The Role of Courtroom Workgroups in Felony Case Dispositions: An Analysis of Workgroup Familiarity and Similarity

Abstract

While pleading guilty has become ubiquitous in criminal trial courts, limited research has focused on the plea process and the factors that influence guilty plea convictions. Numerous theoretical accounts of the plea process highlight the importance of the court actors and their interactions. Based on this research, the current study analyzes the impact of courtroom actor familiarity and similarity on the chosen mode of disposition and the time to disposition. The findings demonstrate that similarity among the actors and familiarity between the prosecutor and judge increase the odds of a plea disposition and reduce the days to disposition. However, familiarity of the defense attorney seems to impede on the informal plea process, such that cases are more likely to proceed to trial when the defense attorney is more familiar with the other actors.

According to the latest statistics for felony sentences in state courts, 94 percent of felony convictions are resolved by a guilty plea (Durose et al. 2009). With so few felony cases proceeding to trial, it can be argued that we do not have an adversarial system of justice, as is often assumed. Although pleading guilty has become ubiquitous, scholars have recognized that there is limited research about the process that leads to guilty pleas, especially in comparison to the amount of research devoted to sentencing (Baumer 2013; Bushway and Forst 2013; Johnson et al. 2014; Ulmer 2012).

Numerous law review articles have offered insights into the non-adversarial nature of the criminal court system, and in almost every account of the plea process, the importance of the actors and their interactions is emphasized. For instance, Alschuler (1968, 1975, 1976) dedicates a separate article to each actor's role in plea bargaining. Bibas (2004), in speculating about the “shadow of a trial” argument for plea bargaining, notes the importance of the actors in the chosen mode of disposition and their incentives to plea. These reviews dovetail with the observations of Blumberg (1967) and Skolnick (1967) regarding the cooperative relations among courtroom actors, and the later work by Eisenstein, Jacob, Flemming, and Nardulli concerning courtroom workgroups (Eisenstein and Jacob 1977; Eisenstein et al. 1988; Flemming et al. 1992; Nardulli et al. 1988). Together, this research makes several propositions regarding the organizational structure of criminal trial courts and the actors within them that can potentially explain the predominance of guilty pleas in the criminal court system.

Despite this fairly large body of research connecting courtroom workgroups to informal case processing methods, there are few quantitative assessments of courtroom workgroup interaction and none of these assessments predict the chosen mode of disposition. To this end, the current study explores the extent to which courtroom actor interactions are related to guilty plea convictions. Several components of the courtroom workgroup, including the familiarity and similarity between the actors, are focused on in an effort to capture the level of interaction between the actors in a given case. It is proposed that increased familiarity and similarity between the actors should facilitate cooperation by increasing the likelihood of a guilty plea disposition and decreasing the time to disposition. This study is the first to develop a measure of familiarity, or repeat interaction, in the context of criminal trial cases and is unique in the sense that all three actors are known for each case in the dataset.

The Criminal Court Organization and Its Actors

From an outsider's perspective, criminal trial courts possess a culture that thrives on the constitutional values of due process, justice, and fairness. In accordance with these values, the actors within the criminal court system seek to attain certain ideological goals respective to their positions. For example, judges attempt to be fair and impartial decision makers, prosecutors strive to keep criminals off the streets, and defense attorneys try to provide the best defense for their clients while safeguarding the rights to which their clients are entitled. Considering their unique roles and goals, it is expected that the interactions between attorneys would be adversarial, with judges as mediators between the two. However, Blumberg (1967) and Skolnick (1967), in their seminal articles, recognized the strong tendency toward cooperation among the actors in the criminal court system, and as Blumberg (1967: 19) stated, the inclination to abandon their “ideological and professional commitments” to service the “higher claims of the court organization.”

These “higher claims” are the bureaucratic priorities and administrative concerns of the court as an organization (Blumberg 1967). Criminal courts, and all organizations for that matter, are troubled by uncertainties and inefficiencies in the workplace, and work toward the organizational goals of reducing uncertainties and increasing efficiency (Thompson 1967). As Albonetti (1986) recognizes, there is little control over the criminalization process in the court system, particularly when a case proceeds to trial. In a trial, the verdict is left to an unpredictable jury. Also, trials require substantial investments of time and effort that impede on the efficiency of case processing (Eisenstein and Jacob 1977). The desire to reduce uncertainty, especially among prosecutors and judges, and efficiently process cases leads to the development of norms and routines in case processing (Albonetti 1986, 1991; Eisenstein and Jacob 1977). One of these norms is resolving cases by pleas of guilt (Albonetti 1999; Blumberg 1967; Cole 1970; Skolnick 1967; Packer 1968).

The actors in the system also have their own incentives to resolve a large number of their cases through the plea process. Prosecutors want to reduce their workloads, ensure convictions, and maintain their public image (Bibas 2004; Hessick, III and Saujani 2002). As Albonetti (1999) recognizes, pleading guilty works in the benefit of the government because it offers a way to convict without the uncertainty of a trial. Similar to prosecutors, public defenders are burdened by large caseloads. They often feel pressure from prosecutors and judges to moves cases quickly to resolution and can fear judicial reprisals against clients for taking cases to trial (Alschuler 1975; Bibas 2004; Hessick, III and Saujani 2002). Privately retained counsel has the added financial incentive as well. Finally, judges are concerned about maintaining their reputation (especially if they are elected), reducing their caseload, and minimizing the number of cases that can get overturned on appeal (Alschuler 1976; Hessick, III and Saujani 2002). Together, the pressures of the organizational goals and the incentive structures induce cooperation among the actors that is expected to lead to informal methods of case processing (Bibas 2004; Blumberg 1967; Feeley 1973; Skolnick 1967).

It is argued by some that the reliance on plea negotiations—while beneficial to the court organization, and at times, to the defendant—can create disadvantages in sentencing among those defendants who proceed to trial. There has been some evidence to suggest that defendants are penalized for going to trial by receiving harsher sentences (Johnson 2003; Ulmer and Bradley 2006; Ulmer et al. 2010). This trial penalty can vary across subcategories of offenders (Johnson 2003; Smith 1986; Ulmer and Bradley 2006; Ulmer et al. 2010) and often depends on the probability of acquittal had the defendants who pled guilty gone to trial (Bushway and Redlich 2012; LaFree 1985; Smith 1986). Ulmer and Bradley (2006: 635) note that “researchers explain guilty plea versus trial sentencing differences as the product of courts rewarding those who plead guilty for behavior or attitudes that courts organizationally value” and that “rewarding those who plead guilty and penalizing those who lose at trial reflects the need for efficiency in case processing.”

