Abstract
Logistic regression was used to assess five different police actions that an investigating police officer can choose to employ when handling a domestic violence call. Each significantly increases the likelihood the prosecutor will file charges: obtain photographs (60 percent); find and arrest the defendant (94 percent); obtain an emergency protective order (87 percent); locate additional witnesses (68 percent); and list more than one criminal charge in the police report (284 percent). Three optional police actions increase the likelihood of criminal conviction: find and arrest the defendant (78 percent); obtain an emergency protective order (102 percent); list more than one charge (142 percent). Survival analysis shows a sixth action, completing the investigation the same day, to significantly increase rates of criminal case filing and also rates of criminal conviction. A strong case, best practices model for the investigation of domestic violence incidents was validated and is presented. Police discretion is discussed. Lawmakers should consider making these optional investigative actions mandatory.
The goal of increasing prosecution and conviction of domestically violent persons is shared by a wide variety of groups including criminal justice professionals, social workers, medical professionals, religious leaders, politicians, advocates and academics. Doing so responds to, and operationalizes values of justice, fairness and service to victims, and it creates a special opportunity for the State to protect children living in violent homes. Sentencing can involve monitored mandates for convicted parent(s) to participate in scheduled and unannounced drug and alcohol testing, and to attend classes such as anger management, parenting skills, even to attend vocational training. Sentencing orders can permanently remove weapons from the home, and mandate regular employment, payment of child support, and victim restitution. Always hanging over the head of the convicted person is the threat that a suspended jail sentence will be imposed for violations of the terms of probation.
Before a domestically violent person can be subjected to post-sentencing mandates they must first be charged with and convicted of at least one crime. Before that can be done a sufficient police investigative report must be received by prosecutors, because it is the information source that they rely upon when deciding whether to file criminal charges, or not. It also provides the basis for prosecution. A report that results in the filing of criminal charges must be preceded by an adequate police investigation. Because dozens of unique circumstances attend any crime, a decision has to be made by the investigating police officer regarding how much effort will be put into detecting, recording, analyzing, and preserving evidence, and how much of it will be allowed to slip away.
Crime scenes are junctions where dozens of phenomenon converge in a unique way at a particular moment in time, in a manner that cannot be exactly duplicated ever again. Variables include the number of people present; their age, sex, gender, culture, intelligence, memory capability, their relationship with each of the principles involved in the crime, position of observation, psychological influences, and the manner in which personal belief systems shape their understanding of what they witnessed; their direct, indirect, or non-involvement in the event; lighting, temperature, time of day, visual, auditory, and physical obstructions; mobility and containment factors such as fences and walls, and so on. For the involved parties there are issues of history, learning, temperament and anger, motive, intoxication, maturity, reciprocity, mental intelligence, and so on. Tangible factors can include actual or improvised weapons, broken and damaged materials, the presence or absence of telephones, as well as historical records such as email and text messages, and so on. Other factors may be influential as well, including level of poverty, unemployment, criminal history and criminal learning, and the shaping influence of having witnessed parental violence, experiencing child abuse, and so on.
The challenge of what to do with so many phenomena led Fisher (2004: 33) to note that individual crime scenes are unique, and thus the investigation of them must be organized according to their particulars. It is impossible, in advance, to envisage the range of investigative possibilities for a given criminal incident (Chatterton, 1979: 85); that is why officer judgment becomes necessary in order to organize the investigation and police response. Investigating officers must create a personalized approach to each unique encounter (Bittner, 1967).
The manner of organization and the sufficiency of a police investigation is somewhat dependent upon the routine work practices of the primary investigating officer, whether minimal, average, or thorough. As Lipsky notes (1980: 13), ‘Police decide who to arrest and whose behavior to overlook’, to which Levenson (2010: 4) adds ‘a police officer’s discretionary decision to arrest or not arrest a person in many cases determines whether a prosecution goes forward’.
The topic of police officer discretion has been disputed in the scholarly literature for at least 250 years. 1 Each side in the argument can be reduced to a basic premise. Those in support of broad police discretion tend to trust officer judgment, and thus they are resistant to service mandates because they believe excess obligations can harm officer initiative and judgment. Those in support of mandates wish to subvert performance mediocrity, something that naturally attends discretion, by establishing performance requirements. Nickels’ (2007) review of the discretion literature from 1960 onward provides a useful overview of many of the current directions of work, and important contributors.
Reading the discretion literature demonstrates more than a few scholars and critics have observed individual police officers can subvert the potential of an investigation if they routinely perform to minimal standards, rather than answering the call to excellence. An example illuminates this potential. For a year the National Institute of Law Enforcement and Criminal Justice studied 4500 police officers at a large municipal agency in the USA, finding that 8 percent of the officers produced over half of the investigations that resulted in criminal convictions, and the investigative work of 5.5 percent were responsible for more than half of all felony conviction. Most surprisingly, 46 percent of the officers did not make a single arrest that year (NILECJ, 1977). 2
The NILECJ findings were replicated by the National Institute of Justice, who measure the work of low conviction rate officers as minimal, describing their average activities such as standing around ‘preserving the crime scene’, and concluding ‘nothing can be done’; not searching for witnesses, and so on. In contrast, the NIJ (1982) describes officers with high conviction rates as those who search the surrounding area in order to locate witnesses and physical evidence rather than ‘preserving the scene’; they use more procedures and techniques for obtaining evidence; they are more likely to effectively interrogate the suspect; they adhere more strongly to the legality of the investigative process; they are more business-like in their procedures; they are more persistent; they have special ways to win the cooperation of reluctant witnesses; and, they are more sensitive to the welfare of witnesses. Ewing (1977: i–ii) has labeled officers of this type as ‘supercops’.
The NILECJ and NIJ findings both suggest the routine work practices of many police officers are inadequate, a possibility that is consistent with complaints raised by victim advocates. Police have been accused of purposefully delaying their response to domestic violence (DV) calls, even for several hours, in order to leave the parties to resolve the situation themselves (Hoctor, 1997), and for being predisposed not to arrest even when one is lawfully warranted (Dobash and Dobash, 1979). Police have been accused of discrimination against DV victims (Buzawa and Buzawa, 2003), of being lenient toward men (Avakame and Fyfe, 2001), and being insufficiently concerned about the safety of women (Stalans and Finn, 2000). And yet, when police enforce the law – something activists have repeatedly demanded – if it was done against the victim’s wishes then the police are accused of undermining the victim’s autonomy (Bailey, 2010).
