Download Behavioral Field Evidence on Psychological and Social Factors in

Behavioral Field Evidence on Psychological and Social Factors in
Dishonesty and Misconduct
Lamar Pierce
Olin Business School
Washington University in St. Louis
Parasuram Balasubramanian
Olin Business School
Washington University in St. Louis
ABSTRACT: We review recent behavioral field evidence on dishonesty and other unethical
behaviors from psychology and related fields. We specifically focus on individual-level
studies that use explicitly behavioral data in natural settings, covering research topics
relevant to psychology from across disciplines. Our review shows both the paucity and
potential of behavioral field evidence on the psychology of dishonesty--although such
research can provide actionable and realistic conclusions, it presents a host of practical and
identification-related challenges that have limited its use. We explain the major
methodological approaches, and discuss the multiple identification challenges for
researchers using archival and other non-experimental data.
Laboratory research on dishonesty and other unethical and illicit behaviors has proven invaluable in
helping to understand the behavioral underpinnings of misconduct. Similarly, survey-based studies
have provided a wealth of data and insights on self-reported dishonesty as well as its motivations,
mechanisms, and prevalence. Yet an emerging stream of research is using behavior data from field
experiments, direct observation, and archival sources to address concerns about the generalizability
of often low-stakes laboratory studies and potentially biased self-reported data. This review details
the current state of this emerging literature in psychology and related fields, and provides guidelines
for future research. We focus specifically on studies that use individual-level behavioral data from
“natural” settings—those where people engage in their typical work or personal activities. Related
reviews on organizational-level misconduct [1] and broader literatures in business and behavioral
ethics [2, 3] are also valuable reading.
Existing Behavioral Field Research on Dishonesty
We first review the existing behavioral field research on topics of interest to psychologists and
behavioral scientists.
Social processes: One of the most promising and important topics on dishonesty is how social processes
influence behavior, with a growing body of work using behavioral field evidence to explore it.
Bucciol et al. [4] used direct observation and interviews to identify how bus passengers travelling with
family members were more likely to have a valid ticket, but not those travelling with friends.
Similarly, a field experiment on customers keeping excessive change in Israeli restaurants found
almost no improved honesty from groups, with the higher average honesty of women exerting little
pressure on their male dining companions [5]. These results suggest that social pressure may
selectively increase honesty, but that the specific social dynamics are critical. Two recent studies of
performance enhancing drugs in baseball [6] and cycling [7] show that social and professional
interactions are critical in disseminating both knowledge and acceptance of illicit drug usage. These
follow an important early study of social processes in sports cheating, where Duggan and Levitt [8]
showed that sumo wrestlers reciprocally throw matches to aid one another in achieving a minimum
win count. It is also consistent with recent work using communication data to examine information
transmission among networks of dishonest parties [9, 10]. This is consistent with a field experiment
by Wenzel [11] that found information on others’ behavior improved tax compliance, as well as
results showing employees become more dishonest when joining dishonest firms [12].
Fairness, equity, and social comparison: Social comparison and related fairness and equity concerns are
also a focus of recent work. Early work by Greenberg [13] was one of the first to address this topic
using behavioral field data, showing increased theft following a pay decrease at two out of three
factories. A related study [14] also showed higher theft when the employer, not coworkers, was the
likely victim. A notable later study by Edelman and Larkin [15] found social comparison as a
motivation among faculty fraudulently downloading their own papers on SSRN. Related to social
comparison is a small set of field studies on socioeconomic class and dishonesty. Although Gino and
Pierce [16] found evidence of dishonest helping within socioeconomic class in mechanics, Balafoutas
et al. [17] find no differences in fraud by taxi drivers across customer income levels. Related work
[18] examines the socioeconomic class of aggressive drivers, although mechanisms linking dishonesty
with personal wealth are difficult to separate.
Moral reminders and preferences: Multiple large-scale field experiments have focused on testing the
efficacy of moral reminders previously established in laboratory studies. Studies of individual
taxpayers [19] and newspaper buyers [20] found that the inclusion of a moral reminder increased
honesty in disclosures and payments. In contrast, a field experiment by Fellner et al. [21] found that
Austrians only improved their honesty in paying TV licensing fees when mailed threats of
enforcement, not when sent moral appeals. These build on an earlier important study of bagel
customers by Levitt [22], who found that payments under the honor system were largely a function of
internal moral preferences. Furthermore, he found that the September 11 terrorist attack significantly
increased honesty in payments, suggesting the power of moral reminders. Related to this, Shu et al.
[23] used a field experiment to show that insurance customers who signed at the top of forms
reported higher annual mileage than those who signed at the bottom, presumably because signing
provided a moral reminder.
Culture: Several recent studies have also found the influence of ethnic or national culture and identity
on dishonest behavior. A foundational study in economics correlated national corruption measures
with the unpaid parking tickets of diplomats [24]. Other papers focused on how interactions within
and across ethnic and national groups can change levels of dishonesty, including favoritism in
Olympic judging [25], ethnic diversity and corruption in Indonesia [26], and stock market fraud in
Kenya [27]. One approach by Bianchi and Mohliver [28] links economic conditions during
executives’ formative periods to stock option backdating.
Professionalism: One growing area of interest is how the professional identity and pro-social motivation
of an expert can clash with her career and financial incentives. Although dishonesty in certain
professions might be expected (e.g., auto mechanics) [29], for others the public’s trust in expert
honesty is critical. Medicine provides several examples, such as how liver transplant surgeons’
financial and prosocial motivations can lead to dishonest patient reporting [30]. Similarly, teachers
who are expected to instill ethical values in children have been shown to cheat when pressured with
strong financial and career incentives [31].
Incentives and control: One of the largest bodies of behavioral field studies centers on extrinsic
motivation from incentives and control—how financial payoffs, monitoring, and penalties can alter
dishonest behavior. Although the majority of this work is in economics [32], the work on monitoring
has particular implications for psychological theories of dishonesty. Monitoring, for example, has
been shown to reduce theft [33, 34], unexcused absenteeism [35], and dishonest reporting [36] in
organizational settings such as call centers, restaurants, schools, and banks. Similarly, recent field
experiments have targeted tax fraud [37, 38] and corruption [39, 40] through the explicit
manipulation of increased monitoring through audits and transparency. Although economic theory
implies the efficacy of monitoring, evidence from a field experiment on factory productivity
monitoring [41] suggests that psychological mechanisms may make monitoring counterproductive in
reducing dishonesty. Behavioral field research that can test the multiple psychological and economic
mechanisms invoked by monitoring is clearly needed.
