Abstract
In the absence of empirical knowledge, public expectations of criminal investigations have been largely informed by fictional and/or dramatised depictions of detective work. To address this issue, incident report data from 243 retrospective burglary and robbery cases were paired with self-report surveys from 40 detectives who indicated the activities and the date they performed them for each of the cases they were assigned. The results suggest that the temporal order of retrospective burglary and robbery criminal investigations is more nuanced than previously acknowledged. Policy implications and areas of future research are discussed from these findings.
Introduction
From novelised crime solvers to television networks dedicated to true crime programming, criminal investigations are an enormous part of popular culture. Unfortunately, literary and cinematic depictions of detective work are often fictional or dramatised for entertainment purposes. Though they are rarely empirically based, crime dramas have become the dominant representation of detective work and have distorted public expectations of criminal investigations (Reiner, 2008). As a consequence of this issue, the CSI effect 1 leads many citizens to believe that detectives can produce evidence, solve cases and obtain convictions for every crime reported to police (see, for example, Schweitzer and Saks, 2007). This belief is so widely held that prosecutors often seek forensic evidence testing, regardless of its probative value, in appeasement of jurors who have come to expect this type of evidence to be presented in court (Erickson, 2015).
The reality is that, while national crime rates have been declining (Federal Bureau of Investigation, 1995–2015), clearance rates in the United States remain lower than public expectations (Pew Research Center, 2017). Though agencies vary (Eck, 1983), in the last 20 years, less than one in four burglaries or robberies were cleared by police. This has come at a substantial cost to law enforcement legitimacy. Horvath et al. (2001), in a representative survey of 3,123 law enforcement agencies nationwide, suggested that clearance rates may have plateaued because detective work has remained relatively unchanged in the last 40 years. As a catalyst for change, research on detective work has been scarce. In fact, studies on criminal investigations, as an area of intrigue, have often been characterised by prior researchers as ‘limited’ (Horvath et al., 2001: 111), ‘neglected’ (Waegel, 1982: 452) and ‘outdated’ (Greenwood et al., 1977: 37). Furthermore, existing research on the temporal ordering of detective activities in criminal investigations tends to rest on modest methodological findings. Consequently, very little is known about how detective activities are ordered in criminal investigations.
The purpose of this study is to overcome the limits of prior research by exploring the ‘black box’ of the criminal investigations process. Forty burglary and robbery detectives with the Houston Police Department identified the cases they were assigned, the activities they performed and the date for their investigative efforts. This information was paired with incident report data from 243 cases assigned to the sampled detectives during a 30-day period of observation. Cases in these analyses were disaggregated by the offence, investigation day and the number of separate days the case was investigated. In doing so, this study seeks to understand the temporal arrangements of detective activities in burglary and robbery investigations and advance our knowledge of detective work for the explicit purpose of improving public safety. Prior to discussing the data, these analyses begin by situating this study in the extant literature on case processing, detective work and criminal investigations.
Background
Case processing in the criminal justice system is typically initiated following a citizen’s call for service, where emergency dispatchers refer cases in need of immediate police attention to patrol officers. Upon arriving on the scene, patrol officers have three priorities: (1) assess emergencies, (2) secure the area and (3) conduct a preliminary investigation (Orthmann and Hess, 2013). Prior research indicates that the quality of preliminary investigations greatly influences case processing outcomes (Glick and Riccio, 1979; Greenwood et al., 1977). Unfortunately, the range of activities that patrol officers are able to perform while at a scene is often dictated by an endless flow of calls for service (Eck, 1983). Eck (1983), for example, reported that patrol officers will spend approximately one hour investigating burglaries and robberies. As a result, many cases do not receive the required attention to be solved during preliminary investigations.
When a case is complex or needs greater attention, it is often referred to detective units for follow-up investigative consideration. Prior to a detective’s involvement, however, investigative administrators review and screen out cases based on formal (and, in some instances, informal) solvability criteria, such as the presence of known or named assailant in the preliminary report (Glick and Riccio, 1979). As an administrative tool, case screening diverts detective attention away from unsolvable cases in an effort to create more opportunities for them to dedicate their time to cases that can reasonably be cleared (Eck, 1983). Criminal cases, therefore, have been scrutinised for detective attention by a citizen, patrol officer and investigative administrator prior to being assigned to a detective.
