the most far-reaching form of oversight, and the ability to contest them the most dangerous to the health of the system. Contesting a recall should be a last resort and something that’s applied only in exceptional cases. Normally, if an official felt a recall was unjustified, they should make that case during the unelection and leave the decision to voters. However, for the exceptional case, officials could contest a recall if they had a clear, fact-based justification for why it had no valid basis. It would then go to reviewers or a legal panel with the necessary specialized expertise, chosen by the opposing sides as discussed earlier.

The biggest difference between those reviewers and others is that potential conflicts of interest would have to be minutely scrutinized. More nebulous loyalties should be considered as well in this case. The tendency to excuse the behavior of colleagues is very strong and it’s one of the biggest problems with oversight of government by government. As with any process involving a balance of rights, ensuring the independence of the arbiters is essential.

To provide an additional dose of independence, a second review track could be composed of a more diffuse pool of people with less but still adequate expertise, along the same lines as selectors who determine the eligibility of candidates for office. A matter as serious as contesting a recall should use both tracks at once, and be resolved in favor of the official only if both agreed.

As I’ve done before, I’ve provided some specific ideas about implementation not necessarily because those are good ways to do it, but mainly to illustrate my meaning. Whatever method is applied, the goal is to ensure the right of public servants to defend themselves against injustice, just like anyone else, and to prevent that from becoming an injustice itself.

Because of the potentially far-reaching consequences of contesting recalls, overuse absolutely must be prevented. If it was the considered opinion of most reviewers involved that an official who lost a bid to prevent a recall did not have grounds to contest in the first place, then that attempt should be considered frivolous. One such mistake seems like an ample allowance.

A structural weakness of oversight by the public is that officials have training, expertise, and familiarity with the system. That gives them the advantage in a contest with a diffuse group of citizens. An official’s ingroup advantage may be smaller against an arm of the government tasked with oversight, but that’s not the issue. The important point is that citizen action is the oversight of last resort so it sets the boundary conditions which must not fail. Since officials have the upper hand in the last-resort situation, it is right to err on the side of protecting oversight against the official’s ability to contest it. Officials must have that option, but it can’t be allowed to develop into something that stifles oversight. It’s another balance between conflicting rights. It may be the most critical balance to the sustainability of a fair system, since that can’t survive without honest and effective government.

Regulation

 

The government’s regulatory functions flow directly from its monopoly on force. What those functions should be depend on the government’s purpose which, in the case being discussed here, is to ensure fairness. People will always strive to gain advantage, and the regulatory aspects of government must channel that into fair competition and away from immunity to the rules that apply to others.

In an ideal world, the regulatory function would scarcely be felt. The rules would be so well tuned that their effects would be self-regulating, just as a heater thermostat is a set-it-and-forget-it device. More complex systems than furnaces aren’t likely to achieve this ideal any time soon, but I mention it to show what I see as the goal of regulation. Like the rest of government, when it’s working properly, one would hardly know it was there.

However, it is regulation that can achieve this zen-like state, not deregulation. This is true even at the simplest levels. A furnace with no thermostat cannot provide a comfortable temperature. In more complex human systems, deregulation is advocated by those powerful enough to take advantage of weaker actors. (Or those who’d like to think they’re that powerful.) It can be an even bigger tool for exploitation than bad regulations. The vital point is that results expose motives regardless of rationalizations. If regulation, or deregulation, promotes concentrations of power or takes control away from individuals, then it’s being used as a tool for special interests and not in the service of a level playing field.

Current discussions of regulations tend to center on environmental or protectionist issues, but that’s because those are rather new, at least as global phenomena. In other respects, regulations are so familiar and obviously essential that most people don’t think of them as regulations at all. Remembering some of them points up how useful this function of government is. Weights and measures have been regulated by governments for thousands of years, and that has always been critical for commerce. Coins and money are a special instance of uniform measures — of value in this case — and even more obviously essential to commerce. The validity of any measure would be lost if it could be diddled to favor an interested party. Impartiality is essential for the wealth-producing effects to operate. That is so clear to everyone at this point that cheating on weights and measures is considered pathetic as well as criminal. And yet, the equally great value of fairness-preserving regulation in other areas generally needs to be explained.

Maintaining a level field is a primary government function, so undue market power is one area that needs attention. The topic overlaps with the discussion about competition and monopolies under Capital in the Money and Work chapter, but it is also an important example of regulation that can be applied only by a government, so I’ll discuss it briefly here as well.

Logically, monopolies shouldn’t exist because there ought to be a natural counterweight. They’re famous for introducing pricing and management inefficiencies that ought to leave plenty of room for competitors. But, people being what they are, market power is used to choke off any competition which could reduce it. The natural counterweight can’t operate, and only forces outside the system, such as government regulations, can halt that process.

One point I’ve tried to make repeatedly is that fairness is not only a nice principle. Its practical effects are to provide stability and wealth. That is no less true in the case of preventing unbalanced market power. Imbalance leads to chaotic consequences during the inevitable readjustment. Having enough market power to tilt the field is not only unfair and not only expensive. It’s also unstable, with all the social disruption that implies. Recognition of that fact has led to the establishment of antitrust laws, but they tend to be applied long after economic power is already concentrated. They assume that the problem is one hundred percent control.

The real evidence of an imbalance is price-setting power. When a business or an industry can charge wildly in excess of their costs, that’s evidence of a market failure and the need for some form of regulation. I’ll call entities who are able to dictate prices “monopolies” for the purposes of discussion, even though they have only a controlling market share, not necessarily one hundred percent of it.

Some monopolies are well understood by now. The classical ones involve massive costs for plants or equipment and very small costs to serve additional customers. Railroads are one example. The initial costs are high enough to prevent new startups from providing competition. Equally classic are natural monopolies where one supplier can do a better job than multiple competing ones. Utilities such as water and power are the usual examples. There’d be no money saved if householders had water mains from three separate companies who competed to supply cheap water. The duplicated infrastructure costs would swamp any theoretical savings.

So far, so good. The need for regulated utilities is well understood. Regulated transport is a bit iffier. The need for a state to handle the roads is accepted, even among the free market religionists in the US, but the need for state regulation and coordination of transport in general is less widely recognized. The free marketeers here can be puzzled about why it takes a week to move things from Coast A to Coast B using the crazy uncoordinated patchwork of private railroads. But, on the whole, the need to enforce limits on pricing power, the need for standards and for coordination is known, if not always well implemented.

The same need for standards, coordination and limits on pricing holds with all other monopolies, too, but this isn’t always grasped, especially in newer technologies. Money is not the only startup cost that can form a barrier to entry. An information industry is a recent phenomenon that started with the printing press and has begun exponential growth after the computer. The main “capital cost” is not equipment but the willingness of people to learn new habits.

Imagine, for instance, that Gutenberg had a lock on printing from left to right. All later presses would have had to convince people to read right to left just so they could get the same book but not printed by Gutenberg. Good luck with that. Similarly with the qwerty keyboard. People learned to use it, it turned out to be a bad layout from the standpoint of fingers, and we’re still using it over a hundred years later. Only one cohort of people would have to take the time to learn a new layout, and that just doesn’t happen. If there had been a modern DMCA-style copyright on that layout, there’d still be nobody making keyboards but Remington.

History shows that people simply won’t repeat a learning curve without huge inducements. And so it also has dozens of examples of companies which had the advantage of being over the learning curve and thus had a natural monopoly as solid as owning the only gold mine. Any monopoly needs regulation. If it works best as a monopoly, then it has to become a regulated utility. If it’s important enough, it may need to be government run to ensure adequate transparency and responsiveness to the users. If it doesn’t need to be a monopoly, then it’s up to regulation to ensure that the learning curve is not a barrier to entry.

Regulation that was doing its job would make sure that no single company could control, for instance, a software interface. As discussed in the chapter on Creativity, an inventor of a quantum improvement in usability should get the benefit of that invention just as with any other. But it cannot become a platform for gouging. Compulsory licensing would be the obvious solution.

Another new technology that’s rapidly getting out of hand is search. Methods of information access aren’t just nice things to have. Without them, the information might as well not be there. A shredded book contains the same information as the regular variety; there’s just no way to get at it. Organizing access to information is the function of nerves and brain. Without any way to make sense of sight or sound or to remember it, an individual would be brainless. Nor would having two or three brains improve matters. If they’re separate, any given bit of information could be in the “wrong” brain unless they all worked together. They have to work as a unit to be useful. Information access is a natural monopoly if there ever was one.

The fact that search is a natural monopoly seems to have escaped too many people. Like a nervous system, without it all the information in cyberspace might just as well not be there. The usual answer is to pooh-pooh the problem, and to say anyone can come up with their own search engine. But having multiple indexers makes as much sense as having multiple water mains. It’s a duplication of effort that actually reduces value at the end.

That’s not to say that different methods can’t exist side by side. Libraries, for instance, organize information in a fundamentally different, subject-oriented way than do search engines, and that facilitates deeper approaches to the material. Left to itself, over time, the need for that approach would lead to the development of the more laborious subject-based indexes in parallel with tag-based ones. But that’s another danger with having a monopoly in the latter kind. Undue influence might not leave any parallel system to itself. To put it as a very loose analogy, word-based methods could become so dominant that mathematics hardly existed. Whole sets of essential mental tools to handle information could fall into disuse.

On the user’s side, there’s a further factor that makes search a natural monopoly. Users don’t change their habits if they can help it. (That’s why there are such heated battles among businesses over getting their icons on computer desktops. If it’s what people are used to and it’s right in front of them, the controlling market share is safe.) Instead of pretending there is no monopoly because people have the physical option to make an initially more effortful choice — something people in the aggregate have shown repeatedly they won’t do — the sustainable solution comes from facing facts. More than one indexer is wasteful, and people won’t use more than one in sufficient numbers to avoid a controlling market share. It’s a natural monopoly. So it needs to be regulated as a public utility or even be part of the government itself.

There are further considerations specifically related to search. Any inequality in access to knowledge leads to an imbalance in a wide array of benefits and disadvantages. Further, an indexer is the gateway to stored knowledge and therefore potentially exercises control over what people can learn. In other words, a search engine can suppress or promote ideas.

Consider just one small example of the disappearance at work. I wanted to remove intercalated video ads from clips on Google’s subsidiary, Youtube. But a Google search on the topic brought up nothing useful in the first three screens’ worth of results. That despite the fact that there’s a Greasemonkey script to do the job [Update, 2014-05-09, the script is no longer there]. Searches for scripts that do other things do bring up Greasemonkey on Google.

Whether or not an adblocker is disappeared down a memory hole may not be important. But the capability to make knowledge disappear is vastly important. Did Google “disappear” it on purpose? Maybe not. Maybe yes. Without access to Google’s inner workings, it’s impossible to know, and that’s the problem.

We’re in the position of not realizing what exists and what doesn’t. Something like that is far more pernicious than straight suppression. That kind of power simply cannot rest with an entity that doesn’t have to answer for its actions. Anything like that has to be under government regulation that enforces full transparency and equality of access.

(None of that even touches on the privacy concerns related to search tools, which, in private hands and without thorough legal safeguards, is another locus of power that can be abused.)

There are two points to this digression into the monopolies of the information age. One is that they should and can be prevented. They have the same anti-competitive, anti-innovative, expensive, and ultimately destabilizing effect of all monopolies. And a good regulator would be breaking them up or controlling them for the public good before they grew so powerful that nothing short of social destruction would ever dislodge them.

My second point is that monopolies may rest on other things besides capital or control of natural resources. All imbalances of market power, whether they’re due to scarce resources, scarce time, ingrained habits, or any other cause need to be recognized for what they are and regulated appropriately.

Regulations as the term is commonly used — environmental, economic, financial, health, and safety — are generally viewed as the government guaranteeing the safety, not the rights, of citizens to the extent possible. But that leads straight into a tangle of competing interests. How safe is safe? Who should be safe? Just citizens? What about foreigners who may get as much or more of the downstream consequences? How high a price are people willing to pay for how much safety?

