- Email: [email protected]
Customary versus technological advancement tests
International Review of Law and Economics 28 (2008) 106–112
Contents lists available at ScienceDirect
International Review of Law and Economics
Customary versus technological advancement tests Bruno Deffains a,∗ , Dominique Demougin b a b
Economics University, Paris 10 and CNRS, France European Business School, 65201 Wiesbaden, Germany
a r t i c l e
i n f o
Keywords: Tort law Standard of care Customary test Technological advancement test
a b s t r a c t In an environment where the optimal level of care is unknown, we ask, under a state-of-the-art defense, which method is better able to induce parties to undertake optimal care. Assuming courts can see a noisy signal of research activities undertaken by a defendant and by some of its competitors, we ask whether courts should use a biased or an unbiased average to compare care. We find that the latter is advantageous. © 2008 Elsevier Inc. All rights reserved.
1. Introduction It is well known that over the course of the last century producers’ liability has become an important topic in most developed economies. The economic justifications for the evolution of producer tort are numerous and well-documented. From a positive perspective, products sophistication has been advanced by Landes and Posner (1987) as well as the ability of the consumer to be correctly informed on the risk of accident (Brown, 1974; Spence, 1977; Viscusi & Moore, 1993). Law and economics scholars have also argued about the excessive activity level under a negligence rule (Calfee & Craswell, 1986; Landes & Posner, 1981; Polinsky, 1980; Shavell, 1980). Moreover, from the point of view of legal theory, some authors propose to justify producer’s liability based upon the risk represented by the consumption of products rather than on the behavior of the producers (for a general discussion see Hylton, 2000). In the same way, the second part of the 20th century has been characterized by a growing acceptance of the notion that more extensive tort liability would serve to compensate parties as an insurance mechanism.1 Nevertheless, some issues remain controversial. One such question is under which circumstances a producer might be able to defend himself? The purpose of this article is to explore the welfare characteristics of a defense based on a state-of-the-art argument. To fix ideas, consider the following fictitious example of assessing the liability of a producer. A pharmaceutical firm has designed and marketed a medication to reduce cholesterol levels. Fifteen years later, a researcher team discovers that this particular medica-
∗ Corresponding author. E-mail addresses: [email protected] (B. Deffains), [email protected] (D. Demougin). 1 For a general discussion of liability rules, see Shavell (2007) and the references therein. 0144-8188/$ – see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.irle.2008.02.006
tion increases the likelihood of cancer. Suppose now that the firm is brought to court by a widow who’s husband was treated with the medication and later on died of cancer. Should, in this example, the full extent of strict liability apply or should courts allow for the possibility of a defense. Indeed, the firm may have been extremely conscientious in doing its research and have included all the scientific knowledge available at the time. It would be questionable as to whether it would be economically and logically sound, in such a case, to make the pharmaceutical firm liable. Instead of increasing producers’ care, could it well reduce overall research activities? Some authors have pointed out such a possible drawback, suggesting that “a broad and unpredictable sweep of liability” could deter innovation. For example, Porter (1990) states that “product liability in the United States is so extreme and uncertain as to retard innovation, it places firms in constant jeopardy of costly and lengthy product liability suits”. He finds the risk of lawsuits is so great, and the consequences so potentially disastrous, that the inevitable result is to exercise more caution in product innovation than in other advanced nations. However, it could be possible to argue that authorizing any form of defense in a producer’s liability case undermines the strict nature of the liability rule, and must consequently be rejected. The two arguments naturally lead to the question as to when a court should deem a defense of state-of-the-art to be reasonable and when it should conclude that a defendant was liable? In practice, product liability cases usually focus on the competing experts debating as to whether the design of the product is sufficiently safe. Not surprisingly, a standard argument brought forward by defense lawyers is that the product of their client and the safety thereof is very similar to competing products by other manufacturers. In this respect, Boyd and Ingberman (1997) suggest two different approaches to this line of defense. According to the first one, firms should avoid liability, if at the time of production the safety of their fabrication process compared favorably to the customary practices in the industry. With this interpretation, the state-of-the-
B. Deffains, D. Demougin / International Review of Law and Economics 28 (2008) 106–112
art defense is a customary practice test, i.e. a relative test of liability (rather than an absolute one). It is almost exclusively designed upon a comparison between the characteristics of the defendant’s production process and the characteristics of its competitors’ process. According to the second interpretation, state-of-the-art should only refer to a more exceptional safety level. This could be interpreted as, demanding of the defendant, to have been at the forefront of technical advancement or capability. We refer to this second criterion as the technological advancement test. Under the first definition, a state-of-the-art production process is one which conforms to the customary practice of the industry. Under the second, simply conforming with custom directly invalidates a state-of-the-art defense. It holds defendants to a more stringent test than simply conformance with industry custom. It does not require the defendant to have the safest technology, but to be significantly better than the average.2 How does the law deal with emerging standards and with customary practices? The development of law is an evolutionary process. What we know from previous court decisions can be creatively adapted to provide answers when new legal issues are encountered. The law has recognized from ancient times that accepting customary practices in commercial disputes is essential to making fair decisions about what should have happened and who is at fault. Customs are generally recognized by the courts only when certain requirements are established to the satisfaction of the court such as the notoriety of practice, the reasonableness of the practice, the publication of the practice and the proof of the practice.3 The jurisprudence is still small, so it is important to develop an economic analysis of how to implement the stateof-the-art defense, comparing technological advancement test and customary practices as we propose in the paper. Put in the context of the economic literature, the state-of-theart defense suggests a form of ex post contest/tournament (see e.g. Lazear & Rosen, 1981) where the safety procedure of the defending firm is compared with a measure of the average practice from a pool of other firms engaged in the same activity. At the time of research, development and production, forward looking firms will anticipate what would happen in the case of accident. Consequently, the ex post contest generates an incentive for undertaking safety activities. In terms of game theory, the state-of-the-art defense induces a Nash equilibrium in the level of care. In some respects, our approach is reminiscent of the analysis of the negligence rule. In a way, it constitutes a new way to analyze
2 In practice, with the two tests, it could be possible to consider that liability is not completely removed when all firms have the same technology. Instead, when all firms seem identically safe, there is a probability that a plaintiff can successfully argue that this customary practice in the industry is, in fact, unreasonably unsafe. This explains why a “race to the bottom” is not a convincing equilibrium as long as courts impose a reasonableness test on the industry custom. The definition of a minimum standard by the regulator can also contribute to control the strategy of the firms. 3 The traditional arguments for adopting an established methodology include:
• Adoption of a standard reduces risks because of the intellectual investment that has gone into developing the standard. • Adoption of a standard enables benchmarking of the project results. If questions are asked later the skill of the experts can be compared against the results in other projects that have adopted the same methodology. • Following a standard facilitates communication. • Adoption of a standard leverages the experience of others through adopting a methodology that others are using at the same time. However, the growing complexity of products and technologies generate new legal problems and impose to find new solutions. The recent debate about the precautionary principle illustrates the debate. In a way, this principle could be interpreted as a technological advancement test.
107
the functioning of liability rules. Usually, when one considers the definition of a “standard of care”, it refers to an absolute level of care (for example the bonus paterfamilias or the Hand formula). Here, our approach combines a traditional model of tort liability with a non-cooperative contest in which, the higher a firm’s level of care compares to that of the other players, the lower are the ensuing expected liability costs.4 Designing the liability rule means that the judicial system needs to determine the number of players to consider, the nature of the state-of-the-art defense, to determine by sample analysis the average level of care and to establish individual liability cases. With this terminology in mind, the question of the appropriate state-of-the-art defense can be reformulated as to whether the contest should be “fair” or “biased”. The question can be generalized is there a contest that is universally optimal or should the state-of-the-art defense depend on some characteristics of the case? If so, what is the dependence? For example, in some cases courts will be able to compare the firm’s behavior to a large pool of other firms. In other situations, perhaps because evaluating care is extremely costly, the pool will be much smaller. Should this affect the way state-of-the-art defense is implemented by courts? For example, should the choice between the “technological advancement” and “customary practice” test depend on the number of firms the defendant is compared with? If so, in which way? A second issue is the level of damage. Our first intuition would suggest that, as damages increase, courts should increase the bias to provide additional incentives since the efficient level of care has been raised. As we will see the intuition is not correct. We derive other unexpected results. For example, we show that an ex post contest with a negative bias may also implement efficient care. This means that a state-of-the-art defense implementing efficient care may compare a firm’s safety level with that of other firms and declare the firm non-liable, even if it produced only 90% of the safety of other firms. To analyze the issue,5 we consider an industry with a large number of firms which are involved in a similar type of research and development project.6 In the case of an accident and ensuing liability suit, courts are assumed to see a noisy signal of the defendant’s care expenditure and similar signals for a number of competing firms. We then assume that courts can decide allocation of damages by comparing these signals. We ask how the court should best proceed? In particular, whether the comparison should be biased or not and, if so, in which direction? From the point of view of minimizing primary costs, we find that there are many ways to define the state-of-the-art defense. However, integrating secondary and tertiary costs consideration, we conclude in favor of the technological advancement test. The remainder of the paper is structured as follows. The next section introduces a model. Section 3 derives the main result. Section 4 develops and discusses comparative static results. The last section offers some concluding remarks.
