I have a strange hobby. I am not sure when it started or why. I think it may have been many years ago when I was thinking about writing a science fiction story based upon a future where eugenic genetic engineering was common.Whatever the origin, I sometimes consider what I would change in the way human beings are designed. For example, I have considered whether or not locating the brain in the head is a good or bad idea. On one hand, it is a rather exposed location, poorly shielded against impact. On the other, a brain located in the torso would likely overheat with the slightest fever. Not to mention that the greater distance from eye to brain would result in slower visual processing times. It is a strange hobby, I know, and one with very little practical application, as I doubt I will ever be in a position to redesign humanity.Then again, it is a relatively harmless hobby, and it amuses my wife to think I spend my time in such odd pursuits.

The reason I mention this is that it provides a perfect example of a case where centralization may have greater costs than benefits. In my thoughts about redesigning humans, other than increasing pressure and particle density in the inner ear to improve balance, the most certain benefit seems to come from redesigning the heart. The basic problem is that there is but a single heart, so either injury or strain can cause total failure, and, unlike lungs, kidneys, eyes and other systems, there is no redundancy. Now, we can't have two hearts, because any lack of synchronization would likely cause one heart to blow the valves out of the other as pressure increased at the wrong moment. However, as an alternative I have imagined a system of muscular bands around the veins and arteries which push the blood. Rather than a single pump, the blood vessels would work more like the intestines, with the blood getting constant little pushes all along the course. It would still have the potential for localized over- and under- pressure, but overall I think it would greatly improve the circulatory system.

But my little thought exercise is not the point. The point is that there are cases where centralization can be harmful. And that is an important point.

You see, when we hear arguments for "health care reform", for example, all we hear are the benefits of centralizing all these functions in a single place. It is implicit in the term (no longer popular) "single payer system". The idea is that by putting everything under the control of a single entity we will remove all the inefficiencies and redundancies caused by multiple systems, as well as enjoy economies of scale, and perhaps discover other benefits currently not obvious due to the distributed system.

And, for those who have but the most superficial understanding of economics and political science, those are persuasive arguments. They are not only persuasive, they have been successful. The same arguments have been used over and over to convince individual voters that the powers traditional vested in the states should be handed over to the federal government, that centralization will lead to efficiency as well as remove impediments that currently prevent us from operating as efficiently as we could.

Basically, the argument is simply an excessive emphasis of economies of scale, arguing that too much distribution both creates redundant services and "waste", in the form of duplicated efforts and services created entirely to serve as an interface between distributed systems. It s also argued that, under a distributed system, some things simply are not possible at all which can be done under a centralized system. And so, the arguments go, if we only put all of these powers in one place, then we will enjoy incredible benefits with little cost.

It is persuasive, or can be, in part because there is some truth to it. There is inevitable redundancy in distributed systems, and there is also cost to interfaces between those systems. And, in some cases, there are economies of scale that can be realized only through centralization. Of course the argument also overlooks many factors, such as the benefits of distribution which can serve to offset the costs, or the fact that economies of scale do not always exist, and sometimes are better realized in smaller systems. It also faisl to ask whether those things that are only possible under a centralized system should be done at all.

So, to act as a counterweight to all those singing the praises of centralization, allow me to offer the counterarguments. I do not offer these to say that centralization s never justified. Clearly even a dedicated federalist mus recognize there are times when a centralized system is justified, or else why have a confederation of states? In fact, if centralization is never beneficial, then why have government at all? So, obviously, there are cases where centralizing functions, to some degree or another, is justified and beneficial. However, as centralization already has so many cheerleaders, I figure we need someone to offer the other side of the argument. And I am assuming that role. As so many of my most cherished beliefs, the free market, federalism, individual autonomy, rest of a belief in the benefits of some degree of decentralization, it seems a natural argument for me.

The first point to make is a simple one. "Economies of scale" has become such a common argument for centralization that it has overshadowed another popular economic phrase which argues for the opposite. That is "division of labor". Now, granted, division of labor is not explicitly and argument for decentralization, but neither are economies of scale strictly an argument for centralization. Both are analogies. However, as centralization has used the concept of economies of scale to imply "bigger is better" and push for more consolidation, I would argue that division of labor pushes in the opposite direction. Rather than consolidating all functions in a single individual or location, it is often better to break them apart, move them away from one another, and let each piece take care of his own concerns. Specialization is a good way to argue against the benefits of centralization.