Methodological Approaches: Field Experiments and Archival
Three principal methodological approaches dominate behavioral field research on dishonesty: direct
observation, randomized field experiments, and archival data analysis. Direct observation involves
actively observing and recording behavior under multiple conditions to infer relationships between
dishonesty and environmental conditions or individual differences. This method lacks the
randomized manipulation of a field experiment, and observation is typically covert to avoid
Hawthorne effects. The broad and random sampling of both honest and dishonest behavior across
conditions and differences is important to avoid selection bias. Examples include researchers actively
watching and recording problematic behavior such as illegal parking [42], bus fare evasion [4],
aggressive driving [18], or bribes paid by truckers in Indonesia [43].
Field experiments typically involve direct observation, but also include random assignment of
manipulations to treatment and control groups, as in laboratory experiments. Although highly
stylized experiments outside of a laboratory are often considered to be “field experiments,” those
conducted in natural behavioral settings are most valuable for understanding behavior in the field.
Such “natural field experiments” study individuals in plausibly normal daily behavior, rather than in
contrived tasks or jobs they would not normally do. Random assignment can occur either as
individuals [23] or groups [20]; field settings often make individual randomization impossible for
practicality reasons or because of the inseparability of organizational or social settings. The strengths
of natural field experiments on dishonesty are threefold: they are immediately generalizable to
specific social or organizational settings, they provide strong causal inference, and they frequently
provide immediate and measurable policy implications for managers or government.
The third methodological approach is the analysis of archival behavioral field data. Archival data on
dishonesty can be easier to acquire than opportunities for true field experiments or even
observational data, particularly in organizational settings where management can be reluctant to allow
researchers to manipulate or interact with the work environment. Archival data present several
important identification challenges, however, because of the lack of random assignment. Causality is
therefore difficult to establish because of selection bias and often cross-sectional data structure. Just
as commonly, omitted variables might explain the relationship between the variable of interest and
dishonesty in a way inconsistent with the proposed theoretical explanation. Researchers therefore
often rely on “natural experiments”—exogenous shocks to some individuals or groups that can
create quasi-treatment and control groups, such as exogenous restaurant monitoring technology
implementation [33].
Measuring Dishonest Behavior
One of the great challenges of studying dishonesty in the field is measurement precision and
accuracy. Dishonesty and other forms of misconduct are intentionally committed under a veil of
secrecy, due to explicit costs from detection and punishment. This yields two substantial problems in
empirical work. First, measures of dishonesty will be inherently imprecise, with low observability
rates across subjects and conditions. So long as this observability is randomly generated, it serves
only as noise in any empirical model, and does not bias any results. The second and larger problem is
when the observability of dishonesty is correlated with some other factor, such as individual
competence, and is thus inaccurate and biased. This most commonly occurs because dishonest acts
are rarely randomly detected and recorded. Instead, the actual detection of dishonest behavior is
usually endogenous to any model of behavior. When we observe data on detected dishonesty, it
reflects many observations of false negatives, and these false negatives almost certainly reflect their
own psychological and economic processes.
Given these measurement challenges, there are several ways in which researchers typically measure
dishonesty using behavioral field data. The simplest is to directly observe the behavior, either by the
researcher or recorded in an archival data set. Jin and Kato [44] provide an excellent example through
a field experiment purchasing baseball cards on eBay—the comparison of reported versus true
condition directly measures dishonesty. Pierce et al. [45] provide an example from archival data,
observing revenue theft by workers in point-of-sales data from a large set of restaurants.
The second common method is to identify suspicious patterns of behavior that are inconsistent with
the known counterfactual of honest behavior. Although this measurement strategy does not allow
researchers to precisely identify individual dishonest acts, it provides probabilistic measures that still
allow for hypothesis testing. Jacob and Levitt [31] provide a classic example of this through
suspicious answer patterns that identify teacher cheating. Pierce and Snyder’s [12, 45] use of
improbably high pass rates from vehicle emissions testing experts provide a good example of fraud in
a firm setting. Within academia, Simonsohn [46] and colleagues [47] provide clear examples of how
counterfactual probability distributions of hypothesis tests can help identify both intentional and
subconscious dishonesty in data reporting. This measurement method can be particularly useful in
the context of a field experiment, where a manipulation that should only affect dishonest people
(such as the risk of a tax audit) can impact reported income and deductions [19, 33, 38].
Other Identification Challenges in Archival Work
Addressing alternative explanations
Although field data may show patterns consistent with the theoretical explanation advanced by
researchers, they carry an additional challenge endemic to non-experimental approaches—alternative
explanations for observed correlations. In a tightly controlled laboratory setting, one can precisely
manipulate the variable of interest, and thereby hold all other factors constant. But in a field setting,
such precision is rarely possible, and identification can be haunted by multiple plausible or even
probable alternative explanations for the results. Convincing archival and observational studies of
dishonesty must not only provide proof for a hypothesis, but must also cast substantial doubt on
alternative hypotheses.
Most significantly, causality is often difficult to establish, and researchers must be careful not to infer
a causal relationship from correlational results. Even panel data, where multiple individuals are
observed across time, often is unable to reveal causal evidence because of independent variables that
are endogenously determined. One promising approaches is to exploit a natural experiment, where
the impact of plausibly exogenous shock on individual behavior can be reasonably called causal [22,
20, 33]. Another is to exploit discontinuous policy or rules through the equivalent of a quasiexperimental regression discontinuity design [8,30, 48]
In many field settings, alternative hypotheses cannot be fully dispelled, and the scope of identification
problems must be weighed against the novelty of the research question and field setting. Yenkey’s
[27] study of ethnicity and stock market fraud in Kenya, for example, is exceptionally novel on both
dimensions. In all cases, researchers should be, and should be encouraged by editors to be, honest
about identification problems in their research.
Identifying psychological (and economic) mechanisms
Perhaps the biggest challenge with behavioral field data is the identification of specific psychological
and economic mechanisms driving dishonest behavior. Theory may propose multiple mechanisms
that could explain observed behavior, yet without the benefit of a controlled experimental setting,
separating these mechanisms can be tricky. Behavioral field data are rarely accompanied by the selfresponse data frequently used by psychologists to measure psychological processes. Consequently,
researchers must instead attempt to identify mechanisms either by using variation in the main effect
through moderators.
