The effect of intensive bail supervision on repeat domestic violence offenders

Most bailbond agents refer to domestic violence suspects as "the bread and butter" of their livelihood. There are two prominent reasons for this claim. First, a large cash bail ($30,000 to $50,000 on average) is typically required for the pretrial release of domestic violence suspects, thus generating extremely high profits for the bailbond agent (about 10% of the full bail amount). Second, because combative spouses are more likely than other offenders to have criminal charges against them dropped by courts, bailbond agents consider domestic violence suspects "safe clients" because they are least likely of all bailees to become an FTA (i.e., failure to appear in court). Hence, the allure of large and easy profits often prompts bailbond agents to obtain the immediate and virtually unconditional release of domestic violence suspects. Unfortunately, bailbond agents seldom consider the consequences that their financially motivated actions will have on the personal safety of domestic violence victims.

Unlike other crime victims, domestic violence victims face unique problems within the realm of cash bail. Specifically, in most jurisdictions in which cash bail is a pretrial release option, suspects charged with domestic violence can use bail to circumvent many pretrial court-mandated programs designed to shield victims from retaliatory or repeat incidences of physical abuse. This problem is noted in the following interview of a California bailbond agent:

A guy could get around all that court stuff like anger mangement or whatever by bailing out early on in the game. What I mean is if a D.V. [domestic violence suspect] bails out really fast they [sic] will go free for about five weeks before the courts will even open up a case on him. Guys who don't make bail and have to stay in will have a case opened on them "pronto" and the court will have a hold of him in about 48 hours. Then there's nothing I can do.

In addition, the unrestricted freedom afforded to domestic violence suspects by the cash bail system increases opportunities for suspects to engage in activities that are potentially dangerous or bothersome to victims: for example, repeat physical attacks, emotional abuse, intimidation, and stalking. According to one female bailbond agent, who deals nearly exclusively in domestic violence cases,

Men I bail out on spousal abuse arrests are really pissed off at their wife or girlfriend and these guys are back on the streets within hours and nobody tells their wife to look out for trouble that's coming back home.

However, bailbond agents possess a very potent, yet seldom-used, legal authority that can be used to counter many of the potentially negative effects that cash bail release may have on domestic violence victims. Namely, bailbond agents have sweeping and unrestricted supervision rights over their clients during the entire pretrial, and trial period (Flemming, 1982). In addition, bailbond agents are authorized to "surrender" or arrest their clients if the contractual (and legal) conditions of bail release have been breached by the bailee (Roth & Wice, 1980). Thus, there is the possibility that bailbond agents could proactively employ their supervision powers as a specific deterrent against future acts of domestic violence committed by their clients.

Hence, from a research standpoint, the problem of cash bail release in the case of domestic violence raises two related questions: First, how, and to what extent, can the existing supervision authority of bailbond agents be expanded to curtail repeat acts of domestic violence committed by bailees? Second, if bailbond agents use their supervision authority in an attempt to protect domestic violence victims, what impact will these efforts have on the incidence and prevalence of repeat domestic violence? The former of the above questions is addressed in the section that follows, and the latter question is the focus of this study.

The Scope of Bail Supervision

Historically, and according to present-day practice, the primary purpose of bail supervision is to ensure that crime suspects appear in court when they are scheduled to do so. The legal foundation of bail supervision is a development of common law and has been called "an ancient and extremely vigorous form of hostageship, which rendered the [bailee] liable to suffer the punishment that was hanging over the head of the released prisoner." (Pollack, p. 589) Federal law and the laws of states that provide for pretrial release via cash bail specify that a bailbond agent has a legal right to supervise "their principal [bailee] at any time, and to this end may pursue and seize him wherever they may find him, even though that may be in another state" (Taylor v. Taintor, 1929). Furthermore, the actions of bailbond agents are not bound by the same legal procedural constraints as are police officers and other public law enforcement officials. For example, warrantless searches and seizures can be executed by bailbond agents whether or not a bailee has committed or is about to commit a criminal offense (Flemming, 1982).

In addition, bail agents have much broader discretion than police and many other public law enforcement officers regarding the decision of when, where, or how to monitor the actions of crime suspects. In most states, bail agents (or their representatives, e.g., "bounty hunters") may supervise an individual at any time, in any place, and by any means that the bailbond agent deems necessary to guard against a suspect's failure to appear in court (Flemming, 1982; Zeisel, 1979). If during the course of the bailbond agent'ssupervision, the suspect's behavior suggests that a failure to appear is imminent, the agent may "surrender" or arrest the suspect without warrants or probable cause.

The supervision of bailees leading to their arrest is a legally sanctioned act so long as the bailbond agent has executed the arrest for "good cause." Recent case law has provided a broad interpretation of acts that constitute good cause for bail arrest. For example, contemporary California law governing bail arrest-which is similar to laws in most states governing the execution of cash bail-stipulates that "good cause for the surrender (i.e., arrest) of the defendant may be established by facts other than the defendant's failure to appear (FTA) in court" (People v. Smith, 1984). Because bail is a private, rather than public, right, the standard applied by courts for demonstrating good cause is far less than that required for "probable cause" in cases of police arrest.