Summarizing criticism of police, Berk and Loseke (1980) note, ‘The message behind the criticisms of police practice in domestic violence disturbances is clear: Police are not doing their job.’ In light of the NILECJ and NIJ findings, criticism of the police might be more accurate if it was directed toward minimal effort officers rather than the entire group. The findings of the NILECJ and NIJ suggest that justice for victims is a craps shoot – will a supercop will be assigned to investigate a particular DV incident, thereby increasing the likelihood of prosecution and conviction, and thus justice for the victim – or not?
Often a victim’s statement is sufficient to justify the arrest of a suspect and the allegation of criminal wrongdoing in a written report. Once probable cause has been established, the decision to employ additional police investigative techniques represents extra work. The NILECJ and NIJ results infer that ordinary police officers avoid extra work; whereas, supercops take it on. What is that extra work? The larger task of this article is to examine six of those optional tasks. Exercising discretion, an officer who investigates a DV crime may choose to use some or all of them – or not.
It is more work to individually interview each child living in the home. Yet, doing so may reveal not only more detail regarding the current incident, but also recent, as-of-yet undetected, additional chargeable crimes. For example, children may disclose one or both parents are habitually drunk in their presence, which is a crime in California; or perhaps that one parent feloniously assaulted the other five days ago, breaking a pool stick in the process. Perhaps the children retrieve the broken pool stick from a garbage can outside, and then the officer can seize it as evidence. It is also extra work to secure an emergency protective order, and to serve it; to obtain photographs now, and again in three days when bruises have fully formed; to track down the defendant for an arrest if (s)he fled the scene prior to the arrival of police; to complete an investigation the same day rather than passing it off to detectives to finish; and to write a report before going home rather than holding it over until the beginning of the next work shift.
Arrest, which is one of the optional investigative techniques under police control, seems to have become a proxy that the academy equates with a sufficiently conducted police investigation (Stanko, 1981 and personal correspondence, 2011–2012). In a sense, arrest has been used as a dichotomous outcome. If an arrest was made, a sufficient police DV investigation is assumed. This article challenges that assumption, suggesting that arrest is only a single member of a much larger constellation of police controlled antecedent actions that together can substantially influence prosecution and conviction. Further, this article challenges the assumption of uniformity of police action, instead suggesting that individual officers can be classified on a continuum of discretion whose poles are marked by minimalism and superwork, those officers being located according to their routine, individual work habits. This article proposes that rates of use of six police controlled antecedents (PCAs) can be employed as classification data for quantitative monitoring of individual officers, and for between-officer and between-agency performance comparisons.
For the most part, the PCAs have been ignored by academics, some entirely. Instead, domestic violence research has tended to focus on explanatory variables such as race, sex, age, socioeconomic level, education, culture, and so on. Like arrest, recidivism has also been studied extensively. Additionally, two dependent variables have largely been ignored: the decision to prosecute, and also criminal conviction. Both are important because they are, after all, the gateway to mandated treatment, monitored behavior, and greater protection for children exposed to domestic violence – following criminal conviction.
The specific purpose of this article is to assess the six PCAs for their relationship to prosecution and conviction. The reason PCAs are referred to in causal terms is because the uni-directional linearity of the criminal justice process permits such a claim. Police work always precedes charging decisions, which in turn always precedes any findings or pleas of guilt, which in turn precede any intervention of the State to punish and rehabilitate the offender(s), and to protect children.
Review of Police Controlled Options
Generally speaking, the prosecutor’s office is handed a finished police investigation that is memorialized in a written report(s). Though sometimes prosecutors utilize investigators to develop more evidence or gather additional witness statements, typically they work with the evidence produced by police. It stands to reason that the variables most likely to influence the decision to prosecute, and also conviction are physically and temporally nearer to the domestic violence event, and therefore lie within the domain of the police investigation. The evidence is fresh, witnesses are still at the scene, investigative opportunities are the greatest, and the willingness of victims, suspects, and witnesses to cooperate is most likely. Therefore, arguably, an effort to increase rates of prosecution and conviction should focus on variables that are present in situ, which can be operationalized by police.
Arrest
This variable has been studied extensively, and is known to have a strong correlation with prosecution and conviction. In an assessment of, among other things, that relationship Garner and Maxwell (2009:55, Table 5) meta-analytically reviewed more than 100 reports of prosecution and conviction rates in the United States since 1995, finding a mean prosecution rate of 30 percent, climbing to 67 percent if an arrest was made, and a conviction rate of 13 percent for all reported offenses, or 35 percent if an arrest was made. These data suggest that arrest has a positive influence upon prosecution and conviction; however, it is possible arrest is a marker of more serious DV cases that would be more likely to be prosecuted and convicted regardless of whether the suspect is taken into custody or not.
Additional witnesses
Testimonial evidence makes or breaks a case. In the absence of solid testimonial evidence, prosecutors are rarely able to convict (NILECJ, 1977: 78–79). How a police officer interacts with a witness is important in setting them up to cooperate with the prosecution. The National Institute of Justice (1982: 14) notes:
The greatest opportunity for obtaining information about an offense exists immediately after the offense has occurred – before witnesses have an opportunity to disappear or to forget. Thus, the police play a vital role in seeing to it that witnesses are located and their cooperation obtained.
In a study of N = 17,534 felony and serious misdemeanor cases, the National Institute of Law Enforcement and Criminal Justice (1977) determined that the conviction rate for cases with two or more witnesses was 35–40 percent, compared to 15–18 percent for cases with zero or one witness (27, Exhibit 3.7). The type of witness is also important. In the same study prosecutors rejected 49 percent of the cases where witnesses were related to the suspect, versus 21 percent if the witness was a stranger. In cases with two or more witnesses, the rejection rate was 30 percent for family, and only 12 percent for strangers (28, Exhibit 3.9).
