10 Questions Need Answered (Please See Ten Questions Attached along with Supporting Documents)

Citation: 13 Am. J. Police 149 1994

Content downloaded/printed from HeinOnline (http://heinonline.org) Wed Jul 29 10:18:46 2015

-- Your use of this HeinOnline PDF indicates your acceptance of HeinOnline's Terms and Conditions of the license agreement available at http://heinonline.org/HOL/License

-- The search text of this PDF is generated from uncorrected OCR text.

American Journal of Police Vol. XIII, No. 3 1994 149

THE RELATIONSHIP BETWEEN EVIDENCE, DETECTIVE EFFORT, AND THE DISPOSITION

OF BURGLARY AND ROBBERY INVESTIGATIONS

Steven G. Brandl University of Wisconsin - Milwaukee

James Frank University of Cincinnati

The police detective is commonly cast by the media and de- tective fiction as a highly skilled and intelligent investigator who, through the use of complex and extraordinary means, is capable of solving virtually any crime (Klockars, 1985; see also Garofalo, 1981). The public, in turn, has largely accepted this portrayal as an accurate reflection of reality (Wilson, 1975). Further, the police organization reflects and perpetuates this institutionalized belief by dividing the in- vestigative task among detectives, who specialize in investigating crimes, and patrol officers, who perform the countless other tasks asso- ciated with police work.

With the appearance of empirical inquiries into the operations of the criminal investigation process, the capabilities of detectives in investigating crimes have been directly questioned. However, not only have these studies been relatively few in number, but they have also produced contradictory findings. For example, Greenwood, Chaiken, and Petersilia (1977), explain that "detective work is not characterized by hard work leading to case solutions . . ." rather, "the cases that get cleared are primarily the easy ones to solve" (p. 112). Thus, "any justi- fication for the work of [detectives] must lie in areas other than crime solution" (p. 141). On the other hand, Eck (1983) argues that "the work of... detectives is extremely important with respect to the subsequent making of follow-up arrests" (p. 243). As a result of the divergent

150 American Journal of Police Vol. XIII, No. 3 1994

findings, definitive conclusions concerning the contributions of detec- tives in solving crimes cannot be drawn.

The analyses reported in this study are intended to cast addi- tional light on the debate concerning the extent to which detective ef- fort contributes to the identification and apprehension of criminals. Specifically, we test hypotheses that link burglary and robbery follow- up investigation arrests to the amount of time spent on follow-up in- vestigations and the strength of the evidence associated with the cases. The study addresses some shortcomings evident in the prior research and provides some empirical evidence where speculation has gone be- fore. In so doing, the study represents an extension of previous re- search.

PREVIOUS RESEARCH

In the first study to examine the contribution of detectives in solving crimes, Isaacs (1967) analyzed data on 1,905 crimes reported to two of 15 Los Angeles Police Department field divisions in January 1966. Isaacs found that of the 1,905 crimes, 336 (18%) resulted in an actual arrest while 146 (8%) resulted in "other clearances" (e.g., victim refused to prosecute, unfounded). Of the 336 cases that resulted in an arrest, only 27 (8%) were as a result of an arrest by a detective. Based on the limited statistical analyses performed on the data, the author concluded that "[o]ne of the most significant characteristics of a case, affecting its chances of being cleared, is whether the suspect was named in the crime report by the victim" (p. 96). For example, of the 1,905 crimes, 1,556 (82%) involved suspects who were not named in the initial crime report. Of these 1,556 cases, 1,375 (88%) were not cleared. Conversely, of the 1,905 crimes, 349 (18%) involved suspects who were named in the crime report. Of these cases, only 48 (14%) were not cleared.

Greenwood (1970), in an analysis of burglary, robbery, and grand larceny crimes that were reported to several precincts of the New York City Police Department during the first six months of 1967, also found that when a suspect could be named by the victim, the chances of an arrest were significantly greater compared to when only a descrip- tion was provided, or other evidence was available. For example, the analyses showed that approximately nine percent of reported robberies resulted in an on-scene arrest while the remaining 91 percent were ini-

American Journal of Police Vol. XIII, No. 3 1994 151

tially unsolved. Of the unsolved cases, the victim could name a suspect in about two percent and could describe the suspect in 67 percent. For those cases with a named suspect, 46 percent resulted in an arrest, while only two percent of the cases with a description resulted in an ar- rest. Once again, circumstances-namely, the victim providing the suspect's name-were portrayed as determining the outcome of the case.

The study conducted by Greenwood, Chaiken, and Petersilia of the RAND Corporation in 1977 supported, and elaborated on, the conclusions of Isaacs (1967) and Greenwood (1970). Their results were derived from analyses of primarily two sets of data. First, based on an analysis of the computer-readable case assignment file main- tained by the Kansas City (MO) Police Department, the researchers provided insight into how detectives spend their working time. In comparing the amount of time spent on uncleared cases with the amount of time spent on cleared cases, they found that the "time spent on uncleared cases [was] significantly higher than time spent prior to clearance on cleared cases with no initial patrol arrest" (p. 115).1 On the basis of this comparison, they conclude that there are basically two types of investigations: "Some investigations are lengthy but do not re- sult in a clearance. Others lead to a clearance after a small amount of work but entail substantial efforts by the investigator once the crime is cleared" (p. 116). Thus, "the cases that get cleared are primarily the easy ones to solve" (p. 112).