If the justification is safety, there’s no sense that low-priced goods from countries with lax labor or environmental laws are a problem. Tariffs aren’t applied to bring them into line with sustainable practices. The regulatory climate for someone else’s economy is not seen as a safety issue. Egregious cases of slavery or child labor are frowned on, but are treated as purely ethical issues. Since immediate physical harm to consumers is not seen as part of the package, regulation takes a back seat to other concerns. Until it turns out that the pollution or CO2 or contaminants do cause a clear and present danger, and then people are angry that the situation was allowed to get so far out of hand that they were harmed.

Approaching regulations as safety issues starts at the wrong end. Sometimes safety can be defined by science, but often there are plenty of gray areas. Then regulations seem like a matter of competing interests. Safety can only be an accidental outcome of that competition unless the people at the likely receiving end of the harm have the most social power.

However, when the goal is fairness, a consistent regulatory framework is much easier to achieve. The greatest good of the greatest number, the least significant overall loss of rights, and the least harm to the greatest number is not that hard to determine in any specific instance. In a transparent system with recall elections, officials who made their decisions on a narrower basis would find themselves out of a job. And safety is a direct result, not an accidental byproduct, of seeking the least harm to the greatest number.

Consider an example. The cotton farmers of Burkina Faso cannot afford mechanization, so their crop cannot compete with US cotton. Is that a safety issue? Of course not. Cotton is their single most important export crop. If the farmers switch to opium, is that a safety issue? Sort of. If the local economy collapses, and the country becomes a haven for terrorists, that’s easy to see as a safety issue. A much worse threat, however, would be the uncontrolled spread of disease. Collapse is not just some academic talking point. The country is near the bottom of poverty lists, and who knows which crisis will be the proverbial last straw. So, once again, is bankrupting their main agricultural export a safety issue? Not right now. So nothing is done.

Is it, however, a fairness issue? Well, obviously. People should not be deprived of a livelihood because they are too poor to buy machines. There are simple things the US could do. Provide regulation so that US cotton was not causing pollution and mining land and water. The remaining difference in price could be mitigated with price supports to Burkinabe cotton. The only loss is to short term profits of US cotton farmers, and where that is a livelihood issue for smallholders assistance could be provided during the readjustment.

There may be better ways of improving justice in the situation, but my point is that even an amateur like me can see some solutions. The consequences of being fair actually benefit the US at least as much as Burkina Faso. (Soaking land in herbicide is not a Good Thing for anyone.) And twenty or thirty years down the road, people both in the US and in Burkina Faso will be safer. The things they’ll be safer from may be different — pollution in one case, social disruption in the other — but that doesn’t change the fact that safety oriented regulation has no grounds for action, but fairness oriented regulation can prevent problems.

Globalization is yet another example where regulation based on fairness facilitates sustainability. Economists like to point out that globalization can have economic benefits, and that’s true, all other things being equal. The problem is other things aren’t equal, and globalization exposes the complications created by double standards. If it applied equally to all, labor would have to be as free to move to better conditions as capital. (See this post for a longer discussion.) Given equal application of the laws, globalization couldn’t be used as a way of avoiding regulations. Tariffs would have to factor in the cost of sustainable production, and then be spent in a way that mitigated the lack of sustainability or fairness in the country (or countries) of origin. There’d be less reward for bad behavior and hence fewer expensive consequences, whether those are contaminants causing disease or social dislocation caused by low wages and lost work. Those things are only free to the companies causing the problem.

Enforcement must be just as good as the regulations or else they’re not worth the paper they’re printed on. The financial meltdown is a stark recent example showing that good regulations are not enough. Maintaining the motivation for enforcement is at least as important. In the US and in many other countries, the regulations existed to preserve the finances of the many rather than the narrow profit potential of the few. In hindsight, the element missing was any desire to enforce those regulations. (E.g. see Krugman’s blog and links therein.)

Enforcement is necessarily a task performed by officials, who are hired precisely to know enough to use foresight so that hindsight is unnecessary. That’s understood everywhere, and yet enforcement is repeatedly the weak link in the chain. That’s because the responsible officials are not really responsible. Rarely do any of them have any personal consequences for falling down on the job. That is the part that has to change.

Officials who enable catastrophe by not doing their jobs commit dereliction of duty, something that has consequences in a system emphasizing feedback and accountability. There has to be personal liability for that. People are quick to understand something when their living depends on it. (With apologies to Upton Sinclair.) The right level of personal consequences for responsible officials is not only fair, it would also provide the motivation to do the job right to begin with and avoid the problem.

For personal responsibility to be justified, the right path to take does need to be well known. I’m not suggesting officials be punished for lacking clairvoyance. I’m saying that when the conclusions of experts exceed that 90% agreement level mentioned earlier, and yet the official ignores the evidence, then they’re criminally culpable. Then a lack of action in the face of that consensus needs to be on the books as a crime, punishable by loss of assets and even jail time, depending on the magnitude of the lapse. Nor should there be any statute of limitations on the crime, because there is none on its effects. The penalties need to be heavy enough to scare officials straight, but not so heavy they’re afraid to do their jobs. Finding the right balance is another area where study is needed to find optimal answers. But regulation is a vital function of government, so it is equally vital to make sure that it’s carried out effectively. Dereliction of duty is no less a crime of state than abuse of power.

Officials who are actually motivated to do their jobs are important in a system which works for all instead of for a few. The point of fair regulation, as of fairness generally, is that avoiding problems is less traumatic than recovering from them. Prevention of problems is an essential function of regulation. Prevention takes effort ahead of time, though, when there’s still the option to be lazy, so it only happens when there’s pressure on the relevant officials. That means there have to be penalties if they don’t take necessary steps. It’s unreasonable to punish a lack of clairvoyance, and yet it’s necessary to punish officials who let a bad situation get worse because it was easier to do nothing.

In short, the government’s regulatory functions should serve to maintain a fair, sustainable environment where double standards can’t develop. Safety or low prices might be a side effect, but the purpose has to be enforcing rights.

Public Works

 

This is one of the few areas where the highest expression of the art of government is more than invisibility. Some of the greatest achievements of humankind come from the type of concerted action only governments can coordinate. The great pyramids of Egypt, medieval European cathedrals, and space flight were all enterprises so huge and so profitless, at least in the beginning, that they couldn’t have been carried out by any smaller entity.

People could — and have — questioned whether those projects were the best use of resources, or whether some of them were worth the awful price in suffering. The answer to the latter is, obviously, no. Where citizens have a voice, they’re not likely to agree to poverty for the sake of a big idea, so that problem won’t (?) arise. But that does leave the other questions: What is worth doing? And who decides?

There’s an irreducible minimum of things worth doing, without which a government is generally understood to be useless. The common example is defense, but on a daily level other functions essential to survival are more important. A police force and court system to resolve disputes is one such function. Commerce requires a safe and usable transportation network and reliable monetary system. In the system described here, running the selection and recall processes would be essential functions. And in a egalitarian system, both basic medical care and basic education have to be rights. It is fundamentally unfair to put a price on life and health, and equally unfair to use poverty to deprive people of the education needed to make a living.

As discussed in the chapter on Care, the definition of “basic” medical care depends on the wealth of the country. (Or of the planet, in a unified world.) In the current rich countries, such as the G20, it simply means medical care. It means everything except beauty treatments. And even those enter a gray area when deviations from normal are severe enough to cause significant social problems. On the other hand, in a country where the majority is living on a dollar a day and tax receipts are minimal, basic medical care may not be able to cover much except programs that cost a tiny fraction of the money they eventually save. That’s things like vaccinations, maternal and neonatal care, simple emergency care, pain relief, treatments against parasites, and prevention programs like providing mosquito netting, clean water, and vitamin A.

The definition of basic education likewise depends on the social context. In a pre-technological agrarian society it might be no more than six grades of school. In an industrial society, it has to include university for those who have the desire and the capacity to make use of it. The principle is that the level of education provided has to equal the level required to be eligible for good jobs, not just bottom tier jobs, in that society.

I’ve gone rather quickly from the irreducible minimum to a level of service that some in the US would consider absurdly high. Who decides what’s appropriate? I see that question as having an objective answer. We are, after all, talking about things that cost money. How much they cost and how much a government receives in taxes are pretty well-known quantities. From that it’s possible to deduce which level of taxes calls for which level of service.

The concept that a government can be held to account for a given level of service, depending on what proportion of GDP it uses in taxes, is not a common one. And the application of the concept is inevitably empirical. It is nonetheless interesting to think about the question in those terms.

As I’ve pointed out, there are several governments who have total tax receipts of about a third of GDP and provide all the basic services plus full medical, educational, retirement, and cultural benefits, as well as providing funds for research. Hence, if a country has somewhere between 30% to 40% of GDP in total tax receipts, it should be providing that level of service. If it isn’t, the citizenry is being cheated somewhere.

On a local note, yes, this does mean the US is not doing well. Total taxes here are around 28%. Japan provides all the services and benefits listed above while taking in 27% of total GDP. The low US level of services implies taxes should be lower, or services should be higher. The US does, of course, maintain a huge military which, among other things, redistributes federal tax dollars and provides some services such as education and medical care to its members. If one factors in the private cost of the tax-funded services elsewhere, US total tax rates would be over 50%. Given the gap between the US and other rich countries, the US is wasting money somewhere … .

On the other hand a poor country, which can’t fairly have much of a tax rate, might have only 5% of an already low GDP to work with. Services then would be limited to the minimums discussed earlier. Even poor governments can provide roads, police, fire protection, commercial regulation to make trade simpler, basic medicine and education. (It should go without saying, but may not, that a poor country may not have the resources to deal with disasters, and may need outside help in those cases.) The only thing that prevents some of the world’s countries from providing the basics now are corruption, incompetence, and/or the insistence of elites on burning up all available money in wars. None of those are essential or unavoidable. Fairness isn’t ruinously expensive, but its opposite is.

At any level of wealth, the receipts from fair levels of taxes dictate given levels of service. Citizens who expect more without paying for it are being unfair, and governments who provide less are no better.

There is now no widespread concept that a government is obligated to provide a given level of service for a given amount of money. Even less is there a concept that taking too much money for a given level of service is a type of crime of state. The government’s monopoly on force in a might-makes-right world means that there aren’t even theoretical limits on stealing by the state. However, when it’s the other way around and right makes might, then those limits flow logically from what is considered a fair tax level.

So far, I’ve been discussing taxes and what they buy on a per-country basis which feels artificial in a modern world where money and information flow without borders, and where we all breathe the same air. A view that considers all people equal necessarily should take a global view of what’s affordable. I haven’t discussed it that way because it seems easier to envision in the familiar categories of nations as we have them now, but there’s nothing in the principles of fair taxation and corresponding services that wouldn’t scale up to a global level.

Finally, there’s the most interesting question. In a country — or a world like ours — which is doing much better than scraping by, what awe-inspiring projects should the government undertake? How much money should be devoted to that? Who decides what’s interesting? Who decides what to spend?

Personally, I’m of the opinion that dreams are essential, and so is following them to the extent possible. I subscribe to Muslihuddin Sadi’s creed:

If of mortal goods you are bereft,
and from your slender store two loaves
alone to you are left,
sell one, and from the dole,
buy hyacinths to feed the soul.

Or, as another wise man once said, “The poor you shall always have with you.” Spend some money on the extraordinary.

So I would answer the question of whether to spend on visionary projects with a resounding “yes.” But I’m not sure whether that’s really necessary for everyone’s soul or just for mine. I’m not sure whether it’s a matter of what people need, and hence whether it should be baked into all state budgets, or whether it’s a matter of opinion that should be decided by vote. If different countries followed different paths in this regard, time would tell how critical a role this really plays.

