perjantaina, toukokuuta 26, 2006

Top-down vs. bottom-up

I just reread an old article by Milton Friedman on the US War on Drugs. In it he develops a strict dichotomy between two forms of economic organization, one centralized and the other decentralized. In his words: "One way is by market mechanisms: from the bottom up. The other way is by command: from the top down." At least to a libertarian, it's rather obvious which approach works and which doesn't.

But why is it that you can't have decentralized government or centralized markets? Why is freedom almost synonymous with markets while centralization invariably seems to lead to
institutionalized coercion and diminishing standards of living?

While I do think that coercion and centralization are bad, and that free markets and liberty go hand in hand, I don't really think the above dichotomy holds generally. There can be decentralized, local government, and it can be necessary for a well functioning society. Well-functioning, centralized markets exist as well, for example in electricity, and they aren't too different from state bureaucracies if you look at them closely enough.

However, as far as economic valuation goes, there is a basic, Hayekian argument in favour of bottom-up and in opposition to top-down organization. The basic institutional economics problem with top-down organization is that hierarchies have an apex by definition, and so if causation flows down such a pyramid, ultimately all of the nodes in the hierarchy will be subservient to the wants, needs and capabilities of the instance at the top. All of the economic actors within the hierarchy could in principle have their own interests, capabilities and resources which could be used to build a tremendously rich economic system, serving and utilizing all those different attributes simultaneously. But if decisions only flow downhill, they will all be conditioned on the attributes of the supreme decider. Information on local, personal or specialized circumstances cannot even in principle be taken into consideration, because such knowledge will not be available at the top; if it was, knowledge would be flowing the up the pyramid, and so the economy couldn't work purely top-down by definition. Thus while there might be millions of simultaneous variables to optimize for optimum welfare in an extended society, the degrees of freedom in a top-down organization will equal the number of degrees of freedom possessed by the top man.

On the other hand, if information and causation flows bottom-up, any piece of information could in theory be present at any level, even if only a small proportion of the sum total could be present at any level simultaneously. The only constraint is the ability of the actors at the different levels to summarize, prune and refine information in ways that minimize glut and maximize relevancy to the higher level actors. This is a significant problem in itself, but unlike top-down organization, it is not a categorical limitation on the degrees of freedom or the amount of relevant information available in the economic system as a whole.

I think this basic pattern is what makes bottom-up economic organization so much more successful than the alternative, even if we forget about practicalities like incentive compatibility: value is subjective, so the above argument shows that you can only produce lots of such value by aggregating individual choices. Going at it the other way around is impossble because as soon as someone above you in a pyramid gets to decide, by definition it will be his values, capabilities and limitations that dictate the outcome. Your potential will be wasted unless it is completely redundant (it never is), your particular knowledge will be lost unless it is entirely shared (it most definitely isn't) and value as you define will not and cannot be produced unless you happen to define it precisely the way your superior does (pretty much impossible, since part of you being separate individuals is your having several interests).

This is nothing new, of course. But then, we can generalize. If we define value by aggregation and then form some sort of consensus opinion at the apex which subsequently gets passed back down, the same problem arises. Since there is just one apex, its internal degrees of freedom once again dictate the space of optimization for the whole economy, regardless of how the state at the apex was arrived at. Thus, it's not just command economies which suffer from the above mentioned problem, but also all kinds of democracies, bureaucracies, private pecking orders and the like. If value is to be produced maximally, value creation has to be highly polycentral, and consensus valuations imposed back down on smaller constituent parts of the economy have to be avoided even if at the root they derive from aggregate, popular opinion.

In the end each hierarchical, political subdivision above the individual simply diminishes the capability of the economy of elucidating what value really is. All that remains is hard individualism, that is, free markets based on individual rights. The only possible deviation from this norm is when you can make choices affecting other people for technical reasons (i.e. technical externalities), which would naturally place you on the next hierarhical level with respect to them. It might be that some form of collective machinery (i.e. a minarchy) is needed to abolish this sort of natural hierarchy. But then that's it; just one, highly limited excuse for government really exists if you want to respect the subjective theory of value.

2 kommenttia:

Anssi kirjoitti...

Why such an arbitrary and violently simplifying model of world as a tree network of information flow? Besides, the amount of information is not all. Top-down has obvious efficiency: fast and coherent actions across the organization.

The point of free markets is freedom, not the market as a flat, idealized non-hierarchical market of a country fair. Free market companies and specialized market cultures ("The Internet transmission capacity market", "electricity market", "market for ISO 9000 certificates") can have any amount of rigid hierarchy, especially when "fast action" is more important than "precise action", or when markets are ripe, low margin, high volume established ones.

What is important, that they are free, i.e. that everybody involved wins. They must not be perfectly Pareto optimal, as long as there are no obvious way to win more without coercing someone.

Sampo Syreeni kirjoitti...

In a market economy the model obviously won't be anywhere near correct. But in a bureaucracy it will be. Either it's a centralized economy or it's not. If more than one person gets the ultimate say, with anarchy or whatnot prevailing between them, then at the level of those persons it already *is* a bottom-up thing, and decentralized at that.

Also, my point was narrowly constructed. It was about value creation, nothing more. If you want to value something, eventually you will have to end up with some summary measure of the valuation. The tree -- or a single rooted directed acyclic graph if you want to generalize -- comes in by definition because the point of final aggregation does have to exist somewhere. Things like actually producing something call for something besides a tree-like organization -- I actually said as much with respect to local government and electricity markets -- but when you want to elucidate traditional universalist concepts of worth, economic demand, or, say, group decision making processes, a tree model is, I think, quite appropriate.

In any case, the analogy does give reason to presume that even given a problem where a rigid centralized structure is the most efficient, it isn't so because of axiological concerns but in spite of them. An efficient bureaucracy might react fast or solve a number of problems with sequential choice and undivided commons (I suspect the main reason firms exist is team production and metering, in the Alchian and Demsetz vein), but at the same time it will also carry a price in terms of being able to create value.

In any case, you're entirely correct, and I too find the pervasiveness of overly simplified tree models in human thought tiresome. For one of the clearest explications of the point, see Christopher Alexander's classic A City is not a Tree.