Photographs
Photographs record the appearance of phenomena as they exist in a unique point in time. They help quantify three dimensional relationships, and they capture details in an instant, permitting greater analysis later when more time is afforded. Because it takes up to 72 hours for bruises to fully form (Langlois and Gresham, 1991; Laposata, 2011: 306; Spitz, 1993: 201–202), a second set of photographs should be ordered for the victim(s) three days after the incident.
Equally important is what photographs do not show, for example, the absence of injuries and bruises, and so on. This will be particularly salient when comparing the two sets of photographs, because the first set may not show injuries that later become visible, seen in the second set. Also, two sets of negative photographs may have exculpatory benefit, particularly when the victim’s statement is questionable.
Only a few studies have examined the relationship between photographs, prosecution, and conviction. Cramer (1999) randomly selected 140 cases from Chesterfield County, Virginia, for analysis of multiple explanatory variables. She reports a significant relationship between photographs and guilt; however, the claim is not quantified. Generalization of the finding is probably not warranted because only n = 10 cases included photographs and there is no indication that other variables were controlled in a manner that permitted sufficient analysis of the explanatory variable.
Another study also reports a higher rate of charging and conviction for cases with photographs (Hester and Westmarland, 2005: 66–67). However, like Cramer, there are a number of problems with the study. First, only percentages are reported without any further statistical analysis; thus, their significance has not been established. Second, apparently there was no attempt to employ case matching or other safeguards against error; thus, it is possible outcome differences merely reflect the likelihood that more serious cases will be accompanied by photographic evidence when presented to prosecutors.
Garcia (2003: 585, Table 3) reports the odds of being found guilty increase 354 percent (p < .02), and of pleading guilty by 514 percent (p < .01) 3 when photographs are taken. Her results are impressive; however, there are several problems with study design. First, officers in the treatment group were not randomly selected; rather, they were volunteers. Second, cases were not randomly selected. Third, treatment group officers were given advanced domestic violence investigative training, including a class to develop their photography skills. There is no indication control group officers were given the same training, minus the photography component. Fourth, additional analysis of Garcia’s data reveals four statistically significant differences between treatment and control group means, meaning they differ on multiple factors that were not controlled for. They are:
Victims investigated by the treatment group required medical attention more often. 4
Suspects investigated by the treatment group had more significant criminal histories. 5
Misdemeanor charges were filed more often in control cases. 6
Felony charges were filed more often in treatment cases. 7
The effect that is reported may be partially or even completely due to differences in case handling by officers who volunteered to be in the treatment group. It is also possible the reported effect is due to the training class, sans the photographic component. It is also possible that any or all of the four population differences may be responsible for outcome differences. Because of design flaws, Garcia’s results should be considered with caution.
Only one well-designed study measuring the relationship between photography and criminal conviction was located. Belknap et al. (2000) assessed N = 2670 court cases. Of those n = 824 resulted in pleas or findings of guilt. Logistic regression on photography showed β = −0.140, SE = 0.224, with an insignificant p value. Thus, there is a disagreement between the findings reported by Garcia and those of Belknap et al.
Listing additional criminal charges
In many circumstances, particularly those involving minor DV, after a police officer has developed sufficient probable cause to allege a single domestic violence crime, lawfully (s)he can discontinue further investigation. It takes more time to develop extra charges. If (s)he chooses to perform the extra work, finding additional crimes, it naturally follows that the prosecutor will therefore be able to file more criminal charges. Filing multiple charges may damage the ‘halo effect’ of the accused. That is the tendency of jurors to give single charge defendants more benefit of doubt; that is, to be more skeptical regarding their guilt. In contrast, jurors are harsher towards multiply charged defendants, seeing them as less likable, less believable, and more dangerous (Greene and Loftus, 1985).
Another reason to file multiple charges is their usefulness during plea bargaining. The defendant might agree to plead guilty to one charge if prosecutors agree to drop another. As Cammack and Garland note (2001: 4–5; see also LaFave, Israel, King and Kerr 2009: 190),
Charging decisions are also sometimes … strategic considerations … a prosecutor may file more counts or more serious charges than she deems appropriate in anticipation that the eventual conviction will be for a reduced charge arrived at through a negotiated plea agreement.
Charging a defendant with more than one crime is correlated to increased rates of conviction. In a study of N = 375,720 federal defendants, Leipold and Abbasi (2006, Table F) compared, among other things, the conviction rate of individuals facing a single charge and those charged with multiple crimes. The results demonstrated that a defendant’s chances of conviction increase by more than 27 percent if multiple charges are filed.
Review of Unassessed, Police Controlled Antecedents
A search of the literature was unable to locate research on two of the optional investigative actions, that is, obtaining an emergency order and rapidly completing the investigation, and their relationship to prosecution and conviction. A few studies that examine somewhat similar variables were located, and are reviewed below.
Emergency protective orders
These orders are distinguished from temporary, permanent, and civil protection orders by how they are obtained, and how long they are valid. In California, when court is closed, police officers telephone the ‘on call judge’ for their jurisdiction. Officers briefly summarize the facts of their investigation, and then request an emergency protective order (EPO). Based upon the officer’s factual summary, and answers to questions posed to the officer, the judge decides whether to grant the request. If the order is granted, the judge dictates the terms to the officer who places them on a standardized EPO form. Typically five items are included in a California EPO:
Move out immediately.
No contact with the victim.
Stay 100 yards away from the victim, family residence, victim’s job and school, and so on.
Award of temporary custody of children.
The date the order expires. 8
A copy of the EPO is given to the victim, and the order is entered into a secure criminal justice database; that makes it immediately available to other law enforcement agencies or courts nationwide. The EPO becomes enforceable once it has been served upon the defendant.
Martin (1994) assessed 23 variables for their relationship to the likelihood of domestic violence prosecution, one being the existence of a permanent DV restraining order (DVRO). Those orders are not the same thing as an EPO because EPOs expire quickly; whereas, permanent orders are in force for at least several years. Also, the presence of a DVRO automatically indicates a history of one or more prior DV incidents; whereas, an EPO contains no such marker. With regard to DVROs and prosecution for a new DV crime, Martin reports a positive correlation.
The importance of seeking EPOs by police, as part of developing a viable prosecutorial strategy, was recommended in an FBI bulletin (Wattendorf, 1996). Though the relationship between DVROs and outcomes such as recidivism have been examined (e.g. Kingsnorth, 2006), no quantitative study of the relationship of EPOs to prosecution and conviction are seen in the literature.