Detective discretion and decision-making
Retrospective criminal investigations follow administrative case assignments and are self-directed by detectives who exercise discretion on if, when, and to what extent a case is worked after it is assigned (Eck, 1983; Greenwood et al., 1977). Furthermore, the autonomy of detective decision-making is unique because it is virtually invisible to internal and external forms of oversight (Waegel, 1981, 1982). On a day-to-day basis, for example, investigative administrators have little knowledge of how detectives are spending their time or the cases they are working (Corsianos, 2003; Pogrebin, 1976). Case progress is most often monitored through informal contact with detectives and, from a supervisory standpoint, detective accountability is primarily based on the timeliness and completeness of investigative reports (Horvath et al., 2001; Waegel, 1982).
The somewhat unfettered scope of detective discretion is balanced by large workloads. In a large national survey of investigative units, Ward (1971) found that detectives were assigned approximately 51 cases a month, which amounted to about one case every two ‘working’ hours. The number of cases worked, however, differs from the number of cases assigned. To this point, Greenwood et al. (1977) commented that ‘although most investigators will have twenty or thirty open cases on their desk at any time, only two or three are really considered active’ (1977: 10). The difference between the number of worked and assigned cases tends to be greater for property crimes, which occur and are assigned at a higher rate (Greenwood et al., 1977; Ward, 1971).
Heavy investigative workloads are often accompanied by delays in initiating retrospective criminal investigations. Fallik (2017), in his analysis of 537 cases that were assigned to 184 detectives with the Houston Police Department, reported that it takes between three and five days for detectives to investigate index crimes following their assignment. Cases with an identified suspect and those that did not end in an arrest were investigated more promptly, to which Fallik (2017) suggested that detectives may be delaying working more complex cases. Furthermore, investigative inaction may be indefinite for some cases. Among 5,647 referred crimes with the Kansas City (MO) Police Department over a one-year period, only 32% of cases were worked by detectives (Greenwood et al., 1977). More recently, Fallik (2016), in his analysis of 564 index crimes, found that 25% of assigned cases did not receive any investigative attention following their assignment during a one-month period of observation.
Additionally, cases that are worked by detectives do not often receive a lot of attention. In 82% of 1,688 burglary and larceny cases that were assigned in a medium-sized Midwestern police department, for example, detective effort was limited to reviewing a patrol officer’s preliminary report and making a few phone calls (Bynum et al., 1982). Similarly, Greenwood et al. (1977) found that 73% of cases were suspended one day after being assigned, 88% were suspended after a week and 98% of investigations were suspended one month after their assignment. Liederback et al. (2011), in their analysis of 351 criminal cases investigated by 21 detectives during a two-month period, suggested that cases are closed promptly because ‘detectives continue to be bogged down’ with ‘administrative tasks and paperwork’ (2011: 57, 59; see also Greenwood et al., 1977). Substantively, detectives tend to have little time to put forth effort in the cases they are assigned, at least not the kind of effort that could result in positive case outcomes.
Within the context of unfettered discretion and limited investigative time, deciding where and how detective efforts are prioritised has been deliberated by two scholars. The ‘skimming’ hypothesis, by Waegel (1981, 1982), is based on observations of a nameless city investigative division and is critical of detectives. It proposes that their decision-making is driven by potential case outcomes, whereby they only work on cases that require the least amount of effort and have the greatest chance of resulting in an arrest or prosecution. Waegel (1981, 1982) goes on to note that investigative administrators support caseload skimming ‘even though they recognise that it ensures that a majority of ordinary cases will never receive a thorough investigation’ (1981: 267–268). Predated support of these anecdotes comes from Greenwood et al. (1977), who found that the ‘cases that get cleared are primarily the easy ones to solve, and that most of a detective’s work is a consequence of the fact that an arrest has been made’ during a preliminary investigation (1977: 112). According to Waegel (1981, 1982), then, investigative outcomes are more attributable to case complexity, rather than the likelihood that an arrest or prosecution will be forthright.
In rebuke of Waegel (1981, 1982), John Eck and the Police Executive Research Forum more systematically observed detective work, how long it takes them to do it, and case processing results in DeKalb County, Georgia, St Petersburg, Florida and Wichita, Kansas. In each of these locations, Eck (1983) paired approximately 5,500 police activity logs with official crime report data from 3,360 burglaries and 320 robberies. Based on these records, several tables (see Tables 5-11 through 5-16 in Eck, 1983) are presented that outline the detective activities that were performed across separate days of burglary and robbery investigations. Tables 1 and 2 collapse these findings by offence and express how often – in the form of a proportion – individual investigative activities were performed on the first, second, third and subsequent days of the investigation. 2 Positive rate changes over time for each investigative activity are noted in these tables with a plus sign.
