HomeMy WebLinkAboutIowa City Police Department Traffic Study -- Chris Barnum, Robert Perfetti, Matt Lint; St. Ambrose University
Iowa City Police Department Traffic Study
2005, 2006, 2007, 2010, 2011 & 2012
Chris Barnum
Robert Perfetti
Matt Lint
St. Ambrose University
Final Revision 5-31-14
Please do not cite or reference this article in any publication or media without the first author’s
permission
Date: June 20, 2014 To: Mr. Thomas Markus From: Chief Sam Hargadine Re: St. Ambrose Study on ICPD Traffic Stops Background In response to some community concerns of racial profiling the ICPD started to collect demographic data on traffic stops in July of 1999. The results of the traffic stop data collection were analyzed in a January 2004 report titled “Traffic Stop Practices of the Iowa City Police Department: January 1 – December 31, 2002.” The research team was from the University of Louisville and this report was frequently referred to as the Louisville study. On or about 2006 the Command Staff was approached by Dr. Christopher Barnum, Associate Professor of Sociology and Criminal Justice and Director of Graduate Studies Masters in Criminal Justice at St. Ambrose University. Dr. Barnum was familiar with the Louisville study and became interested in analyzing ICPD traffic stop data utilizing a differing approach. Dr. Barnum initially indicated a desire to study the data for a six month period of time. After an initial review of the six months period of time, both Dr. Barnum and I determined that a more in-depth analysis was needed in order to better understand operational trends in the department. I maintained the working relationship with Dr. Barnum and provided him data for the years 2005, 2006, 2007, 2010, 2011 and 2012. Unfortunately, a transition to a new data management system resulted in conversion problems that prevent us from analyzing 2008 and 2009 data. Throughout this partnership with Dr. Barnum, our officers were not advised of the study due to the potential of changing behavior patterns. In June of 2013 the City Council passed Resolution 12-320 establishing an Ad Hoc Diversity Committee to study City transit and law enforcement operations as they relate to minority populations. Over the course of six months the Ad Hoc Diversity Committee held 22 Committee meetings. Several community discussion forums were held with community members from diverse backgrounds to discuss and receive feedback about transit and law enforcement operations. During this time a renewed conversation on disproportionate contact with minority populations was sparked. The ICPD took the Ad Hoc Diversity Committee process very seriously and is working hard to implement each of the recommendations of the committee. Based on the community conversation generated by the Ad Hoc Diversity Committee, I worked with Dr. Barnum to incorporate more traffic stop data and finalize his analysis. While this study was initially intended for internal and academic purposes, I now believe it is appropriate to have a public discussion on the topic. By participating in the study, I hope it sends a clear message that the ICPD has taken the issue of disproportional minority contact very seriously in the past and will continue to do so in the future. The Study Attached is a study of ICPD traffic stop data from the years 2005, 2006, 2007, 2010, 2011 and 2012. It is an in depth analysis supervised by Dr. Chris Barnum of St. Ambrose University. He was assisted by graduate students Robert Perfetti and Matt Lint.
June 20, 2014 Page 2
It is important to note that the interpretation of the data is very complex and best explained by Dr. Barnum. The methodology used included observational baseline studies. Over 20 trained observers were stationed at various locations within Iowa City to determine the racial makeup of Iowa City’s drivers. These surveys occurred at different times of the day and over multiple years. Dr. Barnum discusses at length the difference in disproportionality from the data in 2005 – 2007 and 2010 – 2012. Dr. Barnum’s report indicates a notable increase in the disproportionate contacts in particular on the South East side between the two time periods. The numbers jump considerably both among a few officers that were assigned to that area and by the department as a whole. As this was occurring the department was responding to a dramatic increase in violent calls that included two riots, multiple shots fired calls and one homicide. To combat the problem ICPD created a new concentrated zone within the existing beat and patrolled that area extensively. In 2009 reported crime was a significant concern for residents in the Pepperwood, Wetherby, and Grant Wood neighborhoods. A juvenile gang calling themselves the Broadway Goons was believed to be responsible for a significant amount of the reported crime. This area is also well known for its high volume of drug trafficking and weapons offenses. Incidents, many of which gained a lot of media attention, began in the early spring and lasted until late summer. Information gained from arrestees was that the gang was actively recruiting and trying to grow in size. Increased assertive foot patrol efforts were started and directed to the area in an attempt to thwart problems. In October 2009 landlord John Versypt was murdered while working in the hallway of his rental property located in the 1900 block of Broadway. Numerous neighborhood meetings were held to address the issue which included several members of the City Council at that time. These issues were a major factor that led to the passing of the Juvenile Curfew Ordinance and the establishment of the South East Substation. There is no doubt that we intensified directed patrols in the south east portion of town during the later time period. We also asked neighboring jurisdictions to drive through that area if they were driving by anyway. The Iowa State Patrol and Johnson County Sheriff’s Office assisted us with creating a sense of continuous law enforcement presence. The officers with the highest likelihood of disproportionate contact in Dr. Barnum's study were there because they were assigned there by supervisory staff to solve a significant crime problem. It is important to note that crime in this area of town has dropped dramatically as a result of our intensified patrols over the last several years. Presently the Pheasant Ridge/Bartelt Road area saw three very violent shots fired incidents one of which has led to significant injury to an innocent person who was hit as a bullet went through the exterior wall inside to where party goers were assembled. The violence seen this spring on the West side and the concern of residents and neighborhood associations is very much like the concerns expressed by the residents of the South East side of town a couple of years ago. The police department remains committed to see that it does not rise to the level that it did a couple of years ago. Our commitment has included similar resource devotion, including extra patrols and overtime details. While we hope to bring stability to this area, we are certainly more cognizant of the tendency for disproportionate minority contact to occur when engaging in hot spot policing. Ideally, we can bring stability without seeing similar jumps in disproportionate contacts. There are several additional items to keep in mind that are not included in the study but are significant at looking at the entire picture. These include: • Complete CALEA® assessments in 2007, 2010 and 2013. The 2013 assessment team was provided with Dr. Barnum’s report. CALEA® is the Commission on Accreditation for Law Enforcement Agencies. The accreditation process requires compliance with
June 20, 2014 Page 3
rigorous standards that meet the best practices for police agencies in the U.S. and Internationally. Proof of compliance is also required and continually monitored over three year periods. • All traffic stops are videotaped and routine and continued random audits by supervisors have not shown any pattern of biased based policing or unprofessional behavior. • Complaints that have come in claiming racial bias have been taken seriously and are fully investigated by supervisory staff. Any inappropriate behavior has resulted in personnel action. Recommendations Going Forward Going forward the department has reviewed Dr. Barnum’s report with the officers and reiterated that biased based policing is illegal, immoral and if discovered can lead to discipline to include termination. Officers receive legal training once per year specifically on race based traffic stops which outlines the legal and civil penalties they could be exposed to if they engage in racial profiling. Officers have also been through diversity training provided by Chad Simmons of Diversity Focus. It is recommended that this relationship with Diversity Focus be ongoing. Supervisory staff members will continue to randomly review the videos of officers throughout the year for indications of unprofessional, biased based or even unsafe habits. Any violations of policy require documentation and at a minimum corrective counseling. All complaints will continue to be fully investigated. It is recommended that Dr. Barnum be hired to analyze 2013 traffic stop data and compare the data with previous years. Future studies should be conducted to ensure that measures put in place are effective and the disproportionate statistics lowers. I would recommend that at least for the next few years we publish this data as part of the City's Annual Equity Report. This will help demonstrate to the community our commitment to this issue and hopefully will show meaningful progress in the years to come. It is imperative that all officers from the newest recruit to the Chief realize that perceptions are viewed differently based on life’s experiences. Police have to remain vigilant to find unprofessional behavior and take seriously all complaints that are brought to light. Lastly, I want to express my full confidence in the officers and staff in the ICPD. I am personally very proud of their dedication, professionalism and high level of performance. The numbers in Dr. Barnum's study do raise concerns, which I am taking with the utmost seriousness. However, I do not for a minute think the numbers indicate ill motivations. I believe the release of the data is an opportunity for the department to grow and outwardly express our commitment to build relationships and protect all persons in the community with the same high standards of professionalism. I look forward to starting this process with the City Council on June 16th and will make myself available to community groups who may wish to further discuss this issue with me in the coming weeks and months.
2
3
Acknowledgments
We wish to thank the members of the Iowa City Police Department for their cooperation
and invaluable assistance with the transfer of data and the other information they
provided. We especially thank Chief of Police Sam Hargadine, Administrative Services
Captain Rick Wyss, Field Operations Captain Jim Steffen and Jim Baker from Information
and technology. This report would not be possible without their tremendous
cooperation and support. We also thank the many St. Ambrose University students who
participated in various aspects of data collection and analyses.
4
Contents
Executive Summary 7
Chapter One: Levels of Disproportionality 9
Introduction 9
Background 9
The Baseline Problem 10
Methodology 11
Data Sources 11
Observational Baseline Information 11
ICPD Demographic Analyses 2005 & 2007 12
ICPD Demographic Analyses 2010 18
ICPD Demographic Analyses 2011 19
ICPD Demographic Analyses 2012 20
Discussion of 2010 – 2012 ICPD Traffic Stop Demographic Data 21
Two Important Generalizations from 2010 – 2012 21
Beat-Two 21
Beat-Two Baseline Recalibration 21
Observation Recalibration 24
Iowa City Public School Data 27
Summary so far 27
Crime rates and Patrol Procedures 29
Suppositions 31
Summary for this Section 31
Chapter 2: Individual Officer Data 32
The Odds Ratio 32
Disparity Index Ratios for Stops 34
2010-2012 Stop Data 38
Limitations of the Data 39
Summary of 2005 – 2012 Analyses so far: 42
5
Summary of 2010 – 2012 49
Chapter 3: Outcomes Data Analyses 50
2005 Outcomes 51
Citations 51
Arrests 51
Searches 52
Summary Table of Outcomes 53
Stop Outcome Summary 54
Final Summary 55
References 57
Appendix A: Logistic Regression Analyses of Stop Outcomes 59
2005 Logistic Regression Analyses 59
Citations 59
Arrests 59
Consent Requests 59
2006 Logistic Regression Analyses 61
Citations 61
Arrests 61
Consent Requests 61
2007 Logistic Regression Analyses 62
Citations 62
Arrests 62
Consent Requests 62
2010 Logistic Regression Analyses 64
Citations 64
Arrests 64
Consent Requests 64
2011 Logistic Regression Analyses 65
Citations 65
Arrests 65
Consent Requests 65
2012 Logistic Regression Analyses 66
Citations 66
6
Arrests 66
Consent Requests 66
Appendix B: Logistic Regression Analyses: Comparing Racial Differences in Traffic Stops 2005-2007
to 2010-2012 67
Appendix C: Detailed Information for Odds Ratio Analyses 68
2005 Odds Ratios 68
Citations 68
Arrests 68
Searches 69
2006 Odds Ratios 70
Citations 70
Arrests 70
Searches 71
2007 Odds Ratios 72
Citations 72
Arrests 72
Searches 72
2010 Odds Ratios 74
Citations 74
Arrests 74
Searches 75
2011 Odds Ratios 76
Citations 76
Arrests 76
Searches 77
2012 Odds Ratios 78
Citations 78
Arrests 78
Searches 79
Appendix D: HMLM 80
Appendix E: Adapted Time Line of Some Important Events Affecting ICPD during Study Period
82
7
Executive Summary
In response to concerns about the potential for racial bias in the Iowa City Police Department’s traffic
stop activity, the PD began systematically collecting data on traffic stops in approximately 2001.
Recently the City retained our research team to analyze their data. The focus of our investigation was an
assessment of racial disproportionality in the ICPD’s traffic stop activity for stops made in 2005, 2006,
2007, 2010, 2011 and 2012—more than 60,000 stops. The investigation evaluated two broad categories
of police data: (i) the demographic information of drivers stopped by the ICPD and (ii) the outcome or
disposition of a stop.
The methodology used to analyze ICPD’s traffic stop demographics employed a driver-population
baseline fashioned from roadside observations, census data and school enrollment information. A
baseline should be thought of as the proportion of minority drivers on the roads in a given location. The
analysis process is straight forward. It centers on identifying differences between the percentages of
various groups stopped by the ICPD and the baseline information. Any difference between baseline
values and police data signifies disproportionality.
The results of baseline analyses suggested that roughly 10% of the drivers on Iowa City roads were
minority members during the study period. Results also show that between 2005 and 2007 levels of
disproportionality in ICPD stop activity were comparatively low. During this time-period, roughly 14% of
the Iowa City Police Department’s traffic stops involved minority drivers.
However, disproportionality increased in 2010 and then remained stable through 2012. Analyses show
that in 2010 the percentage of minority drivers stopped by ICPD officers increased to roughly 19% and
remained near this level in 2011 and 2012. The analyses also show that the minority-driver baseline
remained essentially constant during this time-frame.
A close examination of ICPD patrol practices suggests that in part, the increase in disproportionality
stemmed from an escalation of patrols in a portion of southeast Iowa City. After a review of various
sources it seems likely that the Iowa City Police Department modified patrol procedures following an
increase in violent crime in the city in 2008 and 2009. These modifications included the establishment of
a new patrol beat located in southeast Iowa City in an area with a comparatively high minority resident
concentration. This new patrol area called “beat-2-A” is rather small. It consists of an area no larger
than few blocks and is geographically much smaller than other ICPD beats. However, the minority
baseline in beat 2-A is significantly higher than in other Iowa City beats.
Individual officer analyses indicate that the officers exhibiting the most disproportionality in traffic stops
were frequently assigned to patrol areas located on the southeast side of Iowa City, or were “float”
officers who were tasked with patrolling high crime areas. Both groups of officers tended to stop higher
proportions of minority drivers than did most of their colleagues. Officers assigned to patrol the small 2-
A beat also tended to stop higher proportions of minority drivers than did officers in other areas of
town. However, this result is expected because the proportion of minority members on the roads in this
area is much higher than in other areas of town and much higher than the 10% minority baseline used
for analysis. Consequently, higher proportions of minority stops for beat 2-A officers do not necessarily
indicate disparity or bias.