It seems to me inescapable that countries which fund interesting projects would wind up being more innovative and interesting places. Miserly countries would then either benefit from that innovation without paying for it, which wouldn’t be fair, or become boring or backward places where fewer and fewer people wanted to live. Imbalances with unfortunate social consequences would then have to be rectified.

As for the ancillary questions of what to spend it on, and how much to spend, those are matters of opinion. Despite that, it’s not clear to me that the best way to decide them is by vote. People never would have voted to fund a space program, yet fifty years later they’ll pay any price for cellphones and they take accurate weather forecasting for granted. I and probably most people have no use for opera, and yet to others it’s one of the highest art forms.

Further, another problem is that budget instability leads to waste. Nobody would vote for waste, but voter anger can have exactly that effect. Jerking budgets around from year to year wastes astronomical sums. A good example of just how much that can cost is the tripling, quadrupling, and more of space program costs after the US Congress is done changing them every year. Voters, on the whole, are not good at the long range planning and broad perspective that large social projects require. I discuss a possible method of selection under Advanced Education. Briefly, I envision a similar system to the one suggested for other fields. A pool of proposals is vetted for validity, not value, and selection is random from among those proposals that make methodological sense. However it’s done, decisions on innovation and creativity have to be informed by innovative and creative visionaries without simultaneously allowing corrupt fools to rush in.

 
– + –

A fair government, like any other, has a monopoly on force, but that is not the source of its strength. The equality of all its citizens, the explicit avoidance of double standards in any sphere, the renunciation of force except to prevent abuse by force, short and direct feedback loops to keep all participants honest, and all of these things supported in a framework of laws — those are the source of its strength. The strength comes from respecting rights, and is destroyed in the moment rights get less respect than might. The biggest obstacle to reaching a true rule of law is that intuition has to be trained to understand that strength is not muscle.

 


So far, I can't make comments appear on archive pages. Please click on the post title or permalink to go to the post page, if you'd like to comment.

Government 1: War and Politics

December 12, 2010, last update

Introduction

 

There’s irony to the fact that in a work on government, the topic itself waits to appear till somewhere in the middle. But government is a dependent variable without any chance of being functional, to say nothing of fair, if the physical and biological preconditions aren’t met. Those preconditions can be satisfied with less effort when people are savvy enough to understand where they fit in the scheme of things, but they can’t be ignored.

It’s customary, or it is now at any rate, to think of government in terms of its three branches, executive, legislative and judicial. But those are really branchlets all in one part of the tree and don’t include important aspects of the whole thing. They are all concerned with the machinery, with how to get things done, and they don’t address the why, what, or who of government.

At its root, the function of government is to coordinate essential common action for the common good. When resources are at a minimum, the function is limited to defense, which is one type of common action. In richer situations, other functions are taken on, generally at the whim of the rulers and without a coherent framework for what actually serves the common good. The priorities are less haphazard in democracies, but still don’t necessarily serve the common good. Wars of aggression come to mind as an example.

The defining characteristic of a government is said to be a monopoly on the legitimate use of force. (Parents are allowed to use force on children in all societies so far — an exception I consider wrong as I discussed in the fourth chapter — but that’s a side issue in the context of this discussion.) That force is used not only against outside threats, but also internal ones, such as criminals. An essential function of government is to protect its citizens from each other. It’s a short step from that to regulatory functions generally, to all the ways in which a government can serve to maintain the people’s rights. Economic, environmental, and safety regulations are all included here. Preventing bad actions is one type function carried out by government for the common good.

The positive functions of government are as important: the achievement of desirable goals, not merely the prevention of disaster. Some of those coordinated actions for the common good are now understood to be just as obvious a function of government as defense. Essential infrastructure is in this category, although people don’t always agree on what is essential. Ensuring basic standards of living, basic levels of health, and security in old age are also coming to be understood as something government is uniquely qualified to do. And then there are the things which fire the imagination and are a quest and an adventure rather than a necessity. These are big or not obviously profitable projects beyond the scope of smaller entities, projects such as support for basic research, for space programs, or for the arts.

Given that government has a great deal to do, it may be worth pointing out which functions it does not have. Its purpose is not to bring happiness or wealth to its citizens, only to bring the preconditions for those desirable things. And it’s certainly not to make a subset of citizens wealthy. Nor is it to make citizens healthy, only to provide the preconditions to make it easy for people to take care of their health. That’s actually a more difficult issue than the others because if health care is government-funded, then a line has to be drawn about how much self-destructive behavior to tolerate. And that runs smack into the right to do anything that doesn’t harm others. It’s a conflict I’ll discuss in the chapter on Care, near the end of the section on medicine.

So, having discussed the preconditions on which government rests, the next step is examining the machinery by which it can implement its functions. What we want from the machinery is for it to interfere with those functions as little as possible and for it to be as effortless and invisible as possible. The business of government is essentially a boring job of solving other people’s problems — or that’s what it is when abuse of power and corruption are removed from the equation. Quite clearly, we have some distance to go before achieving a silent, effortless and effective machine. Below, I’ll offer my ideas on how to get closer to that goal.

Important aspects of the machinery side of the tree which fall outside the traditional executive, legislative, and judicial zones are funding the government and overseeing it. Oversight is a particularly vital issue because the thing being overseen also has a monopoly on force. Given the willingness in enough people to accommodate anything the powerful want to do, it’s essential to counteract that call to power and to make sure that oversight retains the ability to rein in all abuses.

Instead of discussing these topics in their conceptual order above, I’ll take them in the order of their capacity to interfere with people’s lives. That will, I hope, make it easier to imagine how it might work in fact.

Force

 
Between groups

The bloody-minded we will always have with us, so it’s to be expected that even the most advanced human societies will always need to have some sort of defense and police forces. But their function has to be limited to small-scale threats. If an entire population flouts the law, a police force is powerless. It can only stop rare actions, not common ones. Likewise, the defense force of an equitable society would have to be targeted toward stopping occasional attacks by kooks of various kinds, not fullscale wars with equals or, worse yet, more powerful countries.

The reason for that is simple. Fair societies are not compatible with war. War is the ultimate unfairness. Its authority is might, not right. Even self-defense, in other words even a just war, is not compatible with retaining a nice sense of fair play. Wars of self-defense may be forced on a country, but that doesn’t change the facts. Justice and war are simply mutually exclusive.

No, that does not mean that an equitable society is impossible. It’s only impossible if wars are inevitable, and the evidence suggests otherwise. Looking at it from a national level, how many countries have internal wars? Some, but not all. Not even most. Somehow, this so-called inevitable state of war has been eliminated even among large groups within countries. If it can be done for some groups, it can be done for all groups.

It should go without saying, but probably doesn’t, that regressions are always possible. We could regress right back to stone-throwing cave dwellers under the right conditions. That’s not the point. The point is that war has been eliminated between some groups. How that’s done, and the huge wealth effect of doing it are obvious by now. To repeat, if it can be eliminated between some groups, it can be eliminated between all groups. It is not inevitable. Aggression isn’t about to go away, but that particular form of it known as war can and has disappeared in some cases.

Wars happen internationally because on some level enough people think they’re a better idea than the alternative. They want to retain the ability to force agreements rather than come to them. But there’s a price to pay for the method used, and it only makes practical sense if the price is not greater than the prize. Wars make no sense by any objective measure on those grounds. Their only real benefit is on a chest-thumping level, and even that works only at the very beginning, before the price starts to be paid.

For anyone who doubts my statement about price, I should mention that I’m thinking both large scale and long term. In a very narrow view, ignoring most of the costs, one could claim that it’s possible to win something in a war. Take the view down a notch from international to national, and it becomes clear how flawed it is. It’s as if one were thinking on the same level as a Somali warlord. If Somalia had not had a generation with no government, and hadn’t spent the time in the grip of warring gangs, anybody on earth (except maybe a Somali warlord) would agree they’d be much richer and better off in every way. Ironically, even the warlords themselves would be better off. They wouldn’t have to spend a fortune on armed guards, for starters. And yet, unable to think of a way of life different from the hellhole they’re in now, they keep shooting in the hope of winning something in a wasteland they’re reducing to nothing.

That, on an international scale, is where our national minds are. That, on an international scale is the price we also pay. What we’re losing for the planet is the difference between Sweden and Somalia. When the vision of what we’re losing becomes obvious to enough people, and when the pain of not having it becomes big enough, that’s when we’ll decide that an international body to resolve disputes is a better idea than the idiotic notion of trial by ordeal. Then it’ll be obvious what’s the better choice, even if trial by law does mean losing the occasional case. After all, that can happen in trial by ordeal too, but with much more devastation.

So, no, I don’t think that the incompatibility between fairness and war is the end for fair societies. I think it’s ultimately the end for war.

The beginnings of that evolution seem to be happening already. War is becoming less and less of an option between some countries. Even the concept of war between Austria and Italy, or Japan and France, is just laughable. They wouldn’t do it. They’d negotiate each others’ ears off, but they’d come to some resolution without killing each other.

That’s the future. An extrapolation of the past is not.

The fact that the US is closer to the past than the future doesn’t change the trajectory. Likewise with the other large powers, China, Russia, India, Brazil. They all have further to go than anyone would like, but that only means regressions are possible. It still doesn’t change the direction of the path. And even the US seems to feel the need for UN permission before marching into other countries and blowing them up. That’s something I don’t remember seeing ever before. It’s a sign of a potential sea change, if they don’t zig backwards again. The next step would be to actually stop blowing people up. We’ll see if they zag forward and take it.

If violent methods are not an option as a final solution to disputes, then the only alternatives when ordinary channels aren’t working are nonviolent ones. That is not as utopian as the people who fancy themselves hard-headed realists assume. A tally by Karatnycky and Ackerman (2005) (pdf) of the 67 conflicts that took place from 1973 to 2002 indicates that mass nonmilitary action was likelier to overturn a government and — even more interesting — much likelier to lead to lasting and democratic governments afterward. (69% vs 8%, and 23% with some intermediate level of civic involvement.) Armed conflicts, in contrast, tended to result simply in new dictatorships, when they succeeded at all. It is easy to argue with their exact definitions of what is free and how the struggles took place. But short of coming up with an ad hoc argument for every instance, the broader pattern of the effectiveness of nonviolence in both getting results and in getting desirable results is clear.

As the article points out, some of the initially freedom-enhancing outcomes were later subverted. Zunes et al. (1999) cite one notable example in Iran:

Once in power, the Islamic regime proved to abandon its nonviolent methodology, particularly in the period after its dramatic shift to the right in the spring of 1981. However, there was clear recognition of the utilitarian advantages of nonviolent methods by the Islamic opposition while out of power which made their victory possible.

The interesting point here is not that regression is to be expected when nonviolence is used as a tool by those with no commitment to its principles. The interesting thing is that its effectiveness is sufficiently superior that it’s used by such people.

That effectiveness makes little sense intuitively, but that’s because intuition tells us we’re more afraid of being killed than having someone wave a placard at us. Fear tells us that placard-waving does nothing, but violence will get results. But that intuition is wrong. Fear isn’t actually what lends strength to an uprising. What counts is how many people are in it and how much they’re willing to lose. The larger that side is, the more brutal the repression has to be to succeed, until it passes a line where too few soldiers are able to stomach their orders.

That pattern is repeated over and over again, but each time it’s treated as a joke before it succeeds, and if it succeeds it’s assumed to be an exception. The facts become eclipsed by the inability to understand them. The hardheaded realists can go on ignoring reality.

The faulty emotional intuition on nonviolence leads to other mistakes besides incorrect assessments of the chances of future success. It results in blindness to its effects in the present and the past as well. As they say in the news business, if it bleeds, it leads. The more violence, the more “story.” We hear much more about armed struggle, even if it kills people in ones and twos, than we do about movements by thousands that kill nobody. How many people are aware of nonviolent Druze opposition to Israeli occupation in the Golan Heights versus the awareness of armed Palestinian opposition? How many even know that there is a strong nonviolent Palestinian opposition? The same emotional orientation toward threats means that nonviolence is edited out of history and flushed down the memory hole. History, at least the kind non-historians get at school, winds up being about battles. The result is to confirm our faulty intuition, to forget the real forces that caused change, and to fix the impression that violence is decisive when in fact it is merely spectacular.

