Assuring Civil Damages Adequately Deter: A Public Good Experiment

• Published date: 01 June 2014
• DOI: http://doi.org/10.1111/jels.12042
• Date: 01 June 2014
• Author: Theodore Eisenberg,Christoph Engel

Assuring Civil Damages Adequately Deter:

A Public Good Experiment

Theodore Eisenberg and Christoph Engel*

To explore damages rules’ deterrent effect we use a public good experiment to tailor

punishment to rules used in civil litigation. The experimental treatments are analogous to:

(1) damages limited to harm to an individual, (2) damages limited to harm to a group, such

as in class actions, and (3) treble damages. For (1) and (2) we also manipulate the prob-

ability of a player being entitled to claim damages. The treatments with damages limited to

harm to an individual do not prevent deterioration in cooperation over time but deteriora-

tion is slower. In the class action treatment, cooperation is stable over time if the probability

of having to pay damages is sufficiently high. The same holds for the treble damages

treatment. The results persist in variations of (1) and (2) in which the player imposing

damages may have them forfeited with no benefit to her. We can therefore rule out that the

beneficial effect of sanctions hinges on the participants knowing that the player imposing

sanctions cannot intend to enrich herself.

I. Introduction

Does damages liability promote socially desirable behavior? Reviews of tort law’s effects

reach mixed conclusions, with one finding tort law “somewhat successful” in deterring

(Schwartz 1994:443), but a recent experiment finding subjects’ willingness to engage in

risky behavior unaffected by the threat of potential tort liability (Cardi et al. 2012). Empiri-

cal studies assess deterrence by comparing liability with nonliability, such as comparing

alcohol-related motor vehicle fatalities across states that do or do not impose liability on

commercial servers of alcoholic beverages (Sloan et al. 2000). Other studies compare the

degree of available damages by assessing whether tort reforms such as noneconomic

damage caps or higher evidentiary standards for punitive damages are associated with

aggregated behavioral changes (e.g., Rubin & Shepherd 2006).

*Direct correspondence to Christoph Engel, Director, Max-Planck-Institute for Research on Collective Goods,

Kurt-Schumacher-Straße 10, D 53113, Bonn, Germany; email: engle@coll.mpg.de. Engle is also member of the Faculty

of Law and Economics, University of Bonn; Eisenberg was Henry Allen Mark Professor of Law and Adjunct Professor

of Statistical Science, Cornell University.

Helpful comments by Aniol Llorente-Saguer, Thomas Miles, Jeroen van de Ven, Konstantinos Chatziathanasiou,

Marc Jekel, Alexander Stremitzer, anonymous referees, and by audiences at the Conference on Empirical Legal

Studies (2012), the European Law and Economics Association (2012), the Italian Society for Law and Economics

(2012), Tel Aviv University Buchmann School of Law, the Amsterdam Law and Economics Seminar, the University of

Cologne Behavioral Economics Seminar, the University of Bologna, and the University of Colorado Law School are

gratefully acknowledged, as are the programming of the experiment in zTree by Nicolas Meier and the translation of

the instructions into English by Brian Cooper.

bs_bs_banner

Journal of Empirical Legal Studies

Volume 11, Issue 2, 301–349, June 2014

301

An attractive feature of such studies is that they investigate liability rules in action.

External validity is not an issue. Yet however carefully they are done, specifying models of

outcomes, such as fatality rates or damages awards, is challenging. The analyst may be

unsure about omitted variables and reverse causality is often a concern. She is at the mercy

of naturally occurring variation, and cannot fine tune dependent and independent vari-

ables. Panel data usually are not available so one cannot address whether the same subjects

continuously observed over time change their behavior in response to liability experience.

For all these reasons, experimental data are a useful complement. They are, of course, less

“real” than observational data, but random assignment of experimental subjects makes

modeling outcomes straightforward, removes identification problems, and observations

over time enable evaluating the process leading to observed outcomes.

Public good experiments are particularly well suited to overcoming observational

studies’ limitations (e.g., Fehr & Gächter 2000). The experiments provide a continuous

measure of one type of behavior society would want to deter, free riding on others’

efforts to provide public goods. Environmental damage can illustrate troublesome free

riding directly relating to tort (e.g., Hardin 1968). Clean water, clean air, or the proper

disposal of waste may all be modeled as public goods. Everyone benefits (there is no

rivalry in consumption) and no one can be excluded (there is no property right). If some

members of society switch to a more costly alternative to reduce pollution, those reck-

lessly discharging pollutants free ride on their efforts. More generally, rule abiding can

be interpreted as making a contribution to a public good. Frequently, it has a direct

(out-of-pocket) cost. At any rate it has the opportunity cost of the person foregoing the

potential benefit from violating the rule. Many rules, such as traffic rules, directly benefit

undefined numerous individuals. Other rules, such as criminal laws, indirectly benefit a

large group through reduced need to take precautions against rule violations. Antitrust

remedies also illustrate law seeking to promote a public good. Here, workable competi-

tion is the public good.