Some will argue that centralization doe snot mean there will be no division of labor, as labor can still be divided within the centralized authority. However, I would argue that is not entirely true. While at some levels labor will be divided, at the upper levels, the levels where decisions are made, there will be an inevitable call for generalists, and the more functions consolidated the more general they must be. When functions are distributed, each can be managed by a specialist, as each is independent, there is no need for an overall coordinator. Only when you centralize do you find yourself in need  of management that can manage an absurd range of fields. Just think of the number of industries which congress regulates, and ask yourself whether any congressman could really be an expert in each of those fields, or else must rely on staffers, lobbyists, friends and guesswork to make decisions. That is why I say that division of labor argues for decentralization. Centralization requires too much knowledge at the upper levels, and doe snot allow tasks to be separated. And so management becomes too generalist and too inefficient.

A second argument is the one I made about the heart, the problem of a single point of failure. In my present job (computer administrator) there is a near phobia of "single points of failure", everything must be backed up, contain a disaster recovery plan and employ duplicate or triplicate systems to allow continued service during the failure of one or more machines. I have planned not only for the failure of a single machine, or single piece of hardware, but for the failure of an entire data center, or a whole network connection. Yet, in our government, charged with safeguarding the safety of our entire nation, there is a push toward a single point of failure.

Let us think about the recent bank problems. Why did so many banks offer subprime loans which went bad? Because the government encouraged it. I won't get into the real cause here, but whether you think it was foolish regulators encouraging bad loans, Wall Street greed cashing in on government regulations, or inflationary pressures pushing money into guaranteed loans, the single common factor is the government. By creating regulatory agencies which force a single decision on every firm in the industry, they create what is effectively a single point of failure, in which a single bad decision is replicated in every firm.

In a distributed system you can still encounter such events, but they are far less likely. It is conceivable every firm in a market could make the same bad decision, and the tendency to follow the practices of successful firms makes it a bit more likely, but there is also a high probability some maverick firm will do differently, and, more importantly, once the relative merits and disadvantages become obvious, it is more likely other firms will dot he same. In a centralized system, this is far less likely, as the entire system is controlled by a single entity. (And, in a bureaucratic system, where blame avoidance is paramount, admitting a mistake by changing course is often career suicide -- see "The Inevitability of Bureaucratic Management in Government Enterprises","Bureaucratic Management", "The Bureaucratic Mind", "The Inherent Disappointment of Authoritarianism".)

Another example, from a non-economic arena, is found in Jared Diamonds Guns, Germs and Steel (a book with which I disagree as often as I agree, yet which is as thought provoking as Hansen's response Culture and Carnage, which I also recommend.) I am not sure if the argument actually supports Diamond's theory or not, as it is predicated more on straightforward politics than geopolitics, but Diamond does make a case for the role of geography. In any case, leaving aside the meaning in terms of Diamond's thesis, an interesting example comes from the Chinese Empire's age of exploration. During this time, China managed to send expeditions as far as Africa, explored much of the Pacific and generally was set to become a potent colonial power. However,t hanks to political infighting and other matters with which we need not concern ourselves, China turned its back on exploration and adopted an isolationist policy. And, as China was the preeminent power in the region*, largely cut off Asia from the rest of the world, thanks to the Empire being a single point of failure.

What makes this interesting is to compare Europe at the same time. Had any European power decided to adopt an isolationist policy, even the papacy or the Holy Roman Empire, it would not have had the same effect. Thanks to the many nations within Europe, a single decision would have only local impact, the greater flow of history would continue in the other nations. Which means that, unlike China's single point of failure, the distributed governments of Europe could not be changed by a single decision.

Actually, the mention of the papacy points out the one piece of European history which was dominated by a single point of failure. Prior to the rise of protestantism (or maybe the proto-protestant Hussites and Lollards, as Hungary broke with the Church prior to Luther),  the papacy did form a single point of failure**. A perfect example of which would be the stagnation of financial services in Europe. Prior to the Protestant revolution, advances in the mercantile class were, uncharacteristically, not met by developments in finance. Thanks tot he religious prohibitions on usury banking simply did not develop as one would expect. There were some exceptions, whether in states which openly defied the papacy, in those which found subterfuges (such as collecting rent on mortgaged lands until redeemed) or by organizations such as the Templars who could expect to be excused violations of papal regulations. It was only once Protestantism provided an alternative that the Papacy stopped acting as a single point of failure and Papal decisions lost the ability to control the direction of all European history.

Beyond being a single point of failure and minimizing specialization, centralization has other drawbacks.

