PreprintPDF Available

Elevated Police Turnover following the Summer of George Floyd Protests: A Synthetic Control Study *

Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

Several of the largest U.S. police departments reported a sharp increase in officer resignations following massive public protests directed at policing in the summer of 2020. Yet, to date, no study has rigorously assessed the impact of the George Floyd protests on police resignations. We fill this void using 60 months of employment data from a large police department in the western US. Bayesian structural time-series modeling shows that voluntary resignations increased by 279% relative to the synthetic control, and the model predicts that resignations will continue at an elevated level. However, retirements and involuntary separations were not significantly affected during the study period. A retention crisis may diminish police departments' operational capacity to carry out their expected responsibilities. Criminal justice stakeholders must be prepared to confront workforce decline and increased voluntary turnover. Proactive efforts to improve organizational justice for sworn personnel can moderate officer perceptions of public hostility.
Content may be subject to copyright.
Elevated Police Turnover following the Summer of George Floyd
Protests: A Synthetic Control Study
SCOTT M. MOURTGOS, University of Utah
IAN T. ADAMS, University of Utah
JUSTIN NIX, University of Nebraska Omaha
Several of the largest U.S. police departments reported a sharp increase in ocer resignations following massive public
protests directed at policing in the summer of 2020. Yet, to date, no study has rigorously assessed the impact of the George
Floyd protests on police resignations. We ll this void using 60 months of employment data from a large police department
in the western US. Bayesian structural time-series modeling shows that voluntary resignations increased by 279% relative
to the synthetic control, and the model predicts that resignations will continue at an elevated level. However, retirements
and involuntary separations were not signicantly aected during the study period. A retention crisis may diminish po-
lice departments’ operational capacity to carry out their expected responsibilities. Criminal justice stakeholders must be
prepared to confront workforce decline and increased voluntary turnover. Proactive eorts to improve organizational
justice for sworn personnel can moderate ocer perceptions of public hostility.
Keywords: police turnover, Floyd protests, Bayesian structural time series, resignations
Following George Floyd’s death at the knee of Minneapolis police ocer Derek Chauvin on May 25, 2020,
thousands of protests (and in some cases riots) occurred across the United States. As a result, the institution of
policing found itself once again embroiled in demands for reform. Concurrently, a signicant portion of the
rhetoric surrounding policing during this period was exceptionally negative (e.g., “All Cops Are Bastards” or
“ACAB” became a rallying cry among some critics of policing). Some allege that sustained public scrutiny of
this sort sparked an increase in police resignations across the country. For example, over 100 ocers left the
Minneapolis Police Department – more than double the departures that occur in a typical year (Bailey, 2020).
In Portland and San Francisco, it was reported that police ocers were leaving in record numbers (Colton,
2020; Wallace, 2020). e Chicago Police Department experienced a fteen percent increase in retirements in
2020 compared to 2019 (Main & Spielman, 2021). e monthly loss of ocers from the New York City Police
Department (NYPD) was approximately double the amount experienced in 2019, leaving the NYPD with its
lowest headcount in ten years (DeStefano, 2020). Finally, the Seattle Police Department was reportedly on pace
to lose 200 ocers by the end of 2020 in what has been described as a “mass exodus,” leaving the agency with
its fewest police ocers since 1990 (Rantz, 2020).
ese initial reports are concerning. Fewer ocers per capita has been linked to higher crime rates (Ko-
vandzic & Sloan, 2002; Levitt, 1997; Marvell & Moody, 1996; Mello, 2019), and in 2020, many jurisdictions
experienced a dramatic increase in homicides and shootings (Rosenfeld et al., 2021). However, year-to-date
comparisons regarding police turnover can be misleading. Resignations and retirements from police depart-
ments occur every year regardless of public criticism or civil unrest. In fact, a policing ‘workforce crisis’ was
identied well before the events surrounding and following George Floyd’s death (PERF, 2019). Policing has
been dealing with decreased applicants, increased resignations, increased agency competition in recruitment
is is a pre-print version of a manuscript forthcoming at Criminology & Public Policy. Please contact the authors before citing it
for an up-to-date version. Correspondence concerning this article should be addressed to Scott M. Mourtgos, Department of Political
Science, University of Utah, Salt Lake City, UT 84112. Email: version: July 06, 2021
and transfers, and a pending retirement bubble for at least a decade (PERF, 2019; Wilson, 2012; Wilson et al.,
2010). e challenge, then, is to estimate the counterfactual: how many ocers would have retired or resigned
in 2020 absent the George Floyd protests? In other words, is the observed increase in police resignations signif-
icantly larger than we would have otherwise expected? Or is it merely the continuation of an already existing
is study examined the eect of an abrupt shift in the socio-political climate following George Floyd’s
death on police turnover in one large, capital city jurisdiction in the western US. Specically, we considered
three categories of police turnover – resignations, retirements, and involuntary separations – using Bayesian
structural time-series (BSTS) modeling, a time-series quasi-experimental method (Brodersen et al., 2015). Find-
ings indicate that while retirements and involuntary separations did not experience signicant changes during
the post-Floyd period, police resignations increased substantially. Further, probabilistic forecasting models
predict continued elevated rates of police resignations.
Police Turnover
Understanding trends in police turnover is essential. ough the evidence is quite mixed, some studies have
shown that more police per capita is associated with lower levels of crime – particularly violent crime (Chaln
et al., 2020; Chaln & McCrary, 2018; Kovandzic & Sloan, 2002; Levitt, 1997; Marvell & Moody, 1996; Mello,
2019; Mourtgos & Adams, 2019a)1. Specic to police resignations, Hur (2014) found that increased rates of
voluntary ocer turnover are associated with increased rates of violent and property crime. e direct and
indirect costs of police turnover extend beyond crime levels, however. For example, direct costs consist of
expenses required to hire replacements: background checks, psychological assessments, medical assessments,
and training expenses (Hilal & Litsey, 2020). It is estimated that the cost of losing an ocer ranges from one to
ve times the salary of that ocer (Orrick, 2008). Indeed, a police agency’s return on investment in an employee
is lost when that employee quits after receiving training and education at the agency’s expense (Hur, 2014).
Indirectly, when ocers leave, both the organization and the community lose accumulated institutional
knowledge and specic professional knowledge and skills (Hilal & Litsey, 2020; Hur, 2014). Leavism can lead to
decreased quality of service and productivity (Hilal & Litsey, 2020). Often termed a “brain drain,” the reduction
in performance levels and increased operational risks driven by high turnover are just as real as replacement
costs, though they are dicult to quantify (Wilson et al., 2010). On a large enough scale, turnover can hinder
a police agency’s ability to engage in proactive problem-solving and respond to calls for service, which in turn
could contribute to increased crime and further erosion of public trust (Bailey, 2020; MacLean, 2020; Ren et al.,
In sum, high levels of police turnover have harmful eects at the organizational and community levels. Wil-
son (2012, p. 333) succinctly outlined this problem:
Police employee success is a function of ocer experience and ability to make sensible decisions
with little supervision or oversight. By reducing the number of ocers with experience, turnover
inhibits eective decision making. It diminishes the strength and cohesion a department gains
from experienced sta. Agencies with higher turnover and less-experienced ocers suer reduced
productivity and more-frequent complaints. e cost of training sworn police ocers is substan-
tial, particularly in comparison to other elds. e high level of organizational and job-specic
knowledge required of police ocers also means that high turnover can impair organizational
performance and service delivery while replacement personnel are selected and trained.
1at said, upon meta-analyzing 62 studies concerned with the relationship between police force size and crime, Lee et al. (2016,
p. 431, emphasis added) concluded that “the overall eect size for police force size on crime is negative, small, and not statistically
Police agencies often turn to increased monetary compensation to combat increased turnover (Wilson et
al., 2010). Like other organizations, police agencies view increased salaries as a way to encourage employee re-
tention. As Schuck and Rabe-Hemp (2018) found, there is merit to this strategy. Analyzing data from a national
sample of over 2,000 law enforcement agencies, they found that agencies paying higher salaries were less likely
to experience voluntary and involuntary turnover. As the authors note, investing in higher wages upfront saves
organizations from increased costs due to turnover in the long run. Early investment also reduces involuntary
turnover, which is typically driven by misconduct and is costly in terms of public trust and legitimacy.
However, monetary compensation is only one of several factors known to inuence ocer turnover. As
Giblin and Galli (2017, p. 398) note, higher salaries may be used in an attempt to oset less favorable considera-
tions: “[T]he supply of prospective ocers uctuates based on the perceived appeal of…law enforcement. For
example, while Schuck and Rabe-Hemp (2018) found that higher salaries were associated with lower turnover
rates, they also found that police agencies with a higher propensity for dangerous or strained police-citizen en-
counters were more likely to have elevated levels of turnover. is turnover can hamper organizational eorts
to repair damaged relationships between agencies and the public they serve, as the “reality of strained budgets
and long-standing issues in the hiring and retention of ocers will likely pose signicant barriers” to eorts
such as community-oriented policing (Peyton et al., 2019, p. 19897).
Other key factors driving police turnover include loss of trust and condence in leadership (Wilson et al.,
2010; Wolfe & Lawson, 2020) and sustained negative attention (Mourtgos et al., 2020; Nix & Wolfe, 2017; PERF,
2019; Saunders et al., 2019). A given job’s attractiveness depends signicantly on anticipated outcomes (Steers
et al., 2004). Suppose an ocer’s subjective beliefs regarding the non-mutuality of an exchange relationship
with the public, political leaders, or organizational leaders are strongly negative. In that case, there is a feeling
of loss of control and predictability, resulting in task avoidance (Paoline, 2003; Schott & Ritz, 2018), with the
ultimate avoidance being quitting.
