What is good about bureaucracy? It is good at handling a multitude of small tasks, procession forms and the like. It also is a source of jobs. Who were all these door-to-door campaign workers for the political campaigns?
What sorts of informational tasks get served by bureaucracy and what incentive problems are there? Bill Rogerson of Northwestern's economics department was funded by a nice grant from the Pentagon (a geometric form of a bureaucracy) to study just this question. Consider the problem of defense procurement. Congress has general oversight on what defense spending gets undertaken. It can set a general budget and can decide how many units to buy. However, it is very hard for Congress to design or set specific details for how armaments should be. Can you imagine Congress debating how thick the armor should be on a tank or what size ammunition artillery should use or how fast our jets should fly? It would be terrible. So we trust the Pentagon to decide how our weapons should be designed and to suggest what qualities they should have and we leave it to Congress to decide how many the US actually buys.
(Rogerson papers on this topic:
"Simple Menus of Contracts in Cost-Based Procurement and Regulation", American Economic Review 93 (2003): 919-926. Stable URL: http://links.jstor.org/sici?sici=0002-8282%28200306%2993%3A3%3C919%3ASMOCIC%3E2.0.CO%3B2-D
“Economic Incentives and the Defense Procurement Process”, The Journal of Economic Perspectives Vol. 8, No. 4 (Autumn, 1994), pp. 65-90. Stable URL: http://links.jstor.org/sici?sici=0895-3309%28199423%298%3A4%3C65%3AEIATDP%3E2.0.CO%3B2-S )
So what's the problem? Generally, people who work in bureaucracies think what they do is important and good. We would not want this to be otherwise. However, except for odd cases where powerful members of Congress are defending jobs in their districts, it is rare that Congress spends as much on the military as the Pentagon would like it to. In particular, Congress rarely orders as many units of planes, ships or tanks as the military establishment would want. So the military bureaucracy sets the quality of "high-techness" of our armaments and Congress sets the quantity of these armaments we buy. This is a two stage game where the military tries to maximize our military might by setting a high quality with the knowledge that Congress will probably decide to buy few units. Rogerson's work indicates that this process leaves the US with weapons which are very expensive and complex of which we buy few. So we defend ourselves with too few but overly complex weapons. In a world without the need of bureaucracy, we would probably have many more weapons but that each weapon would be less complex and that we would be better and, per unit of force, more cheaply defended.
How do we end up at this sub-optimal situation? People have beliefs about the world and this leads them to choose to work in certain bureaucracies, and to believe that doing more, or spending more in that area would be good for society. (And might also increase their salaries or resources.) Alternatively, people might accept lower wages in exchange for having cooler toys, so as a result, IT departments might be prone to request top of the line computer hardware, even when simpler, cheaper systems would do just fine.
This is the classic principal-agent agent problem. Agents act on behalf of the principal (for instance physicians act for patients) but agents may face incentives to act only to benefit principals.
William Niskanen has such a model of optimal bureaucratic output: Bureaucrats that try to maximize their own budgets will try to request budgets for their agencies such that the total amount they request is equal to the total benefit their agency produces. But TB = TC is not an optimum! Instead, funding for the agency should be cut off beyond the point where its marginal benefit equals its marginal cost.