The Shift Length Experiment

7/30/2019 The Shift Length Experiment

1/62

The

ShifTLengThexperimenT

Wat W Kow About8-, 10-, ad 12-hou Sts polc

POLICEFOUNDATION


2/62

The

ShifT

LengThexperimenTWat W Kow About8-, 10-, ad 12-hou Sts

polcBy

Ka L. Adola, Davd Wsbud,edw e. halto, g Jos, ma Slka

Wt

Ak hta, Jo Sa,Csto Otz, elab Tak

Washington, DC


3/62

The Police Foundation is a national, nonpartisan, nonprot organization dedicated to supportinginnovation and improvement in policing. Established in 1970, the oundation has conducted seminal

research in police behavior, policy, and procedure, and works to transer to local agencies the best newinormation about practices or dealing eectively with a range o important police operational andadministrative concerns. Motivating all o the oundations eorts is the goal o ecient, humane policingthat operates within the ramework o democratic principles and the highest ideals o the nation.

2011 by the Police Foundation.All rights, including translation into other languages, reserved under the Universal Copyright Convention,the Berne Convention or the Protection o Literary and Artistic Works, and the International and PanAmerican Copyright Conventions.

For inormation about using Police Foundation copyrighted material, please visitwww.policeoundation.org/docs/copyright.html.

Police Foundation1201 Connecticut Avenue, NWWashington, DC 20036-2636(202) 833-1460(202) 659-9149 [email protected]

ISBN 978-1-884614-25-5

Library o Congress Control Number: 2011943533

This project was supported by award 2005-FS-BX-0057 rom the National Institute o Justice, Oce oJustice Programs, U.S. Department o Justice. The opinions, ndings, and conclusions or recommendationsexpressed in this publication are those o the authors and do not necessarily refect the views o theDepartment o Justice.

The Shit Length Experiment: What We Know About 8-, 10-, and 12-Hour Shits in Policingis available atwww.policeoundation.org/shitexperiment/.


4/62

TABLe Of COnTenTS

fowod ...........................................................................................................................v

Ackowldts ........................................................................................................... vii

itoducto ..................................................................................................................... 1

Ky fds o t et .................................................................................... 5

T et ................................................................................................................. 7

Method .................................................................................................................... 8Agencies ............................................................... .................................................... 8Sampling Procedures ............................................................................ .................. 8Participation ................................................................... ......................................... 8

Research Design and Methods ................................................................................9Multicenter Trail .............................................................. ........................................ 9Random Assignment .................................................................. ............................. 9Outcomes and Data Collection Methods .............................................................. 9Measures .................................................................................................................11

Laboratory simulations .......................................................................................11Police department data ......................................................................................11Sel-report instruments ........................................................... ........................... 12

Data Analysis .................................................................. ....................................... 14Analytical approach ............................................................... ........................... 14Eect size .................................................................. ....................................... 14

rsults ...........................................................................................................................15

Signicant Findings ........................................................... ...................................... 15Quality o Work Lie (QWL) ................................................................................. 15Sleep Amount ................................................................. ....................................... 15Sleepiness/Fatigue (subjective) .............................................................. ............... 16Alertness .............................................................. ................................................... 17Overtime Worked ............................................................ ...................................... 17


5/62

v

Nonsignicant Findings ...................................................................................................................17Work Perormance and Saety .................................................................... ........................................... 17Health and Stress ....................................................................................................................................18Quality o Personal Lie .........................................................................................................................18Sleep Quality, Fatigue (objective), and Sleep Disorders .............................................................. ........18O-Duty Employment ...........................................................................................................................18

Suay o Coss-idusty rsac o Cossd Wokwks ................................................................ 19

Early Research ..................................................................................................................................... 19

More Recent Findings ....................................................................................................................... 20

Wat W Kow About Cossd Wokwks Law eoct Coad to Ot idusts ...........................................................................................21

Problems with Past Studies ...............................................................................................................21

Summary o Findings by Outcomes .............................................................................................. 22Perormance ............................................................................................................................................22

Sel-initiated activities and eort .......................................................................................................23

Quality o service ...............................................................................................................................23Communication .................................................................................................................................23Cognitive perormance .......................................................................................................................24

Saety .......................................................................................................................................................24Health .....................................................................................................................................................25Quality o Lie ..................................................................... ................................................................... 27Fatigue, Sleep, and Alertness ..................................................................................................................29Overtime and O-Duty Employment ...................................................................................................32Other Outcomes .....................................................................................................................................33

Cocluso ..................................................................................................................................................... 35

Adcs .................................................................................................................................................... 39A. Description o Measures .............................................................................................................. 39

B. Past and Current Psychometric Properties o Measures ........................................................ 42

rcs .................................................................................................................................................... 43

Autos ..........................................................................................................................................................51

illustatos

Tables1. Actual Distribution o Participants Per Cell .................................................................. ..................... 92. List o Measures Within Each Construct ...........................................................................................13

3. Results o GLM Analysis and Associated Eect Sizes ............................................................... ........16Figures

1. Flow o Participants Through Each Stage o the Experiment ................................................... ........ 10


6/62

v

fOreWOrD

Over the last decade, there have been increased calls or evidence-basedpractices in policing. Without the commitment o police leaders toparticipate in scientic research in their agencies, much o what we know

about what worksand what does notin policing would not have beenpossible. The responsibility to maintain order and public saety, the reactive

environment in which the police must oten unction, and resource constraintspose real challenges to scientic inquiry in police agencies. Thus, the decision toparticipate in complex research is not an easy one, even or the most progressivepolice chie. Yet, it is precisely because o these constraints that the advancement oscience in policing is essential.

This experimental eld study o the eects o compressed workweeks (ewerthan ve days) in policing required signicant commitment and cooperationrom police leaders and rank-and-le ocers in the two study sites, Detroit (MI)and Arlington (TX). Few experimental studies in policing have required randomassignment o ocers to conditions that could have as signicant an impact on theirlives as that o the hours and days they work. These agenciestheir leaders, ocers,and police associationsworked diligently to ensure that this study could besuccessully implemented. We acknowledge and applaud the important contributionthey have made in advancing the evidence base in policing research.

When I led the Newark (NJ) Police Department rom 1974 to 1985, the criticalchallenge acing police executives was to do more with less. Fiscal stringency cuta wide swath through the Newark Police Department, as it did many other majorAmerican police departments. In Newark and elsewhere, there were severe cutbacksin personnel, either through attrition or outright layos. There were reductionsin resources ranging rom patrol cars to computer acilities, and constant cautionto hold down costs in all possible ways. Despite those challenges, we agreed toparticipate in two pioneering Police Foundation experimentsoot patrol andear reductionbecause we recognized that opening the department to outsideexamination and experimentation in which police theory and practice wouldbe subjected to the rigors o scientic analysis would produce the best evidenceavailable as to how well they work.

Nevertheless, at that time I could not have imagined allowing ocers to workewer days, even i the total hours were the same. Many personnel practices in


7/62


8/62

v

ACKnOWLeDgmenTS

As in any large-scale project, we owe a debt o gratitude to many people whohelped in developing and implementing this experiment. Most importantly,we thank the Arlington (TX) and Detroit (MI) Police Departments whose

participation in this study allowed us to learn about the impacts o variousshit schedules, which will be o benet to law enorcement executives,

ocers, and scholars.We are grateul to Detroit Police Chie (retired) Ella Bully-Cummings and

Arlington Police Chie Theron Bowman, PhD, or their commitment to advancingthe eld o policing through agency participation in research. The shit work studywas one o the most complex and dicult eld studies to run because it requiredocers to participate in simulations at all times o day and night (and at the endo their shits), and considerable resources (such as oce space, personnel, etc.).Without orward thinking, dedicated leaders such as Chie Bowman and ChieBully-Cummings, much o the important research in policing would not be possible.Similarly, we thank the Detroit Police Ocers Association and the Arlington PoliceAssociation, whose support allowed us to gain participation o willing ocers.

We grateully acknowledge a number o individuals in both policedepartmentskey leaders and support stawho dedicated time and eort toensuring the success o the study. In Detroit, we thank the ollowing individuals:Deputy Chie (ret.) Gail Wilson-Turner; Assistant Chie (ret.) Gary Christian;Commander (ret.) John Autrey, who served as project site coordinator; ChieRalph Godbee (ormerly assistant chie); Marty Bandemer, president o the DetroitPolice Ocers Association; and police personnel at the Casino Gaming Unit, whoshared their oce space with us. In Arlington, we greatly appreciate the eorts oSuzan Cogswell, operations analyst/ocial department ocer liaison coordinator;Assistant Chie James Hawthorne; Lieutenant Osbaldo Flores; Sergeant LaTeshaWatson, PhD; Corporal Keith Scullen; Beth Troy, assistant to the police chie; SharonJones, oce coordinator, South Arlington Police Service Center; and the ocers andsupervisors at South Station who shared their oce space with us.

This project would not have been possible without the support o the NationalInstitute o Justice (NIJ), in particular the Crime Control and Prevention ResearchDivision, headed by Winired Reed. We are also most grateul to our grant monitor,Brett Chapman, PhD, social science analyst, who provided important project


9/62

v

support and participated in our exercise simulations on-site in Detroit. Maggie Heisler, ormer seniorsocial science analyst and our initial grant monitor, demonstrated enthusiasm and involvement in keyproject design issues as did her successor, Akiva Liberman, PhD, who helped iron out design issuesimportant to the scientic merits o the study. We especially thank Thomas Feucht, PhD, executivesenior science advisor, who worked with us in improving upon the proposed study design andpromoted this eort.