Rapid completion of police investigation
A report by the American Bar Association (2004) examined no-drop policies in several major US cities, reporting a significant increase in criminal case filing and conviction rates that they attributed to the refusal of the prosecutors to drop cases when the victims wished them to do so. In that study it was reported that in one city, San Diego, the police department made sure investigative reports were transmitted to the prosecutor within 48 hours for in-custody cases, and five days if no arrest was made. It is possible that some of the increased conviction rate attributed to a no-drop policy was actually due to the speedy submission of investigative reports by police.
The passage of time is an important variable in criminal justice; yet, it is rarely examined. Kurlychek et al. (2006) demonstrate the value of time as an explanatory variable in their study of arrest and recidivism, reporting an inverse relationship between the passage of time and the likelihood of re-offending; that is, the longer individuals go without being arrested again, the more likely they are to not be arrested in the future. Gretton et al. (2001) have employed survival analysis to assess re-offending risk over time for three groups of juvenile offenders, classified according to their scores on psychopathy assessment. The results showed each group has a distinct trajectory upon plotting of survival data. Survival analysis of the relationship between the period of police investigation, prosecution, and conviction has not been reported in the literature.
Hypothesis of Study
Having identified six antecedent police actions that may be selectively employed by a first responding police officer investigating an incident of domestic violence, and two outcomes of interest, prosecution and conviction, the following two hypothesis sets are offered:
A positive relationship will exist between the five non-temporal antecedents and both dependent variables in logistic regression models.
An inverse relationship will exist between the length of the investigative period and both dependent variables, the period beginning the day of the incident and ending with transmission of the police investigation to the prosecutor.
Method
The author was granted concurrent access to the records of a mid-sized California police agency and also the county prosecutor’s office, in order to gather archival data on every case of domestic violence for which a police report was written in the year 2007 (N = 1810). Both agencies divide DV cases into five categories, as described in Table 1. For logistical reasons those categories were utilized during data collection.
Distribution of investigations.
A random number generator was used to select 75 investigations from each of four subcategories. Since there were only 89 DVRO investigations all of them were used. A few investigations had to be discarded for technical reasons, leaving n = 371. Among those five temporal outliers were identified, where investigations remained open at the police department and were not transmitted to the prosecutor’s office for more than one year. Four of the five outliers were for periods less than 400 days; however, the fifth was 758 days. That single outlier more than doubles the scale of the survival time abscissa. In order to limit the period of the abscissa to one year, the five outliers were eliminated, leaving n = 366 investigations. A researcher should not eliminate inconvenient data without reporting the effect of doing so. Comparison of regression outcomes for both sets showed no significant difference; thus, the decision was taken to utilize the smaller set, sans temporal outliers, for all calculations.
All data were collected and coded by the author over the course of two-and-a-half years. Case numbers assigned by the police agency were used to reference the prosecutor database, thus insuring police investigations were properly matched to their prosecutorial counterpart. Data were entered directly into an Open Office spreadsheet. Sixteen (16) data were collected for all N = 1810 investigations. Two hundred forty-two (242) data were collected for the randomly selected subset (n = 366). Among those there are n = 213 dichotomous, and n = 19 continuous explanatory variables. Four types of nominal data were collected (including the case number), and there was one text box for researcher notes. Five types of dates were collected. Calendar days were converted to Julian format using the US Naval Observatory’s online converter. Julian format facilitates easy calculation of the period between two dates. In all, there are 88,572 cells in the password protected spreadsheet. A study of this depth was made possible by the fact that both agencies utilize electronic records keeping. Thus, problems such as having to track down a file in someone’s office, or in a court room, or in a separate storage facility were not encountered.
Data collection was limited to the first responding police officer’s written report. If other police officers assisted at the scene, or if detective follow-up was required later, data were not collected from their reports. The reason is because the focus of this study is an examination of the relationship between the work product of the first responding officer and important prosecutorial outcomes. Conviction was defined by one of three actions: pleading guilty; pleading nolo contendere; 9 or being found guilty at trial of one or more crimes. If a defendant was convicted of any crime it was recorded as a conviction even if, through plea bargaining or other means, the final convicted charge(s) were, or were not a domestic violence crime. That is consistent with a focus on the value of post-conviction sentencing, even if ultimately the charge(s) for which a DV defendant is convicted, or pleads to is not a DV charge. Such occurrences were found to be infrequent. Details will be reported in another article.
The five non-temporal PCAs were analyzed using logistic regression (logit). Logit is the appropriate choice because the two outcomes of interest are dichotomous, that is, the decision to prosecute, and whether the defendant was convicted. Further, logit is able to calculate outcomes from a mixed set of explanatory variables. In this study three of the explanatory variables are dichotomous (whether or not an EPO was obtained, photos were requested, or the defendant was arrested) and two are continuous (the number of witness statements, and the number of criminal charges that are listed in the responding officer’s report).
The sixth variable, the time period an investigation remains open at the police department before it is transmitted to the prosecutor’s office, was assessed using Kaplan-Meier survival analysis. The results were plotted on a graph whose abscissa is measured in days, and whose ordinate is the odds of a positive occurrence of the dependent variable. Inspection was used to interpret the survival graph’s significance. As noted by Hartwig and Dearing (1979: 12), visual inspection of graphs is a powerful analytical tool.
Calculating rates of occurrence for dichotomous outcomes is straightforward if one codes positive occurrences with the nominal number, one (1). Conveniently, it can serve dual purpose as a whole number when employing the rate algorithm, that is, n/N. A dichotomous dummy variable was created for each of the two continuous outcomes. The nominal one (1) was coded if one or more witnesses were located by police, or if more than one crime was alleged in the police report. Frequency rates could then be calculated for the continuous variables as well. This permitted Table 3 (see later) to provide frequencies that can then be directly compared.
All z score tests of means or ratios were of a one-tail design. Probability values were obtained using a z score/p level online calculator provided by WolframAlpha.com. The use of a calculator provides more finely nuanced p level results compared to some traditional printed tables.