4 In a way, the standard is defined in relative terms: When firms face a state-ofthe-art defense, they perceive that they can influence the standard defined by the courts and so, that the standard becomes endogenous. This is true with a customary practice test as well as with a technological advancement test. 5 Notice that our analysis is normative given the fact that courts have to apply a state-of-the-art defense. Whatever the liability system is with or without fault is not the problem. Of course, as most countries have currently a system of strict product liability, the model considers the conditions of application of the defense under this regime. 6 The reason for assuming a large number of firms is just to avoid a cooperative agreement by the firms to keep the level of care low, thereby “forcing” them to play non-cooperatively.
108
B. Deffains, D. Demougin / International Review of Law and Economics 28 (2008) 106–112
2. The model We consider a market with a large number of identical firms all undertaking a similar research and development project, leading to the marketeering of a new product. We denote by c the firm’s expenditures of designing and developing the product safely. Slightly abusing terminology, we refer to these expenditures, hereafter, as the level of care undertaken by the firm. Given care, we denote by p(c) the probability of an accident, with p (c) < 0 and p (c) > 0.7 In case of an accident, the victim suffers the monetary equivalent of a damage L. We assume that L and the function p(c) are known by the firms at the time of decision. If the case goes to trial, the court observes the damage L. In line with the objective of the paper, the court cannot evaluate the level of due care directly.8 Instead, it estimates what the firm should have done, by comparing its behavior to some of its competitors. In a trial, the court can observe a noisy signal of the level of care undertaken by the defending firm. Specifically, we assume that the court sees the monetary signal, x = c + ε, of the firm’s research expenditure.9 For the sake of comparison, the court also collects additional information on the research behavior of competitors of the defendant. Specifically, the court sees a vector (x1 , . . . , xN ) where the variables xi = ci + εi , i = 1, . . . , N are monetary signals of the research expenditure undertaken by the defendant’s competitor i. In the remaining, N will play a significant role. It denotes the sample size of firms to which the defendant will be compared. More generally, it stands for the precision or accuracy which the court will demand of its experts. From the point of view of the judicial system, N, and the way it is decided upon, are decision variables; either it could be externally determined through a legal rule, left to the judge’s or the parties’ discretion. We assume that the error terms in all the respective signals are identically, independently and symmetrically distributed around their mean according to the distribution G(ε). Moreover, we assume that for a given c, the error term is independent of the likelihood of an accident (i.e. a change in c only shifts the mean of the distribution of the x signals, but otherwise leaves the shape of the density function unaffected). We denote respectively by X¯ N and ε¯ N the average over the sample vector (x1 , . . . , xN ) and the average error term. To model the state-of-the-art defense, we suppose that in the case of a lawsuit in the environment described above, courts are asked to compare the defendant’s realization of x with X¯ N . Using this notation, both state-of-the-art defenses under studies can be represented in a similar way by deeming the defendant liable if x ≤ X¯ N + , where varies according to the liability rule.10 Intuitively,
7 Implicitly, all we are requiring is that producers are correctly minimizing costs. For instance, suppose that y1 , . . . , yk are research inputs with associated prices w1 , . . . , wk ; then, the firm’s cost minimization objective becomes
p(c) = min Pr(accident|y1 , . . . , yk )
k
s.t.
y1 ,...,yk
wi yi ≤ c.
captures whether society should expect the defendant to have produced an exceptional effort ( large, i.e. the court is using a technological advancement test) or simply to have conformed to the customary practice of the industry (i.e. = 0).11 3. The firm’s decision problem Anticipating the behavior of courts, a firm’s safety decision can be represented as a cost minimization problem which trades off care costs against expected damage payments. The liability regime induces a Nash game between the firms that implicitly defines, for the case of an accident, the likelihood of failing with a state-of-theart defense. Considering the usual practice in civil liability systems, payments made by the defendant if found liable are set equal to the incurred damage costs L. Altogether, each firm solves min c + p(c)Pr[x ≤ X¯ N + |c, C¯ N ]L, c
(1)
where C¯ N represents the average care level of competitors and Pr[x ≤ X¯ N + |c, C¯ N ] denotes the probability that, in the case of a damage, the firm fails with the state-of-the-art defense. We limit the analysis to the symmetric Nash equilibrium, i.e. to the case where all the firms find it optimal to undertake the same care level denoted c E hereafter. Accordingly, at the Nash equilibrium the average care level of the defendant’s competitors also satisfies C¯ N = c E . Given the foregoing distributional assumptions, the probability of a defendant being found liable, as function of his care level c and the average care level of his competitors c E , becomes Pr[x ≤ X¯ N + |c, c E ] = Pr[ N ≤ c E − c + ], N
(2)
ε − ε¯ N .
= We denote the cumulative distribution funcwhere tion of N by F( N |N) and the density function is f ( N |N). The cdf and pdf of N can both be derived from G(ε) hereafter with F( N |N) and f ( N |N). The foregoing assumptions on the error terms imply that the expected value of N must be zero. In addition, it is easily verified that 2N = (1 + N −1 )ε2 . Naturally, the sample
size of competitors to which the defendant is compared reduces the variance of N .12 We impose two further restrictions: (i) f (·|N) is single peaked and, (ii) F(·|N) satisfies the MLRC property (i.e. (d/d N )[f ( N |N)/(1 − F( N |N))] > 0).13 Altogether, the firm’s care decision problem becomes min C F (c, c E , , N) = c + p(c)F(c E − c + |N)L, c
(3)
where C F (c, c E , , N) stands for the expected costs of a firm undertaking care c in a judicial environment where the state-of-the-art defense is characterized by and N and the firm’s competitors produce average care c E . The symmetric Nash equilibrium is defined when c E is the solution of (3) for each firm. As a result, c E is implicitly characterized by the first-order condition of the firm’s cost minimization problem, i.e.14 : 1 + [p (c E )F(|N) − p(c E )f (|N)]L = 0.
(4)
i=1
8 It is important to note that the benchmark where all the c are observerable (or when the risk of error converges to zero) cannot be considered as a good representation of reality. Following Boyd and Ingberman (1997), it leads to a knife-edge result : if is positive, the technological advancement test yields excessive safety expenditure. If ≤ 0, the “customary practice test creates a race to the bottom in which the equilibrium is to spend nothing on safety” (Boyd & Ingberman, p. 459). 9 The justification of this assumption is that even though courts and experts can observe a firm’s accounts records, the firm can and will most likely use creative book keeping to shift expenditures to its advantage, thereby biasing records. 10 Keep in mind that even though in accordance with legal practice, we consider a strict liability rule with a state-of-the-art-defense, from an analytical point of view, it is also possible to interpret the model in terms of the negligence rule.
11 With < 0, the legal rule would theoretically demand in some cases that courts not to declare the defendant liable, even if the comparison with competitors was unfavorable to him. 12 Geometrically, as the sample size increases, the distribution of N centers more and more around its mean. Note however that even with very large sample, the firm cannot avoid the risk created by its own error term ε. 13 See Milgrom (1981). 14 Observe that the second-order condition of the firm’s problem (p F − 2p f ) + pf is always satisfied for ≤ 0 (note that the terms in brackets are both positive). However, for some > 0 the second-order condition may not be satisfied. This limits the choices of .
