Shootings by Police Officers: Self-Control and More
I stumbled on a recent paper in the Police Quarterly entitled “Quick on the Draw: Assessing the Relationship Between Low Self-Control and Officer-Involved Police Shootings.”
The authors are Christopher M. Donner, Jon Maskaly, Alex R. Piquero, and Wesley G. Jennings from Loyola, U of Texas at Dallas, U of Texas at Dallas and U of South Florida, respectively.
Quoting from the paper:
While the extant literature on police use of deadly force is voluminous, it is fairly limited with regard to the inﬂuence of oﬃcer characteristics. Moreover, this is the ﬁrst known study to explore an individual-level criminological theory(i.e., self-control) in the context of police oﬃcer-involved shootings. In building on previous studies linking low self-control to negative police behavior more generally (Donner et al., 2016; Donner & Jennings, 2014), this study uses data from a sample of 1,935 Philadelphia police oﬃcers to investigate the extent to which Gottfredson and Hirschi’s (1990) general theory can predict oﬃcer-involved shootings speciﬁcally.
Based on theory and related research, it is hypothesized that oﬃcers with lower levels of self-control will be more likely to have used deadly force because police shooting incidents would provide low self-control oﬃcers (those who are more impulsive, self-centered, short-sighted, thrill-seeking, and easily provoked) with an opportunity to engage in a behavior that it is often spontaneous, can provide immediate gratiﬁcation, is adrenaline-inducing, and can provide an outlet for frustration.
Data and Sample
In this study, we use data collected by Greene et al. (2004) for an National Institute of Justice (NIJ)-sponsored study on police integrity in the PPD. The initial collaboration between Temple University and the PPD began in an eﬀort to help create an information system that would assist the PPD with integrity oversight. To aid this process, baseline information concerning possible predictors of negative police behavior was needed. The data set includes background ﬁles, academy training records, and personnel information for 2,094 police oﬃcers across 17 academy classes from 1991 to 1998. Due to missing ﬁles and incomplete academy training among some oﬃcers, the ﬁnal sample of cases included 1,935 oﬃcers. Additional methodological details may be found in Greene et al. (2004).
On average, the sample was almost 27 years of age (range: 18–55), and approximately two thirds of the sample was male. There was virtually equal representation among White (44.5%) and Black (46.0%) oﬃcers, and the sample included a smaller number of Hispanic (7.4%) and other race or ethnicity (2.1%) oﬃcers. The average education level and length of service was 13 and 3 years, respectively. About one ﬁfth (21%) of the sample was married and one tenth (10.9%) had a parent who served in law enforcement. Additional descriptive statistics may be found in Table 1.
The paper goes on:
Greene et al. (2004) were granted access to various databases maintained by the PPD Internal Aﬀairs Division and Police Board of Inquiry. Speciﬁcally, these databases contained information relating to, among other things, citizen complaints, oﬃcer-involved shootings, other internal investigations, and depart-mental disciplinary actions. These data were collected in the Year 2000; thus,oﬃcers in the sample had been out of the police academy for roughly 2 to 9 years. The outcome variable of interest in this study, police shootings, is measured dichotomously (0 = No; 1 = Yes) and reﬂects whether an oﬃcer had ever been involved in a police shooting in which they discharged their ﬁrearm.
The primary independent variable, low self-control, was constructed from selected behavioral indicators contained within an oﬃcer’s Personal Data Questionnaire (PDQ).2 Individuals, who apply to be a Philadelphia Police Oﬃcer and pass the entrance examination, are referred to the Background Unit of the police department. Here, qualiﬁed applicants are given a PDQ.The PDQ collects self-reported background information, including among other things the applicant’s identifying information, family background, residence history, educational history, employment history, credit history, military record, motor vehicle history, adult and juvenile criminal history, and drug-use history. This information is validated through an interview with a background investigator, a full background investigation, and subsequently a polygraph examination.
The paper continues:
On the basis of the available data, we controlled for 11 relevant variables. Age is a continuous variable indicating the oﬃcer’s age and is measured in years. Gender is a dichotomous variable indicating the oﬃcer’s gender (0 =Female;1=Male). Oﬃcer race was originally measured as a nominal-level variable, but was re-coded into three dichotomous non-White racial variables (Black,Hispanic, and Other) with White serving as the reference category. Marital status was originally measured as a nominal-level variable, and it is recoded into a dichotomous variable, which indicates the oﬃcer’s marital status (0 =Not married or separated; 1 =Married, not separated). Military service is a dichotomous variable indicating if the oﬃcer has ever served in the armed forces (0 =No; 1 =Yes). Legacy is a dichotomous variable indicating if the oﬃcer has a parent who was ever a police oﬃcer (0 =No; 1 =Yes). Education is a ratio-level variable and is measured in years of schooling. Length of service is a ratio-level variable indicating an oﬃcer’s length of service (coded in months of service; as of the Year 2000 when the dependent variable information was collected). Handgun safety academy test score and handgun academy qualiﬁcation score are also controlled for as oﬃcers have varying degrees of competence with their primary ﬁrearm, and both are measured as continuous variables. Finally,we control for arrest history prior to employment, and this variable is measured dichotomously (0 =No history of arrest; 1 =History of at least one prior arrest).