The Promise of Mixed Methods
One promising approach to addressing limitations in behavioral field data on dishonesty is to
combine them with laboratory studies or surveys, particularly for the purpose of identifying specific
psychological or economic mechanisms [15, 16, 18, 20]. Researchers can test competing hypothesized
mechanisms in controlled laboratory settings, but must be careful that both the experimental design
and subject pool can be reasonably generalized to the field data setting. This is particularly important
when the field setting involves experts with long-run reputational or incentive concerns, which may
match poorly to undergraduate students playing highly stylized, low-stakes cheating games. Although
scenario studies can help target the specific setting, subjects may not have the expertise and settingspecific knowledge to provide insights on the mechanisms behind observed behavior in the field.
Similarly, surveys can help explicate the decision-making process and emotions, particularly when
directly conducted with the population from which the behavioral field data were drawn.
Conclusion
Behavioral field research presents great opportunities for better understanding the psychological and
economic foundations of dishonesty, yet they present their own set of unique challenges. The lack of
randomized assignment in archival and observational data presents arguably the great challenge
because they make causality difficult to establish and alternative explanations hard to dispel.
Consequently, researchers must approach behavioral field data with an appropriate level of
skepticism, and with the fundamental question of “what else might explain these results?” Results
that are consistent with a theory are not sufficient, because they produce false positive conclusions
that are particularly dangerous in settings involving dishonesty. Instead, results should ideally be
consistent with a theory and inconsistent with alternative theories for observed relationships.
The behavioral field studies discussed here are somewhat arbitrarily bounded by necessity, and are
part of a much broader set of field studies on dishonesty. For example, survey-based can be
extremely valuable to researchers, particularly when embedded in a randomized experimental design
[49].
Given the rapidly expanding literature of laboratory studies on cognitive biases in ethical decisionmaking, there is a tremendous need for further behavioral field evidence on bias in dishonesty. A few
recent studies address this shortage using data from banks [50] and illegal parking [42], but also
demonstrate the difficulty of separating cognitive bias mechanisms from others in field data. Perhaps
most promising is recent work by Rees-Jones [51] that uses a quasi-experimental regression
discontinuity approach to identify how loss aversion influences tax fraud. Further field studies that
can directly identify cognitive biases would provide a substantial contribution to the field of
behavioral ethics.
Finally, behavioral field data on dishonesty bring additional risks and concerns regarding human
subjects protection. Identifiable and private data on dishonesty have the potential to produce
substantial harm, such as job loss, social shame, and even incarceration. Furthermore, field
experiments must be typically limited to interventions believed to decrease dishonesty because of the
potential cost to subjects and other parties.
Annotated Bibliography
[1] Greve HR, Palmer D, Pozner J: Organizations gone wild: The causes, processes, and
consequences of organizational misconduct. The Academy of Management Annals 2010, 4:53–107.
In this review article, Greve et al. critique the literature on misconduct with an emphasis on
organizational causes. They argue for more rigor from researchers in defining and labeling
misconduct; since the definition of misconduct is often implicit in the literature and the role of
social-control agents has been neglected in identifying misconduct. They also examine the
consequences of misconduct for individuals, networks and organizations and propose possible areas
for future research.
[2] Tenbrunsel AE, Smith-­‐Crowe K: 13 Ethical decision making: Where we’ve been and where
we’re going. The Academy of Management Annals 2008, 2:545–607.
Tenbrunsel and Smith-Crowe review and summarize research on ethical decision making in
organizations with specific focus on behavioral or descriptive ethics. Using an inductive approach,
the authors focus on theoretical frameworks that arise from data. To summarize the field and
provide future direction, they develop a model of ethical decision making and a typology of key
dependent variables.
[3] Moore C, Gino F: Approach, ability, aftermath: A psychological process framework of
unethical behavior at work. The Academy of Management Annals 2015, 9:235–289.
Moore and Gino in this review important contributions in moral psychology from over a couple of
decades, and suggest ways for this literature to enrich and inform research on workplace unethical
behavior. They argue that dominant models of workplace unethical behavior of the past two decades
have failed to explain how and why people make the moral decisions they do – from moral
psychology and cognitive neuroscience in particular. The authors explain the role of intuition, affect,
physiology and identity in supporting and informing a more deliberative reasoning process in the
construction and enactment of moral behavior.
[4] Bucciol A, Landini F, Piovesan M: Unethical behavior in the field: Demographic
characteristics and beliefs of the cheater. Journal of Economic Behavior & Organization 2013, 93:248–
257.
In this field study, Bucciol et al. randomly interviewed 541 bus passengers in the city of Reggio
Emilia in Italy. They exploit a high level of fare evasion among passengers to find young individuals,
male and non-European immigrants are more likely to travel without a ticket. They also find a social
component with cheating: traveling with relatives increases probability of holding a valid ticket and
friends have no bearing on unethical behavior while traveling, except around noon when it was
mostly male students. Field interviews were conducted during September to December 2011 in
collaboration with a local NGO.
[5] Azar OH, Yosef S, Bar-Eli M: Do customers return excessive change in a restaurant? Journal
of Economic Behavior & Organization 2013, 93:219–226.
Azar et al. conduct a field experiment in a large restaurant in the center of Israel to study dishonest
behavior in a natural setting. When customers who paid with cash at the restaurant were returned
excessive cash in change, authors found that a majority of these customers did not return the
excessive change. They also found that repeat customers returned the excessive change back to the
restaurant as compared to one-time customers; and among repeat customers, more women returned
excess change as compared to men. The experiment was conducted between March and September
2011 and a total of 192 observations of excessive change and customer behavior were recorded.
[6] Gould ED, Kaplan TR: Learning unethical practices from a co-worker: The peer effect of
Jose Canseco. Labour Economics 2011, 18:338-348.
Gould and Kaplan examine if workers learn productive skills from their co-workers even when those
skills are unethical. The authors use data from the “Baseball Archive”, freely available on the internet
and restrict the sample to seasons between 1970 and 2009. They find performance of players
increased significantly after playing with John Canseco, consistent with his claims that he started a
contagion of steroid use in the 1980’s by influencing his peers and trainers. The authors do not find
any evidence for 30 other power-hitting stars affecting their teammates in the way Canseco did. Their
findings thus suggest that workers not only learn productive skills from their co-workers, but
sometimes those skills may have come from unethical practices.
