A Decade of Decision Making: Prosecutorial Decision Making in Sexual Assault Cases
Round 1
Reviewer 1 Report
The authors of this paper use 9 years of data on sexual assaults reported to a large police department in the midwestern U.S. to examine prosecutorial decisions to file or reject charges. Results are generally consistent with prior research on this topic, although the authors do add to the body of knowledge by exploring the degree to which characteristics of “the ideal victim” influence charging decisions.
My comments on the various sections of the paper can be found below:
Abstract
· Change “less than” to “Fewer than” in the first sentence.
· Statement that “there is little [research] on what happens after this point (reporting to police)” is inconsistent with the body of research and with the authors’ own literature review (which cites numerous studies on prosecutorial decision making in sexual assault cases.
Introduction and Literature review
· In the introductory paragraph, “In the United States, prosecutors great hold . . .”
· The literature review does not address theoretical perspective on prosecutorial decision making. For example, Albonetti (1987) argues that prosecutors’ decisions reflect an attempt to avoid uncertainty by filing charges only if the crime is serious and the evidence is strong. The focal concerns perspective also has been applied to prosecutorial decisions in sexual assault cases.
· The literature review is dated and includes very few empirical studies published in the last decade. They cite O’Neal’s study of decisions in intimate partner violence cases of sexual assault in Los Angeles, but they ignore the other publications that resulted from the Los Angeles study (including Spohn and Tellis’s book, Policing and Prosecuting Sexual Assault, their recent JQ article on untangling police and prosecutorial decisions, and several other studies by that research team. There also are several recent studies by Pattavina and Morabito, who replicated the Spohn and Tellis study in several additional jurisdictions. Also missing is any discussion of Susan Estrich’s work, Real Rape, although it is included in the reference list.
· P. 2, “could all be called into question when by prosecutors.”
· P. 2, what is the meaning of “a higher-level outcome for prosecution”
· P. 2—“this [failure to charge] leads to the suspect being charged with a small fine . . .” not sure what this is referring to, if charges are not filed the suspect would not receive “a small fine” and fines are not “charged” but imposed
· P. 3 “to explore the decision process of the legal system” Please state more clearly the outcome this study was examining
· P. 3 four stage—“if the sexual assault case was considered true by police officers”—I suspect that the authors are referring to the decision to unfound the report. Might change this to “if the police believed that the allegations were neither false nor baseless”
· P. 3, “Some research says that a stranger sexual assault is investigated more strongly, prosecuted more harshly, and is less rejected by prosecutors.” Change to something like “is investigated more thoroughly, is less likely to be rejected by the prosecutor, and results in the filing of more serious charges . . .”
· I understand why only cases involving suspects age 17 or older were included, but why exclude cases with victim under age 17? Many sexual assaults are perpetrated against young teenagers and there is research showing that cases with younger victims are more likely to be prosecuted.
· Section on “ideal victim characteristics” should discuss Estrich’s notion of “real (i.e., aggravated) rape and the ways in which the law differentiates between aggravated and simple rapes.
· In discussing victim credibility on pp. 4-5, the authors state that “Prosecutors weigh heavily on this factor as it is the main foundation for their cases and if the victim shows a lack a credibility, the whole case begins to crumble away. Less investigating is the impact of other ‘ideal victim’ characteristics . . .” Consider revising as “Prosecutors give considerable weight to this factor, as it is highly correlated with the likelihood of conviction; if the victim’s credibility is called into question, the case may begin to fall apart. There have been fewer investigations of the impact of other ‘ideal victim’ characteristics” . .
Methods
· The authors indicate that “the department of study” did have a specialized sexual assault unit for adult crimes. Was this the police department or the prosecutor’s office?
· P. 7—Federal Bureau of Investigations—Investigation, not Investigations
· The authors indicate that some of their cases were sexual assaults (i.e., unwanted touching), but that these cases were not common. How many of these cases were there and were they eliminated from the analysis? It appears that they were because there is no control for sexual assault in the model. Also related to the type of sexual assault, the authors state that anal, oral and vaginal penetration “will not be looked at separately,” and yet there Table 2 does include these three categories (but does not include sexual assault). Also, the total of all the cases in the three categories of sexual assault listed in Table 2 is 738, but there are only 700 cases in the final data set.
· The authors state that the only cases included in their study are those that “were referred to the prosecutor by detectives.” Were all of thse cases referred following arrest of the suspect or were some of them evaluated prior to the suspect’s arrest (see the research by Spohn and Tellis on the pre-arrest charging process in Los Angeles)?
· How were missing data handled? It is clear from Table 2 that not all of the independent variables had a total of 700 cases. Were cases with missing values on any of the independent variables deleted from the multivariate analysis? (This appears to be the case as there are only 556 cases in the analysis reported in Table 4.)