Second, a post hoc analysis of cleared crimes (homicide, ag- gravated assault, felony morals, robbery, burglary, theft, auto theft, and forgery/fraud) from five police departments (Berkeley, Long Beach, Los Angeles, Miami, and Washington D.C.) 2 provided a means by which to describe the method in which cleared cases were solved. Greenwood et al. (1977) found that about 22 percent (38 of 172) of the cleared crimes resulted from an arrest by a patrol officer at the scene of the crime. In another 44 percent (76 of 172), the perpetrator was "known" when the crime report was first taken and, as described by the researchers, "the main job for the investigator in these cases [was] to locate and apprehend the perpetrator, and to assemble evidence ade- quate to charge him" (p. 135). The remaining 34 percent (58 of 172) of all cleared crimes were cleared primarily as a result of "simple routine actions" of detectives (e.g., showing mug shot books, acting on precise and detailed informant tips). According to Greenwood et al. (1977), only about three percent of these cleared crimes could be attributed to

152 American Journal of Police Vol. XIII, No. 3 1994

"special actions" of detectives (e.g., preparation of investigative bul- letins, matching latent prints to inked prints in a file organized by M.O.).

Like Isaacs (1967) and Greenwood (1970), Greenwood et al. (1977) explain that the overwhelming majority of solved crimes are solved because of arrests by patrol officers or because the identity of the perpetrator is known when the crime report reached the detective. "If information that uniquely identifies the perpetrator is not presented at the time the crime is reported, the perpetrator, by and large, will not be subsequently identified" (Greenwood, Chaiken, and Petersilia, 1976:65). Thus, "it is not true in general that the greater the investiga- tive effort devoted to a crime the more likely it will be cleared" (p. 64).

Eck (1983) provides contrast with Isaacs (1967), Greenwood (1970), and Greenwood et al. (1977) by emphasizing, or at least not minimizing, the importance of detective effort in solving crimes. Us- ing data collected from three police departments (DeKalb Co., GA, St. Petersburg, FL, and Wichita, KS), he examined activities performed by detectives and patrol officers to produce crime and/or suspect informa- tion, and the relationship between this information and a suspect being arrested. The study focused on burglaries and robberies. Eck found that certain types of information produced during the preliminary in- vestigation (i.e., information pertaining to the suspect) were predictive of whether or not an arrest was made during the follow-up investigation as were some of the activities performed by detectives during the fol- low-up investigation (e.g., record checks, informant interviews). As such, Eck provides evidence to suggest that circumstances of criminal incidents as well as efforts of detectives contribute to the making of ar- rests during follow-up investigations.

On the basis of his analyses, Eck (1983) proposed what he re- ferred to as the "triage hypothesis" (see also Eck, 1979). According to this hypothesis, cases can be grouped into three categories depending on the strength of the evidence (i.e., favorable circumstances) associ- ated with the case. The hypothesis predicts that, first, cases with weak evidence will rarely conclude with an arrest, regardless of the inves- tigative effort or time allocated to them. Second, cases with moderate evidence, where there are sufficient leads to suggest that the cases could be solved, will often result in an arrest but only if effort is allo- cated to them. Third, cases with strong evidence, where the suspect has been positively identified, will usually result in an arrest, even with only minimal investigative effort. Accordingly, based on this hypothe-

American Journal of Police Vol. XIII, No. 3 1994 153

sis, circumstances determine the outcome of the cases in the first and third group and thus, for these cases, the probability of an arrest does not increase with more time spent. However, effort determines the out- come of the cases in the second group and as a result, for these cases, the probability of an arrest does increase with more time spent on the investigation. Therefore, like Greenwood et al. (1977), Eck implies that, in general, the likelihood of an arrest may not increase with more effort devoted to the case. However, Eck (1983) suggests that this re- lationship--that more time spent on an investigation increases the probability of an arrest--does hold true for cases with moderate suspect information. And, as hypothesized by Eck (1979), it is cases in this second category that have the most time devoted to them.

There are several issues that need to be considered when judging the overall validity of the conclusions offered by the previous research. First, Isaacs (1967) and some of the analyses performed by Greenwood et al. (1977:97-119) used whether cases were "cleared" as the primary outcome variable. 3 A clearance, which is subject to ad- ministrative definition, can be obtained in ways other than by making an arrest (e.g., when the victim refuses to cooperate in the investiga- tion);4 therefore, a "cleared crime" should not be considered synony- mous with a "solved crime." Whether or not at least one arrest was made for the particular crime under investigation appears to be a more accurate indicator of investigative success because it focuses more di- rectly on the work product of detectives rather than on associated ad- ministrative practices or policies (cf. Greenwood, 1970; Eck, 1983; also see Bloch & Bell, 1976). 5 In addition, other analyses performed by Greenwood et al. (1977:121-32) focus only on cleared crimes, thus making it impossible to identify the actual impact of detective effort on case disposition.

