When an agent exert a negative externality such as pollution, it is well known since Pigou that an emission tax equal to the marginal damage could help internalise this effect. Others instruments such as standards or marketable permits are also used. However, it is also clear that an efficient policy to correct externalities implies to control ex post the decisions made by the agents (actual emissions’ level, pollution abatement investments, ...) together with adequate sanctions in case of non compliance. As control is costly for regulators, this implies random auditing of emissions if possible.

Normative analyses have recently shown that requiring firms to self-report their compliance status could be beneficial for regulatory agencies that seek to enforce environmental laws. This permits to audit polluting firms less often as they face certain taxes and are not audited if they declare any non compliance. Nevertheless, they need to be punished more often and this could make self-reporting useless if sanction costs are too high with regards to audit costs. Thus, these analyses show that self-reporting is used as a complement to an audit mechanism.

We think that this result can be due to the fact that regulatory agencies observe nothing before the reports made by firms and before the decision to audit. Actually, we enrich the preceding analyses by incorporating the possibility for the regulator to get an imperfect signal upon damages caused by non-compliance before deciding to inspect. This information can be viewed as coming from victims’ claims or from the firms themselves (increase in production). This could also be the result of a simple costless technology used by the agency to detect imperfectly any non compliance. In this framework, we show that self-reporting and random audit are substitutes : given the relative size of audit and sanctions costs, the optimal control policy consists in taxing on the basis of firms’ reports and nature of signal received or auditing with regards to the signal.

In the last part of the paper, we consider a repeated version of the static game and show the superiority of state-dependent pollution control enforcement with a two group model over static enforcement.