We are sincerely grateul to Bryan Vila, PhD, o Washington State University, who initiated researchon police atigue at NIJ that spawned and complemented a series o high quality research projects onpolice atigue, sleep, health, and stress. We thank the ollowing experts and researchers who providedinput into our study design at the outset: Dr. Laura K. Barger, Harvard Medical School and Brighamand Womens Hospital; Cecil (Buzz) Burchel, PhD, Centers or Disease Control/National Instituteo Occupational Saety and Health; Dr. Steven W. Lockley, Harvard Medical School and Brigham andWomens Hospital; Dr. Thomas C. Neylan, Department o Psychiatry, University o Caliornia, SanFrancisco Medical Center; John Violanti, PhD, Department o Social and Preventive Medicine, StateUniversity o New York at Bualo; Dr. Charles Marmar, chair, Department o Psychiatry, New YorkUniversity, Langone Medical Center; Dr. John Vena, proessor and chair, Department o Epidemiologyand Biostatistics, Norman J. Arnold School o Public Health, University o South Carolina; and Tara A.

Hartley, MPA, MPH, epidemiologist, National Institute or Occupational Saety and Health, Centers orDisease Control and Prevention.

Much appreciation is extended to Richard Bennett, PhD, o American University and chair o thePolice Foundations Institutional Review Board, who oversaw the human subjects aspects o the project.We are grateul to Robert F. Boruch, PhD, University o Pennsylvania, Graduate School o Education andthe Wharton School, who assisted with the complex randomization procedure, and Jose M. Cortina,PhD, George Mason University Psychology Department, or his analytical advice and expertise.

We thank the ollowing Police Foundation sta, consultants, and ellows or their commitmentto and eorts in conducting various phases o this project: Sam Cogen, Anne Corbin, JD, Garth denHeyer, PhD, Vicki den Heyer, Abby Hoyt, Michael Lawrence, Veronica Puryear, PhD, Tyrone Richards,Michael Soelberg, Marketa Tosovka, Michal Tosovk, Hue Williams, Laura Wycko, and Kate Zinsser.

We are grateul to the many Police Foundation student interns who assisted with this project andreceived substantive exposure to practical eld research.

Finally, in order to capture various perormance outcomes, we sought out a variety o simulationexercises and chose those that seemed most relevant and useul or our purposes. As such, we wish toacknowledge those individuals and organizations whose products and services were used in this study.We appreciate the cooperation o IES Interactive Training in Ann Arbor, Michigan. Dean Krutty, IESPresident, provided us with the MILO shooting simulator or use in the Detroit Police Department,as well as support rom sta members Jason LaMons and Mike Hogan who acilitated set up, training,and use o MILO Range scenarios. We thank Ed Hotchkiss o PMI, Inc. in Maryland or assistance andtraining on the use o the Fitness-or-Duty Impairment Screener (FIT). We thank Steve Somers romthe B-PAD Group, Inc. or assistance and training on the use and scoring o the B-PAD. ScantronCorporation did the layout and printing o our police survey, and we thank Esther Byrd and JulieWillmes or their assistance with the development o our Scantron survey booklet, and Bunny Clarkeor providing training on the scanning system.


10/62

1

inTrODUCTiOn

ever since the earliest police orces were established, the schedules and hoursthat police ocers work have been an issue o concern to ocers and chies.Driving these concerns have been issues o saety, health, perormance, quality

o lie, atigue, and eciency. Traditionally, police departments have reliedon a ve-day, eight-hour scheduling ramework with three standard shits

(day, evening, midnight) in each twenty-our-hour period. However, since at leastas early as the 1970s, law enorcement agencies have adopted alternate schedulecongurations. Compressed workweek schedules (CWWs), in which the workweekis shortened and the length o the day is extended, have indeed been popularized inthe last several decades in many industries, including policing.

The traditional ve-day, orty-hour workweek did not become the U.S. standarduntil approximately seventy years ago. Labor unions strongly opposed long workhours that were common in the late eighteenth century but oten to no avail. Bythe turn o the century, however, a number o industries had begun to implementeight-hour workdays (Dankert, Mann, and Northrup 1965). Following the GreatDepression and subsequent legislation associated with the New Deal (the Walsh-Healy Public Contracts Act o 1935 and the Fair Labor Standards Act o 1938), morechanges became possible such that private rms began to implement traditionalve-day, orty-hour workweeks. Around that same time, a ew corporations evenbegan experimenting with a our ten-hour day schedule. By the 1970s, CWWshad gained in popularity, and the Federal Employees Flexible and CompressedWork Schedules Act was enacted into law in 1978. During the 1970s and 1980s,tremendous attention was paid to CWWs.

Almost thirty years ago, in a National Institute o Justice-unded study o workscheduling, researchers surveyed 160 agencies regarding their practices and reportedthat almost 25 percent o departments had implemented 9-, 10-, 11- and even12-hour schedules or one or more shits (Stenzel and Buren 1983). Because nonational data have been reported since that time, the Police Foundation conductedsurveys with a random sample o law enorcement agencies in 2005 and 2009.The results o our national surveys seem to suggest that there is a great variation in


11/62

2 Police Foundation

shit schedules employed in U.S. law enorcement, but there have been little available data on theadvantages and disadvantages associated with these shits.1

Over the years, there has been considerable research to examine the impacts o CWWs andlong working hours across industries, particularly in 24/7 and high-risk operations (e.g., hospitals,production and power plants, utilities, and transportation). Yet, Axelsson (2005, 17) noted that whilemanagement and employees believe the advantages o longer work days outweigh the disadvantages,

it could, perhaps, also be argued that the drawbacks o extended work shits are largely unknown orignored by these groups.

While research on CWWs in policing is quite limited, there has been considerable conjecture aboutthe benets and drawbacks o CWWs and long work hours among law enorcement personnel. Notsurprisingly, law enorcement personnel requently claim that CWWs oer ar more advantages thandisadvantages. Among the many benets espoused are the ability to increase coverage during peakhours o activity, improve ocer job satisaction and morale, increase perormance, reduce responsetime, reduce crime, reduce costs or ocers and agencies (e.g., commuting, overtime, and sick leave),limit atigue, improve teamwork, allow or increased in-service training during periods o overlap,increase days o or personal pursuits/amily activities, and reduce accidents and complaints againstocers (see, e.g., Brown 1974; Cunningham 1982; Durrett 1983; Fournet 1983; Jacques 2010; Strunk

1978; Sundermeier 2008; Vega and Gilbert 1997; Vila, Kenney, Morrison, and Reuland 2000). Manyo these purported benets, however, are ar rom rmly established in the research literature. Due to abelie that such schedules may improve eciency, many law enorcement executives have considered orimplemented CWWs (Oliver 2005; Sundermeier 2008; Vega and Gilbert 1997).

Nevertheless, Cunningham (1990) noted some managers in Canadian law enorcement agencieswere concerned about potential disadvantages associated with CWWs in terms o reduced opportunityor communication with sta, citizen complaints, potential costs, lack o investigative continuity,and lessened identication with the police proession due to time away rom the job. In addition,Melekian (1999) noted potential drawbacks associated with CWWs, such as increased atigue,reduced communication across shits, lessened ability to deal with neighborhood problems, and,most importantly, disengagement rom the job and reduced ability or time to establish and maintain

relationships with the community, thereby detracting rom community policing and job involvement.In the absence o empirical evidence, agencies as well as police unions/associations have

occasionally conducted their own research, albeit oten without the benet o rigorous scienticmethods. As such, when agencies make decisions about scheduling, they oten do so without sucientscientically acquired knowledge. Researchers have routinely noted the many unknown potentialimpacts o CWWs (e.g., deCaruel and Schaan 1990), and scientists and practitioners have called oradditional research on CWWs and optimal shit lengths in law enorcement (e.g., Melekian 1999;Vila 2006).

Moreover, scientists have cautioned about the use o extended and long work hours in positionswhere public health and saety could be threatened (Armstrong-Stassen 1998; Knauth 2007;Macdonald and Bendak 2000; Rosa 1995; Scott and Kittaning 2001). Due to widespread knowledge

o the impact o atigue on saety, policies and requirements have been modied in many ederallyregulated industries. Indeed, according to Vila and colleagues, the well-known impact o atigue onsaety has led the ederal government to regulate the work hours o private, or-prot workerstrainengineers, truck drivers, commercial pilots, and nuclear power plant operatorsbut surprisingly notthe police, the governments most public, sensitive, and routinely controversial service provider

1 A report on the two surveys, entitled Trends in Shit Length: Results o a Random National Survey o Police Agencies(Amendola et al.), is available at www.policeoundation.org/shitexperiment/.


12/62

The Shit Length Experiment 3

(Vila, Morrison, and Kenney 2002, 7). Yet, while law enorcement is raught with considerable risks toocers and the public, examination o the impacts o CWWs in policing has been less requent andoten less rigorous than that conducted in other industries. Furthermore, much o the research acrossindustries, including policing, has been limited by the research designs employed, the methodologiesused, and/or measurement problems, oten leading to contradictory or inconclusive ndings.