The examination of stop outcomes assessed disproportionality in citations, arrests, consent searches and
hit-rates or seizures from consent searches. Univariate odds ratio analyses showed consistent
8
patterns—Iowa City officers disproportionately arrested and (consent) searched minority drivers. On
average across all years of the study the odds were about three times greater that minority drivers
would be arrested on a traffic stop in comparison to others. Likewise, the average odds for consent
searches were about three and a half times greater that ICPD officers would request a search from
minority drivers compared to others, this despite hit rates that were actually lower on average for
minority drivers. In other words, in comparison to others, ICPD officers were more likely to make a
seizure from a nonminority driver as the result of a consent search even though officers were more
likely to request a such a search from a minority driver. Findings also suggest that minority drivers and
nonminority drivers were ticketed at equivalent rates. Multivariate logistic regression analyses show
parallel results. The regression odds ratios were similar in size to those from univariate analyses even
after controlling for officer’s race, officer’s gender, officer’s years of service, officer’s duty assignment,
the time of day, type of traffic violation and the driver’s gender. It should be noted that our analyses
show that many officers were inconsistent in entering information about voluntary consent search
requests with about 50% of officers incorrectly inputting data. This level of inconsistency likely
negatively affects the validity of the findings in this area.
Care should be used when evaluating findings for arrest outcomes. Several important control variables
were not available for inclusion in logistic regression models. Consequently, it’s not possible to evaluate
whether disproportionality in arrest rates was a product of other factors like differences in offense types
or offending rates between demographic categories. Likewise, it is important to emphasize that the
number of cases used for analyses of consent search requests and seizures was much smaller than the
number of cases used in analyses of other stop- outcome variables. This small “n” can affect the validity
of the findings and should be taken into consideration when evaluating results.
Recommendations in Brief
(1) ICPD should continue collecting traffic stop data and repeat this study in one year’s time to
assess trends in disproportionality once officers know their behavior is being monitored.
This analysis should include department level measures of disproportionality as well as an
assessment of individual officers’ traffic stop activity across time and location.
(2) The ICPD should closely monitor officer compliance of data collection to reduce the number
of unknown and missing cases.
(3) ICPD should increase officer training in regards to the proper collection and inputting of data
especially for voluntary search requests
(4) ICPD should modify data collection software so that it becomes practical to collect and
analyze the geographical location of individual stops.
(5) ICPD should also modify data collection software so that it becomes practical to track the
reason for an arrest on traffic stops.
9
Chapter One: Levels of Disproportionality
Introduction
In recent years, US citizens have expressed increasing apprehension about racially biased
policing (sometimes called profiling) in traffic stop activity. Although, many definitions of racially biased
policing exist, most researchers agree that the event occurs when the police use race or ethnicity as a
proxy for suspiciousness when deciding whether to stop or sanction potential targets. Of late, some
Iowa City constituents have communicated concerns that the Iowa City Police Department may be
profiling when interacting with minority members. These concerns generally stem from personal
accounts and anecdotal evidence but persist despite a 2001 University of Louisville study that found no
systematic bias in ICPD officers’ conduct (Edwards, Grossi, Vito & West, 2001). To address this issue the
City of Iowa City asked our research team to develop and implement an analysis of Iowa City Police
traffic stop conduct.
In what follows, we use a two-prong approach to assess ICPD traffic stop activity by focusing on
traffic stop demographics and on the outcome of the stop. The ICPD has been collecting data on officers’
traffic stop behavior for over a decade and has accumulated a substantial amount of raw data.
Interpretation of raw data however can be tricky because the nature of police work is characterized by a
complex array of factors that may legitimately account for disproportionality in police-minority contacts.
In fact, these factors can present issues that cloud interpretation of analyses. Our approach in dealing
with this complexity is straightforward. First, to analyze disproportionality in traffic stops we compare
police stop demographic data to a valid and representative baseline. A baseline is best thought of as the
proportion of minority drivers present on the roads. Second, to assess disproportionality in the outcome
of a stop, we use two statistical techniques, a disparity index predicated on odds-ratios and logistic
regression analyses. The outcome of a stop includes things like whether a citation was issued, an arrest
was made or a search conducted etc. We also look closely at individual officer’s conduct by analyzing
how an officer’s traffic stop information may be affected by work schedules, duty assignments and
neighborhood characteristics.
Background1
Racial disparity within the criminal justice system is an enduring feature of the American
experience. For most of this country’s history, minority members, especially African-Americans have
been overrepresented at nearly all stages of the criminal justice process (Drummond, 1999; Kennedy,
1997; for a contrasting opinion, see DiLulio, 1996; Wilbank, 1987). However, studies conducted over the
past 20 years suggest change. These studies show that the overt use of race in police decision-making
behavior is steadily decreasing (Engel et al., 2002; Sherman, 1980). This trend is likely due in part to
community outrage and legislative action but also it’s partly the result of efforts by police supervisors.
Today most research indicates that police discretionary decision making is predicated more on legal and
situational factors than solely on race (Engel et al., 2002; Mastrofski, Worden, & Snipes, 1995; Riksheim
& Chermak, 1993). Nevertheless, race remains one of the most reliable predictors of attitudes toward
1 Much of this section is adapted from Barnum and Perfetti 2010.
10
the police in America today (Weitzer & Tuch, 2005). African Americans are consistently more likely to
hold negative opinions of the police than are other groups (Hurst, Frank, & Browning, 2000).
Why then, at a time when overt racism by the police seems to be decreasing, do minority
members cling to negative perceptions of the police? In part, the answer may lie in a perception of
double disproportionality—an opinion by minority members that the police tend to energetically enforce
the law against them but fail to adequately enforce the law for them. Certain police and law
enforcement practices may have served to heighten this suspicion. The notable forms of drug courier
profiling that began in the last quarter of the 20th century provide an example.
Profiling in various forms has existed for decades in the United States. However, the practice
became particularly salient in the 1980s when some of the first federally subsidized drug courier
profiling methods were developed and used to train local law enforcement officials. An example of this
activity includes tactics developed in a Drug Enforcement Administration sponsored profiling strategy
called Operation Pipeline. This program was originally designed to stem the flow of drugs that were
being transported from Florida to the metropolitan areas of the Northeast along interstate highways.
Officers participating in this training were taught guidelines for identifying the typical characteristics of
drug couriers. One of these guidelines included race. Using race as an identifier lead to unfortunate
consequences including increased levels of fear and resentment among minority members toward
police, and ultimately to lawsuits and litigation.
The source of the recent interest in racially biased policing in traffic stops is generally traced to
two court cases in the 1990s. Defendants in a New Jersey criminal case, the State of New Jersey vs. Soto
(1996), and plaintiffs in a Maryland civil case, Wilkins vs. Maryland State Police (1993), argued that they
were stopped because of their race rather than their driving. This litigation sparked scholarly interest in
this subject and a spate of other court cases across the country. As a result of this legal action, many
police departments began collecting data on police–citizen contacts. Unfortunately, much of this data
remains untouched.
The Baseline Problem
A key reason for this neglect in data analysis is difficulty in identifying and developing the
essential characteristics of the data. The question of how to develop an effective baseline is one of these
problems. A baseline is a standard for determining the percentage of minority drivers in a given police
jurisdiction who are on the roads at a given time. Investigators compare this benchmark to police traffic
stop data to determine whether the driver’s race was a factor in the officer’s decision to make a traffic
stop. Some methods of benchmarking include using census or DOT information to establish baselines.
These techniques are often ineffective for various reasons, including differences between races in the
amount of time spent driving (driving quantity), racial differences in offending rates and thus police
attention (driving quality), and the racial composition of neighboring communities whose citizens may
travel through the population of interest (driver mobility). More recent innovations, however, use mixed
methodological approaches that combine direct observation with census and other data. These
11
methods have generally established more valid baselines than earlier attempts (e.g., Alpert et al., 2007;
Alpert, Smith, & Dunham, 2004; Lamberth, 2006).
Methodology
In what follows we use a combination of methodologies to evaluate officers’ traffic stop
behavior. First, to establish a baseline we use an applied technique that includes traffic observations and
census data. As noted, the baseline should be thought of as the percentage of minority drivers on the
road in a given area of town. In plain terms, the baseline is a standard that can be used to judge the
percentage of minority drivers that should be stopped by the police when no bias is occurring. Second,
we evaluate post stop outcomes using statistical techniques including logistic regression, hierarchical
linear modeling and a disparity index that is predicated on odds ratio analyses. Finally, we assess
individual officers’ conduct using in-depth analyses of stop outcomes specific to a given officer.
Data Sources
This study examines several years of data that has been collected by the ICPD. The data were
selected from years falling within a period ranging from 2005 through 2012. The ICPD experienced
difficulties with their data collection system in 2008 & 2009. Less than a hundred cases are available for
analyses during these years and we consider this information unreliable so they are not included in the
examination. Our strategy is as follows: we will first analyze older data from 2005 - 2007 and use this
information as a comparison standard when evaluating the more recent data from 2010-2012.
Iowa City street officers record information relevant to self-initiated traffic activity as part of
their regular duties. As noted, the Iowa City Police Department has been collecting traffic stop data for
over a decade. Officers are very familiar with the data-collection routine. When stopping a vehicle,
officers contact the dispatch center who then logs the stop. The officers use their in-car computers to
enter pertinent information at the completion of the stop. The data are then transmitted to the station
where they are centrally stored. For each stop, officers enter data regarding the driver of the vehicle,
the reason for the stop, and demographic information. Officers were unaware that their discretionary
traffic stop behavior was being examined by outside researchers. Consequently, it seems unlikely then
that officers modified their level of discretionary traffic stop behavior during the analysis period over
concerns of increased scrutiny.
Observational Baseline Information.
During the study period, over 20 trained observers monitored traffic in Iowa City. These
individuals were stationed at various locations within each of Iowa City’s four police beats. Several
intersections were designated for observations within each beat. These intersections were chosen at
random prior to the beginning of the study, after being screened for traffic volume and visibility (the
selected intersections were chosen from a pool of relatively busy intersections). The choice of
intersections proved to be less complex than initially thought because the city is comparatively uniform
in terms of the racial composition of neighborhoods. In plain terms, there are no large predominately
minority sections or neighborhoods in town.
12
In fact, an initial examination of data from the 2000 U.S. census (and a reanalysis using 2010
census data) for the percentage of African Americans by block group reveals the following. Iowa City is
made up of roughly 40 block groups. Three of these block groups are populated with the highest
concentrations of African Americans. Two of these areas are located on the southeast side of Iowa City
and one is located on the southwest side. However, in Iowa City the police beats are much larger
geographical areas than are census block groups. Consequently, even in these highest minority
concentration areas, the percentage of African Americans residing in areas located on the rest of the
beat does not exceed 12%. In all other areas of the community, the percentage of African Americans
populating any block group was less than 15.0%. A simultaneous examination of all block groups
strongly suggests that with the exception of the three previously mentioned neighborhoods, on the
whole, African American homes are more or less evenly distributed throughout the community.
We utilized three waves of observations. The initial cohort monitored traffic in 2007, followed
by two more groups that surveyed traffic in 2011 and 2013. For each selected intersection, every traffic
observer made between 200 and 400 traffic observations. Depending on traffic volume, this took
approximately 45 minutes. For the initial rounds of observations, the observers generally examined
traffic in at least one intersection on all four beats in a given session. Consequently, each observation
session lasted roughly 3 or 4 hours. The observers surveyed vehicles to discern the race and gender of
the drivers and conducted their inspections periodically all hours of the day—mornings, afternoons,
evenings, and late nights.
The initial round of observations included data from 14 trained observers. All observers used a
systematic sampling strategy that was dependent on traffic volume. For example, when traffic volume
was light, the observers would attempt to assess race and gender for each vehicle passing through the
intersection. However, when volume was heavier, an assessment was made for a set number of cars
(e.g., every third car) passing through the intersection. Generally, traffic volume was much lighter late at
night than during daytime or evening hours. Therefore, the length of observation periods tended to be
longer at night than during daylight hours. Because the observers worked independently of one another,
the correlation coefficient r was used to assess inter-observer reliability. The assessments from each
observer were compared across all beats. Accordingly, each observer’s observations were compared to
all others. For example, the correspondence of assessments of race across all observation points from
Observer A were compared to those same observation points for Observer B. Observer B’s data were
next compared to observer C’s and so on. This was done for all possible contrasts, for a total of 91
comparisons. The average correlation of assessments between observers was extremely high (r ≈ .9).
This strongly suggests that the roadside observers were independently seeing very similar percentages
of minority and nonminority drivers pass through each observation site.
13
Table 1* Census and observer information
Observations Total Percentage 2010 Census %
White 19,391 88.14 82.5
Black 843 3.83 5.8
Asian 854 3.88 6.9
Other 912 4.15 4.8
Grand total 22,000 100.00 100.00
*χ2 = 148.68. p = .999, r =.989
In the analyses that follow whites and Asians are grouped together and are compared to all other groups
called, “minorities.” We group whites and Asians because previous research strongly suggests that
Asians tend to be disproportionately underrepresented in traffic stops (Novak, 2004; Sheldon, 2001;
Barnum and Perfetti 2010). In other words, the police tend to stop too few Asians in comparison to their
baseline values in the population. And as we shall see shortly, this was indeed the case for Iowa City as
well. Grouping Asians with other minority members then would tend to suppress or hide potential
disproportionality in minority traffic stops.
In the initial round, the observers made an assessment of race for 22,000 drivers between June
and December 2007. Table 1 depicts the findings as well as the parallel 2010 census figures. The
correspondence between the percentages witnessed by the roadside observers and the 2010 census
population percentages is striking; 92. 02% of observers’ assessments were of White or Asian drivers,
whereas 7.98% were minority group members. This closely resembles the 2010 census figures, which
report that 89.4% of Iowa City residents were white or Asian, and 10.6% were members of other racial
groups. In addition, observers found that on each of Iowa City’s four police beats, the average
percentage of whites and Asians was at least 90%, and there was no significant difference in
percentages between daytime and nighttime hours. Based on these findings and the high inter-observer
reliability, it seems reasonable to conclude that at least for initial analyses a valid baseline for Iowa City
driver demographics is 90% white and Asian, and 10 % minority. We will have much more to say about
the baseline in the southeast side of town (called beat-two) in subsequent sections of this paper. We will
also soon describe how the baseline is used in a disparity index to examine traffic stop data.
Summary
White & Asian = 90% of the driving population on Iowa City roads
Minority members = 10% of the driving population on Iowa City roads
ICPD Traffic Stop Demographic Analyses 2005 & 2007
We begin the analyses by looking at demographic information of data resulting ICPD self-initiated traffic
stops in 2005 - 2007. Table 2 gives this information for 2005.