Admittedly, the more brutal the starting point, the less likely nonviolence is to succeed. The people whose struggle awes me the most, the Tibetans, are up against one of the largest and most ruthless opponents on the planet. Even though their nonviolence is right, it’s no guarantee of the happy ending. But — and this is the important point — the more brutal the basic conditions, the less chance that violence can succeed. The less likely it is that anything can succeed in pulling people back from a tightening spiral of self-destruction. Places like the Congo come to mind. The more equitable a society is to start with, the better nonviolent methods will work.

“Common sense” may make it hard to see the facts of the effectiveness of nonviolence, but even common sense knows that ethical treatment is likelier to lead to ethical outcomes than brutality. The fact that armed conflict is incompatible with justice is not a disadvantage to a fair society, so long as it remembers that violence is the hallmark of illegitimacy. The state itself must view the application of violence as fundamentally illegitimate and therefore use its own monopoly on force only in highly codified ways. An analogy is the blanket prohibition against knifing other people, except in the specific set of circumstances that require surgery. I’ll discuss the distinction a bit more below. A refusal to use violence is a clear standard to apply, and it gives the state that applies it an inoculation against one whole route to subversion. When violence is understood to be a hallmark of illegitimacy, then it can’t be used as a tool to take power no matter how good the cause supposedly is, and no matter whether the abuse comes from the state or smaller groups.

Another objection runs along the lines of “What are you going to do about Hitlers in this world of yours?” “What about the communists?” Or “the Americans” if you’re a communist. “What about invasions?” What, in short, about force in the hands of bad guys? If one takes the question down a level, it’s easier to understand what’s being asked. In a neighborhood, for instance, a police force is a more effective response to criminals than citizens all individually arming themselves. The real answer to force in the hands of bad guys is to take it out of their hands in the first place. If a gang is forming, the effective response is not to wait for attack but to prevent it.

On a local level, the effectiveness of police over individual action is obvious. On an international level, it seems laughable because the thugs have power. But that’s common sense playing us false again. Brute force threats are not inevitable. They’re human actions, not natural forces. They’re symptoms of a breakdown in the larger body politic. The best way to deal with them is to avoid the breakdown to begin with, which is not as impossible as it might seem after the fact. The currently peaceful wealthy nations show that dispute resolution by international agreement is possible and they show how much there is to gain by it.

That indicates that objections about the impossibility of using international norms to handle “bad guys” are really something else. An answer that actually solves the problem is to strengthen international institutions until the thugs can’t operate, just as good policing prevents neighborhood crime. However, when people make that objection, international cooperation is the one solution they won’t entertain. That suggests they’re not looking for a solution. Otherwise they’d be willing to consider anything that works. They’re really insisting on continued thuggery, using the excuse that so-and-so did it first.

At its most basic level, the argument about the need to wage war on “bad guys” is built on the flawed logic of assuming the threat, and then insisting there’s only one way to deal with it. The Somali warlord model of international relations is not our only choice.


Force against criminals
 

Just as war is incompatible with fairness, so is any other application of force where might makes right. Therefore force is justifiable only to the extent that it serves what’s right. It’s similar to the paradox of requiring equal tolerance for everyone: the only view that can’t be tolerated is intolerance. Force can’t be used by anyone: it can only be used against someone using it.

Force can be applied only to the extent of protecting everyone’s rights, including the person causing the problem, to the extent compatible with the primary purpose of stopping them from doing violence. The state can use only as much force as is needed to stop criminal behavior and no more.

Because the state does have the monopoly on force, it is more important to guard against its abuse in state hands than any others. Crimes of state are worse than crimes of individuals. Not only do people take their tone from the powerful, but they’re also too ready to overlook bad actions by them. So the corrupting influence of abuse is much greater when done by those in power, and it is much more necessary to be vigilant against people’s willingness to ignore it. An event such as the one where an unarmed, wheelchair-bound man was tasered would be considered a national disgrace in a society that cared about rights. It would be considered a national emergency in a society that cared about rights if people saw tasering as the expected result of insufficient deference to a uniform. That fascist attitude has become surprisingly common in the U.S.

The exercise of state power carries responsibilities as well as limits. When criminals are imprisoned, the state becomes responsible for them. That’s generally recognized, but somehow the obvious implications can be missed. Not only must the state not hurt them, it has taken on the responsibility for preventing others from hurting them as well. The jailers can’t mistreat prisoners. That much is obvious. But neither can people condone the torture of prisoners by other prisoners. The dominant US view that rape in prisons is nobody’s concern is another national emergency. Proper care of prisoners is expensive. People who don’t want the expense can’t, in justice, hold prisoners.

The death penalty is another practice that cannot be defended in a just context. Its justification is supposed to be that is the ultimate deterrent, but a whole mountain of research has shown by now that, if anything, crimes punishable by death are more common in countries with death penalties than without. If the irrevocable act of killing someone serves no legitimate goal, then it can have no place in a just society.

Another problem is the final nature of the punishment. It presumes infallibility in the judicial system, which is ludicrous on the face of it. Fallibility has been shown time and again, and tallied in yet another mountain of research. Executing the innocent is something that simply cannot happen in a society with ambitions of justice.

But perhaps the most insidious effect of having a death penalty is what it says, even more than what it does. The most powerful entity is saying that killing those who defy it is a valid response. And it’s saying that killing them solves something. The state, of course, makes a major point of following due procedure before the end, but that’s not the impressive part of the lesson. People take their tone from those in power, and for those already inclined to step on others’ rights, the due process part isn’t important. What they hear is that even the highest, the mightiest, and the most respectable think that killing is all right. They think it solves something. If it’s true for them, it becomes true for everyone. No amount of lecturing that the “right” to kill works only for judges will convince criminals who want to believe otherwise. By any reasoning, ethical, practical, or psychological, the death penalty is lethal to a fair society.

Decision-making

 
Voting
[Some material also in an earlier post on Democracy Doesn’t Work]

The bad news is that democracy does not seem to work as a way of making sustainable decisions. The evidence is everywhere. The planet is demonstrably spiraling toward disaster through pollution, overpopulation, and tinpot warring dictators, and no democracy anywhere has been able to muster a proportional response. The most advanced have mustered some response, but not enough by any measure. A high failing grade is still a failure.

Given that democracy is the least-bad system out there, that is grim news indeed. But the good news is that we haven’t really tried democracy yet.

As currently practiced, it’s defined as universal suffrage. (Except where it isn’t. Saudi Arabia is not, for some reason, subject to massive boycotts for being anti-democratic even though it enforces a huge system of apartheid.) But defining democracy down to mere voting is a sign of demoralization about achieving the real thing. Democracy is a great deal more than voting. It’s supposed to be a way of expressing the will of the people, and voting is just a means to that end. For that matter, the will of the people is considered a good thing because it’s assumed to be the straightest route to the greatest good of the greatest number. There’s a mass of assumptions to unpack.

Assumption 1: the greatest good of the greatest number is a desirable goal. So far, so good. I think everybody agrees on that (but maybe only because we’re assuming it’s bound to include us).

Assumption 2: Most people will vote in their own self-interest and in the aggregate that will produce the desired result. The greatest good of the greatest number is the outcome of majority rule with no further effort needed. Assumption 2 has not been working so well for us. Possibly that has to do with the further assumption, Assumption 2a, that most people will apply enlightened self-interest. Enlightenment of any kind is in short supply.

The other problem is Assumption 2b, which is that majority rule will produce the best results for everyone, not just for the majority. There is nothing to protect the interests of minorities with whom the majority is not in sympathy.

Assumption 3: Tallying the votes will show what the majority wants. This is such a simple idea it seems more like a statement of fact, and yet reality indicates that it’s the easiest part of the process to rig. Most of the time, people don’t know what they want. Or, to be more precise, they know what they want — a peaceful, happy, interesting life for themselves and their children — but they’re not so sure how to get there. So a few ads can often be enough to tell them, and get them voting for something which may be a route to happiness, but only for the person who funded the ads. Besides the existential difficulty of figuring out what one wants, there’s also a morass of technical pitfalls. Redrawing voting districts can find a majority for almost any point of view desired. Different ways of counting the votes can hand elections to small minorities. Expressing the will of the majority, even if that is a good idea, is far from simple.

However, although democracy’s failed assumptions have become clearer, one of the more optimistic ones is unexpectedly successful. (Unexpected to me, at any rate.) Evidence shows that lots of people are surprisingly much better at guessing right than most of the individuals within the group. This has the catchy name of the “wisdom of crowds,” although wisdom is perhaps overstating the case. It implies that majority rule ought to work a lot better than it does.

I think the reason why it doesn’t lies not far to seek. The experiments demonstrating crowd smarts rely on a couple of preconditions. Everybody in the group must have much the same information. And everybody in the group has to come to their decision independently. The implications for how elections must be run are obvious and rather diametrically opposed to how they’re run now. Just for a start, it implies no campaign advertising and no opinion polling.

I’ve argued repeatedly that people won’t put effort into issues of little direct interest to them. Governing the country is one of those distant activities. It’s a background noise to their primary concerns, and the best thing it can do is be quiet. Achieving a condition of effective yet quiet government takes a lot of knowledge, attention to detail, consideration of implications, and preventive action. In short, it takes professionals. Nobody is going to do the work required in half an hour on Friday night after the barbecue. Even if they had that half hour, they’d find more fun things to do during it.

The fact that people will not spend time on what feels like boring homework means we’re applying democracy at the wrong end of the process. Voters aren’t much good at governing — anyone who lives in California needs no further proof of that. They’re not even good at finding other people who can govern — anyone who lives on this planet can see that.

It isn’t that voters aren’t smart enough. They are. The problem is that hiring is a boring job. People get paid to do that. If they’re not paid, and are not responsible for the performance of the employee, there’s no way enough people will spend time carefully considering a candidate’s resume, studying their past performance, digging up clues about what the candidate really did, and trying to form an accurate assessment of probable future performance. Voters want someone who doesn’t put them to sleep, at least long enough to do their civic duty and decide for whom to vote.

That’s one of those things everybody knows and nobody mentions unless they’re being funny or cynical. As in, for instance, the following ironic comment about a British candidate’s campaign strategy:

David Cameron is direct about how the next election is not one of ideological revolution (which would only flame existing suspicions that the British have about the extent of the modernization of the Tories) but rather who would better manage the economy and government.

Because electing the best technocrat is really inspirational to casual voters.

So voters get what they want, more or less inspirational campaigns. When it turns out that doing the job requires competence, it’s too late.

Voters are no good at electing leaders. They should be unelecting them. (I’ll discuss below alternate ways of finding officeholders.) Figuring out what’s right is hard, but noticing something wrong is easy. We shouldn’t be looking to voters to craft solutions. They’d be much better at smashing messes into small enough pieces to cart away.

Of course, nobody could govern if they had to face recall elections every morning. There needs to be a threshold. The discontent needs to be widespread enough to be more than individual dudgeon or the machinations of interested parties. Ways to measure discontent need study because there’s a fine line between capturing every free-floating anxiety out there and making the process too difficult to be effective at policing officials. Petitions are one traditional way to measure discontent. Whether they’re the best way, I don’t know. (An example of possible implementation is given below.)

If 5% of the voters in the area concerned — city, state, nation, or world — or 500 whichever was larger, signed a petition for recall, and the whole recall effort was volunteer and uncompensated, then the recall election would be held soon after, say six weeks. The concerned parties could present their arguments in voter information booklets. They could have real debates, and answer real unscripted questions from real unscripted people, but they could not advertise during that time (or any other time).

The example is just to give some indication of what I mean. However dissatisfaction is measured, the essential element is that the process must be immunized against influence either by members of the public or by the officials, which is also why the time frame involved should be short. Those wise crowds only appear in the absence of slant.