In an eort to comprehensively address the many potential eects o CWWs in policing in a

systematic way, the Police Foundation conducted an experiment in which ocers were randomlyassigned to shits (8-, 10-, and 12-hour). We examined the independent eects o shit length, takinginto consideration the time o day worked and the variations associated with specic agencies. Becausepast studies have tended to ocus on a limited number o potentially important managerial andindividual considerations, we examined a broad array o outcomes important to the ocers themselvesand the organizations, including ocer stress, sleep, atigue, health, and quality o lie, o-dutyemployment and overtime, and a variety o perormance and saety measures.

In this report, we begin by presenting the key ndings o our experiment and then describethe methodology, the comprehensive array o measures employed, and the results o the analysesconducted in the experiment. Subsequently, we examine cross-industry research on compressedworkweeks, including that rom policing and its connection to our ndings.


13/62


14/62

5

KeY finDingS frOmThe experimenT

Ten-hour shits have advantages over 8-hour shits.

Ten-hour shits appear to oer some advantages over 8-hour shits, both individuallyand organizationally, with no noted disadvantages. For example those ocersworking 10-hour shits got signicantly more sleep per night (over hal an hour)than those on 8-hour shits and had a signicantly higher quality o work lie. Also,those on 10-hour shits worked the least amount o overtime o the three groups,potentially resulting in cost savings.

The benets o 10-hour shits do not extend to 12-hour shits.

Although it may be expected that some advantages associated with 10-hour shitswould inure to those on 12-hour shits, we did not nd that in this study. Forexample, while those on 10-hour shits got signicantly more sleep than those on8-hour shits, the same was not true or those on 12-hour shits.2 Also, those on10-hour shits had a higher reported quality o work lie than those on 8-hour shits,but those on 12-hour shits did not. While those on 12-hour shits worked a lesseramount o overtime than those on 8-hour shits, they still worked more than thoseon 10-hour shits.

Twelve-hour shits may pose saety risks to ocers and the public.

While shit length did not impact saety (e.g., driving, reaction time), those assigned

to 12-hour shits had signicantly lower average levels o alertness at work and weremore sleepy than those on 8-hour shits, something that was not true or thoseon 10-hour shits. Because sleep scientists assert that people underestimate theiratigue levels (e.g., Rosekind and Schwartz 1988), the latter two ndings shouldbe concerning.

2 Although the mean level o sleep or those on 12-hour shits was higher than or those on 8-hourshits, these dierences did not reach statistical signicance.


15/62


16/62

7

The experimenT

We conducted a multisite randomized experiment designed to assess theimpact o compressed workweek schedules (CWWs) on various outcomesin policing using multiple methods and both objective and sel-report

measures. Past ndings related to extended shits and CWWs in otherindustries and especially in policing have been somewhat mixed, in large

part due to limitations o many studies in terms o research designs, methodologicalconcerns, and measurement problems. It was the aim o this research on the impacto CWWs in policing to conduct a randomized experiment that would overcomesome o the limitations o past research and more broadly address a variety ooutcomes within the context o one study. As such, we examined a number ooutcomes o interest, including ocer perormance, saety, health, quality o lie,atigue, sleep, and extra-duty employment, in order to examine the potentialadvantages and disadvantages associated with various shit schedules.

The experiment was conducted in two police departments (Detroit, Michigan,and Arlington, Texas) during the period o January 2007 through June 2009. Aterobtaining volunteers, we randomly assigned ocers to one o three conditions:(a) ve consecutive 8-hour days, (b) our consecutive 10-hour days, and (c) threeconsecutive 12-hour days.4 The agencies agreed to maintain the conditions (shitlength, time o day, and district) throughout the course o the six-month period othe study. We employed a randomized block design, a research design that allowedus to take into account variability within the sites (Detroit and Arlington) andduring dierent shits (day, evening, and midnight), as well as possible interactionsbetween shit length and those actors (see Weisburd and Taxman 2000).

4 In order to ensure all ocers worked eighty hours in each two-week period, ocers assigned to12-hour shits worked three consecutive 12-hour shits in week one, and three consecutive 12-hourshits ollowed by a single 8-hour shit in week two, a conguration which is common amongagencies operating on 12-hour shits.


17/62

8 Police Foundation

Method

Agencies

At the time o the study, both sites were operating on standard eight-hour duty tours with veconsecutive days o work. Both Arlington and Detroit were large enough to ensure minimal impacton regular police operations and to provide a sucient number o cases or our study. As o 2007,the

Detroit Police Department had 3,049 sworn ocers and the Arlington Police Department had 580.5Although the proportion o sworn emale ocers employed in Detroit (27 percent) was higher thanthat o Arlington (18 percent), the proportion o ocers who participated in the study was comparableacross sites. Whereas Detroit had a higher proportion o Black ocers (66 percent), Arlington hadjust 15 percent. There was also a higher proportion o single ocers in Detroit (51.6 percent) than inArlington (36.7 percent). Over 75 percent o participating ocers rom both sites had less than tenyears on the job.

Sampling Procedures

Ocers were recruited or the study on a voluntary basis as is required by ederal regulation and

consistent with institutional review board mandates. In order to be eligible or the study, the ollowingtwo criteria were established or ocers: (1) must be in the patrol operations division (assigned torespond to calls or service); and (2) must not be working oot patrol or on light/restricted duty (toensure all participants were perorming the same general duties). Ocers were inormed that i theyvolunteered or the study, they were agreeing to be assigned to any one o three shits (8-, 10-, or 12-hour). Ocers were also provided with monetary incentives associated with their participation.6

Our power analysis suggested a sample o 300 ocers would be more than sucient to ensurea high level o power or the study, and we obtained 326 volunteers. These ocers were randomlyassigned to one o three conditions: 8-hour shit (n5 109), 10-hour shit (n5 109), or 12-hour shit(n5 108). Our actual sample size was 1287 or Detroit and 147 or Arlington, resulting in a total o275 participants in the study at Time 1. Due to attrition between Time 1 and Time 2, the sample size

or Detroit decreased to 88 participants, while in Arlington it went down to 138, resulting in a totalsample o 226, still sucient to detect medium eect sizes but insucient to detect smaller eects.The nal distribution o participants per cell based on this design is shown in Table 1. Because eveningshits are typically most heavily staed, our study also refected a higher number o evening shitworkers and ewer day shit participants.

Participation

In Figure 1, we present the fow o participants through the experiment rom randomization throughthe outcome measures.

Attrition.There was both voluntary and involuntary attrition in this study. Voluntary attrition occurredwhen ocers chose to drop out o the study either ollowing assignment to their shit or at some otherstage o the study. Involuntary attrition occurred when ocers were discontinued rom the treatmentor excluded rom the analysis or any o the ollowing reasons: (a) the ocer resigned or retired rom

5 These data came rom the2007 Law Enorcement Management and Administrative Statistics, Bureau o Justice Statistics,U.S. Department o Justice. (Reaves, personal communication, September 29, 2010).6 Ocers were provided with $50 payments or completing a survey instrument in both Time 1 and Time 2. In addition,ocers who completed sleep diaries and alertness logs were eligible or a number o $1,000 randomly drawn prizes.7 179 were randomly assigned, and 51 did not ollow through with the treatment.


18/62


19/62

10 Police Foundation

Figure 1. Flow of Participants Through Each Stage of the Experiment


20/62


employment (overtime and o-duty employment). In order to assess these constructs, we used avariety o data collection methods including (a) perormance simulations and other objective measureswe conducted in laboratory settings in the police agencies, (b) daily statistics rom both policedepartments, and (c) surveys and other sel-report instruments we administered. We collected data attwo points in time: beore treatment implementation (the pretest) and at the end o the six-monthstudy period (the posttest).8

Measures

The selection o measures to employ in this study was based on a number o actors including pastdemonstrated reliability and/or validity, delity, ease o administration, and, to a lesser extent,cost. Past research has relied on limited, sel- or supervisory reports o perormance, or inrequentlyoccurring outcomes (e.g., accidents, shootings).9 In order to overcome this problem so that wecould compare all ocers on a variety o outcomes, we used a perormance simulation approachto obtaining data. This approach mimics that o the assessment center, a practical exercise methodused to assess perormance in order to make hiring and promotional decisions across a wide varietyo industries, including law enorcement (see, e.g., Coulton and Feild 1995; Hughes 2006; Krause,Kersting, Heggestad, and Thornton 2006; Thornton and Gibbons 2009). This method allowed us toadminister the same perormance measures under the same conditions across all participating ocers.

Laboratory simulations. Simulation exercises were completed or the pre- and posttest periods.Participants were required to complete ve simulations during the last two to three hours o their workshits to allow researchers to examine outcomes at a point at which atigue was more likely to haveoccurred. To ensure all participants could be scheduled to complete the simulations during the last twoto three hours o their shits, data collection in the sites took rom two to our weeks in Time 1 andabout two weeks in Time 2. The ve simulations, all o which are widely used, include the:

1. Fitness-or-Duty Impairment Screener (FIT): a pupil-response test that detects atigue via ameasure o saccadic velocity (the speed o eye tracking).

2. Behavioral Personnel Assessment Device (B-PAD): a video-based screening tool thatcaptures ocers responses to scenarios to assess interpersonal behaviors.

3. STISIM Drive: a PC-based, high-delity, ully interactive driving simulator or evaluatingsaety by capturing driving mistakes such as accidents, etc.