14
Table 2 Demographic Traffic Stop Information from 2005
Race Total Stops Percentage
White 8394 84%
Black 892 9%
Hispanic 320 3%
Asian 242 2%
Other 127 1%
Unknown 19 .1%
Native 7 .1%
Grand Total 10001 100%
In 2005, the ICPD initiated 10001 traffic stops.2 Of these, roughly 14% involved minority drivers. This
value is moderately higher than the 10% observational/census baseline, meaning that in 2005 the ICPD
stopped about 4% “too many” minority drivers in comparison to baseline values. Keep in mind that
baseline values are estimates of the percentages of drivers on the roads, so 4% over the baseline is not
necessarily a meaningful amount. In order to assess this level of disproportionality further, we use a
series of steps. First, we analyze stops across police beats. Map 1 gives the locations of the four Iowa
City police beats.
2 Only stops where all information was known about driver and stop location were included in the analyses
15
Map 1 Iowa City Police Beats
Three of the four Iowa City police beats are similar size. Only beat number one which is located in the
downtown area of town is smaller than the others. Table 3 below gives the number and percentage of
traffic stops broken out by the race of the driver and the beat where the stop occurred. In the table we
have included an additional beat–five which is used to represent officers who are not assigned to a
specific beat but instead were allowed to “float” city-wide. This designation includes special
enforcement street crime action team (SCAT) officers as well as k-9 patrols and regular patrol officers
who are not assigned to specific beats or areas of responsibility.
16
Table 3 Driver Demographic Traffic Stop Percentages by Beat in 2005*
Race Beat Number Totals
1 2 3 4 5 Stops Percentage
White 1064 2888 2410 1117 693 8394 84%
Black 117 357 142 165 95 892 9%
Hispanic 42 130 56 54 32 320 3%
Asian 45 73 51 40 26 242 2%
Other 20 50 27 18 10 127 1%
Unknown 4 5 3 1 6 19 0%
Native 1 3 2
1 7 0%
Grand Total 1293 3506 2691 1395 863 10001 100%
Min. Percentage 14% 16% 9% 17% 17% 14%
*Does not include 254 traffic stops made by command staff personnel or data where race is unidentified
The bottom row of the table gives the percentages of minority drivers stopped on each beat. The total
percentage for all stops irrespective of beat is highlighted in red. In 2005, disproportionality in traffic
stops was greatest among beat-five officers who floated city wide and those who worked on beats four
and two (and to a lesser degree on beat one). No disproportionality was found for officers working on
beat three. In general levels of disproportionality are relatively modest and more or less evenly
dispersed across the beats. We now evaluate traffic stop information from 2006 and 2007 in a similar
fashion.
Table 4 Demographic Traffic Stop Information from 2006
Race Total Stops Percent
White 9941 82%
Black 1148 9%
Hispanic 463 4%
Asian 289 2%
Native 5 .1%
Other 230 2%
Unknown 27 .1%
Grand Total 12,103 100%
17
Table 5 Minority Stop Percentages by Beat in 2006*
Race Beat Number Totals
1 2 3 4 5 Stops Percentage
White 2177 3745 1960 1008 906 9796 82%
Black 249 499 129 112 148 1137 10%
Hispanic 100 198 53 42 59 452 4%
Asian 54 87 52 53 38 284 2%
Other 56 71 38 37 24 226 1%
Unknown 7 8 8 4 27 <1%
Native 1 1 3 5 <1%
Grand Total 2643 4609 2241 1252 1182 11927 100%
Min. Percentage 15% 17% 10% 15% 20% 15%
* Does not include 176 traffic stops made by command staff personnel or data where race is unidentified
The information from 2006 is similar to 2005. Disproportionality in stops is generally evenly distributed
across beats, although officers on beat-five have higher levels than others.
Table 6 Demographic Traffic Stop Information from 2007
Race Total Stops Percent
White 7105 83%
Black 734 9%
Hispanic 341 4%
Asian 227 3%
Native 3 .1%
Other 105 1%
Unknown 11 .1%
Grand Total 8526 100%
18
Table 7 Minority Stop Percentages by Beat in 2007*
Race Beat Number Totals
1 2 3 4 5 Stops Percentage
White 930 2776 1213 1089 745 8394 83%
Black 121 251 131 89 104 892 9%
Hispanic 38 148 43 34 61 320 4%
Asian 425 66 47 50 25 242 3%
Other 13 31 14 23 21 127 1%
Unknown 2 1 5 1 2 19 <1%
Native
1
2 7 <1%
Grand Total 1129 3273 1454 1286 960 8102 100%
Min. Percentage 15% 13% 13% 11% 19% 14%
*Does not include 424 traffic stops made by command staff personnel or data where race is unidentified
The overall patterns of the 2005 – 2007 data are similar. In each year the levels of disproportionality are
relatively low and disproportionality is greatest among beat-five officers who floated city wide. 3
Two Generalizations from 2005 - 2007
Overall Levels of disproportionality are low
Beat-five officers exhibit highest levels of disproportionality
We use these generalizations to evaluate 2010, 2011 & 2012 ICPD traffic stop data.
ICPD Traffic Stop Demographic Analyses 2010
Table 8 Demographic Traffic Stop Information from 2010
Race Total Stops Percent
White 9311 77%
Black 1527 13%
Hispanic 593 5%
Asian 372 3%
Native 6 .1%
Other 173 1%
Unknown 66 .1%
Grand Total 12048 100%
3 For 2007 data were only available from January 1st – November 12th 2007.
19
Table 9 Minority Stop Percentages by Beat in 2010
Race Beat Number Totals
1 2 3 4 5 Stops Percentage
White 1677 1729 1758 1869 1588 8621 77%
Black 183 451 323 190 285 1432 13%
Hispanic 72 181 118 73 121 565 5%
Asian 60 73 85 62 59 339 3%
Other 26 19 29 42 54 170 2%
Unknown 6 33 1 2 7 49 <1%
Native 1 2 2 5 <1%
Grand Total 2025 2488 2314 2238 2116 11181 100%
Beat Percentage 14% 26% 20% 14% 22% 19%
*Does not include 867 traffic stops made by command staff personnel or data where race is unidentified
The information in the 2010 traffic stop data departs from results seen in earlier years in two important
ways. First, overall levels of disparity have increased from roughly 14% to 19%. Second,
disproportionality on beat-two has noticeably increased by roughly ten percentage points. These trends
continue in the 2011 and 2012 data.
ICPD Demographic Analyses 2011
Row Labels
Table 10 Demographic Traffic Stop Information from 2011
Race Total Stops Percent
White 10124 76%
Black 1489 11%
Hispanic 627 5%
Asian 419 3%
Native 25 .1%
Other 165 1%
Unknown 485 4%
Grand Total 13334 100%
20
Table 11 Minority Stop Percentages by Beat in 2011*
Race Beat Number Totals
1 2 3 4 5 Stops Percentage
White 2262 2663 1599 1993 254 8771 76%
Black 232 682 222 159 65 1360 12%
Hispanic 122 242 100 62 21 547 5%
Asian 94 121 74 68 14 371 3%
Other 34 46 29 18 5 132 1%
Unknown 40 77 86 98 4 305 3%
Native 3 5 1 11 1 21 <1%
Grand Total 2787 3836 2111 2409 364 11507 100%
Min. Percentage 14% 25% 17% 10% 25% 18%
* Does not include 1827 traffic stops made by command staff personnel or data where race is unidentified
ICPD Demographic Analyses 2012
Table 12 Demographic Traffic Stop Information from 2012
Race Total Stops Percent
White 9122 74%
Black 1385 11%
Hispanic 579 5%
Asian 528 4%
Native 52 .1%
Other 194 2%
Unknown 507 4%
Grand Total 12367 100%
Table 13 Minority Stop Percentages by Beat in 2012
Race Beat Number Totals
1 2 3 4 5 Stops Percentage
White 2273 1863 2422 1843 181 8771 75%
Black 251 427 272 284 60 1360 11%
Hispanic 88 172 144 126 19 547 5%
Asian 143 89 125 118 15 371 4%
Other 44 50 58 27 4 132 2%
Unknown 141 40 78 47 2 305 2%
Native 13 8 10 17 2 21 <1%
Grand Total 2953 2469 3109 2462 283 11412 100%
Min. Percentage 13% 25% 15% 18% 29% 18%
* Does not include 955 traffic stops made by command staff personnel or data where race is unidentified
21
Discussion of 2010 – 2012 ICPD Traffic Stop Demographic Data
The information from the tables for 2010 – 2012 diverges from the demographic data from 2005 - 2007
in at least two important ways. First, the overall percentages of minority drivers stopped by the police
were higher in 2010-2012 than the earlier years. For the more recent data, minority stops comprised
roughly 18% or 19% of all stops made by the ICPD. In 2005 - 2007 this percentage equaled roughly 14%.
Given a 10% minority baseline, this suggests that in 2010 – 2012, overall levels of disproportionality
increased from roughly 4% to about 8%. Logistic regression shows this difference is statistically
significant. For this analysis, logistic regression is a statistical technique that evaluates whether specific
“independent variables” are associated with a driver’s race, given that a stop has occurred. Results show
that irrespective of the area of town where a stop occurred, the reason for the stop or the age and
gender of the driver, the year of the stop was associated with an increase in the odds that the driver was
a minority member (given a stop was made). Specifically, results show that a stop made during the 2010
– 2012 timeframe was associated with a roughly 35% increase in the odds that the driver was a minority
member in comparison to 2005-2007 (z = -12.57 p < .001). See appendix B for tables of results.
Second, the percentage of minority drivers stopped dramatically increased in beat-two and to a lesser
extent among beat-five and beat-three officers in 2010-2012 when compared to the earlier years. In
2005 - 2007 the average percentage of minority drivers stopped on beat-two equaled roughly 15%. It
increased by about 10 percentage points during 2010 -2012. The levels of disproportionality on Beat-five
and beat-three increased by about 6% during the same period. Logistic regression shows these changes
were significant (see appendix B for details). Results also show that minority driver stops on the other
beats did not increase in a similar fashion.
Two Important Generalizations from 2010 – 2012
The percentage of minority drivers stopped significantly increased from 2005 – 2007 levels
The increase in the percentage of minority drivers stopped was chiefly driven by significant
increases in minority driver stops on beat-two, beat-three and among officers not assigned to a
beat (designated as beat-five officers).
Beat-Two
As noted, the largest increase in the percentage of minority drivers stopped occurred on beat-two. This
increase may stem from changes in the baseline population—that is, the percentage of minority
members living and driving in the area, or the increase may stem from changes in police conduct. In
what follows we evaluate the likelihood of each of these potential explanations.
Beat-two Baseline Recalibration
In order to assess minority population change we recalibrated the baseline for beat-two. We began with
an examination of the 2010 U. S. Census data for beat-two. Map 2 below gives the percentage of
African-Americans living in each of the five census tracks located within beat-two. It’s clear from map 2
22
that not all the census tracks match-up with beat-two boundaries. The tracts do however give a good
rough estimate of the percentage of African-Americans living on the beat. Map 2 shows that the
majority of African-Americans who reside in beat-two live on the south end of the beat. Approximately
15.79% of the residents living south of US Highway 6 on beat-two are African-American. On the north
side of this demarcation line roughly 6.10% of residents are African-American. The total percentage of
African-Americans living on beat-two equals approximately 10.62%
Given that most of the African-American residents on beat-two live south of Highway 6 we used US
Census block-group data to examine this area more closely. A block-group is a much smaller area than a
census track. Specifically, a block-group consists of clusters of blocks (usually 20 -30) within a given
census track. Map 3 below gives the census block-groups for the area of beat-2 south of Highway 6.
23
Map 2 The percentage of African-Americans living in beat-two 2010 census tracks
15.08
5.44%
1.77%
7.17%
16.7%
North = 6.10%
South = 15.79%
Total = 10.62%
24
Map 3 The percentage of African-Americans living in selected 2010 beat-two block-groups
Map 3 shows that the location of the majority of African-American who reside in beat-two generally live
in an area that is centered around two block-groups located just south and adjacent to US Highway 6.
These two block groups are intersected by Sycamore street. Note the block-group located in the
extreme southeast corner of the map is partially located outside city limits.4
Observation Recalibration: As mentioned earlier, using census data to establish a baseline can be
problematic because the characteristics of the driving population in a given location may not match the
demographics of the residents who live in the area. Research suggests that observational techniques
4 Note: The percentages in maps 2 and 3 are for African-Americans, not all minority members. The percentages for
all minority members would be higher. We chose to use African-Americans rather than all minority members
because US census data do not completely conform with our definition of a minority. For example, a person who
is classified as “two or more races” under the US census and who Asian an white would not be a minority member
using our classification.
19.81%
27.08%
9.45%
11.10%
25
generally provide superior baselines to census data (Alpert et al., 2007; Alpert, Smith, & Dunham, 2004;
Lamberth, 2006). Consequently, we developed a supplemental baseline for beat-two. Subsequent the
original 2007 observation study we conducted two additional rounds of roadside observations in beat-
two. The first of these occurred in April and May 2011 and focused mainly on the north side of the beat
(1100 observations) and the second, was conducted in June and July 2013 on the south end of the beat
and included oversampling in an area near the Broadway apartments (3200 total observations across
the beat). The second study consisted of a total of five observation sites. Maps 4 and 5 give results of
these analyses.
Map 4 Percentages of minority drivers identified by roadside observers in 2011 & 2013
North = 8.83%
South = 11.55%
Total = 10.19%
10.26%
7.46%
13.93% 9.17%
26
The circled areas in map 4 indicate the observation zones. This map shows that about 10% of all
roadside observations were minority drivers. This value is consistent with the earlier 2007 observation
study. Analyses also show that observers saw more minority drivers on the south side of the beat
(11.55%) than on the north side (8.86%). An additional observation area was conducted within the
block-group exhibiting the highest minority resident percentage (see map 3). This zone is located near
the Broadway area of beat-two. Observations here found roughly 40% of all drivers were minority
members, see map 5 below.
Map 5 Percentages of minority drivers identified by roadside observers in 2011 & 2013 including
oversampling in Broadway area
≈ 40.00%
27
Iowa City Public School Data
The information from the supplemental observation studies and census analyses is very consistent with
the original baseline and census findings from 2007. The 2011-13 observation information suggests that
for beat-two as a whole, about 10 or 11% of the drivers are minority members on average across the
entire beat. The census analyses also suggest that the population demographics in beat-two did not
change in a significant way between the years 2007 – 2012.
To further investigate whether minority resident percentages changed on beat-two during the study
period we analyzed Iowa City Public School Enrollment. Table 14 gives the percentages of African-
American students enrolled at Iowa City public schools for beat-two students.5 The table shows that
with the exception of Grant Wood Elementary, African-American enrollment in beat-two generally
remained steady or decreased between the school years of 2005/06 and 2010/11. These findings are
consistent with information from census and observational analyses. Together, the findings suggest that
it’s unlikely that population demographics on beat-two changed in a dramatic way during the study
period.