Increased turnover resulting from a perceived non-mutual relationship between ocers and the public, po-
litical leaders, or organizational leaders aligns with the social psychology literature on exchange theory, equity,
and reciprocity. Social exchange theory contends that individuals attempt to maximize the ratio of benets to
costs in their relationships (ibaut & Kelley, 1959). An exchange-based model assumes individuals desire rela-
tionships in which equity is experienced. Hateld and colleagues (1985) describe equity as a state of aairs in
which the relationship’s benets and costs are proportional to the benets and costs incurred by the other half
of the relationship. Suppose ocers fail to obtain equity in their relationship with their leaders or the public.
For example, ocers are trained to a legal and professional standard regarding their use of force. ere is often,
however, a signicant gap between the legal and professional standards ocers operate under and how some
members of the public, media, and political leaders evaluate their actions (Mourtgos & Adams, 2020; Stoughton
et al., 2020). Individuals who become police ocers tend to do so out of a desire to help others (Moon & Hwang,
2004; White et al., 2010). By doing so, they expose themselves to physical harm (Sierra-Arévalo & Nix, 2020;
Tiesman et al., 2018). If an ocer uses force to protect themselves or someone else, following the legal and
professional standards they have been trained to, but then experience backlash from their leaders or the public,
a perception of inequity may result. In such a case, this may reduce an ocer’s motivation to remain a police
ocer (or at least remain a police ocer in a specic jurisdiction), as it increases their chances of experiencing
negative costs2(Mourtgos et al., 2020). Indeed, a loss of perceived organizational support is linked to an o-
cer’s organizational commitment, job satisfaction, burnout, and turnover intention (Adams & Mastracci, 2019;
Gillet et al., 2013).
2is is not to suggest that backlash from leadership and/or the public is always unwarranted. Public attention to illegitimate
police uses of force is an important democratic tool of governance. Because such acts reect poorly on the profession as a whole, even
uninvolved ocers are subject to its impacts. e salience of use-of-force in the minds of the public can result in increased backlash
against the institution as a whole, which can in turn increase individual ocers’ stress levels.
e adverse impacts from accumulated stress and burnout in law enforcement are well-studied (Dick, 2000;
Mastracci et al., 2012; Mastracci & Mourtgos, 2021; McCarty et al., 2019; Sparger & Giacopassi, 1983). ese
include individual impacts such as increased chronic medical problems, substance abuse disorders, and marital
problems, as well as organizational ones, such as increased job dissatisfaction and leavism. Stress is a multi-
faceted concept. In the modern policing context, it includes the perception among ocers that they are un-
usually vulnerable to political and social forces that might follow from even the lawful use-of-force (Mourtgos
& Adams, 2019b). Negative costs may include damage to one’s reputation, family stress, criminal prosecution,
and loss of livelihood, depending on the circumstances (Mourtgos et al., 2020).
Indeed, Saunders, Kotzias, and Ramchand (2019) argue that researchers have not paid enough attention to
how the socio-political climate aects police ocers. As part of a National Institute of Justice-funded study
examining police suicide, Saunders and colleagues conducted interviews in over one hundred law enforcement
agencies across the United States, representing a wide range of agency type and size. While the socio-political
climate was not the most frequently identied police stressor, those interviewed identied the socio-political
climate as the stressor that has increased the most in recent years. Ocers reported that excessive scrutiny
has led to unfair expectations being placed on ocers and the belief that policing is being set up for failure by
the media. Many of those interviewed expressed an increasing climate of nonsupport from their communities.
Further, interviewees expressed worry regarding how police actions caught on camera would be judged in the
public eye without understanding police training and standards.
To be clear, the events following George Floyd’s murder in May 2020 are not unique in policing (e.g., the
1968 Chicago riots, the 1992 Rodney King riots, and the riots and protests following the fatal shooting of
Michael Brown in Ferguson, Missouri in 2014). Further, ocer perceptions that politicians, the media, and
the public misunderstand, misrepresent, and dislike them are not new (Mourtgos & Adams, 2019b). Decades
ago, Niederhoer (1967) conducted his seminal work on police cynicism and found that a substantial portion
of ocers generally harbored negative feelings toward the public, media, and politicians. More contemporary
research has continued to reveal that many ocers feel they are disliked and not trusted by entities outside their
profession (Moon & Zager, 2007; Nix et al., 2020; Yim & Schafer, 2009).
Claims for profoundly adverse outcomes following similar past crises have been empirically scrutinized,
with studies often indicating such claims are overblown. For example, following the 2014 protests and unrest in
Ferguson, Missouri, commentators claimed that excessive criticism of police had led to a so-called “war on cops”
(MacDonald, 2016). However, there is no compelling evidence that ocers’ working environment became any
more or less dangerous than it was before Ferguson (Maguire et al., 2017; Shjarback & Maguire, 2021; Sierra-
Arévalo & Nix, 2020). However, some evidence suggests that college students became more apprehensive about
a career in law enforcement after Ferguson (Morrow et al., 2019; Todak, 2017). Historically, these crises have
been relatively short-lived without documented substantial increases in police turnover (Rhodes & Tyler, 2019).
Yet, it is certainly possible that the contemporary socio-political climate surrounding policing may have more
staying power due to the salience of social media, the rapid spread of images online, and sustained negative
media attention (Rhodes & Tyler, 2019).
Perhaps a tipping point has been reached, where a hostile socio-political climate has become powerful
enough to drive a retreat from the police profession. Anecdotal reports from various cities indicate that this
may be the case (Bailey, 2020; Colton, 2020; DeStefano, 2020; Main & Spielman, 2021; Wallace, 2020). How-
ever, as noted previously, a “workforce crisis” was identied in policing before the events surrounding George
Floyd’s death (PERF, 2019). For at least a decade, the policing profession has been concerned with stang is-
sues due to decreased applicants, increased resignations, and a pending retirement bubble (PERF, 2019; Wilson,
2012; Wilson et al., 2010). It remains to be seen whether 2020 was an atypical year in terms of ocer retention.
e Current Study
e current study analyzed police ocer turnover from January 2016 through December 2020 in one large
jurisdiction in the western US. e agency provides full-spectrum policing services to an estimated 200,000
residents and is located within a metropolitan area with a population of over one million. e department
employs approximately 600 sworn ocers.
George Floyd died on May 25, 2020, in Minneapolis, Minnesota. On May 30, thousands of individuals
staged a protest in the studied jurisdiction based on the circumstances surrounding Floyd’s death. e protest
turned violent when a large crowd surrounded an occupied police vehicle in the downtown area, forcing the
ocer to abandon it. e vehicle was subsequently ipped and set on re. e ensuing riot resulted in looting
and hundreds of thousands of dollars in property damage (including the agency’s public safety building). Sev-
eral ocers sustained serious injuries, and many more received minor injuries during the hours-long process
of quelling the riot. Ocers from around the state were required to respond to resolve the civil disorder. Fol-
lowing May 30, the National Guard was activated for several weeks to assist in ongoing civil unrest. Between
May 30, 2020, and November 2020, the jurisdiction experienced nearly 300 protests directed at police.
In addition to the socio-political climate surrounding George Floyd’s death, body-worn camera footage
from an ocer-involved shooting in the jurisdiction was publicly released in June. e video upset some in
the community and gained persistent local media attention, increasing the negative socio-political climate sur-
rounding policing locally. After the video footage was released, a city council member publicly declared the
shooting unlawful, despite the investigation not having been completed. e subsequent independent investi-
gation located additional surveillance camera footage clearly showing the individual who was shot had a gun
and pointed it at ocers. e local district attorney ruled the shooting was legally justied, but this determina-
tion prompted yet another violent riot that resulted in additional property damage and ocer injuries.
In early August, the jurisdiction’s mayor announced that the agency would be required to revise its use-
of-force policy within 30 days. Specically, the agency’s policy would be required to go beyond the Consti-
tutionally dened requirements for using deadly force—an ‘objectively reasonable’ standard—and move to a
‘necessary’ standard. is announcement was not well received by the agency’s ocers, and no other surround-
ing agency moved to make similar adjustments to their use-of-force policies.3
We employed secondary data from the department for the analysis. e data consists of monthly counts of resig-
nations, retirements, and involuntary separations (described in more detail below) from January 2016 through
December 2020, providing a total of 60 observations for each measure. With the local civil unrest beginning at
the very end of May 2020, a natural inection point for ascertaining any turnover change is June 2020 (month
#54). is sharp inection point is a critical aspect of the research design. It allows for leveraging a natural exper-
iment’s power to analyze trends during the pre- and post-Floyd periods. When a series of measures are broken
up by introducing an intervention that occurs at a specic point in time, time-series analysis is appropriate
(Shadish et al., 2002). e sharp change in socio-political climate experienced by the agency’s ocers provides
an opportunity for a time-series quasi-experiment to ascertain the impact (or lack thereof) on turnover.
3e revised use-of-force policy was ultimately modeled after the “necessary” language in another state’s use-of-force legislation
(California Assembly Bill 392). Due to a lack of clarity about how a ‘necessary’ standard works at the implementation level, the policy’s
full impact is likely yet to be felt. ough research is underway, empirical assessment of the eects of California’s similar statewide
eort has not yet been available.
As Hur (2014) explains, one should dierentiate between voluntary and involuntary turnover due to dierent
etiologies. While voluntary turnover is a decision made by the employee (whether due to individual perceptions,
organizational characteristics, labor market conditions, or otherwise), involuntary turnover is not. When an
employee is involuntarily terminated, it is because of misconduct (layos may be considered another type of in-
voluntary turnover, but no layos occurred during the study period). Recognizing this, and to allow for a more
nuanced understanding of the socio-political climate’s eect on ocer turnover, three measures of turnover
were constructed: resignations, retirements, and involuntary separations. All three measures were recorded
from January 2016 through December 2020, providing 60 time-series data points for each measurement. Table
1 provides the annual counts for each of the three measures.