B. Deffains, D. Demougin / International Review of Law and Economics 28 (2008) 106–112
109
The condition shows that just as one would expect the firms’ behavior will react both to changes in and N. In particular, implicitly differentiating with respect to yields: ∂c E (p (c E )f (|N) − p(c E )f (|N))L =− , F (c E , c E , , N) ∂ Ccc
(5)
F stands for the second-order condition of the firm’s where Ccc F is necproblem. Since the firm is minimizing expected costs, Ccc essarily positive. Consequently, for ≤ 0, the Nash level of care c E is strictly increasing in . To see this, observe that for ≤ 0, we have f (|N) > 0. Moreover, p (c E ) < 0 and p(c E ), f (|N) > 0, thus in this case, ∂c E /∂ becomes positive. What does this result suggest? Intuitively if = 0, the prisoners’ dilemma implicit in the game between firms and the possibility that a firm may in fact escape liability should induce an underproduction of care. To counteract this natural tendency, one would think that society should direct its courts to behave in such a way as to increase the safety produced by firms. In this respect, biasing the contest by demanding that firms’ care level be above the average may appear reasonable since, according to (5), it would induce firms to increase care at the Nash equilibrium. Accordingly, it would to appear that to minimize primary costs the state-of-the-art defense should be in terms of an exceptional safety effort, thus, requiring a technological advancement test (Boyd & Ingberman, 1997).
4. The regulator’s problem Though the above logic seems intuitively compelling, its heuristic is in fact erroneous. In order to detect the underlying inconsistency, observe that the parties do not minimize primary costs c + p(c)L. Instead, firms minimize their own private cost c + p(c)F(c E − c + |N)L. This distortion has two effects. First, it has a level effect by reducing the likelihood of paying for harm. Second, it has a slope effect by changing the shape of the expected damage cost curve. Our intuition seems to mainly capture the first effect, thereby suggesting that firms should underproduce care. However, the second effect is determinant for the firm’s care decision. From the foregoing section, for the legal rule to have any impact, firms must anticipate the consequence of their choice on expected costs. Thus, the judicial system has to decide the characteristics of the state-of-the-art defense and announce them. In our model, it requires to determine the bias level and the sample size N. In the following, we verify that there are many (, N) combinations inducing firms to undertake the socially efficient care level as Nash equilibrium. Analytically, the first-best solution is defined as the level of care that minimizes the true social costs, i.e. c ∗ = argminc + p(c)L.
(6)
c
Writing the first-order condition of (6) and comparing it with (4) at the point c E = c ∗ yields:
∗
∗
∗
p (c ) = p (c )F(|N) − p(c )f (|N).
(7)
Therefore, any combination of and N that satisfies (7) will in fact induce firms to produce the first-best level of care at the Nash equilibrium.15 For a geometrical representation, it is useful to rewrite (7) in the following way: −p (c ∗ ) f (|N) . = p(c ∗ ) 1 − F(|N)
(8)
In the Fig. 1, N is exogenously given. The upward sloping curve is given by the right-hand side of (8). It is monotonically increasing
Fig. 1. Socially efficient levels of care.
due to the MLRC assumption. The left-hand side of (8) is positive due to p < 0. In the case represented in the figure, a solution in exists.16 Whether (8) has a solution, is a question of the characteristic of the distribution function. For example, in the case of the normal distribution the f (|N)/(1 − F(|N))-curve converges to 0 as → −∞ and becomes unbounded as → +∞. In that case, existence is guaranteed. Moreover, given N, the solution is unique. Changes in N shift the f (|N)/(1 − F(|N))-curve thereby affecting the bias which implements care c ∗ . Consequently, there are many pairs (∗ (N), N) that implement the first-best level of care. The foregoing suggests two initial comments. First, a sole focus on an analysis of biases, as suggested by the comparison between customary and technological advancement, is too limited to correctly evaluate the effect of a state-of-the-art defense. In particular, depending on the specific context, it is possible that a number of tests could accomplish the task. Second, each of the tests require a different appropriate sample size. In the foregoing, we identified the customary test with = 0. It will induce firms to produce first-best care if the sample size, hereafter denoted by N 0 , satisfies17 : f (0|N 0 ) −p (c ∗ ) . = ∗ p(c ) 1 − F(0|N 0 )
Keeping in mind that the distribution of the N is symmetrical around zero, F(0|N) = 1/2 for all N. Therefore, the right-hand side of (9) simplifies to 2f (0|N 0 ). Whether the equation has a solution in N, depends on the exact nature of the N distribution. For example, if the underlying ε are normally distributed, it is easily verified that (9) becomes √ −p (c ∗ ) 2 . (10) = p(c ∗ ) (1 + (1/N 0 )) Since for all N, 0 ≤ (1/N) ≤ 1, (10) has a solution provided that the condition √ −p (c ∗ ) 1 2 (11) ≤ √ ≤ √ p(c ∗ ) is satisfied. Assuming the above condition is satisfied, a customary test with sample size N 0 would implement the first-best. Naturally, a question arises as to whether a technologically advancement test could also induce first-best at the Nash equilibrium. To discuss the issue, consider decreasing N below N 0 . For = 0, note that f (0|N) < f (0|N 0 ) (intuitively reducing N increases
16 15
Assuming the second-order condition holds. See the restrictions in footnote 14.
(9)
17
Obviously assuming the second-order condition holds. Observe that with = 0, the second-order condition is certainly satisfied.
110
B. Deffains, D. Demougin / International Review of Law and Economics 28 (2008) 106–112
sect the y-axis above the f (|N ∗ )/(1 − F(|N ∗ ))-curve. Therefore, implementing the first-best would require to set < 0. In words, the legal rule would demand courts not to declare the defendant liable, even if the comparison with the large sample was (marginally) unfavorable to him. The literature, with respect to the state-of-the-art defense, has focused on the comparison between the customary practice test and the technological advancement test in terms of the bias. In doing so, it has neglected an important dimension. Indeed, for the comparison to be meaningful, it must include a discussion on the sample size associated with the respective tests. The foregoing analysis shows that there is a trade-off between the bias introduced by the courts to establish the state-of-the-art defense, and the sample size of producers which the defendant is compared with. In particular, the technological advancement test necessitates a smaller sample size than the customary practice test. The analysis also suggests, that the comparison should not be solely focussed on primary costs, since it would appear that many combinations of bias and sample size implementing first-best care. There are other aspects based on secondary and tertiary costs that influence the comparison between the two legal practices under study. We list three arguments that speak in favor of a large bias and a small sample size N:
Fig. 2. Bias and sample size.
Fig. 3. Social cost and sample size.
the variance of N , thus, flattens density function of N around its mean). Therefore the f (|N)/(1 − F(|N))-curve intersects the y-axis below the f (|N 0 )/(1 − F(|N 0 ))-curve. Consequently, lowering the sample size below N 0 would require courts to raise the bias or, according to our interpretation of , to use a technological advancement test (see Fig. 2).18 There is another way to understand the result. N 0 was originally defined to satisfy the firm’s first-order condition at c ∗ with set at 0: 1 + [p (c ∗ )F(0|N 0 ) − p(c ∗ )f (0|N 0 )]L = 0.
(12) N0
Consider the effect of an increase in N above in (12). The first term in the square bracket would remain unchanged, since F(0|N) = 1/2 for all N. However, the second term drops, so that the lefthand side of (12) becomes positive. This, in turn, implies that a firm facing a legal rule characterized by = 0 and N > N 0 would find it optimal to increase its care level above c ∗ . Moreover, note that c ∗ + p(c ∗ )F(0|N)L is constant for all N. Thus, the two cost curves must intersect at c ∗ as in Fig. 3. In order to reestablish equality in (7) requires raising . Since F(|N) > 1/2 = F(0|N 0 ), it means that in the case of damages firms are found more often liable thereby increasing expected costs. The observation points to a surprising peculiarity for legal policy. Suppose that, instead of reducing N below N 0 , we were to consider an increase in the sample size. Again, this would shift the f (|N)/(1 − F(|N))-curve. However, now the curve would inter-
18 In the foregoing, the second-order condition is ignored. As we discussed earlier this may limit the size of the feasible and indirectly may provide a lower bound for feasible sample sizes.
