Professional Documents
Culture Documents
Volume 11 Number 1
March 2008 27-49
© 2008 Sage Publications
10.1177/1098611107309625
Identifying Good Cops Early http://pqx.sagepub.com
hosted at
Predicting Recruit Performance http://online.sagepub.com
in the Academy
Michael D. White
John Jay College of Criminal Justice, New York
Good cops always seem to identify the causes of problems and to come up with the
least troublesome ways of solving them. Good cops think ahead and always leave a way
out of any tough situation. Good cops rarely have to resort to the law to solve minor
order-maintenance problems like drunks and noisy kids on the street. Good cops spend
their time finding out about the people and places on their beats instead of lurking at
speed traps or near badly marked stop signs.
(Skolnick & Fyfe, 1993, p. 127)
Introduction
As Skolnick and Fyfe (1993) suggest, the implications of good policing are pro-
found. Police officers who perform their jobs well are more likely to be effective in
preventing and reducing crime, to be respected by people on their beat, and to provide
better service. Unfortunately, police departments have long struggled with how to
27
Author’s Note: Correspondence concerning this article should be addressed to Michael D. White,
Department of Law, Police Science and Criminal Justice Administration, John Jay College of Criminal
Justice, 899, Tenth Avenue, New York, NY 10019; e-mail: miwhite@jjay.cuny.edu.
The article concludes with a discussion of the implications of the results for selection
and recruitment of personnel and for police academy training, as well as the impli-
cations for accurately measuring police officer performance on the street.1
Prior Research
“Activity” is the internal product of police work. It is the statistical measure that the
sergeant uses to judge the productivity of his men, the lieutenant uses to assure himself
that the sergeant is properly directing his men, the captain uses to assure superiors that
he is capably administering his district, and the department administrators use to assure
the public that their taxes are not squandered.
• The dark figure of crime: Since most crime is not reported to police, they have no
opportunity to solve it.
• The complex police role: Much of an officer’s time is devoted to noncriminal matters.
• Police discretion: Officers often do not make arrests when legally authorized to do
so, exercising discretion properly (i.e., good performance often means NOT making
an arrest or generating numbers).
• The root causes of crime: Conventional wisdom among criminologists suggests that
a range of factors contribute to crime causation including poverty, broken homes,
peer influences, social disorganization (just to name few), all of which far exceed
the purview of the police.
Still, there is no one best measure, and departments are struggling to piece together
approaches that can quantify quality. Rather, the optimal approach involves employing
a range of different measures that accurately reflect the diversity of police work and
that tap into the “quality” of policing by drawing on citizens and officers’ peers, yet
still incorporates the more traditional measures associated with the numbers game.
with either poor or exceptional performance, with mixed results. The vast majority
of this work has focused on personality traits measured through tests such as the
Minnesota Multiphasic Personality Inventory (MMPI), the Inwald Personality
Inventory (IPI), the Personality Assessment Inventory (PAI), and the California
Personality Inventory (CPI). These tests generally capture the “Big Five” personality
constructs: extroversion, emotional stability (vs. neuroticism), agreeableness, con-
scientiousness, and openness to experience (Sanders, 2003). Although there has been
some success in linking various subscales of tests to specific negative outcomes such
as termination and turnover (Weiss, Zehner, Davis, Rostow, & DeCoster-Martin,
2005), several researchers have concluded that such tests hold little predictive value,
particularly with regard to identifying good performers (Ash, Slora, & Britton, 1990;
Aylward, 1985; Dwyer, Prien, & Bernard, 1990). Alternatively, recent research by
Sarchione, Cuttler, Muchinsky, and Nelson-Gray (1998) and a meta-analysis by
Ones, Viswesvaran, Cullen, Drees, and Langkamp (2003) indicate that selected
psychological constructs—such as agreeableness and impulse control—may be
useful in predicting a broad range of police behavior.
Prior research has also focused on intelligence and education level as a predictor
of police performance. There is a fair amount of research suggesting that intelligence
(measured by IQ tests) is related to performance in the police academy, though
Burbeck and Furnham (1985) point out that there is little evidence of a relationship
between intelligence and street performance. Both Burkhart (1980) and Rafilson and
Sison (1996) suggest that intelligence testing can serve as a good screener in the
selection and training phases by weeding out those “with the most glaring emotional,
cognitive, or background problems,” but such testing is much less effective at differ-
entiating among those who will perform adequately, above average and exceptionally
(Sanders, 2003, p. 314). In a related area, a substantial amount of research has examined
the relationship between college education and police performance, and again, the results
are mixed: many studies find no difference in performance among college-educated and
non–college-educated officers, while a few others have found an association between
college education and positive performance (i.e., lower rates of citizen complaints;
Walker & Katz, 2002; White, 2007).