Results and Analysis
Comparison of misdemeanor and felony results (seen in Table 2) demonstrates that felony investigations produce filed criminal cases at a significantly higher rate (z = 4.33, p < .00); also, the conviction rate for felonies is significantly greater as well (z = 2.89, p < .00), results that would reasonably be expected. Interestingly, though the case filing rate for DVROs is significantly higher compared to felonies (z = 1.54, p < .06), the conviction rate is nearly identical, suggesting that perhaps some weak DVRO cases are never-the-less filed, perhaps to send a message – via the inconvenience and expense that attends responding to criminal charges – to obey court orders.
Case filing and conviction rates.
In their meta-analysis of domestic violence prosecution and conviction rates, Garner and Maxwell (2009: 55, Tables 5 & 6) report a 30 percent rate of prosecution per reported offense in the USA since 1995, and an average rate of 46 percent for Australia, Canada, Switzerland, and the UK combined. The US conviction rate is 13 percent, and for the other countries 18 percent. Finally, the rate of conviction per prosecution is 50 percent in the USA and 54 percent in the other countries. Data from the present study show higher rates in all three categories, thus demonstrating the mandate of this prosecutor’s office to pursue DV crime aggressively.
Table 3 compares frequency rates between the six PCAs for all three types of DV crime. In the case of misdemeanor versus felony, rates for two antecedents are similar, and are different for the other four. The first similar rate is that of arrest, something that might be expected since the police department under study has a mandatory arrest policy. These data suggest the decision to arrest is taken at a steady rate; however, what is not revealed is the actual number of arrests that could have been effected but were not. For example, a certain percent of the suspects who fled prior to arrival of police could probably be located with a bit of effort.
Frequency of use: Police controlled antecedents.
Notes: aThe presence or absence of an EPO is used by both police and prosecutors to subdivide both misdemeanor and felony investigations/cases. Those same categories were used during data collection; therefore, the frequency of EPO usage cannot be determined from the randomly selected data. Accordingly, the frequency of usage was determined from the full data set, which is comprised of n = 1100 Misdemeanor, n = 621 Felony, and n = 89 DVRO investigations.
From the random sample: n = 137 Misdemeanor, n = 142 Felony, and n = 87 DVRO.
An EPO is not sought when a DVRO is already in place.
The other similar rate, that of identifying and interviewing witnesses, may provide an example of the routine exercise of police discretion. Unlike arrest, which is more closely regulated by mandatory policy, including additional witnesses in the police report is more susceptible to subversion, if the officer is inclined to do so. That is true because in many circumstances the failure to mention witnesses at the scene is not so easily detected. After all, normally a second party is not present taking notes of incidentals that an officer may ignore. That is why a study by Cross and Newbold (2010) is so interesting – they placed observers in police cars, then later compared field notes with dispatch center records, and also DV reporting forms that officers were required to fill out. An excerpt from their article exemplifies how the duty to locate and interview witnesses can be subverted:
Field observations yielded two plain examples of abuse of process. In one case, Communications reported a woman being dragged down the street. When the police arrived, the offender had fled but the victim displayed dishevelment and minor injuries consistent with having been dragged. There were also a number of witnesses. However, when the senior attending officer filled in the POL 400, he stated specifically that there were no injuries and no witnesses present. (Cross and Newbold, 2010: 68, emphasis added)
Thus, the rate of witness identification in the report of the first responding police officer seen in Table 3 may not be an actual measure of the average number of cases where one or more witnesses could be identified and interviewed, but rather the average between two subsets of officers: those who routinely find and interview witnesses, and those who do so less often.
One dissimilarity is the rate at which police officers seek emergency protective orders. The results show they are sought significantly more often in felony cases (z = 9.87, p < .000, n = 1100 misdemeanors, n = 621 felonies). Such is reasonably expected since felony crimes have visible injuries and are more horrific; thus, it is possible that EPOs serve as markers for more serious and injurious DV crime. That possibility then raises a question regarding the presence of EPOs on misdemeanor cases.
It is possible some police officers routinely seek EPOs on misdemeanor cases in an effort to increase the odds of prosecution and conviction, more than to protect a victim from further risk. After all, an EPO is a temporary order issued by a judge after a cursory review of the facts of the case; thus, it is not to be easily dismissed from consideration. Quite possibly the presence of an EPO may pressure a reviewing prosecutor to file charges on a marginally weak police investigation that (s)he might otherwise decline; and, at some level the presence of an EPO may also pressure defendants to plead guilty. These possibilities warrant further study.
Photographs in felony investigations are requested more than twice as often compared to misdemeanors, something one would reasonably expect because misdemeanor investigations generally have few or no visible injuries. 10 This finding suggests that officers may fail to recognize the full importance of pictures beyond preservation of visible injuries, because if they did, photographs would probably be requested more often, without regard to severity of crime. The scene, the emotions, the presence of evidence not recognized at the time but later discerned from review of photographs are but a few reasons to routinely photograph all domestic violence crime scenes, and also the involved parties including children and witnesses.
The number of additional criminal charges alleged in the police report is significantly greater in felony cases (z = 3.30, p < .000). One would reasonably expect this finding because attendant to the more serious injuries will likely be additional, obvious crimes, that is, a broken mirror, the presence of children who are lawful victims by having witnessed domestic violence, threatening voice mail or text messages, telephones damaged so police could not be summoned, and so on. What is not distinguished by these data, though it is most likely conflated within them, is differences between officers who routinely dig for additional crimes, and those who merely act upon the obvious but little more.
The results also show felony investigations are not finished on the same day as often as misdemeanors (z = 2.31, p < .01). One reason first responding police officers turn an unfinished domestic violence investigation over to detectives is to have them locate and interview witnesses who could not be reached. Other reasons could include the need to re-interview victims or witnesses in order to obtain a more detailed, recorded or even videotaped statement, or to have a specially trained forensic interviewer question children. However, not finishing an investigation right away can have devastating consequences upon prosecution and conviction, something discussed in more detail later in this article.
Strong case investigative model
A best practices model for the investigation of domestic violence by first responding police officers is proposed. That model contemplates police will make a good faith/sufficient effort to operationalize each of the six PCAs each time they investigate a DV crime. The model was tested by subjecting five of the PCAs to logistic regression, and the sixth to survival analysis. The results are seen in Tables 4 and 5, and in Figures 1 and 2 (see later).