From these data, Eck (1983) reported that detectives often begin retrospective criminal investigations with several routine procedures, such as reading the account of the responding patrol officer and attempting to contact people mentioned in the preliminary report. He also observed that detective activities often decrease in routineness and become more unpredictable as burglary and robbery investigations progress. Substantively, Eck (1983) found that detective activities in the beginning tend to be victim-oriented and focused on sources of information outside the control of the agency. As they progress, however, activities shift toward being suspect-oriented and directed at sources that are under the control of the agency. This occurs, at least in part, due to the development and exhaustion of leads that prompt new investigative activities. Schroeder and White (2009), in their analysis of 593 homicides in Manhattan between 1996 and 2003, found support for this contention with regard to forensic evidence. They reported that DNA ‘may be a tool of last resort for homicide detectives, a tool that is used only when all else has failed’ (2009: 337).
In selecting where to prioritise their efforts, Eck (1983) concluded that detectives triage their existing caseload based on subjective assessments of cases into three groups: (1) cases that are already solved, (2) cases that will never be solved and (3) cases that might be solved if extensive effort by the detective was put forth. Cases from the first group receive the greatest amount of investigative attention, while cases that will never be solved are neglected altogether. Cases that could be solved but require greater investigative effort are also often deprioritised by detectives. This is problematic, according to Fallik (2016), because investigative outcomes among difficult-to-solve cases were found to be largely dependent on detective efforts. The ‘triage’ hypothesis, as it is known in the literature, differs from the skimming hypothesis because detective decision-making is guided by utilitarian ideals (as opposed to self-interest).
Empirical limits to the skimming and triage hypotheses
Unfortunately, Waegel’s (1981, 1982) and Eck’s (1983) observations of the temporal ordering of burglary and robbery investigations are based on somewhat weak methodological and analytical footing. Waegel’s (1981, 1982) contentions are based on non-systematic descriptions he collected during participant observations. The issues associated with Eck’s (1983) triage hypothesis are much subtler. First, varying lengths of burglary and robbery investigations are not accounted for in his analyses. This is problematic because the temporal patterns of shorter and longer investigations will likely vary and thus should have been analysed separately. Second, investigative effort that occurred on the third and subsequent days of an investigation should be parsed in the data. By aggregating these investigation days, Eck (1983) has confounded the effects of investigations that last longer than three days and inhibited direct comparisons of single-day estimates. Third, Eck’s (1983) analytical strategy for understanding burglaries and robberies does not take into consideration the stage that individual cases are at in the investigative process. The detective effort reported in Tables 1 and 2 are likely, therefore, confounded by detective effort directed at post-arrest case processing objectives like providing aid to prosecutors as they prepare cases for court. Since case processing stages will likely prompt different investigative activities, a more nuanced approach to understanding the temporal patterns of retrospective criminal investigations will hone in on detective effort that occurs in conjunction with specific investigative milestones.
Fourth, Eck’s (1983) data only loosely supports his conclusions about the temporal arrangements of burglary and robbery investigations. Eck (1983), for example, predicts that discussions with detectives (i.e., an activity under the control of the agency) would increase over time, but this pattern was not observed across the totality of burglary investigations in Wichita, Kansas (see Table 1). Additionally, none of the investigative activities increase across the course of robbery investigation in any of the jurisdictions under review (see Table 2). From Eck’s (1983) assertions, we would expect suspect interviews (i.e., a suspect-related activity) and checking departmental records (i.e., an activity under the control of the agency) to increase over time, though neither did in any of the observed jurisdictions.
Eck (1983) Tables 5-11 through 5-13: Proportion of the time detective activities were performed during the first, second, third, and subsequent days of retrospective burglary investigations.
Eck (1983) tables 5-14 through 5-16: Proportion of the time detective activities were performed during the first, second, third, and subsequent days of retrospective robbery investigations.