[7] Palmer D, Yenkey C: Drugs, sweat and gears: An organizational analysis of performance
enhancing drug use in the 2010 Tour de France. Social Forces (forthcoming).
Palmer and Yenkey analyze data on 198 cyclists and 22 teams that competed in the 2010 Tour de
France against the backdrop of increasing, but incomplete social control of doping in cycling. They
theorize two under-explored causes of wrongdoing related to informal peer relationships and to an
organizational participant’s embeddedness in formal organizational structures. Their results suggest
that actors who perform critical organizational roles are more likely to engage in wrongdoing. They
also find that social ties to unpunished offenders increased likelihood of wrongdoing while ties to
severely punished offenders reduced it.
[8] Duggan M, Levitt SD: Winning isn’t everything: Corruption in sumo wrestling. The American
Economic Review 2002, 92:1594-1605.
Duggat and Levitt present statistical analysis documenting match rigging in sumo wrestling. They use
data from more than a decade of Japan’s sumo elite to find overwhelming evidence that match
rigging occurs in the final days of sumo tournaments. The data set contains almost every official
match between January 1989 and January 2000, a total of 32,000 bouts of top rank Japanese sumo
wrestling. With focus on economic analysis of corruption, they find wrestlers win a disproportionate
share of the matches when on the margin, not explained by increased effort. Reciprocity agreements
between stables of wrestlers appear to exist suggesting collusive behavior is not restricted to
individual actors.
[9] Baker WE, Faulkner RR: Social networks and loss of capital. Social Networks 2004, 26:91–111.
Baker and Faulkner use a strategic case in an oil and gas company (Fountain Oil & Gas Company) to
determine the effect of social networks on the probability of loss of capital after investments have
been made. Legal actions against Fountain provided publicly available records at the District
Attorney’s Office of Ventura County, California; in addition they obtain data from the receiver of the
bankruptcy court and Fountain’s financial statements. They find investors with highest risk of loss of
capital with probability 79%, did not use social ties or conduct due diligence; while investors with
lowest risk of loss of capital with probability 14%, effectively used within-network exchange and
conducted due diligence.
[10] Aven B: The paradox of corrupt networks Organization Science (forthcoming).
In this paper, Aven looks at Enron Corporation prior to its bankruptcy using longitudinal data of
organizational crimes, and combines qualitative coding techniques with social network analysis to
study effects of corruption on organization behavior. She identifies three corrupt projects for analysis
using data from government documents and employee testimonies, and also analyzes an email dataset
of communications from Enron for examination of communication of organizational crime. An
important finding of the paper is that corrupt information is a strong determinant of communication
behaviors that coincide with information sharing, but this effect weakens over time. Also it was
found that individuals within corrupt projects would communicate less and share fewer reciprocal
ties than those involved in non-corrupt projects.
[11] Wenzel M: Misperceptions of social norms about tax compliance: From theory to
intervention. Journal of Economic Psychology 2005, 26:862-883.
Wenzel conducts a field study with 1,500 Australian taxpayers in addition to a questionnaire-based
experiment with university students. With the support of the Australian Taxation Office (ATO),
random taxpayers were sent a survey regarding their own views as well as of other taxpayers,
specifically concerning honesty in deduction claims. He argues that taxpayers are likely to estimate
others’ acceptance of tax evasion as being greater than their own; and they find that feedback about
other self-other discrepancy could improve compliance and correct misperception of social norms.
[12] Pierce L, Snyder J: Ethical spillovers in firms: Evidence from vehicle emissions testing.
Management Science 2008, 54:1891–1903.
Pierce and Snyder use a unique data set of over three million vehicle emission tests from the
department of motor vehicles of a large northern state to show how organizations influence unethical
behavior of their employees. The study contains data of all vehicle inspections conducted in 2003
and 2004 for gasoline-powered vehicles under 8,500 lbs. Their results suggest individual ethical
behavior is influenced by the ethics of the employer and when individuals move from facility to
another, their leniency seems to shift to that of the employer. In particular, their findings show
ethical spillovers from firm to individual are strongest at large facilities, corporate chains and weakest
for large-volume inspectors.
[13] Greenberg J: Employee theft as a reaction to underpayment inequity: The hidden cost of
pay cuts. Journal of Applied Psychology 1990, 75:561-568.
In this field experiment, Greenberg studies employee theft rates in a manufacturing plant in a period
when pay was temporarily reduced by 15%. The study includes nonunion employees working for 30
consecutive weeks in one company spread across three manufacturing plants in mid-western US. He
uses two categories of dependent variables to measure theft behavior: actuarial data on employee
theft, and self-report measures tapping some of the processes. Greenberg finds employees when
workers experienced a 15% pay reduction, reported feeling underpaid and stole twice as much as
against when they felt equitably paid. Theft rate reduced and feelings of inequity lessened, when the
basis for pay cuts was fully explained to employees.
[14] Greenberg J: Who stole the money, and when? Individual and situational determinants of
employee theft. Organizational Behavior and Human Decision Processes 2002, 89:985–1003.
Greenberg studies the personal and situational variables affecting employee theft in a real work
environment. The final sample in the experiment consisted 270 customer service representatives
employed by a large financial services company in the US; 142 of the representatives were in an
office with an ethics program while the rest were in a distant city with no ethics program in place. He
found theft was higher with employees having lower moral development, in the office that had no
ethics program and when the company stood to be victimized, instead of coworkers.
[15] Edelman B, Larkin I: Social comparisons and deception across workplace hierarchies:
Field and experimental evidence. Organization Science 2015, 26:78-98.
Edelman and Larkin use field data and scenario-based experiments (M-Turk) to study how
unfavorable social comparisons varyingly engage employees of different hierarchical levels to engage
in deception. For the field study, the authors use data of downloads from SSRN between 2001 and
late 2007 to examine deceptive downloads of academic working papers. Their results provide the first
evidence that the impact of negative social comparison differs across work hierarchies while
confirming the importance of social comparison in the workplace. They find that successful and
long-tenured employees face a greater loss of self-esteem from negative social comparisons and are
more likely to engage in deceptive behavior in response to reported performance that is lower than
that of peers.