Odds of prosecution: Testing the five antecedent model across multiple data subsets.
Key:
χ2 = 63.10, p < .0000, Pseudo R2 = .1416.
χ2 = 44.66, p < .0000, Pseudo R2 = .1243.
DVROs are not included in this subcategory.
χ2 = 42.36, p < .0000, Pseudo R2 = .2237.
χ2 = 11.41, p< .0438, Pseudo R2 = .0753.
An EPO is not issued because a DVRO is already in place.
Odds of conviction: Testing the five antecedent model across multiple data subsets.
Key:
χ2 = 57.96, p < .0000, Pseudo R2 = 1145.
χ2 = .42.63, p < .0000, Pseudo R2 = .1114.
DVROs are not included in this subcategory.
χ2 = 19.56, p < .0015, Pseudo R2 = .1110.
χ2 = 23.32, p < .0003, Pseudo R2 = .1189.
χ2 = 12.18, p < .0161, Pseudo R2 = .1021.
An EPO is not issued because a DVRO is already in place.
Percentage likelihood the prosecutor will file criminal charges.
Percentage likelihood the defendant will plead guilty, nolo contendere, or be convicted.
Listing more than one criminal charge has the greatest effect upon both outcomes, increasing the odds of prosecution by 284 percent and conviction by 142 percent. That makes sense when one considers the fact that in the United States about 95 percent of all criminal cases in local and state courts (Sourcebook of Criminal Justice Statistics, 2002, 2004), and 97.5 percent of all federal prosecutions(US Courts, 2011, Table D-4) 11 are resolved through pleas and not trial. One would reasonably expect police reports that articulate the facts of multiple crimes to be more attractive to prosecutors because they include ‘throw away’ extras, that is, charges that can be dropped in exchange for a guilty plea to a primary offense. When a single charge is listed, such negotiation is not possible. 12
Multiple charges are seen to substantially increase the odds of prosecution for misdemeanor and felony investigations, with the risk of erroneous interpretation increasing somewhat for felony and DVRO cases (p < .092 and p < .110, respectively). That makes sense because felony charges, and charges of a DVRO violation are automatically filed more often, so the influence of multiple charges would be lower. The influence of multiple charges is strongly seen for all five crime categories with regard to conviction. That also makes sense because, as previously described, multiple charges provide ‘throw aways’ to be traded for a guilty plea.
Emergency protective orders exert a similar influence on both dependent measures, increasing the overall odds of prosecution and conviction across categories, with the greatest effect seen in misdemeanor case filing. Whereas felony investigations may demand prosecution based upon photographs, witnesses, charges, arrest, and timeliness of delivery, misdemeanors are less likely to be filed. Recalling little or no visible injury exists in misdemeanor cases, and also that only 14 percent of misdemeanor cases have multiple charges listed, and also that witnesses are identified in only 30 percent of them, the presence of an EPO may serve to provide the nudge needed to reach the threshold of prosecutorial case acceptance.
Interestingly, there is little relationship between EPOs and the odds of felony case prosecution. That may be true because if EPOs serve as a marker of seriousness of an investigation, in the case of felony DV crime EPOs may be redundant to the level of the charge and thus not be needed to suggest seriousness of the offense.
Arrest about doubles the likelihood of case filing by the prosecutor, a result that is similar to the meta-analytical average rate reported by Garner and Maxwell (2009). Their analysis shows about a 127 percent prosecution rate increase when the defendant is arrested (2009: 55, Table 5). 13 The relationship between arrest and prosecution might be somewhat explained by the defendant’s procedural rights. In California an arrested person must be arraigned within 48 hours, arraignment being the point at which prosecution is initiated or the case is dropped. Investigations where the defendant(s) is not arrested are not subject to the same time pressure and may therefore languish while prosecutors handle matters foisted upon them through the act of arresting a suspect.
The relationship between arrest and prosecution holds true for misdemeanor and felony DV cases, but not DVRO. That is likely true because DVRO represents violation of a permanent court order, a challenge to the authority of the court, and is a display of contempt. Thus one expects DVRO cases to be filed.
Interestingly, arrest has no influence upon the likelihood of conviction in misdemeanor cases (p < .156). That may be true because misdemeanor suspects are generally released on bail after a six to eight-hour cooling off period; thus, the value of arrest may not be the act of taking a suspect into custody but rather an elongated custodial period following arrest, which misdemeanants do not typically experience. In the case of felony charges, the time period is probably longer because felony bail is much higher. One would reasonably expect a lower rate of bailing out of jail for suspects booked on felony charges. It is possible that attending court appearances such as a preliminary hearing, while in custody, sometimes in jail clothing and wearing handcuffs and perhaps leg irons – compared to street clothing worn by misdemeanor suspects freed on bail – may in some way have influence upon the likelihood of felony conviction. This possibility should be explored further.
The increase in conviction rate for arrested persons is 78 percent. That is about half of the 169 percent 14 increase that can be calculated from Garner and Maxwell’s (2009) meta-analytical data. The difference between rates may reflect differences in case selection practices, that is, an aggressive DV prosecution policy that pursues all types of cases, including those that have a lower probability of prosecutorial success would be expected to produce a lower conviction rate compared to other prosecutor’s offices that may ‘cherry pick’ cases most likely to result in a conviction. The former type would have a high prosecution rate, and the latter a high conviction rate. It is possible that Garner and Maxwell’s data reflect the routine practice of cherry picking of DV cases by prosecutors in the United States. The meta-analytically derived increase in conviction rate for prosecution of arrested DV suspects in Australia, Canada, Switzerland, and the UK, for cases where an arrest is made is 45 percent. 15 That rate is fairly close to the one calculated in the present study.
Finding and interviewing witnesses also substantially increases case filing though it has no measurable influence upon conviction. The effect is seen to only hold for misdemeanor cases. Recalling that only 14 percent of misdemeanor cases have multiple charges listed, in the absence of expendable charges that prosecutors can dismiss in exchange for a guilty plea to a primary crime, prosecutors may never-the-less be swayed to file a criminal case on a single misdemeanor charge when witnesses are identified in the police report. The diminished value of witnesses over time may explain the lack of relationship between witnesses and the odds of conviction. Witnesses move, memory decays, motivation diminishes, and the risk of intimidation may increase in the time period following filing of criminal charges. The lack of influence of witnesses upon the odds of prosecution suggests that justice delayed works in favor of defendants.