One commonly mentioned when discussing federalism is the lack of ability to test out alternatives. As we know from the free market, having hundreds of competing firms means we will have hundreds of approaches to making a profit. And, among those approaches, some will prove better than others, and will eventually be emulated by other firms. In this way, the market serves as a laboratory where theories can be tested, compared, evaluated, and then adopted by those who think one theory or another has proved itself better. More important, the competition of the market also serves as a means to test the evaluations these firms make, as those who judge efficiency properly will thrive, while those who do not will fail. A similar experience also exists in true federalist systems. States can try out various rules and regulations and compare the results. Those which do best at satisfying citizens will see populations rise, taxes increase and generally will thrive, while those who refuse to adapt will see voters flee and tax revenues dry up.

A centralized system does not have this ability. At best it can test one theory at a time, by changing the laws for everyone. And even then, it is very difficult to evaluate the results. In a distributed system, one entity can try A and one B and the end results can be compared. Yes, there may be some confounding factors, but the results are at least for the same time and roughly the same circumstances. Under a centralized system there is nothing similar. If a central system switches from A to B and see tax revenues rise, was it because of the change? Or because of a hundred other factors? And how tot est? Switch back to the presumably worse policy of B and see if things get worse? That is unlikely. So they stay with A, simply assuming the improvement was due to the change.

As this problem exists, the centralized system tends to be dominated by theoretical debates about changes with little empirical evidence. We see this today in arguments over the impact of gun control and abortion. Without many differences between states due to federal laws, the debate is largely centered on theory since there is simply too little evidence. Even if there were a way to fairly compare the results of a change under a centralized system, this would still be the case, for several reasons. First, a lack of tests. In a distributed system, you can test as many theories as you have entities, less one for a control. Under a centralized system, you can test only one, and doing so means changing the entire system, which, as there are many who oppose such a policy, means forcing it upon those who disagree. And so, as a result, we have debates entirely centered on theories, devoid of most empirical evidence, with decisions based upon those theories changing the course of the entire nation.

Which points out another problem of centralization. As every decision is "winner takes all", with the result of the decision changing circumstances for everyone, debates become very heated. Of course debates are likely to be heated anyway over issues such as abortion or gun control, but with central control debates are heated over issues such as monetary policy and tariffs. In a distributed system, decisions still effect a considerable number, but they also have an alternative. Should local laws prove too disappointing, should they find they are living under a regime they detest, then they can always leave. With multiple laws coexisting, there is always the option of changing one's residence. That is not an option under a central system, and, as a result, debates tend to take on much more significance to the participants then under a more diffuse system.

In addition to these clear drawbacks, there are a number of other advantages and disadvantages which must be balanced. As I discussed at the beginning, the benefits gained from eliminating duplicated efforts and other costs of diffuse systems must be balanced against the benefits that the distributed systems confers, such as improved efficiency from competition, resilience of redundant systems and so on. Going into all of the possibilities here would be impossible, as it would have to be specific to a single situation, and even then exhausting all benefits and costs would be time consuming. So, let us just say that in any case when deciding whether to centralize service sor not, there are a host of factors to consider, not just the simplistic "savings" we often hear touted in debates such we are having over health care.

--------------------------------------------------------------

* Yes, there were other nations and some fairly powerful, but by size alone, China overshadowed all her neighbors. If you look at the history of Korea, southeast Asia, or even Japan, it is obvious that much of the history of those regions depends on events in China. This is not to marginalize those nations. If you think about it, the expulsion of the Turks, or the rise of the Yuan dynasty had effects felt as far as western Europe, so many nations were, to some degree, touched by the events in China. Europe was simply less effected as it was much farther away. The closer nations did not have the same buffer and so were much more strongly influenced by events in China.

** The Eastern Orthodox church was sufficiently distant from western and central Europe that it did not form a viable alternative to the Catholic Church. Especially when the Turks began the push into Europe, absorbing most non-Russian Orthodox states, the Orthodox Church had little influence in most of Europe.

---------------------------------------------------

POSTSCRIPT

My earlier writing on health care is scattered all over this blog. A good place to start would be the list of articles in the postscript to "Bad Economics Part 2 ", as well as my most recent posts on the topic "Reviving Nonsense in the White House", "The GOP Health Care Plan" and "A Different Look at "Health Care Reform"".

Correction: (01/06/2010): I should always read my own posts before publishing. For some reason, I wrote that Hungary broke with the Papacy prior to the Protestant Reformation, when in truth it was Bohemia. The two, as well as Austria, were united by strong family bonds through most of their history, which probably explains the slip, but I would certainly have caught the error had I done even a bit of proofreading. Yet again I am reminded that being lazy is a bad thing. (And yet, I did not bother to correct the countless typos I spotted while reading this, just the big factual error. So, despite decrying laziness, this is still a very lazy correction.)