So they ran a simple logistic regression. Its basic econometrics. Here are the results:
If you’ve never dealt with this type of analysis, here is their interpretation:
Consequently, these data reveal that oﬃcers evincing lower self-control are more likely to use deadly force with their ﬁrearm. Additionally, male oﬃcers(B=1.612, SE =.386, p<.001), Black oﬃcers (B=.755, SE =.247, p<.01), oﬃcers with a parent in the profession (B=.671, SE =.297, p<.05), and more experienced oﬃcers (B=.028, SE =.008, p<.01) were more likely to have been in a police shooting. Conversely, more educated oﬃcers (B=.022,SE =.009, p<.05) were less likely to have been in a police shooting.
Most people don’t read the innards of a paper, which may explain why the conclusion at the very end is more voluminous with the verbage and, simultaneously, far more parsimonious with the findings:
As agents of social control, the police have a very unique role in society. They are empowered to use force—including deadly force—against the citizenry, and incidents of police use of force can have far-reaching consequences. Building upon existing studies, which have linked low self-control to negative police behavior (Donner et al., 2016; Donner & Jennings, 2014), this study utilized data from the PPD in an eﬀort to improve our understanding of police shootings. The results demonstrated empirical validity for Gottfredson and Hirschi’s general theory as low self-control was found to be predictive of oﬃcer shootings.These ﬁndings suggest that police oﬃcers with lower self-control are more likely to be ‘‘quick on the draw,’’ and this result yields important policy implications for police administrators who have a vested interest in making sure that their oﬃcers are extremely prudent in their deadly force decisions.
The conclusions on self-control tie in with my previous post. But there are several variables that are more statistically significant (read: their effects on police shootings were far less likely to be statistical artifacts) than self-control. At least two of them, gender and race, can be applied more broadly than just to police officers. Thus, those two variables can be used a sanity check for the rest of the paper, even if the authors treated those variables the way they would an overbearing relative to whom they owed money.
So let’s do that sanity check, shall we? Now, among Philly police officers males are far more likely to be shooters than women. And when I looked at FBI and Washington Post statistics it turned out that males were also disproportionately likely to be killed by police, to be police personnel, to be police personnel that were killed, to be killers of police officers, and to be homicide “offenders” (to use the anodyne term the FBI seems to prefer). Similarly, this paper finds that Black police officers in Philly are statistically more likely to be shooters than other officers. This also ties in with my earlier post which found that across the US, Black people are more likely to be shot by police, to kill police officers, to be murder offenders, and to be murder victims than one would expect based purely on their share of the population. This suggests that whatever drives a person to shoot someone else (whether your explanation is one that fits with the far right or the far left or anything in between) doesn’t magically go away because a person becomes a cop.
Now, to throw some gasoline onto the fire. Males commit a disproportionate amount of violence. I cannot think of cases, now or in the past, where it hasn’t been true. And we see this same pattern of disproportionate violence by males in our closest primate cousins too. It takes a special person – a something or other studies major, or worse, a professor in the same field – to believe male violence is in any way attributable to anything other than biology and genetics. And biology and genes are less easily “fixable” than other causes. Furthermore, any such fixes that might exist are likely to be more laden with unintended consequences.
So… take this paper which purportedly deals with self-control, and everything else you know about violence, and come up with some policy. Just about everyone would agree that having less violence by both civilians and police officers would be a good thing. How do we get there? Show your work. But be forewarned – if you contradict the data you’ll get a failing grade.
In 1971 Philip Zimbardo conducted an experiment where a group of student volunteers were divided into 2 groups. One group was assigned the role of prisoners and the other group was assigned the role of guards. The experiment had to be stopped after 6 days because the roles had been played out excessively.
Here is a wikipedia criticism of the experiment:
“Some of the guards’ behaviour led to dangerous and psychologically damaging situations. One third of the guards were judged to have exhibited “genuine sadistic tendencies”, while many prisoners were emotionally traumatized; five of them had to be removed from the experiment early.”
Police officers are not born police officers, they are merely playing out the role.
The answer to the problem has to be psychological screening and very carefully crafted training. And the trainers have to be very carefully supervised. I would say that current screening and training are failures.
We should all condemn any police department whose training does not emphasize that when feeling threatened, police officers should take up a defensive position first and only secondarily respond with deadly force. Unfortunately we have been seeing police departments using deadly force first.