There has been very little work examining predictors of police academy perfor-
mance, and again, the emphasis has been on personality traits. Detrick, Chibnall, and
Luebbert (2004) examined the relationships between scores on scales of the Revised
NEO Personality Inventory (NEO PI-R) and a range of academy outcomes and
found that a variety of traits were related to academy outcomes. For example, high
scores on Values and low scores on Excitement-seeking were associated with better
academic performance, while lower Anxiety scores were associated with improved
performance at the firing range. Aamodt (2002), Hirsch, Northrup, and Schmidt
(1986), and Varela (2000) also found that selected personality traits and mental abil-
ity were related to academy performance. Cuttler and Muchinsky (2006) reported
that a life history index—focusing on work history, drug use history and criminal
history—was associated with negative training academy outcomes.
First, well-trained officers are generally better prepared to act decisively and correctly
in a broad spectrum of situations. Second, training results in greater productivity and
effectiveness. Third, training fosters cooperation and unity of purpose. (p. 33)
Rookie officers who leave the academy ill prepared for the realities of police
work are at greater risk of performing poorly during the initial weeks and months on
the street, even if under the guidance of a field training sergeant. Poor performance
among new officers has potential negative implications for police-community rela-
tions, civil liability, and the department’s impact on crime. Alternatively, recruits
who are better prepared for police work seem more likely to take initiative and
engage in positive behavior, earning the respect of the community, coworkers and
supervisors. Quite simply, logic suggests that the top performers in the academy will
be the top performers on the street, while those who struggled in the academy are
more likely to struggle on the street.
Unfortunately, this logical assumption is largely untested, as very few empirical
studies have examined the relationship between academy and street performance.
However, in a recent groundbreaking study of police officers fired by the NYPD,
Fyfe and Kane (2006) explored the relationship between recruit characteristics, per-
formance in the training academy, and later performance on the street. They found
that preemployment history, education, and training are important indicators of
future performance:
Officers whose histories include records of arrest, traffic violations, and failure in other
jobs are more likely than other officers to be involuntarily separated from the NYPD . . .
Officers who hold associate or higher degrees are less likely than those who do not to be
involuntarily separated . . . Those who do well in the Police Academy’s recruit training
program are less likely than marginal recruits to be separated as unsatisfactory proba-
tioners. (Fyfe & Kane, 2006, p. i)
is known about recruits prior to the start of training, which factors predict good
performance? Importantly, this study does not focus on personality traits as predictors
but rather focuses on a range of other officer-related characteristics. Also, whereas
previous research has focused on negative outcomes, this article seeks to identify
predictors of exceptional performance. This analysis, though mostly exploratory in
nature, offers a tentative model for predicting officer performance in the academy
that complements the limited prior work focusing on personality traits. This model can
then serve as a launching point for extending the discussion of police officer performance
on the street—specifically how can it be measured, and once accurately measured, how
it can be predicted?
Limitations
This article suffers from a number of limitations that warrant discussion. First, the
article uses data from one recruit class in one police department, which presents
problems for external validity. Second, the article focuses on a measure of academy
performance that is classroom based and does not include performance in other
aspects of training such as self-defense, firearms and weapons, driving, and scenario-
based training. This approach was taken for a number of reasons. First, the study
police department does not quantify performance in these other parts of training in
the same numeric fashion, which makes it very difficult to tease out exceptional
performance. Second, less than 1% of recruits failed to graduate because of perfor-
mance on these other nonacademic aspects of training. Last, in this academy—and in
most others across the country—the majority of training is completed in the classroom,
and the academic exams serve as the primary measure of recruit performance. This
approach, however, may be inappropriate in departments that place less emphasis on
classroom work and academic performance.
The third limitation involves the data available for analysis. The author was
unable to secure access to personnel files for each recruit, which hold a wealth of
data related to their background, selection and recruitment process (i.e., prior
employment, background check information, etc.). The analyses were limited to
variables collected by the training academy staff that, fortunately, have not been
examined in prior research.