Table 14 Percentage of African-America students in Beat-Two schools
Year SE NW NC Wood Twain Lucas Dist. Total
2005-2006 16.13 14.04 16.02 28.61 45.71 17.81 13.38%
2006-2007 14.39 17.26 10.06 31.89 44.21 19.25 14.42%
2007-2008 19.97 17.54 10.89 36.26 50.38 15.42 16.55%
2008-2009 18.72 18.97 9.75 31.96 45.02 14.86 15.96%
2009-2010 19.17 18.97 11.84 38.23 41.77 15.35 16.16%
2010-2011 17.48 17.58 12.00 39.35 38.68 16.55 16.22%
Map 6 below gives the location of Grant Wood School and summarizes the information from the census,
observation and school analyses. Based on the totality of this information it seems reasonable to
conclude that for most areas of beat-two the minority population and percentage of drivers on the road
equaled roughly 10% during the study period. However, an area located in a southern portion of the
beat (and as indicated in map 5) had a much higher percentage of minority residents and drivers. It
seems likely that in this area 20% or more of the driers on the roads were minority members.
Summary so far
It’s unlikely that the baseline percentage of minority drivers on the road increased in a
significant way during the study period in beat-two.
Consequently, increases in disproportionality for ICPD traffic stops on beat-two likely stem from
changes in patrol procedures.
5 The results from NW Junior High should not be given as much weight as other listed schools because the
boundaries for NW Junior High include only a few blocks of beat-two.
28
As will be outlined below, modifications in patrol procedures likely accounts for changes in the
percentage of minority drivers stopped on beat-two during the study period. These changes include
increased use of focused patrols in the higher minority concentration areas of beat-two. A key question
at this point is, why were ICPD patrol procedures modified? We turn to this question in the next section.
Map 6 Summary of census, observation and school analyses
+ 20 % in this area. 10% or
less elsewhere in beat-two
Grant Wood
Elementary
29
Crime rates and Patrol Procedures
As noted, the analyses thus far suggest that it’s unlikely that the observed increase in disproportionality
of minority drivers stopped by the ICPD that occurred during the study period resulted from a significant
rise in the percentage of minority drivers on the roads. Instead other factors seem more likely to be
responsible for the change.
We believe that a modification of ICPD patrol procedures and tactics—especially on beat-two—
generated increased levels of disproportionality. This change in policing occurred between 2007 and
2010 and was concurrent with a spike in violent crime that occurred in 2008 and 2009.
Chart 1 below gives the rates of violent crime per 100,000 residents in Iowa City between 1999 and
2011. It’s clear that the overall trend in the crime rate during this period is downward. However, in 2008
and 2009 the crime rate sharply increased for a brief period and then resumed its downward trend
through the rest of the decade.6
Chart 1 The estimated violent crime rates per 100,000 residents in Iowa City*
*Source City-Data.com, estimates calculated using decennial census population values estimates
Although the increase in crime in 2008-09 was not large or long lasting, research suggests the spike was
accompanied by a disproportional amount of media coverage (Barnum and Perfetti, 2012; 2013; Perfetti
2013).7 Much of this media coverage framed the “crime problem” in Iowa City as predominately a
6 The following crimes were included as violent crimes in the analyses for chart 1: aggravated assault, murder,
rape, robbery.
7 Here are links that provide a sampling of media stories about increases in Iowa City crime on beat-two during
2008-09. See appendix A for a graph of newspaper coverage of crime that occurred during this time.
http://www.kcrg.com/news/local/44973862.html
0
100
200
300
400
500
600
700
19
9
9
20
0
0
20
0
1
20
0
2
20
0
3
20
0
4
20
0
5
20
0
6
20
0
7
20
0
8
20
0
9
20
1
0
20
1
1
20
1
2
Estimated Violent Crime Rates
Estimated Violent
Crime Rates
30
product of illegal activity occurring on the southeast side of town. Additionally, a substantial amount of
anecdotal evidence suggests that the increase in crime and accompanying media coverage affected law
enforcement behavior. For instance, the ICPD instituted a new patrol beat during this time period. This
new beat (called “beat 2-A’”) is formed from a subsection of the original beat-two and is located on the
south side of the beat. The area designated as +20% concentration of minority residents on map 6
roughly corresponds to beat “2-A.” Secondly, the ICPD opened a police substation in 2010 on beat-two
near this same area. The sub stationed opened in part to address crime problems in the area. Further,
the City of Iowa City instituted a curfew ordinance in December 2009 which according to many media
accounts was enacted in part to deal with the violent crime trend in town especially on the southeast
side.8 Consistent with this, violent crime data for neighborhoods located in beat-two do show higher
rates of violent crime for neighborhoods located on the south side of beat-two than the north side (see
tables 15 and 16 below).9
Table 15* Violent crime rate for neighborhoods located in the south side of beat-two
* South-side beat-two estimates are based on a population estimate that equals 8,710
Table 16* Violent crime rate for neighborhoods located in the north side of beat-two
* North-side beat-two estimates are based on a population that equals 12,093
http://www.press-citizen.com/article/20090512/NEWS01/90512001/Man-arrested-rioting-assault-during-large-
fight
http://coralvillecourier.typepad.com/community/2009/05/five-more-charged-for-mothers-day-brawl---violence-
spills-over-to-city-high.html
8 http://www.kwwl.com/story/11602573/iowa-city-council-to-make-decision-on-curfew-ordinance
http://www.kcrg.com/news/local/59413962.html
http://www.radioiowa.com/2009/09/16/first-reading-of-curfew-ordinance-passed-in-iowa-city/
9 Source IC Press Citizen. The following crimes were included as violent crimes in the analyses for tables 15 & 16:
aggravated assault, arson, forcible rape, kidnapping, murder and robbery.
South Neighborhoods Violent Crime 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Wetherby 35 16 16 8 18 27 25 10 15 13
South Pointe 0 1 0 0 0 0 1 0 0 2
Pepperwood 7 0 0 1 0 0 1 0 0 0
Hilltop 0 0 0 0 0 0 0 0 2 2
Grant Wood 23 11 9 13 25 20 26 19 19 22
South 2 Totals 65 28 25 22 43 47 53 29 36 39
Crime rate for year 746.27 321.47 287.03 252.58 493.69 539.61 608.50 332.95 413.32 447.76
North Neighborhoods Violent Crime 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Village Green 1 1 1 4 1 1 2 2 1 1
Lucas farms 15 6 5 10 8 9 8 6 4 6
South East 21 11 5 7 9 7 7 7 12 7
Longfellow 5 3 0 3 1 2 3 2 0 2
Creek Side 5 6 3 5 7 0 7 6 3 4
Friendship 12 4 7 4 8 5 3 6 1 8
Morningside 5 3 0 2 0 1 2 1 2 1
North 2 totals 64 34 21 35 34 25 32 30 23 29
Crime rate for year 529.23 281.15 173.65 289.42 281.15 206.73 264.62 248.08 190.19 239.81
31
Tables 15 and 16 show that the violent crime rate was notably higher for neighborhoods located on the
south side of beat-two than those located on the north side during the study period.
Suppositions
Based on the analyses so far, our supposition is that the ICPD changed its patrol procedures in response
to perceived increased levels of violent crime on beat-two. The analyses show that the south side of the
beat, especially the Wetherby neighborhood had higher violent crime rates than most other areas of the
city, and that the rates of violent crime in this area were higher in 2008 and 2009 than in the other years
included in the analysis. Moreover, it was during this time frame when the changes in police tactics
occurred. These changes took the form of focused patrols—with more officers patrolling in higher
minority concentration areas (beat 2-A) than had been the case prior to 2008. It seems likely that these
police tactics account for some of the increased minority disproportionality found ICPD traffic stops. It
also seems likely that float officers, including SCAT and k-9 officers concentrated their patrols in these
higher minority population neighborhoods. We will investigate these claims more deeply in the next
section.
Summary for this Section
Observation and census analyses show that the baseline of the percentage of minority drivers
on the roads of Iowa city equaled roughly 10% during the study period
In 2005 - 2007 levels of disproportionality in ICPD stops were comparatively low
Levels of disproportionality significantly increased in 2010 and remained stable though 2012
The increase was not likely due to changes in the proportions of minority drivers on the roads of
Iowa City
Disproportionality increased more on beat-two than other beats during the study period.
ICPD modified patrol procedures in 2008-09 in response to perceived increased violent crime in
Iowa City. These modifications include the formation of a new sub-beat located within beat-two.
This sub-beat is located in an area characterized by a higher percentage of minority residents
than other areas of beat-two (or Iowa City).
32
Chapter 2: Individual Officer Data
In this section we breakout individual officer traffic stop information by beat assignment. A disparity
index, odds ratios and graphs are used to identify officers with higher levels of disproportionality than
their coworkers. Comparisons are made across time, across the entire department and across beat
assignment.
The Odds Ratio
In much of what follows we measure disproportionality using one of two estimators that are predicated
on an odds ratio. Given this, it’s valuable to spend some time becoming acquainted with this estimator.
The odds ratio is a measure of effect size and association. It is useful when comparing two distinct
groups. We use a measure called a disparity index when analyzing traffic stops. This measure compares
stops to baseline values. When assessing the outcome of a stop we use a standard odds ratio measure
which compares the odds of something happening in one group to the odds of it happening in another
group.
Before proceeding let’s define a few terms. A baseline is a standard used to judge disproportionality. It
should be thought of as the percentage of minority drivers who are on the road in a given area, and
consequently as the percentage of minority drivers that should be stopped by the police when no bias is
occurring. If the percentage of minority drivers stopped is either higher or lower than the baseline
percentage then disproportionality is said to occur. The term disproportionality does not necessarily
imply bias or discrimination. In what follows we analyze two essential types of police data: (i) traffic stop
data and (ii) outcome data. As the name implies, stop data deals with comparing the number of stops
made by the police to baseline values. Outcome data gives information about the consequence of a
stop. For example, did the driver receive a ticket? Was s/he arrested? How about searched?
The disparity index used to analyze traffic stops measures the difference in ratios between two groups
and their respective baselines. To illustrate let’s focus on a made-up example. Let’s say the baseline for
a given area of town equals 10%, meaning that we can expect that about 10% of the drivers in this area
are minority members. This value represents the proportion of minority drivers who should be stopped
by the police. It follows then, that the baseline value for white drivers in this area equals 90%. To make
this more concrete, let’s say a given officer makes 100 traffic stops in this area. Further, let’s say that
forty-five of the drivers stopped were minority members while fifty-five were not. Given these values,
the disparity index for this officer equals
(.45/.10) ÷ (.55/.90) = 7.36
This number suggests that for our fictional officer, the odds were more than seven (7.36) times greater
that she would stop a minority driver as a non-minority driver given the baseline values. Please note
that higher odds ratio values signal more minority disproportionality and that a score equal to one
suggests no disproportionality.
33
Now let’s look at the outcome of the stop. Here we’ll use the standard odds ratio to evaluate
disproportionality. To illustrate let’s say that our fictional officer wrote a single ticket to 80 of the 100
drivers she stopped. Let’s also say that forty of these tickets went to minority drivers while forty were
issued nonminority drivers. Given this information, computing the odds ratio for stop outcomes is
straightforward.
Citation
No Yes Total
Minority 5 40 45
W & A 15 40 55
Total 20 80 100
The odds ratios for citations equals (40/5) ÷ (40/15) = 3, meaning that the odds were three times greater
that this officer issued a citation to a minority driver as a white driver. This value is meaningfully greater
than one and so suggests significant disproportionality.
In the charts that follow each officer is represented as a circle. Disparity index values are located on the
horizontal axis. As values move from left to right along this axis levels of disproportionality increase. An
effective strategy to use in examining the charts is to identify officers who: (i) are located on the right
side of the horizontal axis, (ii) who stand out from other officers (iii) who have higher disparity index
values than others and (iv) who consistently have comparatively high values across time and on different
beats.
An important warning: Please keep in mind is that the disparity index is based on an observational
baseline and that the baseline is simply an estimate of the proportion of minority drivers on the roads of
Iowa City. The actual percentages of drivers may be significantly different than the baseline.
Consequently, when evaluating an individual officer’s data, it’s important to evaluate the officer over
time and in comparison to colleagues. This practice is much better than simply focusing on the specific
value of a single disparity index score. In other words, in isolation of context—in particular other
officers’ scores, as well as the target officer’s scores across time and place—a single disparity index
score is not a good indicator of bias. Also, please note that the index values become more valid and
reliable as the number of stops made by the officer increases.
Disparity Index Ratios for Stops
34
Chart 2, disparity index ratios for officers working in 2005
The chart above shows the value of the disparity index score for each ICPD officer making at least
fourteen traffic stops in 2005. This table is useful for identifying officers who stopped disproportionate
percentages of minority drivers (given observational zone baseline values). The estimator is calculated
as described above. Each circle represents an individual officer. The values for the index are given on the
horizontal axis. Higher values suggest more disproportionality and a score equaling one suggests no
disproportionality, meaning that the odds of stopping minority and white/Asian drivers are equal. As a
general rule of thumb a score equal to or greater than three should draw your interest and be examined
more closely. Likewise, scores that appear to be dissimilar from others should also be given special
scrutiny. Also it is very important to remember that disparity values that are based on a large number
of stops are more valid and reliable than those based on fewer stops. On the next page we present a
table that gives the values for officers with a disparity index value greater than three. Interpretation is
direct, for example, the odds are the first officer listed in the table is roughly five times (disparity index =
4.91) more likely to stop a minority driver than a W & A driver given the observational baselines. These
same claims apply for all charts that follow.
2005 Descriptives
Mean 1.71
σ 1.03
Skew 1.45
35
Table 17, officers, disparity index values and beats for 2005
Odds Ratio Beat Stops
4.91 5 51
4.37 2 263
3.70 2 508
3.50 2 50
2.86 4 83
2.55 2 181
2.51 2 261
The data for 2005 show relatively modest amounts of disproportionality. In chart 2 the majority of
officers’ disparity index values cluster around 1.00 (mean = 1.7). Recall that a value equaling one
suggests no disproportionality. Additionally, only four officers have disparity odds ratio values larger
than three.
Chart 3, disparity index ratios for officers working in 2006
2006 Descriptives
Mean 2.00
σ 1.44
Skew 1.56
36
Table 18, officers, disparity index values and beats for 2006
Odds Ratio Beat Stops
6.0 5 25
5.5 2 776
4.95 5 31
4.91 1 51
4.6 2 77
3.5 2 223
3.0 4 40
2.8 1 445
2.7 4 144
2.6 2 417
The disparity index values for 2006 are moderately higher than those for 2005 (mean = 2.0). Several
officers disparity index scores are above three. However of the officers with high values, only one is
based on a large number of stops (n > greater than 100) so caution should be used when interpreting
results.
