The delicate and difficult balance to be struck between autonomy and control was succinctly summarised by T. Moe, who argued that the mechanisms linking the agencies to all the other branches of power are so numerous and complex that they ultimately result in a situation where “no one controls the agency, and yet the agency is under control”.188 This paradox is easy to explain.
Modern administration does not easily tolerate restrictive mandates worded so precisely that they remove any scope for discretion. It is thus in a position of constant interaction between rulemaking and enforcement.189 Consequently, judicial review is never easy, especially when scientific issues are at stake.190At the same time, individual citizens or groups may contest agencies' actions before the courts and possibly win their cases. Such proceedings remain limited however, and cannot ensure that all of society benefits from the agencies' activities. This is particularly true of the EU, where, owing to the very limited possibilities for individuals (natural and legal persons) to bring direct action for annulment before the European courts, decisions can be usefully contested only once they have been implemented in national law.191 As a result, we need a diversified battery of interconnected mechanisms to ensure the right balance between autonomy and control.192
This variety of mechanisms is also required by the operating costs that agencies may generate, which may be administrative or financial. If, by administrative cost, we understand the potential gap between the objectives of the delegator and the work of the agency, a gap which is likely to widen the broader the remit is, and by financial cost, the higher or lower expenditure that might be generated by the intensity and/or scope of the agency's regulatory activity, it is clear that these costs are closely interlinked. A broad remit may reduce the delegator's costs in terms of regulatory production, but will certainly increase the costs of monitoring the agency to ensure that it complies with the terms of its remit, while a very restricted remit will probably have the opposite effect. Hence, the need for an intelligent balance between the degree of independence and appropriate control measures, as illustrated by the great variety both of types of agency and of control mechanisms.
Over the past twenty years, particular authors, drawing essentially on US practice, have developed an impressive series of supervision mechanisms, the typology of which varies greatly according to the relative importance each attributes to one or other of them.193 I have chosen to present these mechanisms in five categories, which seem to me to better reflect the state of thought on the subject: (a) executive oversight, (b) budgetary evaluation, (c) procedural control, (d) judicial review and (e) network coordination. Let us now consider each of these in turn.