4. Psychomotor Vigilance Test (PVT): a PC-based stimulus-response method or assessingatigue via reaction time, lapses (ailure to respond), and alse starts (responses when nostimulus was present).

5. MILO/Range 3000: a shooting simulator used primarily or training purposes to assessshooting perormance in realistic interactive situations using a laser-sighted weapon and video.

Police department data. While our primary ocus was on laboratory simulations or the assessmento perormance, we also gathered departmental data o sel-initiated ocer activity and sick leave orequal time periods.

8 Because it took approximately two weeks to administer the perormance simulations in each agency (we were able torun ten to twelve ocers through the exercises per day), the post measure was done as early as 5 months ater thetreatment but no longer than six months ater.9 The low base rates or inrequently occurring events/critical incidents limit the ability to detect group dierences.


21/62


Sel-report instruments. In addition to the ve simulations, each participant was asked to complete aseries o surveys and other instruments, including:

A short survey completed on the day o the simulations, inclusive o hours worked and timeslept in the previous twenty-our hours, caeine ingested, etc., as well as measures o health andsleep disorders (in Time 2 only).10

A sleep diary and alertness log that were completed or a two-week period prior to thelaboratory simulations.

A 456-item survey, entitled Law Enorcement Ofcer Survey o Work Attitudes, PersonalCharacteristics, Health, Saety, and Quality o Lie.This survey was administered using a Scantronbooklet made up o many previously validated and/or used measures o quality o lie, stress,job satisaction, and sleepiness, among others.

Due to the number o outcomes being assessed in this study, we not only categorized outcomemeasures/constructs but in some cases created composite indices made up o a variety o measuresdesigned to capture the same construct, something that also allowed or an increase in statisticalpower.11 In order to combine measures tapping into the same basic construct (e.g., stress), we

transormed scores on individual instruments into standardized scores. This allowed us to convertmeasures with dierent response scales (scales o 1-5 or 1-7, etc.) into standard deviation units, wherethe mean value is 0 and the standard deviation is 1, and then add their values together to create onecomposite score.

In order to increase reliability and limit systematic variability o both outcomes relying on ratedperormance (interpersonal perormance using the B-PAD and shooting perormance using the MILO/Range 3000), rating training was provided or two separate individuals or the two exercises, and thoseindividuals rated both Time 1 and Time 2 perormance across all participants so as to maintain ratingconsistency.

We assessed the reliability o instruments/scales via Cronbachs alpha, a coecient o reliabilityor internal consistency that indicates how well items within a scale measure a single latent construct.

Scales with alpha levels o .70 and above are generally considered strong and alpha levels between.60 and .70 are, in most cases, considered acceptable (Ary, Jacobs, and Razavieh 1985; Cascio 1991).The majority o our outcome measures demonstrated internal consistency, with Cronbachs alphacoecients over .75 or the latent constructs we used as outcome variables.12 Table 2 presents a list omeasures used to assess each construct. Detailed descriptions o the measures and sample items, whererelevant, are provided in Appendix A, and past psychometric properties o the measures are provided inAppendix B, as are the scales reliability coecients obtained in this study.13

10Health and sleep disorder items were obtained only in Time 2 because the identication o such issues at an earlierstage in the experiment would have imposed ethical responsibilities or disclosure o any identied conditions to the

participants. I disclosure occurred, any treatment sought by participants would have created a conound. Instead, weidentied symptoms associated with sleep disorders in Time 2 and then inormed the participants via letter o ourndings indicating that they should seek urther assessment or diagnosis by trained medical personnel.11 By reducing the number o individual analyses to conduct, statistical power is not urther diluted.12 While we recognize that the Cronbachs alpha levels or the shooting perormance and driving simulator were low (.43and .58, respectively), we elt that the items used in each measure were an accurate refection o the dependent variableo interest in the study. The low alphas on the two scales may be attributed to the samples homogeneity and the resultsmay be justied or use in the analyses (see Bernardi 1994).13 For more details on the measures, reer to the ull technical report o this study, entitled The Impact o Shit Length inPolicing on Perormance, Health, Quality o Lie, Sleep, Fatigue, and Extra-Duty Employment(Amendola et al. 2011), availableat www.policeoundation.org/docs/library.html and www.ncjrs.gov.


22/62


Table 2. List of Measures Within Each Construct

CONSTRUCTS & INSTRUMENTS Measuring Author(s) of Measure, Year

Work Performance & Safety

InterpersonalBehavior

Performance

Compositeofexistingmeasures

8B-PADvignettes TheB-PADGroup,Inc.,1994,1999,2010

Losttemperfrequency Czeisleretal.2005

Driving(STISIM) Safety Rosenthaletal.1995,SystemsTechnology,Inc.

Shooting(MILO/Range3000) Performance IESInteractiveTraining

Self-InitiatedActivitiesPerformance Compositeofdepartmentdata

Arrests,citations/summonses,reports,andstops

Health & Stress

Cardiovascular,GastrointestinalHealth Health Bartonetal.1995(Costasubscale)

Stress

Stress


PoliceStressQuestionnaire McCrearyandThompson2004

WorkEnvironmentInventory Libermanetal.2002

PoliceDailyHasslesScale Hart,Wearing,andHeadey1994

SickLeave Health Departmentdata

Quality of Life

QualityofWorkLife(QWL)

QWL


JobSatisfaction(MSQ) Weiss,Dawis,England,andLofquist1967

ScheduleSatisfaction Tucker,Smith,Macdonald,andFolkard1998

OrganizationCommitment AllenandMeyer1990

JobInvolvementScale Kanungo1982

QualityofPersonalLife(QPL)

QPL


Work-FamilyConictI Carlson,Kacmar,andWilliams2000

Work-FamilyConictII Netemeyer,Boles,andMcMurrian1996

Sleep & Fatigue

SleepAmount,Quality(SleepDiary) SleepAmount Heitmann2006(unpublished) SleepinessComposite(subjectivefatigue)

Sleepiness


SleepAssessment Czeisleretal.2005;Heitmann2006(unpub.)

EpworthSleepinessScale Johns1991,1992

Alertness(AlertnessLog)a Alertness Heitmann2006(Karolinskascale)

Fatigue(objective,physiological)

Fatigue


SaccadicVelocity(FIT) PMI,Inc.

PsychomotorVigilanceTest DingesandPowell1985;Thorneetal.2005

SleepDisorders

SleepDisorders


BerlinSleepApnea(adapted) Netzeretal.1999

Insomnia Heitmann2006

SleepDisorders Czeisleretal.2005

Extra-Duty(sleepdiary)

Off-DutyEmploymentandOvertimeb TotalHours

aTheaveragedailyalertnesslevelondaysworkedwascomputedbasedonthefourteen-dayperiod.14bOvertimewascalculatedastheamountofdepartmentalhoursworkedinexcessofeightyhourspertwo-weekperiod.15

14 While it is likely that alertness levels may decrease as the shit goes on, we were interested in the overall alertnessacross groups or the entire shit. Additional analysis may reveal dierences towards the end o the shit.15 In instances when the hours did not total eighty hours due to the ocer being on vacation, sick, or absent, overtimewas assumed to equal zero or that particular ocers data.


23/62


It is important to note that since many o the outcome measures were derived rom sel-reportitems that assessed ocers attitudes and behavior, we examined whether the ocers were presentingthemselves in a more avorable light by including a measure o social desirability bias. That measureconsisted o a 33-item scale developed by Crowne and Marlowe (1960) to identiy and potentiallyeliminate any cases where an ocer may be responding untruthully. In our sample, the scores rangedrom 4.0 to 29.0 with a mean o 19.1 and a standard deviation o 5.0. All o the participating ocers

ell within the selected criteria (being 2.0 or less standard deviations rom the mean) and, as a result,none o the participants data needed to be excluded rom our study. This is not surprising consideringour sample consisted o police ocers whose jobs are in jeopardy i they are dishonest.

Data Analysis

The results presented in this study are based primarily on comparisons o means or the threetreatment conditions (8-, 10-, and 12-hour shit lengths) and statistical tests to indicate the probabilityo obtaining a dierence between the three groups. Our alpha level or rejection o the null hypothesiswas set atp , .05. Since there were virtually no rigorous past experimental designs testing the impacto shit length, our hypotheses were nondirectional and, as such, we employed two-tailed tests. Missingdata were excluded rom the analysis on a case-by-case basis, so ourn or any o our statistical testsincludes all o the valid cases in the dataset.

Analytical approach. Our analytic approach was an analysis o covariance (ANCOVA), where thepretest measures served as the covariate so as to control or any initial dierences. This analysis wasconducted using General Linear Model (GLM) analyses and the outcome scores were adjusted basedon the pretest measures. Our independent variables, then, were the primary study variable, shit length(8-, 10-, or 12-hour), and our blocking variablestime o day o the shit (day, evening, midnight),and the agency itselas well as any interactions between the shit length and the latter two variables.16

Eect size.Throughout the discussion o the results o this study, we present Cohens(Cohen 1988)eect size index to measure the magnitude o the dierences.17 We relied on Cohens criteria orinterpreting the magnitude o the eects as small (5 .10), medium (5 .25), and large (5 .40),despite the act that Lipsey has argued that eect size values o .10 and larger could easily be opractical signicance (2000, 109). However, we do interpret all eects o .10 as meaningul assuggested by Lipsey.