• First, sampling requires collecting and treating information of different producers’ industrial practices. This is a costly activity (note that from a normative perspective, it is irrelevant who carries these costs).19 Keeping these costs at a minimum, speaks in favor of a small sample and, therefore, a large , i.e. a technological advancement test. • Second, evaluating liability rules also demands considering the impact of damages adjudication on the parties’ activity levels. In the current situation, we saw earlier, that the firms’ expected private costs are smaller than the true social costs, c + p(c)F(|N)L < c + p(c)L. Consequently, the activity level of the defendant will be excessive, compared to the social optimum. In this respect, note that with the customary test F(0|N) = 1/2, whereas with a technological advancement test F((N)|N) > 1/2. Thus this second consideration also speaks in favor of defining a defense, based on a state-of-the-art argument, using the technological advancement test. • Third, if one considers the insurance mechanism inherent to civil liability claims, it seems interesting to remark, that victims will also be better off under a state-of-the-art defense, based on a technological advancement test. The explanation parallels the foregoing argument; with an increase in the Nash equilibrium value of F(.) firms pay damages more frequently. In other words, when state-of-the-art refers to an exceptional safety level, it is more difficult for the defendant to escape liability, being at the forefront of technical advancement. Therefore, partisans of tort liability, as a mean to compensate parties who suffer losses, should prefer a state-of-the-art defense with F(.) as close as possible to unity. Under the above considerations, the state-of-the-art defense should be defined in terms of the technological advancement test with a large bias and a small sample size. The foregoing conclusion must be qualified however since it would necessitate that society find a way to limit the sample size. In practice, this may be prove to be rather difficult in an adversarial environment and in the presence
19 It is important to note that the ability of the judge to control N is probably different under inquisitorial and adversarial procedures. A discussion of this point goes beyond the scope of this paper.
B. Deffains, D. Demougin / International Review of Law and Economics 28 (2008) 106–112
of competing experts whose aims are not to produce an efficient rule, but to represent the interest of their respective party.. 5. Discussion and concluding remarks This paper develops a modelisation of the state-of-the-art defense. From a normative perspective, our analysis could constitute a basis to guide courts applying a state-of-the-art defense. More specifically, on the trade-off between and N, it seems possible to find arguments in favor of the technological advancement test. However, the argument necessitates courts to clearly understand the relationship between the bias and the sample size. In practice, most judicial systems permit some defense based on a state-of-the-art argument. For example, the U.S. restatement Second of Tort § 402 A (1965) defining the products liability principles, permits such a defense requiring a “demonstration that the technology available for the manufacture of a safer finished product with the same characteristics was not feasible”. However, as we documented with the Cunningham v. MacNeal Memorial Hospital case, the state-of-the-art defense remains controversial. In this case, the Illinois Supreme Court rejected the argument that blood products, then in use, complied with the state-of-the-art in terms of safety from hepatitis contamination. The Court stated that “to allow a defense to strict liability on the ground that there is no way, either practical or theoretical, for a defendant to ascertain the existence of impurities. . . would emasculate the doctrine. . . and signal a return to a negligence theory” (Cunningham v. Memorial Hosp. 266 N.E.2d 897 at 902 (1970)). Moreover, the application of the defense raises numerous questions. For instance, when specifically should the defense be applied? In the case O’Brien v. Muskin Corp., the New Jersey Supreme Court, while recognizing the value of evidence as to the state-of-the-art in determining product defect and in applying a risk-utility analysis, held that the defense is not absolute (O’Brien v. Muskin Corp., 463 A.2d, 306 (N.J. 1983)). Further, in Beshada v. Johns Manville, the New Jersey Supreme Court expressly disallowed the state-of-the-art defense to the issue of failure to warn (Beshada v. Johns-Manville Corp., 447 A 2d 539 (N.J. 1982)). These cases concern asbestos litigations that have frequently considered the evolution of scientific knowledge. The same reasoning is generally applied in medical malpractices cases. However, the jurisprudence also refers to customary practice. For instance, the CERCLA imposes joint and several liability for cleaning up contamination caused by hazardous substances of responsible parties. One of the defense allowed is for the defendant to provide he had no reason to know of the existence of hazardous substances and made all “appropriate inquiry into the previous owner and uses of the property consistent with good commercial or customary practice.” 42 U.S.C. sec. 9601. We observe a similar debate in the European context. The council directive 85/374/EEC concerning the liability for defective products states that a producer will not be liable if the producer proves that the state of scientific and technical knowledge at the time the producer placed the product on the market was not such as to enable the defect to be discovered. The jurisprudence is still limited, but it appears, for instance, that courts’ decisions concerning liability of pharmaceutical firms adopt a definition based on scientific knowledge to evaluate the producer’s liability. In these cases, the definition of the state-of-the-art considers the technological advancement of the industry as expressed in scientific publications. The paper demonstrates that the two interpretations of the state-of-the-art defense in civil liability claims, the customary practice test and the technological advancement test, could be – contrary to the intuition – equivalent in terms of firms’ efforts to promote safer technologies. In analytical terms, the reason is that, a tradeoff exists between the bias introduced by the courts to establish
111
the state-of-the-art, relatively to the average level of care and the size of the sample. Of course, it is difficult to determine the optimal combination. However, if the observation of technical characteristics of an industry is difficult for the courts, implying costly and long experts’ investigations, it could be better to refer to small samples. In this case, a normative analysis would conclude in favor of the technological advancement test because it does not require to collect exhaustive information about firms. Moreover, this system presents advantages in terms of injurers’ activity levels and of victims’ compensation. Considering the fear expressed by Porter (1990) that product liability becomes so extreme and uncertain as to retard innovation, it appears that the problem is not the evolution towards strict liability. It is essentially explained by the US judicial environment: importance of pain and suffering awards, punitive damages, role of civil juries, contingent fees for lawyers, class actions and the standard of proof based on the preponderance of evidence. The state-of-the-art defense is generally discussed under a strict liability rule. The context which has influenced this choice is, of course, the one of product liability in most of countries. However, our main conclusion concerns liability rules in general. In a sense, the problem of the state-of-the-art defense is similar to the definition of a due care standard under a negligence rule. The problem could easily be extended this way, because a due care standard could be defined in reference to the current practices of people or to an exceptional level of effort exerted by them. Finally, our analysis demonstrates that there are theoretically many pairs (∗ (N), N) that can implement the first-best. However, from a practical point of view, we have seen that the interest of the society could be to define the largest possible value for F(.). This implies a small N and an important bias under a technological advancement test. In this case, the solution will tend towards a strict liability system. So, the position of the Illinois Supreme Court in Cunningham v. MacNeal Memorial Hospital about the risk of “emasculation of the doctrine” could be well justified on economic grounds. Another important practical consideration concerns the way firms are selected to be included in the sample. In the model, we have implicitly supposed that they are randomly chosen. However, this is not necessarily true. In practice, an injurer found liable based on the signals of some competitors (chosen by the court or by the expert) may sometimes want to present alternative signals based on other competitors that are more favorable to his case. This constitutes a natural extension of the model beyond the scope of our analysis. In addition, the nature of the procedures should also be studied. For example, our analysis presupposes that the court controls the procedures and, in particular, the way experts select the sample. In that respect, there is certainly a significant difference between common law and civil law countries. Acknowledgement The authors would like to thank two anonymous referees for comments on an earlier version of this paper. References Boyd, J., & Ingberman, D. A. (1997). Should “Relative Safety” be a test of product liability? Journal of Legal Studies, 26(2), 433–473. Brown, J. P. (1974). Product liability: The case of an asset with random life. American Economic Review, 64(1), 149–161. Calfee, R., & Craswell, J. E. (1986). Deterrence and uncertain legal standards (with John E. Calfee). Journal of Law Economics & Organization, 2(2), 279–303. Hylton, K. (2000). The theory of tort doctrine and the restatement of torts. Boston University School of Law Research Paper No. 07. Available at SSRN: http://ssrn.com/abstract=242925 or doi:10.2139/ssrn.242925. Landes, W. M., & Posner, R. (1981). The positive economic theory of tort law. Georgia Law Review, 15, 851–924. Landes, W. M., & Posner, R. (1987). The economic structure of tort law. Cambridge University Press.
112
B. Deffains, D. Demougin / International Review of Law and Economics 28 (2008) 106–112
Lazear, E., & Rosen, S. (1981). Rank order tournaments as optimum labor contracts. Journal of Political Economy, 89, 841–861. Milgrom, P. R. (1981). Good news and bad news: Representation theorems and applications. Bell Journal of Economics, 12(2), 380–391. Polinsky, A. M. (1980). Strict liability vs. negligence in a market setting. American Economic Review, 70, 363–370. Porter, M. E. (1990). The competitive advantage of nations. The Free Press.
Shavell, S. (1980). Strict liability versus negligence. Journal of Legal Studies, 9, 1–25. Shavell, S. (2007). Economic analysis of accident law (new edition). Harvard University Press. Spence, M. (1977). Consumer misperceptions, product failure and producer liability. Review of Economic Studies, 44, 561–572. Viscusi, W. K., & Moore, M. J. (1993). Product liability, research and development, and innovation. Journal of Political Economy, 101(1), 161–184.