Second, although Greenwood (1970) and Eck (1983) use "arrest" as the outcome variable, it is not clear whether they recognize that detectives are sometimes constrained in their abilities to make an arrest even when a suspect has been positively identified (see Eck, 1983:202-4). For example, a victim may not desire that the police con- duct an investigation of the criminal incident. In another instance, it may be determined (either by the police or the victim) that a reported crime did not actually occur. It would be inappropriate to expect the police to pursue charges and make arrests in these types of cases and thus, it would be misleading to consider these types of cases as "no ar- rest" cases in the statistical analyses. It would seem more appropriate

154 American Journal of Police Vol. XIII, No. 3 1994

to consider such cases "constrained cases" and treat them independent of the "arrest" versus "no arrest" classification.

Third, while much of the previous research suggests that the evidence-namely, suspect information-available at the conclusion of the initial investigation determines the ultimate disposition of the case (Isaacs, 1967; Greenwood, 1970; Greenwood et al., 1977), the com- plexity of the construct has often escaped its operationalization. Say- ing that a suspect is "described," (Eck, 1983), "named" (Isaacs, 1967; Greenwood, 1970) or "known" (Greenwood et al., 1977) without speci- fying how, may invite erroneous conclusions. For example, a suspect could be "named" as a result of a victim guess or because an eyewit- ness saw the individual committing the crime. There is clearly varia- tion in the strength of the information in these two situations but it is not captured with the simple "named" versus "not named" classifica- tion. Thus, to adequately determine the strength of suspect informa- tion, it is necessary to specify the basis of the information.

Finally, measures of detective effort have often been suspect in terms of their validity. Greenwood et al. (1977) used "hours spent" but activities that took less than one-half hour were not included in the estimate. 6 In addition, the activity categories on which to record time spent prior to an arrest were rather limited ("interrogation," "interview," "report writing," and "crime scene investigation"). Eck (1983) used, as the primary indicator of investigative effort, whether certain activities were performed in the investigation, but this measure does not account for the variation in the amount of time spent on spe- cific activities (e.g., an interview that lasts two hours versus one that takes ten minutes). Most troubling, however, is that some analyses do not include any measure of detective effort but inferences about the contributions of detectives in solving crimes are made nonetheless (Isaacs, 1967; Greenwood, 1970; Greenwood et al., 1977:121-132).

In this study, we attempt to address these issues. We focus on cases that were closed with either an "arrest" or "no arrest" and we ex- plicitly realize that, independent of detective effort and strength of evi- dence considerations, detectives are sometimes constrained in their ability to make arrests even when a suspect has been identified. We use a measure of suspect information that enables us to test Eck's (1983) triage hypothesis. Finally, we include in the analyses an appro- priate measure of detective effort. In making these methodological and conceptual refinements, the study may extend our understanding of the relationship between evidence, detective effort, and case dispositions.

American Journal of Police Vol. XIII, No. 3 1994 155

METHOD

Study Site

The data for this study were obtained from a medium-sized midwestern municipal police department. At the time of the study, the department employed 245 sworn officers and 102 civilians. Thirty of the 245 officers were assigned to the investigations division. In 1989 the department served a community of approximately 130,000 people; 33 percent of whom were non-white.

Sample

The process of selecting cases for this study consisted of sev- eral steps. Initially, a list of all the burglaries and robberies that oc- curred between July 1, 1989 and June 30, 1990 was obtained from the department's computer system. This list identified 339 robberies and 1,674 burglaries. All of these cases, which were in chronological order within the records bureau, were then reviewed and several categories of cases were eliminated.

First, all of the cases that were not assigned to detectives were excluded (12 robberies, 1,219 burglaries).7 Second, all of the reports which stated that an arrest was made prior to case assignment (i.e., during the initial investigation) were excluded (32 robberies, 73 bur- glaries). Third, cases that were assigned to detectives only after infor- mation became available that would allow the cases to be immediately "closed" (defined below) were excluded (e.g., a person confessed to committing a crime while being questioned about another crime) (no robberies, 32 burglaries). Fourth, cases that were missing from the files were excluded from the population (two robberies, 33 burglaries). As a result of the case selection procedure, 292 robbery cases and 317 burglary cases were included in the sample (N = 609). For each of the 609 cases, the initial and follow-up investigation reports were reviewed and a case data coding form was completed.