The disparity index information for 2007 is given on the following page.
37
Chart 4, disparity index for officers working in 2007
2007 Descriptives
Mean 1.75
σ 1.07
Skew 1.31
Table 19, officers, disparity index scores and beats for 2007
Odds Ratio Beat Stops
5.17 2 359
3.98 1 186
3.78 3 216
3.77 2 159
3.29 4 56
2.94 1 65
2.83 5 380
The data for 2007 are very similar to those for 2005. The 2007 information shows only modest levels of
disproportionality with most officers’ values clustered around 1.0 (mean = 1.75). Only five officers’
disparity odds ratios were larger than three. Incidentally, no officers in 2007 with odds ratio scores
above three had similarly high scores (disparity index values over three) in 2005 or visa-versa.
38
2010-2012 Stop Data
Chart 5, disparity index for officers working in 2010
2010 Descriptives
Mean 2.56
σ 1.81
Skew 1.52
Table 20, officers, disparity index values and beats for 2010*
Odds Ratio Beat Stops
9.00 5 70
7.41 2 186
6.14 2 69
6.03 2 137
5.75 3 231
5.31 4 264
4.91 2 266
4.53 5 233
4.42 2 367
4.22 2 47
3.78 3 493
3.60 2 35
* Officers highlighted in red were assigned to beat 2A; officers
highlighted in green worked the beat occasionally
39
The data from 2010 show a marked increase in disproportionality compared to data from 2005 – 2007.
Examination of chart 5 shows twelve officers have disparity index values greater than three. The
arithmetic mean of the entire distribution of disparity index values equals 2.56 and is clearly higher than
those from 2005 – 2007. Table 20 above lists the officers whose disparity index values are greater than
three. Nine of these twelve officers were assigned to beat-two or as beat-five float officers.
These data make apparent that much of the increase in disproportionality in 2010 disparity index is
driven by those assigned to beat-two. It is important to note that the officers whose information is
highlighted in red were assigned to beat 2-A fulltime. Information highlighted in green is from officers
who worked beat-2A at least some of the time. Recall that beat 2-A is a special beat that was developed
in 2010 to deal with perceived increases in crime on the southeast side of Iowa City. Six officers listed in
table 17 were assigned to this beat at least some of the time in 2010.
As noted, the census and observational baseline analyses show that the percentage of minority
residents and drivers in the area demarcated by beat 2-A were significantly higher than in other areas of
beat-two. In fact, observational analyses suggest that minority baseline values for beat 2-A were as high
as 40%. Consequently, the 10% minority driver baseline used for other areas of beat-two is not valid
or appropriate for officers making stops solely in beat 2-A. Simply put, using the 10% baseline for an
officer working only in this area would dramatically increase the officer’s odds ratio value and give a
false impression of levels of disproportionality
Limitations of the Data
There are two important limitations with the ICPD traffic stop data: first, is it is not possible to
determine the location of individual traffic stops and second, although we know the beat assignments of
officers, it is not possible to know where on the beat an officer spent most of his/her time.
Consequently, we cannot know the proportion of stops an officer made in a specific location or area of a
beat or know how much time the officer spent in an area looking for a stop. This means that for beat-
two officers it is not possible to know the percentage of time a given officer spent patrolling beat 2-A or
the number of stops the officer made in this area.
The individual officer data for 2011 and 2012 follow. Summary and interpretation will follow the
presentation of results for both years.
40
Chart 6, disparity index for officers working in 2011
Descriptives 2011
Mean 2.31
σ 1.74
Skew 2.03
Table 21, officers, disparity index values and beats for 2011
Odds Ratio Beat Stops
9.00 5 22
7.43 2 418
6.88 2 337
6.08 3 129
5.73 5 18
5.27 2 203
5.20 3 112
4.45 2 248
4.15 5 171
3.38 1 22
3.13 2 190
* Officers highlighted in red were sometimes assigned to beat 2A
41
Chart 7, disparity index for officers working in 2012
Descriptives 2012
Mean 2.32
σ 1.54
Skew 1.99
Table 22, officers, disparity index values and beats for 2012
Odds Ratio Beat Stops
9.33 2 55
5.59 2-A 261
4.76 5 52
4.37 2 266
4.29 3 96
4.22 1 144
4.16 2 313
3.90 5 139
3.82 2 218
3.76 † 112
3.61 2 199
3.50 ‡ 26
3.38 4 461
3.38 2 282
* Officers highlighted in red were sometimes assigned to beat 2A
† investigator, ‡ deidentified
42
The disparity index data for 2010 – 2012 show a clear pattern. The mean disparity index values for each
year are appreciably higher than those from 2005 – 2007 (see Appendix D HMLM section for a statistical
analysis of differences). An examination of individual officers with the highest disparity index values
(greater than three) shows that the majority of these officers were assigned to beat-two or beat-five.
Summary of 2005 – 2012 Analyses so far:
Levels of disproportionality among ICPD officers were comparatively low in 2005 – 2007
Levels of disproportionality significantly increased in 2010 and remained stable in 2011 and
2012 (see appendix D).
In general, officers assigned to beat-two or beat-five demonstrated the highest levels of
disproportionality in 2010 – 2012 traffic stops.
Next, we look more closely at beat-two and beat-five officers’ disparity index values for 2010 – 2012.
43
Chart 8, disparity index for beat 2 officers working in 2010
Descriptives 2010 beat 2
Mean 3.89
σ 1.83
Skew .48
Table 23, officers, disparity index values for beat 2 in 2010*
Odds Ratio Beat Stops
7.41 2-A 186
6.15 2-A 69
6.03 2-A 137
4.91 2-A 266
4.42 2-A 367
4.22 2 47
3.60 2 35
2.76 2 196
2.66 2-A 269
2.33 2 102
2.12 2 291
1.75 2 159
1.29 2 183
* Officers highlighted in red were sometimes assigned to beat 2A
44
Chart 9, disparity index for beat 5 officers working in 2010
Descriptives 2010 beat 5
Mean 3.69
σ 2.50
Skew 1.55
Table 24, officers, disparity index values for beat 5 in 2010
Odds Ratio Beat Stops
9.00 70
4.53 233
3.06 323
2.79 283
2.66 35
2.2 56
2.12 189
1.68 918
Analyses show that in 2010 the disparity index values for officers assigned to work beat 2-A were higher
than other beat-two officers who were not designated to work solely in this area.
45
Chart 10, disparity index for beat 2 officers working in 2011
Descriptives 2011 beat 2
Mean 3.26
σ 1.96
Skew 1.15
Table 25, officers, disparity index values for beat 2 in 2011*
Odds Ratio Beat Stops
7.427948 2-A 418
6.879581 2-A 337
5.273438 2 203
4.445783 2 248
3.12766 2 190
2.616279 2 333
2.595092 2 210
2.273684 2 238
2.076923 2 128
2.076923 2 80
1.979253 2 294
1.774038 2 249
1.738636 2 210
1.431818 2 204
* Officers highlighted in red were sometimes assigned to beat 2A
46
Chart 11, disparity index for beat 5 officers working in 2011
Descriptives 2011 beat 5
Mean 5.04
σ 3.21
Skew .107
Table 26, officers, disparity index values for beat 5 in 2011
Odds ratio Beat Stops
9.0 22
5.73 18
1.30 142
4.15 171
Again the 2011 data make clear that the disparity index values for beat 2-A officers were higher than the
ratios for beat-two officers not designated to work beat 2-A and the values for some beat-five were also
higher than other beat-two officers.
47
Chart 12, disparity index for beat 2 officers working in 2012
Descriptives 2012 beat 2
Mean 3.55
σ 2.29
Skew 1.25
Table 27, officers, disparity index values for beat 2 in 2012
Odds Ratio Beat Stops
9.33 2 55
5.59 2-A 263
4.37 2 270
4.16 2 315
3.82 2 219
3.61 2 202
3.38 2 284
2.56 2 293
1.94 2 126
1.69 2 171
1.10 2 302
1.02 2 149
* Officers highlighted in red were sometimes assigned to beat 2A
48
Chart 13, disparity index for beat 5 officers working in 2012
Descriptives 2012 beat 5
Mean 3.24
σ 1.32
Skew .089
Table 28, officers, disparity index values for beat 5 in 2012
Odds Ratio Beat Stops
4.76 52
3.90 139
2.48 74
1.84 59
49
Summary of 2010 – 2012
ICPD traffic stop disproportionality for 2010-2012 data increased in comparison with 2005 – 2007 levels.
The analyses suggest that much of this increase stemmed from an intensification of focused patrols in an
area of southeast Iowa City characterized by higher minority-resident concentrations than other areas of
town. This location is known as beat 2-A and was implemented as a patrol area in 2010. Since then, a
small number of officers have been assigned to patrol only this beat. Additionally, evidence suggests
that beat-five officers (especially street crime action team or SCAT officers) have frequently focused
their patrols in this area. SCAT officers are tasked with patrolling high crime areas.
Data for individual officers shows that in general, the disparity index values for officers assigned to beat
2-A and many beat-five SCAT officers are higher than the values for officers not designated to work
solely in this area of town. As noted previously, the percent of minority drivers and residents in beat 2-A
is considerably higher than in other areas of town. Consequently, the 10% baseline value used to
calculate individual officer disparity index values is not valid for officers whose patrol areas are
limited to this beat. In fact, using the 10% baseline for officers whose patrol areas are circumscribed by
beat 2-A would significantly inflate their disparity index values.
However, it’s also important to emphasize that several officers not assigned to beat 2-A or SCAT
demonstrated high levels of disproportionality in comparison to their colleagues. Although many of
these officers were assigned to beat-two, some were assigned to beats located in other areas of the city.
It’s also important to mention that not all beat-two or beat-five officers demonstrated high levels of
disproportionality in traffic stops in comparison to colleagues. In fact, the disparity index values for
roughly one half of all beat-two and beat-five officers were lower than 3.0.
Knowing that some beat-two officers exhibited disparity index values while others did not begs an
important question. Why the difference? Two possibilities seem reasonable. First, perhaps beat-two
officers with low values tended to avoid the locations on their beat with high minority concentrations
(like beat 2-A) and simply focused their attention elsewhere. If so, these officers would be making traffic
stops solely in locations where baseline values for minority drivers were lower. Or second, perhaps
although not specifically assigned to beat 2-A, the beat-two officers with higher disparity index values
may have focused their attention on the small area known as beat-2A which is located within their beat
(perhaps because they believed crime was more likely to occur in 2-A). More analysis is needed to
adjudicate between these two possibilities. However, in order to effectively evaluate the likelihood of
each possibility it is necessary to know the precise location of each traffic stop made by officers working
in beat-two. This information is needed to determine if officers with higher disparity index values were
stopping cars more frequently in beat 2-A than other officers. As noted above, this type of analyses is
not possible with these data because exact locations of stops were not provided.
50
Chapter 3 Outcome Data Analyses
In this chapter we examine traffic stops outcomes by looking for disproportionality in citations,
searches, arrests and seizures. The analyses include both univariate odds ratios and multivariate
regression techniques (see appendix A for detailed logistic regression. See Appendix C for detailed
univariate odds ratio analyses).
Outcome analysis provides information about the consequence of a stop. In basic terms, it tells us what
happened to drivers once they were stopped. Our focus is on whether minority drivers were more likely
to receive some sort of sanction (like a ticket) than white/Asian drivers. Assessments include analyses
for citations, arrests, search requests and hit rates—or the rate that a seizure of contraband or evidence
occurred during a consent search.
Unlike the analyses for traffic stops, an investigation of stop outcomes is not dependent on population
baseline characteristics. Outcome assessment simply compares two or more groups using the
proportion of traffic stops as the comparison benchmark. So as an example, let’s say a given officer
stopped ten drivers all for the same offense—running a red light. Here the benchmark is the ten stops.
Let’s also say that five of these drivers were white/Asian and five were minority members. The analysis
simply compares the officer’s outcomes to the stop baseline. Since in this example five drivers from each
demographic violated the law, we’d expect the officer to issue an equal number of tickets to each group.
However, if the officer issued only one ticket to white/Asian drivers but five to minority drivers, this
disparity may suggest bias.
In nearly all instances however, the situation is not as simple as the example above. Officers do not
generally stop drivers for just one type of offense. Instead, officers usually stop drivers for a variety of
reasons, including moving violations, equipment violations, reasonable suspicion and so forth. This adds
a degree of complexity to the analyses. Multivariate statistical techniques like logistic regression and
HMLM are useful in these contexts. These techniques enable researchers to statistically control (or set
aside) potential explanatory variables that are not of interest.
The tables below present summary data for the odds ratio analyses, appendix A provides tables from
logistic regression analyses for outcomes.
Our presentation strategy is as follows. Immediately below we present an example of a complete odds
ratio analyses of data from 2005 to illustrate the process. Following this we present a summary table of
the final results for all years followed by a discussion of the findings. A detailed analysis of odds ratios
for all years can be found in appendix C.
51
2005 Outcomes
Citations
Citations No Yes Total Percent of Stops
Minority 831 530 1361 14%
W & A 4592 4044 8636 86%
Total 5423 4574 9997 100%
* 5 cases missing data
2005 Odds Ratio for citations = .724 (1.38)
Received Citations No Yes
Minority Percent Cited 61% 39%
W & A Percent Cited 53% 47%
Interpretation: in 2005 given that a citation was issued, the odds were 1.35 times higher that a
white/Asian driver would receive a ticket than would a minority driver.
Arrests
Arrests No Yes Total
Minority 1230 131 1361
W & A 8288 348 8636
Grand Total 9518 479 9997
* 5 cases missing data
2005 Odds Ratio for arrests = 2.54
Arrests No Yes
Minority Percent Arrested 90% 10%
W & A Percent Arrested 96% 4%
Interpretation: given that an arrest was made, the odds were 2.5 times greater that a minority driver
would be arrested during a traffic stop than would a W & A driver in 2005.
52
Searches
Consent Request No Yes Total
Minority 1299 61 1360
W & A 8479 157 8636
Grand Total 9778 218 9996
* 6 cases missing data
2005 Odds Ratio for consent search requests = 2.54
Consent Search Requests No Yes
Minority Percent Requested 96% 4%
W & A Percent Requested 98% 2%
Interpretation: given that a search request was made, the odds were 2.5 times greater that an officer
would request to search a car driven by a minority member than a car driven by a W & A driver in 2005.