Contents lists available at ScienceDirect
International Review of Law and Economics
Customary versus technological advancement tests Bruno Deffains a,∗ , Dominique Demougin b a b
Economics University, Paris 10 and CNRS, France European Business School, 65201 Wiesbaden, Germany
a r t i c l e
i n f o
Keywords: Tort law Standard of care Customary test Technological advancement test
a b s t r a c t In an environment where the optimal level of care is unknown, we ask, under a state-of-the-art defense, which method is better able to induce parties to undertake optimal care. Assuming courts can see a noisy signal of research activities undertaken by a defendant and by some of its competitors, we ask whether courts should use a biased or an unbiased average to compare care. We find that the latter is advantageous. © 2008 Elsevier Inc. All rights reserved.
1. Introduction It is well known that over the course of the last century producers’ liability has become an important topic in most developed economies. The economic justifications for the evolution of producer tort are numerous and well-documented. From a positive perspective, products sophistication has been advanced by Landes and Posner (1987) as well as the ability of the consumer to be correctly informed on the risk of accident (Brown, 1974; Spence, 1977; Viscusi & Moore, 1993). Law and economics scholars have also argued about the excessive activity level under a negligence rule (Calfee & Craswell, 1986; Landes & Posner, 1981; Polinsky, 1980; Shavell, 1980). Moreover, from the point of view of legal theory, some authors propose to justify producer’s liability based upon the risk represented by the consumption of products rather than on the behavior of the producers (for a general discussion see Hylton, 2000). In the same way, the second part of the 20th century has been characterized by a growing acceptance of the notion that more extensive tort liability would serve to compensate parties as an insurance mechanism.1 Nevertheless, some issues remain controversial. One such question is under which circumstances a producer might be able to defend himself? The purpose of this article is to explore the welfare characteristics of a defense based on a state-of-the-art argument. To fix ideas, consider the following fictitious example of assessing the liability of a producer. A pharmaceutical firm has designed and marketed a medication to reduce cholesterol levels. Fifteen years later, a researcher team discovers that this particular medica-
∗ Corresponding author. E-mail addresses: [email protected] (B. Deffains), [email protected] (D. Demougin). 1 For a general discussion of liability rules, see Shavell (2007) and the references therein. 0144-8188/$ – see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.irle.2008.02.006
tion increases the likelihood of cancer. Suppose now that the firm is brought to court by a widow who’s husband was treated with the medication and later on died of cancer. Should, in this example, the full extent of strict liability apply or should courts allow for the possibility of a defense. Indeed, the firm may have been extremely conscientious in doing its research and have included all the scientific knowledge available at the time. It would be questionable as to whether it would be economically and logically sound, in such a case, to make the pharmaceutical firm liable. Instead of increasing producers’ care, could it well reduce overall research activities? Some authors have pointed out such a possible drawback, suggesting that “a broad and unpredictable sweep of liability” could deter innovation. For example, Porter (1990) states that “product liability in the United States is so extreme and uncertain as to retard innovation, it places firms in constant jeopardy of costly and lengthy product liability suits”. He finds the risk of lawsuits is so great, and the consequences so potentially disastrous, that the inevitable result is to exercise more caution in product innovation than in other advanced nations. However, it could be possible to argue that authorizing any form of defense in a producer’s liability case undermines the strict nature of the liability rule, and must consequently be rejected. The two arguments naturally lead to the question as to when a court should deem a defense of state-of-the-art to be reasonable and when it should conclude that a defendant was liable? In practice, product liability cases usually focus on the competing experts debating as to whether the design of the product is sufficiently safe. Not surprisingly, a standard argument brought forward by defense lawyers is that the product of their client and the safety thereof is very similar to competing products by other manufacturers. In this respect, Boyd and Ingberman (1997) suggest two different approaches to this line of defense. According to the first one, firms should avoid liability, if at the time of production the safety of their fabrication process compared favorably to the customary practices in the industry. With this interpretation, the state-of-the-
B. Deffains, D. Demougin / International Review of Law and Economics 28 (2008) 106–112
art defense is a customary practice test, i.e. a relative test of liability (rather than an absolute one). It is almost exclusively designed upon a comparison between the characteristics of the defendant’s production process and the characteristics of its competitors’ process. According to the second interpretation, state-of-the-art should only refer to a more exceptional safety level. This could be interpreted as, demanding of the defendant, to have been at the forefront of technical advancement or capability. We refer to this second criterion as the technological advancement test. Under the first definition, a state-of-the-art production process is one which conforms to the customary practice of the industry. Under the second, simply conforming with custom directly invalidates a state-of-the-art defense. It holds defendants to a more stringent test than simply conformance with industry custom. It does not require the defendant to have the safest technology, but to be significantly better than the average.2 How does the law deal with emerging standards and with customary practices? The development of law is an evolutionary process. What we know from previous court decisions can be creatively adapted to provide answers when new legal issues are encountered. The law has recognized from ancient times that accepting customary practices in commercial disputes is essential to making fair decisions about what should have happened and who is at fault. Customs are generally recognized by the courts only when certain requirements are established to the satisfaction of the court such as the notoriety of practice, the reasonableness of the practice, the publication of the practice and the proof of the practice.3 The jurisprudence is still small, so it is important to develop an economic analysis of how to implement the stateof-the-art defense, comparing technological advancement test and customary practices as we propose in the paper. Put in the context of the economic literature, the state-of-theart defense suggests a form of ex post contest/tournament (see e.g. Lazear & Rosen, 1981) where the safety procedure of the defending firm is compared with a measure of the average practice from a pool of other firms engaged in the same activity. At the time of research, development and production, forward looking firms will anticipate what would happen in the case of accident. Consequently, the ex post contest generates an incentive for undertaking safety activities. In terms of game theory, the state-of-the-art defense induces a Nash equilibrium in the level of care. In some respects, our approach is reminiscent of the analysis of the negligence rule. In a way, it constitutes a new way to analyze
2 In practice, with the two tests, it could be possible to consider that liability is not completely removed when all firms have the same technology. Instead, when all firms seem identically safe, there is a probability that a plaintiff can successfully argue that this customary practice in the industry is, in fact, unreasonably unsafe. This explains why a “race to the bottom” is not a convincing equilibrium as long as courts impose a reasonableness test on the industry custom. The definition of a minimum standard by the regulator can also contribute to control the strategy of the firms. 3 The traditional arguments for adopting an established methodology include:
• Adoption of a standard reduces risks because of the intellectual investment that has gone into developing the standard. • Adoption of a standard enables benchmarking of the project results. If questions are asked later the skill of the experts can be compared against the results in other projects that have adopted the same methodology. • Following a standard facilitates communication. • Adoption of a standard leverages the experience of others through adopting a methodology that others are using at the same time. However, the growing complexity of products and technologies generate new legal problems and impose to find new solutions. The recent debate about the precautionary principle illustrates the debate. In a way, this principle could be interpreted as a technological advancement test.
107
the functioning of liability rules. Usually, when one considers the definition of a “standard of care”, it refers to an absolute level of care (for example the bonus paterfamilias or the Hand formula). Here, our approach combines a traditional model of tort liability with a non-cooperative contest in which, the higher a firm’s level of care compares to that of the other players, the lower are the ensuing expected liability costs.4 Designing the liability rule means that the judicial system needs to determine the number of players to consider, the nature of the state-of-the-art defense, to determine by sample analysis the average level of care and to establish individual liability cases. With this terminology in mind, the question of the appropriate state-of-the-art defense can be reformulated as to whether the contest should be “fair” or “biased”. The question can be generalized is there a contest that is universally optimal or should the state-of-the-art defense depend on some characteristics of the case? If so, what is the dependence? For example, in some cases courts will be able to compare the firm’s behavior to a large pool of other firms. In other situations, perhaps because evaluating care is extremely costly, the pool will be much smaller. Should this affect the way state-of-the-art defense is implemented by courts? For example, should the choice between the “technological advancement” and “customary practice” test depend on the number of firms the defendant is compared with? If so, in which way? A second issue is the level of damage. Our first intuition would suggest that, as damages increase, courts should increase the bias to provide additional incentives since the efficient level of care has been raised. As we will see the intuition is not correct. We derive other unexpected results. For example, we show that an ex post contest with a negative bias may also implement efficient care. This means that a state-of-the-art defense implementing efficient care may compare a firm’s safety level with that of other firms and declare the firm non-liable, even if it produced only 90% of the safety of other firms. To analyze the issue,5 we consider an industry with a large number of firms which are involved in a similar type of research and development project.6 In the case of an accident and ensuing liability suit, courts are assumed to see a noisy signal of the defendant’s care expenditure and similar signals for a number of competing firms. We then assume that courts can decide allocation of damages by comparing these signals. We ask how the court should best proceed? In particular, whether the comparison should be biased or not and, if so, in which direction? From the point of view of minimizing primary costs, we find that there are many ways to define the state-of-the-art defense. However, integrating secondary and tertiary costs consideration, we conclude in favor of the technological advancement test. The remainder of the paper is structured as follows. The next section introduces a model. Section 3 derives the main result. Section 4 develops and discusses comparative static results. The last section offers some concluding remarks.