Courtroom Workgroups and Cooperation

Several scholars have highlighted important aspects of courtroom workgroup interaction that are supposed to make cooperation among the actors easier, and thus facilitate negotiation strategies. According to Eisenstein and Jacob (1977), uncertainty in negotiating is reduced by familiarity among the actors. The more workgroup members are familiar with one another, the better they can negotiate and avoid the formalities of adversarial proceedings. The actors most familiar with one another are the lawyer-regulars, or repeat players, who engage in many similar cases over time and have more opportunities to develop informal relations with the other court actors (Bibas 2004; Blumberg 1967; Eisenstein and Jacob 1977; Galanter 1974). Longer, established relationships between the actors can reduce the likelihood of formal case processing through adjudication and litigation (Eisenstein and Jacob 1977; Galanter 1974; Ulmer 1995).

Studies have quantitatively assessed the effect of repeat pairings in civil cases. For instance, Colvin (2011) created a dummy variable to identify cases in his sample that were repeat employer-arbitrator cases. He found that employers were more successful in arbitration cases involving repeat employer-arbitrator pairings. More closely tied to the propositions of Eisenstein and Jacob (1977), Johnston and Waldfogel (2002) looked at the influence of repeat attorney interaction on cooperation in civil cases, operationalized as the speed of case processing and the tendency to settle. Similar to the strategy used by Colvin (2011), they created a dummy variable to identify cases during the sample in which pairs of lawyers appear together more than once. They considered the actual number of interactions during the sample as well. Their results indicate that cases involving attorneys who interact repeatedly are less likely to be tried and are resolved faster (i.e., the duration between filing and termination is shorter). However, the statistical effect of repeat pairings on criminal case dispositions is still unknown.

In addition to familiarity, Ulmer (1995: 599) recognizes that “robust shared pasts and the identities forged and sedimented through them facilitate negotiation as an interaction strategy.” According to him, “common pasts can also be used to manage uncertainty” (Ulmer 1995: 588). Commonalities or similarities between actors can induce greater cooperation, since people tend to “value the contributions of similar others” (Hinds et al. 2000; Hoskins Haynes et al. 2010). In terms of “common pasts,” prior research has considered whether attorneys in a case went to the same college and/or law school. Johnston and Waldfogel (2002) accounted for this nuance. Unexpectedly, they found that civil cases involving college or law school classmates were more likely to be tried.

Hoskins Haynes et al. (2010) also accounted for college and law school similarity in their study of criminal cases in Pennsylvania. However, they focused on the decision to incarcerate, impose fines, and impose restitution. Their measures of college and law school similarity were at the county level, reflecting the percentage of judges who attended Pennsylvania colleges or law schools if the district attorney attended college or law school in Pennsylvania. Law school similarity was negatively related to the decision to incarcerate and impose restitution, while college similarity was positively related to the decision to incarcerate and impose fines. In addition to college and/or law school similarity, they also considered race, gender, age, and political party similarity. These measures represent the percent of judges in a county that have the same race, gender, age (±5 years), and political party, respectively, as the district attorney in that county. Gender, age, and political party similarity were all negatively related to the decision to incarcerate.

Some additional studies have also considered race representation among workgroups at the district and county level and whether increased minority representation affects minority defendant outcomes. Ward et al. (2009) assessed whether disparities in sentencing outcomes between Black and White defendants are reduced in districts with greater Black representation among judges and prosecutors combined. They found that at lower levels of Black workgroup representation, Black offenders were slightly more likely to be incarcerated compared to White offenders, and vice versa. In a similar fashion, King et al. (2010) looked at whether sentences for Black and Hispanic defendants varied by the percent of Blacks and Hispanics employed as lawyers in a county. They found that defendants had lower odds of an incarceration sentence and received shorter sentences when the county had a higher percentage of Black and Hispanic attorneys. There was also evidence to suggest that racial and ethnic disparities in the probability of incarceration and sentence length decreased as the percentage of Black and Hispanic attorneys in a county increased.

Similar to the former line of research, there are studies that have considered similarities between the judge and the defendant in a case and whether their common pasts influence case outcomes. These studies have an advantage in that they link judges to their respective cases. In an early study, Spohn (1990) found that both Black and White judges sentence Black defendants harsher than White defendants. More recently, Johnson (2006) found that minority judges were less likely to incarcerate Black and Hispanic offenders but sentenced Black offenders to longer periods of incarceration. Unfortunately, scholars have not yet assessed the effect of workgroup similarities among actors in criminal cases on the mode of disposition, as a method of determining whether common pasts among the actors facilitate cooperation and negotiation.

The Plea Process

As previously mentioned, very little research focuses on the plea process. There are no quantitative studies to date that have looked at the direct impact of courtroom workgroup dynamics on guilty plea convictions. Nardulli et al. (1988) come closest in their linear assessment of charge reductions and sentences in plea cases with familiar workgroups, or where the same judge, prosecutor, and defense attorney handled at least five cases together in the county sample. However, several qualitative studies, including the theoretical and descriptive research of Eisenstein and Jacob (1977), Eisenstein et al. (1988), and Flemming et al. (1992), have found evidence of a relationship between workgroup dynamics and plea dispositions. For instance, Ulmer's (1995) interviews and observational work in three Pennsylvania counties revealed that workgroup familiarity and stability condition strong informal case processing norms and heavy reliance on plea negotiations. Not only shared pasts, but also common pasts, seemed to help manage uncertainties in decision making. While this research has been both insightful and important, corroborating quantitative evidence can help in the development of our understanding of courtroom workgroups (Maruna 2010).

Existing quantitative assessments of the mode of disposition are mostly limited to studying the effects of case and defendant characteristics. These studies suggest that plea and sentence negotiations are less likely when the offense is more serious and the offender has a more extensive prior record (Albonetti, 1990; Frenzel and Ball 2008; Meyer and Gray 1997). Blacks seem to be consistently less likely to enter into both negotiated and non-negotiated pleas and would rather take their cases to trial (Albonetti 1990; Frenzel and Ball 2008; Meyer and Gray 1997). Cases in which there are larger numbers of witnesses and the accused used a weapon and/or was detained are more likely to go to trial, while physical evidence and confessing to the offense increases the likelihood of a plea of guilt (Albonetti 1990).

The Current Study

The current study is designed to explore the relationship between courtroom workgroup interaction and guilty plea convictions. It is expected that higher levels of familiarity and similarity between the judge, prosecutor, and defense attorney in a case will induce greater cooperation. This nuance will be reflected as an increase in plea dispositions and a decrease in the time to disposition in cases that involve more familiar and similar court actors. This is the first study to derive a measure of familiarity among actors in criminal cases. The data used is also unique in that the judge, prosecutor, and defense attorney can be identified for each criminal case.