According to recent case law defining the legal scope of bail supervision, bailbond agents possess the authority to proactively supervise bailees regardless of whether or not the bail agent believes that his or her bailee is about to flee or is a "danger to the community." (see, e.g., People v. Smith, 1998). In regard to domestic violence offenders, the law appears to grant bailbond agents the unbridled authority to proactively supervise any and all classes of suspects-regardless of whether or not they are at risk of flight or reoffending. Furthermore, it appears conceivable under current existing case law that bailbond agents may even use their legal powers to make preemptive arrests of bailees exhibiting behavior suggestive of a future act of repeat domestic violence.

In the landmark case governing bail administration the United States, Taylor v. Taintor (1929), the U.S. Supreme Court reaffirmed the broad powers of supervision and arrest granted to bail agents:

The Sureties [i.e., bail agents] whenever they choose to do so may seize him and deliver him up in their discharge and if this cannot be done at once, they may imprison him until it can be done. They may pursue him to another state, may arrest him on the Sabbath, and if necessary Break and Enter his house for that purpose. The seizure is not made by virtue of a new process, none is needed, it is likened to the rearrest by the sheriff for an escaping prisoner. (emphasis added)

Hence, the broad and sweeping powers of bail supervision allow bailbond agents to more closely monitor bailees, especially those who have been arrested for domestic violence. If employed correctly, the somewhat random and unpredictable nature of bail supervision may prove to deter domestic violence reoffending among suspects who are granted pretrial release via cash bail.

However, in the present-day practice of bail, the idea of proactive supervision as a tool for deterring future acts of domestic violence among bailees and their victims is a new one. As the following interviews and observations with practicing bail agents clearly suggest, bail supervision for purposes other than guarding against FTAs simply does not exist in the day-to-day administration of bail. In regard to domestic violence cases, bail supervision is seldom (if ever) employed against domestic violence suspects because their risk of flight is extremely low (for further discussion on this topic, see Wice, 1974).

Cash Bail and Domestic Violence: Voices From the Inside

In the section that follows, "inside information" (see, e.g., Bogdan & Taylor, 1975) obtained from various bailbond agents is presented to gain a preliminary understanding of status quo attitudes and practices within the bail industry toward the supervision and handling of domestic violence cases, suspects, and their victims. This qualitative examination is based on unstructured interviews between the researcher (who was a licensed bailbond agent at the time of study and acting as a participant observer) and active bail agents. The interview field setting included field, office, and jail locations. All information was derived with an objective of purposive sampling. Thus, dialogues and observations were obtained from bail agents and domestic violence cases that were deemed by the researcher to be "typical" of the industry normative standards. The primary focus or research question guiding this preliminary field investigation is, "How do bailbond agents view and carry out their role as the court appointed guardian of domestic violence suspects?"

Qualitative Findings

Interestingly, most bailbond agents appear very knowledgeable about domestic violence cases and are willing to speak quite freely and candidly about how they handle typical domestic violence clients on a day-to-day basis. Perhaps this is because, as previously stated, the domestic violence suspect is a frequent and much sought after client for bailbond practitioners. As one bailbond agent notes:

We all role out the red carpet for Domestics. It used to be that a dude on a "domestic" has a $5,000 or maybe $10,000 bail. But now, after O. J. Simpson almost every county [in California] has jacked the bail up to $50,000 to try to keep 'em in for a while. So Domestics are big money. For a Domestic, I'll let them make payments, I'll drive them where they need to go . . . shit, I'd even lend em' the bailbond premium if I knew they'd pay it back. I wouldn't do anything like that for a hype or some other jacked up knucklehead [sic].

Although apparently motivated to do "whatever it takes" to acquire domestic violence clientele, bailbond agents seem to have little concern for issues that fall outside the realm of profit making. The following statements obtained from two separate bailbond agents illustrate this common industry sentiment:

BAILBOND AGENT X: I can't wait for those hot days in July and August, especially in the evening around 9:00 or 10:00 on the weekend. When the phone rings, I know it's a guy who's got nailed for fight' in with his woman. . . . Marriage and the weather is like the stock market. . . . The hotter it gets, the more money I make from "happy couples" boxing at home.

BAILBOND AGENT Y: You know, but lucky for me, there are folks out there that should not be married. There are men out there who have accounts with me, just like at the liquor store. Everytime they get popped for hitting their wife, I just go out to the jail and get them out. They just keep coming back and filling my pockets.