4 In a way, the standard is defined in relative terms: When firms face a state-ofthe-art defense, they perceive that they can influence the standard defined by the courts and so, that the standard becomes endogenous. This is true with a customary practice test as well as with a technological advancement test. 5 Notice that our analysis is normative given the fact that courts have to apply a state-of-the-art defense. Whatever the liability system is with or without fault is not the problem. Of course, as most countries have currently a system of strict product liability, the model considers the conditions of application of the defense under this regime. 6 The reason for assuming a large number of firms is just to avoid a cooperative agreement by the firms to keep the level of care low, thereby “forcing” them to play non-cooperatively.
108
B. Deffains, D. Demougin / International Review of Law and Economics 28 (2008) 106–112
2. The model We consider a market with a large number of identical firms all undertaking a similar research and development project, leading to the marketeering of a new product. We denote by c the firm’s expenditures of designing and developing the product safely. Slightly abusing terminology, we refer to these expenditures, hereafter, as the level of care undertaken by the firm. Given care, we denote by p(c) the probability of an accident, with p (c) < 0 and p (c) > 0.7 In case of an accident, the victim suffers the monetary equivalent of a damage L. We assume that L and the function p(c) are known by the firms at the time of decision. If the case goes to trial, the court observes the damage L. In line with the objective of the paper, the court cannot evaluate the level of due care directly.8 Instead, it estimates what the firm should have done, by comparing its behavior to some of its competitors. In a trial, the court can observe a noisy signal of the level of care undertaken by the defending firm. Specifically, we assume that the court sees the monetary signal, x = c + ε, of the firm’s research expenditure.9 For the sake of comparison, the court also collects additional information on the research behavior of competitors of the defendant. Specifically, the court sees a vector (x1 , . . . , xN ) where the variables xi = ci + εi , i = 1, . . . , N are monetary signals of the research expenditure undertaken by the defendant’s competitor i. In the remaining, N will play a significant role. It denotes the sample size of firms to which the defendant will be compared. More generally, it stands for the precision or accuracy which the court will demand of its experts. From the point of view of the judicial system, N, and the way it is decided upon, are decision variables; either it could be externally determined through a legal rule, left to the judge’s or the parties’ discretion. We assume that the error terms in all the respective signals are identically, independently and symmetrically distributed around their mean according to the distribution G(ε). Moreover, we assume that for a given c, the error term is independent of the likelihood of an accident (i.e. a change in c only shifts the mean of the distribution of the x signals, but otherwise leaves the shape of the density function unaffected). We denote respectively by X¯ N and ε¯ N the average over the sample vector (x1 , . . . , xN ) and the average error term. To model the state-of-the-art defense, we suppose that in the case of a lawsuit in the environment described above, courts are asked to compare the defendant’s realization of x with X¯ N . Using this notation, both state-of-the-art defenses under studies can be represented in a similar way by deeming the defendant liable if x ≤ X¯ N + , where varies according to the liability rule.10 Intuitively,
7 Implicitly, all we are requiring is that producers are correctly minimizing costs. For instance, suppose that y1 , . . . , yk are research inputs with associated prices w1 , . . . , wk ; then, the firm’s cost minimization objective becomes
p(c) = min Pr(accident|y1 , . . . , yk )
k
s.t.
y1 ,...,yk
wi yi ≤ c.
captures whether society should expect the defendant to have produced an exceptional effort ( large, i.e. the court is using a technological advancement test) or simply to have conformed to the customary practice of the industry (i.e. = 0).11 3. The firm’s decision problem Anticipating the behavior of courts, a firm’s safety decision can be represented as a cost minimization problem which trades off care costs against expected damage payments. The liability regime induces a Nash game between the firms that implicitly defines, for the case of an accident, the likelihood of failing with a state-of-theart defense. Considering the usual practice in civil liability systems, payments made by the defendant if found liable are set equal to the incurred damage costs L. Altogether, each firm solves min c + p(c)Pr[x ≤ X¯ N + |c, C¯ N ]L, c
(1)
where C¯ N represents the average care level of competitors and Pr[x ≤ X¯ N + |c, C¯ N ] denotes the probability that, in the case of a damage, the firm fails with the state-of-the-art defense. We limit the analysis to the symmetric Nash equilibrium, i.e. to the case where all the firms find it optimal to undertake the same care level denoted c E hereafter. Accordingly, at the Nash equilibrium the average care level of the defendant’s competitors also satisfies C¯ N = c E . Given the foregoing distributional assumptions, the probability of a defendant being found liable, as function of his care level c and the average care level of his competitors c E , becomes Pr[x ≤ X¯ N + |c, c E ] = Pr[ N ≤ c E − c + ], N
(2)
ε − ε¯ N .
= We denote the cumulative distribution funcwhere tion of N by F( N |N) and the density function is f ( N |N). The cdf and pdf of N can both be derived from G(ε) hereafter with F( N |N) and f ( N |N). The foregoing assumptions on the error terms imply that the expected value of N must be zero. In addition, it is easily verified that 2N = (1 + N −1 )ε2 . Naturally, the sample
size of competitors to which the defendant is compared reduces the variance of N .12 We impose two further restrictions: (i) f (·|N) is single peaked and, (ii) F(·|N) satisfies the MLRC property (i.e. (d/d N )[f ( N |N)/(1 − F( N |N))] > 0).13 Altogether, the firm’s care decision problem becomes min C F (c, c E , , N) = c + p(c)F(c E − c + |N)L, c
(3)
where C F (c, c E , , N) stands for the expected costs of a firm undertaking care c in a judicial environment where the state-of-the-art defense is characterized by and N and the firm’s competitors produce average care c E . The symmetric Nash equilibrium is defined when c E is the solution of (3) for each firm. As a result, c E is implicitly characterized by the first-order condition of the firm’s cost minimization problem, i.e.14 : 1 + [p (c E )F(|N) − p(c E )f (|N)]L = 0.
(4)
i=1
8 It is important to note that the benchmark where all the c are observerable (or when the risk of error converges to zero) cannot be considered as a good representation of reality. Following Boyd and Ingberman (1997), it leads to a knife-edge result : if is positive, the technological advancement test yields excessive safety expenditure. If ≤ 0, the “customary practice test creates a race to the bottom in which the equilibrium is to spend nothing on safety” (Boyd & Ingberman, p. 459). 9 The justification of this assumption is that even though courts and experts can observe a firm’s accounts records, the firm can and will most likely use creative book keeping to shift expenditures to its advantage, thereby biasing records. 10 Keep in mind that even though in accordance with legal practice, we consider a strict liability rule with a state-of-the-art-defense, from an analytical point of view, it is also possible to interpret the model in terms of the negligence rule.
11 With < 0, the legal rule would theoretically demand in some cases that courts not to declare the defendant liable, even if the comparison with competitors was unfavorable to him. 12 Geometrically, as the sample size increases, the distribution of N centers more and more around its mean. Note however that even with very large sample, the firm cannot avoid the risk created by its own error term ε. 13 See Milgrom (1981). 14 Observe that the second-order condition of the firm’s problem (p F − 2p f ) + pf is always satisfied for ≤ 0 (note that the terms in brackets are both positive). However, for some > 0 the second-order condition may not be satisfied. This limits the choices of .
