Bureaucracy's strength is in getting repeatable results out of large numbers of unreliable human beings. There are places where that's the best strategy.

But as soon as you're trying to do something unique or one-off or under-specified, bureaucracy impedes more than it helps. Which is probably why large bureaucratic organizations often get out-innovated by small organizations with far less resources at their disposal.

Bureaucracy is a solution to a coordination problem that I would argue doesn't really exist anymore. Software is, in one sense, bureaucracy codified in mechanical form. It replaces and obsoletes human bureaucracy. The classic bureaucrat essentially applies a rulebook to his input information and passes output information to someone else. He is a human computer.

Even the evolution of war leaves open the question whether bureaucracy is an asset or a liability going forward. War is getting faster and more targeted.

At least with a human there's the possibility they could be persuaded, reasoned with, or even corrupted.

A computer on the other hand is utterly intractable. I'm not sure how that would help with unique, one off or under-specified problems.

If you want your bureaucracy to be more adaptable, I'd argue you need to give it more flexibility not less.

>> A computer on the other hand is utterly intractable

I'm sure several people on here would disagree with that, as there are plenty of people good at "convincing" computers to do things they shouldn't be doing, or to do things that their owners don't want them to. You just have to be more of a black hat, less of a schmmoozer....

Yeah, should've probably put a caveat in that!

I'd argue that even a hacked system is intractable though, as it still rigidly follows the rules no matter the outcome. That's usually what black hats are exploiting in the first place.

If the problem with bureaucracy is rigidity, then replacing human agency with computed agency would only exacerbate that in my opinion.

For a fictional point of reference to the above sentiment: remember HAL 9000 in 2001: a Space Odyssey, and the explanation of why it went homicidal as given in 2010: Odyssey Two. It had two directives: preserving human life and the mission; the first conflicted with its ability to ensure the other's success once the humans started plotting to turn HAL off. In the movie (at least), it was implied that the mission's importance was imparted by Presidential authority, and was given a higher priority than anything else.

I disagree. Bureaucracy is inherently a job of managing people, in all their imperfect emotional people-like glory. Bureaucrats don't just put numbers into a spreadsheet and get out an answer, they balance the needs and wants of competing stakeholders to allow society to function effectively a collective.

Someone balancing needs without numbers to back up their costly actions is why higher-order bureaucracies exist.

A (governmental) bureau is responsible for a delegated area of licensing (pre-emptive control), regulatory oversight (ongoing control), and "criminal" investigation (retributive action for violation), as granted by its enabling authority.

Bureaucracy is the discipline/science/art of managing a bureau, which isn't producing competitive market-action revenue (profits), so can only control efficiency by containing costs, which itself is an action that doesn't produce market revenue.

Oversight of a bureau (ascendant bureaucracy) exists solely to make sure some metric of cost containment and compliance with enabling authority is maintained within each bureau itself.

In short, bureaucracy is the regulatory management of regulatory management.

And yet, the (modern) classic symbol of intractable bureaucratic hell is the green screen.

Von Mises analyzed Bureaucracy in a book aptly named (in English translation): Bureaucracy. Bureaucracy is a mechanism used to control costs in an enterprise absent measurable economic profitability (cost/revenue efficiency). In private profit-seeking enterprises with sufficient scale to have division of labor, subsets of the organization may be deemed "cost centers", these are the ones that get the most bureaucratic over time. Even regulatory compliance departments main goal is to constrain the costs of fines from non-compliance.

The omni-present danger in bureaucratic management is that the mechanisms to maintain and manage the bureaucracy are themselves non-profit seeking, and tend to motivate growth in the bureaucracy. In private enterprises (barring regulatory compliance requirements), the overall cost of its bureaucratic functions impedes its profitability, and should tend to limit the size of the "cost center" budgets to maintain market profitability (or a failed enterprise).

For government functions, the overall cost of the bureaucracy can only be paid for by tax receipts and public debt (and license and fines, which as the Chief Justic pointed out are effectively tax receipts). If there is no political will to constrain one of both of those, the bureaucracy will grow unbounded.

Anyway, the short of it has nothing to do with marshaling resources, it has to do with constraining costs absent profit.

> a military which does not use bureaucracy loses wars to ones that do.

You couldn't be more wrong about this. The decentralized Vietnamese beat the (by-then) bureaucratic US.