· Table 2 should list both categories (and the associated descriptive statistics) for the binary independent variables. For example, the Daytime variable should be relabeled as “Time of Assault” with two categories—daytime and nighttime. Witnesses present should have two categories, “yes,” and “no.” Sex pf the victim should include both males and females, and so on.
· Regarding criminal record of the victim and suspect, did the police investigation always include a criminal history check on the victim? Or did you assume that if no criminal history was mentioned in the report that the victim did not have a criminal history? How was criminal history measured—arrest or conviction or something else? On p. 8 the authors indicate that “being a suspect in a case or being arrested was counted.” What is the rationale for this unusual conceptualization?
Results
· Table 2 indicates that column two includes both the % and the SD for each IV. The SD’s are not presented.
· Bivariate results—Table 2 is confusing and does not actually present statistics on the bivariate results between the IVs and the DV. Instead, the authors present the correlation between each category of the IV and the DV. For example, there are three X2 tests for Assault type—one for anal penetration, one for vaginal penetration, and one for oral penetration. Rather than presenting the correlation between type of assault and charging, on other words, the authors test for whether cases involving anal penetration have a higher likelihood of charging—but, a higher likelihood compared to what? Cases involving vaginal penetration and those involving oral penetration? How should one interpret the p values for victim/offender relationship, all of which are statistically significant. The authors state on p. 11 that “all relationship status types individually showed a significant relationship with prosecutor acceptance.” But what does this really tell us? What we want to know if whether relationship type (all three considered together) had a significant effect on charging. Later in the bivariate results section, the authors seem to switch their interpretation of the results; noting that the race (i.e, Black) of the victim approached statistical significance, they then state that “cases with black victims saw higher percentages of prosecution (what does this mean?) compared to white and other race victims.” But this conclusion is based only on the percentages themselves and not on the chi square analysis, which does not address the overall effect of race on charging.
· Multivariate results—I have to question the validity of these results, given the aforementioned issues with the bivariate analysis and the fact that the only IVs include in the MV analysis were those “that were significant at and around a p-value of 0.10.” First, what does “at and around” mean? Second, although they say that they include all variables that were significant in the bivariate analysis, they have distorted the meaning of “variable.” Independent variable victim/offender relationship is measured as romantic partner, other known offender, and stranger. However, because only romantic partner had a p-value of .10 in Table 2, it is the only relationship type included in the MV analysis. In other words, partners are compared to a reference category that includes both acquaintances and strangers. The fact that partners is not significant in the MV results presented in Table 4 may reflect this—cases involving romantic partners may be less likely to result in charging than those involving strangers, but this cannot be deduced from the analysis, as partners are compared to acquaintances and strangers and there is no comparison of partners versus strangers or partner s versus other known individuals. To further exacerbate the problems in the MV analysis, not all of the variables with p-values of .10 or less are included—known offenders, strangers, other residence, injuries present, and under the influence also have p-values of .10 or lower. Were these variables not included because they are captured in the “ideal victim” construct? If so, this should be spelled out in the methods section.
· To address these methological issues, I suggest that the authors:
First, run a MV model that includes all the variables listed in Table 2 and that does not include a variable for the ideal victim characteristics. Do not exclude variables that did not have a p-value of .10 in the bivariate analysis, as variables that were significant in the bivariate analysis may not be significant once all of the variables are included. Similarly, variables that were not significant in the bivariate analysis may have a significant effect on the DV in the MV analysis (you won’t know until you include all of the theoretically relevant variables in the analysis.
Second, replicate the above analysis with the ideal victim scale and without the redundant variables (i.e., those included in the IV scale).
Third, replicate the first analysis without the IV scale but including the individuals indicators of the ideal victim.
I think that this paper has promise, but the authors must attend to (and correct) the methodological problems in this version.
No comments.
Author Response
Please see attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The paper provides an analysis of 9 years of data (which is fantastic) on variables that influence a prosecutor moving forward with a sexual assault/rape case in the USA.
Comments by section:
1. Abstract is clear and concise but would be stronger if more information on the methodology is provided ie sample size, over what time - including the years.
2. Introduction
Line 31+ When you state - "The 30 current study will attempt to uncover the considerations of prosecutors, both legal and extralegal when determining to pursue charges on a sexual assault case." I think it is more honest to say build on existing knowledge - whilst the research is limited it is not non-existent.
2.1 Literature review
The literature review is well written and cites some really alarming statistics - for example, "Cases that involved intimate partners decreased odds of gaining approval for felony charges by 98% compared with cases involving strangers." "Alderden & Ullman (2012), and, Lisak & Miller’s (2002) report on undetected rapists, found that in a sample of college men, approximately 76 men (4.04% of their sample) had engaged in an average of 5.8 rapes each, without any criminal justice intervention."
