Decentralization by Function and Location

Under what conditions is decentralization of facilities rational for a client-centered system of service or administration, and when is great centralization more cost-effective? Four aspects of decentralization are identified: (1) pluralization of facilities (e.g., service points or channels); (2) dispersal in geographic space; (3) specialization by function or kind of service; and (4) adaptation to the specific requirements of each case through repeated feedback passes or negotiating queries. Different functions or kinds of service are treated as being located in a function space; the distances among them correspond to the number and cost of adjustment steps which a server or a service facility needs to shift from one function to another. For calculating the optimal number of facilities, n, the most important variables, calculated as averages, are the service load; the geographic distance; the cost of time spent in transporting a request and the response to it, and the transport speed; the functional distance or number of functions; the cost of time in adjusting to the function requested; and the adjustment speed; and the total fixed cost per facility. Other important variables are the number of negotiating queries per request, and an indicator of the value of time (i.e., of a speedier response). Under certain assumptions, such as neglecting economies or diseconomies of scale, the optimum number of facilities for a system with a single function then varies as the square root of the service load; this is also the case for a system with a single location but several functions. For a system with several functions and locations, the optimal number of facilities varies as the two-thirds power of the service load. Long-term trends may be toward decentralization when service loads and the costs of service time grow faster than capital costs and transport and adjustment speeds, as seems likely for the next several decades. Where the opposite conditions prevail, cost-effectiveness should favor centralization, such as perhaps in some earlier periods, and possibly in the more distant future. Possible applications through eventual elaborations of these oversimple models are suggested for fire, ambulance, and police, school and local government systems; hospitals and clinics; libraries, telephone exchanges, and computer systems; laboratories, repair shops, and mail order houses; law courts and judicial districts; and the numbers of administrative districts and sovereign states.