The US army during WWII was defined by its ability to delegate to the lowest possible level of the officer hierarchy and this was one of the major reasons it was able to beat the rigid and highly-centralized Nazis.

The Confederacy came within a stones throw of winning the Civil War, which would have been the equivalent of two guys in their garage taking down one of the worlds largest Corporations. They lost because of the overwhelming differences in resources, but their agility and ability to promote talent allowed them to run circles around the North for the first two years.

Winfield Scott and Hernando Cortes both won in Mexico because they distanced themselves from their respective governments. They were allowed complete freedom from bureaucracy because they went out of communication range of the politicians.

Sun Tzu actually made a name for himself by beating a major neighboring kingdom with a tiny, but agile force.

I could keep going...

The Civil War is a better example than Vietnam -- not only because public support died in the US with Nixon's "Guam doctrine", but because it wasn't a particularly existential struggle for the US -- that is, my point was that there is a "natural selection" of social institutions based on military defeats. (Also, South Vietnam was an interesting question in its own right: to what extent is a strong-man or dictator a "bureaucrat"? Chinese civilizations are a grand example of a clear bureaucracy with a central dictator, so the answer isn't trivially either "not at all" or "absolutely.")

Just to be clear: I am not arguing against agility and I think that the central agile ability -- the ability to fail cheaply -- is probably the most important feature in many areas, including some military conquests. (One of the most interesting applications of cheap failure is interface design. If you can test a new idea cheaply it allows feedback loops which make you feel like you're really "interacting with something.")

In any case, you've given me some food for thought, and thanks for that.

The Civil War is an excellent example, and the best example of the ones I gave.

I think one of the problems people have with evaluating the efficiency of government is that most of the projects government engages in are necessarily very complex. The point of government research is to work on projects that are too blue-sky to be immediately interesting to industry, but the flip side of this is that such projects are inherentl difficult to manage. DOD projects are notorious for this. Every big DOD RFP asks for something literally nobody has ever built before. When I worked at a company that contracted for DARPA, the government literally didn't know what it wanted us to build because our work would inform their opinion of what was possible to build. You can't hold these sorts of projects to the same standard as implementing an invoicing system for a company that's basically the same as every other invoicing system ever made.

Ok, how about a welfare system then, it seems to me that a government welfare ministry should distribute about 90% of what it takes in spending 10% on administration, when I looked into my government it was spending 55% on administration.

Put this into perspective, once the US finished paying interest on debt, medicare and welfare there's no money left over for anything else.

Science programs are one thing, but the government can't deliver 70 year old services with reasonable efficiency.

It depends entirely on the structure of the welfare program. For example, the veterans affairs administration provides services, mostly health services, directly to veterans. In that case, you'd expect 100% to be overhead, because VA doctors and hospitals are paid out of overhead.

Most federal welfare programs, however, involve cutting a check. Look at the Social Security Administration. It has a budget of about $12.7 billion, and administers $700-800 billion in benefits to 50-60 million people: http://www.ssa.gov/pressoffice/basicfact.htm

That amounts to an overhead of 1.6%, or ~$250 per person served.

Or, look at it another way. Total federal payroll is about $180 billion, for 2.15 million federal employees (a number that has fluctuated between 1.8 and 2.2 million for the last half century despite nearly a doubling of the population). That's about 4.7% of the budget, or 1.2% of GDP.

Most of the federal government is not overhead. 95% of it is just taking in tax revenue from certain people and cutting checks to other people.

You could get that kind of efficiency, or get closer to it anyway, if you just cut people a check. It turns out that's more effective at combating poverty than most social programs anyway. But it's politically impossible.

Bureaucracy is a solution to a very difficult scaling problem. Dunbar's number is around the limit to the largest group that can naturally work together. Above that you need something else to scale, and bureaucracy is one of these things.

The main problem with a bureaucracy is that they appear to have a tendency to almost always decrease in efficiency over time:

http://en.wikipedia.org/wiki/Parkinson%27s_law

Agreed. To some extend bureaucracy exchanges negative liberty (freedom from constraint) for positive liberty (the enabling force). The exchange has some overhead not only in terms of money, but we also seem to pay more negative liberty to obtain the ``same'' amount of positive liberty.

Haha, that sounds like the perfect competition for a startup country :P.