Daylight Saving Time and Police Harassment 1
Running Head: DAYLIGHT SAVING TIME AND POLICE HARASSMENT
Working Paper
Law and Error: Daylight Saving Time and Police Harassment
David T. Wagner
Lundquist College of Business
University of Oregon
Christopher M. Barnes and Cristiano Guarana
Foster School of Business
University of Washington
This manuscript is a working paper. Please do not circulate without the permission of the
authors.
Daylight Saving Time and Police Harassment 2
Abstract
In this study we examine decision-making among law enforcement officers, paying particular
attention to instances of police officer harassment of suspects. With two large databases – one
from the Los Angeles Police Department (LAPD) and the other from the Federal Bureau of
Investigation (FBI) – we examine how the shift to daylight saving time impacts law enforcement
officer decisions. Notably, we find that on the Monday after the shift to daylight saving time—
which is known to produce a sleep decrement—officers are more prone to exhibit racial bias in
their harassment of suspects, suggesting that the officers have less self-regulatory ability to
suppress racial biases when sleep deprived. We briefly discuss the implications of our findings
for individual decision making and federal and organizational policy.
Daylight Saving Time and Police Harassment 3
LAW AND ERROR: DAYLIGHT SAVING TIME AND POLICE HARASSMENT
Human decision making is often plagued by errors that manifest themselves in the form
of heuristics and biases (Tversky & Kahneman, 1974). Although decision making errors in any
context are likely to yield inefficiency or suboptimal outcomes (e.g., Stauffer & Buckley, 2005),
the context in which decisions are made can have a leveraging effect on the magnitude of the
consequences of the decision errors. One highly visible and highly critical context for such
biases is law enforcement work. Law enforcement is an especially salient context in which to
study biases because law enforcement requires the exercise of a substantial amount of human
judgment in contexts that are demanding and fast-paced. Understandably, these factors at times
lead to errors, some of which can be attributed to cognitive biases. Indeed, when a police officer
errs – for instance, shooting and killing an unarmed individual – accusations of bias often follow
(Wines, 2014).
Although such dramatic events often garner headlines and dominate public discussion,
other police decision-making errors are less dramatic, more numerous, and can occur in run-ofthe-mill settings. Further contributing to the potential impact of such a large number of decisions
is that fact that many of these decisions afford decision makers (in this case police officers)
substantial discretion. The existence of such discretion gives birth to the possibility that such
decisions could be interpreted as the inappropriate exertion of power or authority, even to the
extent that these actions could be deemed police harassment. For our purposes we define police
harassment as discretionary behavior – such as stops, searches, or arrests – that are not necessary
given the facts of the situation, and that result in no subsequent legal action toward the suspect.
Examples of this form of harassment could include automobile searches that fail to uncover
contraband, arrests that are ultimately not prosecuted for lack of sufficient cause or gravity.
Daylight Saving Time and Police Harassment 4
Although the forms of police harassment described herein are certainly less grave than
fatal shootings, they can nonetheless incur costs to the individual. For example, being falsely
arrested can result in substantial social costs such as humiliation or stigma (Newman, 2006).
Thus, not only do the costs of law enforcement decision-making failures result in grave
outcomes such as death or maiming, but they can spur police harassment that results in
psychological, social, and other personal costs to civilians.
The incidence of such decision failures might reasonably be chalked up as an
unavoidable and accepted cost of living in an organized society that enjoys social goods such a
police protection. Such a characterization of these flaws might be reasonable if the costs were
equitably distributed across the population. However, repeated studies suggest this is not the
case. For example, a one-year study of metropolitan police officers in the city of Los Angeles
found that police officers stopped and unnecessarily searched or frisked a disproportionate
number of racial minorities (Ayres & Borowsky, 2008). Meta-analytic results of experimental
studies have found corroborating evidence indicating that jurors tend to treat Black defendants
more harshly, both in terms of verdicts and punishments, than they do White defendants
(Mitchell, Haw, Pfeifer, & Meissner, 2005). What these findings indicate is that researchers have
already uncovered some level of racial bias in the American justice system.
Researchers suggest that much of these effects can be attributed to individual level biases
(Plant & Devine, 2009). However, there is also reason to believe that less examined yet
systematic macro-level factors could also contribute to the incidence of police harassment. The
reason for this supposition is that, although most people have some level of bias toward others,
they are generally able to self-regulate their behavior such that these biases are not evidently
manifest in their behavior (Plant & Devine, 2009). However, when individuals’ self-regulatory
Daylight Saving Time and Police Harassment 5
abilities become depleted or are otherwise hampered due to a mismatch between self-regulatory
demands and the individuals’ current temporal environment (Bodenhausen, 1990), they are less
able to suppress their biases, making the biases much more likely to surface (Ghumman &
Barnes, 2013). While many intra-individual factors can drive or deplete an individual’s state selfregulatory ability (Baumeister, Vohs, & Tice, 2007), recent studies have uncovered that sleep
influences self-control (Barnes, Schaubroeck, Huth, & Ghumman, 2011; Wagner, Barnes, Lim,
& Ferris, 2012). This opens the question of what factors influences sleep in a manner that would
shape the way that individuals suppress their biases.
In many countries across the globe, daylight saving time marks the shift from standard
time to daylight saving time with a phase advance, which is the turning ahead of the clocks by
one hour. When the clocks shift to daylight saving time, people’s circadian rhythms do not
immediately respond to the change (Kantermann, Juda, Merrow, & Roenneberg, 2007), meaning
that people’s sleep-wake cycles no longer match the pacing of the clock. The implication of this
is that people are likely to experience a disruption to their sleep, meaning that on average, people
get approximately forty minutes less sleep the Sunday to Monday night after the spring change to
daylight saving (Barnes & Wagner, 2009). In light of the findings that individuals lack adequate
self-regulation to suppress their prejudicial judgments or behaviors when sleep deprived
(Ghumman & Barnes, 2013), the clock shift associated with daylight saving time begins to look
more insidious than saving, causing us to more closely consider the potentially harmful macrolevel factor.
In the sections that follow, we elaborate upon the nature of police harassment and then
explore the mechanisms connecting the shift to daylight saving time to these behaviors. We put
Daylight Saving Time and Police Harassment 6
forward hypotheses articulating these relationships and test them with two quasi-experiments.