While qualitative research of courtroom workgroups has its advantages, including the “awareness of situational and contextual factors and concerns that are often missed in survey research,” quantitative research also has its own strengths to contribute to the courtroom workgroup literature (Maruna 2010: 127). Often, quantitative research is seen as more transparent, replicable, objective, and generalizable. It is argued that “statistical techniques allow for the elimination of confounding influences and better assess cause and effect relationships among variables” (Maruna 2010: 128). Therefore, it is advantageous to test theoretical propositions both qualitatively and quantitatively to determine (1) if the findings from both types of studies corroborate and (2) how each type of research can inform the other. While the primary goal of the paper is to use quantitative methods to assess courtroom workgroups, interviews from a judge, prosecutor, and defense attorney in the courthouse under study are incorporated to help explain and expand on the statistical findings.

Data

The data were collected from public defender case files in one of the larger counties in FL. Case, defendant, and attorney information was gathered for 500 felony plea cases and 411 felony trial cases using a case-control sampling design. Trial cases were oversampled to allow for greater variation in the mode of disposition given that case disposition is virtually a constant in most data sets. The sample of trials actually represents the population of closed cases in the public defender's office initiated between 2002 and 2010 that were disposed of by a trial, while the sample of pleas composes a random sample of 500 closed cases in the public defender's office initiated between 2002 and 2010 that ended in either a plea of guilty or a plea of no contest.¹

The sampling frame was developed by the records department in the public defender's office. For the trials, a list of case numbers was provided for all felony trial cases initiated from 2002 to 2010 that were closed. All the trials in the list were included as part of the data collection. Cases in which the defendant went to trial on some counts and pled guilty on others were excluded so that all cases in the data set reflect a single chosen mode of disposition. The case information was coded on a case-by-case basis by accessing the case records through the public defender office's computer database. Once the trials were collected, the records department provided a list of all felony case numbers—pleas and trials—initiated from 2002 to 2010 (N = 32,832). To obtain a sample of pleas, a random sample of 500 case numbers was drawn from this list. Within the initial random sample, there were several trial cases that had already been coded. These cases were substituted with randomly drawn replacements until a sample of 500 pleas was obtained. The case level information was once again coded on a case-by-case basis using the computer database.² After eliminating cases with missing data, the final sample size is 905 cases—408 trials and 497 pleas.

The Court Community

The cases were handled in a circuit courthouse in FL and represent cases from a single county within that circuit. In the public defender's office, cases were usually assigned based on offense type to a group of felony public defenders, but the felony division has changed over the years. For instance, the public defender interviewed indicated she was originally paired with one prosecutor, but at the time of the interview, the division had recently decided to randomize pairings with prosecutors. Attorneys were typically assigned to work regularly with a particular judge. The turnover in each of the offices, assignment to new positions, and election of new judges could easily alter this dynamic. The public defenders were responsible for handling both pleas and trials. Particular weeks within the court calendar were designated for trials, and judges typically alternated between trial weeks and hearing weeks. Within the dataset, the same prosecutor, defense attorney, and judge handled the case after arraignment—during both the hearings and into sentencing—with a few exceptions. In about 1.9 percent, 1.1 percent, and 1.7 percent of the convictions (n = 748), the prosecutor, defense attorney, and judge changed during the sentencing phase, respectively.

From 2002 to 2005, there were 15 judgeships in the circuit court, and from 2006 to 2010, there were 16 judgeships. The number of circuit judges varies from year to year depending on the population and caseload of the area. They are elected in nonpartisan contested elections and serve for 6 year terms. The state attorney and public defender are also elected, but in partisan elections, and serve 4 year terms. Elections for circuit judges were held in 2002, 2004, 2006, 2008, and 2010. Elections for the State Attorney and Public Defender were held in 2004 and 2008. In both elections, the Public Defender went uncontested, while the State Attorney only went uncontested in 2008. The cases in the data set were handled by 66 different prosecutors, 118 different defense attorneys, and 36 different judges.

During the years of the study, the courthouse operated within a fairly liberal circuit, with only about 24–25 percent of the population in the circuit voting Republican. There were six prisons within the circuit that could house up to 8,328 inmates at maximum capacity. Of the institutions, there is only one female prison, and one of the six prisons did not open until the summer of 2005. There were two significant changes related to sentencing laws over the time period. Beginning in July 2001, the responsibility for preparing guidelines and criminal punishment code scoresheets for felony defendants was placed solely with the State Attorney's Office. The scoresheet numerically records the level of the offender's crime and is used at the offender's sentencing. In 2005, penalties were enhanced for sexual crimes against children and lewd and lascivious molestation of a child.

Variables

Dependent Variables

Similar to Johnston and Waldfogel (2002), cooperation is operationalized in two ways. First, the tendency to settle is captured in the mode of disposition. Plea Disposition is coded such that cases disposed of by a plea of guilty or no contest are coded 1, while cases disposed of through a jury or bench trial are coded 0.³ As denoted in Table 1, 54.9 percent of the cases were resolved by a plea of guilty or no contest and 45.1 percent of the cases were resolved by a bench or jury trial. These percentages are a reflection of the case-control sampling strategy utilized.

Table 1. Descriptive Statistics

Abbreviations: s.d., standard deviation; Min, minimum; Max, maximum.

Second, the speed of case processing is reflected as the Time to Disposition, which is the total number of days from the day the defendant was arrested to the day s/he was declared guilty or not guilty (either through a plea or after the culmination of a trial). The natural log of this measure was used to create a more normal distribution. Lower values indicate greater efficiency and cooperation in case processing. The average logged number of days to disposition among the cases in the data set was 5.063.

Key Independent Variables

Based on prior research, two aspects of court actor interaction are considered: familiarity and similarity. The degree of familiarity and similarity between the actors is not only measured for the full workgroup in each case, but also for the (1) prosecutor and defense attorney pairing and (2) prosecutor and judge pairing. This strategy is employed since the early phases of case processing are heavily dependent on the prosecutor and his or her affiliations with the other actors (Bushway and Forst 2013; Cole 1970; Eisenstein and Jacob 1977; Skolnick 1967).

There are two different measures of familiarity utilized. The first measure is continuous and reflects the number of interactions between the workgroup or pair in a case. The first instance of an interaction is coded as 0. In the next interaction between the workgroup or pair, the case is given a 1. This pattern is continued for all subsequent interactions (2, 3, 4, and so forth). Essentially, these measures reflect a count of Workgroup Familiarity, Prosecutor-Defense Familiarity, and Prosecutor-Judge Familiarity. On average, the same workgroup came in contact about twice, the same prosecutor and defense attorney came in contact about three times, and the same prosecutor and judge came in contact about four times within the data set (these values take into consideration that the first encounter is coded 0).