Results
The results are presented in three stages. First, there is a brief descriptive discus-
sion of the characteristics of the recruit class. This is followed by review of the
results from multiple regression with the continuous-level final average as the
dependent variable. Finally, results from logistic regression and CHAID with the top
performers dependent variable (90% or higher) are presented.
Descriptive Analysis
Recruits in the July 2003 academy class are primarily male (84%) and White (52%),
though one quarter are Hispanic and 16% are African American (see Table 1). The
average age of recruits is 26 with just 20% more than age 30. Two thirds are residents
of the city, and 7% have a military background. An additional 7% are from the
department’s Cadet Corps program, which functions much like the national Police
Cadet Corps program (with more intensive recruitment and selection, financial support,
and a required period of service following graduation). The department requires that
recruits have at least 60 college credits, though both military and prior law enforce-
ment experience can be substituted for the college requirement. The mean number
of college credits for the recruit class is 80.67. Only 10% had no college credits at
the start of the academy; about three quarters had 60 or more and almost one third
had 120 or more credits (a 4-year degree). Finally, the department requires each
recruit to take a reading test as part of the academy training. The average reading
level among recruits is 11.5, with almost two thirds of the class at a 12th-grade reading
level or higher. Interestingly, there is a sizeable minority of the recruit class with
reading ability below a 10th-grade level (15%).
Table 1
Selected Characteristics of July 2003 Recruits
Characteristics 2003 (%, n)
Gender
Male 83.9 (1,305)
Female 16.1 (251)
Race
White 52.2 (812)
Black 15.8 (246)
Hispanic 25.1 (390)
Other 7.0 (108)
100.0 (1,556)
City resident
No 36.6 (570)
Yes 63.4 (986)
100.0 (1,556)
Age
21-25 51.8 (805)
26-30 27.8 (432)
31-35 17.3 (269)
36+ 3.0 (47)
100.0 (1,553)
Mean 26.4
Cadet Corps
No 92.9 (1,446)
Yes 7.1 (110)
100.0 (1,556)
Military service
No 92.7 (1,443)
Yes 7.3 (113)
100.0 (1,556)
Reading level
Less than 7th 4.8 (73)
7th-9th 10.1 (153)
10th-12th 22.0 (334)
12th+ 63.1 (956)
100.0 (1,516)
Mean level 11.5
(continued)
Table 1 (continued)
Characteristics 2003 (%, n)
College credits
None 10.2 (141)
1-60 14.2 (197)
61-119 46.2 (639)
120+ 29.4 (406)
Mean credits 80.67
Table 2 also shows the nested model with only those predictors that are signifi-
cantly related to the outcome measure. Although statistically significant, parameters
indicate that the model is relatively weak and explains less than 4% of the variance
in the performance measure.
Table 2
Results From Multiple Regression With Final Academy
Average, July 2003 Recruit Class
B SE t Beta p value
Predictor Variables
Race 0.337 0.499 0.675 .020 .500
Sex –3.285 1.234 –2.663 –.073 .008
Age –0.236 0.102 –2.317 –.064 .021
City resident (no or yes) 1.596 0.996 1.603 .047 .109
Reading grade level 1.371 0.224 6.119 .175 .000
Number of college credits –0.001 0.010 –0.145 –.004 .885
Cadet Corps (no or yes) 4.415 1.858 –2.376 –.067 .018
Military experience (no or yes) 0.437 1.801 0.243 .007 .808
Constant 82.327 5.535 14.873 .000
Model statistics
F 8.552
Significance 0.000
R-squared 0.049
df 8
N 1,324
Predictor Variables
Sex –2.775 1.168 –2.377 –.061 .018
Age –0.215 0.097 –2.205 –.056 .028
Reading grade level 1.307 0.205 6.378 .164 .000
Cadet Corps (no or yes) 3.732 1.688 –2.211 –.057 .027
Constant 83.109 4.767 17.434 .000
Model statistics
F 14.379
Significance 0.000
R-squared 0.037
df 4
N 1,514
separating the recruit class into two cells: those who have a reading level of 12th grade
or higher (n = 956, 61.44% of the total) and those with less than a 12th-grade reading
level (n = 600, 38.56% of the total). The split is made based on differences in the
dependent variable: 35.36% of those with 12th grade or higher reading levels posted
an average of 90% or higher, compared with just 11.5% of those with lower reading
levels. An additional split is made from the 12th grade or higher reading level cell
based on race: 39.88% of recruits who were White or Asian or Other with a reading
level of 12th grade or higher posted an average of 90 or higher, compared with 24.65%
of Black and Hispanic recruits with similar reading levels. On the other side of the tree,
the Cadet Corps variable splits from the lower reading level node. Twenty percent of
those in the Cadet Corps with lower than 12th-grade reading levels were in the top
performer category, compared with 10.47% of non–Cadet Corps recruits with similar