There is, however, a keen awareness among bailbond agents of the potential dangers that are created by "bailing out" a domestic violence suspect. Many agents express person disdain for these suspects, as well as sympathy for the victims of abusive domestic relationships. The typical bailbond agent's personal reaction to observing the results of a domestic encounter gone bad, according to one male agent, is as follows:

I think I can speak for most bailbondsmen by saying that you never get used to actually seeing the results of a bad domestic fight. All the time women would come into my office and bail out their husband or boyfriend, and sometimes for the second time, and be all bruised and cut up. Everytime I see a woman come through the door wearing sunglasses at night and a scarf covering her face, I know what she's here for. The bastards who do this really have a problem, and maybe some of them go home after getting out of jail and do it again. I think some of them need to stay in jail, but I'm not the judge or jury . . . and they are innocent until proven guilty . . . so it's just my job to bail them out.

Finally, in the practice of bail there is an evident feeling of "helplessness" among bailbond agents regarding the ability of the bail industry to provide some sort of assistance or protection to domestic violence victims. As one bailbond agent suggested, the standard method of screening and issuing bailbonds does not take victim matters into account:

When a bondsman takes an application for bail for domestic violence, most of the time we don't deal directly with the suspect. We usually complete the bond with the suspect's family or friends. Sometimes we see the victim if she's a wife or girlfriend. If things go right, we may never even see the guy we bail out. . . . But our job is to make sure the guy gets out of jail and goes to court during his scheduled court dates. We never see or say anything to the guy we bail out unless he doesn't show up to court. That's standard practice in the business. What can I say or do to stop one of my clients from going back home and beating his wife again? That's a job for a social worker or cop, not a bailbondsman.

As the information presented in the above section indicates clearly, the bailbond profession appears focused on the profits generated by domestic violence suspects; yet some bailbond agents seem aware of and are sensitive to the needs of domestic violence victims. Moreover, from a legal standpoint, bailbond agents ostensibly possess the ability and authority to proactively supervise their clients. However, as information presented in this preliminary examination indicates, bailbond agents rarely (if ever) supervise domestic violence suspects under their care, and they are unaware that such proactive measures are within the legal or practical scope of their occupation.

Police Arrest vs. Bail Supervision

The ability of proactive bail supervision to curb repeat domestic violence falls within the theoretical realm of the specific deterrence doctrine (Berk, Campbell, Klap, & Western, 1992). Briefly, specific deterrence theory specifies that individuals will make a rational choice not to commit crime when it is relatively certain that their untoward actions will result in punishment that is perceived to be swift, certain, and severe (Zimring, 1973). Thus far, most studies of specific deterrence effects on domestic violence have focused on police arrest (see, e.g., Sherman & Berk, 1984). Within the context of police arrest and domestic violence, specific deterrence theory predicts that punishment in the form of arrest (or threat of arrest) by police will serve to reduce the likelihood of repeat violence. Although ample scientific evidence exists to both support and refute the latter proposition (Berk & Newton, 1985), the theoretical importance of police arrest in domestic violence situations lies in what (Sherman & Berk, 1984) refers to as a "swift imposition of a sanction of temporary incarceration" (p. 264).

Unfortunately, the very nature of how police come to arrest quarreling couples calls into question the specific deterrence doctrine's "swiftness" requirement for domestic violence cases. That is, police maintain a highly reactive stance in the discovery of domestic violence incidents. Because of the intimate quality of relations between couples who quarrel, police must rely on reluctant victims, neighbors, and bystanders to report domestic violence incidents. Research based estimates on the number of domestic violence incidents not reported to police range from 30% to 60% (Dobash & Dobash, 1992).

Even when domestic violence incidents are reported, they tend not to be reported in a timely manner. Hours, days, weeks or more may pass before an ongoing pattern or even single incident of domestic violence is brought to the attention of police (Goldberg & Tomlanovich, 1984). Thus, for a characteristically underreported crime such as domestic violence, the ability of police arrest to effectuate a level of punitive cost that outweighs the potential benefits of criminality in the offender's mind is perhaps weaker than similar costs of arrest associated with more highly reported crimes. The following statement obtained from a domestic violence offender interviewed for this study illustrates the aforementioned shortcomings of police arrest:

Shit no, I don't want to get arrested by the cops again if me and my lady have some more trouble. But just because we have a phone, doesn't mean she is going to use it again if we have more trouble. Just because my neighbors have a phone doesn't mean they are going to use it. You see, my lady and my neighbors both know that you can call the cops on me once . . . but you better not let it fuck'in happen a second time!

However, from a theoretical standpoint, there is reason to assume that bail supervision may create a stronger specific deterrent effect on repeat domestic violence than that observed in prior studies for traditional police arrest. More specifically, the perceived "certainty" of rearrest may be heightened among would-be domestic violence offenders who are subjected to random, independent, and individualized behavior checks by bailbond agents. When, where, and how bailbond agents detect future untoward domestic acts is completely unpredictable to suspects released on cash bond. The likelihood of police arrest, on the other hand, is perhaps more predictable (and, perhaps, less certain) to offenders who threaten their victims into not calling the police.