156 American Journal of Police Vol. XIII, No. 3 1994

Measures

The independent variables in this study were (1) strength of suspect information at the conclusion of the initial investigation and (2) amount of detective time spent on the follow-up investigation. Data on "suspect information" were obtained from the initial investigation re- port. The variable incorporated what was known about the culprit (description, identification, name) and how this information was pro- duced (on the basis of a victim or witness guess, an eyewitness to the crime, or some other witness account). A description refers to infor- mation about the physical characteristics of the culprit while an identi- fication means that the witness could recognize the culprit if seen again. It is expected that this measurement scheme captures more of the construct's complexity than a simple tally of how many suspect-re- lated solvability factors were present at the conclusion of the initial in- vestigation (e.g., suspect named? suspect described? etc.) or stating that the suspect is "described," "known" or "named" without specifying how (e.g., a guess? an eyewitness account? as in Isaacs, 1967; Green- wood, 1970; Greenwood et al., 1977; and Eck, 1983). Although one might question the use of "suspect information" as the sole measure of "evidence," previous research has consistently shown that suspect in- formation is the most powerful, if not the only, evidence consideration vis-A-vis the solvability of burglaries and robberies (Greenwood et al., 1977; see also Bloch & Bell, 1976).

To be congruent with, and test, the "triage" hypothesis identi- fied earlier, the suspect information variable had three values: (1) weak suspect information (no significant suspect information available, a suspect was described and could be identified on the basis of a [victim] guess only, a suspect could be described by an eyewitness to the crime, a suspect could be named by a guess); (2) moderate suspect informa- tion (a culprit could be described and identified by an eyewitness to the crime, the culprit could be named by being seen in the area at about the time the crime occurred, the culprit could be named on the basis of a guess and supported by other witness information, the culprit could be named by an individual who saw the culprit with the stolen property or heard the culprit confess to committing the crime, or the culprit could be named in some other way [i.e., through patrol activities]); and (3) strong suspect information (the culprit could be named by an eyewit- ness to the crime).

American Journal of Police Vol. XIII, No. 3 1994 157

The operational definition of "suspect information" was de- veloped on the basis of observations and discussions with the burglary and robbery detectives in the department. 8 The information classified as "moderate suspect information" consistently constituted "good leads" for detectives to follow; the information was of better quality and provided more direction in terms of activities to perform than the suspect information defined as "weak." However, "moderate suspect information" did not allow the detectives to conclude with any degree of certainty that the suspect could/would be arrested-as was the case if "strong suspect information" was present.

Data on the amount of time spent on follow-up investigations were obtained from "turn-back sheets" on which detectives recorded the amount of time spent on, and the activities performed in, 9 the in- vestigation from the point when the initial investigation report was re- ceived until the case was closed (e.g., an arrest warrant was issued, all leads were exhausted, the crime was unfounded, victim did not wish to pursue the complaint, etc.). This measure did not include any time spent on the investigation subsequent to an arrest or any other case clo- sure. The completion of the turn-back sheets was a standard procedure of the department.

10

The dependent variable, case closure status, was also deter- mined from the turn-back sheets. Detectives could assign one of sev- eral statuses to a case in order to terminate (i.e., "close") an investiga- tion: (1) "No Further Investigation" (NFI; all leads were exhausted and the investigation could go no further); (2) "Prosecutor Refused to Is- sue" (PRF; a suspect was identified but the prosecutor's office would not issue an arrest warrant); (3) "Unfounded" (UNF; a reported crime was determined not to have been committed); (4) "Victim Canceled In- vestigation" (VCI; the victim no longer wished to pursue the complaint, did not cooperate in the investigation, or did not wish to prosecute the offense); (5) "ClosedCOther Prosecution" (COP; the accused was not charged with the present crime but was charged with a different crime); or (6) "Warrant Pending Arrest" (WPA; an offender was identified and a warrant was issued by the prosecutor's office, or a juvenile offender was referred to probate court).

For most of the analyses, case closure status was dummy coded as "arrest made" versus "no arrest made." The "arrest" category consisted of (a) cases in which an arrest warrant was issued (WPA) and (b) cases that were investigated by detectives and the identity of the suspect was positively known but for some (unspecified) reason, the

158 American Journal of Police Vol. XIII, No. 3 1994

suspect was not charged with the crime (COP). The "no arrest" cate- gory consisted of cases in which (a) all leads were exhausted (NFI) and (b) the prosecutor declined to issue an arrest warrant (PRF). For rea- sons identified earlier, cases in which (a) the victim canceled the in- vestigation (VCI) and (b) the crime was determined not to have oc- curred (UNF) were considered "constrained cases" and excluded from most of the analyses.

FINDINGS

Table 1 provides descriptive data on the uncollapsed values of suspect information and case status for burglary and robbery cases and also shows the mean amount of time spent on each category of cases. Of particular importance in Table 1 are the data concerning the char- acteristics of the "constrained cases"-cases that are excluded from the subsequent analyses. It is seen that relatively few burglary and robbery cases were closed with the "UNF" (Unfounded) status; only 14 of 317 (4.4%) burglaries and 11 of 279 (3.9%) robberies were closed in this manner. Given the relative rarity of this case closure status, it does not appear, at least on its face, that there was a widespread unjustified practice of unfounding (at least burglary and robbery) crimes in this department. As a result, there is little reason to believe that bias is in- troduced when these cases are excluded from the subsequent analysis. Overall, 64 of the 317 (20.2%) burglaries and 97 of the 279 (34.8%) robberies were excluded as a result of being closed with either the "iNIF" or "VCI" status.