Table 3
Results From Logistic Regression With Top
Performers, July 2003 Recruit Class
B SE Wald Odd-Ratio p value
Predictor Variables
Sex –0.112 .200 0.315 0.894 .574
Age –0.006 .134 0.002 0.994 .962
City resident 0.174 .142 1.495 1.190 .221
Cadet Corps –0.453 .281 2.599 0.636 .107
Military experience 0.200 .256 0.613 1.222 .434
Reading level 1.343 .172 61.336 3.831 .000
College credits –0.048 .143 0.112 0.953 .738
Race 0.467 .156 8.973 1.596 .003
Constant –1.532 .612 6.268 0.216 .012
Model statistics
Log likelihood 1411.947
R-squared (Cox and Snell) 0.087
Chi-square 120.087
df 8
Significance 0.000
N 1,492
Predictor Variables
Reading level 1.264 .155 66.732 3.539 .000
Race 0.579 .135 18.351 1.785 .000
Constant –2.288 .150 232.495 0.101 .000
Model statistics
Log likelihood 1600.775
R-squared (Cox and Snell) 0.082
Chi-square 127.936
df 2
Significance 0.000
N 1,325
reading levels. Table 4 summarizes the results of the CHAID analysis illustrating the
characteristics of each of the four termination cells, as well as the percentage of the
total, n, and percentage of the dependent variable (ranging from 10% to 40%).
The results from logistic regression and CHAID are quite similar. Clearly, the
best predictor of being a top performer is reading level, specifically, reading at the
12th-grade level or higher. Race emerges in both sets of analyses, but the CHAID
analysis offers insight regarding the nature of the relationship. Recruits who were
White or Asian or Other were more likely than Blacks and Hispanics to be top
performers, but only when examining recruits with a high reading level. Race was
not associated with performance among those with reading levels below the 12th grade.
And finally, the CHAID analysis shows that recruits in the Cadet Corps outperformed
their colleagues, but only among those with lower reading levels.
Figure 1
CHAID Tree for Top Performers
top performers- final avg of 90 or more
Node 0
Category % n
no, less than 90 73.84 1149
yes, 90 or higher 26.16 407
Total (100.00) 1556
Node 1 Node 2
Category % n Category % n
no, less than 90 64.64 618 no, less than 90 88.50 531
yes, 90 or higher 35.36 338 yes, 90 or higher 11.50 69
Total (61.44) 956 Total (38.56) 600
Table 4
Summary of CHAID End Groups
Description of End Group N % of Total % Top Performer
Discussion
Summary
Despite its obvious importance, departments have long struggled with how to
quantify good police performance. Skolnick and Fyfe (1993) suggest that reliance on
the “numbers game” means that the skilled police officer often becomes completely
invisible in his or her department: preventing crime, de-escalating potentially violent
encounters, respected by the community and colleagues, but rarely producing numbers.
You know what they figure in headquarters? “Hey, nothing ever happens on this guy’s
beat. There’s no crime, no traffic problem, no noisy kids. We put him out there and he
shows us nothing. Never makes any arrests, no tickets. Nothing. Zip. What are we paying
him for? Let’s take him out of there and put him where he’s needed. Let him help the
guy in the next post. There’s lots of crime there and no matter how many arrests and
tickets we give out, the people still complain that they don’t get enough protection. Isn’t
it funny how two streets so close to each other can be so different?”
Hah! If the bosses ever came to look, they’d see that the streets are the same, but that
the difference is the cops. The bosses can tell you what kind of numbers look good in a
report, but they wouldn’t know a good cop on a bet. (Skolnick & Fyfe, 1993, p. 128)
This article suggests that there are lessons to be learned regarding these perfor-
mance measurement problems by focusing on the earlier stage of the academy,
where there are well-established measures of police performance. Using a large
recruit class from one police department, the article uses bivariate and multivariate
analysis to identify predictors of police performance among both all recruits as well
as the top performers—those in the top quarter of the class. The article adds to the
scant existing literature on the topic by examining nonpersonality-related character-
istics and focusing on “good” performance. Findings suggest that there are identifi-
able predictors of police performance including recruit reading level, being in the
Cadet Corps program, age, sex, and, in some cases, race. Just as important, several
notable variables were not associated with academy performance, including military
experience, college credits, and city residency. The implications of these findings for
police recruitment, selection, and training, as well as for measuring performance on
the street are discussed below.