2005 Odds Ratio for hit rates = .624 (1.60)
Search Hits No Yes Total
Minority Hits 54 7 61
W & A 130 27 157
Grand Total 184 34 218
Minority Hits 89% 11%
W & A Hits 83% 17%
Interpretation: given that an item was seized, compared to W & A drivers, the odds were 2.5 times
greater that an officer would request a search from a minority driver during a traffic stop in 2005;
however in the same year the odds were 1.60 times greater that an officer would seize evidence or
contraband as a result of the search requested of W & A drivers as opposed to minority drivers. In plain
terms minority drivers were subjected to more search requests but when voluntary searches were
conducted, the hit rates were higher when requested from W & A drivers.
A summary table for each year of the study follows. See appendix C for individual tables for the data
analyzed during 2005 -2012.
53
Summary Table of Outcomes
Odds ratios for outcomes by year
Citations Minority Odds
2005 0.72
2006 0.67
2007 0.86
2010 1.18
2011 1.38
2012 1.44
Arrests ----
2005 2.54
2006 2.82
2007 2.61
2010 3.08
2011 3.18
2012 2.55
Search Requests ----
2005 2.54
2006 3.42
2007 5.62
2010 2.75
2011 3.89
2012 2.44
Hit Rates ----
2005 0.62
2006 1.20
2007 0.34
2010 0.44
2011 0.78
2012 0.87
54
Stop Outcome Summary
The purpose of the analyses of stop outcomes was to evaluate disproportionality in citations, arrests,
consent searches and seizures from consent searches. The univariate odds ratio analyses showed
consistent patterns—Iowa City officers disproportionately arrested and asked for consent to search
minority drivers across all years of the study. On average the odds were about 2.80 times greater that
minority drivers would be arrested on a traffic stop in comparison to others. Likewise, the odds were
roughly 3.45 times greater that ICPD officers would request a search from minority drivers compared to
others, this despite “hit rates” that were actually higher for non-minority drivers. Results also suggest
that white/Asian and minority drivers were ticketed at similar rates. Multivariate logistic regression
show similar results. The regression odds ratios are similar in size to those from univariate analyses even
after controlling for officer’s race, officer’s gender, officer’s years of service, officer’s duty assignment,
the time of day, moving violation, equipment violation and the driver’s gender.
It’s important to emphasize that across most years of the study the hit rates that resulted from consent
searches were actually lower for minority drivers than for a white/Asian driver. So although officers
were more likely to ask minority drivers for permission to search, they were more successful in seizing
contraband and evidence from white/Asian drivers.
A final word about searches: We recently surveyed officers to check compliance and accuracy of the
inputting of search request data. The results suggest that ICPD officers were inconsistent in entering
information about search requests. Specifically, roughly 50% of officers correctly input each search
request made. These officers input data each time they made a search request. However, about 50% of
the officers incorrectly entered this information. Instead of entering a request each time an attempt was
made, these officers input a search request only after being granted consent for the search by the
driver. Moreover, it is not possible to know which type of search requests are present for a given search
in this data set. This information should be considered when interpreting search request information.
A final word about arrests: the findings show that across the study period the odds were greater that a
minority driver would be arrested on traffic stop than a white/Asian driver. However, caution should be
used when interpreting this result because important control variables could not be included in logistic
regression models. Most importantly, information was not available for the reason for arrest during a
traffic stop. Consequently, it is unknown whether minority drivers were more likely to be arrested for
low discretion offenses such as bench warrants, driving while under suspension and operating while
intoxicated. Officers have very little discretion when deciding whether to affect an arrest for these types
of offenses. It was not possible to test for differences in offending rates between racial groups for these
types of offenses—which could theoretically account for some of the observed disproportionality—
because the data set does not include this information.
55
Final Summary
This study looked for disproportionality in traffic stops made by the Iowa City Police Department during
2005, 2006, 2007, 2010, 2011 and 2012—more than 60,000 stops. The investigation analyzed two broad
categories of discretionary police conduct: (i) a made traffic stop and (ii) the outcome or disposition of a
stop. The methodology used to analyze ICPD traffic stops employed a driver-population baseline
fashioned from roadside observations, census data and school enrollment information. The
observational portion of the baseline centered on observations from people who surveyed traffic in
Iowa City to determine the race and gender of drivers on the roads. These observers monitored traffic at
various times between 2007 and 2013 and made roughly 25,000 total observations. The methodology
used in assessing ICPD officers’ traffic stop data is straight forward. It centered on identifying
differences between the PD’s traffic stop information and the baseline. Any difference between baseline
values and police data signified disproportionality.
The results of baseline analyses suggested that roughly 10% of the drivers on Iowa City roads were
minority members during the study period. Results also show that between 2005 and 2007 levels of
disproportionality in ICPD stop data were comparatively low. During this time-period, roughly 15% of
the Iowa City Police Department’s traffic stops involved minority drivers.
However, disproportionality significantly increased in 2010 and then remained stable through 2012.
Analyses show that in 2010 the percentage of minority drivers stopped by ICPD officers increased to
roughly 19% and remained near this level in 2011 and 2012. The analyses also show that the minority-
driver baseline remained constant during this time-frame.
A close examination of ICPD patrol practices suggests that the increase in disproportionality stemmed
from an intensification of directed patrols in a portion of southeast Iowa City. After a review of various
sources it seems likely that the Iowa City Police Department modified patrol procedures following an
increase in violent crime in the city in 2008 and 2009. These modifications included the establishment of
a new patrol beat located in southeast Iowa City in an area with a higher minority resident
concentration than other areas of town. This beat—called “2-A” is rather small. It consists of an area no
larger than few blocks and is geographically much smaller than other ICPD beats. However, the minority
baseline in beat 2-A is significantly higher than in other Iowa City beats.
Individual officer analyses indicate that the officers exhibiting the most disproportionality in traffic stops
were frequently assigned to patrol areas located on the southeast side of Iowa City, or were “float”
officers who were tasked with patrolling high crime areas. Both groups of officers tended to stop higher
proportions of minority drivers than did most of their colleagues. Officers assigned to patrol the small 2-
A beat also stopped higher proportions of minority drivers than did other officers. However, for these
officers this result should be discounted because of the higher minority baselines in this area.
Consequently, higher proportions of minority stops for beat 2-A officers do not necessarily indicate
disproportionality or bias.
The examination of stop outcomes assessed disproportionality in citations, arrests, consent searches and
hit-rates or seizures from consent searches. Univariate odds ratio analyses showed consistent
patterns—Iowa City officers disproportionately arrested and (consent) searched minority drivers. On
average across all years of the study the odds were about 2.80 times greater that minority drivers would
be arrested on a traffic stop in comparison to others. Likewise, the odds were roughly 3.45 times greater
that ICPD officers would request a search from minority drivers compared to others, this despite hit
56
rates that were actually on average higher for non-minority drivers. Findings also suggest that minority
drivers and others were ticketed at equivalent rates. Multivariate logistic regression analyses show
parallel results. The regression odds ratios were similar in size to those from univariate analyses even
after controlling for officer’s race, officer’s gender, officer’s years of service, officer’s duty assignment,
the time of day, moving violation, equipment violation and the driver’s gender.
Care should be used when evaluating findings for arrest outcomes. Several important control variables
were not available for inclusion in logistic regression models. Consequently, it’s not possible to evaluate
whether disproportionality in arrest rates was a product of differential offending rates between
demographic categories. Likewise, it is important to emphasize that the number of cases used for
analyses of consent search requests and seizures was much smaller than the number of cases used in
analyses of other stop- outcome variables. This small “n” should be taken into consideration when
evaluating results.
57
REFERENCES
Alpert, Geoffrey P. and Michael R. Smith. 2007. Investigating Racial Profiling By The Miami-
Dade Police Department: A Multimethod Approach. Criminology & Public Policy 6: 25
– 56.
Barnum, Christopher C and Robert Perfetti (under review): “Public Perceptions of Crime and
Racially Biased Policing.” Criminology
Barnum, Christopher C and Robert Perfetti 2010: “Race Sensitive Choices by Police Officers in
Traffic Encounters: Three Conceptual Models.” Police Quarterly 13: 180 – 208.
Barnum, Christopher C and Robert Perfetti. (2013, August) “Racial Migration: Moral Panic in
America‟s Heartland.” To be presented at the Annual Meeting of the Society for the
Study of Social Problems, New York, NY.
DiLulio, John J. Jr. 1996, Spring “My Black Crime Problem, and Ours.” City Journal. Retrieved
July 18, 2008, from http://www.city-journal.org/html/6_2_my_black.html
Drummond, Tammerlin. 1999, Monday, Jun. 14 It‟s Not Just In New Jersey. Time magazine.
Retrieved from Http://www.time.com/time/magazine/article/0,9171,991207,00.html
Edwards, T. D., Grossi, E. L., Vito, G. F., & West, A. D. (2002). Traffic stop practices of the
Louisville Police Department: January 15-December 31, 2001. Louisville, KY:
Department of Justice Administration, University of Louisville
Engel, Robin Shepard., Calnon, Jennifer.M. and Bernard, Thomas.J., 2002. Theory and racial
profiling: Shortcomings and future directions in research. Justice Quarterly 19, pp. 249–
273
Hurst Yolander G, James Frank, and Sandra L. Browning. 2000. The attitudes of juveniles
toward the police: A comparison of black and white youth. Policing an International
Journal of Police Strategies and Management, 23(1), 37-53. Retrieved from
www.csa.com
Kennedy, Randall. 1997. Race, Crime, and the Law (1st ed). New York: Pantheon Books.
Lamberth, John. 2006, Data Collection and Benchmarking of The Bias Policing Project.
Lamberth Consulting. December.
Mastrofski, Stephen D., Robert E. Worden and Jeffrey B. Snipes. 1995. Law Enforcement in a
Time of Community Policing. Criminology, 33(4), 539-563. Retrieved from
www.csa.com
Novak, Kenneth J. 2004, March. Disparity and Racial Profiling in Traffic Enforcement. Police
Quarterly, 7(1), 65-96. Retrieved June 26, 2008, doi:10.1177/1098611102250359
58
Riksheim Eric.C, and Steven M. Chermak. 1993. Causes of police behavior revisited. Journal of
Criminal Justice, 21(4), 353-382. Retrieved from www.csa.com
Ruiz, J 2009, „State v. Soto‟, in HT Greene, & SL Gabbidon (eds), Encyclopedia of race and
crime, SAGE Publications, Inc., Thousand Oaks, CA, pp. 769-71, viewed 14 July 2013,
doi: 10.4135/9781412971928.n314.
Shelden, Randall G. 2001. Controlling the Dangerous Classes: A Critical Introduction to the
History of Criminal Justice. Allyn & Bacon
Sherman, Lawrence W. 1980. The Effects of Police Reform on Political Culture: Three Case
Studies,” pp. 37-57 in David M. Peterson, Ed., The Police: Strategies and Outcomes in
Law Enforcement. Beverly Hills, California: Sage Publications, Inc.
Weitzer, Ronald., and Tuch, Steven A. 2005, September. Determinants of Public Satisfaction
with the Police. Police Quarterly, 8(3), 279-297. Retrieved July 18, 2008,
doi:10.1177/1098611104271106
Wilbanks, William. 1987. The myth of a racist criminal justice system. Monterey, CA:
Brooks/Cole. Retrieved from www.csa.com
Wilkins v. Maryland State Police, Civil Action No. CCB–93–483, Maryland Federal District
Court 1993. “Driving While Black: A Statistician Proves That Prejudice Still Rules the
Road,” Washington Post, August 16, 1999, at C1.
59
Appendix A
Logistic Regression Analyses of Stop Outcomes
2005 Logistic Regression Analyses (minority coded as 0)
Citations B S.E. Exp(B)
Officer’s race* -0.638 0.172 0.529
Officer’s gender* 0.505 0.115 1.657
Years of service* 0.03 0.003 1.031
Assignment* 0.01 0.003 1.011
Daytime stop* 1.605 0.048 4.976
Moving violation 0.025 0.074 1.025
Equip violation* -0.714 0.077 0.49
Male driver 0.071 0.047 1.073
W & A driver -0.028 0.067 0.972
Constant -1.11 0.22
* p < .01
Arrests B S.E. Wald Exp(B)
Officer’s race ** -0.62 0.246 6.359 0.538
Officer’s gender* 0.554 0.281 3.893 1.741
Years of service** -0.02 0.007 7.455 0.98
assignment 0.007 0.006 1.22 1.007
Daytime stop -1.687 0.132 163.483 0.185
Moving violation -0.184 0.155 1.405 0.832
Equip violation** -0.484 0.162 8.969 0.616
Male driver** 0.49 0.109 20.076 1.632
W & A driver** -0.747 0.111 44.956 0.474
Constant -1.644 0.406 16.436
**p < .01, *P < .05
Consent Request B S.E. Wald Exp(B)
Officer’s race 17.241 3.23E+03 0 3.08E+07
Officer’s gender** -0.991 0.211 22.026 0.371
Years of service** -0.117 0.015 62.443 0.889
Assignment 0.012 0.008 2.02 1.012
Daytime stop** -0.792 0.159 24.939 0.453
Moving violation* -0.494 0.221 4.993 0.61
Equip violation 0.138 0.22 0.395 1.148
Male driver** 0.531 0.16 10.943 1.7
W & A driver** -0.582 0.158 13.582 0.559
Constant -18.613 3.23E+03
**p < .01, *P < .05
60
Interpretation: the results of logistic regression are consistent with odds ratio analyses. Even after
controlling for several important alternative explanations, results show that in comparison to W & A
drivers, the odds were essentially equal that minority drivers would receive a ticket. However, the odds
were greater minority drivers would l be arrested (2.11) and have an officer ask to search the vehicle
(1.78).
61
2006 Logistic Regression Analyses (minority coded as 0)
Citations B S.E. Exp(B)
Assignment*** -.1299 .0165 .878
Daytime stop*** 1.348 .0149 3.851
Moving violation* .128 .063 1.137
Equip violation*** -.555 .0633 .574
Male driver -.005 .0408 .994
W & A driver*** .221 .0555 1.246
Constant -.6634 .0954
* p < .05. *** p < .001
Arrests B S.E. Exp(B)
Assignment -.031 .0301 .964
Daytime stop*** -1.258 .0996 .248
Moving violation*** -1.308 .1291 .270
Equip violation*** -1.04 .1306 .352
Male driver*** .3724 .0971 1.451
W & A driver*** -08583 .0971 .4238
Constant -.8981 .1741
***p < .001
Consent Request B S.E. Exp(B)
Assignment** -.121 .0431 .885
Daytime stop*** -.590 .1093 .554
Moving violation* .374 .167 1.454
Equip violation*** .838 .167 2.312
Male driver*** .953 .137 2.595
W & A driver*** -1.092 .111 .335
Constant -3.28 .249
***p < .001, **P < .01, *p < .05
Interpretation: the results of logistic regression are consistent with odds ratio analyses. Even after
controlling for several important alternative explanations, results show that in comparison to W & A
drivers, the odds were slightly greater that a white/Asian driver would receive a ticket (1.24) but the
odds were greater that a minority driver would be arrested (2.33) and have an officer ask to search the
vehicle (2.98).