One more safeguard is needed. Consciousness of scrutiny can produce good behavior, but if the scrutiny only appears on a schedule, the same becomes true of the good behavior. Officials should be subject to random checks for which they can’t prepare. Effective methodology needs study and trial in practice, because until it’s been tried there is no fund of experience on what works. As an example of what I mean, something like random “snap unelections” might be useful. These might list those in the top five percent for complaints, even without formal petitions against them.

I am not trying to argue in favor of a particular method. Effective methods can best be found by means of longitudinal studies. The point is that the most effective methods known at the time are the ones to apply, and they should be able to make officials do their jobs in the interests of the greatest good all the time, not once every few years.

This is also the place to discuss some of the minutiae of the voting process: delimitation of districts and methods of counting votes.

The whole process of drawing districts has reached ludicrous nonfunctionality in the US. I’m not familiar with the situation elsewhere, but it’s such an obvious way to enable politicians to choose their voters, rather than the other way around, that it probably crops up everywhere when it’s not actively stopped. And stopped it must be. It makes a mockery of the whole idea behind democracy, whether the voting is used for elections or unelections.

Political districts are simply a matter of population and ease of access to the seat of government. (And polling places, if the technology is at a level where neither mail nor electronics is reliable enough.) That is a problem in geography, not politics. Districts need to be drawn by technicians and scientists with degrees in geographical information systems, not by politicians or judges or other amateurs. In very spiky situations with concerns of bias, three separate groups of technicians could draw the lines, one group from each side and another one from halfway around the world, and the results agreed upon, or averaged, or whatever other solution works to provide an objective outcome. But however it’s done, this critical step in government can’t itself be part of the political process.

One objection voiced against purely population-based districts in the US is that some have been drawn to protect minorities, and drawing them otherwise will reduce those protections. This is an attempt to solve a problem by not addressing it. It’s like the old joke about dropping a coin around the corner, but looking for it under the streetlight because it’s easier to see there. Minority rights need protection. There’s no question about that (and I’ll discuss some methods in a moment). So they should be directly protected. It makes no sense to hope that some kind of protection might emerge from an irrelevant kabuki dance on another topic. The fact that the dance is feasible under the current political system is as relevant to the desired outcome as the streetlight is to finding the coin.

Methods of tallying votes is another bit of arcana that has come to wider attention in the US after the problems with the 2000, 2002, 2004, 2006, and 2008 elections. (It’s getting hard to pretend that these are all unfortunate exceptions with no pattern.)

The bottom line is that the method of adding up the votes can produce a variety of results, as the graphic to the right shows. The example is taken from an article in Science News, Nov. 2, 2002, by Erica Klarreich. Fifteen people have beverage preferences. Six prefer milk first, wine second, and beer third; five prefer beer first, wine second, and milk third; and four prefer wine first, beer second, and milk third. (Example from work she cites by Saari.)

Plurality voting is conceptually easy, it’s very commonly used, and it’s the worst at expressing the will of the voters. (In the example, 6 first place votes for milk is greater than 5 for beer, which is greater than 4 for wine.) It differs from majority voting in that the latter requires the winning choice to have more first place votes than the other choices combined, in other words to have more than half the total votes. There is no majority winner in the example, and a second election between the two top candidates would be needed in a majority vote system.

In instant runoff voting, the choice with fewest first place votes is eliminated, and those voters’ second choice is used in the tally. (In that case beer would win because 6 votes for milk is less than 5 plus 4 for beer.)

Intuitively, it seems that the third system, which allows voters to rank their choices, ought to capture people’s intentions, but the actual result is to favor the least-disliked alternative. (In the example, wine is liked only by 4, but disliked by nobody. So if first place choices have 2 pts, second, 1 pt, and third, 0, then the tally is: 12 for milk, which is less than 14 for beer, which is less than 19 for wine.) I’ve seen this method at work in an academic department where it was mainly notable for favoring the blandest candidate, since that one had the fewest enemies.

A peculiar paradox can occur in multi-round voting that wouldn’t occur in single round, according to Saari. This is true whether or not the second round happens on another date or immediately after the first tally, as with instant-runoff. Increased popularity can actually cause a frontrunner to go down a notch in the tally and lose the election. The process is described in Klarreich’s article. Paradoxical outcomes, where the most popular choice doesn’t win, have been studied mathematically by Saari who’s pointed out that plurality voting does lead to the most paradoxes by far.

There is also a method called cumulative voting. In the easiest-to-use variant, each voter gets as many votes as there are alternatives. If it’s a simple up or down issue, like a recall, the number of votes per person might be one or some agreed upon number for the sake of minority protection (see below). The voter can then distribute those votes however they like. Cumulative voting was studied extensively by Lani Guinier, who published The Tyranny of the Majority (1994), among many other writings. More recently, it’s also been studied by Brams (2001, 2003) (pdf), and other publications. As Brams notes in 2003, “The chief reason for its nonadoption in public elections, and by some societies, seems to be a lack of key “insider” support.”

In the example diagrammed above, cumulative voting would give three votes to each voter. A beer fanatic could put all three votes on beer. Somebody who likes all three drinks equally could give one to each. The outcome of approval voting is not possible to predict mathematically since the distribution of votes depends on the voters. And that is as it should be.

Approval voting has been applied in practice most notably, as far as I’m concerned, in the Australian Territory of Norfolk Island. That’s the closest it comes to being used in an autonomous administrative unit. In general, cumulative voting is applied where minorities are powerful enough to insist on being heard, such as corporate boards, or, in the US, where it’s mandated by the federal government to rectify injustices under the Voting Rights Act. That indicates its effectiveness at thwarting majority dictatorship. However, there are also potential disadvantages in that the system can be vulnerable to insincere tactical voting. Given the degree of coordination required for tactical voting, cumulative methods could be more subject to gaming in the smallest elections. In large-scale ones, it seems to me that the benefit of mitigating majority dictatorship far outweighs the small likelihood of successfully coordinating insincere votes.

Scholars of voting methods could, no doubt, come up with further improvements that would make the likelihood of tactical voting even smaller. Furthermore, to be successful, insincere voters would need to have good advance information about likely election results. But in the system described here, opinion polling is not allowed because it interferes with the independence of decision-making.

When voters need to decide among an array of possible policies rather than candidates, it is especially important to use voting methods that reflect voters’ wishes as closely as possible. Since determining policy is one of the major uses of voting in the system I’m envisioning, it’s critical to use something more accurate than pluralities.

Another advantage of cumulative voting is the relative simplicity of designing understandable ballots. In the example to the right, there are three bubbles to fill in next to each choice. It’s simple to point out that putting all one’s votes on beer, say, would use up the votes and would weight beer the most heavily.

One big problem with majority rule, even assuming that large groups have a way of making the right decision, is that “majority,” by itself does not mean “large group.” Even a mere unelection with only two choices could deliver a bogus result if only a tiny minority of voters bother to cast a ballot. Then an even tinier minority decides the issue, and there’s no crowd to generate wisdom. A quorum is an essential requirement of any election that’s supposed to be fair.

The quorum, I think, should be set quite high, such as 75% or 80% of the population voting in the particular election. In a fair society, registering to vote and voting must be as effortless as is consistent with honesty. If most of the population can’t be bothered to express an opinion when doing so is as simple as returning a postage-paid envelope, then the issue in question didn’t need a vote. It stays as it was, and the election officials try to figure out how to prevent frivolous elections from being called.

Elections could be called by petition or by some other indication of popular will or by the relevant officials, and could be held at any time, although not oftener than once per quarter. Government should be unobtrusive to avoid citizen fatigue, and having elections every few days would not satisfy that requirement. Since a primary function of elections is oversight, not leadership, the ad hoc timing is intentional. It would reduce effectiveness if they were regularly scheduled.

I’m assuming it goes without saying, but perhaps I should be explicit about the remaining aspects of voting. The process must be as easy as is compatible with 1) secrecy of the ballot, 2) fair voting 3) transparency of the process, and 4) maintaining tamper-proof records that can be recounted later.


Minority Protections
 

The issue of minorities is a structural weakness in democracies. Majority rule is a defining feature of democracies, so they’ll necessarily be less than ideal for minorities. Saying “Get used to it” is not a solution. Majority rule is not some desirable goal in itself. Its function is to serve the greatest good of the greatest number. However, people tend to ignore any problems that aren’t their own, and so a minority can get stepped on for no reason at all. Minority protections counteract obliviousness in majorities, and they enable minorities to require consideration of their legitimate needs. Those protections are vital because without them fairness is lost for some, and since it only exists when shared by all, then it’s lost for all. “Fairness” limited to a few is just privilege, and a dictatorship of the majority is still a dictatorship.

Various methods of protecting minority rights have generally been designed with a view to a specific minority. For instance, part of the purpose behind the number and distribution of Senators in the US was to make sure the less-populous states had louder voices in the Senate. At the turn of the 1800s when the system took shape, the important minority was farmers. Farmers remain a minority and they remain important, but they were and are far from the only minority group whose rights need protection.

What’s needed is a method to give any minority a voice. As is generally the case, what’s right is also what’s necessary for long term survival. Ignored minorities are endangering democracies, or preventing their birth in a pattern that’s far too common. Sectarian, tribal, ethnic, and racial strife are endemic at this point. Often it’s fomented by politicians with ulterior motives, but they wouldn’t be successful without a fertile substrate for their nasty work. By facilitating gradual change, democracies are supposed to prevent the need for violent overthrow, but their inability to accommodate minorities is leading to predictable results.

The most promising idea I’ve heard for giving minorities a voice is cumulative voting, discussed by Lani Guinier (1994), which I described earlier. Although it would not help a minority against an equally committed majority, which except in unusual circumstances is as it should be, it would help a focused minority against a diffuse majority. People who felt strongly about an issue, could concentrate all their votes on one choice and potentially prevail. An advantage of building minority protections into the vote tallying method is that minorities don’t have to be defined either beforehand or for the purposes of the vote. Any group that feels very strongly about an issue has a way to make its voice heard.

(There is a point of view that democracies are defined by majority rule, and any dilution of it is somehow wrong. Apparently, there’s something to this effect in Roberts Rules of Order. That idea could only be valid if there’s no difference between democracy and majority dictatorship. If the idea behind democracy is fairness, which is the only meaning that makes it a desirable form of government to most people, then one has to do what it takes to prevent majority dictatorship. That necessarily involves curbing the power of the majority.)

Sometimes, a minority might need greater protection than that afforded by cumulative voting. For instance, consider the case of a small group speaking a rare language who wanted schools to be taught in their mother tongue as well as in the dominant language. A country’s official language is hardly a matter of principle. It’s something that could be appropriately decided by vote. And yet, if it is, all minor languages will die out. Whether or not one sees the value of diversity (I happen to think it’s huge), the fact is that it doesn’t hurt anyone for the smaller groups to preserve their languages. In order to do it, though, they may need actual preference, not just improved equality, to ensure their ability to live as they like.

A similar situation is an ethnic group which is outnumbered in its home area by immigrants. (For a while that was true of Fijians and Fijian-Indians, for instance. Or, as another example, the American Indians and all the later arrivals.) Although the new place is now home to the immigrants, and it’s not as if they necessarily have a country to go back to, yet it’s just not right for a people and a culture to be obliterated at home by the weight of a majority. In that case, too, I could see a weighted cumulative voting system that protects the cultural viability of the original inhabitants.

It’s possible that majority rule would be less of a problem in a fair society, since all matters of rights are decided by law, not by vote. Only matters of opinion and preference could be subject to votes, so minority protections in voting serve as an added safeguard, not as the only one.


Officeholders
 

We’re used to the concept that positions requiring special skills are filled by appointees selected on that basis. The heads of a nation’s financial oversight, environmental oversight, or air traffic control are not elected. Judges are appointed, sometimes for life in the hope of ensuring their independence as well as their knowledge. Those jurisdictions where judges are elected (there are some, at least in the US) tend to show up in the news as examples of appalling ignorance and bad judgment. It would be self-evidently foolish to expect voters to evaluate someone’s qualifications to run, say, the food safety division.