Similar to Colvin's (2011) study, the second measure is a dummy variable, such that cases in which all three lawyers appeared together more than once are coded 1 and 0 otherwise to represent Workgroup Familiarity. Two corresponding measures are used to reflect Prosecutor-Defense Familiarity and Prosecutor-Judge Familiarity, whereby cases in which the prosecutor and defense attorney appeared together more than once are coded 1 and 0 otherwise, and cases in which the prosecutor and judge appeared together more than once are coded 1 and 0 otherwise. Colvin (2011) recognizes scholars who have argued that the first instance in which the workgroup or pair appears together should not be classified as a repeat interaction. Therefore, the first instance of an interaction (1) among the workgroup, (2) between the prosecutor and defense attorney, and (3) between the prosecutor and judge is coded 0. As indicated in Table 1, about 32 percent of the cases involved a repeat workgroup, while about 51 percent and 69 percent of the cases involved a repeat prosecutor-defense pairing and a repeat prosecutor-judge pairing, respectively.

It is important to note that these familiarity measures likely underestimate the full extent of the interactions that took place, especially considering that the study only includes a sample of cases that were initiated between 2002 and 2010, not the full population. The dummy version of the measure can account for some of this underestimation, but it assumes familiarity based on two interactions. As stated by Johnston and Waldfogel (2002: 48), these measures of familiarity should be viewed “not literally but rather as noisy measures of the true frequency of joint appearance.” Because of the possible underestimation, the current study is biased against finding an effect of familiarity on the chosen mode of disposition.

Four similarities among the actors are taken into account. The first is law school similarity. To capture Workgroup Law School Similarity, cases in which all the actors within the workgroup went to the same law-school are coded 1. In addition, cases in which the prosecutor and defense attorney went to the same law-school are coded 1 and 0 otherwise to represent Prosecutor-Defense Law School Similarity, and cases in which the prosecutor and judge went to the same law-school are coded 1 and 0 otherwise to represent Prosecutor-Judge Law School Similarity. In 17.9 percent of the cases, the workgroup went to the same law school, while in 27.5 percent of the cases the prosecutor and defense attorney went to the same law school and in 38.3 percent of the cases the prosecutor and judge went to the same law school.

The second is race similarity. Once again, dummy variables are created to designate whether the actors within the workgroup were the same race (1 = the workgroup was all Black or all non-Black and 0 = the workgroup was a combination of Black and non-Black participants), whether the prosecutor and defense attorney were the same race (1 = both actors were Black or both actors were non-Black and 0 = one actor was Black and one actor was non-Black), and whether the prosecutor and judge were the same race (1 = both actors were Black or both actors were non-Black and 0 = one actor was Black and one actor was non-Black). These measures are denoted as Workgroup Race Similarity, Prosecutor-Defense Race Similarity, and Prosecutor-Judge Race Similarity, respectively. The workgroup was the same race in 85.3 percent of the cases, while the prosecutor and defense attorney were the same race in 86.7 percent of the cases and the prosecutor and judge were the same race in 95.9 percent of the cases.

The third is sex similarity. Workgroup Sex Similarity designates whether the workgroup in a case was all the same sex (1 = the workgroup was all male or all female and 0 = the workgroup was a combination of male and female participants). Additionally, Prosecutor-Defense Sex Similarity denotes whether the prosecutor and defense attorney in a case were the same sex (1 = both actors were male or both actors were female and 0 = one actor was male and one actor was female), and Prosecutor-Judge Sex Similarity indicates whether the prosecutor and judge in a case were the same sex (1 = both actors were male or both actors were female and 0 = one actor was male and one actor was female). In about 38 percent of the cases, the workgroup was the same sex. The prosecutor and defense attorney were the same sex in 56.2 percent of the cases, and the prosecutor and judge were the same sex in about 60 percent of the cases.

The fourth similarity measure is experience similarity. Prior research has considered the age and/or experience of the court actors in affecting case processing (Bibas 2004; Eisenstein and Jacob 1977; Hoskins Haynes et al. 2010; Johnston and Waldfogel 2002). Eisenstein et al. (1988) also recognized that large generation gaps between judges, prosecutors, and defense attorneys can inhibit communication. The years of experience for the prosecutor, defense attorney, and judge in each case were calculated by subtracting the year the court actor was sworn into the FL bar from the year the case was resolved. The absolute value of this subtraction was utilized so that larger values represent a greater difference in experience between actors. Because of the nature of the calculations, the resulting variables are actually measures of dissimilarity. The difference in years of experience between the prosecutor and defense attorney is denoted Prosecutor-Defense Experience Difference and the difference in years of experience between the prosecutor and judge is denoted Prosecutor-Judge Experience Difference. The average difference in experience between the prosecutor and defense attorney was about 11 years and the average difference in experience between the prosecutor and judge was about 18 years.

Workgroup Experience Difference was calculated by adding the absolute values of the differences in years of experience between the (1) prosecutor and defense attorney, (2) judge and prosecutor, and (3) judge and defense attorney. Once again, greater values reflect a greater difference in experience among the workgroup members. For example, if the prosecutor and defense attorney had a difference in experience of 5 years, the judge and prosecutor had a difference in experience of 10 years, and the judge and defense attorney had a difference in experience of 8 years, the total Workgroup Experience Difference would be 23 years. The average difference in experience among the full workgroup was about 43 years.

As a final measure, all four aspects of similarity were combined into indices. Because the experience measures reflect differences, the indices were also designed to capture dissimilarity. The measures of law school, race, and sex similarity were recoded to reflect this dissimilarity. These measures, along with the experience difference measures, were combined into three mean standardized indices, and are denoted Workgroup Dissimilarity Index, Prosecutor-Defense Dissimilarity Index, and Prosecutor-Judge Dissimilarity Index. These measures are designed to capture the cumulative impact of working with generally different people (i.e., larger experience gap, different law schools, different race, and different sex).

Control Variables

Several confounding factors are taken into account. Offense Level is a score given to each case based on the severity of the primary (or most serious) charge against the defendant in the case. Using FL's criminal punishment code, the level of the offense was determined based on the statute the offense violated. The average offense was at a level of about 5, with 10 being the highest offense level. Dummy variables were also created to indicate the primary offense type—Violent Offense, Property Offense, Drug Offense, or Other Offense. Violent Offense is used as the reference category. The FL Statutes were consulted for the purposes of grouping the offenses into each of these categories.

Prior Record is a measure of the defendant's criminal history. The court from which the data were collected had a pre-sentence investigation interview of the defendant in the file of the case. Defendants were given a 3 if they had no prior convictions, 0 if they had misdemeanor convictions, −1 if they had a case pending or more than one failure to appear, −2 if they had prior felony convictions, and −2 if they had been incarcerated in prison within the past 5 years. As a result, scores ranged from −5 to 3. For instance, if a defendant had a Prior Record of −5, s/he had a case pending or more than one failure to appear, prior felony convictions, and a prison stay in the past 5 years. For ease of interpretation, the scale was recoded so that 1 is equivalent to a score of 3 and 8 is equivalent to a score of −5. Therefore, Prior Record ranges from a score of 1–8, with higher values representing a more extensive prior record.