In addition to assessing public good behavior, public good experiments’ subjects

repeatedly interact over time. This enables observing the dynamic individual-level process

of behavior, followed by sanction, followed by possibly modified behavior, further possible

sanction, and so on. Public good experiments therefore allow assessing whether partici-

pants change behavior in response to previous damages decisions. More generally, the

experiments address whether subjects learn to adjust their behavior to the presence of the

experimental institution, or risk that the institution becomes unsustainable. In a repeated

game, one can also simultaneously investigate outcomes (How effective are damages at

deterring socially undesirable behavior?) and process (Why do [do not] damages serve

their stated purpose?). Use of public good experiments to explore the effects of legal rules

is growing (Grechenig et al. 2010; Engel & Kurschilgen 2011).

This article is the first to use a public good experiment to study the deterrence and

cooperation-producing effects of different civil damage regimes. To explore damage rules’

influence on behavior, we tailor allowable sanctions, referred to herein as damages, to rules

analogous to actual civil litigation rules. In each period of the multiperiod experiment, a

random draw first determines, with equal probability, whether damages may be claimed. If

they may, a specified participant, the active player, may obtain damages from a targeted

302 Eisenberg and Engel

participant, subject to a different damages limitation in each treatment. The experiment’s

three treatments are analogous to: (1) damages to an individual litigant in the classical

bipolar lawsuit, (2) damages to a group available in aggregate litigation such as class actions,

and (3) damages beyond actual harm to victims, which we implement as treble damages, as,

for instance, in antitrust suits. We are interested in which treatment yields the greatest

contributions to the public good, in particular in the long run.

Studying the deterrent effect of different damages rules via a public good experiment

can help inform about the expected performance of a range of actual damages situations,

but caution is necessary with respect to the degree to which such an experiment corre-

sponds to the reality of any particular damages area, such as tort. Our experiment’s

treatments vary the degree and probability of damages, as do many real-world legal rules

such as allowing recovery for harm to a class, or allowing damages beyond actual harm, such

as treble damages or punitive damages, but the experiment is, of course, not fully analogous

to actual specific damages rules or procedural devices relating to damages, such as class

actions. Nor can the experiment replicate the richness of real-world behavior, such as

decisions to take actual harm-producing precautions. We defer further discussion of such

external validity concerns until after describing the experiment and reporting its results.

Regardless of the strength of the analogy between our treatments and actual legal rules, we

believe this is the first public good experiment even to try to relate cooperative behavior

over time to real-world civil damages rules.

Since we are interested in damages’ deterrent effect, it is natural to investigate the

effect of varying the level of damages; in criminological parlance, we manipulate severity.

Across treatments, the active player has power to inflict varying harm on passive players. In

the parallel discourse in criminal law, it usually is found that certainty looms larger than

severity (Antunes & Hunt 1973; Cook 1980; Paternoster 1987; Nagin 1998; Pogarsky 2002;

Durlauf & Nagin 2011a, 2011b). To observe the effect of certainty, we replicate treatments

(1) and (2), but make it twice as likely that a victim may claim damages.1The sequence of

treatments also allows us to test a third explanation. Potential freeriders might be sensitive

to the expected value of the sanction, that is, to the product of its probability with its

amount.

In a replication of treatments (1) and (2), we give the active player the additional

option to destroy via forfeiture some or all of the targeted participant’s period income, with

no benefit to the active player, as in the case of societally imposed punishment. The

maximum amount that the active player can cause to be forfeited equals the maximum

damage amount that can be claimed in these two treatments. This forfeit option is analo-

gous to reporting misbehavior to the authorities, with the expectation of society-level

punishment being imposed, rather than suing in tort or contract in the hope of obtaining

individual-level economic benefit via damages for the plaintiff.

In all our treatments, assuming that everybody maximizes income, is risk neutral, and

that this is common knowledge, a unique equilibrium exists in which all participants keep

1We cannot do so for treatment (3) since this would change the equilibrium. Unlike the equilibrium in treatments

(1) and (2), it would not be a selfish individual’s best response to contribute her entire endowment.

Assuring Civil Damages Adequately Deter 303