The most important objective for a police officer is NOT to be able to go home to his family after his shift. It is to serve and defend the community that employs him. Any police officer who says otherwise should be fired.
Juries around the country have been acquitting police officers who were caught in the act of killing a non threatening citizen. I don’t believe that they are all racist. Perhaps the problem is that our widely held belief systems will not allow us to condemn someone who was thrust into a situation not of his own making? Prosecutors need to use a more thoughtful approach. Perhaps training officers should be put in the witness chair so that they can defend ‘unfortunate’ homicide, for hours on end.
But maybe the best that we can hope for is that police officers who overreact will be fired and placed on a federal list of ex police officers who should not be employed in a public safety position. Laws would have to be changed to force the firing and enforce that list.
Actually in Jim’s comment there is the nub of the problem, “We should all condemn any police department whose training does not emphasize that when feeling threatened, police officers should take up a defensive position first and only secondarily respond with deadly force. ” Taking a defensive posture is not the male role typically it is to be aggressive not defensive, as taking a defensive posture can lead to one being called a coward. (not on police forces but in society in general). So the issue is the societal norm for male reactions to threats.
The paper cited spends considerable space that directly addresses your general “come up with some policy” in the section “Policy Implications:
I’ll just copy the 1st part of that section:
Our ﬁndings also yield important policy implications for police administrators In particular, department screening eﬀorts should consider paying more attention to the sorts of behavioral markers that may be reﬂective of lower self-control, such as implementing more stringent hiring criteria (Palmiotto, 2001;Sechrest & Burns, 1992), the use of more rigorous background investigations(Palmiotto, 2001), and an increased use of psychological exams and interviews(Arrigo & Claussen, 2003; Cochrane, Tett, & Vandercreek, 2003). Stringent disqualifying criteria and rigorous background investigations are important because they help police administrators conﬁrm if an applicant is suitable for employment. Psychological evaluations, such as personality inventories and structured interviews with trained clinicians, can also be helpful in detecting both behavioral and attitudinal indicators of low self-control. The Minnesota Multiphasic Personality Inventory, for example, includes several items that tap into Gottfredson and Hirschi’s (1990) dimensions of low self-control (e.g.,impulsivity, temper, and thrill-seeking).”
It goes on with many more recommendations.
What I’m curious about though is what Mr. Kimel “comes up with” as policy. “How do we get there? Show your work. But be forewarned – if you contradict the data you’ll get a failing grade.”
Re: JimH, no racism can’t be the whole problem. Consider this weekend’s headline from CNN featuring the fatal shooting of a white female yoga instructor in Minneapolis MN from Australia. The police officers who are involved in the shooting were both wearing body cams but neither had them turned on. Not much has been disclosed on the case yet but it highlights how much still has to be done, e.g. Ensuring officers actually turn on devices that are used to guarantee they act in accordance with their training and department policy.
Why is there an “Other Race” (with a high average co-efficient but high SE) and not “Caucasian” (which surely would be the largest group). Something stinks here. (Oh and talk with Robert Waldmann about statistical significance).
Oh and Mike, by the way gender and race are completely different factors as far as genetics are concerned. Gender is at least several hundred million years and can be definitely determined by a simple test. Race is not clearly defined and is very in evolutionary terms very young.
So what are you suggesting? Only arm female police officers? It would seem to be indicated.
So wouldn’t there be a benefit to identifying that one third (or more?) and weeding them out? Take a look at the table in the post again.
I’m not sure I agree. But I would note – based on the paper’s findings (see the table and the paragraph quoted immediately below it), it you conclude that about males you also conclude that about Black people. As I said, make sure whatever you state conforms with the data. If two variables point in one direction, you cannot make a statement about one variable and not the other unless you have an argument to make about statistical artifacts. So think carefully about whether you want to make the argument you did.
The authors’ analysis doesn’t seem to be wrong, but they are mostly ignoring their own results.
Their policy conclusions (and most of the paper) seem to ignore all the variables that are more significant than self-control. Put another way – if you follow their own findings, reaching the conclusions they did is like worrying about a yapping chihuahua on the sidewalk across the street instead of the pit bull clamped onto your leg.
The table I quoted would seem to suggest that a disproportionate share of shootings are not what you are comfortable calling racially motivated shootings. Now, that isn’t the impression you’d get from watching the news. So perhaps a) the data and/or the analysis is wrong, and we have to contend with a bunch of white racist cops gunning down Black people. Alternatively, b) the news is wrong and being misreported, in which case we have to contend with something that looks quite the opposite and which, presumably, cannot be solved through methods designed to reduce white racist cops shooting Black people. I would suspect that what works in case a does not work in case b. I would go further and suggest that in general, if you are trying to tackle issue a, but the real issue is b, the solutions to a you implement will at a minimum have unintended consequences if not make b worse.