We conclude with a discussion of our findings for theory, practice, and public policy.
Unnecessary Stops, Searches and Arrests
Law enforcement officers have discretion with regard to whom they will stop, whom they
will search, and whom they will arrest. In most cases, officers only have legal authority to stop
someone who has broken a law or who otherwise presents probable cause of having broken a
law. However, stop and frisk laws allow law enforcement particular latitude in who they will
stop and search. In fact, if a law enforcement officer has reasonable suspicion that an individual
is breaking a law, the officer can stop and search that person and his or her property. This
suggests that there is substantial judgment that feeds into law enforcement decision making.
Although many of these discretionary searches prove successful at identifying and eliminating
contraband, a recent study of police stops and searches found that there was significant racial
bias in who was stopped and in who among those stopped was subsequently searched (Ayres &
Borowsky, 2008).
In addition to the racial bias evident in traffic stops resulting in the search of drivers and
their vehicles, bias might also be evident in police officer decisions to arrest people for
inconsequential violations of the law. For example, many times a civilian might break a law,
such as jaywalking on a quiet street, or playing one’s music too loudly. In such situations, police
officers have a wide range of actions in which they could engage to address the violation. One
option is for the officer to ignore the violation on the basis of its immaterial nature, whereas
another option would be to ask the civilian to correct the behavior and not engage the civilian
further. Yet another option is to treat the violation explicitly as allowed by the law, which could
include stopping the civilian, formally citing the individual, and even placing the individual
Daylight Saving Time and Police Harassment 7
under arrest. Because police officers have the legal authority to carry out any of the above
approaches to addressing breaches of the law, their responses to such (and all) situations entails
substantial discretion. However, despite the legal force that could attend any of these actions,
local norms may dictate a particular course of action in response to a given violation. This
suggests that there may be a tension between what is legally defensible and what is socially
normative. The normative nature of these arrests is often addressed by the complementary arm of
the justice system that includes legal prosecutors.
This tension is directly addressed by a legal system that is designed to counterbalance the
actions taken by law enforcement, with the aim of enhancing both civil security and justice.
Naturally then, occasions should arise wherein the legal system is able to correct faulty decisions
or unnecessary arrests made by law enforcement officials. One way to accomplish this is for the
prosecutor – who ensures that criminals are tried and, if appropriate, punished – to
“exceptionally clear” an individual instead of legally trying them for their crime. This means that
the individual, while presumably guilty of committing a crime, is released because the crime was
so insignificant as to not merit full prosecution by the legal system. In other words, individuals
who are exceptionally cleared for breaking laws have generally made the type of relatively
benign violation of the law that generally goes unpunished. In light of these normative behaviors
which would typically lead a police officer to forego arrest or formal charges against a civilian
who has committed an inconsequential infraction, we consider those instances in which a civilian
is arrested, but later exceptionally cleared, as cases of police harassment. This means that,
although the suspect appeared to have violated some law, the police officer’s behavior was so far
outside the norms of typical law enforcement as to be construed as harassment (Federal Bureau
of Investigation, 2009).
Daylight Saving Time and Police Harassment 8
Daylight Saving Time and Police Harassment
Any number of reasons could lead to an outcome in which the prosecution of a suspected
offender is declined, and these outcomes could be experienced by anyone, regardless of race.
However, related research gives us reason to believe that there may be systematic factors that
result in a disproportionate number of black suspects whose cases are exceptionally cleared
without prosecution. In other words, there may be reason to believe that black individuals are
systematically harassed more than white individuals. Although this outright finding has been
suggested by other studies (e.g., Ayres & Borowsky, 2008), we propose that macro-level factors
may also play a role in the decision-making processes that lead black citizens to be
disproportionately harassed by police.
The basis for this expectation is that people may harbor some level of bias in their
internal decision making, simply due to their life experiences or upbringing (Dovidio,
Kawakami, & Gaertner, 2002). However, people are typically motivated by either internal or
external forces to inhibit the expression of these biases (Plant & Devine, 2009) either because
they disagree with these biases, or because they recognize that such biases are not socially
appropriate. One factor that determines the efficacy of these impulse-restraining efforts is the
individual’s available stock of self-regulatory resources. Self-regulation deals with the ability to
restrain impulses that are preferable or spontaneous or in the short term in favor of decisions that
have larger and typically longer-term benefits. Self-regulatory resources are depleted throughout
the day as people make these decisions and because everyone makes decisions throughout the
day that deplete these resources, it is necessary to restore one’s resources on a daily basis
(Baumeister et al., 2007). Due to the regulation-restoring effects of sleep, it follows that the
disruption of sleep would be associated with a decrement in an individual’s ability to regulate
Daylight Saving Time and Police Harassment 9
behavior, and empirical research has borne out this prediction (e.g., Christian & Ellis, 2011;
Davis & Leo, 2012; Lanaj et al., 2014).
Although these studies highlight how insufficient sleep can result in impulsive or underregulated behavior, they largely focus on individual behaviors or decisions that impact selfregulation. By contrast, the present study seeks to uncover more insidious forces that not only
affect individuals, but that do so at a national level. Policy governing daylight saving time is just
such a force. The practice of advancing the clocks in a way that shifts light from morning to
night represents a shock to people’s biological rhythms. Because people cannot immediately
shift their internal clocks to match the new pacing of the external clock, sleep patterns no longer
match the pattern of daily activity, which often results in the truncation of sleep or the disruption
of established sleep patterns in a way that leaves individuals less well rested (Czeisler et al.,
1999). Estimates of this effect have shown the Americans tend to sleep 40 minutes less on the
Sunday to Monday night following the spring shift to daylight saving time (Barnes & Wagner,
2009; Kantermann, Juda, Merrow, & Roenneberg, 2007). Although this amount of sleep
deprivation might seem benign, such nominal changes in sleep quantity have been found to
dramatically influence individual behavior (Harrison & Horne, 2000).