Yet when it comes to running the whole state, we revert to a Romantic notion of the purity of the amateur. A heart full of good intentions is supposed to work better than a skill set. For all I know, that might even be true. We’ll never know because there is no way to reliably find these pure amateurs. Set aside worries about purity for the moment, and consider elections. There is no universe in which amateurs emerge victorious from a system that requires a small or large fortune and a life remade to fit it. We’re getting professionals. They’re just professionals at something other than governing.

I realize there’s nothing terribly new about those insights, and that the objection to trying anything else is that it will be worse. It’s back to saying that “this is the least-bad system out there.” That’s fine as far as it goes, but it’s no reason to stop trying to do better.

I’m suggesting that professionals are needed if we want professional-level workmanship. There are indeed two big pitfalls which stop people from trying that obvious solution. The first is the selection process. Judging skills is very difficult. A thorough evaluation takes much more time than people are generally willing to give, so the tendency is to go to a magic number system. Get a degree from the right school, and you’re in. Take a test that generates a grade, and we’re all set. The Imperial Chinese used elements of a test-based bureaucracy for hundreds of years before they fell, and all it led to was the most ossified system on the planet.

The selectors themselves are another weak link in the selection process. It takes one to know one, and the selectors would have to be at least as adept at governing as the pool of candidates. If they were, though, why would they be hanging around, reading applications? And that doesn’t even touch on the problems of bias or corruptibility. At least with hordes of voters, the thinking goes, it’s not possible to buy all of them and any bias cancels out. (It doesn’t necessarily, but that’s a whole different topic.)

The second pitfall is oversight. Once a person has the power granted by a government position, how do you make sure they’re doing their jobs? If they aren’t, how do you prevent the elite cliques from covering for each other? This factor has the steepest uphill climb against the tendency to forgive all things to the powerful until it’s too late.

It may, however, be possible to preserve enough randomness to prevent fossilization, enough expertise to prevent incompetence, and large enough numbers among the selectors to preserve both any wisdom the crowd might have and prevent corruption.

The thing to keep firmly in mind is that randomness is acceptable. It’s even good. We like to tell ourselves that our selection processes, whatever they happen to be, are set up to find the best. And then naturally we declare the person we have found to be the best, which proves that the process works. On the evidence, this is a circular fairy tale. A comforting one, but nonetheless nonfunctional.

There is an element of randomness in any selection process now operating, whether it’s hiring employees, ranking songs on the web, or picking the market’s stock favorites. Consider elections. Many factors aren’t controlled, and although that’s not the same as random, it might be better if it was. For instance, who runs for election to begin with has little to do with job qualifications. So we already have randomness. It’s merely unplanned.

I’ll give an example of a selection process just to illustrate what I mean. Obviously it hasn’t been tested, and it’s only by a process of trial, error, and study of the evidence that a really workable solution can be found. For low level offices such as city councils or county positions, people who felt they were qualified could put their names up for consideration. They’d be expected to show experience relevant to the position. If it was managerial, they’d be expected to show some talent at managing, even if the evidence came from volunteer work managing a kindergarten. If it required specialized knowledge, they’d be expected to show training and some experience in that field.

The selectors would not choose the actual officials. They would choose who among those who volunteered their names actually had plausible-sounding qualifications. This part of the process would work best in a computer-saturated society. Maybe it would only work in a computer-saturated society, because the pool of selectors should be as large as possible and their ranking of the candidates should be highly redundant. In other words, many selectors should each look at candidates independently and rank them as plausible or implausible. They shouldn’t know what ranks have been assigned by other selectors. Not all selectors should have to look at all candidates, but only at as many as they had the energy to read up on properly. Candidates with multiple “passes”and few or no “fails” would advance to the next stage.

The selectors themselves would be chosen from among those in the population who had the qualifications to evaluate the backgrounds of candidates. It’s the same general idea as the selection of the pool of reviewers for academic journal articles. There’s a large population with known qualifications. They can opt out of the task, but they don’t have to do anything special to opt in. Here again computers are probably essential. The database of people with the background to select candidates for given offices would be huge. The database could be compiled from an amalgam of job descriptions, both employed and self-employed, and of school records showing who received which degrees. If each application is evaluated by a few hundred people, then one could hope the crowd of independent and less-than-totally ignorant people should show a bit of wisdom.

The idea is that by having as large a pool of selectors as is consistent with the basic knowledge needed to be one, the pitfall of ignorant selectors is avoided. Possibly people would be willing to participate in the process out of a sense of civic duty and volunteerism, much as they do now on computer help forums or in the search for stardust in aerogels. If that’s not so, then there could be a formal request system, rather like jury duty now. Qualified people would receive notification to evaluate, say, three applications, with a minimal penalty if they ignore it.

Once the selectors had narrowed down the pool of candidates to those plausibly capable of doing the job —and that’s all they would be doing, they wouldn’t be ranking them — then the actual officeholder should be selected by lottery.

I know that selection by lottery sounds odd, but it’s not really that different from current systems except for two things. One is that the pool of candidates would be narrowed to those who might be able to do the job. Two is that the process would not select according to criteria that have no relevance to the job. Both of these would be improvements. Furthermore, a lottery system would prevent at least some selector bias, and it would prevent the formation of impervious elite cliques. A plain old lottery would have as good a chance of finding the best person for the job as most of our current methods, and a much better chance if the pool was narrowed down at the beginning to those who had a chance of competence at the job. There would also be the added advantage that we couldn’t fool ourselves into thinking we were so bright that the person selected necessarily had to be the best.

(This system differs in a couple of important ways from random appointments in, for instance, the ancient Athenian concept of democracy or from sortition. (Sortition? Who comes up with these terms? “Random selection” would be too easy to understand, I suppose.) The random selection operates on a not-random pool of people with some expertise. The offices in question are those requiring expertise to administer specific functions and are not to find representatives who form a legislative body. Elections and voting are still part of the process, but they function to provide oversight instead of to select government officials.)

In a society of the far future, when fairness has been the standard for so long that all people really are equal, then the composition of the population of officeholders may cease to matter. We’re far from there yet. So one other use to which a lottery system could be put is to ensure that those in office reflect the population they’re supposed to govern. The pool from which the officials are chosen could be limited to those coming from the necessary groups. Half the positions could be filled from the pool of males, half from females, and those in whatever mix of races, ethnicities, castes, or whatever, is relevant. I’m not suggesting that people from different groups can’t feel for each other. But it’s a fact in this stupidly stratified world of ours that they often don’t, and that having people in government from otherwise marginalized groups makes a beneficial difference.

Another advantage to this system is that the candidates would not have to be good at anything except the jobs they’re doing and hope to do (and at expressing that in their background materials). They wouldn’t have to be able to stomach the campaign process in order to govern. They wouldn’t have to know the right people in order to run. They wouldn’t have to figure out which parts of themselves to sell to raise the money campaigning takes.

Higher offices, those at the level of province, state, nation, continent, or world, would be selected in much the same way with the difference that the pool would only be open to those who had proven themselves by, for instance, at least five successful years on a lower rung. The time should be long enough to make sure that an official couldn’t create problems and then escape the consequences by moving up. The track record needs to be complete enough to form a basis for judgment.

Success would be defined as complaints not exceeding some low percentage (to be discussed in the Oversight section), and, certainly, a lack of successful recall elections. It would also mean effective exercise of foresight and prevention of larger consequences in problems that arose, and it would mean a well-managed department and high morale among the personnel who worked for that official in their lower rung capacity.

This would tend to make high officials relatively elderly, since some time has to elapse before they can put their names in for national or international jobs. I would see that as a good thing, since it introduces a natural element of term limits. Ideally, the system would not depend on rapid advances in anti-aging technology to fill the highest slots.

The purpose of selectors at the higher rungs would be, again, checking the backgrounds of the officials who had expressed an interest in a higher position and removing the names of those who clearly did not qualify. The selectors shouldn’t, in this case, select who is in the pool but rather who is out of it. The idea behind that is to reduce the effects of selector bias.

An official’s life under this system would not be a sinecure. The following section on Oversight will show that’s it’s even less of a walk in the park than it seems so far. When the perils of a position are great, then the compensation has to be proportionate. Miserliness about the salaries of government officials would be a big mistake. They’re being asked to perform a difficult, demanding, and generally thankless job, and asking them to do it for peanuts is counterproductive. Their jobs, since they hold the well-being of the population in trust, are more important than private sector jobs. As such, they ought to be compensated proportionately more than private sector executives in comparable positions of responsibility. As I’ll discuss in the chapter on Money and Work, I’m of the opinion that pay in the business world should have some limits. With that caveat, the compensation of government officials should exceed their business counterparts by some percentage, like 10% or 20%, over the average (mode) in the private sector. The tendency of too many voters to resent any rewards for their public servants has to be curbed.

One potential problem, even in the ideal situation of completely competent technocrats, is that nobody would think creatively. There’d be no great leaders, just good ones. Although that would be an improvement, it still feels like an unnecessary compromise to renounce the outstanding just to avoid the awful.

Maybe the continual and inevitable influx of outsiders in a lottery-based system would be enough for new ideas to percolate in. If that looked to be insufficent, then I would imagine that the two most straightforward points to change would be the candidate pool and the stringency of the selection process. Whether it’s best to loosen requirements or tighten them depends on what longitudinal studies of the results of different methods show in practice. Possibly, evidence might show other factors or other methods entirely are most associated with a selection process that facilitates outstanding governing at any given level.

Whatever the method, it’s important to guard against being too clever, primarily because that’s the likeliest excuse for trying to shift the balance of power, but also for its own sake. We, meaning we humans, really aren’t that good at looking into each others’ souls and finding the best person for, well, anything. It’s better to trust to the luck of the draw. That is, after all, how we’ve selected our leaders for all of human history, but without the benefit of requiring any evidence of qualification at all.


Precognitive decisions
 

There’s also a cognitive aspect to decision making that any society aspiring to sustainability, and therefore rationality, must consider. I’ve stressed the need for informed citizenry and education, but all of that is layered on top of much quicker and less conscious modes of thinking. Rational decision-making requires awareness of that level, where it’s pushing, and whether that’s the desired direction. I’ll go over some of the symptoms and factors involved.

Decisions require commitment to a given path and therefore a loss of other options. Without a sense that the decision is right, the process is uncertain and unpleasant. In the interests of avoiding that, we do what we can to feel sure, and that’s where the trouble starts.

That conviction of being right can come from ignorance, as in the Dunning-Kruger effect (which should be a joke, but isn’t). That research showed something most people have observed: the incompetent rate their own abilities more highly than the skilled, precisely because they don’t know enough to realize how little they know. Education may not fix that, because the ability to absorb the information may be lacking.

Or the conviction can come from knowing it all. Confidence in being highly trained can lead to ignorance of bias. Education cannot fix that. Examples can be found in any field requiring a high level of skill where, for instance, experts hold on to pet theories against the evidence or where there is discrimination against women or minorities. In other words, all of them. I’ll mention two examples.

Orchestras hired men because they played better. Once hiring committees started having auditions behind screens, where the actual music was the only information available to them, they hired more women (Goldin & Rouse, 2000). Recently, women actually do better than men in a blind selection process, implying that not only were the experts wrong in their assumptions about men, but also that women as a whole are more than equally good. They’re better. (A not unexpected outcome when mediocre women are less likely to attempt careers than mediocre men.) But it came as a big shock to many of those doing the hiring that something besides their superior knowledge of music was influencing their decisions. (I’d actually be surprised if quite a number didn’t think there was something wrong with blind selection rather than with their decision-making.) Unawareness is just that. Unawareness. Saying “I know I would never do that” only confirms the lack of awareness if one has just finished doing precisely that.