The Study

The goal of this study is to explore if, and to what extent, bail supervision can be used to reduce repeat acts of domestic violence committed by persons released on cash bail. The research objectives are twofold: (1) to conceptualize and implement an experimental form of bail supervision that requires intensive, proactive supervision of domestic violence bailees, and (2) to test the impact of this supervision through a fully randomized experimental design. The study began in June 1996 and was carried out in two phases lasting approximately 3 years.

Planning Phase

The first or "planning" phase took place during the summer of 1996 and involved the solicitation and orientation of bail agencies for study participation. Approximately 25 agencies participated in an informational meeting sponsored by the researcher at an annual private bail conference attended by bail agents representing both northern and southern California. The point of this meeting was to recruit bail agents as volunteer study participants by presenting to them the potential financial benefits of providing "a unique method of bail supervision" that could increase the professional status of bail agents, as well as increase their customer base. (The basic "profit pitch" was that bail agencies could attract more female clients by offering proactive domestic violence supervision.) Following the meeting and after several weeks of follow-up discussions, five agencies agreed initially to become study participants. However, one of these agencies later withdrew its participation on the grounds that the study might create losses in business profits because of unwarranted bail forfeitures. An additional concern was that potential male clients would avoid agencies with strict supervision requirements. These concerns were also echoed by bail agencies that declined outright to participate in the study.

Next, bail agents employed by the remaining four participant agencies were given a two-hour presentation and demonstration by the researcher describing proposed experimental study procedures. New agents hired subsequent to these sessions were also provided with similar information. In addition, each participating agent was given a series of practice mock cases, which were evaluated and supervised for procedural accuracy by research staff members before handling real study cases.

Experimentation Phase

Phase two, or the "experimentation" phase, began during the first week of August 1996. Within the four cooperating bail agencies, all subjects who were bailed out for felony domestic violence were randomly assigned to either an experimental "intensive supervision" group or "regular supervision" control group. The selection of experimental and control cases was carried out by individual bailbond agents at the time of initial contact with the representative(s) of a domestic violence suspect applying for bail release. Upon discovering that bail was set for a domestic violence offense, the bail agent then selected a color-coded bail application from a special file used only for study cases. The file contained white and blue forms, which represented regular and intensive supervision group assignment, respectively. Each bail application was numbered and drawn sequentially. Randomization of the study bail applications was determined by sequentially numbering four sets of 150 blue and white applications and then ordering the forms according to a random number series generated specifically for each set of applications. In all, four sets of 150 randomized applications (total N = 600; each set consisting of 75 white and blue forms) were provided to each of the participating bail agencies.

Biweekly quality checks were performed by the research staff to ensure that bail agents did not draw study applications out of sequence, thus jeopardizing the study's internal validity. Quality checks consisted of cross-checking times, dates, and numeric sequences of completed bail applications to determine if forms were used out of order. Checks revealed that on five separate occasions the randomization process was violated by bail agents. However, each of these incidents involved agents forgetting to use experimental application forms; hence, the original random ordering of the study applications remained intact. Problems such as agents refusing to administer treatment conditions to specific clients (similar to "selection bias" problems encountered by Sherman and Berk [1984] with police randomly applying arrest) were not encountered because bail agents (unlike police) very seldom know specific details regarding offenders or their offense at the time a decision to approve bail is made.

After subjects were randomly assigned to their respective experimental and control conditions, bail agents met individually with their bailees for a brief informational interview (approximately 20-30 minutes in length) regarding the terms and conditions of bail release. With the exception of informational content, the location, length, style, and rapport for interviews conducted with experimental and control subjects was identical. Agents then read to subjects a standardized set of either experimental or control instructions corresponding to the particular group of assignment. Both instructional sets were scripted. The control instructions contained content regarding financial terms and obligations of the bail agreement such as monthly payments (if any), use of cash or property collateral holdings, and so forth. For control subjects, no mention of the potential for intensive supervision was mentioned. (It is standard practice for bail agents to discuss only the financial terms of bail release with an arrestee; it is usually never the case that a bail agent will discuss, or even mention, the possibility of supervision, let alone intensive supervision.)

However, instructions for the experimental condition were designed to inform subjects of the intensive supervision treatment they would receive. Specifically, experimental subjects were led to perceive that the supervision would consist of random checks, and that bailbond agents had the discretion to invoke bail arrest if they had "good reason" to believe that a repeat domestic violence incident had occurred or was likely to occur. An sample of this dialogue is provided below:

As a bail agent, one of our primary responsibilities is community safety. We are not police officers, but the law in this state makes me personally responsible for your conduct while you are out on bail. I have the obligation to supervise you anywhere and anytime, and if I have any reason to believe you have committed or are thinking about committing another domestic violence crime or any other crime I will have to arrest you and surrender you to the police. And my arrest powers are much greater than the police. . . . I can arrest you at any time and in any place without a warrant. . . . I can arrest you whenever I think you might be a danger to someone else. This includes following people, harassing people, yelling and threatening people or just plain being nasty to people.