Photographs increase case filing by an additional 60 percent for combined misdemeanor and felony cases, at a marginal probability level (p < .09). Thus, it is not surprising that for the smaller subsets of misdemeanors alone, or felonies alone, no significance is obtained. Also, photographs have no significant relationship to the odds of conviction. These results replicate the finding of Belknap et al. (2000), and cast further doubt upon the effect reported by Garcia (2003).
At first the finding of no relationship between photographs and conviction might seem peculiar until one considers the value of photographs, that being to show a jury the injuries of the victim(s), the chaos of the scene, the presence of children, and so on. The need for photographs in a courtroom trial are minimal because most cases are resolved through plea bargaining; thus the lack of influence upon conviction is to be expected. If rates of jury trial increase, the value of photographs should increase as well.
The graph in Figure 1 demonstrates an inverse relationship between the amount of time an investigation is open at the police department and the odds of prosecution. The decay or probability is rapid, with a 25 percent reduction in just a few days, and 50 percent in less than a month. This finding strongly suggests that first responding police officers should complete their entire investigation the same day, or as soon as possible, and not pass it off to detectives for follow-up.
A similar decay function is seen in Figure 2, indicating that the likelihood of a guilty plea, or conviction drops off rapidly if the completion of the investigation is delayed, even for just a few days. Individual regression of a series of dichotomous dummy variables created to measure various time periods, in days, found that the period of greatest opportunity for prosecution exists when the written report of the first responding police officer is transmitted to prosecutors in 10 days or less (χ2 = 8.08, p < .0045, β = .668, z = 2.80, p < .01, odds = 1.95, a 95 percent increase); however, regressed within a model containing the other five PCAs the period of the first 10 days (i.e. first 240 hours, aka T240) does not attain statistical significance because of collinearity. That result tells us time is an important and integral part of all five PCAs; that is, they should all be carried out rapidly. A similar result is seen when the likelihood of conviction is logistically regressed on the five PCAs and T240 (χ2 = 11.24, p < .0008, β = .712, z = 3.33, p < .00, odds = 2.04, a 104 percent increase).
Evaluation of study hypothesis
A strong positive relationship was hypothesized between the five non-temporal PCAs and both outcomes when assessed in regression models. With regard to case filing, the prediction was validated; however, two of the PCAs show no relationship with conviction: witnesses and photographs. The other three non-temporal PCAs showed significant relationship with conviction, as predicted. Thus, with regard to the first model, the hypothesis is entirely accepted. For the second model it is partially accepted for those variables that are shown to be significant.
A strong inverse relationship was hypothesized between the length of time an investigation remains open at the police department and both outcomes. Both predictions were demonstrated to be valid and thus the alternative hypothesis is accepted.
Discussion
It was not the purpose of this study to test an a priori proposition, that is, a self-evident truth that more thorough police investigations would be expected to positively correlate to increased criminal case filing and conviction. Kerlinger (1973: 5–6) reminds us that what is assumed to be self-evident by one a priorist is not to another, whose reasoning and assumptions may be quite different. Thus, what scholars believe to be true about domestic violence, and also the investigation of it by police should be subjected to inquiry and quantitative falsification. More than once an academic belief has been repudiated by scientific findings. Thus, in fact, this was a study that justifies some tentative a posteriori conclusions.
One could question whether these data were selected on a dependent variable, being those incidents of DV that came to the attention of the police. Such would only be a concern if the results were generalized to all domestic violence incidents, including those that do not come to the attention of the police; however, such is not the case. This is not a study of domestic violence, per se, but rather the manner in which police investigation influences the prosecution of it. Thus, generalization is limited to police investigation of domestic violence. The two dichotomous outcomes examined in this study are case filing and conviction; neither were used as a selection criterion. Therefore, variance in outcomes can be compared to explanatory variables and inferences of relationship can be made, so long as they are otherwise statistically and methodologically warranted (King et al., 1994: 129–130).
This study assessed six police controlled antecedents for their power in a strong case investigative model. A work of this type has not been published before. It is hoped the present effort will draw attention to these and other police controlled antecedents, and that they will all be studied further. More attention should be focused on the choices made by police officers who are called to the scene of a domestic violence incident, because as this study demonstrates, those choices can have a substantial impact upon the mechanics of prosecutorial justice.
A trickier issue to discuss is the ongoing argument about discretion versus mandates. Though the professional capability, initiative, education, discretion perspective is appreciated, and it may be an important means to accomplish some change in police behavior, the view of this author is different. It is shaped by observations that were made while working as a police officer. Generally, two types of fellows were observed with regard to how thoroughly domestic violence incidents were investigated: Those who routinely perform excellent work, and the much larger group – those who routinely perform minimal work. These observations are consistent with the two officer categories described by the NIJ (1982) and NILECJ (1977) studies.
The author has, as a police officer, witnessed attempts to effect change in police practice via the education route, either through less formal ‘roll call’ classes, or scheduled bi-annual ‘advanced officer training’ seminars. Three examples pertinent to domestic violence come to mind. In each case officers were educated about a DV related matter, but mostly failed to make any changes to their investigative routines until later when, individually, discretionary use was replaced by an organizational mandate. Only then did rates of DV arrest, provision of an information card to DV victims, and completion of a standardized DV information reporting form substantially increase. Mandates compelled compliance where education and discretion had failed.
The proposition that greater training may improve professional willingness to employ new measures such as operationalization of the PCAs seems to assume a deficiency in police capability that must be overcome through training. However, such is not the case with regard to the six PCAs. That is because rookie police officers are not matriculated out of training status until they have repeatedly demonstrated the ability to locate and interview witnesses, deeply investigate for multiple charges, locate and arrest suspects, call a judge and secure an EPO, and either take photographs or request a technician to take them, and to write a comprehensive written report documenting a thorough investigation. Thus, any matriculated police officer is experienced in deployment of all of the PCAs and further training is not needed. Besides, with the exception of EPOs, the other five PCAs are skills that are operationalized on a daily basis across a wide variety of crimes; thus, they are well practiced, and are intimately familiar to each police officer.