Implications
The findings from multivariate analysis raise some interesting questions regarding
the factors associated with performance in the police academy. Clearly, recruit read-
ing level was the best predictor of performance, as measured in this analysis. Reading
level was both highly predictive of overall performance and of top performance in the
July 2003 recruit class. This relationship was also tested with data from the previous
recruit class (July 2002), and the findings are identical.7 There were also a number
of bivariate relationships with reading level that did not emerge in the multivariate
analysis. For example, males had significantly higher reading levels than females
(11.57 vs. 11.22; t = 2.416, p = .016), Whites had higher reading levels than recruits
of other races (12.18 vs. 10.58 for Blacks, 10.93 for Hispanics, 10.90 for Asian or
Other; F = 61.814, p < .000), younger recruits had higher reading levels than older
recruits (r = –.076, p = .003), city residents had lower reading levels than outside res-
idents (11.12 vs. 12.19, t = –10.861, p < 000), and Cadet Corps recruits had lower
reading levels than regular recruits (10.39 vs. 11.60, t = –5.851, p < .000). In some
cases, differences in reading level explained differences in performance among these
categories of recruits, and in other cases, not. Nevertheless, recruitment and training
staff in the study department may wish to investigate this relationship further, possibly
incorporating it into the earlier selection process (recruits now take the test at the
beginning of training) and also using it to target recruits in the academy who may
require additional assistance and/or tutoring. Moreover, writing and communication
are two critically important—though unheralded—skills that are prerequisites for
good policing. Reading level is typically related to writing and communications abilities,
thereby increasing its relevance to the recruitment, selection, and training processes.
Although reading level is an important predictor of academy performance, the
findings indicate that college education—measured as the number of college credits—
is not. There are persuasive arguments on both sides of the debate regarding college
education requirements for police, and perhaps the most convincing argument on the
“against” side is that research has failed to identify a consistent relationship between
performance and college education. Of course, this prior research suffers from the
performance measurement problems described earlier; specifically, how can we con-
sider the relationship between performance and education if we are unable to effectively
measure performance? Yet, if there is one area of policing where college education
would seem to emerge as significant, it would be the academy. This is the stage of
policing most like the college experience, including classroom time, instruction in the
traditional college pedagogy, and college-like exams. In fact, this article examined
performance in terms of classroom outcomes only. College education, however,
failed to emerge as a factor related to performance.
There are a number of potential explanations for this finding. First, it may simply
be that college education is not related to performance in the academy (or on the
street). Second, it is possible that college education is related to performance but the
analyses have failed to capture the relationship. Or perhaps college education is
related to better police performance but not in a way that can be quantified in class-
room performance. Or maybe the number of college credits a recruit possesses is not
reflective of the recruits’ intelligence or educational level (perhaps reading level is a
better proxy). Regardless of the explanation, the relationship between college edu-
cation and performance in the academy did not emerge in the analyses. As a result,
the police department may wish to consider why it desires a college-educated police
force. Will college-educated recruits be more mature? Will they be more tolerant
of different cultures and religions? Will they be more intelligent, or better critical
thinkers? Clearly, the potential benefits of a college education are multifaceted and
complex, but if the department employs the college credit requirement in hopes of
drawing a more intelligent recruit class, their efforts may be more fruitful if they
drop the college credit requirement and instead administer reading level tests to
applicants as part of the recruitment and selection process.
There are a number of other issues raised from the findings presented here. Military
experience and residency requirements are staples of the traditional recruitment
and selection process. Applicants with military experience are often given preferential
status, or in some cases that experience can be substituted for educational require-
ments (e.g., college credits). The rationale for this is that policing and the military
share many attributes, and those recruits with military experience have already been
trained and exposed to those aspects, thereby making them more attractive as candi-
dates for police officer positions. Over the past two decades, however, a number of
scholars and academics have criticized the militarization of American policing, arguing
that it is incompatible with prevailing philosophies of policing (e.g., community policing),
and that the “cops as soldiers” mentality can actually contribute to strained community
relations and can cause police misconduct (see Skolnick & Fyfe, 1993). There is no
empirical evidence to suggest that former military personnel make better police offi-
cers than those without military experience, and the findings here support this posi-
tion. Military experience was not associated with performance in the police academy.