[16] Gino F, Pierce L: Robin Hood under the hood: Wealth-based discrimination in illicit
customer help. Organization Science 2010, 21:1176–1194.
Gino and Pierce combine data from the field and a laboratory experiment to propose that envy and
empathy lead employees to discriminate in illicitly helping customers based on customer wealth. In
the field study, they use a database of over six million emission tests from a metropolitan area (of a
large northern state) from 2001 to 2004 and identify relative levels of leniency for individual
inspectors and customer wealth based on their ownership of luxury versus standard vehicles. The
results show that individuals appear to be influenced by social comparisons when choosing to engage
in illicit behavior. They find a significant number of inspectors treated standard vehicles differently
than luxury cars, and majority of such inspectors appeared to be illegally helping customers who
exhibited less wealth.
[17] Balafoutas L, Beck A, Kerschbamer R, Sutter M: What drives taxi drivers? A field experiment
on fraud in a market for credence goods. The Review of Economic Studies 2013, 80:876–891.
Balafoutas et al. conduct a field experiment on taxi rides in Athens, Greece to measure fraud and
examine influence of passengers’ presumed income and information on the extent of fraud. Primarily
they find passengers who do not have information about optimal routes are taken on longer detours
while lack of information about the local tariff system increases the likelihood of cheating of inflated
bills by about fifteen percentage points. They use five experimenters switching between two different
income roles and three different information roles, collecting observations during two weeks in July
2010 and one week in March 2012.
[18] Piff PK, Stancato DM, Côté S, Mendoza-Denton R, Keltner D: Higher social class predicts
increased unethical behavior. Proceedings of the National Academy of Sciences of the United States of
America 2012, 109:4086-4091.
Using two field experiments and five laboratory experiments, Piff et al. argue that upper-class
individuals behave more unethically than lower-class individuals. In the first field study, the authors
investigate a busy four-way intersection with four stop signs to see if upper-class drivers were more
likely to cut off other vehicles. In the second field study, they tested if pedestrians at a crosswalk are
more likely to be cut off by upper-class drivers. Authors collected data for 274 vehicles in the first
study and 152 vehicles in the second, both in the San Francisco Bay Area.
[19] Bott K, Cappelen AW, Sørensen EØ, Tungodden B: You’ve got mail: A randomised field
experiment on tax evasion. NHH Norwegian School of Economics 2014.
Bott et al. conduct a randomized field experiment in Norway with taxpayers who according to the tax
administration may have misreported their foreign income. The authors limited the sample to 18,000
taxpayers for the income year 2011 since these taxpayers had a substantial negative deviation between
self-reported foreign income and third party reports. They find that including a moral appeal in a
letter from the tax administration nearly doubled the average foreign income reported, as compared
to a base letter without such an appeal. The moral appeal helps increasing the amount of reported
foreign income while the probability of detection increases the number of taxpayers who report
foreign income.
[20] Pruckner GJ, Sausgruber R: Honesty on the streets: A field study on newspaper
purchasing. Journal of the European Economic Association 2013, 11:661-679.
Pruckner and Sausgruber conducted a field experiment and two complementary studies to examine
honesty in the honor system of newspapers selling on the street in Austria. They find that a moral
reminder increases the level of honesty in payments but the same message has no effect on an
individual’s honesty. The authors collect data in three different ways – field experiment at 40 booth
locations over 6 days; a quasi-field experiment at 250 locations over 7 weeks and lastly using a survey
of customers who had taken a newspaper from a booth.
[21] Fellner G, Sausgruber R, Traxler C. Testing enforcement strategies in the field: Threat,
moral appeal and social information. Journal of the European Economic Association 2013, 11:634-660.
In a natural field experiment, Fellner et al. evaluate alternative strategies to enforce compliance with
the law. They use the setting of the radio license fees that funds public broadcasters in Austria, to
manipulate mailings sent out by GIS (Gebuhren Info Service) to potential license fee evaders. The
mailing contained information about a citizen’s payment duty, size of the fee and maximum fine
imposable in case of detection. The results show that the mailings had a positive impact on fee
payments in the treatment group as compared to an untreated control group. The authors suggest
that mailing may signal surveillance and thus an increased perceived risk of sanction.
[22] Levitt SD: White-collar crime writ small: A case study of bagels, donuts and the honor
system. The American Economic Review 2006, 96:290-294.
Levitt in this decade long field study recorded results of nearly 75,000 deliveries of bagels and donuts
to corporate offices. In particular, he analyzed the discrepancy between prices of goods consumed
and actual payments deposited in a lock box. The data reflects deliveries to offices over a 12-year
period from 1993 to 2004 along with a list of prices and a lockbox where customers make payments
on the honor system. He finds that payment rates fall in response to increases in the posted price,
suggesting some marginal consumers to be willing to sustain the moral cost associated paying less
than full price or not stealing goods. He also finds payment rates are higher when bagels and donuts
are uneaten, suggesting that the marginal consumer pays a lower share of the posted price than the
inframarginal consumer.
[23] Shu LL, Mazar N, Gino F, Ariely D, Bazerman MH: Signing at the beginning makes ethics
salient and decreases dishonest self-reports in comparison to signing at the end. Proceedings of
the National Academy of Sciences 2012, 109:15197–15200.
Shu et al. use field and laboratory experiments to find that signing before an opportunity instead of
after, makes ethics salient and significantly reduces dishonesty. The authors conduct the field
experiment with an insurance company in southeastern United States, with customers required to
sign an honesty statement at the beginning of an insurance form instead at the end. They collected
data from 13,488 insurance policies for a total of 20,741 cars. They find that asking customers to sign
at beginning of the form led to a 10.25% increase in calculated miles in comparison to signing at the
end; suggesting that customers were actually checking the odometer when they signed at the
beginning.
[24] Fisman R, Miguel E: Corruption, norms, and legal enforcement: Evidence from diplomatic
parking tickets. Journal of Political Economy 2007, 115:1020–1048.
Fisman and Miguel analyze parking behavior of United Nations officials in Manhattan to study
cultural norms and legal enforcement in controlling corruption. The authors use field data of parking
violations covered from November 1997 to November 2005 from New York City Department of
Finance, Kaufmann et al. data on country corruption levels and also employ country-level data for
economic, political and social characteristics. The sample was restricted to all countries that had 1998
populations greater than 500,000. They find a strong effect of corruption norms that diplomats from
high-corruption countries accumulated significantly more unpaid parking violations. In 2002 when
enforcement authorities acquired the right to confiscate diplomatic license plates of violators,
following which unpaid violations by diplomats dropped sharply in response.
