The skewing effect of outcome evidence
| Published date | 01 October 2023 |
| DOI | http://doi.org/10.1177/13657127231187056 |
| Author | Omer Pelled |
| Date | 01 October 2023 |
| Subject Matter | Articles |
The skewing effect of outcome
evidence
Omer Pelled
Faculty of Law, Bar-Ilan University, Israel
Abstract
Behaviours are primarily regulated to reduce the risks of a negative outcome to others. This
article discusses the use of outcomes as evidence of violations of a legal standard (outcome
evidence). The current debate over outcome evidence centres around limited rationality.
Opponents argue that factfinders’estimations are distorted by hindsight bias, while supporters
argue that factfinders properly update the probability of fault, given information about the out-
come. The article adopts the rationality assumptions and argues that factfinders should never-
theless disregard outcome evidence in most cases unless the outcome can provide evidence
that works for or against the defendant or when the law creates inefficient incentives to com-
ply with the legal standard, then using adverse outcomes as evidence may help solve the prob-
lem of undercompliance. The article further shows that when evidence cannot be excluded,
changes to the law governing primary behaviour are warranted to account for the distortion-
ary effect of outcome evidence.
Keywords
cost benefit analysis, error, hindsight bias, incentives, outcome evidence, uncertainty
Introduction
The law often regulates continuous aspects of behaviour, such as driving speed, drivers’alcohol con-
sumption, polluters’emission levels and noise, to name just a few. The regulation process comprises
two stages: setting the regulation, during which the legislator defines permitted and prohibited spectrums
of the behaviour, and the enforcement stage, during which a factfinder (e.g. courts or police) detects and
sanctions violations. For example, when regulating driving speed on a particular road, the local transpor-
tation department decides on a speed limit. Later, local police, or other transport enforcement agencies,
must assess the deriving speed of a suspected violator to enforce that limit.
During the enforcement stage, factfinders rely on evidence to determine how parties have acted. As
evidence may relay inaccurate information, there is a risk that factfinders will overestimate or
Corresponding author:
Omer Pelled, Faculty of Law, Bar-Ilan University, Ramat Gan 5290002, Israel.
E-mail: omer.pelled@biu.ac.il
Article
The International Journal of
Evidence & Proof
2023, Vol. 27(4) 307–324
© The Author(s) 2023
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/13657127231187056
journals.sagepub.com/home/epj
underestimate the conduct. For example, the police might use a radar detector to identify speed limit vio-
lations. When the radar is calibrated correctly, it provides an accurate assessment of the speed on averag e,
but in each reading, it might overestimate or underestimate the speed. Since the assessment of the conduct
is inaccurate, actors face uncertainty about enforcement.
Among other evidence, factfinders can learn about the actor’s behaviour from the outcome of that
behaviour. For example, the occurrence of a road accident may be suggestive of driving speed. We
refer to this type of evidence as outcome evidence. Behavioural economists have argued that factfinders
suffer from irrational hindsight bias and, as a result, systematically overestimate the probability of the
actor’s fault when they are presented with outcome evidence (Jolls et al., 1998; Kamin and
Rachlinski, 1995; Labine, 1996; Sunstein, 2000). These scholars have suggested changing the legal
standard or increasing the evidentiary threshold for liability to correct the bias.
1
Rational choice theorists
have argued in response that the so-called hindsight bias is not an irrational cognitive bias but rather a
rational Bayesian updating of the probabilities based on additional evidence (Kelman et al., 1998;
Posner, 1998, 1999). As the argument goes—when the factfinder has imperfect information about a
party’s conduct, and the factfinder knows that the conduct affects the probability of an adverse
outcome, a rational factfinder updates its prior estimation of failure to comply with the legal standard
based on the knowledge that the adverse outcome has occurred. The article does not attempt to settle
this argument. Instead, it aims to reconcile these two opposing accounts of outcome evidence and
shows that even if using outcome evidence is rational, it skews the court’s perception of the actor’s
conduct and distorts the incentives to comply with the legal standard.
To start the analysis, we should consider how uncertainty generally affects the incentives to comply
with a legal standard. Calfee and Craswell (1984, 1986) show that when actors face uncertainty about the
legal ramifications of their conduct, it affects deterrence and may lead to overcompliance or undercom-
pliance. Uncertainty distorts incentives in two opposite ways. First, when actors know that their actions
are detected with some noise, the difference between the expected sanction when conforming to the
standard and when violating it gets smaller, reducing the incentives to comply. Second, actors know
that they can affect evidence production and reduce the probability of being sanctioned by increasing
their compliance efforts, thus increasing incentives to comply. The net incentive to overcomply or under-
comply depends on the distribution of errors in the factfinder’s assessment.
The distribution of errors in assessing the actor’s conduct is determined by the availability of evidence
and the rules governing evidence admissibility and weight. Since the distribution of errors affects com-
pliance efforts, one possible objective of evidence law is to promote efficient incentives. This article
focuses on how the admissibility of certain probative evidence may distort the incentives by changing
the mean of the distribution. When the mean and median of errors equal to zero, factfinders are
unbiased—they do not systematically overestimate or underestimate the actor’s conduct. When the
mean error is positive, factfinders tend to overestimate the actor’s conduct. The skewness in the distribu-
tion of errors affects the incentives in an obvious way—when the factfinders tend to err in favour of an
actor, the actor has less incentive to comply and vice versa. For example, if factfinders underestimate
driving speed, all else being equal, drivers have less incentive to comply with the speed limit
(Dari-Mattaiacci, 2005).
The question, then, is what type of evidence can produce this skewed distribution of errors, assum-
ing that factfinders are rational and well informed about the signal the evidence provide. At first
glance, it seems that errors should never be skewed to one side, i.e. even though factfinders might
err in any given case, their assessment is accurate on average. Rational factfinders that know the dis-
tribution of errors can offset any systemic bias in the evidence as part of their evaluation. For example,
1. In extreme cases, Arkes and Schipani (1994) suggested the to deal with hindsight bias, injurers should be immune from liability,
for example, in cases where the business judgment rule applies.
308 The International Journal of Evidence & Proof 27(4)
To continue reading
Request your trial