All subjects were required to sign a written statement acknowledging the fact that they understood the conditions of their bail and that they agreed to follow them to the best of their abilities.

The day after release on bond, each experimental subject was provided with a form of bail supervision conceived and created especially for this study termed "Intensive Bail Supervision (IBS). Briefly, the idea of IBS involved bailbond agents making unplanned telephonic and personal contact visits with subjects during times and in places where a repeat domestic violence incident was most probable. Usually, this entailed phoning or visiting the suspect at home in the evening between the hours of 6 P.M. and 9 P.M. Some contacts, however, were made at the bailee's place of employment, when home visits were not possible. Two contacts were make per week (one phone, one in-person). Because of the high potential for domestic violence to occur on weekends, one of the two contacts was scheduled for either Friday or Saturday.

For both experimental and control subjects, the postarrest follow-up period for this study was 6 weeks from the time of arrest. Although it may have been possible to track some suspects for a longer time period, a shorter follow-up was selected on the basis of experimental validity concerns. First, repeat incidents of domestic violence are most likely to occur during the month following the presenting incident. Thus, the bulk of repeat incidents will be captured by the study's official subsequent arrest dependent variable. Second, the abbreviated follow-up minimized attrition from the experimental and control groups. Second, and most important, the potential for the treatment effects of the bail supervision experimental treatment to be contaminated or confounded by extraneous factors was minimized by the shorter 6 week follow-up period. This is because persons released on cash bail within the counties under study are afforded a 5 to 6 week grace period before having to make their first court appearance. By comparison, persons not released on cash bail are required to start their case proceedings within 36 hours of arrest. Therefore, during this "free" period provided by cash bail release, study subjects were not affected by potentially sources of internal invalidity such as jail detention, plea bargains, or specific release conditions involving mandatory attendance in counseling or anger management programs.

Contrary to fears of most bail agents participating in the study, the threat of intensive supervision, and perhaps bail arrest, for experimental subjects did not result in as one agent bemoaned "having to arrest a bunch of guys for just screaming at their lady." During the entire 3-year period of study, only four experimental subjects were rearrested by agents during the 6 week follow-up period. (These arrests were the result of victim complaints indicating pretrial flight of the suspect rather than proactive anti-domestic violence measures taken by bailbond agents.) Thus, the threat of bail arrest stemming from intensive supervision for experimental subjects did not impose a financial burden on participating bail agencies as was expected by bail agents at the outset of the study.

Data, Sample, and Variables

Data and Sample

Study data come from pretrial release information gathered by bailbond agents from client suspects at two points in time. The first set of information consists of various background characteristics of the arrestee, the victim, and the arrestee's family and is collected by bail agents as part of the application process for a bail bond. For the average case, this data was gathered from the arrestee and/or the arrestee's family 3 to 8 hours following arrest. This initial report is used by the bail agent to decide which suspects are "good risks" and "bad risks" for pretrial flight. The secondary data source in this study consisted of limited case outcome data provided to the bail agent by prosecutors and courts following the close of an arrestee's case. Included in this report were the status of dropped cases, plea bargains, trial verdicts, new convictions and/or arrests, and other official court-related case outcome data.

In addition, bail data reported here are derived from both official and interview sources; however, victim interviews are not included. Because cash bail is an offender-based enterprise, bailbond agents rarely make personal contact with victims who are involved in the subject offense. Therefore, unofficial victim accounts of domestic violence incidents occurring subsequently to a subject's arrest are not included in the present study.

The initial study sample consisted of 552 persons requesting and ultimately obtaining cash bail for release from custody in Los Angeles or Orange Counties following a felony arrest for domestic violence as defined by California Penal Code 273.5 (i.e., felony domestic violence). The final sample was composed of approximately 90% (n = 496) of the original sample, after attrition because of dropped cases/charges (n = 32) and bail absconding (n = 14).

Independent Variables

Independent variables consist of the offender's marital status (married = 1; not married, living together, friends/lovers = 0); employment (employed = 1; unemployed = 0); prior domestic violence (prior D.V. offenses = 1; no prior D.V. offenses = 0). One dummy-coded experimental outcome measure that reflects the randomized effects of intensive bail supervision (1 reflecting the experimental condition; 0 reflecting the control condition) was created. Race/ethnicity was broken down into four categories, based on the characteristics of the sample: African American, Caucasian, Latino, and Mexican National. Separate nominal categories were created to reflect Latino and Mexican National race/ethnicity. This decision was guided by substantive and empirical preliminary analyses of race effects suggesting several statistically significant (p < 0.05) differences between Latino and Mexican National subsamples for several outcome predictors. For purposes of this study, Mexican Nationals are operationally defined as persons who were citizens of Mexico and who did not maintain U.S. citizenship at the time they received bail for domestic violence. Latino subjects, on the other hand, were persons of Mexican, South American, or Central American descent who were U.S. citizens at the time they received bail.