Number of Counts is a continuous variable indicating the total number of counts against the defendant in each case. If the case involved an offense where the defendant used a Weapon, the case is coded 1, otherwise, the case is coded 0. Since Prior Record does not take into account whether the defendant was on probation or parole, a dichotomous variable was created to indicate whether the defendant was on probation or parole at the time of the arrest. If the defendant was on Probation/Parole, the case is coded 1 and 0 otherwise. The last legal variable accounted for is whether the defendant was detained in the case. Cases in which the defendant was Detained in jail prior to disposition are coded 1, while cases where the defendant bonded out, received a pretrial release, or was released on recognizance (ROR) are coded 0.⁴

Additional extralegal variables are identified as potential confounders. The court overwhelmingly coded defendants as either Black or White. Because of the possibility that some of the defendants coded White could be Hispanic, race is dichotomized as Black (1) and non-Black (0).⁵ Sex is also taken into consideration, with cases involving Male defendants coded 1 and those involving female defendants coded 0. Age is a continuous measure of the defendant's age in years. Lastly, dummy variables were created for the years the cases were resolved with 2002 as the reference category.⁶

Analysis

Because cases are nested within workgroups and pairs, random intercepts models are utilized, such that the intercepts in each of the models are allowed to vary across workgroups and pairs. Hierarchical logistic regression is used for the Plea Disposition models and hierarchical linear regression is used for the Time to Disposition models. Results are reported separately for the full workgroup and the actor pairings. The 905 cases in the sample are nested within 619 workgroups, 441 prosecutor-defense pairings, and 279 prosecutor-judge pairings. Multi-level modeling accounts for the cases that were handled by the same workgroup or pairs and may have correlated residual errors (Britt 2000; Ulmer and Johnson 2004; Ulmer et al. 2010). As noted by Ulmer and Johnson (2004: 152), multi-level models also adjust “the degrees of freedom to correctly represent the number of level 2 units” and overcome “the aggregation bias that can occur when a variable takes on different meanings at different levels of analysis.” Similar to these prior studies, all variables are centered at the grand mean to appropriately assess the impact of the cluster-level variables (Enders and Tofighi 2007; Raudenbush and Bryk, 2002).

Three different models are reported to predict Plea Disposition and Time to Disposition. In Model 1, the count versions of the familiarity measure are used. Model 2 incorporates the dissimilarity indices, as opposed to the separate measures of law school similarity, race similarity, sex similarity, and experience difference. Model 3 reports the dummy versions of the familiarity measure. As previously noted, each of the models includes dummy variables for the year the case was resolved, but these variables are not reported in the results tables shown for a clearer presentation of the results. Year dummies are used to account for any external influences in a particular year that could have affected case processing in that year.

Findings

Table 1 reports the descriptive statistics for the measures disaggregated by the mode of disposition. Contrary to expectations, the means of the Workgroup Familiarity measures seem to suggest that more familiar workgroups take their cases to trial. The count of Prosecutor-Defense Familiarity indicates otherwise, but the standard deviation around the mean for the plea cases shows substantial variation, and the dummy version of the measure once again suggests that more familiar pairings are likely to proceed to trial. In accordance with the theoretical propositions, familiar prosecutors and judges do seem to rely more on plea convictions. Overall, the descriptive statistics for the similarity measures follow expectations as well, such that more similar workgroups are more likely to plea.

Tables 2-4 present the hierarchical logistic regressions and hierarchical linear regressions analyzing the effects of workgroup, prosecutor-defense, and prosecutor-judge familiarity and similarity on the mode of disposition and the time to disposition. The confounders have consistent effects across the models. Cases involving more serious offenses and defendants with more extensive prior records were less likely to be resolved by a plea. Given that offense seriousness decreases the likelihood of a plea, it is not surprising that cases where the defendant committed a property, drug, or other crime were more likely to be pled than cases where the defendant committed a violent crime. Cases involving older, Black, male defendants were less likely to result in a guilty plea conviction, while cases with defendants who were on probation/parole when arrested and who were detained had greater odds of being disposed by a plea. With regards to the Time to Disposition, cases that were (1) resolved by a plea, (2) involved a property, drug, or other offense, and (3) had a defendant who was detained were brought to resolution quicker, while more serious cases and cases with a larger number of counts experienced a longer time to disposition.

Table 2. Hierarchical Models of Plea Disposition and Time to Disposition on Workgroup Familiarity and Similarity

Notes: n = 905; Models include controls for the year the case was resolved, with 2002 as the reference category; All variables were grand-mean centered; Unstandardized coefficients are reported.

^* p ≤ 0.05, ^**p ≤ 0.01, ^***p ≤ 0.001

Table 3. Hierarchical Models of Plea Disposition and Time to Disposition on Prosecutor-Defense Familiarity and Similarity

Notes: n = 905; Models include controls for the year the case was resolved, with 2002 as the reference category; All variables were grand-mean centered; Unstandardized coefficients are reported.

^* p ≤ 0.05, ^**p ≤ 0.01, ^***p ≤ 0.001.

Table 4. Hierarchical Models of Plea Disposition and Time to Disposition on Judge-Prosecutor Familiarity and Similarity

Notes: n = 905; Models include controls for the year the case was resolved, with 2002 as the reference category; All variables were grand-mean centered; Unstandardized coefficients are reported.

^* p ≤ 0.05, ^**p ≤ 0.01, ^***p ≤ 0.001

Table 2 reports the results pertaining to the workgroups. In the unconditional model for Plea Disposition (not reported), the variance (3.228) and standard error (1.008) suggest that there is significant variation in the intercept across workgroups, justifying the muli-level approach. By adding the case, defendant, and workgroup predictors to the models, about 74 percent to 81 percent of the variation in the intercept is explained, depending on the model. In Model 1, the pattern in the descriptive statistics is confirmed, such that greater familiarity among the workgroup decreased the odds of a plea by about 16 percent (OR = 0.842, p ≤ 0.05). This relationship remains consistent across the models, with familiarity among the workgroup decreasing the odds of a plea disposition by about 16 percent in Model 2 (OR = 0.844, p ≤ 0.05) and about 36 percent in Model 3 (OR = 0.641, p ≤ 0.05). As shown in Models 1 and 2, cases handled by a workgroup that was the same sex had greater odds of being resolved by a plea of guilt or no contest (about 72 percent and 69 percent, respectively), while a greater difference in experience among the actors decreased the odds of a guilty plea conviction by about 3 percent. While law school and race similarity alone did not have a significant impact on the mode of disposition, Model 2 shows that cases involving a generally dissimilar workgroup were more likely to proceed to trial (OR = 0.522, p ≤ 0.001).