16 I the block by treatment interactions did not achieve statistical signicance at the .05 level, we excluded them romthe analyses; see Fleiss (1982) who actually suggests an even more conservative alpha level op , .01.17 Cohens eect size index ormula is the square root o the Eta squared (2) divided by 1 minus 2 (Cohen 1988,280288). 2 is calculated as the ratio o the eect variance (SSeect) to the total variance (SStotal) 25 SSeect / SStotal.The value or the SStotal in the ormula includes the SS or each o the eects and the error term, but does not include theSS or the intercept in the GLM models. Note the 2 column in one version o SPSS provided only partial 2 output. Assuch, 2 was manually calculated.


24/62


25/62


Table 3. Results of GLM Analysis and Associated Effect Sizes

OUTCOME MEASURE F (df)a P Effect size: Cohens f

Performance and Safety

InterpersonalCommunication 1.73(2,214) .180 .12

Driving .264(2,214) .768 .04 Shooting .481(2,111) .619 .08

Self-initiatedActivity 1.22(2,252) .298 .07

Health and Stressb

CardiovascularHealth .007(2,222) .993 .01

GastrointestinalProblems .809(2,202) .447 .08

WorkStress .319(2,197) .727 .03

SickLeave .16(2,271) .854 .03

Quality of Life

QualityofPersonalLife .303(2,192) .739 .04

QualityofWorkLifeComposite 3.94 (2,197) .021 .16

ShiftLength*Site 4.76 (2,197) .010 .19

Sleep, Fatigue, Alertness

AverageSleepAmount 3.23 (2,147) .043 .19

AverageSleepQuality .865(2,147) .423 .09

Sleepiness(subjectivefatigue) 5.75 (2,222) .004 .20

Alertness 4.42 (2,132) .014 .21


Fatigue(FIT)(objective) .098(2,201) .906 .02

Fatigue(PVT)(objective) 1.49(2,214) .228 .11

SleepDisordersc .208(2,224) .812 .04

Extra-Duty Employment

Off-DutyEmployment .241(2,146) .786 .05

Overtime 15.42 (2,145) .000 .42


aTheFwascalculatedtakingintoaccountthepretestmeasure,siteofstudy,andtimeofday.bGastrointestinalandcardiovascularproblemswerenotmeasuredduringthepretest.cSleepdisorderswerenotassessedduringthepretest.

greater (mean 5 7.86) than the average hours o sleep among ocers on the 8-hour shit (mean 57.27, p5 .036), but not the 12-hour group (mean 5 7.63, p5 ns).

Sleepiness/Fatigue (subjective)

To assess sleepiness, we included items rom the Harvard Study o Work Hours (Czeisler et al. 2005) andour sleep consultant, as well as the entire Epworth Sleepiness Scale. There was a signicant eect o shitlength on the sleepiness composite, F(2,222)5 5.75, p5 .004. Testing the magnitude o the sleepinessconstruct resulted in a small to medium eect (5 .20) or shit length. A pairwise comparison test o


26/62


the adjusted group means revealed the average level o sleepiness or ocers in the 12-hour shit (0.771)was signicantly higher than or those on the 8-hour shit (mean5 -0.721, p 5 .003).

Alertness

There was a signicant eect or alertness based on shit length, F(2,132) 5 4.42, p5 .014,representing a small to medium eect size (5 .21). A pairwise comparison test o the adjusted groupmeans revealed the average level o alertness or ocers in the 12-hour shit was signicantly lower(mean 5 6.11) than the average alertness levels among ocers on the 8-hour (mean 5 6.74, p 5 .012),but not the 10-hour shit (mean 5 6.31, p5 ns).

There was also a signicant treatment by site interaction or level o alertness, F(2,132) 5 6.01,p 5 .003. In Arlington, the main eect o shit length was F(2,91) 5 8.47, p5 .000, which translates toan eect size o5 .29 (medium). Those working 12-hour shits were signicantly less alert (mean 56.10) than those on the 10-hour shit (mean 5 6.74, p5 .000) and the 8-hour group (mean 5 6.53,p5 .037). While the 8-hour ocers in Detroit appeared more alert than those on either compressedschedule, those dierences were not statistically signicant.

Overtime WorkedWhen considering the amount o overtime hours taken by ocers, there was a signicant dierenceacross groups or shit length F(2,145) 5 15.42, p5 .000, which represents a large eect (5 .42). Apairwise comparison o the group means adjusted or the eect o the pretest overtime hours revealedthe average amount o overtime among ocers in the 8-hour shit (5.75 hours) was signicantlyhigher than the average hours or ocers in the 10-hour (mean 5 0.97 hours, p5 .000) and 12-hour(mean 5 1.89 hours, p5 .000).

There was, however, a signicant interaction eect or shit length by study site when consideringamount o overtime, F(2,145) 5 5.86, p 5 .004. While the analysis revealed ocers on 8-hour shits inboth sites worked the most amount o overtime, those in Detroit worked considerably more (mean 58.76)19 than those in Arlington (mean 5 2.74 hours). In Detroit, the main eect o shit length was

F(2,44) 5 7.39, p 5 .002, which translates to an eect size o5 .53 (large). The 8-hour group hadsignicantly more overtime (mean 5 9.01 hours) than both the 10-hour group (mean 5 1.49, p5.002) and 12-hour group (mean 5 3.02, p5 .013), indicating either type o compressed schedule inDetroit results in less overtime than or 8-hour shits. In Arlington, the main eect o shit length wasF(2,104)5 3.03, p5 .053,20 which translates to an eect size o5 .25 (medium). Similar to Detroit,the ndings in Arlington were all in the same direction, although the paired comparisons did not reachstatistical signicance.21

Nonsignificant FindingsWork Perormance and Saety

We did not identiy any dierences across shit length groups or any o our perormance and saetymeasures (shooting perormance, driving saety, sel-initiated activity on the job, or interpersonal

19 Adjusted or pretest overtime hours.20 Just reaching statistical signicance.21The 8-hour group also had the most overtime (mean 5 2.54 hours) as compared to those on 10-hour shits who hadthe least (mean5 0.75, p5 .064, approaching signicance), and the 12-hour shit ell in between (mean5 1.04 hours,p5 ns).


27/62


behavior in simulated encounters with suspects or the general public). However, it is important tonote that or composites in which the measures were not internally consistent (shooting and drivingperormance), it is not known whether shit length may have an eect on specic items within eachcomposite measure.22

Health and Stress

Shit length did not have an impact on cardiovascular health or gastrointestinal problems,our composite stress measure, or the amount o sick leave taken. It should be pointed out thatcardiovascular health is infuenced by a number o actors and problems tend to build over time.Because our study only looked at changes over a six-month period, we cannot rule out the possibilitythat longer-term eects may occur. However, there is reason to believe the use o sick leave,gastrointestinal upsets, and subjective experiences o stress would be maniest in the shorter term (sixmonths), based on the short-term nature o these measures and prior research on health and stress.

Quality o Personal Lie

Due to the comprehensive nature o this study, we had to limit the number o outcome variables we

examined. As such, while many studies have examined various aspects o personal lie such as leisuretime, in our study we chose to ocus on the more job-relevant concern o the impact o shit schedules(shit length, number o days worked) on amily via measures o work-amily confict. No statisticallysignicant dierences were identied on work-amily confict based on the shit length.

Sleep Quality, Fatigue (objective), and Sleep Disorders

While we did identiy meaningul dierences in the amount o sleep obtained based on the shitlength (those on 10-hour shits got signicantly more sleep than those on 8-hour shits), there wereno observed dierences in the reported quality o sleep, inconsistent with some prior ndings.Additionally, the length o the shit did not dierentiate those with reported sleep disorders or thoseor whom there was an indicator o potential sleep apnea. Although we noted important concerns with

regard to sleepiness and lack o alertness or those on 12-hour shits, there were no dierences betweenshit length groups or physiological (objective) measures o atigue (saccadic velocity o pupil usingthe FIT and reaction time using the PVT).

O-Duty Employment

Finally, while we ound a very meaningul dierence across shit lengths with regard to the amounto overtime worked, it is perhaps surprising we did not nd a signicant dierence based on o-duty employment. While those on 10-hour shits had a lower number o hours o o-duty work, thatnding was not statistically signicant.

22 Composite measures were used to assess broader constructs (e.g., driving saety) in order to maximize power. The actthat the driving and shooting composites demonstrated low internal consistency may indicate that the items comprisingeach composite are independent. The analysis o each item independently would have reduced power signicantly.Consequently, it is possible that certain types o risky driving behaviors (lane deviations versus speeding) may beindependently impacted by shit length, whereas as a group they were not.


28/62

19

SUmmArY Of CrOSS-inDUSTrYreSeArCh On COmpreSSeDWOrKWeeKS

Early Research

During the 1970s, there was increased interest in CWWs across industries (includinglaw enorcement), but much o what was known about them refected anecdotalinormation, opinions, or data derived rom indirect (Calvasina and Boxx 1975)or subjective measures. Some examples include highly subjective measures such asusing chies observations o amily relations (Durrett 1983), or amily spendingassociated with CWWs (Gavney, Calderwood, and Knowles 1979; Goodale andAagaard 1975).