Police departments have traditionally required their officers to live within the
jurisdictional boundaries of the community they serve. The rationale for the resi-
dency requirement centers on the belief that officers who live in the community will
feel more connected and will have better knowledge of the residents, businesses, and
goings-on; and this connection will make them better police officers. In recent years,
however, many departments have dropped residency requirements because of their
impact on recruitment—it necessarily limits those who can apply—and because of
concerns raised by police unions and fraternal police organizations (i.e., residency
requirements unfairly limit officers’ freedom and may cause undue financial burdens).
Moreover, prior research has failed to establish a connection between residency
requirements and improved police performance, and the findings here support that
position. More than one third of the officers in the July 2003 recruit class were not
residents of the study city, and their academy performance was no better or worse
than recruits who lived in the city.
A number of demographic factors were related to performance in the academy
including officer age, race, and gender. Younger recruits outperformed their older coun-
terparts, and the training staff may wish to further investigate this relationship. Perhaps
older recruits could be offered additional support and assistance—such as tutoring—to
help improve their performance in the academy. The findings that men outperformed
women and Whites outperformed recruits of other races are a bit more troubling.
Kappeler, Sluder, and Alpert (1998) note that policing has traditionally been an occu-
pation for working- and middle-class White males with conservative values, and they
suggest that the nature of police academy training serves to weed out applicants who
do not conform to those norms or characteristics. Kappeler et al. (1998) state,
In order to be selected for employment, police applicants must demonstrate that they
conform to a select set of middle-class norms and values . . . The police academy
refines the cohort again by weeding out those who do not conform to the demands of
paramilitary training . . . in part due to the traditional police selection [and training]
process, the vast majority of people in policing have been and are today, middle-class
White males. (pp. 89-90)
Though some of the differences in performance among men and women and Whites,
Blacks, and Hispanics were explained by differences in reading level, the significance
of race and gender as predictors of academy performance—when controlling for
reading level—supports Kappeler et al.’s (1998) argument. Although there may be
other explanations, it is also possible that there are still inherent attributes of the academy
training curriculum that favor the values and characteristics of White males and make it
more difficult for women and minorities to excel. Given concerns about affirmative
action and the importance of drawing an applicant pool that is representative of the
community (see, e.g., Commission on Accreditation for Law Enforcement Agencies
standards), this is an issue that the training staff needs to revisit.
Last, the study department operates a Cadet Corps program that functions much
like the national Police Corps program. The program intends to draw a more highly
qualified applicant pool, offers a more rigorous and competitive selection process,
and provides financial incentives for those who are accepted into the program (the
national program originally offered up to $30,000 for college education) as long as
a minimum period of service is completed (4 years in the national program). The
intent of the program—both local and national—is to draw the “best of the best,” and
the findings here suggest that the local program is accomplishing its objectives.
Recruits in the Cadet Corps program outperformed other recruits, particularly when
controlling for differences in reading level. The success of the program may warrant
discussions regarding expansion, or applying elements of the Cadet Corps process to
the traditional recruitment and selection procedures.
(college credits, military experience, and residency). Each of these findings should
be explored further to determine, first, if the findings emerge in other recruit classes (in
this department and others), and, second, if these factors also predict street performance
(both during and after field training). The reading level finding is particularly inter-
esting because of its implications for college education requirements. Perhaps the
elusive connection between a college education (and knowledge) and performance
is better captured by focusing on reading level, though race and gender differences in
reading level would have to be addressed.8 The relationships between race, gender,
and performance also warrant a closer look. With regard to street performance, do
White males outperform their female and minority counterparts? There is currently no
empirical evidence to suggest that this is the case. If differences in academy performance
by gender and race do NOT extend into street performance, then Kappeler et al.’s
(1998) argument regarding the self-selection bias of the recruitment, selection, and
training processes would seem to be on the mark. If this is the case, the academy
training curriculum would need a substantial reworking to eliminate gender and racial
biases. And last, if the nonsignificant findings involving college credits, residency,
and military experience carry over to the street, there would be persuasive evidence
suggesting their removal as requirements for employment—at least for this department.