Resignations consist of monthly counts of ocers voluntarily leaving the police department before becoming
eligible for retirement.
Retirements consist of monthly counts of ocers voluntarily leaving the police department after being eligible
for retirement. Based solely on counts, it is impossible to ascertain the reason why an individual retired (i.e.,
low job satisfaction, age, or otherwise). In this particular agency, ocers are qualied for retirement at either
20 or 25 years, dependent upon their entrance into law enforcement employment. Accordingly, retirements
were analyzed separately from resignations.
Involunatary Separations
Involuntary resignations consist of monthly counts of ocer terminations. While possible that a socio-political
climate calling for additional accountability could lead to an increased level of ocer terminations4, the etiology
of any change in this type of turnover is dierent from voluntary resignations or retirements. Accordingly, it
was analyzed separately.
Table 1: 2016 - 2020 Turnover Measures
Year Resignations Retirements Involuntary Separations
2016 5 21 2
2017 9 22 3
2018 18 29 4
2019 19 21 5
2020 37 26 8
Bayesian Structural Time Series (BSTS) modeling was used for the analysis. BSTS models are ideal because they
are exible and modular. ey allow the investigator to determine whether short- or long-term predictions
are more important, to account for seasonal eects, and whether to include regressors (Scott, 2017). Further,
4e agency reports there were no involuntary separations in 2020 related to allegations of police misconduct during the protests.
by working in a Bayesian framework, investigators can better acknowledge and incorporate uncertainty into
statistical models and discuss outcomes in terms of probabilities (Mourtgos & Adams, 2021).
e analysis proceeds in three steps. First, BSTS models are estimated for each measure. Each model’s
time series is decomposed, and resulting plots are examined for any apparent change in turnover beginning
in June 2020 for all three measures. Second, a causal impact analysis is performed (explained in more detail
below) to assess the post-Floyd socio-political climate’s eect on police turnover. Finally, a probabilistic one-
year turnover forecast is estimated.
BSTS Models for Resignations, Retirements, and Involuntary Separations
BSTS models are best described as observation equations, linking observed data with an observed latent state
and transition equation. e transition equation describes the development of the latent state over time. e
observation equation is dened as
 
 
where is a scalar observation, is an output vector, and is the unobserved latent state. e transition
equation is dened as
   
where is the unobserved latent state, is the transition matrix, and is the control matrix. allows
the inclusion of seasonality in the analysis. e error terms (and ) are Gaussian and independent.
Use of BSTS models are ideal for social science (Brodersen et al., 2015) and have been used in other crimino-
logical studies to examine the eect of mandatory sexual assault kit testing policies on rape arrests (Mourtgos
et al., 2021); the ecacy of COVID-19 vaccinations within a police workforce (Mourtgos & Adams, 2021); vio-
lence reduction as a result of gang arrests (Ratclie et al., 2017); the deterrent eect of unarmed private patrols
(Liu & Fabbri, 2016); and the eects alcohol licensing policies have on crime (de Vocht et al., 2017). Further,
BSTS models are preferrable to Auto Regressive Integrated Moving Average (ARIMA) models when examining
low-volume count data, as we do here (Ratclie et al., 2017).
Model estimation and validation
A BSTS model was estimated for each of the measures with a local linear trend state component and a sea-
sonal state component, allowing for a monthly seasonal parameter. A local linear trend assumes that both the
mean and the slope of the trend follow random walks (Scott, 2020). Ten thousand Markov chain Monte Carlo
(MCMC) iterations were simulated to t each model.
Several post-estimation diagnostics were analyzed for all three models. First, autocorrelation plots were
examined. Autocorrelation was not a problem in any of the three models. Second, a plot of the posterior distri-
bution residuals was generated for each of the three models, showing the residuals to be normally distributed.
ird, one-step-ahead prediction errors for each model were calculated. A time series has t observed times
points  . Predictions are made for the model at times ,, etc. If we denote the steps
ahead forecast of   with data  by , then  is the prediction of   based on data
up to and including time . e one-step-ahead prediction error, then, is described as   
and is the amount by which the model diers from the actual value once it becomes available. e mean one-
step-ahead prediction errors for resignations, retirements, and involuntary separations are 0.14, -0.20, and
0.08, respectively. As explained below, the mean prediction error for resignations reported here is inated.
However, even keeping this in mind, the average one-step-ahead prediction errors are minimal, indicating a
good predictive t for all three models.
Decomposition and assessment
Figure 1 shows the individual state components (i.e., trend and seasonal components) for each model. Decom-
posing and plotting trend and seasonal components separately allow one to see trends in time series data more
clearly without seasonal noise. A regression line is added to each measure’s trend plot. is addition allows for
easy assessment of whether a signicant increase occurred in any turnover measures beginning in June 2020
(the dashed vertical line).
A few observations can be made from Figure 1. First, all three measures show a positive trend over the entire
period, lending credence to past concerns of a growing ‘workforce crisis’ in policing (PERF, 2019; Wilson, 2012;
Wilson et al., 2010). Second, retirements and involuntary separations did not experience any noticeable change
in trends in the post-Floyd period. ird, one can see a substantial jump in resignations above the regression
line immediately in the post-Floyd period5.
Recall that the mean one-step-ahead prediction error is inated for the resignations model. With such a
noticeable jump in resignations following the post-Floyd period, one would not expect the BSTS model to ade-
quately predict resignations during that time. e model’s predictions are based on the preceding years without
the apparent exogenous shock. Accordingly, the mean one-step-ahead prediction error was recalculated for
the resignation model from January 2016 through May 2020 and June 2020 through December 2020. When
measured separately, the mean prediction error for the resignation model is -0.11 and 1.96, respectively—an
absolute dierence of 2.07. is measure is a further indication that the post-Floyd period had a signicant
eect on ocer resignations. Based on the above observations, causal impact analysis is appropriate.
Causal Impact Analysis
BSTS models can be used to infer causal impact by predicting the counterfactual treatment response in a syn-
thetic control that would have occurred if no intervention had taken place. e synthetic control is constructed
using the experimental group’s time-series behavior before and after the intervention and combining control
variables predictive of the target series before the intervention (Brodersen et al., 2015).
e use of synthetic controls is not new in criminal justice research (e.g., Becker & Klößner, 2017; Donohue
et al., 2019; Mourtgos et al., 2021; Mourtgos & Adams, 2021; Saunders et al., 2015). Historically, however, the
5One can observe an increase in the decomposed resignation time-series trend line before the post-Floyd period. is does not
mean that a substantial rise in resignations began to occur before the post-Floyd period. Instead, this is due to the plot visualizing
the model’s trend rather than actual values. at is, as with any trend measure, the values before and after a specic point in time
aect the trend across the entire period, not necessarily indicating what occurred during that particular point in time. In this case,
November 2020 through February 2021 was a low resignation period, with zero resignations. One can see this uncharacteristically
quiet resignation period pulling the decomposed trend line below the regression line. However, from March 2021 through May 2021,
resignations begin to occur again, with a total of four resignations (one, two, and one, respectively), thus resulting in an increasing trend
line. We do not know of any plausible explanation for the zero resignations occurring from November through February and expect it
to be naturally occurring ‘noise.’ Beginning in June 2020, resignations begin a signicant jump with 29 resignations in total—ve, four,
zero, six, eleven, and three, respectively. e ve resignations in June 2021 alone total more than the previous three months combined
(seven months if one includes November 2020 through February 2021 when zero resignations occurred) or any other single month
in the entire preceding time-series. e sustained substantial increase forces the decomposed trend line to elevate rapidly, aecting
when the increasing trend appears to begin. However, when one examines the absolute values rather than the trend line, it is clear that
a substantial jump in resignations did not start until June 2020 (i.e., the post-Floyd period). e decomposed BSTS models presented
in Figure 1 are not meant to provide causal inference. Instead, they are meant for preliminary analysis, which indicates a sustained
increase in resignations in the post-Floyd period. Using the estimated BSTS models that Figure 1 visualizes, the Causal Impact analysis
provides the causal inference portion of our analysis.
Figure 1: “Resignation, Retirement, and Involuntary Separation Trend”
selection of comparison groups has often been made “on the basis of subjective measures of anity between
aected and unaected units” (Abadie et al., 2010, p. 493). Accordingly, as synthetic models have matured, data-
driven procedures to construct synthetic controls have been advocated for and adopted, requiring researchers
to demonstrate the appropriateness of their selected units for comparison.
Yet, even as counterfactual methods have advanced in sophistication, the construction of a counterfactual
still typically relies on units of analysis endogenous to the entity that receives the intervention of interest. For
example, Wu and colleagues (2021) construct a synthetic counterfactual to test the impact of recreational mar-
ijuana legalization on crime in Oregon by using a pool of counties within states that had not legalized mar-
ijuana. Donohue and colleagues (2019) follow a similar approach to test the eect of right-to-carry laws on
violent crimes, as do Saunders and colleagues (2015) to test the ecacy of focused deterrence on crime but at
the neighborhood (instead of state) level. While this method has been shown to be robust (Becker & Klößner,
2017), it still requires comparing the experimental unit to a synthetic control constructed from nonequivalent
controls. at is, there is an assumption that unknown confounding variables between the experimental unit
and nonequivalent controls (e.g., states, neighborhoods, police departments, etc.) are being accounted for in the
statistical model.
