Liability rules, Pigouvian taxes and combinations of the two approaches

From an efficiency point of view there are two problems associated with the legal rules analysed in the previous section. First, legal rules such as strict liability and negligence rule involve compensation payments flowing from injurers to victims. This compensation alters marginal costs and incentives in each market, and in competitive markets this affects entry and exit decisions (and therefore overall production) in each industry, leading to inefficient entry and exit. Secondly, the legal rules do not (and cannot) take into account the level of activity in each market, and so ignore information that is vital for efficiency of the long run competitive equilibrium in each industry.

In contrast, an efficient Pigouvian tax simply sets a care-dependent specific tax on firms in industry i equal to the marginal harm at the optimum. In other words, consider a tax which is set at at tt (xt) = Q*H (xt). Firms in industry i can minimise their tax bill by choosing the efficient level of care. Thus, marginal costs in industry i under this tax would be c + wix* + Q* H (x*). This would induce efficient production in industry i, so Qt - Q*.

Firms in industry v would still be harmed, with harm equal to Q*H (x*) per unit of the good. But there is no need to compensate firms for this harm. Indeed, as condition (4.13) shows, facing firms in industry v with this harm is exactly what is needed for efficient outcomes in that industry. A no liability rule (that is, a rule which does not compensate victims), combined with a tax of tt (xi) = Q*H (x{) therefore produces efficient outcomes.

Notice that the optimal Pigouvian tax really involves both a tax and a legal rule. This suggests that other combinations of taxes and legal rules may be efficient. This turns out to be the case. Consider, for example, a strict liability rule combined with a tax on industry v. This would also produce efficient outcomes here. To see this, recall that under a rule of strict liability, firms in industry i choose the efficient level of care per unit of the good, which increases marginal cost in the industry and results in efficient production overall, conditional on Qv. But under this rule, firms in industry v would be fully compensated for all harm, which reduces their marginal costs and results in an inefficiently high level of production in industry v, and a correspondingly low level of production in industry i.

Therefore, to control excessive production of Qv in industry v under this legal rule, a tax equal to Q*H (xt) would need to be set. This would raise marginal costs in industry v to c + Q*H(x{), inducing efficient production decisions in that industry (since firms in industry i would choose the efficient level of care under a strict liability rule), as well as efficient production decisions in industry i.

Efficient control of accidental harm is often seen as an 'all or nothing' choice between Pigouvian taxes and legal rules. This is the wrong way to think about Pigouvian taxes. In fact, as this section has shown, a traditional Pigouvian tax really consists of two instruments: a tax on the injurer combined with a legal rule of no liability. But other combinations of taxes and legal rules may produce the same outcomes, and there may be no reason to prefer one over the other. The incentive and efficiency effects of the tax will depend on the characteristics of the legal rule it is being combined with.

Finally, we conclude this section with a word of caution regarding Pigouvian taxes. Note also that our criticism of legal rules - namely, that they ignore levels of production in each industry - can also be levelled at Pigouvian taxes. The taxes explored in this section - and Pigouvian taxes in general - have significant informational requirements. In the above examples, the government must know either Q* or Q*, as well as the shape of the H (x{) function, and must be able to observe the level of care taken by firms in industry i. In a world of imperfect information there is no reason to expect that the 'correct' Pigouvian tax will be chosen, and therefore no reason to expect that the Pigouvian approach will yield superior outcomes to (say) a negligence rule.

< Prev   CONTENTS   Source   Next >