The interclass correlation for the unconditional model predicting Time to Disposition (not reported) revealed that about 46 percent of the variance is due to differences between workgroups and 54 percent of the variance is due to within workgroup variation (.313/(.313+.360)). The inclusion of the case, defendant, and workgroup predictors explains about 58 percent to 62 percent of the variation in the intercept. While familiarity decreased the time to disposition, as would be expected, the effect of familiarity is not significant. However, cases involving a workgroup of similar sex came to a significantly quicker resolution (b = −0.236, p ≤ 0.001), and cases involving a generally dissimilar workgroup were resolved significantly slower (b = 0.074, p ≤ 0.01), as would be predicted.

Table 3 focuses on the prosecutor and defense attorney pairings. Similar to the workgroup models, the variance (3.707) and standard error (0.905) of the unconditional model for Plea Disposition (not reported) suggest that there is significant variation in the intercept across prosecutor-defense attorney pairings. Addition of the case, defendant, and group predictors explains about 30 percent to 44 percent of this variation. The findings in Models 1 and 3 demonstrate that having a prosecutor and defense attorney who were the same sex substantially increased the odds of a plea disposition (OR = 1.994, p ≤ 0.05, and OR = 1.950, p ≤ 0.01). Unlike the workgroup models, differences in experience between the prosecutor and defense attorney did not affect the mode of disposition, and general dissimilarity between the prosecutor and defense attorney did not influence whether the case was resolved by a plea or trial. As a count, familiarity did not have an impact on the mode of disposition, but the dichotomous measure used in Model 3 suggests that cases involving a familiar prosecutor and defense attorney had lower odds of being disposed by a plea (OR = 0.448, p ≤ 0.001). Similar to the workgroup models, this result is unexpected based on the theoretical propositions.

According to the unconditional model for the Time to Disposition (not reported), 40 percent of the variance is due to differences between prosecutor-defense attorney pairings and 60 percent of the variance is due to within prosecutor-defense attorney variation (0.251/(0.251 + 0.370)). The case, defendant, and group factors explain about 51 percent to 54 percent of this variation. Across all three models, familiarity between the prosecutor and defense attorney in a case decreased the time it took to dispose of the case (b = −0.022, p ≤ 0.01, b = −0.024, p ≤ 0.01, b = −0.089, p ≤ 0.05, respectively). Also, sex similarity between the prosecutor and defense attorney significantly reduced the time to disposition (b = −0.157, p ≤ 0.01, b = −0.163, p ≤ 0.01, respectively). Law school similarity, race similarity, difference in experience, and general dissimilarity between the prosecutor and defense attorney did not have an impact on the speed of case processing.

Lastly, Table 4 reports the results concerning the prosecutor and judge pairings. Similar to the previous models, the unconditional model for Plea Disposition (not reported) has a variance (2.098) and standard error (0.580) that indicates significant variation in the intercept across prosecutor-judge pairings. The case, defendant, and group factors are able to explain about 60 percent to 62 percent of this variation. Unlike the other two tables, sex similarity between the prosecutor and judge did not influence the manner in which the case was resolved, along with law school similarity and race similarity. However, cases handled by a prosecutor and judge with a greater difference in experience were less likely to be disposed by a plea (OR = 0.937, p ≤ 0.001 and OR = 0.940, p ≤ 0.001), and having a generally dissimilar workgroup decreased the odds of a guilty plea by about 45 percent (OR = 0.554, p ≤ 0.001). Similar to the results for the prosecutor-defense pairings, the count version of the familiarity measure between prosecutors and judges did not have a significant impact on the mode of disposition (see Model 1) but the dichotomous measure did. Model 3 shows that cases involving a more familiar prosecutor and judge had greater odds of resulting in a plea conviction (OR = 1.559, p ≤ 0.05). Unlike the previous familiarity findings, this finding is in line with theoretical expectations.

The unconditional model for the Time to Disposition (not reported) once again indicates that there are differences between prosecutor-judge pairings. About 37 percent of the variance in the Time to Disposition is due to these differences, and about 63 percent of the variance is due to within prosecutor-judge variation (0.248/(0.248 + 0.418)). Despite this finding, none of the prosecutor-judge familiarity and similarity measures are significant predictors of the time to case disposition. However, inclusion of the case, defendant, and prosecutor-judge measures did explain about 63 percent to 65 percent of the variation in the intercept.

Supplementary Analyses

Two alternate specifications of the models were examined. First, an additional model for each of the tables was analyzed whereby the first instance in which the workgroup or pair appeared together was classified as a repeat interaction (coded 1 instead of 0). The results were identical to those reported in Model 3 of Tables 2-4. Second, weights were added to adjust for the sampling design. When considering the sampling strategy utilized, there is evidence to suggest that the intercept may no longer be a valid estimate (King and Langsche 2001, 2002; Halipern and Wiintainer 2003). As a precaution, the models were re-estimated with weights that adjusted the sample to reflect the actual proportion of cases that were resolved by a plea or trial during the 2002–2012 time period. It has been recognized that weighted and unweighted estimates may differ when the sampling rates depend upon the outcome, as is the case here (Korn and Graubard 1995: 24). Ultimately, weighted estimators tend to be more variable than unweighted estimators (Korn and Graubard 1995).

The models that account for the weights were re-analyzed using logistic and OLS regression, accounting for potential clustering among the workgroups and pairs. This was done since there were not separate level 1 and level 2 weights to apply to the hierarchical models. With hierarchical modeling, the weights used are designed to adjust for cluster sampling (i.e., sampling workgroups and then cases within workgroups), which was not the case here. The model comparisons are reported in the Supporting Information Online Appendix. Supporting Information Table A1 shows the differences across the Plea Disposition models. Overall, the results were substantively similar, however, there were some differences. In the prosecutor-defense models, the coefficients for sex similarity were no longer significant at the 0.05 significance level. Also, the dummy version of prosecutor-judge familiarity became non-significant in the weighted models. Supporting Information Table A2 presents the differences in the Time to Disposition models. Once again, the results were substantively similar, with some exceptions. In Model 2 of the workgroup analyses, familiarity among the actors became a negative and significant predictor of the days to disposition. In addition, experience differences among the actors in the workgroup models and prosecutor-judge models became significant and positive predictors of the time to disposition at the 0.05 significance level.

Interviews with Court Actors

Interviews with a prosecutor, judge, and defense attorney within the county studied were conducted prior to data collection. These interviews add some context to the statistical findings. Overall, the attorneys and judge emphasized the necessity of the plea process to handle the large amount of caseload. The defense attorney noted that at one point in time, she was working through about 85 cases at once. The judge was clear in recognizing the importance of cooperation between the offices in facilitating plea negotiations. In reference to prosecutors and defense attorneys, the judge said,

I think when they don't get along, it has an impact on the overall resolution of their cases, which impacts me. In other words, if they're not able to come to a negotiation on any of their cases, all of sudden we have a bunch of cases that have to go to trial and frankly the system can't handle everything being tried. We just don't have time to do that.