I didn't find the use of Christie's 'parameters, an “ideal victim” of sexual assault, particularly helpful to frame the analysis, and the heavy focus on the victim over the perpetrator very limiting. I believe that the field and policing/justice has moved beyond the view that - 'the “ideal victim” of sexual assault would be a young virginal woman on her way home from caring for her grandmother who is assaulted by a large man unknown to her'. The role of perpetrator characteristics in prosecuting decisions also needs attention - by focusing on the victim, we continue to perpetrate our patriarchal bias and inequity in our data analysis.
Your findings support this - especially when you tested the reliability of the 'ideal victim' scale. At this stage, I think you should have dropped the ideal victim discourse and used the literature to guide the selection of your variables.
Your methodology would have been much stronger if you had done some key informant interviews with prosecutors - and as a result you might have created a more valid and reliable scale - something which the field needs. Maybe you still can - you have an incredible dataset!
On reading some of the stats and knowing that you have a 9-year data set, I was looking forward to seeing if the stats have changed from what was cited. However, this is not a big part of the discussion.
2. Present Study
It is not clear in the main body of the text what the department of study is. It is in the abstract - but please add it to the paper as well. There is no mention of ethical approval - and/or ethical challenges that needed to be overcome. This must be added.
There is no mention on how you did finally decide on the variables - was it only based on the literature/the ideal victim? Or did the researchers do it based on the raw data?
2.1 Sample
Who were the researchers and how was it that they spent 18 months in the department? Seems ethically and logistically complicated. How were they supported emotionally? Reviewing rape cases for 18 months can be extremely impactful on someone's well-being and health.
Coding the variables - was this done independently and then did researchers come back and decide together? Did you only use the variables listed in the table? And again, how did you decide on the variables in the end?
There are a number of data limitations noted in this section in the footnotes, but they are not noted in the limitations section - and they should be.
4. Discussion
This is where the discourse on 'ideal victim' characteristics is confusing - as ''only three of the ten variables were related to prosecutor decision to accept charges.' And the 'ideal victim' as originally presented does not match the variables used in the analysis.
What was super interesting for me, are the three variables generated through the final model - ie suspected “date rape” drug; delayed reporting; and history of sexual assault charges. Two of which are related to perpetration tactics and perpetration characteristics. I would love to see more discussion on these - especially the use of date rape drugs; how there may be much more of this going on and how prosecutors are hesitant to make charges in cases where date rape drugs are suspected - the reasons for which you outline in the paper - but what can we do about it!?
I also found the finding that DNA evidence and SANE play a limited role in prosecutorial decisions - especially given how newsworthy DNA testing backlogs can be, and how expensive DNA testing is - especially for poorer countries and/or states.
4.2 Policy recommendations
This section is super weak and not evidence-informed.
First, you say that 'deciding to report a rape to police officials is a big step forward in the healing process for many victims'. Please reference this – as making a report to the police may not be a healing step, big or otherwise - but rather a necessary step for survival and/or to access services – health, insurance, etc – and as you note can be - often is - extremely time-consuming, traumatic and lacking in justice.
You also talk about the important role delayed reporting appears to play in the decision process for prosecutors. Delayed reporting is not surprising given how awful it often is to report a rape case. Your recommendation and its focus on the victim having to change their behavior rather than looking at the systemic barriers to reporting – very few rapes are stranger rapes, many women just want the violence to stop, and many women from low SES are often financially dependent on their partners – so we need to really think about the policy implications from a trauma-informed victim-centered view.
I suggest great caution before you make any policy recommendation especially if there is no evidence of their efficacy. Public Announcements can be harmful to women. You are suggesting we adopt PSAs - telling women to report their rape – I don’t know any study on the impact of PSAs on the likelihood of a woman reporting their rape to the police in a timely way or reporting it at all. Do you? If yes, cite it - and using an example of efficacious programs that are not relevant to the current topic is not helpful and at the very least could waste precious public health funds and at the very worst cause harm to women.
Finally, you state - 'While many significant results were found in this study, this is still one of only a limited number of studies investigating prosecutor decision-making in sexual assault cases.' This may be so, but it is hard to prove if only US literature is cited. I am aware of studies from elsewhere in the world – both rich and poor countries - that track justice and the critical role prosecutors play, and the variables that influence them. The paper would be greatly strengthened if you extend your evidence review for evidence beyond just US borders.
The paper is interesting but I think needs more work before it can be published.
The paper needs a light edit as some sentences are unclear - overall the paper is very well written.
Author Response
Please see attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The manuscript is much improved. Kudos to the authors for taking the reviewers' comments and suggestions seriously.
Reviewer 2 Report
Thank you for your thoughtful responses to my review.