Dependent Variable

Also included is a single dependent measure of repeat offending, which was dummy-coded 1 for offenders who had been rearrested for a subsequent domestic violence incident during the study follow-up period and 0 for offenders who had not. This measure is based on official police records rather than on unofficial self-report offender or victim accounts (as used by Sherman et al., 1992). Furthermore, Berk, Campbell, Klap, & Western (1992) argue that official arrest variables are preferable to those derived from unofficial sources because of a high degree of response error, recall bias, and potential invalidity associated with self-reported domestic violence measures.

Findings

Descriptive Results

Table 1 presents descriptive statistics for variables used in the analysis. Within-category percentages for the criterion measure "rearrest for domestic violence" reveal that about 12% of the study sample (n = 58) was rearrested for at least one violent act against their partner during the study follow-up period. Percentage breakdowns for independent variables show that the majority of experimental subjects were married (69%; n = 343), employed (58%; n = 290), and had at least one previous arrest (52%; n = 254) for domestic violence. Compared to similar studies, this study sample appears to have an overrepresentation of married and employed participants. This is, perhaps, due to selection bias that is created by sampling subjects from a more affluent subpopulation of domestic violence offenders who are financially able to obtain pretrial release via cash bail. The race/ethnicity of the sample is overrepresented by persons who are either Latinos (42%; n = 211) or Mexican Nationals (30%; n = 151), which reflects the specific population characteristics of the communities in which the participating bail agencies were located.

Experimental Effect

Maximum Likelihood coefficients (L^sub b^) for various logistic regression models of the log odds of being rearrested for domestic violence are presented in Table 2. The first model examines the relationship between the experimental effects of intensive bail supervision and the rearrest dependent measure. Past research suggests mixed results for similar analyses: statistically significant positive and negative relationships, as well as insignificant relationships, have all been discovered. Model 1 shows the results of regressing the log odds of being rearrested on the dummy variable representing the independent effect of intensive bail supervision. The coefficient associated with the IBS effect parameter is negative and significant (L^sub b^ = -1.29; p < 0.01). This finding suggests the existence of a deterrent effect for intensive bail supervision. By converting the raw coefficient to an odds ratio, the finding implies the odds of being rearrested for domestic violence are about 3.6 times smaller for subjects under the IBS experimental supervision as for those subjects who were under regular supervision. From a substantive viewpoint, this result implies that the experience of being supervised while on bail for domestic violence provides a specific deterrent effect that is additional to the deterrent effect, if any, imposed by arrest alone.

Demographic Effects

Model 2 tests whether or not the observed experimental effect for intensive bail supervision persists after controlling for four additional demographic predictors: prior domestic violence arrest, marital status, employment, and race/ethnicity. Although the addition of these terms makes a highly significant contribution to explaining repeat domestic violence (X^sup 2^^sub Model 2^ - X^sup 2^^sub Model 1^ = 55.19, df = 6, p < 0.001), the intensive bail supervision effect remains significant as well (L^sub b^ = - 1.06; p < 0.01). Of the demographic predictor effects, the prior arrest variable has the strongest impact on rearrest criterion. This finding is congruent with results of numerous studies that have shown prior domestic violence activity to be among the strongest predictors of subsequent combative encounters between couples.

Within the demographic control variable set, an interesting finding regarding the second significant predictor (race/ethnicity) is worth noting. Specifically, the results in Model 2 reveal that subjects who are Mexican Nationals are significantly (L^sub b^ = 1.28; p < 0.01) more likely than their Caucasian, African American or Latino counterparts to be rearrested for domestic violence. Conversion of these findings to odds ratios further indicates that the likelihood of being rearrested for domestic violence for Mexican National subjects is approximately 3 times that discovered for subjects of other race/ethnicity classifications. In contrast, numerous past studies have shown a significant positive relationship between African American race/ethnicity and repeat domestic violence. However, the inability of the present study to reproduce race/ethnicity effects evidenced in previous domestic violence research may be the result of the disproportionately large number of Latinos and Mexican Nationals included in this study's sample.

Interaction Effects

Model 3 includes a set of interaction variables created from product terms that combine the experimental IBS predictor with all demographic predictors. Taken together, these interaction effects were designed to test the "stakes-in-conformity hypothesis" applied to domestic violence offending by Sherman et al. (1992). This theoretical paradigm, based on Toby's (1957) social control model, asserts that formal specific deterrent effects have the potential to increase domestic violence among some subclasses of offenders and to reduce violence among others. According to Sherman et al. (1992), who examined the deterrent effect of arrest on repeat domestic violence, punitive measures taken against domestic violence offenders are least likely to deter (and perhaps will even exacerbate) tendencies to quarrel among persons who lack community and/or familial bonds. On the other hand, "individuals subject to social control in jobs and marriages are more likely to be deterred by legal sanctions [i.e., arrest] than are those without such stakes in conformity" (p. 681).