Particularly germane to the present study is a series of results uncovered by Ghumman
and Barnes (2013). In a series of studies, these scholars found that even relatively modest
variations in sleep had meaningful outcomes related to prejudice. For instance, they found that
naturally occurring variation in sleepiness predicted the extent to which subjects would freely
describe a Muslim woman pictured in a photograph with stereotypical comments. In a follow up
study, they found that when asked to assess the suitability of a job applicant for a managerial
position, sleepier subjects found the candidate to be less capable to performing the job, but only
Daylight Saving Time and Police Harassment 10
when the candidate was black (there was no effect of subject sleepiness on these assessments
when evaluating a white candidate). Further highlighting the importance of sleep in this process,
was their final study, which showed that individuals with pre-existing racial biases scored much
higher on a measure of modern racism than did biased individuals who were well rested. The
explanation for these findings is that, although many historical indicators of prejudice might no
longer exist in most societies, many people appear to simply be suppressing their prejudice.
Given the relationship between self-regulatory strength and impulsive behavior, we suspect that
when people are deprived of sufficient sleep, they are likely to exhibit lower self-regulatory
strength, and this vacuum of self-control makes the expression of impulsive prejudicial behavior
much more likely.
In light of the evidence that inadequate sleep is likely to result in greater prejudicial
behavior, and that the national shift to daylight saving time results in a modest amount of sleep
loss on the Sunday to Monday night following the phase advance, we expect that prejudicial
behaviors will be more likely on the Monday following the shift to daylight saving time.
Although this prejudice can take many forms, in the context of law enforcement, we expect that
some of this prejudice will be evident in police harassment of individuals who police officers
might typically ignore or treat more leniently. On the basis of this theory and research, we
propose that the macro-level shift to daylight saving time will have an impact on the incidence of
police harassment, especially with regard to black suspects.
Hypothesis 1: A greater incidence of police harassment will occur on the Monday after
the shift to daylight saving time than on comparison days.
Hypothesis 2: Blacks will be more likely to experience police harassment on the Monday
after the shift to daylight saving time than on comparison days and White suspects will
not experience such an increase in police harassment on this day.
Daylight Saving Time and Police Harassment 11
To test our hypotheses we present two studies, each utilizing the daylight saving time
quasi-experiment to examine the impact of lost sleep on police harassment. In the first study we
draw from a database of traffic stops by police officers in the Los Angeles, California Police
Department (LAPD) to examine our first example of police harassment – automobile and
personal searches from which no contraband was found – for black as opposed to white drivers.
In the second quasi-experiment we draw from a database of arrests across the entire United
States, compiled by the Federal Bureau of Investigation (FBI), which indicates the outcomes of
every arrest made by every law enforcement agency across the country for the years 1991 to
2011. With these data we are able to assess the extent of police harassment, operationalized as
arrests after which the individuals were exceptionally cleared because prosecutors declined to
prosecute the suspects, with emphasis given to the difference in harassment directed toward
black as opposed to white suspects, on the Monday following the shift to daylight saving time.
STUDY 1
The City of Los Angeles and the U.S. Department of Justice authorized the LAPD to
systematically collect data on stops in order to investigate whether racial profiling by police
officers occurred beyond anecdotal evidence. The data collection effort started on July 1, 2003
and finished on June 30, 2004 (the City’s fiscal year). After every stop police officers filled out
reports containing detailed information regarding stops, searches, outcomes, officers’
demographics, suspects’ demographics, and reporting districts. The initial dataset was made
publicly available by Professor Ian Ayres at http://islandia.law.yale.edu/ayres/indexcivil.htm.
The data collection effort resulted in a database of 799,814 stops conducted by 6,027
LAPD police officers. Officers had an average age of 35 years (SD=7.47), and were primarily
White (40.2%) or Hispanic (38.4%), with a substantial number of officers reporting Black
Daylight Saving Time and Police Harassment 12
(10.9%) or Asian (8.5%) ethnicity. The racial composition of suspects was 39% Hispanic, 29.2%
White, 23.5% Black, 6.5% Asian, and 1.6% members of “other” races. Of the 799,814 stops,
208,743 (26.1% of stops) resulted in the search of the individual or the individual’s vehicle, and
of these searches, 66,351 (8.3% of stops) resulted in the discovery of weapons, drugs, alcohol,
stolen vehicle or property, evidence of a crime, illicit money, or other contraband. In order to
maintain consistency between our first and second studies, we focused on searches conducted on
White and Black suspects only (99,982 searches, and 27,674 discoveries).
Measures
Police harassment. After conducting searches, police officers had to report whether they
found any weapons, drugs, alcohol, a stolen vehicle or property, evidence of a crime, illicit
money, or other contraband. We operationalize police harassment as a search that did not
uncover any form of contraband. Although police work is likely to have some level of
uncertainty, a systematic bias toward searching certain types of drivers or pedestrians could
result in the accumulation of searches that fail to uncover any contraband. Because police
officers have discretion over whether they will conduct a search, especially when it is not
immediately evident that illegal activity is underway, a failed search is suggestive of police
harassment because it could be conducted even without clear indication that laws have been
broken or that contraband is present. Events were coded as “1” if the stop was deemed
harassment and “0” if the police officer discovered any form of contraband, suggesting that the
stop was warranted and thus not considered harassment.
Daylight saving. As noted above, previous research (Barnes & Wagner, 2009) indicates
that it is the Monday after the shift to daylight saving time in which people are short on sleep
(i.e., Sunday night they sleep less). Thus, our focus is on the Monday immediately following the
Daylight Saving Time and Police Harassment 13
Spring time change. In 2004 the first Monday following the shift to daylight saving time was
April, 5th. We coded this day as “1” and the remaining days in the database as “0.”
Race. Black suspects were coded as “1,” whereas White suspects were coded as “0.”
Control variables. In our analyses we controlled for several variables that could have an
impact on the incidence of searches or police harassment. Geographic regions with a relatively
higher proportion of Black residents affords officers more opportunities to engage in harassment
toward Black drivers or pedestrians, and thus we created a set of 17 dummy variables to account
for geographic region effects. There could also be a number of justifications for an officer to stop
an individual, and these could differentially impact the likelihood of the officer searching the
stopped individual. For example, it seems reasonable to expect that an individual who is stopped
because he or she fled from police is more likely to be carrying contraband, as opposed to
someone who is driving with a faulty turn signal or some other municipal code violation. We
therefore controlled for the reason for the stop by including a set of 10 dummy variables.
