The first-order conditions of the problem are Cn=0 Vn=1+ En -Cn-Pna(sa+sp)=0 Since second-order conditions are satisfied, we derive the optimal contract (m*, n*). The socially harmful activity is decreasing in the policy parameters (o, Sa, sp), whereas the productive effort is increasing in those same parame- ters(because Cmn>0) As in the usual framework(Polinsky and Shavell, 2000), we consider social welfare to be the sum of the payoffs of the employer and of the management minus the social damage caused by the socially harmful activity. Social welfare is given by W=m+n+e(n)-C(m, n)-P(m)H-h where H is social harm. Notice that the difference between the government's objective and the employer's is the social damage. By setting Sa +s H/o, the government can make the employer's objective identical to its own Nevertheless this is not a first best outcome because enforcement is costly Becker, 1968) It is not very relevant who is actually punished since management and ployer can bargain er ante and reallocate sanctions. It is equally effective to set sa=H/o and sp=0 or sp=H/o and sa=0. Furthermore, individual liability of management alone induces efficient behavior Corporate liability is not needed or necessary unless there wealth onstraint that limits sa. Suppose there is a binding liquidity constraint so that sa=o<H/o. Then, we should have sp=H/o-o to fully internalize social damage Corporate liability is justified on the grounds that managers do not have enough wealth to pay for social damage(Polinsky and Shavell 1993; Shavell,1997). n our model, the principal is the government, not the corporation. The corporation and its management team are the agents. There is virtually no distinction between corporation and management because their interests can be aligned at no cost. Once the alignment of interests is costly, the manager is the agent, but the corporation becomes a supervisor or a quasi-enforcerThe first-order conditions of the problem are: Vm = 1 − Cm = 0 (4) Vn = 1 + En − Cn − Pnσ(sa + sp) = 0 (5) Since second-order conditions are satisfied, we derive the optimal contract hm∗ , n∗ i. The socially harmful activity is decreasing in the policy parameters hσ, sa, spi, whereas the productive effort is increasing in those same parame￾ters (because Cmn > 0). As in the usual framework (Polinsky and Shavell, 2000), we consider social welfare to be the sum of the payoffs of the employer and of the management minus the social damage caused by the socially harmful activity. Social welfare is given by: W = m + n + E(n) − C(m, n) − P(n)H − k (6) where H is social harm. Notice that the difference between the government’s objective and the employer’s is the social damage. By setting sa + sp = H/σ, the government can make the employer’s objective identical to its own. Nevertheless this is not a first best outcome because enforcement is costly (Becker, 1968). It is not very relevant who is actually punished since management and employer can bargain ex ante and reallocate sanctions. It is equally effective to set sa = H/σ and sp = 0 or sp = H/σ and sa = 0. Furthermore, individual liability of management alone induces efficient behavior. Corporate liability is not needed or necessary unless there is a wealth constraint that limits sa. Suppose there is a binding liquidity constraint so that sa = ¯ω < H/σ. Then, we should have sp = H/σ − ω¯ to fully internalize social damage. Corporate liability is justified on the grounds that managers do not have enough wealth to pay for social damage (Polinsky and Shavell, 1993; Shavell, 1997). In our model, the principal is the government, not the corporation. The corporation and its management team are the agents. There is virtually no distinction between corporation and management because their interests can be aligned at no cost. Once the alignment of interests is costly, the manager is the agent, but the corporation becomes a supervisor or a quasi-enforcer. 7