Fifth and perhaps most concerning is that looking at activity rate increases and decreases over time is an overly simplistic way of understanding the temporal patterns of retrospective criminal investigations. With three data points, as in Eck (1983), investigative activities may be performed at their greatest rate in the middle of an investigation and decline thereafter, in a downward-shaped parabola. In robbery investigations, collecting physical evidence demonstrates this effect in each of the jurisdictions evaluated by Eck (1983). Additionally, an activity may be performed at the greatest rate at the beginning and end of an investigation, in an upward-shaped parabola. Detective discussions with supervisors in robbery investigations demonstrate this pattern in each of the jurisdictions evaluated by Eck (1983) as well. Regrettably, none of these patterns were discussed in Eck (1983). Furthermore, with a greater number of data points, like in the current analyses, more patterns of criminal investigations may emerge.
Finally, over 30 years have passed since Waegel’s (1981, 1982) and Eck’s (1983) publications. While this, in itself, does not mean that their conclusions should be dismissed, it is important to note that the empirical status of these studies has not been replicated. This is primarily due to a shifting interest in how victim, suspect and case characteristics impact case attrition in the criminal justice system (e.g., Bouffard, 2000; Decker, 1993; Spohn and Spears, 1995); however, the contemporary discourse on criminal investigations often neglects the role of detectives in case processing outcomes, which Waegel (1981, 1982) and Eck (1983) understood to be important. Nevertheless, much has changed in criminal investigations since these publications. Advancements in communication, information processing and forensic technologies, for example, suggest that it is time to revisit the ‘black box’ of retrospective criminal investigations.
Methods
These analyses seek to overcome the issues noted above by offering a more refined look at the investigative process. The data utilised in the current study were gathered by the Houston Police Department (HPD) in Houston, Texas. At the time the data were collected, the HPD employed approximately 5,300 sworn officers and served one of the largest cities in the United States of America (boasting over two million citizens). In the investigative divisions at the HPD, three subcommand branches pursue crimes known to police, including technology services, special and criminal investigations. These subcommand branches house 12 investigative divisions that have detectives and support staff who focus on offenders, offences and victims. Each of the investigative divisions at the HPD have ongoing reactive and proactive investigations. Reactive, in this sense, means that the investigation was in response to public demands for service. Alternatively, proactive investigations are initiated by detectives. For our purposes, reactive and proactive investigative units were parsed in these data and these analyses focus on the former to mirror Eck (1983).
Between April 22 and June 20, 2013 (i.e., 60 days of observation), HPD detectives collected self-report activity log data from each of their investigations. Like Eck (1983), these analyses focus on burglary and robbery investigations 3 and, as such, over a quarter of detectives assigned with burglary and robbery investigative duties with the HPD participated in this data collection effort (n = 22, 28.9% and n = 18, 27.7%, respectively). The 40 observed detectives responded to a solicitation to participate in the study via a list-serve of active detectives that the HPD provided to the researchers. Participating detectives did so voluntarily and could withdraw from the data collection at any time and for any reason without repercussions, though none did.
Detectives that contributed to the study were tasked with identifying the cases they worked and investigative actions they took from a list of 19 provided activities. Like many other police departments (see Horvath et al., 2001), the HPD was not collecting detective effort-related data prior to these analyses. As such, the detective activity measures found in this study were identified by consulting the available criminal investigations literature and HPD investigative command staff. Consequently, there is substantial overlap between the measures employed here and those found in Eck (1983; see Tables 1 and 2).
Effort that did not pertain to any specific case, such as training or sick leave, was not analysed in this study. Additionally and unlike Eck (1983), these analyses hone in on the temporal patterns of a specific investigative landmark (i.e., effort that follows case assignment up to an arrest in the case). Subsequent investigative effort was omitted from these analyses, including time spent making an arrest and appearing in court, as these measures were found to be temporally inconsistent with the arrest disposition. This resulted in 17 detective effort measures, including attempting to locate someone, closing a case out, conducting physical surveillance, conferring with HPD personnel, non-HPD personnel and the prosecutor, checking a database, processing physical evidence, reviewing the casefile, running a warrant, conducting a suspect interview, travelling, conducting a victim interview, writing supplemental reports and an ‘other’ category that allowed detectives to specify an activity not found among these items. 4
To ensure activity log validity, daily counts for each detective were inspected for irregularities. In general, detectives reported a credible amount of daily effort (e.g., 8 hours a day, ± 1 hour). When detectives reported 12 or more hours of daily effort, HPD investigative command staff corroborated the accuracy of the activity log. This was an infrequent occurrence and in all instances investigative command staff personnel verified the time-task log in question. Additionally, detective activity log coding was reviewed for patterns that would indicate a disregard for the activity log coding scheme, though none of the burglary or robbery detectives in these data were questioned in this regard.