[25] Zitzewitz E: Does transparency reduce favoritism and corruption? Evidence from the
reform of figure skating judging. Journal of Sports Economics 2014, 15:3–30.
In this field study, Zitzewitz examines data from the International Skating Union (ISU) following
vote trade scandals in the 1998 and 2002 Olympics. The ISU introduced a number of changes to its
judging system in order to disrupt collusion by groups of judges. For the study, the author considers
three samples: one from prior to 2002 scandal, one for events in 2002-04 which retained the 6.0
system and having judge anonymity, and the third sample with events from both the code of points
system and judge anonymity. As a result of aggregate favoritism and corruption (vote trading), the
results suggest that the compatriot-judge effect increased slightly after the reforms.
[26] Olken BA: Corruption and the costs of redistribution: Micro evidence from Indonesia.
Journal of Public Economics 2006, 90:853–870.
Olken examines corruption in a large Indonesian anti-poverty program that distributed subsidized
rice to poor households. In particular, he compares survey data on the amount of rice actually
received by households with the administrative data on amount of rice distributed. The survey
analysis is done using two datasets, one a detailed survey of 100 villages and the other a nationally
representative survey of 200 Indonesian districts. The author finds ethnically heterogeneous and
sparsely populated areas have higher likelihood of missing rice. He also finds at least 18% of rice in
the program disappeared due to corruption, substantial enough to make a welfare enhancing program
become welfare reducing on net.
[27] Yenkey CB: Tribes and (dis)trust. University of Chicago Working Paper. Available at SSRN:
http://ssrn.com/abstract=2562539.
In this paper, Yenkey examines the socially diverse and contentious population of investors in
Kenya’s emerging stock market. The author uses data from Nairobi Securities Exchange (NSE) back
office databases maintained by Central Depository and Settlement Corporation (CDSC) and archival
records from Kenya Investor’s Compensation Fund. He analyzes investor-level data to suggest that
investors choose a same-tribe stockbroker and avoid rival-tribe brokers, particularly in areas of intertribal violence and where one’s tribe is a minority. On the other hand, a corrupt stockholder is more
likely to steal from clients of his own tribe instead of rival tribes; and the stockbrokers are twice as
likely to steal from their own tribe members at the top of the wealth distribution rather than the
bottom.
[28] Bianchi E, Mohliver AC: Do good times breed cheats? Entering the workforce in an
economic boom predicts later unethical behavior. London Business School Working Paper.
Bianchi in this study argues that during an economic boom, entering the workforce will increase the
likelihood that people will engage in ethically questionable practices later in their careers. The authors
use data from 14,872 executives and directors of publicly traded companies in the US between 1996
and 2005; executives who received at least one stock option grant during this period. They find that
people who began their careers in economically prosperous times are more likely to backdate stock
options later in their careers. They also suggest that at a formative period in life, environmental
conditions can affect ethical impulses for a long time.
[29] Schneider HS: Agency problems and reputation in expert services: Evidence from auto
repair. The Journal of Industrial Economics 2012, 60:406–433.
With undercover visits to auto repair garages using a test vehicle, Schneider conducts a field
experiment to examine how asymmetric information between motorists and mechanics over auto
repair service quality affects outcomes. With the help of mechanics working for a Canadian public
interest group, data is derived from a study conducted at forty independent auto repair shops in four
towns in two Connecticut counties. The author also uses data from repair receipts in 51 undercover
garage visits conducted by the Canadian group. Results are largely consistent with large agency
problems that the mechanic is not likely to present efficient service whether a customer represents
one-time or repeat business.
[30] Snyder J: Gaming the liver transplant market. Journal of Law, Economics, and Organization 2010,
26:546–568.
In a field study of the liver transplant market, Snyder uses a policy change to examine changes in ICU
(intensive care unit) admission behavior. The policy change in effect after March 1, 2002 mandated
that liver transplant waiting lists were not to be influenced by ICUs. The author uses data provided
by the United Network for Organ Sharing (UNOS), a comprehensive database on all liver transplants
performed in the US from mid-1987 till the end of 2008. The results suggest that post-reform, areas
with multiple transplant centers saw a reduction in ICU usage. In competitive areas, he finds that in
order for the sickest patients to receive a liver, the waiting list seems to be manipulated.
[31] Jacob BA, Levitt SD: Rotten apples: An investigation of the prevalence and predictors of
teacher cheating. The Quarterly Journal of Economics 2003, 118:843-877.
Jacob and Levitt use data from Chicago public schools to find serious cases of teacher or
administrator cheating on standardized tests each year in at least 4-5 percent of elementary school
classrooms. To detect teacher cheating, they develop an algorithm that combines information on
expected score fluctuations and suspicious patterns of student answers in a classroom. Data for this
study comes from Chicago Public Schools administration which includes detailed question-answers
for every student in grades 3 to 8 taking the Iowa Test of Basic Skills (ITBS) from 1993 to 2000.
Their results primarily suggest that teacher cheating appears responsive to relatively minor changes in
incentives.
[32] Zitzewitz E: Forensic economics. Journal of Economic Literature 2012, 50: 731-769.
Zitzewitz reviews a new emerging field of forensic economics that uncovers evidence on hidden
behavior in a variety of domains. This article focuses on uncovering evidence of wrongdoing across
different research questions, drawing out common threads and opening possibilities of further
research across fields. The author chooses to deal with economically important hidden behavior
which may usually be considered least unethical; examples include road builders skimping on
materials, violations of UN sanctions, unnecessary heart surgeries and racial biases in employment
decisions.
[33] Pierce L, Snow DC, McAfee A: Cleaning house: The impact of information technology
monitoring on employee theft and productivity. Management Science (forthcoming).
Pierce et al. look at how firm investments in technology-based employee monitoring impacts both
misconduct and productivity. The data for the field study obtained from a restaurant system IT
vendor includes five restaurant chains (firms) with a total of 392 restaurant locations in 38 American
states from March 2010 to February 2012. The data includes weakly theft data in addition to all
transactions at each restaurant for the two year period. They find significant treatment effects in
reduced theft and improved productivity driven by changed worker behavior rather than worker
turnover. The results in the study suggest that employee misconduct is not solely a function of
individual differences in ethics or morality, but can be influenced by managerial policies that can
benefit both firm and individual.