Results for Model 3 reveal that the addition of interaction variables to the model provides yet another significant contribution to the prediction of domestic violence rearrest (X^sup 2^^sub Model 3^ - X^sup 2^^sub Model 2^ = 34.59, df = 7, p < 0.001). Two significant interaction terms were discovered, both of which lend support to the stakes-in-conformity hypothesis. The employment interaction, which produces the strongest effect, indicates that the odds of domestic violence recidivism were significantly reduced (L^sub b^ = -3.92; p < 0.001) among subjects who were under intensive bail supervision and were employed during the time they were being supervised. Similar joint effects among subsequent domestic violence, employment, and formal arrest were discovered by Sherman et al. (1992) In that latter study, and in the present study, social commitment in the form of employment appears to condition the relationship between behavioral sanctioning and repetitive physical abuse between couples.

The last, and perhaps the most interesting, finding is the significant interaction term representing Mexican National race/ethnicity (L^sub b^ = -2.94; p < 0.01). Ostensibly, the propensity for Mexican National subjects to engage in repeat domestic violence is reversed among Mexican National subjects who are placed under intensive bail supervision. For these subjects, intensive bail supervision appears to produce the most profound increases in the perceived costs of rearrest for domestic violence. This deterrence relationship, like that discovered earlier for employment, may be explained by a unique desire of Mexican National subjects to maintain established community ties and stakes in conformity (see the next section for a more detailed substantive explanation of this finding).

Conclusion

Bail Supervision and the Sword of Damocles

If indeed bail supervision serves to reduce repeat domestic violence among suspects released on cash bail, as is suggested by the findings from this study, what is the theoretical basis of this result? Within the realm of specific deterrence theory is one plausible explanation, which has been previously explored in the domestic violence literature. This conceptual framework, known as the "Sword of Damocles Hypothesis," was created by Dunford (1989) to explain how the "threat of formal arrest" may deter repeat domestic violence to a greater degree than the sanction of formal arrest (also see Dunford, 1992).

Specifically, in the Omaha Offender Absent Study, Dunford (1989) found that domestic violence suspects who fled the crime scene and were issued arrest warrants committed significantly fewer repeat acts of domestic violence than their counterparts who were arrested and booked into jail. Dunford suspected this finding was because of a more potent specific deterrent effect associated with offenders under warrant who perceived an ongoing pressure of imminent arrest because they felt they did not know when or where they would be apprehended by police. Thus, the primary assumption of Dunford's theory is that the infinite risk of "the element of surprise arrest" is a potentially stronger, longer lasting deterrent than are the effects of punishment imposed by the finite act of arrest.

In regard to this study's findings, the random and unpredictable nature of bail supervision may have the potential to re-create the same specific deterrent effect observed by Dunford for arrest warrants. First, both bail supervision and the arrest warrant process pose a lingering, constant "risk" of detection and apprehension to suspects who are contemplating subsequent acts of domestic violence. However, specific deterrent effects of bail supervision may be even stronger than those of official sanctions, such as arrest warrants, because of the proactive stance taken by bailbond agents during the postarrest period. For example, many suspects are aware that police often do not actively pursue domestic violence offenders who have a warrant for their arrest. In comparison, the relationship between bailbond agents and their domestic violence bailees seems to be much more restrictive. This conclusion is supported by the following dialogue obtained from an offender supervised in this study:

Sometimes I feel like a puppet on a string. If I fuck up, the bondsman's going to pull the string. He [the bailbond agent] is not a cop or noth'in like that. But that is probably worse. With him, I got no rights. Shit, my woman told me that I can't even have a funk'in beer or she'll call the bondsman and tell him I'm too drunk to go to court. . . . I think in a lot of ways I'm better off deal'in with the cops.

An additional deterrent imposed by bail supervision may come from the domestic violence offender's perception of the bailbond industry. First, many offenders expressed concerns about the financial honesty of bailbond agents:

These guys are blood suckers and they will suck your bucks dry if you give 'em a chance. I've been through this bail game before. I give them $3,000 bucks and signed my aunt's house over to 'em. I go to court every time I'm supposed to and boom . . . they look for any reason to haul me into to jail just before I get sentenced and take my money and the house. I'm not gonna cause any shit because this time I'm not gonna give them any reason to take my shit.

Second, many offenders felt that bailbond agents had the potential to be more physically abusive than police in conducting an arrest-if necessary:

You don't want to mess with the bail cops. You know the bounty hunters. Especially if you got busted for beat'in up a lady. If the lady calls for help, those guys will beat the shit out of you, and roll you up in the trunk of their car to the jail. They're a thousand time worse than the police.