To complement detective sampling procedures, Figure 1 displays case sampling considerations among temporally inconsistent criminal investigations. In the best-case scenario, for example, criminal investigations would start and end in conjunction with the period of observation (M1; Ideal). This allows researchers to properly define their study’s parameters. In reality, however, criminal investigations may not be assigned, initiated (M2; Premature) or completed (M3; Delayed) in concurrence with the period of observation. An investigation may also start prior to and last well beyond the observational period (M4; Prolonged). Furthermore, it is possible that cases will be worked intermittently during the period of observation (M5; Gap; Eck, 1983). Investigations, therefore, lack uniformity in their commencement, conclusion, and length (M6; Varied). 5 These variations influence the measurement of detective effort.
Temporal sampling technique for detective effort in burglaries and robberies.
To account for the temporal variations in criminal investigations, the forthcoming analyses utilise a unique longitudinal case sampling technique. In conjunction with detective activity logs, only offences occurring and cases assigned between 22 April and 21 May 2013 were incorporated into these analyses. This 30-day period overlaps with the first half of the detective observational period. This sampling procedure is necessary because a crime that occurs on the last day of the observational period (i.e., 19 June 2013) would likely have few, if any, instances showing detective effort. Including these types of cases might lead to (inaccurately) concluding that the detective had shirked on their responsibility to investigate the case. In actuality, however, the detective’s effort was outside the period of observation and, therefore, not found in these data. This selection procedure ensures that at least 30 (up to 60) days of detective effort were available in these analyses and provides observational parity among the detective effort estimates.
Though Leiderbach et al. (2011) and Fallik (2017) utilised similar temporal case sampling procedures, a number of significant points must be explored in greater detail. To be included in these analyses, for example, the crime and assignment of the case to a detective had to occur within the case sampling period (t0–t4; 22 April – 21 May 2013). This dual criterion was chosen because cold case temporal patterns are likely different than contemporary cases and unassigned cases are not typically investigated by reactive detectives. Though the exclusion of these cases will likely diminish our ability to observe the totality of a detective’s concurrent investigative effort, that objective is beyond the scope of these analyses, in which the unit of analysis is cases, not detectives. Furthermore, the reliability of our findings would be greatly influenced if the typical criminal investigation lasted longer than the observational period. Greenwood et al. (1977), however, found that 84% of cases are cleared within a single week (see also Eck, 1983). The 30-day minimum period of detective effort observations found in these data is four times greater than is typically required for case clearances. Finally, these analyses are not intended to estimate investigative effort throughout a case’s life-course. Any number of unforeseeable events could affect a detective’s effort in a case over time and it is impossible for researchers to predict these occurrences. These analyses, however, can be said to observe the temporal patterns of detective activities for at least a 30-day period of observation.
Results
The number of days that burglaries and robberies were investigated during the period of analysis is displayed in Table 3. Burglaries found in these data (n = 99, 40.7%) were investigated between zero and 11 days; however, only two cases (or 2.0% of the burglaries sampled) were investigated for more than four days, while the bulk of burglaries were investigated in one day (n = 30, 30.3%). Many burglaries, however, were never investigated during the period of analysis (n = 28, 28.3%). Robberies found in these data (n = 144) were investigated between 0 and 14 days, though investigations lasting seven or more days were rare (n = 7, 4.9%). Furthermore, slightly less than half of the robberies found in these data were not worked (n = 36, 25.0%) or worked in a single day (n = 29, 20.1%). Unfortunately, cases that were not worked, single-day investigations and investigative lengths with fewer than five cases do not provide adequate observational variation that is necessary to observe trends. These cases, therefore, are not found in the subsequent analyses (n = 123, 50.6%).
Number of separate days burglaries and robberies were investigated.