For example, if we were to build a synthetic counterfactual for turnover based on similarly sized police
agencies within the US, we would be making the assumption that (1) dierences between those agencies and
the communities they serve do not aect the dierences observed (if any) with the experimental agency, or (2)
the statistical model is adequately accounting for the dierences. Police agency culture, practices, surround-
ing communities, and political idiosyncrasies vary greatly (Bottoms & Tankebe, 2012; Crank, 2014; Paoline,
2003). And while the more conventional synthetic control method (Abadie et al., 2010) has been shown to be
robust (Becker & Klößner, 2017) and can arguably account for this variation, we believe one advantage of our
methodological approach is not having to invoke this assumption. at is, we are able to build the synthetic
counterfactual using the time series behavior of the experimental group prior to and after the intervention, as
well as predictors specic to the agency being studied.
Model Predictors
A key component of utilizing BSTS models for causal impact analysis is the identication of predictors. Predic-
tors are linearly regressed onto the observed values of the target series, creating the composite synthetic control
series (along with the experimental group’s time-series behavior before and after the intervention). Based on
the analysis above, the retirement and involuntary separation measures are used for building the synthetic con-
trol for causal impact analysis. Using a Markov Chain Monte Carlo (MCMC) algorithm, computation of the
posterior distribution of the counterfactual time-series is accomplished. A Gibbs sampler and Kalman lter are
used to simulate from a Markov chain with a stationary distribution for model parameters    . e
posterior incremental eect is described by 
   , where    describe the counter-
factual response. e resulting posterior density is conditional only on the observed data and the priors. Param-
eter estimates and the inclusion or exclusion of covariates with static regression coecients are all integrated
out (de Vocht et al., 2017).
By placing a spike-and-slab prior over coecients, the model can choose the proper variables to construct
the synthetic control through Bayesian model averaging techniques. is ensures a particular set of covariates
do not have to be committed to, nor point estimates of their coecients, avoiding arbitrary selection of co-
variates and overtting (Brodersen et al., 2015). All predictors are weighted by averaging of marginal inclusion
probabilities for each regression coecient (Dablander, 2019; de Vocht et al., 2017).
e pre-intervention time series began in January 2016 and ended in May 2020. e post-Floyd time-series
started in June of 2020 and continued through December of 2020. Accordingly, the pre-intervention series is
53 months in length, with a post-intervention span of 7 months. Ten thousand MCMC samples were drawn
for the analysis, with 95% posterior distribution credible intervals generated6. Results for the model assessing
change in resignations are presented in Figure 2.
Figure 2: “Eect of Post-Floyd Socio-Political Climate on Police Resignations”
Figure 2 contains three panels. e panel labeled “original” shows the data and a counterfactual prediction
for the post-treatment period. e counterfactual prediction is the horizontal dashed line, with a correspond-
ing 95% credible interval surrounding it. e solid line represents the observed data. e panel labeled “point-
wise” represents the pointwise causal eect, as estimated by the model. at is, it shows the dierence between
observed data and counterfactual predictions. e panel labeled “cumulative” visualizes the intervention post-
treatments cumulative eect by summing the second panel’s pointwise contributions. e dashed vertical line
represents the intervention date (i.e., June 2020).
Table 2: Causal Impact Posterior Inference
Parameter Average SD 95% CI
Actual Resignations Post-Intervention 4.71
Predicted Resignations Post-Intervention 1.24 0.42 [0.43, 2.1]
6e frequentist analog to a credible interval is a condence interval, yet the two concepts are statistically dierent. In a frequentist
paradigm, condence intervals are based on repeated sampling theory. A 95% condence interval indicates that if the same experiment
is repeated ad innitum, the unknown, but xed coecient will fall within it. A credible interval can be interpreted as the probability
that the population parameter is between the upper and lower bounds of the credible interval, based on the available information.
Absolute Eect 3.47 0.42 [2.65, 4.28]
Relative Eect 2.79 0.34 [214%, 344%]
Posterior probability of a causal eect > .99
As indicated in Table 2, the post-Floyd socio-political climate is associated with increased police ocer
resignations at the studied police department. During the post-intervention period, the study had an average
of ve (4.71) ocers resigning per month. By contrast, in the absence of an intervention, a counterfactual
average of one (1.24) ocer resignations per month would be expected. Subtracting this prediction from the
observed response yields an estimate of the absolute causal eect the intervention had on resignations (3.47),
resulting in a 279% relative increase in ocer resignations. e posterior probability of the observed eect is
>.99. In other words, there is an extremely high probability that the post-Floyd socio-political environment
exerted a positive eect on ocer resignations at the studied police department.
Assessing the eect of past events is not the only interest of this study. Of practical consequence is forecasting
continued ocer resignations into the future. e BSTS model for resignations can be used to provide a prob-
abilistic forecast for future resignations. Figure 3 plots the resignation forecast for one year beyond December
In Figure 3, the blue line represents the median number of resignations the forecast model predicts. e
dashed lines represent 95% credible intervals. e one-year forecast indicates continuing increases in resig-
nations, predicting a median value of ve (5.43) ocer resignations per month. Noteworthy are the credible
intervals in the forecast, allowing for a greater probability of increased resignations than a decrease in resig-
nations. According to the forecast, it is more probable that the studied department will experience increasing
resignations over the next year than decreasing resignations.
Forecasts are not destiny. Any number of external factors (the economy, pay increases, political shifts, etc.)
could aect resignations (negatively or positively). e forecast cannot account for these possibilities. Rather,
the forecast takes the BSTS model in its entirety (both pre- and post-Floyd) and estimates a probabilistic projec-
tion for the next 12 months based on the preceding data. Moreover, at some point one would expect a threshold
to be met, at which ocers prompted to leave by the post-Floyd socio-political environment have done so, and
the studied agency returns to ‘normal’ attrition levels7.
One way to assess the above-listed forecasting limitations is by comparing the forecast with continued ob-
servations. We obtained an additional two months of resignation data since the one-year forecast was projected
(January and February 2021). e one-year forecast posterior distributions estimated a median value of ve
(5.019) resignations in January 2021 and ve (5.396) resignations in February 2021. Actual resignations in each
month were four additional ocers. e observed values are below the projected resignation values, possibly
indicating a pending reduction in resignations. However, the error is low. us, the studied agency may not
yet be out of the woods regarding continued resignations.
7We are not able to investigate how the post-Floyd socio-political environment aects recruitment. While a resignation maximum
threshold will likely be met at some point, recruitment could potentially continue to suer. However, this is an empirical question –
one which this and other agencies should be assessing.
Figure 3: “1-Year Forecast of Ocer Resignations”
For the past decade, there has been growing concern about a “workforce crisis” in policing, characterized by
an increase in ocers leaving the profession (through resignation or retirement), as well as fewer individu-
als applying to become police ocers (PERF, 2019; Wilson, 2012; Wilson et al., 2010). Since the fallout from
the fatal police shooting of Michael Brown in Ferguson, Missouri, in August 2014, these concerns have only
grown. Yet, the impact of the socio-political climate on police recruitment remains unclear. While Morrow
and colleagues (2019, 2020) provide evidence of college students being more reticent to become police ocers
post-Ferguson (see also Todak, 2017), Rhodes and Tyler (2019) found no evidence of a post-Ferguson decrease
in police applicants in one major metropolitan police department.
While empirical research on police recruitment is scant, empirical research on police turnover is even more
lacking. A better understanding of police turnover dynamics is crucial as it aects both police organizations
and the communities they serve. In the present study, we examined the eect of a sharp change in the socio-
political climate on ocer turnover in a large metropolitan police department. We show the socio-political
environment did not signicantly aect retirements or involuntary separations (although both showed an in-
creasing trend across the full period, lending credence to previous concerns about a shrinking police workforce).
However, shortly after protests surrounding George Floyd’s murder began, there was a 279% increase in ocer
resignations compared to the synthetic counterfactual.
Empirical evidence of such a substantial increase in police resignations, paired with anecdotal accounts
from other agencies country-wide (Bailey, 2020; DeStefano, 2020; Main & Spielman, 2021; Rantz, 2020; Wallace,
2020), is concerning for many reasons. A large exodus of experienced ocers from a police department can
result in a “brain drain,” where institutional knowledge, along with job experience and skill, are lost from a
department. is loss can reduce performance levels and increase operational risks as newer, less experienced
ocers replace veterans (Hilal & Litsey, 2020; Hur, 2014; Wilson, 2012). Prior research indicates that experience
is associated with decreased verbal and physical force (Paoline & Terrill, 2007) and better performance (Sanders,
2008). at being said, prior research also points to experienced ocers having lower levels of job satisfaction
(Johnson, 2012) and higher levels of cynicism (Hickman et al., 2004). Future research must continue to study
the correlates and consequences of police turnover, particularly in agencies where turnover is characterized by
rapid replacement of veterans with rookie ocers.
On a large scale, turnover can lead to a police agency’s inability to respond to calls for service, increased
crime, and erosion of public trust (Bailey, 2020; Oliveira et al., 2020). Indeed, some studies indicate that more
police ocers are associated with less crime (Kovandzic & Sloan, 2002; Levitt, 1997; Marvell & Moody, 1996;
c.f. Lee et al., 2016). Recent research suggests a correlation between police force size and community violence,
such that each additional ten ocers abates approximately one homicide (Chaln et al., 2020). e authors
also note that “although the total reduction in homicide is roughly equal across Black and white victims, the
decline in homicide is twice as large for Black victims in per capita terms” (p. 4). us, a rapid departure of
police ocers from a jurisdiction may have a disproportionately negative impact on communities of color (Nix
& Wolfe, 2020). Using RAND’s cost of crime calculator and open-source crime data from the studied agency’s
website, the experienced loss of 33 ocers to resignation can be expected to result in an $11.64 million increase
in crime costs over a year’s period, with the value of each individual ocer estimated at $357,533 (RAND, 2021).