[34] Nagin DS, Rebitzer JB, Sanders S, Taylor LJ: Monitoring, motivation, and management:
The determinants of opportunistic behavior in a field experiment. The American Economic Review
2002, 92:850-873.
In a field experiment, Nagin et al. examine the rational cheater model by observing how
experimentally induced variation in monitoring employees of a telephone call center influences
opportunism. The experiment was run at a telephone marketing firm which was geographically
dispersed at 16 sites; the job of the employees at these sites was to call potential donors and request
contributions. Cheating is detected by measuring suspicious bad calls (SBC) during a week. The
authors find a considerable fraction of employees behave according to predictions of the rational
cheater model – employees engage in malfeasance at a higher rate in response to reduced monitoring.
Evidence from a survey of employees who responded to reduced monitoring shows that these
employees perceived their employer as being uncaring and unfair.
[35] Duflo E, Hanna R, Ryan S: Incentives work: Getting teachers to come to school. The
American Economic Review 2012, 102: 1241-1278.
Duflo et al. use a randomized field experiment and a structural model to examine if monitoring and
financial incentives can increase learning and reduce teacher absenteeism. Out of 113 schools in
India, the authors randomly choose 57 to verify a teacher’s daily attendance through photographs
with the help of time and date stamps; and salary was a non-linear function of attendance. In the
schools with treatment effect, they found teacher absence rate fell from 42 percent to 21 percent
compared to non-treatment schools. They suggest that teacher response was almost entirely due to
financial incentives while the program improved child learning.
[36] Hertzberg A, Liberti J, Paravisini D: Information and incentives inside the firm: Evidence
from loan officer rotation. The Journal of Finance 2010, 65:795–828.
Hertzberg et al. in this field study of a commercial bank argue that reassigning tasks among agents
can alleviate moral hazard in communication. The authors study loan officers of a commercial bank
to find the effects of officer rotation policy on their reporting behavior. From the internal records of
a large multinational US bank operating in Argentina, the authors analyze the reporting behavior of
loan officers in the small and medium business division. The sample covers 7-year period from
December 1997 to 2004, 1248 firms and 100 loan officers. They find that reports are more accurate
when an officer anticipates rotation, contains more bad news about borrower’s repayment prospects
and affects decisions of bank lending.
[37] Slemrod J, Blumenthal M, Christian C: Taxpayer response to an increased probability of
audit: Evidence from a controlled experiment in Minnesota. Journal of Public Economics 2001,
79:455–483.
Slemrod et al. conduct a controlled experiment regarding income tax compliance of 1,724 randomly
selected Minnesota taxpayers in 1995. This control group of 1,724 taxpayers were informed by a
letter that they their file would be closely examined. The Minnesota Department of Revenue
provided data of taxpayers for income years 1993 and 1994. They find that low and middle-income
taxpayers on average increased their tax payments compared to the previous year, indicating the
presence of noncompliance. They also find reason to suspect that high-income taxpayers may react
by reporting lesser income compared to the previous year, probably perceiving that final outcomes
are based on initially reported incomes.
[38] Kleven HJ, Knudsen MB, Kreiner CT, Pedersen S, Saez E: Unwilling or unable to cheat?
Evidence from a tax audit experiment in Denmark. Econometrica 2011, 79:651-692.
Kleven et al. analyze a tax enforcement field experiment of over 40,000 individual incomes in
Denmark. In the experiment, one group (half) of tax filers were randomly selected for an audit in the
second year and in the following year, tax filers in two groups were randomly sent threat of audit
letters. The authors use data from the Danish tax administration’s (SKAT) Business Object Database
for incomes reported during 2006 and 2007, which includes all income items from third-party reports
and are pre-populated, filed and audited tax returns for each year and taxpayer. The authors find
substantial tax evasion rate for self-reported income as compared to third party reporting. They also
find that random prior audits and threat-of-audit letters have significant effect on self-reported
income but no effect on third-party reported income.
[39] Olken BA: Monitoring corruption: Evidence from a field experiment in Indonesia. Journal
of Political Economy 2007, 115: 200-249.
Olken conducts a randomized field experiment in Indonesia on monitoring corruption in over 600
Indonesian village road projects. Data for this study comes from World Bank funded Kecamatan
Development Projects (KDP) in two of Indonesia’s most populous provinces of East Java and
Central Java, and collected between September 2003 and August 2004. He finds that even in a highly
corrupt environment, increasing government audits can play an important role in reducing
corruption. In particular, he finds an eight percentage points reduction in missing expenditures, as
measured by discrepancies between official project costs and an independent engineers’ estimate of
costs.
[40] Peisakhin L: Transparency and corruption: Evidence from India. Journal of Law and Economics
2012, 55:129–149.
Peisakhin examines the relative effectiveness of using the freedom-of-information law, compared to
bribery, for middle and lower class individuals applying for public services. The author uses data
from field experiments with New Delhi’s middle and low income residents and slum dwellers, and
finds that applicants who use the freedom-of-information law attain the same rate of success and
those who employ bribery, suggesting that the law erases class differences in procuring public
services.
[41] Bernstein ES: The transparency paradox a role for privacy in organizational learning and
operational control. Administrative Science Quarterly 2012, 57:181–216.
Bernstein examines the impact of transparent organizational design on workers’ productivity and
performance using a field experiment and data from embedded participant-observers at the second
largest mobile phone factory in China. Two of the sixteen production lines were used for
experimental condition where curtains were installed to instill privacy, while the author tracked
hourly production and quality data using the company’s management information systems. The
findings from the experiment argues for the transparency paradox, that zones of privacy under
certain conditions increases performance; under observability, employees conceal their activities
through codes and other costly means inhibiting performance.
[42] Yap AJ, Wazlawek AS, Lucas BJ, Cuddy AJC, Carney DR: The ergonomics of dishonesty:
The effect of incidental posture on stealing, cheating, and traffic violations. Psychological Science
2013, 24:2281–2289.