The second area where these findings offer food for thought involves taking
advantage of academy performance to inform assignment and monitoring in field
training assignments. There was a wide range of performance in the academy, with
just over one quarter being identified as top performers. Perhaps field training officer
assignments could be made based on academy performance so that rookies on the
lower end of the performance scale are assigned to the most experienced and accom-
plished field training officers in an effort to bring them up to speed. Or perhaps the
field training curriculum itself could be modified on the basis of academy performance
so that low performers get a more intensive, longer curriculum while top performers
get an expedited curriculum (without compromising the field training experience).
Last, Walker (2005) talks extensively about the value of early intervention systems
as indicators of police performance, both good and bad, and how the system can be
used to enhance performance evaluations: “In addition to correcting performance
problems, EI system experts increasingly recognize that they also have the capacity
to identify and reward good performance” (p. 118). If indeed there is a link between
academy and street performance, information regarding low and exceptional academy
performance could be included in the early intervention system.
Conclusion
This article examines performance in the police academy using data from one
large recruit class, indicating that there are identifiable predictors of performance.
Given the limitations of the study, the findings have a number of implications for
recruitment, selection, and training, particularly with regard to reading level, college
education, demographic characteristics, and residency and military requirements.
The article also seeks to inform the discussion regarding the measurement of officer
performance on the job, and perhaps it raises more questions than it answers in this
area. The assumed link between academy performance and street performance is largely
untested, and this article seeks to take a first step toward investigating that assumption
by examining the academy experience and identifying factors related to performance
at that earlier stage. The next step—one that is far more challenging—involves apply-
ing the findings from the academy to both field training and street performance.
Notes
1. This article focuses on measuring police performance at the individual officer level only, not at the
department level.
2. For example, the officer evaluation employed by the New York City Police Department (NYPD)
requires that the supervising sergeant rate the officer on more than 25 different dimensions including
performance areas and behavioral dimensions.
3. In simpler terms, are police officers more like plumbers and electricians, where the required skills
are learned through on-the-job training, or more like doctors and lawyers where skills are learned in an
educational environment followed by supervised field experience (i.e., residency)?
4. In cases where a recruit “failed out” and did not take an exam, the recruit was assigned a “0” as an
exam score. This 0 would then be calculated in the overall exam average.
5. The training lasts 6 months, running 5 days per week, 8 hr per day (it is not a residential academy—
recruits go home at night and on weekends).
6. The final exam is comprehensive, covering all material from the academy training. Although
recruits could technically “fail” the previous three exams, only failure to pass the final exam resulted in
failure to graduate from the academy. Recruits had to score 70 or higher on the final exam to pass. This
article uses the average of all four exams as the indicator of performance, because it represents a broader
picture of recruit performance over the entire 6-month academy experience.
7. The data from the previous year are incomplete and had to be excluded from the analysis.
8. There are other aspects of a college education that are purported to be related to performance, such
as maturity and tolerance of other cultures, which are likely not related to reading level.
References
Aamodt, M. G. (2002). The end of a 10-year journey: Research in law enforcement selection. Paper pre-
sented at the annual meeting of the Society for Police and Criminal Psychology, Corpus Christi, TX.
American Bar Association. (1980). Standards relating to the urban police function (2nd ed.). Boston:
Little, Brown.
Ash, P., Slora, K. B., & Britton, C. F. (1990). Police agency officer selection practices. Journal of Police
Science and Administration, 17, 258-269.
Aylward, J. (1985). Psychological testing and police selection. Journal of Police Science and Administration,
13, 201-210.
Bayley, D. H., & Bittner, E. (1984). Learning the skills of policing. Law and Contemporary Problems, 47, 35-59.
Burbeck, E., & Furnham, A. (1985). Police officer selection: A critical review of the literature. Journal of
Police Science and Administration, 13(1), 58-69.
Burkhart, B. (1980). Conceptual issues in the development of police selection procedures. Professional
Psychology, 11, 121-129.
Commission on Accreditation for Law Enforcement Agencies. (1994). Standards. Fairfax, VA: Author.
Cuttler, M. J., & Muchinsky, P. M. (2006). Prediction of law enforcement training performance and dys-
functional job performance with general mental ability, personality, and life history variables. Criminal
Justice and Behavior, 33, 3-25.
Detrick, P., Chibnall, J. T., & Luebbert, M. C. (2004). The revised NEO personality inventory as predictor
of police academy performance. Criminal Justice and Behavior, 31, 676-694.
Dwyer, W. O., Prien, E. P., & Bernard, J. L. (1990). Psychological screening of law enforcement officers:
A case of job relatedness. Journal of Police Science and Administration, 17, 176-182.