B. Deffains, D. Demougin / International Review of Law and Economics 28 (2008) 106–112
109
The condition shows that just as one would expect the firms’ behavior will react both to changes in and N. In particular, implicitly differentiating with respect to yields: ∂c E (p (c E )f (|N) − p(c E )f (|N))L =− , F (c E , c E , , N) ∂ Ccc
(5)
F stands for the second-order condition of the firm’s where Ccc F is necproblem. Since the firm is minimizing expected costs, Ccc essarily positive. Consequently, for ≤ 0, the Nash level of care c E is strictly increasing in . To see this, observe that for ≤ 0, we have f (|N) > 0. Moreover, p (c E ) < 0 and p(c E ), f (|N) > 0, thus in this case, ∂c E /∂ becomes positive. What does this result suggest? Intuitively if = 0, the prisoners’ dilemma implicit in the game between firms and the possibility that a firm may in fact escape liability should induce an underproduction of care. To counteract this natural tendency, one would think that society should direct its courts to behave in such a way as to increase the safety produced by firms. In this respect, biasing the contest by demanding that firms’ care level be above the average may appear reasonable since, according to (5), it would induce firms to increase care at the Nash equilibrium. Accordingly, it would to appear that to minimize primary costs the state-of-the-art defense should be in terms of an exceptional safety effort, thus, requiring a technological advancement test (Boyd & Ingberman, 1997).
4. The regulator’s problem Though the above logic seems intuitively compelling, its heuristic is in fact erroneous. In order to detect the underlying inconsistency, observe that the parties do not minimize primary costs c + p(c)L. Instead, firms minimize their own private cost c + p(c)F(c E − c + |N)L. This distortion has two effects. First, it has a level effect by reducing the likelihood of paying for harm. Second, it has a slope effect by changing the shape of the expected damage cost curve. Our intuition seems to mainly capture the first effect, thereby suggesting that firms should underproduce care. However, the second effect is determinant for the firm’s care decision. From the foregoing section, for the legal rule to have any impact, firms must anticipate the consequence of their choice on expected costs. Thus, the judicial system has to decide the characteristics of the state-of-the-art defense and announce them. In our model, it requires to determine the bias level and the sample size N. In the following, we verify that there are many (, N) combinations inducing firms to undertake the socially efficient care level as Nash equilibrium. Analytically, the first-best solution is defined as the level of care that minimizes the true social costs, i.e. c ∗ = argminc + p(c)L.
(6)
c
Writing the first-order condition of (6) and comparing it with (4) at the point c E = c ∗ yields:
∗
∗
∗
p (c ) = p (c )F(|N) − p(c )f (|N).
(7)
Therefore, any combination of and N that satisfies (7) will in fact induce firms to produce the first-best level of care at the Nash equilibrium.15 For a geometrical representation, it is useful to rewrite (7) in the following way: −p (c ∗ ) f (|N) . = p(c ∗ ) 1 − F(|N)
(8)
In the Fig. 1, N is exogenously given. The upward sloping curve is given by the right-hand side of (8). It is monotonically increasing
Fig. 1. Socially efficient levels of care.
due to the MLRC assumption. The left-hand side of (8) is positive due to p < 0. In the case represented in the figure, a solution in exists.16 Whether (8) has a solution, is a question of the characteristic of the distribution function. For example, in the case of the normal distribution the f (|N)/(1 − F(|N))-curve converges to 0 as → −∞ and becomes unbounded as → +∞. In that case, existence is guaranteed. Moreover, given N, the solution is unique. Changes in N shift the f (|N)/(1 − F(|N))-curve thereby affecting the bias which implements care c ∗ . Consequently, there are many pairs (∗ (N), N) that implement the first-best level of care. The foregoing suggests two initial comments. First, a sole focus on an analysis of biases, as suggested by the comparison between customary and technological advancement, is too limited to correctly evaluate the effect of a state-of-the-art defense. In particular, depending on the specific context, it is possible that a number of tests could accomplish the task. Second, each of the tests require a different appropriate sample size. In the foregoing, we identified the customary test with = 0. It will induce firms to produce first-best care if the sample size, hereafter denoted by N 0 , satisfies17 : f (0|N 0 ) −p (c ∗ ) . = ∗ p(c ) 1 − F(0|N 0 )
Keeping in mind that the distribution of the N is symmetrical around zero, F(0|N) = 1/2 for all N. Therefore, the right-hand side of (9) simplifies to 2f (0|N 0 ). Whether the equation has a solution in N, depends on the exact nature of the N distribution. For example, if the underlying ε are normally distributed, it is easily verified that (9) becomes √ −p (c ∗ ) 2 . (10) = p(c ∗ ) (1 + (1/N 0 )) Since for all N, 0 ≤ (1/N) ≤ 1, (10) has a solution provided that the condition √ −p (c ∗ ) 1 2 (11) ≤ √ ≤ √ p(c ∗ ) is satisfied. Assuming the above condition is satisfied, a customary test with sample size N 0 would implement the first-best. Naturally, a question arises as to whether a technologically advancement test could also induce first-best at the Nash equilibrium. To discuss the issue, consider decreasing N below N 0 . For = 0, note that f (0|N) < f (0|N 0 ) (intuitively reducing N increases
16 15
Assuming the second-order condition holds. See the restrictions in footnote 14.
(9)
17
Obviously assuming the second-order condition holds. Observe that with = 0, the second-order condition is certainly satisfied.
110
B. Deffains, D. Demougin / International Review of Law and Economics 28 (2008) 106–112
sect the y-axis above the f (|N ∗ )/(1 − F(|N ∗ ))-curve. Therefore, implementing the first-best would require to set < 0. In words, the legal rule would demand courts not to declare the defendant liable, even if the comparison with the large sample was (marginally) unfavorable to him. The literature, with respect to the state-of-the-art defense, has focused on the comparison between the customary practice test and the technological advancement test in terms of the bias. In doing so, it has neglected an important dimension. Indeed, for the comparison to be meaningful, it must include a discussion on the sample size associated with the respective tests. The foregoing analysis shows that there is a trade-off between the bias introduced by the courts to establish the state-of-the-art defense, and the sample size of producers which the defendant is compared with. In particular, the technological advancement test necessitates a smaller sample size than the customary practice test. The analysis also suggests, that the comparison should not be solely focussed on primary costs, since it would appear that many combinations of bias and sample size implementing first-best care. There are other aspects based on secondary and tertiary costs that influence the comparison between the two legal practices under study. We list three arguments that speak in favor of a large bias and a small sample size N:
Fig. 2. Bias and sample size.
Fig. 3. Social cost and sample size.
the variance of N , thus, flattens density function of N around its mean). Therefore the f (|N)/(1 − F(|N))-curve intersects the y-axis below the f (|N 0 )/(1 − F(|N 0 ))-curve. Consequently, lowering the sample size below N 0 would require courts to raise the bias or, according to our interpretation of , to use a technological advancement test (see Fig. 2).18 There is another way to understand the result. N 0 was originally defined to satisfy the firm’s first-order condition at c ∗ with set at 0: 1 + [p (c ∗ )F(0|N 0 ) − p(c ∗ )f (0|N 0 )]L = 0.
(12) N0
Consider the effect of an increase in N above in (12). The first term in the square bracket would remain unchanged, since F(0|N) = 1/2 for all N. However, the second term drops, so that the lefthand side of (12) becomes positive. This, in turn, implies that a firm facing a legal rule characterized by = 0 and N > N 0 would find it optimal to increase its care level above c ∗ . Moreover, note that c ∗ + p(c ∗ )F(0|N)L is constant for all N. Thus, the two cost curves must intersect at c ∗ as in Fig. 3. In order to reestablish equality in (7) requires raising . Since F(|N) > 1/2 = F(0|N 0 ), it means that in the case of damages firms are found more often liable thereby increasing expected costs. The observation points to a surprising peculiarity for legal policy. Suppose that, instead of reducing N below N 0 , we were to consider an increase in the sample size. Again, this would shift the f (|N)/(1 − F(|N))-curve. However, now the curve would inter-
18 In the foregoing, the second-order condition is ignored. As we discussed earlier this may limit the size of the feasible and indirectly may provide a lower bound for feasible sample sizes.