Yap et al. in this article study the impact of physical environments on dishonest behavior using a field
experiment, two laboratory studies and one observational field study. In the first field experiment,
eighty-eight members were recruited from South Station in Boston and outside Columbia University
to study the relationship between stretching and impression formation. In the observational field
study to test the real-world generalizability of the incidental-posture effect, the authors use
information of legally parked vehicles blocked by double-blocked cars. Overall, the four studies
together provide evidence that expansive postures inadvertently lead to feeling more powerful and
these feelings can lead to dishonest behavior.
[43] Olken BA, Barron P: The simple economics of extortion: Evidence from trucking in Aceh.
Journal of Political Economy 2009, 117:417–452.
In this field study, Indonesian truck drivers are accompanied by surveyors on 304 trips during which
they observe over 6,000 illegal payments to soldiers, police and weigh station attendants. During the
trips, drivers spent an average USD 40 per trip on bribes, extortion and protection payments
accounting to 13 percent of total cost of the trip. Two major routes were studied and data was
collected between November 2005 and July 2006. The authors use plausible exogenous changes in
number of checkpoints to show market structure affects the level of illegal payments. They find that
a reduction in number of checkpoints for bribes resulted in an increase in average bribes at the
remaining points along a route.
[44] Jin GZ, Kato A: Price, quality, and reputation: Evidence from an online field experiment.
The RAND Journal of Economics 2006, 37:983–1005.
Jin and Kato in this online field experiment look at internet auctions to study the link between price,
quality, seller claims and seller reputation. The purchase and grading procedures of ungraded baseball
cards were designed to mimic actual practices in the market. Also to avoid changes in value of cards
due to players’ performance, the authors restricted the purchase of baseball cards from December
2001 to March 2002. They find some buyers in the market are not only misled by incredible claims of
quality, but also pay higher prices for lower quality and are often defrauded. They also find online
seller reputation is effective for identifying good-faith sellers; high-claim sellers often target lessexperienced buyers and reputable sellers provide better quality conditional on completed auctions.
[45] Pierce L, Snyder JA: Unethical demand and employee turnover. Journal of Business Ethics 2015,
doi: 10.1007/s10551-013-2018-2.
Pierce and Snyder examine how the unethical behavior of employees are influenced by their
organizations. They use data from over three million vehicle emissions tests in the northeastern US
to find strong evidence of ethical spillovers from firms to individuals. The results suggest that
employee behavior can be influenced by managers using formal norms and incentives, but employees
have persistent ethics that limits the magnitude of influence. The authors find that these spillovers
are strongest at corporate chains and large facilities, and weakest for large-volume inspectors.
[46] Simonsohn U: Just post it: The lesson from two cases of fabricated data detected by
statistics alone. Psychological Science 2013, 24:1875–1888.
Simonsohn in this article describes two cases of suspected academic fraud exclusively using statistical
analysis of reported means and standard deviations. Analyzing raw data of published results
confirmed his initial suspicions of data fabrication and suspected fraud. He argues that availability of
raw data makes the detection of fraud easier and more diagnostic, thus making the task of fabrication
more difficult and intimidating. The author investigates the following two articles – Sanna, Chang,
Miceli and Lundberg (2011) and Smeesters and Liu (2011).
[47] Simmons JP, Nelson LD, Simonsohn U: False-positive psychology: Undisclosed flexibility
in data collection and analysis allows presenting anything as significant. Psychological Science
2011, 22:1359–1366.
Simmons et al. argue that empirical work in psychology is very easy to publish “statistically
significant” evidence that can contain any hypothesis. They show that flexibility in data collection,
analysis and reporting dramatically increases false-positive rates, which is in spite of a low rate of
false-positive findings (≤ 0.05). The authors suggest an effective disclosure-based simple and low
cost solution to the problem, which involves six concrete requirements for authors and four
guidelines for reviewers.
[48] Bennett VM, Pierce L, Snyder JA, Toffel MW: Customer-driven misconduct: How
competition corrupts business practices. Management Science 2013, 59: 1725-1742.
The authors exploit variation in local competition among vehicle emissions testing facilities to
demonstrate that the demand for fraudulent leniency from customers pressures expert inspectors to
illegally pass gross polluting cars. The paper argues that under intense competition for customers,
firms will be driven to cross ethical and legal boundaries when dishonesty is a valued element of
quality for customers. The authors use individual-level test and inspector data, combined with
discontinuous pass thresholds and geographically isolated small-town markets, make causal
arguments about competition and fraud.
[49] Mayer DM, Nurmohamed S, Treviño LK, Shapiro DL, Schminke M: Encouraging employees
to report unethical conduct internally: It takes a village. Organizational Behavior and Human
Decision Processes 2013, 121:89–103.
Using two field studies and a laboratory study, Mayer et al. examine how supervisory ethical
leadership within an organization is associated with employees reporting unethical conduct. One field
study was conducted in a multinational company headquartered in the US, with participants who
were newly-hired employees responsible for selling food to new and current customers in their
assigned delivery areas. In the second field study, the authors sample 33,756 employees of 16
manufacturing and technology companies with the question if an employee has observed unethical
behavior in the past; a total of 6,554 employees responded ‘yes’ to the question. They find that both
supervisor and coworker have an influence on employees’ internal whistle-blowing.
[50] Derfler-Rozin R, Moore C, Staats BR: Enhancing ethical behavior through task variety.
London Business School Working Paper.
Derfler-Rozin et al. use data from the field and a laboratory experiment to argue that task design
affects rule breaking in the workplace. For the field study, they use data from the home loan
applications-processing operations of a Japanese bank collected between June 2007 and December
2009. In order to capture rule breaking, the authors examine whether employees take a lunch break
that exceeds the length permitted under bank regulation and test it against task variety. They find that
task variety supports rule compliance through deliberative thinking, when conceptualized in terms of
increasing number of switches between different types of subtasks.
[51] Rees-Jones A: Loss aversion motivates tax sheltering: Evidence from U.S. tax returns.
University of Pennsylvania Working Paper.
Rees-Jones uses data from the 1979-190 IRS Panel of Individual Tax Returns to show evidence that
loss aversion affects how taxpayers treat losses differently from gains in tax sheltering decisions. He
shows evidences of bunching and shifting in the distribution of tax returns near zero that cannot be
explained by structural elements of the tax code and is consistent with prospect theory. Although
legal and illegal sheltering methods cannot be fully disentangled, the identified behavioral patterns
almost certainly reflect widespread dishonesty motivated by loss aversion, given past evidence on
base rates of tax fraud.