Tables 2 and 3 provide data on the collapsed categories of suspect information and case status for burglaries and robberies, re- spectively. The mean amount of time spent on each category of cases is also presented. Not surprisingly, Table 2 shows that the likelihood of a burglary investigation concluding with an arrest increases with stronger suspect information. Of the 133 cases with weak suspect in- formation, 13 (10.8%) concluded with an arrest. This percentage in- creases to 33.7 percent with moderate suspect information, and to 62.1 percent with strong suspect information. The correlation between sus- pect information and case status is moderately strong (r = .43; p < .001).11

Anerican Journal of Police Vol. XIII, No. 3 1994 159

Table 1 SUSPECT INFORMATION, TIME SPENT, AND CASE STATUSES:

DESCRIPTIVE STATISTICS FOR BURGLARY AND ROBBERY FOLLOW-UP INVESTIGATIONS

Mean Time CASE STATUS Susp. on Case Arrest No Arrest Constrained Info (in hours) S.D. Made Made Case

WPA COP NFiL PRF UNF VCI Totals

No Sig Info 3.8 5.6 1 (3.1) (2.2) (0)

Des/Id-guess 2.7 .9 0 (-) (-) (0)

Des-eyewit 3.1 2.7 0 (3.2) (6.0) (0)

Named- 2.8 2.8 12 guess (-) (-) (0)

Des/Id- 5.3 2.8 3 eyewit (5.2) (6.2) (30)

Named- 3.2 1.6 4 saw in area (-) (-) (0)

Named- 5.2 5.5 5 guess w/ (6.1) (-) (0) other support

Named-saw 6.1 11.1 5 w/prop or (7.9) (2.1) (1) heard confess

Named- 6.0 4.4 8 other (6.7) (3.4) (3)

Named- 2.9 2.6 23 eyewit (2.8) (1.5) (9)

TOTALS 61

MEAN TIME SPENT ON CASE (in hours)

S.D.

0 34 1 (1) (19) (1)

0 2 0 (0) (0) (0)

0 9 0 (0) (34) (0)

0 69 5 (0) (0) (0)

1 12 1 (11) (76) (3)

0 11 3 (0) (0) (0)

1 9 0 (0) (1) (0)

0 11 1 (0) (0) (0)

1 7 0 (0) (0) (0)

0 6 8 (0) (2) (3)

3 170 19 (41) (12) (130) (7) 6.8 6.2 3.2 3.4

(8.6) (9.8) (4.0) (5.4)

7.6 3.1 3.3 3.2 (6.1) (9.7) (5.8) (2.9)

Notes: (1) entries in crosstabulation are raw frequencies; (2) entries in parentheses are for robberies; (3) missing data are excluded from the analyses.

43 (28) 2

(0) 9

(40) 101

(0)

18 (181) 20 (0)

20

(1)

21 (2)

17 (5) 66

(35)

2 5 (3) (4) 0 0

(0) (0)

0 0 (1) (5)

5 10 (0) (0)

0 1 (5) (56)

0 2 (0) (0)

1 4 (0) (0)

3 1 (1) (0)

0 1 (0) (2)

3 26 (2) (19)

14 50 (11) (86)

2.1 1.9 (3.0) (2.2) 2.2 1.7

(1.5) (1.4)

160 American Journal of Police Vol. XIII, No. 3 1994

Table 2 RELATIONSHIP BETWEEN CASE OUTCOMES, STRENGTH OF SUSPECT

INFORMATION, AND TIME SPENT FOR BURGLARY FOLLOW-UP INVESTIGATIONS

Case Status_ 0= 1= Mean Time

N No Arrest Arrest Spent (hours)

Strength of Suspect Information l=Weak 133 120(3.01) 13 (6.41) 3.35 2=Moderate 83 55 (3.87) 28 (8.75) 5.52 3=Strong 37 14 (2.56) 23 (4.44) 3.73

N 253 189 (3.23) 64 (6.72)

Notes: (1) entries in parentheses are mean amount of time spent on each set of cases (in hours), other entries are raw frequencies; (2) missing data are ex- cluded from the analyses.

Also apparent from Table 2 is that, overall, more time was spent on burglary cases that concluded with an arrest than on cases that did not conclude with an arrest (3.23 versus 6.72 hours; t = -3.65; p < .000; t-tests not tabled). Overall, there is a slight relationship between time spent on burglary follow-up investigations and case status (r = .31; p < .001). In addition, most time was spent on cases with moderate suspect information (5.52 hours).