More analogous to this study is a recent study published by Piza and Chillar (2020), who found that when
the Newark Police Department laid o 13% of its ocers in 2010, violent and property crime promptly in-
creased. Further, overall crime and violent crime became progressively worse each year following the police
layos. While the cause of the turnover examined by Piza and Chillar is dierent from the reason examined in
this study (i.e., layos versus resignations), the two studies inform each other as they are both concerned with
sudden, substantial increases in turnover, rather than gentle ebbs across time. It was beyond our study’s scope
to ascertain the causal eect of rapid resignations on crime in the jurisdiction. However, it should be noted
that – similar to other cities across the United States – this jurisdiction experienced a 22% increase in violent
crime and a 25% increase in property crime in 2020. is could be due at least in part to the agency abruptly
becoming understaed in the summer of 2020, but of course, it might also be a function of societal disruptions
caused by the COVID-19 pandemic (Rosenfeld & Lopez, 2020).
Concerns regarding police turnover and crime are compounded by the realities of how police hiring works.
For starters, a months-long hiring process is required to ensure high standards are met. e hiring process typ-
ically consists of a psychological exam, medical screening, interview, background check, and more. Once hired,
though, police ocers are not “road-ready. ey must graduate from a police academy and then complete a
eld training phase. While the length of academy and eld training varies from jurisdiction to jurisdiction, it
takes approximately ten months to have an ocer ready to function as a solo police ocer in the agency we
studied. Our forecasting analysis is instructive when considering the ability of the agency to recover resigna-
tions through increased hiring. Assuming that the police department could recruit and hire 33 ocers (approx-
imately 6% of its entire sworn sta) to replace those who resigned (i.e., not including the ocers who retired
or were terminated during the same period), it would take almost a year for those 33 new ocers to ll the
vacancies. In the meantime, during every month of the year it takes to train the new ocers, the above forecast-
ing model predicts an additional ve ocers will be resigning, equating to a continuing net loss. Accordingly,
for this agency to adequately address their turnover, they will need to hire well ahead of their authorized size
to experience substantial stang improvements. Considering that retirements and involuntary separations
will likely also continue to occur, it is clear that it will take some time until the agency can return to its full
strength. While forecasting is not destiny, this agency, and others experiencing similar trends, have good cause
for concern.
Financial Costs
ere are signicant nancial costs associated with police turnover. ese costs include the actual hiring pro-
cess and new ocers’ equipment (Hilal & Litsey, 2020). It is estimated that the cost of losing an ocer ranges
from one to ve times the salary of that employee (Orrick, 2008). ere have been well-publicized calls to “de-
fund” or even “abolish the police” since George Floyd’s death (Kaba, 2020; Vitale, 2017). While the modern
concept is relatively ill-dened, the potential impacts of “defunding” could resemble what has historically oc-
curred following signicant police budget cuts.. For example, during the Great Recession many cities were
forced to decrease police stang out of nancial necessity (Wilson et al., 2010). Along with the decrease in o-
cers working in the community, more crimes went unsolved, community outreach programs suered, and re-
sponse times increased. Moreover, ocers were required to work more overtime, and use-of-force complaints
rose (Weichselbaum & Lewis, 2020). Supporting research further shows that when Newark Police Department
laid o approximately 13% of its police force in response to the 2008 recession, violent, property, and overall
crime increased in response (Piza & Chillar, 2020), with the well-known costs of crime passed onto the city.
Similar problems may be compounded contemporarily with an exceptionally hostile socio-political climate
surrounding policing, as ocers experiencing an antagonistic relationship with the public more strongly en-
dorse coercive tactics and use more physical force (Marier & Moule, 2019; Muir, 1977; Terrill et al., 2003). If
violent crime continues to increase, we may enter a negative feedback cycle, as increased victimization is asso-
ciated with decreased condence in the police (Ren et al., 2005). While reasonable people can disagree on the
merits of “reimagining policing,” a socio-political climate that results in substantial increases in police turnover
may be failing to remember what recent history taught us.
Perhaps public monies invested in policing are better suited for other crime reduction eorts or community
outreach programs, as proponents of the #DefundePolice movement contend. If cities move forward with
“defunding” their police departments, it will be imperative that they (1) have a plan for how the money saved
will be reallocated, (2) establish clear goals and objectives, and (3) rigorously evaluate outcomes.
Long-term Eects
Finally, a substantial increase in turnover not only can have immediate negative eects on the organization
and community, it can also have negative consequences throughout the organization for decades to come by
disrupting workforce cohorts. As Wilson and Heinonen (2012) demonstrate, balance in a police agency’s work-
force structure is important. For example, suppose there is a high level of turnover at the senior level. In that
case, the organization may have to progress junior ocers to higher ranks before they possess the skills or expe-
rience to fulll those duties. Likewise, needing to quickly replace large numbers of personnel results in a large
pool of junior ocers, thus inating the pool of individuals looking to eventually progress to higher ranks. If
progression comes too slowly because of a large cohort competing for a small number of senior positions, the
organization can face a future of frustrated and less satised members, possibly resulting in increased future
Wilson and Heinonen (2012) provide several detailed progressions of workforce structure imbalances and
the resultant negative consequences. Due to data limitations, we cannot determine if the experienced turnover
at the studied department may result in a workforce structure imbalance. However, we call attention to an
increased likelihood of these imbalances resulting from sudden, increased turnover and suggest this is an area
demanding attention from policing scholars.
Addressing a Resignation Crisis
So, what can be done to reduce police resignations given the current socio-political climate? We oer sev-
eral suggestions. First, more police agencies should publish data (e.g., use-of-force, stops, arrests, stang)
promptly—either on their website or repositories like the Police Data Initiative. Such transparency is critical
for policing in the 21st Century (President’s Task Force on 21st Century Policing, 2015). Generally speaking,
open access to police data can help nurture trust between the police and the community. When a controversial
incident (like the killing of George Floyd or the shooting of Jacob Blake) occurs, journalists and citizens can
quickly ascertain how the police in their community perform in regards to use-of-force, racial disparities, and
other outcomes, rather than being left to assume issues in one city are representative of all cities.
Second, police executives and their political sovereigns must carefully balance their responsibilities to in-
vestigate and discipline misconduct quickly, while doing so in a way that is likely to be perceived by ocers as
fair (Nix & Wolfe, 2017). ere is a statistically signicant and sizeable relationship between perceived orga-
nizational justice and desirable ocer attitudes and behaviors (Wolfe & Lawson, 2020). Recent work suggests
that ocers who perceive greater organizational justice are less sensitive to the ill eects of heightened pub-
lic scrutiny (Nix & Wolfe, 2016). Further, they are less likely to “de-police” in response to a legitimacy crisis
(Mourtgos & Adams, 2019b) and instead appear to remain committed to working with residents to solve prob-
lems in their communities (Wolfe & Nix, 2016). More generally, research indicates that ocers who believe
their agencies treat them fairly are less cynical (Bradford & Quinton, 2014), hold more favorable views toward
the public (Myhill & Bradford, 2013; Tankebe, 2014; Trinkner et al., 2016), and are less likely to engage in mis-
conduct (Wolfe & Lawson, 2020). Research expects that organizationally-just policy has secondary eects that
may help agencies retain ocers during times of crisis, through lower levels of depression (Y. Wu et al., 2017),
distress and maladaptive behaviors (Trinkner et al., 2016), and work-related burnout among ocers (Adams &
Mastracci, 2019).
In other words, while agencies have little control over how nationalized stories of police misconduct aect
public opinion, leaders can select and train their front-line supervisors to provide a buer between community
hostility and their sworn personnel (Wolfe et al., 2018). President Obama’s Taskforce on 21st Century Policing
recognized that organizational justice must precede enhanced trust between police and the communities that
depend upon them (President’s Task Force on 21st Century Policing, 2015). e organizational literature is
vast (see Wolfe & Lawson, 2020, for a recent meta-analysis), but some general approaches to improving organi-
zational justice include giving room for ocers to express their feelings about current events aecting them;
revisiting and revising policies with an eye towards clarity of procedure and expectations; enhancing due pro-
cess for ocers facing investigation; and ensuring that resulting discipline is consistent and fair. Executive
leadership can model these behaviors, train their supervisors to follow suit, and work to ensure their ocers
carry those same behaviors out to their street-level behavior and decision making.
Finally, politicians and journalists should resist going along with poorly framed narratives. e use of de-
scriptors like “epidemic” to describe civilian deaths resulting from .0002% of all police-citizen encounters (Har-
rell & Davis, 2020; Washington Post, 2021) is at best misleading, at worst dangerous. e framing of how often
and why police-involved fatalities occur can have powerful eects on the way individuals think about police
use-of-force specically, and policing more generally. Labeling every police-involved death as the result of “po-
lice violence” wrongly suggests the ocer is the sole cause of every police-involved fatality, diminishing the
citizen’s contribution to the outcome (Nix, 2020).