Another example is the prevalence of tall male CEOs. The hiring committees spend months on these decisions, examining all sorts of minutia. Yet somehow, the end result of all the conscious thought is that tall is better. That’s true for a chimpanzee trying to lead a troop. There’s not a shred of evidence to say it’s true for the skills a CEO needs.

That is an important point. The emotional reaction, the “first impression,” the judgment made in those first few minutes before there’s any chance that evidence could come into the result, that emotional reaction is not merely a confounding factor that takes effort to discount. It takes precedence over reason. Reason then serves to justify the decision already taken. The results tell the story, not the convictions.

And that relates directly to the “intuitive betrayal hypothesis” (Simmons & Nelson, 2006), which refers to acting against intuition. The “hypothesis predicts that people will be more confident in their final decisions when they choose the intuitive option than when they choose a nonintuitive alternative.” Not only is that first impression obtained without relevant evidence, but subsequent evidence that contradicts it is discounted because of the confidence conveyed by gut feelings. The emotional confidence doesn’t arise when unsupported by intuition, so counterintuitive decisions never feel “right,” no matter how much evidence supports them.

Gut feelings operate with the same level of analysis available a million years ago. That doesn’t make them useless. They may work better than expected for avoiding danger or finding a sex partner. But they lack the ability to parse modern situations that didn’t exist then. And yet it doesn’t feel right to go against them. In short, we have a problem.

(I know this runs contrary to much of the recent motivational press about trusting gut feelings. That’s a welcome message because it confirms what we want to do anyway. On the evidence, however, it does not lead to reality based decisions. Don’t be surprised if your intuition says otherwise.)

Either way, whether from ignorance or smugness, it’s important not to assume that people will make decisions based on the best available information. They will do that when it’s the only available information. Not being omniscient, we don’t necessarily know what the best is, but the totally extraneous can be easier to determine. Thus, for instance, politicians’ hair styles can be confidently stated to have no predictive value for indicating their legislative abilities. Everybody, including all voters, knows that. And yet, when voters have that information, it becomes a big enough factor to predict the winner. (pdf downloads on click) (Huber, Hill, and Lenz. 2012).

We need to consciously deprive ourselves of the information used in “first impressions” and force ourselves to consider evidence first, and then force ourselves to reconcile any large incongruity that crops up between between intuition and evidence-based choices. We need to force ourselves to be aware of the information coming from intuition and to consciously examine whether it’s pushing in a fair and sensible direction.

There are some people who can examine their assumptions and force themselves to ignore intuition and go purely by evidence. That’s the essence of scholarly work, and it takes most people years of training. Even with all that training, scholars are notorious for holding to erroneous pet theories. Without all that training, the odds are much worse. So, although everyone who reads this will know they, personally, have no trouble discounting emotion, in the aggregate there are too few who can do that. Even with all the education in the world, rationality must be actively and consciously given priority and extra weight if it’s to have a chance of taking its optimum place in decision making.


Laws
 

The biggest breakthrough in creating an egalitarian legal system came in the days of the ancient Romans. The laws, they said, should be written. That way they were accessible to everyone. They applied equally to everyone (officially), and all people had the same information as to what they were.

The only part of this noble ideal that has been preserved, now that most people can read, is that the laws are written down. To make up for that, they’re incomprehensible, inaccessible to anyone outside a select priesthood, and applied unequally based on access to that priesthood. It’s an egregious example of how to deprive people of their rights without raising any alarm until it’s too late. That’s not knowledge that’s been lost since the time of the ancient Romans. More recently, Thomas Jefferson made the same point.

I know that the official justification for legal language is precision. If the laws were written in English then, God help us, we’d have to figure out exactly which implication out of many was intended. Yet, oddly enough, nobody seems to know what the stuff written in legalese means until it is so-called “tested in the courts.” And then the decision handed down becomes a precedent, that is, an example of what that law means in practice. In order not to get lost in a forest of possible interpretations, it’s customary to follow these precedents. Yet, officially, the reason we need legalese is because otherwise there could be different possible interpretations.

Furthermore, tell me which has the more unequivocal meaning: George Orwell’s piece on Politics and the English Language or any legal document you care to dig up. The whole argument that legalese is somehow more precise is nothing but a smokescreen. Its precision is an illusion generated by the common consent of an ingroup to exclude outsiders. The meaning of words is always a matter of common consent. When the meaning pertains to the laws of the land, that consent truly needs to be common, and not limited to a priesthood.

Laws in a fair society would have to meet a plain language standard. There are various ways to measure adequate clarity. One, for instance, would be to make available Orwell’s essay above, or one of Krugman’s newspaper columns, or any other clear piece of writing about difficult concepts, and see whether people who understood them correctly also understood what a new law said. If people who have no trouble understanding plain language misunderstand the new law, it needs to be rewritten until it says what it’s trying to say. Areas of the law that apply only to a specialized subset of people, maritime law for instance, would need to be comprehensible to the target group, not necessarily to those who aren’t relevant. Plain language laws are one more thing that would be much easier to do in a computer-saturated society. A few hundred independent readers would be plenty to get an accurate assessment of comprehensibility.

Another aspect of clarity besides plain speech is plain organization. At this point new laws which are revisions of older laws which expand on yet older laws which rest on seriously old laws written in an idiom so out of date they need to be translated are all tucked away in separate books of law and referenced in a Gordian knot of precedents. People get degrees for their ability to track laws through thickets much worse than that sentence. That situation, too, is rife with opportunities for conferring advantage on those in the know. It’s not accceptable in a fair society whose legal system cannot work better for some than for others.

The situation is analogous to the issues faced by programmers trying to track the changing code of a program. They’ve come up with version control systems to handle it. A very small subset of what they do is also in document management options that can track changes to text. As Karl Fogel notes, version control has application to a wide range of situations in a world where many people make changes to the work of other people. Staying for now with legal codes, a version control system identifies

  • who introduced or changed which item, whether line or paragraph or bill,
  • which other items it relates to, influences, requires changes in, or negates
  • and which items are currently proposed, being reviewed, or accepted into the main body of code.
Just the first line, showing who authored which bits of proposed legislation, would be a large change from the current opaque system. In the software world, one way to check authorship is the command “git blame.” Computer geeks have a sense of humor, but it’s nonetheless true.

A further improvement needs to be made to an eventual legal version control system. The branches and webs of related laws need to be easy to find, search, access, and understand by ordinary users with no previous training. That will take invention of new methods by experts in the presentation of information, because as far as I know, no such thing is available in any field yet.

I’ve been discussing laws as if they arrive by immaculate conception, which was unavoidable since I’ve eliminated the elected class of legislators. Who, in this system, are the lawmakers?

There are different aspects to the process of creating laws. Proposing a law to address an overlooked or new problem is just one step. Then it must be checked for redundancy. If not redundant, it needs to be written clearly, then debated, and finally passed. The checking and writing parts are specialist functions that require skilled officials as well as general input from the public.

The best originators of laws are hard to pin down. Philosopher kings, plain old kings, and priests have always made a mess of it, so far. Democracies generally have settled on the idea that there’s really no way to know. Greengrocers or dancers or doctors of law can be elected to office and their interest in the job is supposed to be a hopeful sign that they’ll write laws worth having. Their aides help them with the legalistic bits. That system has worked better than the wise men system, so why not go with it all the way? Anyone can propose a law if they see the need for it.

Again, I’m serious. Allowing anyone to propose laws is no different from allowing legislators to be elected regardless of background for the purpose of making laws. The only difference is that by making the process open there’s the chance that the Queen and King Solomons among us may contribute. There’s the chance, of course, that the really stupid will also contribute, but we seem to be surviving their help in the national assemblies of the world. Furthermore, under an open version control system, proposing the law is just the first step. The checking and debating (in papers, or internet fora, or real ones) will have the attention of professionals as well as the general public, and that would likely be a more stringent selective process than what we have now. The professionals would also have the responsibility of formally writing it down and checking the text for clarity. Redundant or superfluous laws could be weeded out at this stage, as could ones that contradict rights or accepted laws.

There are some areas of law which are necessarily specialized, but the modifications required in this system are minor. Plain language in that case means plain to the people who will be using the law. The reviewers will be those affected by it. The professionals checking its wording and validity would have the relevant specialized background. The principle, however, remains the same.

Then, last, comes the decision whether to make the new law part of the active code. I see that as another matter best left to the wisdom of a large but qualified crowd. The selectors of laws would be all those with the professional qualifications to know whether the new law agreed with guaranteed rights. A quorum would be a few hundred, and the majority favorable should be much larger than half, such as three quarters or even more. Their function is to determine constitutionality, and if they’re rather evenly split, that’s not a good sign. As with my other examples of methods, there may be better ways to achieve the goal of a streamlined, functional, adaptable, and equitable legal code. The point isn’t the method but the goal.

Moving on from the making of law to its practice, there is at least one systemic lapse that increasingly prevents justice instead of ensuring it. The time required to reach a verdict is appropriate only for bristlecone pine trees. Justice delayed is justice denied, and by that measure alone there’s very little justice to be had. This fact has not been lost on corporations who have been using lawsuits as a weapon. The Intel-AMD lawsuit comes to mind as an example. (1, 2, and links in those articles.) The few decisions actually handed down have said that Intel did abuse its market position. But then Intel appeals again, or they delay trial by finding new emails, or some other thing. Even when it’s the US Supreme Court, Intel disagrees with the decision and is appealing. I wasn’t aware that it was possible to appeal a decision by the Supremes, but one thing is eminently clear. Intel continues the disputed practices and AMD loses by it. Meanwhile, the courts labor over stalling tactics as if they meant something. (Update 2009-11-11: the endless suit was finally settled out of court. AMD settled for much less than their real damages, no doubt because they gave up hope of justice instead of delays.) By such means a good thing, due process, is turned into its opposite. The situation is so egregious that, on occasion, the lawyers themselves point it out.

Simple cases should be decided in days or weeks, medium ones in months, and complicated ones in less than a year. The legal system needs deadlines. It will improve, not hurt, the cause of justice if they have to focus on the merits of cases instead of every possible minutia that can be tacked onto the billable hours. There is no human dispute so complicated it can’t be resolved in that time unless there’s determined obfuscation. Even the US Supreme Court feels that a lawyer can adequately summarize a case in half an hour. All aspects of hearing a case can be given short and reasonable time limits.

The way cases are tried in our current system, with judge(s), lawyers, juries and the lot does not necessarily have to work against fairness. But from what I’ve seen myself, the jury system seems rather hit or miss. The deliberations for one case I was on were surprisingly short because one of the men had booked a fishing trip for the afternoon on the assumption he wouldn’t actually be called for duty. I happened to feel we came to the right decision as it was, but those who didn’t were definitely quashed by an agenda that had nothing to do with the law. There are thousands of stories along those lines and they are disturbing, but possibly they could be fixed. Maybe the system would also work better in a plain language context with short trials that didn’t waste everyone’s time. Below is an example of an alternative way to structure a judicial system.

Legal professionals are selected by the methods already outlined or by earning a degree. In other words, selectors can decide that some combination of legal training and experience has provided knowledge of the law and of rights equivalent to a degree.

Amateur juries are replaced by panels of legally trained people that are chosen with input from the two sides. Each side chooses from one to three of these professionals, depending on the importance of the case or the point in the appeals process. Those professionals, call them “judges” perhaps, then agree on another one to three, for a total of from three to nine on a panel.

The panel familiarize themselves with the case, hear the arguments of the advocates or the two opposing sides if they haven’t hired lawyers, ask the two sides questions, and ultimately hand down a decision and a sentence, if any. The voices of the “middle” members of the panel could be weighted more heavily if necessary through a cumulative system, as in voting. Two further appeals with newly selected, and maybe larger and more stringently qualified, groups would be allowed, but if they all decided the same way, then the decision would stand. If they did not, the majority of decisions after a further two appeals, i.e. three out of five, would stand.