In 1981, however, Ronen and Primps (1981) reviewed the 1970s literature onCWWs across industries and ound ourteen studies in which various impacts oCWWs were examined. In reviewing this early work, they ound employee reactionsand attitudes toward CWWs were mostly avorable (e.g., Goodale and Aagaard 1975;Nord and Costigan 1973; Northrup, Wilson, and Rose 1979; Poor and Steele 1970).Most workers on compressed schedules had higher job satisaction (e.g., Hodgeand Tellier 1975; Ivancevich 1974), reported more leisure time, and believed thecompressed schedule had beneted their marriage and/or social lie (Goodale andAagaard 1975). However, atigue was reportedly higher or those on compressedschedules as well (e.g., Hodge and Tellier 1975; Maklan 1977; Poor and Steele 1970;Volle, Brisson, Prusse, Tanaka, and Doyon 1979).

When examining perormance, however, Ronen and Primps (1981) reportedmixed results. Benets noted included improvements in supervisory ratings operormance (e.g., Foster, Latack, and Reindl 1979; Ivancevich 1974), reductions inabsenteeism (e.g., Foster et al. 1979; Goodale and Aagaard 1975; Nord and Costigan1973), and a 10 percent reduction in overtime or those on 10-hour shits (Goodaleand Aagaard 1975). Others, however, reported disadvantages associated with CWWsin terms o supervisory perceptions o productivity, quality, service to others, andwork coordination (Goodale and Aagaard 1975).


29/62


In 1990, Moores conducted a meta-analysis o the eects o CWWs and ound that or the twostudies done in the 1980s, there was a moderate eect on absenteeism, and or the three in the 1970s,there was a small eect, whereas CWWs were associated with decreased absenteeism. For the six studieso productivity conducted with blue-collar workers, the mean eect was just 0.10. When consideringjob satisaction, however, Moores (1990) ound a large eect (1.16) across ve studies, temperedby an increase in atigue with a mean eect o -0.35. Using a more rudimentary counting approach,

Moores (1990) reported that in all studies o satisaction (n 5 37) and productivity (n5 42) therewere increases associated with CWWs and decreases in absenteeism (n5 34), turnover (n5 20), andovertime (n5 4).

More Recent Findings

While there were many limitations o early research, a more recent review o the impacts o CWWsrelied on teen rigorous studies (Josten, Ng-A-Tham, and Thierry 2003). In that review, researchersnoted positive eects were more requently reported in studies prior to and including 1982, whereaslater studies tended to nd more negative eects associated with CWWs. Across earlier studies,employees perceptions were more avorable than supervisors regarding CWWs (Cohen and Gadon

1978; Ronen and Primps 1981), as were those o younger workers.Since the earlier research o the 1970s and 1980s, a growing body o research has accumulated

on compressed schedules mainly due to concerns over saety or other important considerations ina variety o industries. Harrington (1994), or example, noted the increased amount o studies onCWWs in the early 1990s. Indeed, over the last couple o decades, research has been conducted acrossa variety o domains including, but not limited to, medicine (see, e.g., Burke 2003; Fitzpatrick, While,and Roberts 1999; McGettrick and ONeill 2006); manuacturing and mining (see, e.g., Duchon,Smith, Keran, and Koehler 1994, 1997; Northrup 1991); utilities and power plants (see, e.g., Mitchelland Williamson 2000; Rosa 1992, 1993); and transportation, including trucking (see Aamodt 2010;Hamelin 1987), railroads (Hrm, Sallinen, Ranta, Mutanen, and Mller 2002; Sallinen et al. 2005),and aviation (Schroder, Rosa, and Witt 1998).

Overall, when considering compressed schedules, the ndings have suggested: (a) employeesgenerally and overwhelmingly avor them (e.g., Armstrong-Stassen 1998; Axelsson 2005; Bendak 2003;Dowd, Oakley, French, Fischer, and Storm 1994; Duchon et al. 1997; Duchon, Keran, and Smith 1994;Dunham, Pierce, and Castaeda 1987; Facer and Wadsworth 2010; Lowden, Kecklund, Axelsson, andAkerstedt 1998; Pierce and Dunham 1992; Rosa and Colligan 1992); (b) compressed schedules areassociated with improvements in home and personal lie including increased leisure, personal, andamily time or greater satisaction (e.g., Armstrong-Stassen 1998; Knauth 2007; Lowden et al. 1998;McGettrick and ONeill 2006; Mitchell and Williamson 2000; Facer and Wadsworth 2010), as well asreduced work-amily confict (Facer and Wadsworth 2008, 2010); (c) longer days tend to be associatedwith greater atigue (e.g., Armstrong-Stassen 1998; Bendak 2003; Knauth 2007); and (d) those onCWWs oten get more or better sleep than those on traditional, 8-hour schedules (e.g., Duchon et al.1997; Mitchell and Williamson 2000).23 While these ndings were mostly derived rom sel-reporteddata, they tended to be consistent regardless o the research designs employed or the scientic rigor othose studies.

23 However, Axelsson (2005) noted that sleep duration is impacted by shit type, start and end times o shit, andrecovery time between shits.


30/62


31/62


32/62


33/62


the community associated with CWWs, Cunningham later (1989) ound that police managers eltthey were more likely to see their ocers more regularly when they were assigned to 8-hour shits ascompared to 12-hour shits. Furthermore, Cunningham (1990) argued longer hours could infuenceocer interactions with the public. Recently, DiMambro (2008) reported that a number o agenciesthat had adopted 12-hour shits indicated a deleterious eect upon communication among ocersor between ocers and their supervisors. This appears to be consistent with much earlier research by

Brown (1974), who reported that almost hal o the agencies he surveyed that had adopted 12-hourshits reported greater diculty with cross-shit communication. Because these ndings are based onsubjective reports, they should be interpreted with increased scrutiny.

Research rom other industries has also been mixed. For example, McGettrick and ONeill (2006)reported poorer communication among medical sta when working CWWs, whereas others haveshown improvements in internal communication (Johnson and Sharit 2001). On the other hand, Facerand Wadsworth (2008, 2010) noted those on CWWs had an improved ability to interact with citizens,but these ndings too were based on subjective data.

In our randomized experiment, we examined interpersonal interactions with communitymembers in a less subjective, laboratory-based exercise and ound no dierences based on shit lengthgroupings.25 However, we did not examine internal communication associated with CWWs.

Cognitive perormance.Findings related to cognitive perormance across industries have been mixedas well. There is some evidence that CWWs are associated with lower cognitive perormance (e.g.,grammatical reasoning, reaction time, motor abilities) when comparing workers on 12-hour shits tothose on 8-hour shits (e.g., Duchon et al. 1994; Rosa and Bonnet 1993; Rosa and Colligan 1992).In a longitudinal study o control room operators at a continuous processing plant, workers on 12-hour shits displayed poorer perormance on a series o cognitive, perceptual, and motor skills testsas compared to those on 8-hour shits (Breaugh 1983). These studies suggest detrimental impacts o12-hour shits.

On the other hand, others have not ound dierences across shit lengths when examining criticalthinking skills (e.g., Washburn 1991; Bernreuter and Sullivan 1995), or cognitive unctioning among

nurses (e.g., Campolo et al. 1998; Fields and Loveridge 1988). When considering the dierencesbetween 8- versus 10-hour shits among air trac control specialists on cognitive tasks such as reactiontime or digit addition, researchers also did not nd any signicant dierences (Schroeder et al. 1998).In our examination o reaction time using the Psychomotor Vigilance Test in our sample o policeocers, we also did not detect any cognitive dierences across shit length groups.

Saety

While the research on saety in policing has ocused on atigue rather than shit length (see, e.g.,Vila 2000), research across industries has suggested increased accidents, incidents, or accident riskassociated with longer shits. However, some o that research has tended to ocus on particularly longshits. For example, Caruso et al. (2004) conducted a systematic review and concluded that injuries

increased or those working long hours, especially or very long shits and when 12-hour shits arecombined with more than orty hours o work per week. Additionally, in a study o truck drivers,Hamelin (1987) ound that accident risk was quite high ater driving or more than eleven hours.Furthermore, in a review o duty-period extensions or the Federal Aviation Administration (FAA),the Battelle Memorial Institute (1998) indicated shits at or above twelve hours are associated with

25 Our measure o interpersonal behavior was an expert rating using the B-PAD. While ratings are typically consideredsubjective, this measure was much more objective because the rater received extensive training by the B-PAD Group, therater did not know the participants, and we relied on one rater across all participants to minimize rating errors.


34/62


35/62


36/62


37/62


38/62


39/62


40/62


41/62


42/62


or those on 8-, 10-, and 12-hour shits. Surprisingly, we ound those ocers working 8-hour shitsaveraged ve times the amount o overtime than those on 10-hour shits, and three times that o thoseworking 12-hour shits. However, the dierences across shit lengths or o-duty employment werenot statistically signicant. Our results are consistent with ndings in other industries. For example,some have noted decreases in paid overtime or those on CWWs (Facer and Wadsworth 2010), which isconsistent with an earlier nding by Foster et al. (1979) who ound a 33 percent reduction in overtime

or those on CWWs.