Table 4 displays the proportion of times burglary detectives performed individual investigative activities in two-, three- and four-day-long investigations. A plus sign notes day-to-day proportional increases in the performance of investigative activities. Among burglary investigations that lasted two days (n = 23, 23.2%), three investigative activities were never performed by detectives (i.e., conducting physical surveillance, conferring with non-HPD personnel and the prosecutor). Most investigative activities were performed at a greater rate from Day 1 to Day 2 of the investigation (n = 9), while conferring with other HPD personnel and travelling declined as two-day investigations progressed. Among the activities that were performed in two-day investigations, only interviews with the victim occurred at a constant rate (34.8%). Two-day investigations, however, offer fewer data points to evaluate, which limits our ability to interpret the broader temporal patterns of criminal investigations. Rate of change is, therefore, better understood and subsequently explored among longer burglary investigations.
Proportion of the time detective activities were performed during two-, three- and four-day retrospective burglary investigations.
Ten (10.1%) burglaries were investigated over a three-day period. Conducting physical surveillance and running a warrant were not performed among any of these cases. Furthermore, and unlike two-day investigations, none of the investigative activities increased over the course of three-day-long investigations. Alternatively, the rate of conducting a database check, conferring with other HPD personnel, reviewing the casefile, and suspect interviews declined each day as three-day-long burglary investigations progressed. Additionally, with more data points a new pattern emerged among some of the investigative activities, including travelling victim interviews and writing supplemental reports. These investigative activities are shaped like a downward-opening parabola, with a positive vertex occurring during the second day of the investigation. This suggests that the rate these investigative activities were performed is greatest during the middle of an investigation and least likely to be performed at the beginning and end of an investigation. With only three data points, however, many other investigative activities still do not demonstrate easily discernable patterns (n = 6).
Approximately 1 in 16 burglaries found in these data were investigated over four separate days (n = 6, 6.1%). Conducting physical surveillance and running a warrant were not observed among these cases; however, the proportion of cases in which the detective processed physical evidence increased from Day 2 to Day 4 of the burglary investigation. Alternatively, victim interviews declined over the course of four-day-long investigations. Like three-day burglary investigations, several investigative activities (n = 7) were shaped like a downward-opening parabola, with a positive vertex in the middle of the investigation. The investigative activities that demonstrated this effect, however, were dissimilar across investigative lengths. Additionally, three investigative activities (i.e., database checks, reviewing the casefile, and travelling) were shaped like an upward-opening parabola. This effect was not observed among shorter burglary investigative periods but suggests that these detective activities are least likely to occur in the middle and most likely to be performed in the beginning and end of an investigation. Finally among four-day burglary investigations, there was no overt pattern observed for running a warrant and writing supplementary reports among these cases. Across the varying burglary investigation lengths, the only pattern found in these data was that none of the detectives reported that they conducted physical surveillance as part of two-, three- and four-day-long investigations.
The rate activities were performed by robbery detectives in two-, three-, four-, five- and six-day-long investigations is displayed in Table 5. There were several activities (n = 5) during two-day robbery investigations (n = 24, 16.7%) that were not performed. Conferring with non-HPD personnel remained constant but a rare occurrence (4.2%), while the proportion of several other investigative activities (n = 5) increased across two-day robbery investigations. Alternatively, the proportion of cases with database checks, reviewing the casefile, suspect, and victim interviews all declined from Day 1 to Day 2 in these robbery investigations. Like two-day burglary investigations, however, the rate of activity change is better reflected among longer criminal investigations that have a greater number of data points and, are subsequently explored.
Proportion of the time detective activities were performed during two-, three-, four-, five- and six-day retrospective robbery investigations.
Among robberies that were investigated on three separate days (n = 16, 11.1%), three investigative activities were not performed (i.e., conducting physical surveillance, closing a case out, and running a warrant). Alternatively, the rate at which detectives conducted database checks, conferred with the prosecutor and wrote supplementary reports increased from day-to-day among three-day robbery investigations. Furthermore, the performance of several investigative activities (n = 6) declined as a three-day robbery investigations progressed. Processing physical evidence was least likely to occur at the beginning and end but peaked during the middle of three-day robbery investigations, in a downward-opening-shaped parabola. This finding runs contrary to Schroeder and White (2009), who suggested that DNA evidence would be processed at the greatest rate near the end of an investigation. Additionally, detective travel remained constant (12.5%) throughout three-day-long investigations.