Poorly framed narratives foster a ‘no-win’ socio-political climate for police ocers. For example, the initial
video footage of the Jacob Blake shooting in Kenosha, Wisconsin, was inconclusive. While the local police union
asked for the public to withhold judgment until the investigation was complete, the narrative of an unarmed
Black man being shot in the back while entering his vehicle quickly spread across the nation (McLaughlin &
Vera, 2020). Even after an independent investigation found that Jacob Blake was armed with a knife while ac-
tively resisting a lawful arrest—Jacob Blake also gave a public interview admitting to being armed with a knife
(Winsor et al., 2021)—prominent news agencies continued to describe the incident as a police ocer shoot-
ing an unarmed Black man (Wulfsohn, 2021). While this incident’s eects on police retention and morale in
Kenosha remain to be seen, the rushed and inaccurately framed narrative led to rioting in Kenosha, exacerbat-
ing an already hostile socio-political environment. is incident also serves as a reminder that police ocers
are not the only ones impacted by such poorly framed narratives: the damage to buildings and businesses from
the Kenosha riots was estimated at $50 million (Flores, 2020).
To be clear, we are not advocating for indiscriminate defense of all police behavior. All stakeholders must
ardently oppose police misconduct. However, political and organizational leaders (if not news agencies) have
an ethical responsibility to their employees and communities to lead with patience and reason. For politicians
and police executives, this is accomplished by ensuring comprehensive and independent investigations and
addressing any misconduct appropriately when and if it is found. For journalists, it means, at a minimum, doing
their due diligence to speak with researchers who can situate incidents within the proper broader context. For
their part, researchers who have subject matter expertise should pick up the phone – or click “reply” – when
journalists contact them about stories. In this regard, entities like the Crime and Justice Research Alliance are
an excellent resource – they can help connect journalists with researchers, and provide training for researchers
interested in learning about how to communicate their research to broader audiences.
Some limitations impact our ability to condently generalize this study’s ndings, suggesting the need for fur-
ther research in varying contexts. At a broad level, the debate over potential linkages between protests and
police turnover does not end with the ndings presented here. Undoubtedly, there are dierences among
states, jurisdictions, and law enforcement modalities that were not addressable in the current study. Of con-
siderable interest: did cities that experienced signicant protesting suer higher ocer turnover rates than
nearby/similar jurisdictions that did not experience such incidents? More evidence is required to determine
if the large eect detected here can be generalized to other jurisdictions. While the advent of social media
and telecommunication technology allows an event in Minneapolis to be viewed by individuals from Portland,
Oregon to Portland, Maine, local responses to these ‘national’ events varies. While George Floyd’s death had a
substantial eect on policing in Minneapolis (Bailey, 2020), it may have had less of an eect on police turnover
in other areas of the country. Some cities with early re-imaginings of policing have since re-litigated the ques-
tion, with some locales moving from an initial “defunding” position to much smaller budget decreases, or even
substantive increases to police budgeting (Akinnibi et al., 2021). e current study cannot address this class of
limitation due to the lack of timely and accessible police stang data, but we urge future scholars to consider a
design capable of answering it.
We are also unable to point to the exact reasons individual ocers chose to resign. As Paoline and Gao (2020)
show, there is substantial variation in how satised and dissatised ocers perceive various dimensions of their
work. For example, satised ocers appear to prioritize public perception and respect more than dissatised
ocers, while dissatised ocers seem more attentive to factors such as pay and benets. is variation could
conceivably explain which ocers are more likely to resign in the face of circumstances similar to the protests
of 2020. ere may also be variation in resignations by ocer demographics, including tenure length, age,
race and ethnicity, and gender. Again, data limitations at the agency level preclude our own analysis at the
individual level. We resist speculation without investigation and suggest future scholarship to pursue related
research questions to understand personal motivation better. Agencies themselves should also consider using
detailed exit and “stay” interviews to build an evidence base of what is, and is not, driving resignations in their
organization. Researchers can assist in analyzing such data, but the agencies themselves should begin with
proactive eorts to collect the data in the rst instance.
We must also withhold any speculation regarding potential impacts on law enforcement recruitment in the
face of shifting socio-political winds. We were not able to secure access to recruitment data for this study. Fur-
ther research in this area, particularly regarding any impact from the 2020 protests, is strongly recommended
in order to present a balanced assessment of the overall employment crisis in policing.
Our study’s single-agency nature leaves at least one other outstanding question: Are ocers permanently
leaving the profession, or just relocating to other departments? It may be that ocers in our study chose to
leave a sizeable, capital city police jurisdiction that has been the focal point of local protests and riots for the
perceived safety (professional and political, if not physical) of nearby suburban or rural departments (Bradley,
2020; PERF, 2019). Locally, at least twenty geographically contiguous jurisdictions reported shortages in their
hiring and retention eorts over the past several years. Our conversations with local chiefs in these agencies,
who report hiring back ocers who had previously left, provide anecdotal support that at least some of the
studied agency’s turnover may be attributable to this phenomenon.
Additionally, we were unable to control for the potential impact of COVID-19 on police ocer turnover in
our model. e biological exposure risks of policing are genuine (Mourtgos & Adams, 2021), and it is entirely
possible that pandemic-related stress among ocers in this agency accumulated over time, reaching a tipping
point during the protests. It is also possible that economic anxiety about being unemployed protected against
pandemic-related leavism in law enforcement. Ultimately, we are not aware of any contemporaneous accounts
of ocers leaving the profession due to pandemic-related concerns, but given the likelihood of increased ocer
stress during the pandemic, the possibility should not be dismissed without further investigation (Stogner et
al., 2020). Indeed, the intertwining of pandemic and protest in 2020 and their eects on crime and the justice
system are likely to be the focus of ongoing academic debate.
Finally, while our ndings point to a concerning spike in voluntary resignations, we withhold judging all po-
lice turnover as having a negative impact. As discussed earlier, there are valid reasons to design hiring, training,
discipline, and separation systems to ensure police employees can deliver the critical public service expected
of them. It is conceivable that within the population of ocers who resign and retire are some ocers who
should have been involuntarily separated. Under some analyses, less sudden and severe stang pressure can
produce pro-organizational outcomes, such as more eective resource allocation (Wilson & Weiss, 2012). It
is well outside the scope of this study to make such determinations, and future work should aim to unpack
competing explanations as more evidence from other contexts emerge.
ere have been many assertions from police leaders across the US that in the wake of sustained policing
protests, their agencies have experienced a signicant uptick in ocer turnover throughout the latter half of
2020. ere have also been claims that excessive criticism has hampered or even harmed police agencies and
their sworn employees in previous police legitimacy crises. While previous analyses have usually shown those
claims to be overblown or even non-existent (Maguire et al., 2017; Shjarback & Maguire, 2021; Sierra-Arévalo
& Nix, 2020), the present study suggests that with respect to police ocer retention, there is cause for legitimate
concern. Our analysis of 60 months of data shows that while resignations and involuntary separations were
not signicantly altered in the post-Floyd period, voluntary resignations were. Ocers resigned at a 279% in-
creased rate relative to the synthetic control expectations, and this rise has a posterior probability greater than
While questions remain about whether resigned ocers are leaving this agency for law enforcement em-
ployment elsewhere, or are leaving the profession altogether, the forecast model for this particular agency is
alarming. e greatly increased resignation rate is forecasted to continue. e increased loss of ocers to res-
ignation is not easily remedied through recruitment, and places signicant budgetary pressure on the agency,
with the costs of hiring a single ocer estimated to be up to ve times the annual salary for that position. Fur-
ther, a decit of ocers may contribute to an already increased violent crime rate in the studied agency, which
has already become local headline news. e potential cohort eects of a broad-based retention crisis in polic-
ing could ripple across communities for years into the future. We hope researchers will rigorously evaluate the
relationship between police stang levels and community crime and disorder in the months and years ahead.
Adams, I. T., & Mastracci, S. H. (2019). Police Body-Worn Cameras: Eects on Ocers’ Burnout and Perceived
Organizational Support. Police Quarterly, 22(1), 5–30.
Bailey, H. (2020, November 13). Minneapolis violence surges as police ocers leave department in droves.
Washington Post.
Bradford, B., & Quinton, P. (2014). Self-legitimacy, police culture and support for democratic policing in an
English constabulary. British Journal of Criminology, 54(6), 1023–1046.
Brodersen, K. H., Gallusser, F., Koehler, J., Remy, N., & Scott, S. L. (2015). Inferring causal impact using
bayesian structural time-series models. Annals of Applied Statistics, 9(1), 247–274.
Chaln, A., Hansen, B., Weisburst, E. K., & Williams Jr., M. C. (2020). Police force size and civilian race
[Working Paper].
Colton, E. (2020, August 19). “Just the beginning”: San Francisco police ocers leaving department in record
numbers. Washington Examiner.
Dablander, F. (2019). Spike-and-Slab. In
de Vocht, F., Tilling, K., Pliakas, T., Angus, C., Egan, M., Brennan, A., Campbell, R., & Hickman, M. (2017).
e intervention eect of local alcohol licensing policies on hospital admission and crime: A natural experiment
using a novel Bayesian synthetic time-series method. Journal of Epidemiology & Community Health, 71(9),
DeStefano, A. M. (2020, October 8). NYPD: Head count lowest in 10 years. Newsday.
Flores, T. (2020, September 9). Damage due to rioting, unrest in Kenosha tops $50 million; 2,000 Guard
assisted here. Kenosha News.
unrest- in- kenosha- tops- 50- million- 2- 000- guard- assisted- here/article_26473ec9-c08a-5490- 9d09-
Giblin, M. J., & Galli, P. M. (2017). Compensation as a police candidate attraction tool: An organizational-
level analysis. Police Quarterly, 20(4), 397–419.
Gillet, N., Huart, I., Colombat, P., & Fouquereau, E. (2013). Perceived organizational support, motivation,
and engagement among police ocers. Professional Psychology: Research and Practice, 44(1), 46–55. https:
Harrell, E., & Davis, E. (2020). Contacts between police and the public, 2018—Statistical tables (NCJ 255730).