Thus, as indicated above, the air of "mystique" that surrounds the relationship between domestic violence suspects and the bailbond agents who supervise them may serve to enhance the specific deterrent effect advanced in the Sword of Damocles Hypothesis. Moreover, the perceived costs to offenders contemplating repeat domestic violence may be heightened by fears over what may happen to them when the "sword" is in the hands of bailbond agents rather than the police.

Bail Supervision and Stakes in Conformity

As originally conceived by Sherman et al. (1992), the stakes-in-conformity hypothesis assumes that persons sufficiently tied to conventional society by factors such as employment and marriage will be deterred by formal legal sanctions (e.g., arrest). This is because these persons have the most to lose by being classified and stigmatized as an "offender" by their family, friends, employers, and the community at large. By substituting the deterrent effect of bail supervision for formal arrest, findings from this study lend support to the theoretical notion by Sherman et al. (1992) that the relationship between specific deterrence and repeat domestic violence is conditioned by social commitment factors. First, it was discovered here that suspects who were employed and supervised were less likely than any other class of subjects to be rearrested for domestic violence. However, as the following interview suggests, a common theme echoed by employed study participants involved fear over loss of employment resulting from increased contact with bailbond agents:

The whole reason I bailed myself out was to get to work. I work at the garbage company and here if you miss one day somebody else will take your job. Also if that bail guy tells my boss anything-it's all over for me. I just don't want that guy coming around my work to check up on me.

Second, it was discovered that Mexican Nationals were also less likely than suspects of other races/ethnicities to be rearrested for an additional violent encounters if they were assigned to the intensive bail experimental condition. Interestingly, Mexican Nationals not assigned to intensive bail supervision were among the most likely study participants to be rearrested for domestic violence. This finding can also be interpreted within the theoretical context of a stakes-in-conformity framework. More specifically, the most important community tie for Mexican National participants was the privilege of living and working in the United States. Many study participants feared that bail supervision would lead to a certain rearrest for domestic violence, thus severing their well-established American community ties. The following interview demonstrates the nature of typical concerns expressed by Mexican National study participants who were subjected to bail supervision:

Where I come from a man is not muy hombre if he lets his wife tell him what to do. In Mexico, I would never go to jail for slapping my wife.

But I'm here [in the United States] and I want to stay here. When the fianza policeman comes around I get scared because I think he will bring me back to jail for nothing. I promise him I won't do anything to her if he doesn't send me back to Mexico. I need my job, my kids, and I don't have $3,000 to pay a coyote to get me back across the border.

Taken together, the above joint relationships involving social ties predictors (i.e., employment and Mexican Nationalism) support the general conclusion that bail supervision has a greater deterrent effect for persons with stakes in conformity. However, in the present study, the fear of personal loss resulting in rearrest for domestic violence appears to increase the costs of punishment much more than other factors such as personal shame or being stigmatized as an "offender."

Discussion

Generally speaking, bailbond agents have a very narrow view of their function in the criminal justice system. Most of them consider their job to consist of only two functions: "getting people out of jail" and "making sure people go to court." The former of these tasks is how bail agents earn their livelihood; and if they are unsuccessful at the latter task, they will lose their livelihood. The idea of using the power of cash bail to provide a vital public service such as protecting victims of domestic violence is very foreign to the average bailbond agent. As one of the bail agents participating in this study noted:

It really wasn't a hell of a lot of work to do the intensive supervision. But I really can't see myself offering this service to everybody who gets bailed out on a domestic violence change. Look, in this business it's all about the money. If doing this type of service just takes up time without making any money, nobody in the bail business is going to do it. We're not cops or social workers-we're businessmen.

Second, bailbond agents have little financial incentive to use their power to prevent clients from being rearrested for domestic violence. When a bailee is taken back into custody for a subsequent offense, the client is no longer a flight risk, and the bail agent is not required by law to return the client's cash premium. And as previously mentioned, domestic violence offenders are the industry's prize catch because they carry high bails and low flight risks. Simply put, repeat domestic violence is a cash cow for bailbond agents.

However, contrary to popular belief, the finding of a deterrent effect for intensive bail supervision is a potential financial asset to bailbond agents. After all, it is often the female victim of a domestic violence incident who initially bails out her abusive partner after he has been taken into custody. When reminded of this fact, the "skeptical" bailbond agent who provided the above interview stated the following:

You know I never thought of it like that. If I put in my ads that I give this "intensive bail supervision" it would draw all of those women who look for bail after the police takes their man away. You know, a special service for domestic violence clients. I like it.

Thus, as intimated by the latter statement, bailbond agents may have interest in exploring and applying the concept of intensive bail supervision if the technique promises financial gain. More important, regardless of their underlying motive, bailbond agents who employ intensive bail supervision will perhaps play a vital role in the reduction of repeat domestic violence. This, in turn, will greatly assist the bailbond industry's current efforts to reinvent itself as a necessary and professional component of the criminal justice system.