Other Outcomes

Some researchers have identied other organizational outcomes associated with compressed schedules,such as reduced commuting costs (e.g., Price 1981). For example, State o Utah employees surveyed byFacer and Wadsworth (2010) also reported reduced commuting costs or those on 4/10 schedules, alogical nding given ewer days at work and one that is consistent with assertions made by many whopromote such schedules. Sundo and Fujii (2005) reported commute times may be urther reduced onCWWs due to non-peak hour commutes. Facer and Wadsworth (2010) also noted that when the Stateo Utah examined energy consumption associated with a 4/10 compressed schedule, they noted over a10 percent decrease in energy use or an overall statewide reduction o $502,000. Others have reportedthat CWWs result in decreased leave and absenteeism (Facer and Wadsworth 2010; Foster et al. 1979).Hung (2006) suggests a potential cost savings with CWWs, but it appears to be based on minimizingsta levels. Although this is not based on 24/7 operations and is hypothetical rather than actual, theauthor has previously documented savings in commuting costs (Hung 1996).


43/62


44/62


45/62


46/62


and yet our ndings were consistent with some past research. Overall, this study has revealed thatthere can be benets associated with CWWs, particularly the 4/40 conguration (10-hour shits).While those benets do not inure to those on 12-hour shits, our ndings do not raise signicantconcerns associated with the implementation o 12-hour shits. However, given the ndings acrossindustries and our ndings regarding alertness and sleepiness, agencies considering or currently using12-hour (or longer) shits should careully monitor ocers total work hours including o-duty and

overtime work, ensure there is sucient recovery time between shits, routinely query ocers regardingtheir perceived levels o atigue (at various times throughout the shit), and make adjustments toactivities accordingly. It should also be emphasized that simply implementing an 8- or 10-hour shitis not sucient or reducing atigue, improving quality o lie, or reducing risks to saety, health, andperormance. Many agencies continue to employ double shits in certain circumstances, rely on rapidlyrotating shits (something that has long shown to be detrimental on many levels), or routinely holdover certain personnel or more than our hours in a typical day. Such activities should be avoidedand agencies should instead begin to ocus on ocer wellness, saety, and health by providing variousprograms and implementing and enorcing policies designed to protect the saety and health o ocersand their amilies, as well as that o the general public.


47/62


48/62


49/62


50/62


51/62


52/62

43

referenCeS

Aamodt, M.G. (2010). Industrial Organizational Psychology: An Applied Approach (6th ed.).Belmont, CA: Wadsworth CENGAGE Learning.

kerstedt, T. (1997). Readily available countermeasures against operator atigue. Paper presented

at the International Conerence on Managing Fatigue in Transportation, Tampa, FL.Alberta Human Resources and Employment. (2004). Fatigue, extended work hours, and

saety in the workplace. Ergonomics, 15, 110.

Allen, N., & Meyer, J. P. (1990). The measurement and antecedents o aective, continuance,and normative commitment to the organization.Journal o Occupational Psychology, 63,118.

Amendola, K. L., Weisburd, D., Hamilton, E., Jones, G., Slipka, M., Heitmann, A., Shane,J., Ortiz, C., & Tarkghen, E. (2011). The Impact o Shit Length in Policing on Perormance,Health, Quality o Lie, Sleep, Fatigue, and Extra-Duty Employment.Washington, DC: PoliceFoundation.

Andrusaitis, S. F., Oliveira, R. P., & Filho, T. E. P. B. (2006). Study o the prevalence and riskactors or low back pain in truck drivers in the state o So Paulo, Brazil. Clinics, 61(6),

503510. doi: 10.1590/S1807-59322006000600003Armstrong-Stassen, M. (1998). Alternative work arrangements: Meeting the challenges.

Canadian Psychology/ Psychologie Canadienne, 39(12), 108123. doi: 10.1037/h0086799

Ary, D., Jacobs, L.C., & Razavieh, A. (1985). Introduction to research in education (3rd ed.).New York, NY: Holt, Rinehart and Winston.

Axelsson, J. (2005). Long shits, short rests and vulnerability to shit work. (Doctoral Dissertation)Stockholm University: Stockholm. Retrieved rom http://www.su.diva- portal.org/smash/

get/diva2:194082/FULLTEXT01

Baltes, B. B., Briggs, T. E., Hu, J. W., Wright, J. A., & Neuman, G. A. (1999). Flexible andcompressed workweek schedules: A meta-analysis o their eects on work-related criteria.

Journal o Applied Psychology, 84(4), 496513. doi:10.1037/0021-9010.84.4.496

Bambra, C. L., Whitehead, M. M., Sowden, A. J., Akers, J., & Petticrew, M. P. (2008). A

hard days night? The eects o compressed working week interventions on the healthand work-lie balance o shit workers: A systematic review.Journal o Epidemiology &Community Health, 62, 764777. doi: 10.1136/jech.2007.067249

Barter Trenholm, S. B. (1997). The satisaction o police ofcers and their spouses with 12-hourshit work schedules (Doctoral Dissertation). Memorial University o Newoundland,Canada.

Barton, J., Costa, G., Smitt, L. R., Spelten, E. R., Totterdell, P. A., & Folkard, S. (1995). TheStandard Shitwork Index Manual: A battery o questionnaires or assessing shitwork-related problems. Work and Stress, 9, 330.


53/62


Battelle Memorial Institute, JIL Inormation Systems (1998).An overview o the scientifc literature concerning atigue,sleep, and the circadian cycle. Retrieved rom http://c.alpa.org/internet/projects/tdt/backgr/batelle.htm

Bayley, D. H., & Worden, R. F. (1998). Police overtime: An examination o key issues. National Institute o Justice inResearch in Brie. Retrieved rom http://www.ncjrs.gov/pdles/167572.pd

Bendak, S. (2003). 12-hour workdays: Current knowledge and uture directions. Work & Stress, 17(4), 321336.doi: 10.1080/02678370310001643478

Bennet, S. A. (2003). Flight crew stress and atigue in low-cost commercial operations: An appraisal. InternationalJournal o Risk Assessment and Management, 4(2/3), 207231. doi:10.1504/IJRAM.2003.003528

Bernardi, R. A. (1994). Validating research results when Cronbachs Alpha is below .70: A methodologicalprocedure. Educational and Psychological Measurement, 52, 766775.

Bernreuter, M., & Sullivan, M. (1995). Survey and critique o studies related to shit length variations innursing rom 1970 to 1993. International Journal or Nursing Studies, 32(2), 188197. doi:10.1016/0020-7489(94)00026-G

Bommer, W. H., Johnson, J. L., Rich, G. A., Podsako, P. M., & MacKenzie, S. B. (1995). On the interchangeabilityo objective and subjective measures o employee perormance: A meta-analysis. Personnel Psychology, 48(3),587602. doi: 10.1111/j.1744- 6570.1995.tb01772.x

B-PAD. (n.d.). The Challenge o selecting tomorrows peace ofcers. Sonoma, CA: The B-PAD Group, Inc.

The B-PAD Group (1994, 1999, 2010). http://www.bpad.com/bPAD_howBPADWorks.html

Brown, P. (1974). Cycle scheduling or eight and ten hour days. The Police Chie, 3841.Breaugh, J. (1983). The 12-hour work day: Diering employee reactions. Personnel Psychology, 36, 277288.

Burke, R. J. (2003). Length o shit, work outcomes, and psychological well-being o nursing sta. InternationalJournal o Public Administration, 26(14), 16371646. doi:10.1081/PAD-120024415

Calvasina, E. J., & Boxx, W. R. (1975). Eciency o workers on the our-day workweek.Academy o ManagementJournal, 18(3), 604610. doi:10.2307/255689

Campolo, M., Pugh, J., Thompson, L., & Wallace, M. (1998). Pioneering the 12-hour shit in Australiaimplementation and limitations.Australian Critical Care, 11(4), 112115. doi:10.1016/S1036-7314(98)70496-5

Carlson, D. S., Kacmar, K. M., & Williams, L. J. (2000). Construction and initial validation o a multidimensionalmeasure o work-amily confict.Journal o Vocational Behavior, 56(2), 249276. doi:10.1006/jvbe.1999.1713

Caruso, C. C., Hitchcock, E. M., Dick, R. B., Russo, J. M., & Schmit, J. M. (2004). Overtime and extended work shits:Recent fndings on illness, injuries, and health behaviors (DHHS Publication No. 2004143). Cincinnati, OH:

National Institute or Occupational Saety and Health.Cascio, W. F. (1991). Applied psychology in personnel management (4th ed.). Englewood Clis, NJ: Prentice Hall.

Chervin, R. D., & Aldrich, M. S. (1999). The Epworth Sleepiness Scale may not refect objective measures osleepiness or sleep apnea.Neurology, 52(1), 125131.

Cochrane, G. (2001, July). The eects o sleep deprivation. FBI Law Enorcement Bulletin, 2225.

Cohen, J. (1988). Statistical power analysis or the behavior sciences (2nd ed.). Hillsdale, NJ: Lawrence ErlbaumAssociates, Inc.

Cohen, A. R. & Gadon, H. (1978).Alternative work schedules: Integrating individual and organizational needs. Reading,MA: Addison-Wesley.

Coulton, G. F. & Feild, H. S. (1995). Using assessment centers in selecting entry-level police Ocers: Extravaganceor justied expense? Public Personnel Management, 24(2), 223254.

Cramer, W. (2007, April). Fighting atigue in manuacturing. Saety Solutions. Retrieved rom http://www.saetysolutions.net.au/articles/976-Fighting-atigue-in-manuacturing

Crowder, R. H., Jr. (1982). The our-day ten-hour workweek. Personnel Journal, 61, 2628.