Likewise, the emergence of violence during the police-civilian interaction, and the existence of a
warrant for the individual’s arrest, are both likely to lead to a greater incidence of searches after
stops by police, and thus we controlled for these factors. Other factors that might play into
decisions to search or not search stopped suspects include daily, weekly, and seasonal timing –
i.e., day of week, week of year, and time of day (Kouchaki & Smith, 2014) – and thus we created
dummy variables for these factors. Finally, we controlled for officer race and age.
Results and discussion
We tested our hypotheses via logistic regression. Table 1 shows the means, standard
deviations, and correlations for the hypothesized variables. Police harassment and race are
Daylight Saving Time and Police Harassment 14
positively correlated, whereas police harassment and daylight saving time are not significantly
related.
Hypothesis 1 predicted that daylight saving time would be positively related to the
incidence of police harassment. After controlling for the variables described above, we found no
support for the hypothesized main effect (exp b = 1.02, SE = .15, p = .88), failing to support
Hypothesis 1. Next, we tested for Hypothesis 2, which predicted that the shift to daylight saving
time interacts with suspect race to predict police harassment. We adopted a stepwise approach to
test for this hypothesis (Aiken & West, 1991; Cohen & Cohen, 1983). In step 1, we entered the
control variables. In step 2, we added the independent variable (i.e., daylight saving time) and the
moderator (suspect race), which explained a significant amount of variance (Δχ2 = 55.13, p˂.01).
The results showed that suspect race has a significant effect on unnecessary stops and searches
(exp b = 1.16, SE = .09, p ˂ .01), which is consistent with the findings of Ayres and Borowsky
(2008). Finally, in step 3, we entered the interaction term. This final model explained additional
variance (Δχ2 = 3.83, p˂.05) with results indicating that the shift to daylight saving time interacts
with suspect race to predict police harassment (exp b = 1.76, SE = .29, p ˂ .05; see Table 2). In
order to check for the direction of the interaction and facilitate interpretation of this finding, we
calculated the probabilities of officers harassing White and Black suspects on the Monday
following the shift to daylight saving time and on other days. These analyses indicate that there
was a 75.59% probability that the stops of Black suspects on the Monday following the shift to
daylight saving time were classified as harassment in comparison to 71.26% of the stops on other
days that were considered harassment; by contrast, there was a 60.47% probability that stops of
White suspects on the Monday following the shift to daylight saving time could be considered
Daylight Saving Time and Police Harassment 15
harassment whereas there was a 68.22% probability that stops of White suspects on other days
were considered police harassment (see Figure 1).
What these findings indicate is that Black suspects, while always more likely to be
harassed even if they carry no contraband, are 6.1% more likely to be harassed on the Monday
after the shift to daylight saving time than on other days. By comparison, White suspects were
actually 11.4% less likely to be harassed the day after the time change than they were on other
days. What these findings suggest is that police harassment – in this case, officer decisions to
unnecessarily stop and search drivers and their cars – is not only influenced by inherent racial
bias, as suggested by the findings of Ayres and Borowsky (2008), but that these tendencies are
exacerbated by the shift to daylight saving time. The likely reason for this is that the macro-level
influence of changing the clocks and disrupting people’s circadian rhythms results in reduced
and lower-quality sleep, which hampers individuals’ self-regulatory ability. Such inability to
self-regulate can mean that racial biases are significantly more likely to manifest themselves
through individuals’ behaviors (Ghumman & Barnes, 2013). This means that law enforcement
officers working hard at their jobs are being undermined by national policies that result in the
enhancement of prejudicial behavior.
STUDY 2
Although the findings of Study 1 highlight how employee racial bias can be magnified on
the day following the shift to daylight saving time, there may be particular features of the LAPD
or the Los Angeles population in general, or of the time period during which the data were
collected, that could call into question whether the findings from Study 1 are endemic, or
whether they generalize to a broader population. We therefore conducted a second study with a
Daylight Saving Time and Police Harassment 16
different and much larger population, and over a much longer time period, with a different
operationalization of police harassment.
The National Incidence-Based Reporting System (NIBRS) is an incident-based reporting
system for crimes known by the police. This publicly available national dataset
(http://www.icpsr.umich.edu/icpsrweb/NACJD/series/128) is compiled and formatted by the
Federal Bureau of Investigation and includes detailed information on the incident, victim,
offender, and arrestee from January 1st, 1991 to December 31st, 2011. This database consisted of
61,662,778 incidents across 21 years. Of the suspects in these incidents, 20,717,446 were White,
10,610,403 were Black, 200,821 were American Indian, and 530,728 were Asian (29,603,390
reports did not include race data). Twenty five percent of the suspects were female and the
average age was 30.02 years (SD=14.38). Similar to Study 1, we conducted analyses using White
and Black suspects only; we also excluded states and territories that do not implement daylight
saving time (Arizona, Hawaii, American Samoa, Guam, Puerto Rico, and the Virgin Islands),
which resulted in a final sample of 29,650,041 incidents. There were 78,348 incidents on the
Mondays following the shift to daylight saving time.
Not only does the NIBRS provide data relating to the nature of the incident itself, but it
also provides information about subsequent prosecution for alleged crimes. Although in most
instances the incidents logged in the database result in some sort of judicial process, a small
number of these cases are “exceptionally cleared” because the prosecutor declines the
opportunity to prosecute the individual. What this means is that when the prosecutor reviews the
particulars of the case there exists evidence of probable cause, indicating that it was likely a law
was broken, yet despite this evidence the prosecutor forgoes prosecution. There could be many
reasons for such a decision, but examples offered by the FBI suggest that in many cases these
Daylight Saving Time and Police Harassment 17
decisions are made because the violation was so trivial that the prosecutor did not deem it
worthwhile to bring the case before a judge. Instead, the prosecutor recommends that the suspect
be released without any punitive action.
Although suspects exceptionally cleared in this way are nominally guilty of violating the
law, the violation is so trivial as to not be considered a significant violation of societal norms.
