Suppose that an industrial lobby, denoted by \(I\), has developed a new product. If the new product is marketed, lobby \(I\) will obtain a benefit \(b\) from selling it. However, the new product may also have environmental and/or health detrimental effects. If the new product is marketed, it is assumed that an environmental lobby, denoted by \(E\), will bear a damage \(\delta\).¹¹It is assumed here that that the two lobbies truly and fully represent their members’ interests. The problem of free-riding is considered in section 6. By assumption, only lobby \(I\) observes \(\delta\) before the product is marketed. However, it is common knowledge that \(\delta\) is uniformly distributed on \(\left[0,1\right]\).²²By assumption, the root of all asymmetric information in the model comes from the fact that only the industrial lobby, as the developer of the new product, is in a position to test the associated detrimental externalities. Clearly, the latter directly determine the damage \(\delta\), but are much less likely to influence the benefit \(b\). This justifies why we postulate that the benefit is public information, whereas the damage is private information. ³³The situation where neither lobby observes \(\delta\) is not considered, because the lobbying activities then always foster a less efficient decision-making by the government. The situation where both lobbies \(I\) and \(E\) observe \(\delta\) is considered in section 6.

The new product needs to be officially approved before lobby \(I\) can market it. Two ways of doing so will be considered and compared below. In the first one, the regulator simply decides according to his prior beliefs using a cost-benefit analysis. In the second one, the regulator’s decision depends on the two lobbyists’ efforts, according to a simultaneous Tullock contest (Tullock, 1980). Formally, lobbies \(I\) and \(E\) simultaneously bid nonnegative values, respectively denoted by \(x\) and \(y\), and the regulator approves the new product with the probability⁴⁴Tullock (1980) uses the general contest success function \(x^{r}/\left(x^{r}+y^{r}\right)\). We consider here the special case where \(r=1\), which is commonly used and referred to as the ”lottery contest” in the literature (Konrad, 2007).

Finally, we assume that lobby \(I\) can be held liable for damage and that a strict liability rule prevails. In other words, if lobby \(I\) sells the product and lobby \(E\) bears a damage \(\delta\), lobby \(I\) will be asked in court to pay \(\delta\) to lobby \(E\). Furthermore, suppose that the maximum amount that lobby \(I\) can pay is limited to \(k\), which can represent either a statutory limit on damages or his level of assets.⁵⁵With the latter case, the fact that lobby \(I\) can pay a judgment of only \(k\) reflects the implicit assumption that \(b\) cannot be paid back (since otherwise, the available assets would be \(b+k\)). In the literature (Shavell, 1984; Shavell, 2005), this is usually justified saying that \(b\) is a utility benefit. This justification does not apply here, where \(b\) is a monetary benefit by definition. Here, an alternative justification would be that \(b\) is distributed to the shareholders as dividends before the judgment. Hence, assuming that \(k<1\), lobby \(I\) will sometimes pay \(k\) instead of \(\delta\), with \(k<\delta\).