Hateld, E., Traupmann, J., Sprecher, S., Utne, M., & Hay, J. (1985). Equity and intimate relationships: Recent
research. In Compatible and incompatible relationships (pp. 1–27). Springer-Verlag.
Hickman, M. J., Piquero, N. L., & Piquero, A. R. (2004). e validity of Niederhoer’s cynicism scale. Journal
of Criminal Justice, 32(1), 1–13.
Hilal, S., & Litsey, B. (2020). Reducing police turnover: Recommendations for the law enforcement agency.
International Journal of Police Science & Management, 22(1), 73–83.
Hur, Y. (2014). Turnover, voluntary turnover, and organizational performance: Evidence from municipal
police departments. Public Administration Quarterly, 37(1), 3–35.
Jackman, T. (2020, November 21). Homicides skyrocket across U.S. during pandemic, while robberies and
rapes plummet. Washington Post.
Johnson, R. R. (2012). Police Ocer Job Satisfaction: A Multidimensional Analysis. Police Quarterly, 15(2),
Kaba, M. (2020, June 12). Yes, we mean literally abolish the police. e New York Times. https://www.nyti
Kovandzic, T. V., & Sloan, J. J. (2002). Police levels and crime rates revisited: A county-level analysis from
Florida (1980–1998). Journal of Criminal Justice, 30, 65–76.
Levitt, S. D. (1997). Using electoral cycles in police hiring to estimate the eect of police on crime. American
Economic Review, 87, 270–290.
MacDonald, H. (2016). e war on cops: How the new attack on law and order makes everyone less safe.
Encounter Books.
MacLean, C. E. (2020). Improving African American condence in law enforcement: Recruit to optimize
procedural justice, not racial quotas. International Journal of Police Science & Management, 146135572097469.
Maguire, E. R., Nix, J., & Campbell, B. A. (2017). A war on cops?: e eects of Ferguson on the number of
U.S. police ocers murdered in the line of duty. Justice Quarterly, 34(5), 739–758.
Main, F., & Spielman, F. (2021, January 15). Police retirements grow in Chicago, New York, Minneapolis
amid anti-police backlash. Chicago Sun-Times.
Marier, C. J., & Moule, R. K. (2019). Feeling blue: Ocer perceptions of public antipathy predict police
occupational norms. American Journal of Criminal Justice, 44(5), 836–857.
Marvell, T. B., & Moody, C. E. (1996). Specication problems, police levels, and crime rates. Criminology,
34, 609–646.
McLaughlin, E. C., & Vera, A. (2020, August 24). Wisconsin police shoot Black man as children watch from
a vehicle, attorney says. CNN.
Mello, S. (2019). More COPS, less crime. Journal of Public Economics, 172, 174–200.
Moon, B., & Hwang, E. (2004). e reasons for choosing a career in policing among South Korean cadets.
Journal of Criminal Justice, 32, 223–229.
Moon, B., & Zager, L. J. (2007). Police ocers’ attitudes toward citizen support: Focus on individual, or-
ganizational and neighborhood characteristic factors. Policing: An International Journal of Police Stategies &
Management, 30, 484–497.
Morrow, W. J., Vickovic, S. G., Dario, L. M., & Shjarback, J. A. (2019). Examining a Ferguson Eect on
college students’ motivation to become police ocers. Journal of Criminal Justice Education, 30(4), 585–605.
Morrow, W. J., Vickovic, S. G., & Shjarback, J. A. (2020). Motivation to enter the police profession in the
post-Ferguson era: An exploratory analysis of procedural justice. Criminal Justice Studies, 1–21. https://doi.
Mourtgos, S. M., & Adams, I. T. (2019a). e eect of prosecutorial actions on deterrence: A county-level
analysis. Criminal Justice Policy Review.
Mourtgos, S. M., & Adams, I. T. (2019b). e rhetoric of de-policing: Evaluating open-ended survey re-
sponses from police ocers with machine learning-based structural topic modeling. Journal of Criminal Jus-
Mourtgos, S. M., & Adams, I. T. (2020). Assessing public perceptions of police use-of-force: Legal reason-
ableness and community standards. Justice Quarterly, 37(5), 869–899.
Mourtgos, S. M., Mayer, R. C., Wise, R. A., & O’Rourke, H. (2020). e Overlooked Perspective of Police
Trust in the Public: Measurement and Eects on Police Job Behaviors. Criminal Justice Policy Review, 31(5),
Muir, W. K. (1977). Police: Streetcorner politicians. University of Chicago Press.
Myhill, A., & Bradford, B. (2013). Overcoming cop culture? Organizational justice and police ocers’ atti-
tudes toward the public. Policing: An International Journal of Police Strategies & Management, 36(2), 338–356.
Niederhoer, A. (1967). Behind the shield: e police in urban society. Doubleday.
Nix, J. (2020). On the challenges associated with the study of police use of deadly force in the United States:
A response to Schwartz & Jahn. PLoS ONE, 15(7).
Nix, J., Pickett, J. T., & Wolfe, S. E. (2020). Testing a theoretical model of perceived audience legitimacy:
e neglected linkage in the dialogic model of police–community relations. Journal of Research in Crime and
Delinquency, 57(2), 217–259.
Nix, J., & Wolfe, S. E. (2016). Sensitivity to the Ferguson Eect: e role of managerial organizational justice.
Journal of Criminal Justice, 47, 12–20.
Nix, J., & Wolfe, S. E. (2017). e impact of negative publicity on police self-legitimacy. Justice Quarterly,
34(1), 84–108.
Nix, J., & Wolfe, S. E. (2020, June 18). Defunding or disbanding the police is a dangerous idea if done hastily.
Washington Post.
Oliveira, T. R., Jackson, J., Murphy, K., & Bradford, B. (2020). Are Trustworthiness and Legitimacy ‘Hard
to Win, Easy to Lose’? A Longitudinal Test of the Asymmetry esis of Police-Citizen Contact. Journal of
Quantitative Criminology.
Orrick, D. (2008). Recruitment, retention, and turnover of police personnel: Reliable, practical, and eective
solutions. Charles C. omas Publisher, Ltd.
Paoline, E. A. (2003). Taking stock: Toward a richer understanding of police culture. Journal of Criminal
Justice, 31(3), 199–214.
Paoline, E. A., & Gau, J. M. (2020). An Empirical Assessment of the Sources of Police Job Satisfaction. Police
Quarterly, 23(1), 55–81.
Paoline, E. A., & Terrill, W. (2007). Police education, experience, and the use of force. Criminal Justice and
Behavior, 34(2), 179–196.
PERF. (2019). e workforce crisis, and what police agencies are doing about it. Police Executive Research
Peyton, K., Sierra-Arévalo, M., & Rand, D. G. (2019). A eld experiment on community policing and police
legitimacy. Proceedings of the National Academy of Sciences, 116(40), 19894–19898.
Piza, E. L., & Chillar, V. F. (2020). e eect of police layos on crime: A natural experiment involving New
Jersey’s two largest cities. Justice Evaluation Journal, 1–19.
President’s Task Force on 21st Century Policing. (2015). Final Report of the President’s Task Force on 21st
Century Policing.
Rantz, J. (2020, December 28). Rantz: Seattle Police on pace to lose nearly 200 ocers in mass exodus this
year. MyNorthwest.Com.
Ren, L., Cao, L., Lovrich, N., & Ganey, M. (2005). Linking condence in the police with the performance
of the police: Community policing can make a dierence. Journal of Criminal Justice, 33(1), 55–66. https:
Rhodes, T. N., & Tyler, D. H. (2019). Is It cool to be a cop? Exploring the dierential impact of Ferguson on
police applicants. Policing: A Journal of Policy and Practice.
Rosenfeld, R., & Lopez, E. (2020). Pandemic, social unrest, and crime in U.S. cities.
Sanders, B. A. (2008). Using personality traits to predict police ocer performance. Policing: An Interna-
tional Journal of Police Strategies & Management, 31(1), 129–147.
Saunders, J., Kotzias, V., & Ramchand, R. (2019). Contemporary police stress: e impact of the evolving
socio-political context. Criminology, Criminal Justice, Law & Society, 20(1), 35–52.
Schott, C., & Ritz, A. (2018). e dark sides of public service motivation: A multi-level theoretical frame-
work. Perspectives on Public Management and Governance, 1(1), 29–42.
Schuck, A. M., & Rabe-Hemp, C. E. (2018). Investing in people: Salary and turnover in policing. Policing:
An International Journal, 41(1), 113–128.
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Quasi-experiments: Interrupted time-series designs.
Experimental and quasi-experimental designs for generalized causal inference.
Shjarback, J. A., & Maguire, E. R. (2021). Extending research on the “war on cops”: e eects of Ferguson
on nonfatal assaults against U.S. police ocers. Crime & Delinquency, 67(1), 3–26.
Sierra-Arévalo, M., & Nix, J. (2020). Gun victimization in the line of duty: Fatal and non-fatal rearm
assaults on police ocers in the United States, 2014-2019. Criminology & Public Policy. https://jnix.netlify.a
Steers, R. M., Mowday, R. T., & Shapiro, D. L. (2004). e future of work motivation theory. e Academy
of Management Review, 29(3), 379–387.
Tankebe, J. (2014). Rightful authority: Exploring the structure of police self-legitimacy. In Crime, justice
and social order: Essays in honour of A. E. Bottoms (pp. 1–30). Oxford University Press.
Terrill, W., Paoline, E. A., & Manning, P. K. (2003). Police culture and coercion. Criminology, 41(4), 1003–
ibaut, J. W., & Kelley, H. H. (1959). e social psychology of groups. John Wiley & Sons, Inc.