As quoted in the post:
It is fairly standard statistics to take the largest group and make it the baseline.
If your only concern is to reduce shootings by police officers, then that would be indicated, as you note. You can do better still by not having police at all. But I suspect society wants something more from police than just not shooting people.
Amazingly enough, there are cretins who would dispute that. I did make allusions to that in the post.
Um, no. As a general rule, race is shorthand for “where your ancestors are from” and that can be determined pretty accurately.
And just as the two genders adapted in different ways, people who lived in area X and people who lived in area Y developed along different avenues. I’ve noted many times before – when my wife was pregnant, the fetus was given some genetic tests and not others based entirely on where our ancestors were from.
I can’t comment on how local police do their business, but I did serve 25 as a federal LEO and I know that one big change after 9/11 was the elimination of what was call “The Use of Force Model” in federal LEO training. The model trained officers to respond with only that amount of force appropriate for the situation: i.e., an unarmed suspect should always be handled with non-lethal detention techniques. It was drummed into us so that it became second nature that drawing a firearm was only appropriate when a “reasonable person” viewing the circumstances would conclude that there was an imminent possibility of deadly force being used against the officer or a bystander.
Training is a big modifier of behavior in the field. In the Minnesota shooting, it is pretty clear to this “reasonable person” and retired LEO that the officer’s training was woefully lacking.
The objective of the study and analysis was to test the hypothesis that officer self control is significant element of officer involved shootings:
” Based on theory and related research, it is hypothesized that oﬃcers with lower levels of self-control will be more likely to have used deadly force because police shooting incidents would provide low self-control oﬃcers (those who are more impulsive, self-centered, short-sighted, thrill-seeking, and easily provoked) with an opportunity to engage in a behavior that it is often spontaneous, can provide immediate gratiﬁcation, is adrenaline-inducing, and can provide an outlet for frustration.”
THAT was the entire point of the study and analysis. It was not designed to test or rank all reasons for officer involved shootings as the text of the study clearly explains describes.
That would require far more information than the study data was able to obtain as the text also explains.
You are simply taking other peripheral findings from the study, which are not by themselves sufficient to draw substantiated conclusions … for example that Black officers coefficient was 075 or so is relative to white officers which the study clearly stated as used a the reference case… e.g. coefficient = 1.
The study simply affirmed that the hypothesis is supported by a single city’s time limited data. It was not designed to do any thing else, nor were the authors stepping into other subjects since their study design was not sufficient to draw conclusions on any other subject … basically because no other hypothesis was descriptively being tested.
The other information is something that now a new hypothesis can be described and tested.. and so perhaps you should request such be done. Also, note the authors stated clearly that the single city limited time period cannot be used to describe a general US condition, although the hypothesis is that “self control” is a perennial and time independent variable…
Thanks for your comment.
You’d have made one heck of a newspaper. I can imagine the tirade upon one of your writers:
“You were sent to the Ford Theater last night to review a play! Did I tell you to write about the President getting shot? No, I most certainly did not. I don’t care that it happened four feet away from you, and that your coat got splattered in blood. You had no business taking your eyes off the stage. I’ve got a newspaper to publish. The readers want to know about the play. You have twelve minutes to get me a review or your are fired!”
“As a general rule, race is shorthand for “where your ancestors are from” and that can be determined pretty accurately. ”
No its not and a surprising number of people have ancestors from multiple places (of course if you go far enough back all are from Africa).
“So wouldn’t there be a benefit to identifying that one third (or more?) and weeding them out?”
Yes, as I wrote before “The answer to the problem has to be psychological screening and very carefully crafted training.”
So I doubt that psychological screening alone would solve this problem.
And I don’t believe that psychological screening and good training will completely counter the tendencies noted in the Philip Zimbardo experiment. There will always be room for police officers to fill in the training gaps while playing out their role.
There has to be a mandatory weeding out process to deal with the very few police officers who physically abuse the citizenry. (Including shooting them.)
In the Minnesota case, the two police officers were approaching the scene of reported criminal activity and they did not turn on their body cameras! Perhaps the loss of a months pay would cause police officers to remember better!
I would happily see cameras in classrooms and public locations as well. I do agree that cops should be required to have cameras on just about all the time when in the job.
Got it. So if you took fifty people living in, say, Japan 2,000 years ago, and then ten thousand of their descendants eighty years ago, what percentage of those descendants do you expect you would have classified as Asian, White, Black, Native American, etc?
Repeat the experiment using fifty people living in what is now Northern Canada two thousand years ago. Repeat it again using fifty people living in what is now the Democratic Republic of the Congo two thousand years ago. Do it yet again using 50 people living in Denmark 2,000 years ago.
Do you really believe that the average adult in most parts of the world cannot ace the tests I suggested?