Table 3 shows that, as with burglaries, the likelihood of a rob- bery investigation concluding with an arrest increases with stronger suspect information. Of the 55 cases with weak suspect information, only one (1) (1.8%) concluded with an arrest. Of the 121 cases with moderate suspect information, 43 (35.5%) concluded with an arrest. Of the 14 cases with strong suspect information, nine (9) (64.3%) con- cluded with an arrest. There is a moderately strong relationship be- tween suspect information and case status (r = .41; p < .001).12

American Journal of Police Vol. XIII, No. 3 1994 161

Table 3 RELATIONSHIP BETWEEN CASE OUTCOMES, STRENGTH OF SUSPECT

INFORMATION, AND TIME SPENT FOR ROBBERY FOLLOW-UP INVESTIGATIONS

Case Status 0= 1= Mean Time

N No Arrest Arrest Spent (hours)

Strength of Suspect Information l=Weak 55 54 (3.32) 1(6.50) 3.38 2=Moderate 121 78 (4.63) 43 (10.35) 6.66 3=Strong 14 5(3.40) 9(3.51) 3.47

N = 190 137 (4.07) 53 (9.12)

Notes: (1) entries in parentheses are mean amount of time spent on each set of cases (in hours), other entries are raw frequencies; (2) missing data are excluded from the analyses.

Table 3 also reveals that, overall, more time was spent on rob- bery cases that concluded with an arrest than on cases that did not (4.07 versus 9.12 hours; t = -4.68, p < .000). Overall, there is a slight rela- tionship between time spent and case status (r = .35; p < .001). In ad- dition, most time is spent on cases with moderate suspect information (6.66 hours).

To determine the effect of time spent on the likelihood of an arrest being made within each category of suspect information, and thus to directly test the triage hypothesis, logistic regression analyses were used. Separate analyses were conducted on burglaries and rob- beries. In each model, a pair of interactive terms (or "slope dummy variables;" see Hanushek and Jackson, 1977) was included. Each model took the following form:

Y = a + blX 1 + b2 X2 + b 3 X3 + b4 X4 + b5 X5 + e

where a = a constant X, = time spent X2 = time spent if suspect information is weak, 0 otherwise X 3 = time spent if suspect information is strong, 0 otherwise X4 = 1 if suspect information is weak, 0 otherwise X5 = 1 if suspect information is strong, 0 otherwise

162 American Journal of Police Vol. XIII, No. 3 1994

The results can be interpreted as follows: b1 is the estimated effect of effort on case disposition (O=no arrest; l=arrest) in cases with moderate suspect information; the sum of bl+b 2 is the estimated effect of effort in cases with weak suspect information; and the sum bl+b3 is the esti- mated effect of effort in cases with strong suspect information. b 2 re- flects the size of the difference in the effects of effort between moder- ate and weak suspect information cases and b3 reflects the size of the difference in the effects of effort between moderate and strong suspect information cases. The effects of suspect information on case disposi- tion are reflected in b4 and b5 .

Table 4 LOGIsTIc REGRESSION OF TIME SPENT AS A DETERMINANT OF

ARRESTS IN BURGLARY AND ROBBERY FOLLOW-UP INVESTIGATIONS

Regressors Burglaries Robberies

(X 1) Time Spent .235** .226** (.075) (.061)

(X2 ) Time Spent/Weak Cases -.102 -. 171 (.093) (.117)

(X3 ) Time Spent/Strong Cases .047 -. 119 (.189) (.555)

(X4 ) Suspect Info/Weak -.857 -2.204 (.615) (1.261)

(X5) Suspect Info/Strong 1.497* 2.247 (.772) (2.022)

Constant -1.930** -2.029** (.465) (.425)

N 253 190

-2 log Likelihood 215.61 160.74

Notes: (1) unstandardized coefficients reported with standard errors in parentheses; (2) * p <.05; ** p <.01, one-tailed test; (3) missing data are excluded from the analyses.

As seen in Table 4, the amount of time spent on the burglary follow-up investigation is a significant predictor of whether an arrest

American Journal of Police Vol. XIII, No. 3 1994 163

was made in the investigation for cases with moderate suspect infor- mation (b 1 , p=.000). Separate logistic regression procedures (not tabled), using time spent on weak cases and strong cases as baseline measures show that the effect of effort in weak burglary cases is also significant (bl+b2 , p=.007) but the effect of effort in strong cases falls just short of a conventional level of significance (bl+b3 , p=.052). While the impact of effort on whether an arrest was made appears the strongest in burglary cases with moderate suspect information,

13 the re- sults reported in Table 4 show that the size of the difference in the ef- fects of effort between moderate and weak cases (b2 ), and between moderate and strong cases (b 3) are not significant (p=. 136 and p=.402; respectively).

14

As for robberies, the amount of time spent on the follow-up investigation is, as with burglaries, a significant predictor of whether an arrest was made in the investigation for cases with moderate suspect in- formation (b1 , p=.000). Separate logistic regression procedures (not tabled) demonstrate that the effect of effort on robbery cases with weak and strong suspect information is not significant (bl+b2 , p=.292; bl+b3 , p=.443, respectively). However, it is seen in Table 4 that the size of the difference in the effects of effort between moderate and weak cases (b2 ) and between moderate and strong cases (b3 ) are not significant (p=.071 and p=.415, respectively).