Conducting physical surveillance and closing the case out continued to be unreported by detectives among four-day robbery investigations (n = 15, 10.4%). The rate at which detectives processed physical evidence and ran warrants, however, increased over the course of four-day robbery investigations. Additionally, detectives were less likely to review the casefile and interview suspects as four-day-long robbery investigations progressed. Several investigative activities were performed at their greatest rate during the middle of a four-day investigation, in a downward-opening-shaped parabola, including conducting a database check, conferring with non-HPD personnel, the prosecutor, travelling and writing supplemental reports. These investigative activities were also least likely to be performed at the beginning and end of a four-day-long robbery investigation. Alternatively, attempting to locate someone and conferring with other HPD personal were least likely to be performed in the middle of an investigation but were more likely to be performed at the beginning and end of a four-day-long robbery investigation, in an upward-opening-shaped parabola. The rate at which victim interviews were performed among four-day robbery investigations, however, did not demonstrate a consistent trend among these cases.
During five-day-long robbery investigations (n = 9, 6.3%), detectives did not close out the case or run warrants in these data. Furthermore, as these investigations progressed detectives were less likely to confer with the prosecutor and review the casefile. Detectives, however, were more likely to write supplemental reports in the middle of an investigation but least likely to perform this task at the beginning and end of five-day robbery investigations. This trend was shaped like a downward-opening parabola. Alternatively, conducting database checks, conferring with other HPD personnel and travelling were least likely to be performed in the middle and most likely to occur at the beginning and end of a five-day-long robbery investigation, in an upward-opening-shaped parabola. Several investigative activities (n = 6), however, showed no trend in their performance across five-day-long robbery investigations.
Finally, several investigative activities (n = 5) during six-day-long robbery investigations (n = 8, 5.6%) were not found in these data. Attempting to locate someone, conducting a database check and processing physical evidence tended to peak in the beginning of these cases and decline as six-day robbery investigations progressed. Conferring with non-HPD personnel, suspect interviews and travelling were most likely to occur during the middle of an investigation and least likely to be performed at the beginning and end of six-day-long robbery investigations, in a downward-shaped parabola. Alternatively, interviews with victims were the least likely to be performed in the middle and most likely to occur at the beginning and end of six-day-long robbery investigations, in an upward-shaped parabola. Finally, there was no clear observable pattern for two investigative activities (i.e., conferring with the prosecutor and writing supplemental reports) among six-day-long robbery investigations.
Unlike burglaries, there are some trends among robbery periods of investigation. First, detectives did not close out their cases in any of these robbery cases. Likewise, robbery detectives in only one instance conducted physical surveillance or ran a warrant in these data. Finally, with the exception of six-day-long robbery investigations, the rate at which casefiles were reviewed tended to decline during the course of two-, three-, four- and five-day-long robbery investigations.
Discussion
Perceptions of criminal investigations are largely based on fictional and sensationalised interpretations of detective work, but this has come at a considerable cost to police legitimacy and public safety (Reiner, 2008). The purpose of these analyses was to explore the ‘black box’ of the criminal investigations process. Prior research has suggested that detectives exercise a great deal of discretion that is virtually invisible to oversight (Pogrebin, 1976; Waegel, 1981, 1982). Discretion is balanced by heavy workloads that encourage detectives to skim or triage their efforts. Unfortunately, suppositions about the temporal arrangements of criminal investigations have been based on shaky methodological and analytical grounds. To overcome these issues, 40 detectives with the Houston Police Department voluntarily reported the date, activities and cases they were assigned during a 60-day period. Participating detectives were assigned 243 burglary and robbery cases that were analysed in this study. The results show that a quarter of the burglaries and robberies in this sample received no investigative attention during the period of analysis. Another quarter of these cases were investigated in a single day. The remaining investigations – where the temporal patterns of burglaries and robberies could be evaluated – were found to be considerably different (i.e., there was little consistency in the temporal arrangements of detective activities across differing investigative periods). As a consequence, few observable patterns were found in these data.
Policy implications
The results presented are consistent with prior research that finds that detectives do not have a lot of time to investigate the cases they are assigned (see Greenwood et al., 1977; Ward, 1971). Departments, however, have several opportunities during a case’s development in the criminal justice system to better balance detective workloads. Patrol officers, for example, should have a larger role in the investigative process. To this point, Liederbach et al. (2011) commented that streamlining the criminal investigations process would likely involve ‘concurrent efforts to restructure patrol officer time’ (2011: 61). Similarly, Fallik (2016) proposed that patrol officers should carry out all investigations with a known suspect to their resolution.