Fyfe, J. J., & Kane, R. (2006). Bad cops: A study of career-ending misconduct among New York City
police officers (Final report). Washington, DC: National Institute of Justice.
Gottfredson, S. D. (1987). Prediction: An overview of selected methodological issues. In D. M. Gottfredson &
M. Tonry (Eds.), Prediction and classification: Criminal justice decision making (pp. 21-51). Chicago:
University of Chicago Press.
Hirsch, H. R., Northrup, L. C., & Schmidt, F. L. (1986). Validity generalization results for law enforcement
occupations. Personnel Psychology, 39, 399-420.
Jones, P. R., Harris, P. W., Fader, J., & Grubstein, L. (2001). Identifying chronic juvenile offenders. Justice
Quarterly, 18, 479-507.
Kappeler, V. E., Sluder, R. D., & Alpert, G. P. (1998). Forces of deviance: Understanding the dark side of
policing. Prospect Heights, IL: Waveland Press.
McKenzie, I. K. (2002). Distance learning for criminal justice professionals in the United Kingdom:
Development, quality assurance and pedagogical properties. Journal of Criminal Justice Education,
13, 231-251.
Memory, J. (2001). Teaching patrol officer problem solutions in academic criminal justice courses.
Journal of Criminal Justice Education, 12, 213-230.
Ones, D. S., Viswesvaran, C., Cullen, M. J., Drees, S. A., & Langkamp, K. (2003). Personality and police
officer behaviors: A comprehensive meta-analysis. In S. W. Spilberg & D. S. Ones (Chairs),
Personality work behaviors of police officers. Symposium conducted at the 18th annual meeting of
the Society for Industrial and Organizational Psychology, Orlando, FL.
Pollock, J. M., & Becker, R. F. (1995). Law enforcement ethics: Using officers’ dilemmas as a teaching
tool. Journal of Criminal Justice Education, 6(1), 1-20.
Rafilson, F., & Sison, R. (1996). Seven criterion-related validity studies conducted with the national
police officer selection test. Psychological Reports, 78, 163-176.
Rubinstein, J. (1973). City police. New York: Farrar, Straus and Giroux.
Sanders, B. A. (2003). Maybe there’s no such thing as a “good cop”: Organizational challenges in selecting
quality officers. Policing: An International Journal of Police Strategies and Management, 26, 313-328.
Sarchione, C. D., Cuttler, M. J., Muchinsky, P. M., & Nelson-Gray, R. O. (1998). Prediction of dysfunctional
job behaviors among law enforcement officers. Journal of Applied Psychology, 83, 904-912.
Scott, E. J. (1981). Calls for service: Citizen demand and initial police response. Washington, DC:
Government Printing Office.
Simon, F. H. (1971). Prediction methods in criminology. London: Her Majesty’s Stationery Office.
Skolnick, J. H., & Fyfe, J. J. (1993). Above the law: Police and the excessive use of force. New York: Free Press.
Tarling, R., & Perry, J. A. (1985). Statistical models in criminological prediction. In D. Farrington & R. Tarling
(Eds.), Prediction in criminology (pp. 210-231). New York: SUNY press.
Varela, J. G. (2000). Meta-analysis of the predictive validity of personality testing in law enforcement
employment settings. Unpublished doctoral dissertation, University of Alabama.
Walker, S. (2005). The new world of police accountability. Thousand Oaks, CA: Sage.
Walker, S., & Katz, C. M. (2002). The police in America: An introduction. New York: McGraw-Hill.
Weiss, W. U., Zehner, S. N., Davis, R. B., Rostow, C., & DeCoster-Martin, E. (2005). Problematic police
performance and the Personality Assessment Inventory. Journal of Police and Criminal Psychology,
20(1), 16-21.
White, M. D. (2007). Current issues and controversies in policing. Boston: Allyn & Bacon.
Wilson, J. Q. (1968). Varieties of police behavior. Cambridge, MA: Harvard University Press.
Michael D. White, PhD, is an associate professor in the Department of Law, Police Science and Criminal
Justice Administration at the John Jay College of Criminal Justice, and Deputy Director of the Research
and Evaluation Center at John Jay. His primary research interests involve the police, including use of
force, training, and performance measurement. His recent work has been published in Journal of
Research in Crime and Delinquency, Journal of Experimental Criminology, Policing: An International
Journal of Police Strategies and Management, and Police Quarterly.