Insufficiency of understanding is not the issue and training is not the remedy. In essence, the argument presented in this article is that there is nothing special about the skills and capabilities of ‘supercops’, other than those individual officers are routinely thorough. Any cop can be a supercop with regard to DV investigation – all that is needed is the willingness to perform above minimum requirements. Further, the willingness to be thorough is a universal skill not limited to the United States borders, or to US-style policing. That is true because the key component to thorough policing, whether in a western metropolis, on an indigenous reservation, in a theocracy, or even in lawless lands, is not defined by the tools at hand but by the mind and heart of the individual police officer. A deep sense of justice, unbridled curiosity, innovative tenacity, and also some bravery – coupled together with a notebook and pencil, a camera, and a few paper bags in which to place evidence – these more than sufficiently equip any police officer, anywhere, to investigate any type of domestic violence crime. In the absence of a personal drive to investigative excellence, mandates are believed to be the only means by which routine minimalists can be compelled to sufficiently deploy all of their professional skills each and every time they investigate a domestic violence incident.
Performance incentives might encourage routinely low effort officers to choose to employ the PCAs more often; however and unfortunately, in the public sector incentive based performance is not really possible. A rewards based management style that offers incentives in an attempt to improve achievement may work in private industry, but it is inconsistent with the expectations of citizens who believe excellence is the sworn duty of police officers and that no extra reward beyond a paycheck is necessary or appropriate.
The benefits of employment of the six PCAs will not be realized if minimalist officers are not compelled to employ them, or to explain in their written report why one or more could not be operationalized. There is a historical fact that suggests this proposition may be true. Only with legal mandates did police officers begin to investigate domestic violence and make arrests. Before the 1980s that was not done much. A slew of new laws and rules in the 1980s and 1990s changed all that. Then as now, education failed and mandates became necessary.
Suppose the PCAs were mandated – would an over-use syndrome develop? That is, would their effectiveness be diminished in proportion to their increased use? The answer is already known for one of them. Arrest has been mandated for years, and yet it continues to strongly correlate with odds of prosecution and conviction. Garner and Maxwell’s (2009: 55, Table 5) meta-analytical comparison of pre-1995 rates with those in 1995 and later, when mandates were in place in most of the USA, shows prosecutions for arrested DV suspects increased from 53 percent to 67 percent, with convictions increasing from 31 percent to 35 percent. In other words, in the era when mandatory arrest was implemented an over-use phenomenon did not emerge – in fact, just the opposite occurred: The rate of prosecution increased. What is unknown is whether an over-use fatigue will be noted, over time, for any of the other PCAs after they have been mandated for a while. That remains to be seen.
If one is to err, it seems better to err on the side of victims by attempting to raise embarrassingly low rates of prosecution and conviction – both in the USA and abroad. Though concerns about discretion and professionalism and over-use syndromes are all very important, it is argued here that the comforts and freedoms of police officers should be subordinated to the best interests of victims, especially child victims, and also to justice.
Categorically discretion contemplates between-officer differences; yet, generally the performance of police officers is treated as a constant variable in most DV studies, resulting in a rather one dimensional view of police work. Also, perhaps too much so, victim or suspect characteristics are generally the explanatory variables that are studied. Together then, these foci have left a gaping hole in the literature because insufficient attention has been paid to individual officer routines, between officer differences, and to the many investigative options that each officer is capable of deploying when called to investigate a DV incident. This article is an attempt to re-focus domestic violence research attention away from demographics and theoretical arguments and instead direct them toward police related variables – particularly those that can be manipulated by officer decisions and organizational mandates.
What has been left dangling in this article is the need to quantitatively partition individual officers according to their routine levels of effort, something that can be done utilizing the PCAs, and then compare their rates of criminal case filing, and criminal conviction. Including these data and their analysis, along with the requisite literature review would double the size of this article. Those tests have been run and as one might anticipate, significant differences were indeed found; in due course they will be presented in another article that is currently being written.
There are limits to the applicability of the present work. First, there are thousands of prosecutorial agencies spread among the more than 3000 counties within the 50 US states; each operating under different state and local laws, rules, priorities, budgets, and theoretical understandings about domestic violence. Additional limitations exist when generalizing to an international audience. The results and recommendations from the present study may have a fairly close fit to many locales, but, undoubtedly there will also be some variance.
This article is being published in a criminology and criminal justice journal with an international audience. Some readers may be in countries with different approaches to criminal investigation and prosecution, for example, places with judge/prosecutor offices, or perhaps where justice is dispensed more rapidly according to local custom. Never-the-less, violence is a human condition and the fundamentals of investigation and proof are fairly universal – what differs are matters such as who conducts the investigation, the weight given to eyewitness testimony, various types of evidence, and the rules of admissibility, legal process, and judgment. Thus, it is believed that the six fundamental police actions evaluated in this article should have fairly widespread, transnational application.
Second, the sample size was not small but neither was it large. In part that is due to the demanding nature of thoroughly reading and coding police reports and prosecution files on multiple occasions over several years. Thus, as is true for many researchers, use of random sampling was necessary. Generalization of sample findings to the population from which they were drawn is permitted when the sample was randomly selected, is of a sufficient minimum size, and when proper statistical analysis is employed. Those were all done.
Third, there is the potential for authorial bias. That is true because the author is a former police officer who specialized in domestic violence investigation. It is felt that potential authorial bias was naturally kept to a minimum due to the nature of quantitative data, which does not need to be interpreted before it is classified and recorded, for example, either the suspect is male or female, or yes/no there was a charge of vandalism accompanying the main DV charge, and so on.
Lastly, when new phenomena are described for the first time they should be subjected to rigorous falsification efforts by others in order to determine their replicability, and also their rates of occurrence under other circumstances in different locations. Until that is done, these data and conclusions should be viewed with appropriate caution.
These results have policy implications. Police agencies may wish to consider adopting, and states and nations mandating the strong case investigative model that is validated by these findings. Requiring first responding police officers to employ the six PCAs, or to explain in their report why one or more were not used should produce better quality investigations, more witnesses, more charges, and elevated rates of criminal case filing, and convictions. 16 By keeping comparative records supervisors can identify officers who routinely make low use of PCAs so that they can be remediated, disciplined, or if need be – they can be discharged from their office as unfit.