Furthermore, departments can relieve detective workloads by relying more on civilian workers. Civilians can aid retrospective criminal investigations by performing unskilled or highly specialised tasks, such as clerical work and forensic analyses (Maguire and King, 2004). Unfortunately, Horvath et al. (2001) found that civilian personnel represent approximately 3% of persons employed by law enforcement agencies and only a third of departments with civilian employees have them assigned to investigative divisions. The civilianisation movement in contemporary policing has been somewhat inhibited by the costs associated with hiring additional staff and uncertainty of civilian suitability with investigative work (Liederbach et al., 2011). Retrospective criminal investigations have, therefore, been underserved in this capacity. In either case, departments should monitor case processing outcomes so that shifts in departmental resources do not displace workload imbalances elsewhere in the criminal justice process.
Additionally, these analyses observed that the temporal arrangements of criminal investigations tend to be more nuanced than previously acknowledged in the extant literature. Though this study did not represent a direct test of the skimming or triage hypotheses, it does call into question whether distinct configurations of detective work exist. In addition to challenging existing explanations of detective decision-making, dissimilarities in criminal case processing also cause practical difficulties. Investigative administrators and detectives, for example, cannot know, at the onset or even in the middle of an investigation, what will come of the cases they are working. This makes it difficult, if not impossible, for them to budget their time appropriately or forecast where additional resources may be needed or reappropriated. Though this is certainly frustrating, an important reading of these results is that it would be inappropriate for departments to develop strict case processing polices because there does not appear to be a normal or standard way that burglary or robbery retrospective criminal investigations are carried out.
Though researchers may prefer emergent, definitive and dramatic results, ‘no pattern’ findings do tell citizens something important about the core nature of criminal investigations. That is that no two cases are alike. Contrary to the CSI effect, therefore, citizens should be patient and more understanding when their case appears to have stalled with detectives. This prescription is inconsistent with their expectations; however, these results suggest that criminal investigations are often less predictable than what is depicted in entertainment-based interpretations of detective work.
Limitations and areas of future research
Despite the progress made in this study, future research could further advance this field of inquiry by addressing the limitations found in these data. The information presented here, for example, was based on 243 burglary and robbery cases assigned to detectives within the Houston Police Department. The somewhat small sample size analysed may be less sensitive to deviations and, therefore, hinder our ability to observe changes in the temporal patterns of criminal investigations (i.e., an issue of statistical conclusion validity). Likewise, our focus in these analyses was on burglaries and robberies, to be directly comparable to Waegel (1981, 1982) and Eck (1983); however, a more complete understanding of detective decision-making would capture a greater number offences. The temporal patterns of aggravated assault, arson, forcible rape, homicide, larceny and motor vehicle theft criminal investigations, for example, have not been explored here or in other research. Substantively, a larger sample, spread among other offences, would quell these issues.
Additionally, other important within-case differences are not observed in these analyses. Though this study parsed criminal investigations by the offence, the length of the investigation and where the investigation was temporally, prior research (Eck, 1983; Greenwood et al., 1977; Waegel, 1981) finds that the suspect status at the time the case was assigned will promote different investigative activities and, as a consequence, could confound our observations. Similarly, retrospective criminal investigations are naturally nested within a hierarchical system that has levels of aggregation that are uniquely defined. Between-unit differences among detectives, divisions and departments are not observed in these data but are ripe for hierarchical analyses. How long investigative activities were performed and how well detectives execute activities, for example, were not captured in these analyses but could have important implications for our understandings of detective decision-making.
Finally, these analyses focused on an important phase in the development of a case in the criminal justice system; however, prior to and following retrospective criminal investigations are other important landmarks. Understanding the temporal arrangements of preliminary investigations, for example, may better contextualise patrol officer decision-making. Similarly, contextualising the role of detectives in post-arrest case processing is important to understanding the totality of detective workloads. Future analyses should look to explore these areas of case processing to better understand the temporal patterns of case processing in the criminal justice system.
Conclusion
Despite these limitations, this study systematically explored burglary and robbery criminal investigations in an effort to better understand detective work. In doing so, several methodological and analytical issues that have hindered the prior extant literature from deriving empirically sound hypothesis about detective decision-making were overcome. The burglary and robbery investigations observed in these data were found to be much more nuanced than prior empirical research has suggested. As a consequence, this exploration has promoted new and exciting areas for the next generation of scholarship to address.
Footnotes
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.