The public defender also mentioned that it was beneficial to “play nice” with the prosecutors because it would not do the client any good “to get into a pissy match with the prosecutors.”

Despite this recognition, there seemed to be some animosity between the prosecutor and public defender offices. The prosecutor mentioned that they got along with public defenders but they are “kind of a pain.” He called them “believers” who were trying to “stand against the injustice and the imperialistic society that we live in.” He said, “we damn the poor and they're here to help us. They call themselves the public defenders, but we defend the public, they defend criminals.” The public defender shared similar sentiments regarding prosecutors. She mentioned that “sometimes you can develop a rapport with the prosecuting attorney if you feel they've always done things correctly and they've always shot it to you straight. But, sometimes you'll get one that you feel like they've thrown you under the bus for no good reason, and then all of a sudden there's a reason to be a little bit leery about what they tell you.”

There was also some indication that the plea process was not always smooth and easy between the offices. The public defender mentioned that a particular prosecutor she worked with often did not want to try to initiate plea negotiations again if the initial negotiation was rejected by the judge. In these instances, she would often plea straight to the judge. The public defender also noted that “some prosecutors are kind of passive, aggressive. Some are just, you know, it's just gonna be set for trial and I'm withdrawing all my offers.” Essentially, it comes down to the idea that “the philosophies are different. State attorneys are more prosecutorial. They want to see all criminals go to jail. And public defenders are probably a little more lenient.”

Aside from this disjunction between the two offices, it also appeared that the defense attorney had much less informal relationships with judges than the prosecutor. The defense attorney said that “as far as the judge goes, we don't really have that much of a relationship, except in court…Every once in a while, you'll see them out and about, and you talk, but we never really talk off the record about any cases.” Alternatively, the prosecutor, at one point, referred to some of the judges as his “friends” and that sometimes he feels the need to speak with some of his judge friends when they are not handling a situation correctly in court. Specifically, he was referring to the necessity to let the judge know when s/he was not on the same page as the prosecutor and disrupting negotiations. This coincides with what the public defender mentioned earlier about pleading to the judge. The prosecutor said,

Like some defendants, they won't take your offer, and they'll just plea straight up to the judge…So we'll recommend 30 years, they'll recommend probation, and the judge will give them 8…Well why would you negotiate with us ever if you go to the judge and he's gonna give you a better deal?

Ultimately, while each office recognized the need to work with the other office, there were clearly distinct ideologies in the courthouse under study. The prosecutor and defense attorney established these differences and how it could be difficult to overcome some of those differences. In Ulmer's (1995) study of social pasts in criminal courts, he recognizes one of the counties that had “robust shared pasts of ideological conflict” (598). In this context, open pleas were more common and their shared pasts only “mobilized conflictive strategies and coalition formation” (Ulmer 1995: 598). Based on the interviews, it appears that the courthouse studied may have encountered similar issues between prosecutors and defense attorneys that could lead to conflict instead of negotiation. In this instance, their common pasts could ease some of the disagreement and help with negotiation, as the findings seem to indicate.

Discussion and Conclusion

The current study was designed to assess the impact of courtroom actor interaction on the mode of disposition. It was suggested that increased familiarity and similarity between the actors should facilitate cooperation, which would be manifested in an increase in guilty plea dispositions versus trial dispositions and shorter time to disposition. The results demonstrate that the level of interaction among the court actors in a given case is an important factor to consider in case processing. There are two main conclusions that can be drawn from the findings presented.

First, similarity among the court actors influenced the method of case resolution. In particular, sex similarity consistently emerged as a significant predictor, such that it increased the likelihood of a plea and decreased the time to case disposition. Also, a greater difference in experience and a general dissimilarity among the actors seemed to impede on informal and efficient case processing. Therefore, as theorized, certain commonalities between the actors facilitated cooperation, such that common pasts could have helped manage uncertainties and increase efficiency (Hinds et al. 2000; Hoskins Haynes et al. 2010; Ulmer 1995). As Eisenstein et al. (1988) suggest, it also appeared that similar levels of experience among the actors helped with communication and negotiation.

Second, it is evident that increased participation of the defense attorney may actually serve as a barrier against informal case processing. While there was some evidence to suggest that familiarity between the prosecutor and judge increased the likelihood of a case being resolved by a plea, this was not true when familiarity with the defense attorney was taken into consideration. The relationship between the prosecutor and judge follows theoretical expectations. Repeat interaction between the judge and prosecutor seemed to create an understanding between the two parties conducive to an informal case processing system (Eisenstein and Jacob 1977; Galanter 1974). Considering that an argument could be made that these two actors have the most to lose by going to trial—in terms of their reputation—this finding may not be surprising. However, a familiar defense attorney seems to impede upon the plea process, which is not theoretically predicted. Although, familiarity between the prosecutor and defense attorney did lead to more efficient case processing, which would be expected.

It is not exactly clear why these findings regarding familiarity with the defense attorney emerged. It could be that less familiar defense attorneys may feel the need to go along with the judge and prosecutor as a measure of good will that can help them in their future cases, and more familiar defense attorneys do not feel this pressure (Hessick and Saujani 2002). Also, the sample includes cases represented by public defenders. Skolnick (1967) suggests that cooperation is sometimes more common with private defense attorneys. In addition, Mather (1988) argues that defense attorneys who are more familiar might be handling more serious or more complicated cases that are less likely to result in a plea. In the data, there is a negative correlation between workgroup familiarity and offense level (r = −0.019) and prosecutor-defense familiarity and offense level (r = −0.131), suggesting that Mather's (1988) argument does not apply here. However, familiar workgroups did handle a greater percentage of the violent crimes (38 percent vs. 34 percent).

The interviews could also add insight into this unexpected relationship. Countless studies have shown the importance of both organizational and environmental components of courthouses in affecting case processing (Johnson 2005, 2006; Johnson et al. 2008; Kramer and Ulmer 2002; Ulmer 1997; Ulmer and Bradley 2006; Ulmer and Johnson 2004; Ulmer et al. 2010, 2011). Because the current study only looks at one courthouse, the influence of organizational and environmental factors could not be considered. Despite this limitation, the interviews provide some context to the environment within the courthouse studied. It appears that there were shared pasts between court actors but that these shared pasts were of ideological conflict, particularly with the prosecutor and public defender offices. This conflict between actors may have impeded the negotiation process. Common pasts, though, could have helped overcome some of that conflict.

Situating these key findings into the broader theoretical understanding of courts as organizations, it becomes clear that the actors are an important component of case processing, as has been emphasized by organizational theorists (Blumberg 1967; Eisenstein and Jacob 1977; Skolnick 1967). Based on the few interviews conducted, it appears that the actors are driven by organizational concerns, such as reducing the caseload, and there is an awareness of the uncertainties that trials create. In terms of the argument that workgroup familiarity and similarity can induce cooperation, there is support for the idea that common pasts can help the negotiation process. The findings regarding familiarity, or shared pasts, are less clear and are not entirely consistent with theoretical expectations. These findings call attention to the importance of considering the courthouse environment and ideologies among the actors. As Ulmer (1995) suggests, the degree to which shared pasts can help manage uncertainties may depend on particular relationships between the court actors that extend beyond the number of interactions.