62
2007 Logistic Regression (minority code as 1)
Citations B S.E. Exp(B)
Officer’s race -0.348 0.225 0.706
Officer’s gender** 0.704 0.145 2.021
Years of service** 0.062 0.004 1.064
Assignment* -0.028 0.012 0.972
Daytime stop** 1.127 0.069 3.087
Moving violation** 0.616 0.107 1.851
Equip violation 0.095 0.108 1.1
Male driver -0.014 0.063 0.986
W & A driver** 0.262 0.091 1.3
Constant -2.744 0.199
* p < .5, ** p < .01
Arrest B S.E. Exp(B)
Officer’s race -0.634 0.732 0.53
Officer’s gender -0.207 0.26 0.813
Years of service** -0.049 0.01 0.952
Assignment -0.047 0.033 0.954
Daytime stop** -1.069 0.155 0.343
Moving violation** -0.712 0.224 0.491
Equip violation** -0.999 0.232 0.368
Male driver** 0.853 0.162 2.346
W & A driver** 0.747 0.153 2.111
Constant -1.625 0.411
* p < .5, ** p < .01
Search Request B S.E. Exp(B)
Officer’s race 0.33 0.632 1.391
Officer’s gender** -1.07 0.358 0.343
Years of service** 0.035 0.016 1.036
Assignment* 0.031 0.013 1.032
Daytime stop** -1.7 0.287 0.183
Moving violation -0.203 0.368 0.816
Equip violation** -0.177 0.373 0.838
Male driver** 1.531 0.356 4.623
W & A driver** 1.501 0.228 4.484
Constant -4.374 0.584
* p < .5, ** p < .01
Interpretation: the results of logistic regression are consistent with odds ratio analyses. Even after
controlling for several important alternative explanations, results show that in comparison to W & A
drivers, the odds were roughly equal minority driver would receive a ticket (1.3) but the odds were
63
greater that a minority driver would be arrested (2.11) and have an officer ask to search the vehicle
(4.84).
64
2010 Logistic Regression (minority coded as 0)
Citations B S.E. Exp(B)
Officer’s race 0.047 0.118 1.048
Officer’s gender -0.066 0.138 0.936
Years of service** 0.033 0.003 1.033
Assignment** -0.01 0.001 0.99
Daytime stop** -0.867 0.054 0.42
Moving violation** 0.329 0.087 1.39
Equip violation** -0.332 0.087 0.718
Male driver 0.047 0.048 1.049
W & A driver** -0.423 0.059 0.655
Constant -0.777 0.201
* p < .05; ** p < .01
Arrests B S.E. Exp(B)
Officer’s race** -0.63 0.198 0.532
Officer’s gender 0.185 0.306 1.203
Years of service** -0.021 0.008 0.979
Assignment 0 0.003 1
Daytime stop** 0.657 0.118 1.93
Moving violation** -1.54 0.148 0.214
Equip violation** -1.72 0.149 0.179
Male driver* 0.276 0.113 1.318
W & A driver** -0.951 0.109 0.386
Constant -1.025 0.393
* p < .05; ** p < .01
Search Requests B S.E. Exp(B)
Officer’s race* 1.775 0.714 5.902
Officer’s gender -0.104 0.319 0.901
Years of service* -0.021 0.01 0.979
Assignment -0.001 0.003 0.999
Daytime stop** 0.817 0.15 2.264
Moving violation** -0.796 0.217 0.451
Equip violation** -0.636 0.21 0.53
Male driver** 0.721 0.154 2.057
W & A driver** -0.856 0.135 0.425
Constant -4.856 0.818
* p < .05; ** p < .01
Interpretation: the results of logistic regression are consistent with odds ratio analyses. Even after
controlling for several important alternative explanations, results show that in comparison to W & A
drivers, the odds were greater that minority drivers would receive a ticket (1.52) would be arrested (2.6)
and would have an officer ask to search the vehicle (2.354).
65
2011 Logistic regression (minority coded as 0)
Citation B S.E. Exp(B)
Officer’s race 0.154 0.089 1.166
Officer’s gender** 0.677 0.168 1.967
Years of service** 0.031 0.003 1.031
Assignment ** -0.016 0.001 0.984
Daytime stop** 0.454 0.051 1.574
Moving violation** 0.209 0.08 1.232
Equip violation** -0.782 0.082 0.458
Male driver** -0.003 0 0.997
W & A driver** -0.583 0.056 0.558
Constant -1.597 0.21
** p < .01
Arrests B S.E. Exp(B)
Officer’s race -0.318 0.19 0.728
Officer’s gender 0.266 0.346 1.305
Years of service 0.012 0.007 1.012
Assignment -0.001 0.002 0.999
Daytime stop** -1.035 0.115 0.355
Moving violation** -1.149 0.14 0.317
Equip violation** -1.099 0.139 0.333
Male driver* 0.003 0.001 1.003
W & A driver** -0.928 0.1 0.395
Constant -1.334 0.422
** p < .01
Search requests B S.E. Exp(B)
Officer’s race* 0.76 0.326 2.139
Officer’s gender 0.049 0.346 1.05
Years of service -0.008 0.008 0.992
Assignment -0.003 0.003 0.997
Daytime stop** -0.646 0.127 0.524
Moving violation -0.012 0.179 0.988
Equip violation 0.016 0.177 1.016
Male driver 0.001 0.001 1.001
W & A driver** -1.284 0.112 0.277
Constant -3.134 0.514
* p < .05; ** p < .01
Interpretation: the results of logistic regression are consistent with odds ratio analyses. Even after
controlling for several important alternative explanations, results show that in comparison to W & A
drivers, the odds were greater that minority drivers would receive a ticket (1.79) be arrested (2.53) and
have an officer ask to search the vehicle (3.61).
66
2012 Logistic Regression (minority coded as 0)
Citations B S.E. Exp(B)
Officer’s race 0.083 0.108 1.087
Officer’s gender** 0.589 0.121 1.803
Years of service 0.005 0.003 1.005
Assignment** -0.01 0.002 0.99
Daytime stop** 0.649 0.055 1.914
Moving violation** -0.371 0.087 0.69
Equip violation** 0.363 0.088 1.437
Male driver** 0.181 0.048 1.199
W & A driver** -0.49 0.056 0.613
Constant -2.104 0.197
** p < .01
Arrest B S.E. Exp(B)
Officer’s race** -0.506 0.183 0.603
Officer’s gender 0.443 0.329 1.557
Years of service 0.003 0.009 1.003
Assignment 0.002 0.003 1.002
Daytime stop** -1.318 0.137 0.268
Moving violation** -1.161 0.14 0.313
Equip violation** -1.367 0.146 0.255
Male driver** 0.425 0.104 1.529
W & A driver** -0.764 0.1 0.466
Constant -1.286 0.411
** p < .01
Search Request B S.E. Exp(B)
Officer’s race** 1.564 0.583 4.776
Officer’s gender 0.413 0.39 1.511
Years of service 0.014 0.011 1.014
Assignment -0.01 0.007 0.99
Daytime stop** -1.234 0.18 0.291
Moving violation -0.345 0.21 0.708
Equip violation -0.103 0.21 0.902
Male driver** 0.661 0.142 1.937
W & A driver ** -0.754 0.128 0.471
Constant -4.998 0.744
** p < .01
Interpretation: the results of logistic regression are consistent with odds ratio analyses. Even after
controlling for several important alternative explanations, results show that in comparison to W & A
drivers, the odds were greater that minority drivers would receive a ticket (1.63) be arrested (2.15) and
have an officer ask to search the vehicle (2.12).
67
Appendix B
Logistic Regression Analyses: Comparing Racial Differences in Traffic Stops 2005-2007 to 2010-2012
Logistic Regression for all Beats Comparing 2005-2007 to 2010-2012
Driver’s Race=DV B S.E.
Year of Study*** -.3059 .024
Male Driver*** -.195 0..24
Assignment .003 .0009
Moving violation*** .523 .0398
Equip violation .098 .0400
Male driver*** 0.071 0.047
Daytime Stop*** .277 .0243
Constant 1.413
n 53100
*** p < .001 (DV=minority driver coded as 0) Note: year of study is an indicator variable
with 2010-2012 coded as 1
Logistic Regression for individual beats comparing 2005-2007 to 2010-2012
Driver's Race=DV B S.E. Exp(B) n
Year of Study Beat-1 0.0841 0.0576 1.087 9821
Year of Study Beat-2*** -0.5121 0.0258 0.599 16314
Year of Study Beat-3*** -0.5791 0.0564 0.560 11592
Year of Study Beat-4 -0.1371 0.0627 0.871 8212
Year of Study Beat-5*** -0.3569 0.0893 0.693 4876
*** p < .001 (DV=minority driver coded as 0) Note: year of study is an indicator variable with 2010-2012 coded as 1. The control
variables used are the same as the analysis above but are not listed in this table
68
Appendix C
Detailed Information for Odds Ratio Analyses
2005 Citations
Citations No Yes Total Percent of Stops
Minority 831 530 1361 14%
W & A 4592 4044 8636 86%
Total 5423 4574 9997 100%
* 5 cases missing data
2005 Odds Ratio for citations = .724 (1.38)
Received Citations No Yes
Minority Percent Cited 61% 39%
W & A Percent Cited 53% 47%
Interpretation: in 2005 given that a citation was issued, the odds were 1.35 times higher that a
white/Asian driver would receive a ticket than would a minority driver.
Arrests
Arrests No Yes Total
Minority 1230 131 1361
W & A 8288 348 8636
Grand Total 9518 479 9997
* 5 cases missing data
2005 Odds Ratio for arrests = 2.54
Arrests No Yes
Minority Percent Arrested 90% 10%
W & A Percent Arrested 96% 4%
Interpretation: given that an arrest was made, the odds were 2.5 times greater that a minority driver
would be arrested during a traffic stop than would a W & A driver in 2005.
69
Searches
Consent Request No Yes Total
Minority 1299 61 1360
W & A 8479 157 8636
Grand Total 9778 218 9996
* 6 cases missing data
2005 Odds Ratio for consent search requests = 2.54
Consent Search Requests No Yes
Minority Percent Requested 96% 4%
W & A Percent Requested 98% 2%
Interpretation: given that a search was requested, the odds were 2.5 times greater that an officer would
request to search a car driven by a minority member than a car driven by a W & A driver in 2005.
2005 Odds Ratio for hit rates = .624 (1.60)
Search Hits No Yes Total
Minority Hits 54 7 61
W & A 130 27 157
Grand Total 184 34 218
Minority Hits 89% 11%
W & A Hits 83% 17%
Interpretation: compared to W & A drivers, the odds were 2.5 times greater that an officer would
request a search from a minority driver during a traffic stop in 2005; however in the same year the odds
were 1.60 times greater that an officer would find evidence or contraband as a result of the search
requested of W & A drivers as opposed to minority drivers. In plain terms minority drivers were
subjected to more search requests but when voluntary searches were conducted, the hit rates were
higher when requested from W & A drivers.
70
2006 Outcomes
Citations
Citations No Yes Total Percent of Stops
Minority 1137 718 1855 15%
W & A 5302 4928 10230 85%
Total 6439 5646 12085 100%
2006 Odds Ratio for citations = .67 (1.49)
Received Citations No Yes
Minority Percent Cited 62% 38%
W & A Percent Cited 52% 48%
Interpretation: in 2006 given that a citation was issued, the odds were 1.49 times higher that a
white/Asian driver would receive a ticket than would a minority driver.
Arrests
Arrests No Yes Total
Minority 1675 180 1855
W & A 9855 375 10230
Grand Total 11530 555 12085
2006 Odds Ratio for arrests = 2.82
Arrests No Yes
Minority Percent Arrested 90% 10%
W & A Percent Arrested 96% 4%
Interpretation: given that an arrest was made, the odds were 2.8 times greater that a minority driver
would be arrested during a traffic stop than would a W & A driver in 2006.
71
Searches
Consent Request No Yes Total
Minority 1714 141 1855
W & A 9990 240 10230
Grand Total 11530 381 12085
* 6 cases missing data
2006 Odds Ratio for consent search requests = 3.42
Consent Search Requests No Yes
Minority Percent Requested 92% 8%
W & A Percent Requested 98% 2%
Interpretation: given that a search request was made, the odds were 3.4 times greater that an officer
would request to search a car driven by a minority member than a car driven by a W & A driver in 2006.
2006 Odds Ratio for hit rates = 1.20
Search Hits No Yes Total
Minority Hits 121 20 141
W & A 211 29 240
Grand Total 332 49 381
Minority Hits 86% 14%
W & A Hits 87% 13%
Interpretation: compared to W & A drivers, the odds were 3.4 times greater that an officer would
request a search from a minority driver during a traffic stop in 2006 and in the same year the odds were
1.20 times greater that an officer would find evidence or contraband as a result of the search requested
of W & A drivers as opposed to minority drivers. In plain terms minority drivers were subjected to more
search requests and when voluntary searches were conducted, the hit rates were higher when
requested from minority.
72
2007 Outcomes
Citations
Citations No Yes Total Percent of Stops
Minority 690 493 1183 13.8%
W & A 3949 3383 7332 86.2%
Total 4639 3876 8515 100%
2007 Odds Ratio for citations = .979 (1.02)
Received Citations No Yes
Minority Percent Cited 58% 42%
W & A Percent Cited 54% 46%
Interpretation: given that a citation was issued, the odds were 1.02 times greater that W & A drivers
would receive a citation during a traffic stop than would a minority driver in 2007.
Arrests
Arrests No Yes Total
Minority 1085 98 1183
W & A 7073 259 7332
Grand Total 8158 357 8515
2007 Odds Ratio for arrests = 2.47
Arrests No Yes
Minority Percent Arrested 92% 8%
W & A Percent Arrested 96% 4%
Interpretation: given that an arrest was made, the odds were 2.47 times greater that a minority driver
would be arrested during a traffic stop than would a W & A driver in 2007.
73
Searches
Consent Request No Yes Total
Minority 1120 63 1183
W & A 7249 83 7332
Grand Total 8369 146 8515
2007 Odds Ratio for consent search requests = 5.67
Consent Search Requests No Yes
Minority Percent Requested 95% 5%
W & A Percent Requested 99% 1%
Interpretation: given that a search request was made, the odds were 5.67 times greater that an officer
would request to search a car driven by a minority member than a car driven by a W & A driver in 2007.