Examples might include a driver exceeding a speed limit by only a few miles an hour,
jaywalking on a quiet street, or listening to one’s music a bit too loudly. Each of these acts could
violate formal laws, but social norms would suggest that police action against such behavior is
unlikely. Nonetheless, in such instances police have the legal authority to take action. Thus, the
existence of legal authority to act, but social norms to abide, present a scenario in which police
officer behavior is largely left to the officer’s discretion. Such a situation might be termed a
weak situation (Mischel, 1977), in which an individuals’ underlying personality, preferences, or
attitudes might be more commonly manifest than in situations wherein police officers’
behavioral reactions might be constrained (e.g., when an armed suspect fires at police officers
and escapes in a stolen car, armed pursuit is the likely course of action).
Given this latitude in possible response to these ignorable offenses, observing police
behavior in such situations offers a window into police harassment as it exemplifies individuals’
behaviors in situations that do not strongly constrain action. Accordingly, in Study 2 we examine
police harassment as the extent to which police officers arrest individuals for whom prosecution
is subsequently declined and who are released without further legal action. In other words, we
operationalize police harassment as the arrest of any individual for reasons that legal and social
norms dictate did not merit arrest.
Measures
Daylight Saving Time and Police Harassment 18
Police harassment. Police harassment was operationalized as incidents with an outcome
of “prosecution declined.” Events were coded as “1” if prosecution was declined (i.e., police
harassment) and “0” otherwise.
Daylight saving. In the same manner as in Study 1, Mondays following the shift to
daylight saving time were coded as “1” and the remaining days were coded as “0.”
Race. Black suspects were coded as “1,” whereas White suspects were coded as “0.”
Control variables. As in Study 1 and consistent with other studies of the consequences of
daylight saving time, we controlled for daily, weekly, and yearly effects, taking into account
whether the arrest happened during the day or at night, and the day of week, week of year, and
the year of the arrest. We also controlled for the age and gender of the suspect.
Results and discussion
We applied logistic regression equations with conditional fixed effects using STATA to
test the hypotheses (conditional fixed effect set to week of the year)1. Correlations, means,
standard deviations, and correlations for the hypothesized variables are provided in Table 2.
Hypothesis 1 suggested that the shift to daylight saving time influences the incidence of police
harassment. After controlling for the variables described above, there was no main effect of
daylight saving time on police harassment (exp b = 1.01, SE = .02, p = .51), thus Hypothesis 1
was not supported. Next, we tested Hypothesis 2, which predicts that the shift to daylight saving
time interacts with suspect race to predict police harassment. We adopted a stepwise approach to
test this hypothesis (Aiken & West, 1991; Cohen & Cohen, 1983). In step 1, we entered the
1
We could also use a fixed-effects model simply by including dummy variables for week of the
year. However, such a standard fixed-effects model yields inconsistent estimates of the
coefficients because of the “incidental parameters problem” (Cameron & Trivedi, 1998) unless
the number of units is fixed and relatively small. Thus, the conditional method used in our
primary analyses takes into consideration unobserved heterogeneity and thus effectively controls
for week of the year effects.
Daylight Saving Time and Police Harassment 19
control variables. In step 2, we added the independent variable (i.e., daylight saving time) and the
moderator (suspect race), which explained a significant amount of variance (Δχ2 = 2,782.69, p ˂
.01). The results revealed that suspect race has a significant effect on police harassment (exp b =
1.11, SE = .01, p ˂ .01), again consistent with past research. Finally, in step 3, we entered the
interaction between daylight saving time and suspect race. This final model explained
significantly more variance than the prior models (Δχ2 = 11.57, p ˂ .01) and revealed that
daylight saving time interacts with suspect race to predict unnecessary arrests (exp b = 1.13, SE
= .04, p ˂ .01; see Table 2). In order facilitate interpretation of this finding, we calculated the
probabilities that arrests of White and Black suspects would be considered harassment on the
Monday following the shift to daylight saving time as compared to other days. Our results
indicate that 4.18% of the arrests of Black suspects on the Monday following the shift to daylight
saving are considered police harassment in comparison to a 3.84% of the arrest of Black suspects
on other days. By contrast, only 3.35% of White suspects arrested on the Monday following the
change to daylight saving time were victims of police harassment as compared to a 3.46% of the
White suspects arrested on other days (see Figure 2). What this indicates is that Black suspects
are 8.9% more likely to be harassed by police on the Monday after the shift to daylight saving
time, whereas White suspects are 2.9% less likely to be harassed by police on the Monday
following the time change.
GENERAL DISCUSSION
The findings of our two quasi-experimental field studies corroborate past evidence
suggesting that police are more likely to harass Black than White suspects (Ayres & Borowsky,
2008). There are many explanations for why this racial bias exists, and many of these reside at
the individual level. However, the novel findings of the present paper reveal that in addition to
Daylight Saving Time and Police Harassment 20
these race-related biases, police harassment is especially common toward Black suspects (and
less common toward White suspects) when police officers’ self-regulatory resources are depleted
by macro-level factors that hamper sleep. These findings are important for several reasons. First,
our findings direct part of the blame for police harassment at largely ignored macro-level cause.
Even though police harassment is eventually determined by the officer’s behavior, the fact that
these behaviors are influenced, and potential biases unrestrained, by at macro-level policy that
affects nearly all residents of the United States and many countries around the world is cause for
alarm. Although there are individual-level interventions that can be employed to mitigate the
harmful consequences of the sleep lost due to daylight saving time (e.g., Barnes, 2012), the
policy itself is typically endorsed at the national level. Hence, progress on this front may require
a reexamination of the practice by policy makers in order for our society to advance past this
apparently harmful practice.
This reexamination should also consider recent evidence suggesting that the shift to
daylight saving time is counterproductive and can hamper valued outcomes in many contexts.
For instance, the time change has been linked to a greater incidence of automobile accidents
(Coren, 1996) and mining injuries (Barnes & Wagner, 2009), and has even seeped into white
collar contexts such as motivation and cyberloafing (Wagner et al., 2012). Given the
complexities that characterize many decisions made by law enforcement officers, the lost sleep
associated with daylight saving time is particularly concerning given that complex decisions are
more drastically affected by sleep than are simple decisions (Schnyer, Zeithamova, & Williams,
2009).