15

CONCLUSION

Our findings cast some additional light on the relationship between detective effort, strength of the evidence associated with the case, and whether an arrest is made in the investigation. Specifically, the analyses show that in burglary and robbery cases with moderate suspect information, the probability of an arrest increased significantly with more time spent. In fact, the relationship between time spent and case outcome is the strongest in the cases with moderate suspect infor- mation. However, the magnitude of the effects across cases with varying strengths of suspect information is not significant; the effect of time spent on case outcome in moderate cases is not significantly dif- ferent than the effect of time spent on case outcome in either weak or strong suspect information cases. Accordingly, while our findings of- fer only limited and qualified support for Eck's (1983) triage hypothe- ses, the results do contradict Greenwood et al. (1977) and others who

164 American Journal of Police Vol. XIII, No. 3 1994

have suggested that detective effort has no effect on whether crimes are solved. When considering the triage hypothesis in relation to the find- ings reported here however, one must be cognizant of the fact that in the study department, the (burglary) cases with the weakest suspect in- formation are screened out. If such cases had been assigned to detec- tives, then one might expect to find greater support for the triage hy- pothesis-i.e., find the effects of effort in weak cases to be quite small and thus, differ significantly from the effects of effort in cases with moderate suspect information.

Consistent with the expectation of Eck (1979), the analyses show that more time is spent on cases with moderate suspect informa- tion than on cases with weak or strong suspect information. This is perhaps because of the detectives' reasonable expectation that with just a little more information these cases could conclude with an arrest. The analyses also show that, overall, significantly more time was spent on burglary and robbery cases that were solved than on cases that were not solved. This directly contradicts the conclusion of Greenwood et al. (1977) that "the cases that get cleared are primarily the easy ones to solve" (p. 112).

As a result of this study, some directions for future research have also been illuminated. First, it would be useful to ask about the extent to which the findings of this study are generalizable to the in- vestigation of other types of crimes and to other organizational con- texts. On the basis of the data analyzed here, it is seen that there are few differences between burglaries and robberies; however, there is lit- tle justification for this belief vis-A-vis other crime types. As a result, the question becomes: In what types of crime does detective effort have more or less impact on whether an arrest is made? It would also be useful to examine the relationships of interest across different organi- zational settings with varying organizational structures, workloads, and/or policies (especially case screening). In essence, the generaliz- ability of the findings in these regards remain empirical issues.

Second, while this study focused on the impact of time spent on case status in burglary and robbery investigations, the study does not, because of data limitations, provide guidance as to what activities should consume investigative time. Accordingly, future efforts could be directed at identifying how detectives should most productively spend their case-related investigative time.

Finally, because the evidence associated with cases appears to limit (but not eliminate) the extent to which detectives can affect case

American Journal of Police Vol. XIII, No. 3 1994 165

outcomes, and because patrol officers are initially responsible for iden- tifying and collecting evidence, it is important to examine the determi- nants of time spent on initial investigations and the corresponding role of evidence in this calculus. Studies focused on these issues could add another dimension to our understanding of the criminal investigation process.

NOTES

1. The researchers considered a full range of crime types in these analy- ses including crimes against persons (e.g., homicide, rape, robbery, kidnapping), crimes against property (e.g., auto theft, theft from auto, burglary, larceny), other destructive acts (e.g., arson, vandalism) and acts of fraud (e.g., shoplifting, forgery). The same general finding was uncovered across most of the crime types including burglaries and robberies.

2. The reported findings were also confirmed with data from the Kansas City Police Department.

3. However, Greenwood et al. (1977) do not specify what constitutes a "clearance" in these analyses.

4. The difference between the number of clearances and the number of arrests can be substantial. For example, Eck (1983) found that in one agency, 58 percent of the burglaries that were cleared were as a result of an arrest. In another site, 100 percent of cleared cases resulted from an arrest. Greenwood (1970) found that of the burglary cases that were cleared, only 26 percent were by an actual arrest. Of the robberies that were cleared, 60 percent were as a result of an arrest.

5. See Sherman and Glick (1984) for a discussion of the validity and re- liability of arrest statistics.

6. In assessing the validity of their data, Greenwood et al. (1977) ex- plain that "some case-related activity presumably takes place in blocks of less than a half-hour and therefore goes unreported" (p. 104).

7. For a case to be assigned to a detective for a follow-up investigation it had to survive two screening stages. First, the department operated a Differential Police Response (DPR) strategy of call management.

166 American Journal of Police Vol. XIII, No. 3 1994

As part of this strategy, certain types of complaints, which were identified by policy, were taken over the telephone by police tele- phone operators. Complaints handled in this manner were very rarely brought to the attention of the investigations division. Second, de- tective sergeants had the responsibility for screening cases that re- ceived an initial investigation and assigning only those deemed ap- propriate for follow-up investigations. Accordingly, of the 12 rob- beries not assigned, all 12 were screened out by the detective sergeants. Of the 1,219 burglaries not assigned, 697 received tele- phone reports and 540 were screened out by the detective sergeants.