Tiesman, H. M., Gwilliam, M., Konda, S., Rojek, J., & Marsh, S. (2018). Nonfatal injuries to law enforcement
ocers: A rise in assaults. American Journal of Preventive Medicine, 54(4), 503–509.
Todak, N. (2017). e decision to become a police ocer in a legitimacy crisis. Women & Criminal Justice,
27(4), 250–270.
Trinkner, R., Tyler, T. R., & Go, P. A. (2016). Justice from within: e relations between a procedurally just
organizational climate and police organizational eciency, endorsement of democratic policing, and ocer
well-being. Psychology, Public Policy, and Law, 22(2), 158–172.
Vitale, A. S. (2017). e end of policing. Verso Books.
Wallace, D. (2020, December 23). Portland cops resigning at “unprecedented” level midcareer to take low-
paying jobs elsewhere: Report [Text.Article]. Fox News; Fox News.
Washington Post. (2021). Fatal force: Police shootings database. Fatal Force. https://www.washingtonpost
Weichselbaum, S., & Lewis, N. (2020, June 9). Support for defunding the police department is growing.
Here’s why it’s not a silver bullet. e Marshall Project.
White, M. D., Cooper, J. A., Saunders, J., & Raganella, A. J. (2010). Motivations for becoming a police ocer:
Re-assessing ocer attitudes and job satisfaction after six years on the street. Journal of Criminal Justice, 38(4),
Wilson, J. M. (2012). Articulating the dynamic police stang challenge: An examination of supply and
demand. Policing: An International Journal of Police Strategies & Management, 35(2), 327–355. https://doi.or
Wilson, J. M., Dalton, E., Scheer, C., & Grammich, C. A. (2010). Police recruitment and retention for the new
millennium (p. 150). RAND Corporation.
Winsor, M., Brady, E., & Wash, S. (2021, January 14). ABC News exclusive: Jacob Blake speaks out for 1st
time since being shot by police. ABC News.
Wolfe, S. E., & Lawson, S. G. (2020). e organizational justice eect among criminal justice employees: A
meta-analysis. Criminology, 58(4), 619–644.
Wolfe, S. E., & Nix, J. (2016). e alleged “Ferguson Eect” and police willingness to engage in community
partnership. Law and Human Behavior, 40(1), 1–10.
Wolfe, S. E., Nix, J., & Campbell, B. A. (2018). Police managers’ self-control and support for organizational
justice. Law & Human Behavior, 42(1), 71–82.
Wu, Y., Sun, I. Y., Chang, C. K.-M., & Hsu, K. K.-L. (2017). Procedural justice received and given: Supervisory
treatment, emotional states, and behavioral compliance among Taiwanese police ocers. Criminal Justice and
Behavior, 44(7), 963–982.
Wulfsohn, J. (2021, January 5). Washington Post slammed for claiming Jacob Blake was “unarmed” during
altercation with Kenosha police [Text.Article]. Fox News; Fox News.
Yim, Y., & Schafer, B. D. (2009). Police and their perceived image: How community inuence ocers’ job
satisfaction. Police Practice and Research, 10, 17–29.
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
Although a common maxim among many practitioners argues that police departments should recruit their way out of the African American confidence race gap by hiring more minority officers, that maxim is unfounded and redirects our recruitment efforts away from hiring to ensure procedural justice and police effectiveness—the two most powerful determinants of African American confidence in the police. The author’s nationwide survey revealed that African Americans living in cities with more racially representative law enforcement agencies were no more confident in local law enforcement than those living in cities where African Americans were underrepresented. That same survey proved, instead, that African American confidence is far higher where local police forces deliver procedural justice and effective policing than where local police forces are merely racially representative. This article presents the survey findings and explores the policy implications for law enforcement recruitment.
Full-text available
Objectives Test the asymmetry thesis of police-citizen contact that police trustworthiness and legitimacy are affected more by negative than by positive experiences of interactions with legal agents by analyzing changes in attitudes towards the police after an encounter with the police. Test whether prior attitudes moderate the impact of contact on changes in attitudes towards the police. Methods A two-wave panel survey of a nationally representative sample of Australian adults measured people’s beliefs about police trustworthiness (procedural fairness and effectiveness), their duty to obey the police, their contact with the police between the two waves, and their evaluation of those encounters in terms of process and outcome. Analysis is carried out using autoregressive structural equation modeling and latent moderated structural models. Results The association between both process and outcome evaluation of police-citizen encounters and changes in attitudes towards the police is asymmetrical for trust in police effectiveness, symmetrical for trust in procedural fairness, and asymmetrical (in the opposite direction expected) for duty to obey the police. Little evidence of heterogeneity in the association between encounters and trust in procedural fairness and duty to obey, but prior levels of perceived effectiveness moderate the association between outcome evaluation and changes in trust in police effectiveness. Conclusions The association between police-citizen encounters and attitudes towards the police may not be as asymmetrical as previously thought, particularly for changes in trust in procedural fairness and legitimacy. Policy implications include considering public-police interactions as ‘teachable moments’ and potential sources for enhancing police trustworthiness and legitimacy.
Full-text available
In response to Gabriel Schwartz and Jaquelyn Jahn’s descriptive study, “Mapping fatal police violence across U.S. metropolitan areas: Overall rates and racial/ethnic inequalities, 2013–2017,” I provide three reflections. First, the framing of this issue is vitally important. Second, police-involved fatalities represent a nonrandom sample of all incidents involving police use of deadly force (i.e., physical force that causes or is likely to cause death), and unfortunately, we lack comprehensive data on use of deadly force that does not result in fatalities. Finally, to make sense of who is killed by the police, researchers must also identify who was exposed to the risk of being killed by the police.
Full-text available
Maintaining the police workforce level is one of the most salient challenges facing law enforcement today. In the long run, both the supply of and demand for qualified officers are changing in a time of increasing attrition, expanding law-enforcement responsibilities, and decreasing resources. These contribute to the difficulties that many agencies report in creating a workforce that represents the demographics of their communities, that is committed to providing its employees the opportunity for long-term police careers, and that effectively implements community policing. These difficulties, perhaps surprisingly, have persisted through recent recessionary times and may become more challenging as the economy improves. Department resources have continued to decrease and responsibilities to increase, with agencies being asked to do more police work with fewer resources. Some agencies report staffing shortages (a small number still claim continuing drops in applications). Others question whether the long-term commitment of applicants and current officers will persist in times of economic improvement. To help address these challenges and provide lessons for the law-enforcement community, this monograph does not offer any new data but rather summarizes for police practitioners lessons on recruiting and retaining diverse, effective workforces. It provides a means for local officials to identify what has been tried elsewhere and what might be applicable in their own communities. It is a broad analysis of issues confronting many agencies and how these have developed over time. Each agency will face unique circumstances that it must consider in workforce planning; this monograph can provide guidance on relevant lessons learned elsewhere. Focusing specifically on recruitment and retention, this guide does not address other key personnel planning issues, such as specific ways of determining staffing need, allocation of staffing resources, distribution of staff attributes and qualities, and opportunity costs and implications of implementing recruiting and retention strategies on other parts of the organization, which are also important to development of an effective, efficient workforce. Challenges might vary by agency, especially by local conditions confronting each agency.
Full-text available
Organizational justice has been shown to be an important predictor of criminal justice employees' work-related perceptions, attitudes, and behaviors. This study takes stock of the organizational justice effect on criminal justice employees' work outcomes by subjecting the literature to a meta-analysis. Multilevel modeling based on 1,924 effect size estimates drawn from 143 studies (95 independent datasets) was used to establish the empirical status of the organizational justice effect. The results indicate a sizeable relationship between organizational justice and justice system employee work outcomes (Mz = .256, CI = [.230, .283]). The findings also demonstrate that the organizational justice effect size varies slightly across several methodological variations. Specifically, the organizational justice effect size is larger in studies that measure the concept with scales that contain survey items tapping into all four dimensions of justice. Also, we found that outcome type, presence of confounding mechanisms, research design, and sample characteristics moderate the justice effect. We conclude that organizational justice theory is a useful framework for developing a more theoretically informed understanding of justice system employees' work outcomes. We discuss the theoretical implications of the meta-analytic findings and avenues for future research based on the results.
The current study tests the effect of police layoffs on crime through a natural experiment involving Newark and Jersey City, New Jersey’s two largest cities. In response to severe budget shortfalls resulting from the economic recession beginning in 2008, officials in both cities seriously considered police layoffs as a potential component of their cutback strategies. The Newark Police Department terminated 13% of the police force in late 2010 while Jersey City officials averted any layoffs from occurring. The current study uses monthly Part 1 crime counts spanning from 2006 to 2015 to measure the effect of the police layoffs on crime in Newark. Findings of time series generalized least squares regression models indicate the police layoffs were associated with significant increases of overall crime, violent crime, and property crime in Newark as compared to Jersey City in the post-layoffs period. Supplemental analyses found the overall crime and violent crime increases become progressively more pronounced each year following the police layoffs.
The process-based model of regulation has become a dominant framework for understanding why people obey the law. Grounded in past and present research, the current study explores how the process-based model of regulation may explain motivational differences for entering the police profession in the post-Ferguson time era: a time period in which police agencies are expressing increased difficulties recruiting and hiring new officers. Some commentators have attributed these challenges to the heightened levels of public and media scrutiny. Considering the hiring challenges faced by law enforcement agencies, the current study investigates whether students' (i.e. prospective police officers) motivation to enter the police profession is influenced by the procedurally just treatment of individuals by the police. Using a sample of college students from two universities, the findings provide support for the aforementioned inquiry. The implications of these findings are contextualized against the backdrop of the Ferguson Effect, procedural justice, and hiring challenges in the police profession.