• First, sampling requires collecting and treating information of different producers’ industrial practices. This is a costly activity (note that from a normative perspective, it is irrelevant who carries these costs).19 Keeping these costs at a minimum, speaks in favor of a small sample and, therefore, a large , i.e. a technological advancement test. • Second, evaluating liability rules also demands considering the impact of damages adjudication on the parties’ activity levels. In the current situation, we saw earlier, that the firms’ expected private costs are smaller than the true social costs, c + p(c)F(|N)L < c + p(c)L. Consequently, the activity level of the defendant will be excessive, compared to the social optimum. In this respect, note that with the customary test F(0|N) = 1/2, whereas with a technological advancement test F((N)|N) > 1/2. Thus this second consideration also speaks in favor of defining a defense, based on a state-of-the-art argument, using the technological advancement test. • Third, if one considers the insurance mechanism inherent to civil liability claims, it seems interesting to remark, that victims will also be better off under a state-of-the-art defense, based on a technological advancement test. The explanation parallels the foregoing argument; with an increase in the Nash equilibrium value of F(.) firms pay damages more frequently. In other words, when state-of-the-art refers to an exceptional safety level, it is more difficult for the defendant to escape liability, being at the forefront of technical advancement. Therefore, partisans of tort liability, as a mean to compensate parties who suffer losses, should prefer a state-of-the-art defense with F(.) as close as possible to unity. Under the above considerations, the state-of-the-art defense should be defined in terms of the technological advancement test with a large bias and a small sample size. The foregoing conclusion must be qualified however since it would necessitate that society find a way to limit the sample size. In practice, this may be prove to be rather difficult in an adversarial environment and in the presence
19 It is important to note that the ability of the judge to control N is probably different under inquisitorial and adversarial procedures. A discussion of this point goes beyond the scope of this paper.
B. Deffains, D. Demougin / International Review of Law and Economics 28 (2008) 106–112
of competing experts whose aims are not to produce an efficient rule, but to represent the interest of their respective party.. 5. Discussion and concluding remarks This paper develops a modelisation of the state-of-the-art defense. From a normative perspective, our analysis could constitute a basis to guide courts applying a state-of-the-art defense. More specifically, on the trade-off between and N, it seems possible to find arguments in favor of the technological advancement test. However, the argument necessitates courts to clearly understand the relationship between the bias and the sample size. In practice, most judicial systems permit some defense based on a state-of-the-art argument. For example, the U.S. restatement Second of Tort § 402 A (1965) defining the products liability principles, permits such a defense requiring a “demonstration that the technology available for the manufacture of a safer finished product with the same characteristics was not feasible”. However, as we documented with the Cunningham v. MacNeal Memorial Hospital case, the state-of-the-art defense remains controversial. In this case, the Illinois Supreme Court rejected the argument that blood products, then in use, complied with the state-of-the-art in terms of safety from hepatitis contamination. The Court stated that “to allow a defense to strict liability on the ground that there is no way, either practical or theoretical, for a defendant to ascertain the existence of impurities. . . would emasculate the doctrine. . . and signal a return to a negligence theory” (Cunningham v. Memorial Hosp. 266 N.E.2d 897 at 902 (1970)). Moreover, the application of the defense raises numerous questions. For instance, when specifically should the defense be applied? In the case O’Brien v. Muskin Corp., the New Jersey Supreme Court, while recognizing the value of evidence as to the state-of-the-art in determining product defect and in applying a risk-utility analysis, held that the defense is not absolute (O’Brien v. Muskin Corp., 463 A.2d, 306 (N.J. 1983)). Further, in Beshada v. Johns Manville, the New Jersey Supreme Court expressly disallowed the state-of-the-art defense to the issue of failure to warn (Beshada v. Johns-Manville Corp., 447 A 2d 539 (N.J. 1982)). These cases concern asbestos litigations that have frequently considered the evolution of scientific knowledge. The same reasoning is generally applied in medical malpractices cases. However, the jurisprudence also refers to customary practice. For instance, the CERCLA imposes joint and several liability for cleaning up contamination caused by hazardous substances of responsible parties. One of the defense allowed is for the defendant to provide he had no reason to know of the existence of hazardous substances and made all “appropriate inquiry into the previous owner and uses of the property consistent with good commercial or customary practice.” 42 U.S.C. sec. 9601. We observe a similar debate in the European context. The council directive 85/374/EEC concerning the liability for defective products states that a producer will not be liable if the producer proves that the state of scientific and technical knowledge at the time the producer placed the product on the market was not such as to enable the defect to be discovered. The jurisprudence is still limited, but it appears, for instance, that courts’ decisions concerning liability of pharmaceutical firms adopt a definition based on scientific knowledge to evaluate the producer’s liability. In these cases, the definition of the state-of-the-art considers the technological advancement of the industry as expressed in scientific publications. The paper demonstrates that the two interpretations of the state-of-the-art defense in civil liability claims, the customary practice test and the technological advancement test, could be – contrary to the intuition – equivalent in terms of firms’ efforts to promote safer technologies. In analytical terms, the reason is that, a tradeoff exists between the bias introduced by the courts to establish
111
the state-of-the-art, relatively to the average level of care and the size of the sample. Of course, it is difficult to determine the optimal combination. However, if the observation of technical characteristics of an industry is difficult for the courts, implying costly and long experts’ investigations, it could be better to refer to small samples. In this case, a normative analysis would conclude in favor of the technological advancement test because it does not require to collect exhaustive information about firms. Moreover, this system presents advantages in terms of injurers’ activity levels and of victims’ compensation. Considering the fear expressed by Porter (1990) that product liability becomes so extreme and uncertain as to retard innovation, it appears that the problem is not the evolution towards strict liability. It is essentially explained by the US judicial environment: importance of pain and suffering awards, punitive damages, role of civil juries, contingent fees for lawyers, class actions and the standard of proof based on the preponderance of evidence. The state-of-the-art defense is generally discussed under a strict liability rule. The context which has influenced this choice is, of course, the one of product liability in most of countries. However, our main conclusion concerns liability rules in general. In a sense, the problem of the state-of-the-art defense is similar to the definition of a due care standard under a negligence rule. The problem could easily be extended this way, because a due care standard could be defined in reference to the current practices of people or to an exceptional level of effort exerted by them. Finally, our analysis demonstrates that there are theoretically many pairs (∗ (N), N) that can implement the first-best. However, from a practical point of view, we have seen that the interest of the society could be to define the largest possible value for F(.). This implies a small N and an important bias under a technological advancement test. In this case, the solution will tend towards a strict liability system. So, the position of the Illinois Supreme Court in Cunningham v. MacNeal Memorial Hospital about the risk of “emasculation of the doctrine” could be well justified on economic grounds. Another important practical consideration concerns the way firms are selected to be included in the sample. In the model, we have implicitly supposed that they are randomly chosen. However, this is not necessarily true. In practice, an injurer found liable based on the signals of some competitors (chosen by the court or by the expert) may sometimes want to present alternative signals based on other competitors that are more favorable to his case. This constitutes a natural extension of the model beyond the scope of our analysis. In addition, the nature of the procedures should also be studied. For example, our analysis presupposes that the court controls the procedures and, in particular, the way experts select the sample. In that respect, there is certainly a significant difference between common law and civil law countries. Acknowledgement The authors would like to thank two anonymous referees for comments on an earlier version of this paper. References Boyd, J., & Ingberman, D. A. (1997). Should “Relative Safety” be a test of product liability? Journal of Legal Studies, 26(2), 433–473. Brown, J. P. (1974). Product liability: The case of an asset with random life. American Economic Review, 64(1), 149–161. Calfee, R., & Craswell, J. E. (1986). Deterrence and uncertain legal standards (with John E. Calfee). Journal of Law Economics & Organization, 2(2), 279–303. Hylton, K. (2000). The theory of tort doctrine and the restatement of torts. Boston University School of Law Research Paper No. 07. Available at SSRN: http://ssrn.com/abstract=242925 or doi:10.2139/ssrn.242925. Landes, W. M., & Posner, R. (1981). The positive economic theory of tort law. Georgia Law Review, 15, 851–924. Landes, W. M., & Posner, R. (1987). The economic structure of tort law. Cambridge University Press.
112
B. Deffains, D. Demougin / International Review of Law and Economics 28 (2008) 106–112
Lazear, E., & Rosen, S. (1981). Rank order tournaments as optimum labor contracts. Journal of Political Economy, 89, 841–861. Milgrom, P. R. (1981). Good news and bad news: Representation theorems and applications. Bell Journal of Economics, 12(2), 380–391. Polinsky, A. M. (1980). Strict liability vs. negligence in a market setting. American Economic Review, 70, 363–370. Porter, M. E. (1990). The competitive advantage of nations. The Free Press.
Shavell, S. (1980). Strict liability versus negligence. Journal of Legal Studies, 9, 1–25. Shavell, S. (2007). Economic analysis of accident law (new edition). Harvard University Press. Spence, M. (1977). Consumer misperceptions, product failure and producer liability. Review of Economic Studies, 44, 561–572. Viscusi, W. K., & Moore, M. J. (1993). Product liability, research and development, and innovation. Journal of Political Economy, 101(1), 161–184.