A second reason these findings are important is that they raise awareness of police
harassment that could be viewed as rather benign, but that nonetheless generates costs in terms of
Daylight Saving Time and Police Harassment 21
victims’ physical, mental, and social well-being. Although spending an evening in jail is most
assuredly better than winding up dead from undeserved gunshot wounds, as has been the case
with Amadou Diallo (Cooper, 1999) and others, the personal and social costs associated with
mistreatment by law enforcement are not trivial (Perr, 1988; Newman, 2006). In light of these
costs, which could be difficult to monetize and even more difficult to track, the cumulative cost
of daylight saving time policy should be reassessed and put in balance against its ambiguous
benefits (Downing, 2005).
A third reason our findings are important is that, although the costs of police harassment
to victims should be thoroughly considered, law enforcement officers engage in demanding and
potentially dangerous work and thus any consideration of how to mitigate victim costs must also
be weighed against the demands of the function of law enforcement. Because of the complexities
of police work, policy makers should be especially attentive to how policies affecting sleep
might hamper decision making in this critical component of our society. Indeed, not only have
the present findings highlighted that the shift to daylight saving time reduces the self-regulatory
strength to suppress racial prejudice, but evidence also exists that this time change, and a lack of
sleep in general, makes people less aware of the moral elements to everyday situations. This
suggests that not only must police officers exert themselves to appropriately eliminate bias, but
they are already placed in an unfavorable position with regard to the information they need to
make appropriate and effective decisions.
As evidence continues to mount, it is increasingly difficult to contest with the argument
that daylight saving time policy has significant costs that range from public health and accidents
(Barnes & Wagner, 2009; Janszky & Ljung, 2008), to the performance of individuals and
publicly traded stocks (Kamstra, Kramer, & Levi, 2000; Wagner et al., 2012). As the present
Daylight Saving Time and Police Harassment 22
study uncovers the effects of daylight saving time policy that spill into the public square, the
urgency to address this policy escalates.
In conclusion, we candidly acknowledge that “all professionals make mistakes; surgeons
operate on the wrong kidney, lawyers mess up cross-examinations, accountants overlook
legitimate tax deductions” (Toby, 2000, p. 42). However, errors in police work are often critical,
and sometimes fatal such as the shooting death of an innocent man (Cooper, 1999). Of course,
any sort of police harassment carries a cost (Toby, 2000). Indeed, not only do these actions have
adverse consequence for suspected criminals, but also for the officers who make the bad
decisions, as legal and social consequences can plague those who make a bad decision in any
given moment (Manos, 1967). Although living in a mistake free world is assuredly unattainable,
by addressing systemic factors that result in poor decision making, such as national policy that
leaves people sleep deprived, we can systematically make our world a better place and lend the
support to the men and women who strive to enforce the laws that provide a structure for our
society. Hence, in light of the mixed support for the benefits of daylight saving time policy
(Downing, 2005), the growing body of evidence against daylight saving time and the
manipulation of clock time in particular suggests that the time may have come to change our
policies rather than continuing to change our clocks.
Daylight Saving Time and Police Harassment 23
REFERENCES
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions.
Thousand Oaks, CA: Sage Publications, Inc.
Ayres, I., & Borowsky, J. (2008). Racial profiling and the LAPD: A study of racially disparate
outcomes in the Los Angeles Police Department. ACLU of Southern California.
Barnes, C. M. (2012). Working in our sleep: Sleep and self-regulation in organizations.
Organizational Psychology Review, 2: 234-257.
Barnes, C. M., & Wagner, D. T. (2009). Changing to daylight saving time cuts into sleep and
increases workplace injuries. Journal of Applied Psychology, 94: 1305-1317.
Barnes, C. M., Schaubroeck, J., Huth, M., & Ghumman, S. (2011). Lack of sleep and unethical
conduct. Organizational Behavior and Human Decision Processes, 115: 169-180.
Basner, M., Rubinstein, J., Fomberstein, K. M., Coble, M. C., Ecker, A., Avinash, D., & Dinges,
D. F. (2008). Effects of night work, sleep loss and time on task on simulated threat
detection performance. Sleep, 31: 1251-1259.
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analysis for the
behavioral sciences. Hillsdale, New Jersey: Lawrence Erlbaum Associates, Inc.
Cooper, M. (February 5, 1999). Officers in Bronx fire 41 shots, and an unarmed man is killed.
The New York Times.
Coren, S. (1996). Daylight savings time and traffic accidents. New England Journal of
Medicine, 334: 924-925.
Czeisler, C. A. et al. (1999). Stability, precision, and near-24-hour period of the human circadian
pacemaker. Science, 284: 2177-2181.
Daylight Saving Time and Police Harassment 24
Downing, M. (2009). Spring forward: The annual madness of daylight saving time. Berkeley,
CA: Counterpoint Press.
Ghumman, S., & Barnes, C. M. (2013). Sleep and prejudice: A resource recovery approach.
Journal of Applied Social Psychology, 43: 166-178.
Kantermann, T., Juda, M., Merrow, M., & Roenneberg, T. (2007). The human circadian clock’s
seasonal adjustment is disrupted by daylight saving time. Current Biology, 17: 19962000.
Manos, J. M. (1967). Police liability for false arrest or imprisonment. Cleveland-Marshall Law
Review, 16: 415-427.
Mischel, W. (1977). On the future of personality measurement. American Psychologist, 32: 246254.
Newman, S. H. (2006). Proving probable cause: Allocating the burden of proof in false arrest
claims in §1983. The University of Chicago Law Review, 73: 347-376.
Perr, I. N. (1988). Claims of psychiatric injury after alleged false arrest. Journal of Forensic
Sciences, 33: 21-34.
Schnyer, D. M., Zeithamova, D., & Williams, V. (2009). Decision-making under conditions of
sleep deprivation: Cognitive and neural consequences. Military Psychology, 21: S36-S45.
Stauffer, J. M., & Buckley, M. R. (2005). The existence and nature of racial bias in supervisory
ratings. Journal of Applied Psychology, 90: 586-591.