8. For a period of 30 weeks, from September 1990 to March 1991, de- tectives were observed by the first author for 400 hours (approximately 50 eight-hour shifts). Time was split equally between the crimes against property squad (burglaries) and the crimes against persons squad (robberies). The primary focus of the observations was on examining how detectives made time allocation decisions (see Brandl, forthcoming). The observations informed judgments as to how detectives considered and acted upon evidence, as well as other case information, and provided a means by which to judge the valid- ity of the self-report measure of time spent (see note number 10).

9. Listed activities were similar to those identified by Eck (1983). They included interviewing victims, witnesses, informants, suspects, and others, conducting neighborhood canvasses, searching the crime scene, submitting evidence for analysis, searching departmental records or other files, conducting photo line-ups and physical line- ups, and showing mug-shot books. Although it could have been a useful analytic addition, time spent on each activity was not collected for this study.

10. The validity of the self-report measure of time spent may be subject to criticism (Manning, 1980; Waegel, 1981). However, on the basis of the observations, it is believed that little error in this measure was intentionally introduced by the detectives. While detectives were re- quired to record the amount of time spent on the investigation, the in- formation was not monitored or organizationally reported in any fashion. There was little feedback to the detectives or supervisors on the basis of these statistics. Detectives were not evaluated on this in- formation. However, distortion was probably introduced through memory errorsCwhich resulted when the activities performed and time spent were forgotten before completion of the case logCand cal- culation errorsCwhich resulted from the difficulties of allocating time

American Journal of Police Vol. XIII, No. 3 1994 167

spent on certain activities to particular investigations (e.g., driving time, discussions with other personnel in the department, etc.).

11. When suspect information is kept in its raw form (i.e., 0 through 9), the correlation between suspect information and case status is .45, p < .001.

12. When suspect information is kept in its raw form (i.e., 0 through 9), the correlation between suspect information and case status remains at .41, p < .001.

13. One can rank the impact of variables based on their level of signifi- cance only when the variables are measured on the same scale, as is the case here (Johnson, Johnson, & Buse, 1987).

14. When "strength of suspect information" is coded so that "a name pro- vided by a [victim] guess" is considered moderate rather than weak evidence and "a culprit named in some other way" is considered strong rather than moderate evidence, the results change minimally. In addition, when victim type (individual/business), dollar value of property loss (in dollars), and physical evidence present at the con- clusion of the initial investigation (yes/no) are used as control vari- ables in the analyses, essentially the same results are obtained.

15. Once again, when "strength of suspect information" is coded so that "suspect described by an eye-witness" is considered moderate instead of weak suspect information and "culprit could be named in some other way" is classified as strong instead of moderate suspect infor- mation, the results change minimally. In addition, substantially the same results are obtained when victim type, dollar value of property loss, physical evidence present, as well as weapon used in commis- sion of offense (yes/no), and degree of victim injury (none, minor, se- rious) are used as controls in the analyses on robberies.

REFERENCES

Bloch, P. and J. Bell (1976). Managing Investigations: The Rochester System. Washington, DC: Police Foundation.

Brandl, S. (forthcoming). 'The Impact of Case Characteristics on Detective Decision Making." Justice Quarterly.

Eck, J. (1983). Solving Crimes: The Investigation of Burglary and Robbery. Washington, DC: Police Executive Research Forum.

168 American Journal of Police Vol. XIII, No. 3 1994

(1979). Managing Case Assignments: The Burglary Investigation Decision Model Replication. Washington, DC: Police Executive Research Fo- rum.

Garofalo, J. (1981). "Crime and the Mass Media: A Selective Review of Re- search." Journal of Research in Crime and Delinquency, 18:319-50.

Greenwood, P. (1970). An Analysis of the Apprehension Activities of the New York City Police Department. New York, NY: Rand.

Greenwood, P., J. Chaiken, and J. Petersilia (1976). "Response to: An Evalua- tion of the Rand Corporation's Analysis of the Criminal Investigation Process." Police Chief, 12:62-71.

(1977). The Criminal Investigation Process. Lexington, MA: Heath.

Hanushek, E. and J. Jackson (1977). Statistical Methods for Social Scientists. New York, NY: Academic Press.

Isaacs, H. (1967). "Police Operations: The Apprehension Process." In Presi- dent's Commission on Law Enforcement and Administration of Justice, Task Force Report: Science and Technology. Washington, DC: United States Gov- ernment Printing Office.

Johnson, A., M. Johnson, and R. Buse (1987). Econometrics: Basic and Ap- plied. New York, NY: Macmillan.

Klockars, C. (1985). The Idea of Police. Newbury Park, CA: Sage Publica- tions.

Manning, P. (1980). The Narc's Game: Organizational and Informational Limits on Drug Law Enforcement. Cambridge, MA: MIT Press.

Sherman, L. and B. Glick (1984). The Quality of Police Arrest Statistics. Washington, D.C.: Police Foundation.

Waegel, W. (1981). "Case Routinization of Investigative Police Work." Social Problems, 28:263-75.

Wilson, J. (1975). Thinking About Crime. New York, NY: Vintage.