Even though the current system could, theoretically, work fairly, in practice it has enough problems to make alternate methods worth testing. My notion of one such alternative is in the example to the right. It retains the somewhat random element seen in juries, and the training requirement for professional members.

All aspects of the system need to be tax-funded. Nobody can be financially beholden to either of the two sides. That’s essential to reduce the influence of wealth on the whole legal system. The court apparatus and all the legal professionals, including the lawyers for each side, must be publicly funded. Acquiring extra legal help based on ability to pay doesn’t fit with a fair system.

I realize that the first thing wealthier people would do is hire somebody who calls themselves anything but a lawyer and proceed from there. That has to be illegal and punishable because it subverts fairness for everyone. The model to follow might be medical licensing. A license would attach to being any part of the legal profession, and one of its requirements would be to receive no outside funds of any kind. Those found offering any kind of legal advice without a license would be fined some multiple of whatever they received from their clients, with increasing penalties if they kept doing it. It really is that important to keep money out of the courts. That should be obvious from the current state of justice.

My thinking is that a system as in the example, with ad hoc collections of professionals, would cut out much of the chaff from the legal process by obviating the charade supposedly needed to keep legally inexperienced jurors unbiased. Yet it would preserve the presence of expert assistance for each of the two sides, and even add voices for them when the final deliberations were in progress. Because the legal personnel are assembled partly by the parties involved, a class of judges with undue personal power should not develop.

I know that such a minimally stodgy system seems very odd, very unlegal. Stodginess is such a constant feature in courts that it’s absence feels like the absence of law itself, or at least the absence of due process. And yet, what we see is the slow, deliberate pace being used to deny due process, not assist it.

The real issue, however, is that whatever form the legal resolution of disputes takes, whether slow or not, it should have the hallmarks of fairness. It should be equally accessible to all, it should treat everyone equally, and it should accomplish that with a minimum of effort, anguish, and expense for all equally.


Administration and Taxes
 

These are the practical, boring parts of government. Nameless flunkies are supposed to just do them somewhere out of sight.

In many ways it’s true. They are practical and boring … in the same way as bricks. If they’re done wrong, the whole building comes down.

In a sustainable system, they need as much attention as principles. It’s quite possible they need more attention because they form the actual structure. They not only keep the whole edifice standing, they’re also most people’s actual point of contact with the government. Last, and far from least, precisely because they’re boring they’re an easy place to start the creeping abuse of power without causing enough resistance. So, mundane or not, it is very important for the administrative aspects to facilitate fairness instead of counteracting it.

I’ve already touched on the relation of politics to voting districts, but there is a more general sense in which the organization of a country’s — or the world’s — subunits can foster equal treatment for all. Currently, bigger is better because it gives the central government control over more resources, taxes, and people to conscript into armies. That is really the only reason to deny self-determination to smaller groups. The Turks don’t want control over some of the Kurds because they value their culture so much. The Chinese haven’t colonized Tibet or Uighur because they admire the scenery.

Administrative borders delimited according to varying criteria would be more conducive to self-determination and peace. There is no necessary minimum size for a group based on language, culture, or a shared philosophy. A viable economic unit, on the other hand, requires several million people to have enough complexity to be stable. But there is no requirement that those two have to be the same unit. Small groups could run their own schools and video stations in their own languages plus the main language of the larger group. Small units could together form one viable economic unit. Those could work together for their common benefit vis-a-vis other similar large groups.

In terms of borders, this is not that different from the nested sets of units within and between nations. The difference is that self-determination is not limited to the biggest units but is spread to the smallest units compatible with a system that doesn’t waste people’s taxes. Things that need to stay coordinated over large areas, such as transportation or roads, are wastefully inefficient when handled by many smaller units, as, for instance, railroads in the US. Things that are best handled on a local level, such as membership in a time zone, cause inefficiency when imposed from far away, as, for instance, the single time zone in all of China. The bigger the unit, the less its function is actual administration and the more it is arbitration between its members and with other similar groups. People could identify with nations or any group that suited them, but administration should devolve to the lowest level where it can be accomplished efficiently.

Taxation is a supremely prickly subject, perhaps even more so than national power. The couple of points I’d like to make are that fairness requires an equal burden, and that any burden must have a limit or it becomes unfair no matter how equal it is.

An equal burden is not measured by the amount of tax. It is measured by the payment capacity of the taxpayer. That should go without saying, but if it did we wouldn’t have people suggesting that flat taxes are the “fair” solution. Paying a 50% tax for someone who makes $10,000,000 per year will have no effect on their life. It’ll mean they put $5,000,000 less in the bank. But someone who makes $10,830 the 2009 poverty level wage in the US would die of starvation if they paid $5400 in taxes. The burden is just as disproportionate if the amount is 1%, although the consequences for the poor might be less vicious. I say “might” because at a 1% tax rate government services would be near-nonexistent and there would be no safety net at all. An equal burden means that the proportion of tax paid relative to income should have an approximately equivalent felt impact across income levels.

How high a burden is too high is a matter of opinion. As a matter of data, the countries that provide significant services for the taxes they collect — medical care, education, retirement benefits, and the like — take around 40% of their GDP on average in all taxes. (That and much other tax data at the OECD.) Some of the countries in that situation, such as Australia, achieve 30% levels. Japan reaches 27%.

(Just for comparison, the US takes about 28%, but only the elderly receive funded medical care and only school children may get free education. If the money US citizens have to spend on the services provided elsewhere was added in, they’d have the highest or near-highest tax rates of any developed country. The per capita medical spending here is $7200, college spending is around $20,000 per student, which works out to around $1300 per capita. Adding in only those two items puts the US in the 46% rate, but with much worse outcomes than, say, Norway, at 44%.)

So, speaking in round numbers, the data suggest that a government can fulfill even very advanced functions on about one third of the GDP. Given also, that large numbers of people from many different cultures receive services and pay that amount without feeling wronged, an amount in that range could be considered a rule of thumb. In other words, taxes, as a matter of law and rights, could not go higher than a fixed percentage, like 35%, and for that they would have to provide advanced benefits. Better service for less cost is, of course, always an improvement. For each reduction in benefits, taxes would be reduced by law to a lower percentage. The government would be limited in how much money it could take in, at any given level of service. How that tax burden is distributed among businesses, asset owners, and income has to follow the principle that the burden is shared equally among people. In other words, a business is not a person and its taxes would be evaluated in light of how they affect the owners, workers, and customers. In the interests of transparency, the taxes should be evident, not hidden in other costs.

Last, I wanted to discuss some aspects of bureaucracy. The system I’m suggesting would depend heavily of technocrats, bureaucrats, professionals, or whatever you’d like to call them. Controlling bureaucracies is therefore essential. Unfortunately, there’s no clear path to follow. If there’s an example of a bureaucracy, whether public, private, military, educational, or artistic, that didn’t grow until it became an end in itself, I haven’t heard of it.

I’d be willing to bet there’s a very simple reason for that. It’s human nature to admire status, and that accrues with numbers of people and size of budget. It does not accrue from doing an excellent job with the fewest resources. And the size and budgets of departments depend partly on the departments themselves. The need for new hires is determined within the bureaucracy, not outside of it. The very people high enough up the scale to control hiring are the ones who gain respect, promotions, and better pay based on the importance of their empire. Bureaucracies can’t be controlled until that factor is consciously identified and counteracted.

If that reward structure was changed, there might be a chance of controlling bureaucratic growth. What that new structure should be is something management experts have been studying and obviously need to continue studying. There’s still a shortage of effective methods in the wild. Of course, the first hurdle is that the people who know how to work the current system don’t want effective methods, but that’s the usual problem of unseating vested interests. That part is not specific to bureaucracies as such. If it were to turn out that no matter what, bureaucracies can always pervert any rules and keep growing, then my whole concept of a government of specialists would be unworkable. Some other way than what I’m envisioning would have to be found to keep concentrations of power to their optimal minimum.

I’ll give an example of how I imagine limits to bureaucracy might work. Again, not because this is necessarily the best way but to illustrate my meaning.

First, there’s data on what proportion of expenses are spent on administrative and infrastructure costs in various types of organizations. For instance, Medicare has been much discussed lately. Krugman notes that it has 2% administrative costs against the private insurers’ 11%. Others give the figure as 3% or 6%. Among non-profits, food banks have near-zero administrative costs, whereas museums, given the exacting nature of their requirements, tend to be around 15%. Research institutions, whether university, federal, industrial, or military, have administrative and infrastructure costs in the low 30% range. As with so many things discussed so far, there is no one right answer. But, for a given situation, it is possible to find out what people spend on administrative personnel and infrastructure, to compare that to what the more effective institutions in that class spend, and to set realistic limits. Those limits should probably be set somewhat higher than the best-in-class. We’re not all geniuses, and the first priority is good service, not a government which is the low price leader.

The limits as a percent of the budget would be set outside the bureaucracy in question, and based on data collected across the whole functionally comparable sector. The percentage amount would not be subject to appeal. The bureaucracy would have to account for its spending, but it wouldn’t be submitting budget requests. They’d do their jobs with a fixed amount of money and there’d be nothing to request.

If a department’s functions changed — if, for instance, a breakthrough like the invention of PCR meant the Patent Office would be processing vastly more inventions based on that — then the department would submit evidence of the change to the equivalent of a General Accounting Office which would decide whether or how their budget needed to be adjusted.

From an outsider’s perspective, one big source of bloat in bureaucracies is management layers. Ranks, like budgets and pay scales, should also be limited externally to the flattest structure compatible with effective management. Ranks should not be at the discretion of the department head any more than staffing.

Forms are an ever-present aspect of bureaucracies and probably the main avenue of interaction with them for the general public. Commenting on such a picayune thing as forms may seem odd, but abuses start in the small stuff. Just as more bodies means more status for the very person doing the hiring, similarly paperwork is a way for bureaucrats to act like they’re doing a job. What makes it bad is that they are the ones with sole discretion about the forms they generate. In a system that expects bureaucrats to run things, forms will likely metastasize unless their growth and proliferation are stopped.

Paperwork is a symptom of a much larger issue. It indicates whether government serves the citizens or the continued employment of its officials. If bureaucrats are really to be public servants, then an important factor in their jobs is to consult the convenience of their masters. That means forms must be as few in number and as simple and non-repetitive to fill out as is compatible with getting the job done optimally.

There should be explicit limits on the number and length of forms in any given department. If a new form is needed, the department then has to figure out which old one is obsolete. There should also be explicit limits on how long it takes a naive user to complete all the paperwork necessary to achieve a given result. (The US government prints estimates of how long it will take to complete its forms. No limits, just estimates. It’s the beginning of the right idea, but the estimates are so low they can only be explained by time travel.) What the limits should be on the time it takes to fill out forms is a matter of opinion. However, if it was decided by vote, I and everyone else would always vote for zero as the appropriate amount of time. I’m not sure what the fairest and most effective consensus method should be to decide that question, but whatever it turns out to be, setting limits on forms is essential.

Last, one of the things bureaucracies are famous for is ducking responsibility. As I noted in earlier chapters, leaving that loophole open is not a minor lapse. It goes right to the heart of a fair system that requires accountability. Government by committee is not compatible with fairness. Government by passing responsibility up to higher levels is also not compatible. Responsibility and the control relevant to it needs to rest with the lowest level official who can effectively carry it. Responsibilities and control must be clearly demarcated to prevent officials from passing it up the line … or down to someone without enough control over the decision. The person who controls the decisionmaking is also the one who must sign off on it, and who will be held responsible if fault becomes an issue.

– + –

In summary, decision-making in an equitable and sustainable system would differ from our current one in fundamental ways. Changes range from the far-reaching to the mundane, from the end of war and opaque laws to disciplined bureaucracies.

[Continued in Government 2: Oversight]
 

 

 

 

 

 

 

 

So far, I can't make comments appear on archive pages. Please click on the post title or permalink to go to the post page, if you'd like to comment.