Based on this assessment, there are several considerations to be made by future studies using an organizational approach to explain the criminal court system. First, there are additional aspects of familiarity that should be taken into account. While this is the first study to quantify familiarity among the court actors in criminal cases, the measures used are far from perfect. Of primary importance is the fact that the measures were derived based on only a sample of cases, and as a result, do not capture all interactions that occurred in the years under study. Therefore, as mentioned above, the study is biased against finding an effect of familiarity (Johnston and Waldfogel 2002).It also true that the familiarity, and even similarity, measures are influenced by the sampling design chosen. Future research may want to consider other sampling designs, such as propensity score matching, and the possibility of using a full population of cases, although court data including the actors is not always easily accessible for a full population. The familiarity measures are also based on the quantity of interactions, not the quality of interactions. As the interviews seem to indicate, this could make a big difference within the plea process. Particularly, it would be important to understand the ideologies of the actors and to determine whether there is conflict among ideologies within the workgroups (Ulmer 1995). Lastly, Eisenstein and Jacob (1977) stressed the importance of workgroup stability. According to them, familiarity depends on the stability of the workgroups themselves, since more interactions will occur when there is less change in workgroup actors. While the familiarity measure may capture some of this stability, the interviews suggest it may be important to develop a separate measure of workgroup stability that captures the level of cohesion, versus conflict, between the group members.

Second, aside from familiarity, there are two additional measurement limitations. The first is the disposition outcome. Johnson (2003) calls attention to how workgroup dynamics and discretion can differ across various modes of conviction, including non-negotiated pleas, negotiated pleas, bench trials, and jury trials. Johnson (2003) confirms that the effects of extralegal factors in sentencing departures vary across the different modes of conviction. These findings suggest that future research should determine whether the findings reported here are robust across these various distinctions. The second measurement limitation is the lack of information pertaining to the evidence in the case. Prior studies have suggested that factors related to the evidence (e.g., number of witnesses, confessing to the offense, and physical evidence) could impact the mode of disposition (Albonetti 1990; Kutateladze et al. 2016). While these measures were not available in the current study, it can be argued that the availability of evidence is often dependent on the seriousness of the case. Therefore, controlling for the offense level could account for some of the effect of evidence on the chosen mode of disposition. Still, future research should consider its impact and whether the workgroup interaction effects remain after controlling for the evidence in the case.

Third, similar studies of courtroom actor interaction should continue to be conducted in other courthouses. The current findings are limited in generalizability because of the focus on cases in a circuit courthouse of one county in FL. Internal operations and external environmental constraints are different across courthouses and regions. For example, the effects found may vary depending on the level of bureaucratization in the courthouse. Dixon (1995) noted that plea negotiations and workgroup discretion may be more prevalent in more bureaucratized courthouse environments. Baumer (2012) has also called attention to the importance of considering the effect of particular policies on case processing, which can vary by state. For instance, Harris and Jesilow (2000) found that Three Strikes Laws have limited the discretion of workgroups and undermined plea bargaining by making it difficult to predict case outcomes. All of these factors can influence workgroup interaction and cooperation.

Finally, the cases in this study come from a public defender's office. Bibas (2004) acknowledges that private defense attorneys have different incentive structures, including the desire to bill more hours. Alschuler (1975) also brings attention to the importance of a defense attorney learning the strength of the prosecutor's case in the plea negotiation process, and that prosecutors may be more willing to disclose their evidence to public defenders than private attorneys. While Alschuler (1975) seems to suggest that this can lead to more plea deals with public defenders, the opposite might actually be true. Because public defenders know more about the case, familiar defenders may feel more confident about the trial prospects, while private defense attorneys, who know less about the case, will be more inclined to negotiate a plea. Therefore, consideration of private defense attorneys may reveal different findings than those presented.

In addition to potential expansions of the current research, the findings also call attention to the importance of considering additional avenues, particularly in reference to workgroup discretion. Focal concerns theory suggests that stereotypes and biases can influence assessments of defendant blameworthiness, defendant dangerousness, and practical constraints of the courthouse (Steffensmeier et al. 1998). While this study focused mostly on organizational measures of familiarity and similarity, it is important to consider actor discretion (Bushway and Forst 2013). Eisenstein et al. (1988) note that attitudes toward punishment and due process differed from person to person and position to position in the courthouses they studied. Pollitz Worden (1995) has also considered where judges fall on the due process/crime control scale, judge's cynicism toward wrongdoers, and judges’ attitudes toward plea bargaining in case processing. Albonetti (1991) argues that judges manage uncertainties by developing patterned responses, and as a result, judges may rely on stereotypes that link extralegal factors to the likelihood of future offending. The attitudes and biases of the court actors can impact the plea process, especially with its informalities, and should be taken into account.

It is also necessary to consider the race and gender of the actors further. While the current study could only differentiate between Black and non-Black actors, recent sentencing research has called attention to the importance of considering both race and ethnicity (Demuth 2003; Kutateladze et al. forthcoming). The lack of an effect of race similarity may be partially due to this nuance. In addition, race similarity between actors and defendants may affect case processing, as Spohn (1990) and Johnson (2006) suggest. Baker et al. (2015) found that white female defendants who shared the same race as the prosecutor in their case perceived the courts to be more procedurally just. It has been suggested that these perceptions could influence the chosen mode of disposition (Albonetti 1990). As discussed, there is also reason to suspect that the racial and gender breakdown of the workgroup may make a difference in case processing (Farrell et al. 2009; King et al. 2010; Ward et al. 2009). Cases involving a predominantly White or a predominantly male workgroup may be handled differently than cases involving a predominantly non-White or predominantly female workgroup.

Lastly, the role of the defendant's race and gender in predicting the mode of disposition should be highlighted. Cases involving Black and male defendants were more likely to be resolved by a trial. This particular finding needs to be explored further. It begs the question as to why so many Blacks and males proceed to trial, even when controlling for offense seriousness and prior record. It is quite possible that Blacks, and particularly Black males, may be getting less of a value for their plea (Abrams 2011; Bushway and Redlich 2012; Piehl and Bushway 2007; Smith 1986). It is also possible, as referenced above, that this finding could be related to the discretion given to court actors and their respective attitudes and biases. Few studies have looked at the “unwarranted disparity in plea bargaining decisions,” but it seems an important avenue for future research (Frenzel and Ball 2008:61; Kutateladze et al. 2016; Kutateladze et al. 2014).