Toby, J. (2000). Are police the enemy? Society, 37: 38-42.
Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases.
Science, 185: 1124-1131.
Daylight Saving Time and Police Harassment 25
Wagner, D. T., Barnes, C. M., Lim, V. K. G., & Ferris, D. L. (2012). Lost sleep and
cyberloafing: Evidence from the laboratory and a daylight saving time quasi-experiment.
Journal of Applied Psychology, 97:1068-1076.
Daylight Saving Time and Police Harassment 26
TABLE 1
Means, Standard Deviations and Correlations (LAPD Study 1) ª
Mean
SD
1
1. Police Harassment
.72
.45
2. Daylight Saving
.01
.05
-.01
3. Race
.69
.69
.08*
2
-.01
ª N = 99,982; Police Harassment = 1, Control = 0; Daylight Saving = 1, Control = 0; Black
Suspect = 1, White Suspect = 0; * p < .01.
Daylight Saving Time and Police Harassment 27
TABLE 2
Logistic Regression of Daylight Saving Time and Suspect Race on Police Harassment
(LAPD Study1) ª
Constant
Area Dummy set
Rampart Area
Southwest Area
Hollenbeck Area
Harbor Area
Hollywood Area
Wilshire Area
West LA Area
Van Nuys Area
West Valley Area
Northeast Area
77th Street Area
Newton Area
Pacific Area
North Hollywood Area
Foothill Area
Devonshire Area
Southeast Area
Reason for Stop Dummy set
Equipment and Registration Violation
Call for Service
Municipal Code Violation
Consensual Stop
Department Briefing
Health and Safety Code Violation
Any Other Stop
Pedestrian Code Violation
Penal Code Violation
Suspect Flight
Day of the Week Dummy Set
Monday
Tuesday
Wednesday
Thursday
Friday
Saturday
Week of the Year Dummy Set
Week 2
Week 3
Week 4
Week 5
Step 1
B
Exp(B)
.89**
2.43
B
.76
Step 2
Exp(B)
2.15
B
.76
Step 3
Exp(B)
2.15
.23**
.81**
.22**
.40**
.24**
.44**
.08
.01
-.35**
-.02
.98**
.44**
.22**
.15**
-.09*
.04
.58**
1.25
2.24
1.24
1.50
1.27
1.55
1.09
1.01
.71
.98
2.65
1.55
1.25
1.16
.92
1.04
1.78
.25**
.79**
.27*
.46**
.30**
.44**
.14
.09
-.25**
.04
.96**
.42**
.28**
.23**
-.02
.11**
.56**
1.29
2.20
1.31
1.58
1.34
1.56
1.15
1.09
.78
1.04
2.60
1.52
1.32
1.26
.98
1.12
1.75
.25**
.79**
.27**
.46**
.30**
.44**
.14**
.08**
-.25**
.04
.96**
.42**
.28**
.23**
-.02
.11**
.56**
1.29
2.20
1.31
1.58
1.34
1.56
1.15
1.09
.78
1.04
2.60
1.52
1.32
1.26
.98
1.12
1.75
.01
.95**
.68**
.74**
.86**
-.37**
.44**
.97**
.41**
-.07
1.01
2.58
1.96
2.10
2.35
.69
1.56
2.65
.01
.38
.01
.95**
.66**
.73**
.85**
-.38**
.44**
.96**
.40**
-.08
1.00
2.59
1.94
2.07
2.33
.69
1.55
2.62
1.49
.92
.01
.95**
.66**
.73**
.85**
-.38**
.44**
.96**
.40**
-.08
1.00
2.59
1.94
2.07
2.33
.69
1.55
2.62
1.49
.92
-.01
-.05
-.03
-.01
-.05
-.07*
.99
.95
.97
.99
.96
.93
-.01
-.06
-.03
-.02
-.05
-.07*
.99
.95
.97
.99
.95
.93
-.01
-.05
-.03
-.01
-.05
-.07*
.99
.95
.97
.99
.95
.93
-.02
.05
.12
.01
.98
1.06
1.12
1.01
-.02
.05
.12
.01
.98
1.05
1.12
1.01
-.02
.05
.12
.01
.98
1.05
1.12
1.01
Daylight Saving Time and Police Harassment 28
Week 6
Week 7
Week 8
Week 9
Week 10
Week 11
Week 12
Week 13
Week 14
Week 15
Week 16
Week 17
Week 18
Week 19
Week 20
Week 21
Week 22
Week 23
Week 24
Week 25
Week 26
Week 27
Week 28
Week 29
Week 30
Week 31
Week 32
Week 33
Week 34
Week 35
Week 36
Week 37
Week 38
Week 39
Week 40
Week 41
Week 42
Week 43
Week 44
Week 45
Week 46
Week 47
Week 48
Week 49
Week 50
Week 51
Week 52
Week 53
Night Stop
Violent
Warrant
-.02
.11
-.07
-.06
-.07
-.01
-.05
-.08
-.15
-.10
-.21
-.26**
-.39**
-.48**
-.32**
-.43**
-.31**
-.35**
-.46**
-.34**
-.26**
-.24**
-.37**
-.40**
-.29**
-.38**
-.57**
-.46**
-.56**
-.48**
-.47**
-.48**
-.42**
-.50**
-.51**
-.65**
-.50**
-.48**
-.49**
-.46**
-.38**
-.60**
-.64**
-.57**
-.67**
-.64**
-.62**
-.79**
.26**
-.25**
-1.06**
.98
1.12
.93
.95
.93
.99
.95
.92
.86
.91
.81
.77
.68
.62
.73
.65
.73
.70
.63
.72
.77
.79
.69
.67
.75
.68
.57
.63
.57
.62
.63
.62
.66
.61
.60
.52
.60
.62
.62
.63
.68
.55
.53
.56
.51
.53
.54
.45
1.30
.78
.35
-.02
.11
-.08
-.06
-.07
-.02
-.05
-.09
-.16
-.10
-.22*
-.26**
-.39**
-.48**
-.31**
-.43**
-.31**
-.35**
-.46**
-.34**
-.26**
-.24**
-.37**