2007 Odds Ratio for hit rates = .735 (1.37)
Search Hits No Yes Total
Minority Hits 53 10 63
W & A 66 17 83
Grand Total 119 270 146
Minority Hits 84% 16%
W & A Hits 80% 20%
Interpretation: compared to W & A drivers, the odds were 5.67 times greater that an officer would
request a search from a minority driver during a traffic stop in 2007; however in the same year the odds
were 1.37 times greater that an officer would find evidence or contraband as a result of the search
requested of W & A drivers as opposed to minority drivers. In plain terms minority drivers were
subjected to more search requests but when voluntary searches were conducted, the hit rates were
higher when requested from W & A drivers.
74
2010 Outcomes
Citations
Citations No Yes Total Percent of Stops
Minority 1680 619 2299 19.2%
W & A 7395 2288 9683 80.8%
Total 9075 2907 11982 100%
2010 Odds Ratio for citations = 1.19
Received Citations No Yes
Minority Percent Cited 73% 27%
W & A Percent Cited 76% 24%
Interpretation: given that a citation was issued, the odds were 1.19 times greater that minority drivers
would receive a citation during a traffic stop than will a W & A driver in 2010.
Arrests
Arrests No Yes Total
Minority 2124 175 2299
W & A 9435 248 9683
Grand Total 11559 423 11982
2010 Odds Ratio for arrests = 3.13
Arrests No Yes
Minority Percent Arrested 92% 8%
W & A Percent Arrested 97% 3%
Interpretation: given that an arrest was made, the odds were 3.13 times greater that a minority driver
would be arrested during a traffic stop than a W & A driver in 2010.
75
Searches
Consent Request No Yes Total
Minority 2190 109 2299
W & A 9509 174 9683
Grand Total 11699 283 11982
2010 Odds Ratio for consent search requests = 2.72
Consent Search Requests No Yes
Minority Percent Requested 95% 5%
W & A Percent Requested 98% 2%
Interpretation: given that a search request was made, the odds were 2.72 times greater that an officer
would request to search a car driven by a minority member than a car driven by a W & A driver in 2010.
Search Hits (Requests) No Yes Total
Minority Hits 96 13 109
W & A 137 37 174
Grand Total 233 50 283
Minority Hits 88% 12%
W & A Hits 79% 21%
2010 Odds Ratio for hit rates = .50 (1.99)
Interpretation: compared to W & A drivers, the odds were 2.72 times greater that an officer would
request a search from a minority driver during a traffic stop in 2010; however in the same year the odds
were 1.99 times greater that an officer would find evidence or contraband as a result of the search of W
& A drivers as opposed to minority drivers. In plain terms minority drivers were subjected to more
search requests but when voluntary searches were conducted, the hit rates were higher when
requested from W & A drivers.
76
2011 Outcomes
Citations
Citations No Yes Total Percent of Stops
Minority 1627 679 2306 18.0%
W & A 8093 2450 10543 82.0%
Total 9720 3129 12849 100%
*485 cases missing data
2011 Odds Ratio for citations = 1.38
Received Citations No Yes
Minority Percent Cited 71% 29%
W & A Percent Cited 77% 23%
Interpretation: given that a citation was issued, the odds were 1.38 times greater that minority drivers
would receive a citation during a traffic stop than would a W & A driver in 2011.
Arrests
Arrests No Yes Total
Minority 2111 195 2306
W & A 10245 298 10543
Grand Total 12356 493 12849
* 485 cases missing data
2011 Odds Ratio for arrests = 3.18
Arrests No Yes
Minority Percent Arrested 92% 8%
W & A Percent Arrested 97% 3%
Interpretation: given that an arrest was made, the odds were 3.18 times greater that a minority driver
would be arrested during a traffic stop than a W & A driver in 2011
77
Searches
Consent Request No Yes Total
Minority 2144 162 2306
W & A 10342 201 10543
Grand Total 12486 363 12849
*485 cases missing data
2011 Odds Ratio for consent search requests = 3.89
Consent Search Requests No Yes
Minority Percent Requested 93% 7%
W & A Percent Requested 98% 2%
Interpretation: given that a search request was made, the odds were 3.89 times greater that an officer
would request to search a car driven by a minority member than a car driven by a W & A driver in 2011.
Search Hits (Requests) No Yes Total
Minority Hits 109 53 162
W & A 124 77 201
Grand Total 233 130 363
Minority Hits 67% 33%
W & A Hits 62% 38%
2011 Odds Ratio for hit rates = .78 (1.27)
Interpretation: compared to W & A drivers, the odds were 2.89 times greater that an officer would
request a search from a minority driver during a traffic stop in 2011; however in the same year the odds
were 1.27 times greater that an officer would find evidence or contraband as a result of the search
requests of W & A drivers as opposed to minority drivers. In plain terms minority drivers were subjected
to more search requests but when voluntary searches were conducted, the hit rates were higher when
requested from W & A drivers.
78
2012 Outcomes
Citations
Citations No Yes Total Percent of Stops
Minority 1681 597 2278 19.0%
W & A 7736 1914 9650 81.0%
Total 9417 2511 11928 100%
*439 cases missing data
2012 Odds Ratio for citations = 1.44
Received Citations No Yes
Minority Percent Cited 74% 26%
W & A Percent Cited 80% 20%
Interpretation: given that a citation was issued, the odds were 1.44 times greater that minority drivers
would receive a citation during a traffic stop than will a W & A driver in 2012.
Arrests
Arrests No Yes Total
Minority 2097 181 2278
W & A 9334 316 9650
Grand Total 11431 497 11928
* 439 cases missing data
2012 Odds Ratio for arrests = 2.55
Arrests No Yes
Minority Percent Arrested 92% 8%
W & A Percent Arrested 97% 3%
Interpretation: given that an arrest was made, the odds were 2.55 times greater that a minority driver
would be arrested during a traffic stop than a W & A driver in 2012.
79
Searches
Consent Request No Yes Total
Minority 2176 102 2278
W & A 9468 182 9650
Grand Total 11644 284 11928
*439 cases missing data
2012 Odds Ratio for consent search requests = 2.44
Consent Search Requests No Yes
Minority Percent Requested 96% 4%
W & A Percent Requested 98% 2%
Interpretation: given that a search request was made, the odds were 2.44 times greater that an officer
would request to search a car driven by a minority member than a car driven by a W & A driver in 2012.
Search Hits (Requests) No Yes Total
Minority Hits 35 67 102
W & A 57 125 182
Grand Total 92 192 284
Minority Hits 34% 66%
W & A Hits 31% 69%
2012 Odds Ratio for hit rates = .87 (1.15)
Interpretation: compared to W & A drivers, the odds were 2.44 times greater that an officer would
request a search from a minority driver during a traffic stop in 2012; however in the same year the odds
were 1.15 times greater that an officer would find evidence or contraband as a result of the search
requested of W & A drivers as opposed to minority drivers. In plain terms minority drivers were
subjected to more search requests but when voluntary searches were conducted, the hit rates were
higher when requested from W & A drivers.
80
APPENDIX D
HMLM
We use hierarchical multivariate linear modeling (HMLM) to investigate the effects of time on levels of
disproportionality in individual officers’ disparity indexes. Statistical hierarchies are common in data and
usually consist of units grouped at different levels. For the present analysis, this structure came about
because the same individuals were measured on more than one occasion during the study period.
Consequently, we treat multiple observations on each officer as nested within the officer.
When measurements are repeated on the same participants the measurement repetitions (called
occasions) are level-1 units and the participants are level-2 units. We model a linear relationship
between the year of the study and a given officer’s disparity index. This simple model is appropriate for
data like ours because there are only a few observations per officer and the time period between
observations is short (Bryk & Raudenbush, 1992). The model takes the form of a linear growth model,
where the year of the study is treated as an age metric. This variable is grand-mean-centered so it
describes the difference in years between a given year of the study period and the midpoint of the study
(2009). Both the intercept and the time parameter vary at level-2 as a function of characteristics of the
officer.
Equation 1 specifies the level-1 model for this investigation.
Yij = π0j + π1j(time)ij + π2j(beat) + rij (1)
This equation models a linear relationship between time elapsed during the study period, the beat or
area of the town and a given officer’s disparity index. In equation 1, the symbol Yij represents the value
of officer j’s disparity index at time i, π0j is the average level of disparity across occurrences for a given
officer, it represents the officer’s effect on the disparity index, π1j is the change in levels of disparity
across occurrences that is due to time period for a given officer, π2j is the change in levels of disparity
across occurrences that are due to the area of town an officer is working, this is a time varying covariate
and rij is the unique effect of a given occurrence for a particular officer. We assume that the errors are
independent and normally distributed with a common variance. Equations 2, 3 and 4 model how the
stage of an officer’s career mediates the effect of time on disparity. The seniority variable is defined as
the maximum number of years an officer has worked on the street at the end of the study period.10
π0j = 00 + 01(years of service)j + u0j (2)
At level-2 the average level of disparity across occurrences of the study for an officer (π0j) is a function of
the average level of disparity across all officers ( 00); plus the amount of disparity that is a function of
the officers’ years of service, ( 01); and a unique individual component of disparity that is due to a given
officer (u0i) this is formulated as the difference between the officer’s mean change in disparity and 00.
10 It was unreasonable to include other officer level characteristics such as age or race for this analysis because nearly all the
officers were white males. This limited the variance in the data and made estimates unreliable.
81
π1j = 10 + 11(years of service)j + u1j (3)
The parameter 10 represents the average change in disparity across all officers that is a function of the
time period of the study. This coefficient denotes the effect of time on disparity. The parameter 11 is
the amount of change in disparity that results from an interaction between an officer’s years of service
and time period. Finally, u1j is an error term representing the unique portion of the change in disparity
that is due to a given officer.
π2j = 20 + u2j (4)
The parameter 20 represents the average change in disparity across all officers that is a function of area
of town. This coefficient denotes the effect of a beat on disparity. The parameter u2i is an error term
representing the unique portion of the change in disparity that is due to a given officer.
The table below gives the estimated fixed effects results of HMLM analysis. The table includes results of
estimates of three models: (i) a control model consisting of the intercept parameter only, (ii) a restricted
model consisting of the intercept and slope parameters and (iii) a full model that includes all the
parameters.
Summary for HMLM analysis
Fixed Effects Model 1 Model 2 Model 3
Coefficients Coefficients Coefficients
Net Effects Officers (intercept)
00 0.582 (0.057)*** 0.566(0.059)*** 0.818(0.099)***
01 -- -- -0.0229(0.006)***
Net Effects of Time (slope)
10 -- 0.317(0.080)*** 0.579(0.163)***
11 -- -- -0.0223(0.0104)*
20 - 0.0421(0.058) 0.0412(0.057)
Deviance 376.8 366.2 349.1
n 76 76 76
*p < .05, **p <.01, ***p < .001
The results of HMLM suggest the following: changes in time during the study period are associated with
significant increases in levels of disproportionality, as reflected by officers’ disparity indexes net of area
of town. In the control model the estimated mean disparity across all officers 00) is significantly
different from zero at 0.582. This result serves as a rough and ready indicator that can be used to see if
there is traffic stop disparity in the data, 00’s value suggests there is. Model 2, the restricted model, is
used as a preliminary test of a change in disparity levels across occasions of the study. This model is
analogous to independent t-tests, but this test takes into consideration the nested nature of the data.
Results show that that the intercept 00 equals 0.566 and is significantly different from zero. This value
represents the logged average level of disparity across all officers when the difference between the year
of the study and the grand mean equals zero (the mid-point of the study). The slope parameter 10 is
also significant. This implies that the level of disparity increases over the occasions of the analysis, for a
unit change in year of the study the logged disparity index increases 0..317 units. The slope
82
parameter 20 which indicates the net effects of a beat or area of town on officers’ disparity indexes is
not significant. Finally, the full model tests the net effects of time and officer seniority on disparity. The
two of the three slope parameters in this model are significant. 10, represents the degree to which the
average level of disparity changes as a function of time across occasions of the study, a year change in
time brings a 0.818 unit increase in the average logged level of disparity units net the other variables.
11, is the coefficient for an interaction effect. It indicates whether the stage of an officer’s career
mediates the effect of time on disparity. Results show that a one year increase in seniority reduces the
effect of time by 0.022 logged units. This implies that the year of the study (before or after 2009) had
more impact on less experienced officers than veteran officers. The parameter 20 is not significant. This
suggests that the area an officer worked did not have a net significant effect on levels of
disproportionality. Finally, the analysis for the intercept coefficients, 00 and 01 show that net baseline
levels of disparity across officers are not affected by job seniority. The value of 00, indicates that a
significant amount of disparity remains even after the effects of seniority and news stories are taken
into account. The significant parameter 01, implies that seniority has a net effect on levels of disparity,
meaning that less senior officers have higher disparity indexes than more seasoned officers regardless of
the time period of the study.
83
Appendix E
Adapted Time Line of Some Important Events Affecting ICPD during Study Period
2006 2007 2008 2009
“Groper” appears Suepple Murders
Downtown Drinking & Assaults ‘Mother’s Day Riot”
October 2006, increasing September 2007 with an arrest made July 19, 2008 —The “Groper,” an
assailant who sneaks up behind women, pushes them down, and gropes them before fleeing. Almost 40
cases reported. “Law-enforcement authorities have stressed that they’re pouring resources into solving
these cases.”
“Local police deal with open cases, some take years,” Daily Iowan, REGINA ZILBERMINTS, MARCH 11,
2009, http://www.dailyiowan.com/2009/03/11/Metro/10537.html
2006-2010—Downtown underage drinking and violence crackdown. “In response to a string of random
and seemingly unrelated assaults involving men in the downtown area, Iowa City and UI police are
collaborating to assign more officers to the Pedestrian Mall, where many attacks have occurred.
“Violence tests police,” BY REGINA ZILBERMINTS | APRIL 15, 2009 7:38 AM,
http://www.dailyiowan.com/2009/03/11/Metro/10537.html
2008—Suepple Murders. “Iowa banker facing federal embezzlement and money laundering charges
murdered his wife and four young children in their home before killing himself. . . .”
“Indicted Banker's Desperate Murder-Suicide,” ABC News, DAVID SCHOETZ
March 26, 2008, http://abcnews.go.com/US/story?id=4521545&page=1
May, 2009—The “Mothers Day Riot.” Violent fights that broke out in Southeast Iowa City later dubbed
the Mother’s Day riot.
A1 Number of crime stories published in IC Press Citizen during the study period*