Crowne, D. P., & Marlowe, D. (1960). A new scale o social desirability independent o psychopathology.Journalo Consulting Psychology, 24(4), 349354. doi:10.1037/h0047358

Cruz, C., Rocco. P. D., & Hackworth, C. (2000). Eects o quick rotating shit schedules on the health andadjustment o air trac controllers.Aviation, Space, and Environmental Medicine, 71(4), 400407.

Cunningham, J. B. (1981). Exploring the impact o a ten-hour compressed shit schedule.Journal o OccupationalBehaviour, 2(3), 217222. doi:10.1002/job.4030020307


54/62


55/62


Fleiss, J. L. (1982). Multicentre clinical trials: Bradord Hills contribution and some subsequent developments.Statistics in Medicine, 1(4), 353359. doi:10.1002/sim.4780010413

Folkard, S., & Lombardi, D. A. (2004, April).Modeling the impact o the components o long work hours on injuries andaccidents. Paper presented at the National Conerence: United Kingdom.

Folkard, S., & Tucker, P. (2003). Shit work, saety and productivity. Occupational Medicine, 53(2), 95101.doi:10.1093/occmed/kqg047

Foster, L. W., Latack, J. C., & Reindl, L. J. (1979). Eects and promises o the shortened work week. Paperpresented at the 39th annual meeting o the Academy o Management, Atlanta, Georgia.

Four-day week gets more pay, The. (1971). Business Week. July, p. 33.

Fournet, M. G. (1983). The 12-hour shit: A workable alternative. The Police Chie, 5759.

Garbarino, S., Nobili, L., Beelke, M., Balestra, V., Cordelli, A., & Ferrilo, F. (2002). Sleep disorders anddaytime sleepiness in state police shitworkers.Archives o Environmental Health, 57(2), 167173. doi:10.1080/00039890209602932

Gavney, R., Calderwood, J., & Knowles, L. (1979, February). Attitude o patrol ocers and wives toward a our-dayworkweek. The Police Chie, 3335.

Goodale, J. G., & Aagaard, A. K. (1975). Factors relating to varying reactions to the 4-day workweek.Journal oApplied Psychology, 60(1), 3338. doi:10.1037/h0076345

Grosswald, B. (2004). The eects o shit work on amily satisaction. Families in Society, 85(3), 413424. doi:

10.160611044-3894.1507Hamelin, P. (1987). Lorry drivers time habits in work and their involvement in trac accidents. Ergonomics,

30(9), 13231333. doi: 10.1080/00140138708966026

Hrm, M., Sallinen, M., Ranta, R., Mutanen, P., & Mller, K. (2002). The eect o an irregular shit system onsleepiness at work in train drivers and railway trac controllers.Journal o Sleep Research, 11(2), 141151.doi:10.1046/j.1365-2869.2002.00294.x

Harrington, J.M. (1994). Shit work and health: A critical review o the literatures on working hours.Annals oAcademy o Medicine Singapore, 23(5), 699705.

Hart, A., & Krall, S. (2007). Productivity: Do 89 hour shits make a dierence?Annals o Emergency Medicine,50(3), S69S70. doi: 10.1016/j.annemergmed.2007.06.372

Hart, P. M., Wearing, A. J., & Headey, B. (1994). Perceived quality o lie, personality, and work experiences:Construct validation o the police daily hassles and uplits scales. Criminal Justice and Behavior, 21(3), 283

311. doi: 10.117710093854894021003001Hodge, B. J., & Tellier, R. D. (1975). Employee reactions to the our-day week. Caliornia Management Review,

18(1), 2530.

Hughes, F. (2006, August). Does the benet outweigh the cost? Using assessment centers in selecting middlemanagers. The Police Chie, 106111.

Hung, R. (1996). Using compressed workweeks to reduce work commuting. Transportation Research Part A: Policyand Practice, 30(1), 1119. doi: 10.1016-0965-8564(95)00012-7

Hung, R. (2006). Using compressed workweeks to save labour cost. European Journal o Operational Research, 170,319322. doi: 10.1016/j.ejor.2004.09.043

Ivancevich, J. M. (1974). Eects o the shorter workweek on selected satisaction and perormance measures.Journal o Applied Psychology, 59(6), 717721. doi: 10.1037/h0037504

Ivancevich, J. M., & Lyon, H. L. (1977). The shortened workweek: A eld experiment.Journal o Applied Psychology,62(1), 3437. doi:10.1037/0021-9010.62.1.34

Jacques, E. (2010).A resh look at 12-hour shits. Redord, MI: Police Ocers Association o Michigan. Retrievedrom: http://www.poam.net/main/journal/resh-look-at-12-hourshits.html

Jeanmonod, R., Brook, C., Winther, M., Pathak, S., & Boyd, M. (2008). Declining resident productivity over timein the emergency department.Annals o Emergency Medicine, 51(4), doi:10.1016/j.annemergmed.2008.01.065

Johns, M. W. (1991). A new method or measuring daytime sleepiness: The Epworth sleepiness scale. Sleep, 14(6),540545.

Johns, M. W. (1992). Reliability and actor analysis o the Epworth sleepiness scale. Sleep, 15(4), 376381.


56/62


Johnson, M. D., & Sharit, J. (2001). Impact o a change rom an 8-h to a 12-h shit schedule on workers andoccupational injury rates. International Journal o Industrial Ergonomics, 27(5), 303319. doi:10.1016/S0169-8141(00)00058-5

Josten, E. J. C., Ng-A-Tham, J. E. E., & Thierry, H. (2003). The eects o extended workdays on atigue, health,perormance and satisaction in nursing.Journal o Advanced Nursing, 44(6), 643652. doi:10.1046/j.0309-2402.2003.02854.x

Kanner, A. D., Coyne, J. C., Schaeer, C., & Lazarus, R. (1981). Comparison o two modes o stress measurement:Daily hassles and uplits versus major lie events.Journal o Behavioral Medicine, 4(1), 139. doi:10.1007/BF00844845

Kanungo, R. N. (1982). Measurement o job and work involvement.Journal o Applied Psychology, 67(3), 341349.doi:10.1037/0021-9010.67.3.341

Knauth, P. (2007). Extended work periods. Industrial Health, 45(1), 125136. doi:10.2486/indhealth.45.125

Knutson, K. L., Rathouz, P. J., Yan, L. L., Liu, K., & Lauderdale, D. S. (2006). Stability o the Pittsburgh SleepQuality Index and the Epworth Sleepiness Questionnaire over 1 year in early middle-aged adults: TheCARDIA Study. Sleep, 29(11), 15031506.

Krause, D. E., Kersting, M., Heggestad, E. D., & Thornton, G. C. (2006). Incremental validity o assessment centerratings over cognitive ability tests: A study at the executive management level. International jounal o Selectionand Assessment, 14(4), 360371.

Laundry, B. R., & Lees, R. E. M. (1991). Industrial accident experience o one company on 8- and 12-hour shitsystems.Journal o Occupational Medicine,33(8), 903. doi:10.1097/00043764-199108000-00018

Lees, R. E. M. & Laundry, B. R. (1989). Comparison o reported workplace morbidity in 8-hour and 12-hour shitsin one plant.Journal o Social Occupational Medicine, 39(3), 8184. doi:10.1093/occmed/39.3.81

Liberman, A. M., Best, S. R., Metzler, T. J., Fagan, J. A., Weiss, D. S., & Marmar, C. R. (2002). Routine occupationalstress and psychological distress in police. Policing: An International Journal o Police Strategies & Management,

25(2), 421439. doi:10.1108/13639510210429446

Lilley, R., Feyer, A. M., Kirk, P., & Gander, P. (2002). A survey o orest workers in New Zealand. Do hours o work,rest, and recovery play a role in accidents and injury?Journal o Saety Research, 33(1), 5371. doi:10.1016/S0022-4375(02)00003-8

Lipscomb, J. A., Trinko, A. M., Geiger-Brown, J., & Brady, B. (2002). Work-schedule characteristics and reportedmusculoskeletal disorders or registered nurses. Scandinavian Journal o Work, Environment & Health, 28(6),394401.

Lipsey, M. (2000). Statistical conclusion validity or intervention research: A signicant (p, .05) problem. In L.Bickman (Ed.), Validity and social experimentation: Donald Campbells legacy(pp. 101120). Thousand Oaks, CA:Sage.

Loudoun, R. (2008). Balancing shitwork and lie outside work: Do 12-hour shits make a dierence?AppliedErgonomics, 39, 572579. doi: 10.1016/j.apergo.2007.12.004

Lowden, A., Kecklund, G., Axelsson, J., & Akerstedt, T. (1998). Change rom an 8-hour shit to a 12-hour shit,attitudes, sleep, sleepiness and perormance. Scandinavian Journal o Work, Environment, & Health, 24(3),6975.

Macdonald, W., & Bendak, S. (2000). Eects o workload level and 8-versus 12-h workday duration on testbattery perormance. International Journal o Industrial Ergonomics, 26(3), 399416. doi:10.1016/S0169-8141(00)00015-9

Maklan, D. M. (1977). The our-day workweek: Blue-collar adjustment to a nonconventional arrangement o work andleisure. New York, NY: Praeger Publisher Inc.

McClay, J. (2008). Comparison o ten-hou

Documents

The Shift Length Experiment