-.40**
-.29**
-.38**
-.57**
-.46**
-.56**
-.48**
-.47**
-.48**
-.42**
-.50**
-.51**
-.65**
-.50**
-.48**
-.49**
-.46**
-.38**
-.60**
-.64**
-.57**
-.67**
-.64**
-.62**
-.79**
.26**
-.26**
-1.06**
.99
1.12
.92
.95
.93
.98
.95
.92
.86
.90
.81
.77
.68
.62
.73
.65
.73
.71
.63
.72
.77
.79
.69
.67
.75
.69
.57
.63
.57
.62
.63
.62
.66
.61
.60
.52
.61
.62
.61
.63
.68
.55
.53
.57
.51
.53
.54
.45
1.30
.77
.35
-.02
.11
-.08
-.06
-.07
-.02
-.05
-.09
-.15
-.10
-.22*
-.26**
-.39**
-.48**
-.31**
-.43**
-.31**
-.35**
-.46**
-.34**
-.26**
-.24**
-.37**
-.40**
-.29**
-.38**
-.57**
-.46**
-.56**
-.48**
-.46**
-.48**
-.42**
-.50**
-.51**
-.65**
-.50**
-.48**
-.49**
-.46**
-.38**
-.60**
-.63**
-.57**
-.67**
-.64**
-.62**
-.79**
.26**
-.26**
-1.06**
.98
1.12
.93
.95
.93
.98
.95
.92
.86
.90
.81
.77
.68
.62
.73
.65
.73
.70
.63
.71
.77
.79
.69
.67
.75
.68
.57
.63
.57
.62
.63
.62
.66
.61
.60
.52
.61
.62
.61
.63
.68
.55
.53
.57
.51
.53
.54
.45
1.30
.77
.35
Daylight Saving Time and Police Harassment 29
Officer Race Dummy Set
Black
Asian
Hispanic
Officer Age
Daylight Saving
Suspect Race
Daylight Saving x Suspect Race
Chi-square
Δ Chi-square
-2Loglikelihood
-.02
.05*
.17**
-.01**
.98
1.05
1.19
.99
8,511.26
109,322.86
-.03
.05
.17**
-.01**
.02
.15**
.97
1.05
1.19
.99
1.02
1.16
8,566.39
55.13**
109,267.73
-.03
.97
.05
1.05
.17**
1.19
-.01**
.99
-.34
.71
.14**
1.15
.56*
1.76
8,570.21
3.83*
109,263.91
ª N = 99,982; b Police Harassment = 1, Control = 0; Daylight Saving = 1, Control = 0; Black
Suspect = 1, White Suspect = 0; ** p < .01, * p < .05
Daylight Saving Time and Police Harassment 30
TABLE 3
Means, Standard Deviations and Correlations (FBI Study 2) ª
Mean
SD
1
1. Police Harassment
.04
.20
2. Daylight Saving
.003
.05
.0007
3. Race
.33
.47
.009
2
.00
ª N = 29,650,041; Police Harassment = 1, Control = 0; Daylight Saving = 1, Control = 0; Black
Suspect = 1, White Suspect = 0; ** p < .01.
Daylight Saving Time and Police Harassment 31
TABLE 4
Logistic Regression of Daylight Saving Time and Suspect Race on Police Harassment (FBI
Study 2) ª
Constant
Year Dummy Set
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
Week of the Year Fixed Effects
Day of the Week Dummy Set
Monday
Tuesday
Wednesday
Thursday
Friday
Saturday
Night Stop
Suspect Age
Suspect Gender
Suspect Race
Daylight Saving
Daylight Saving x Suspect Race
Chi-square
Δ Chi-square
-2Loglikelihood
Step 1
Step 2
B
Exp(B)
B
Exp(B)
-3.27**
.04
-3.33**
0.04
Step 3
B
Exp(B)
-3.33**
.04
.35**
.45**
.38**
.20**
.18**
.17**
.18**
.21**
.19**
.20**
.17**
.16**
.23**
.23**
.15**
.07**
.06**
.02
.04**
.04**
Yes
1.42
1.57
1.46
1.22
1.20
1.19
1.19
1.23
1.21
1.22
1.18
1.17
1.26
1.26
1.17
1.08
1.07
1.02
1.04
1.04
.36**
.47**
.40**
.23**
.21**
.20**
.21**
.24**
.21**
.23**
.19**
.18**
.26**
.26**
.18**
.10**
.09**
.04**
.06**
.07**
Yes
1.43
1.60
1.49
1.25
1.23
1.22
1.23
1.27
1.24
1.25
1.21
1.20
1.30
1.29
1.19
1.10
1.09
1.04
1.07
1.07
.36**
.47**
.40**
.23**
.21**
.20**
.21**
.24**
.21**
.23**
.19**
.18**
.26**
.26**
.18**
.10**
.09**
.04**
.06**
.07**
Yes
-.01**
-.04**
-.04**
-.05**
-.06**
-.07**
-.10**
.01**
-.11**
.99
.96
.96
.95
.94
.94
.91
1.01
.89
-.01**
-.04**
-.05**
-.06**
-.06**
-.07**
-.10**
.01**
-.12**
.11**
.01
0.99
0.96
0.96
0.95
0.94
0.94
0.91
1.01
0.89
1.11
1.01
-.01**
.99
-.04**
.96
-.05**
.96
-.06**
.95
-.06**
.94
-.07**
.94
-.10**
.91
.01**
1.01
-.12**
.89
.11**
1.11
-.03
0.97
.12**
1.13
28,392.99
11.57**
9,435,997
25,598.73
9,438,809.2
28,381.42
2,782.69**
9,436,007
1.43
1.60
1.49
1.25
1.23
1.22
1.23
1.27
1.24
1.25
1.21
1.20
1.30
1.29
1.19
1.10
1.09
1.04
1.07
1.07
Note. ª N = 29,650,041; b False Arrests = 1, Control = 0; c Daylight Savings = 1, Control = 0; d
Black Suspect=1, White Suspect= 0; ** p < .01.
Daylight Saving Time and Police Harassment 32
FIGURE 1
Probabilities of Police Harassment (LAPD Study 1)
Daylight Saving Time and Police Harassment 33
FIGURE 2
Probabilities of Police Harassment (FBI Study 2)