The Role of Private Legal Representation and the Implicit Effect of Defendants’ Demographic Characteristics in Setting Bail and Obtaining Pretrial Release

Abstract

Misdemeanor bail schedules use the seriousness of the alleged crime and the defendant’s criminal history. Court officials contend the schedules ensure that only legal factors are considered for bail. Using longitudinal data from randomly selected misdemeanor cases filed during the first quarter of 2012 in Harris County, Texas we use ordinary least squares (OLS) to test the influence of demographic characteristics of defendants and their ability to retain private representation on the process of setting the monetary amount for bail. We find evidence of implicit bias and a distinct advantage of private representation. Using logit analysis, we estimate how demographic factors and monetary bail influence the probability of pretrial release, and present point estimates of making bail by demographic type. Demographics matter but retaining private counsel increases the probability of bail by sevenfold. Policy recommendations include expanding the use of public defenders and allowing them access to clients much earlier in the process.

Keywords

implicit bias wealth-based bail bail probability pretrial release policy

Introduction

The nexus between money and pretrial release in modern misdemeanor court systems has its origins in medieval England. Crime was considered an act between private individuals, and conviction usually meant the accused made financial payment to the grieved. The English were as concerned then about the accused fleeing the community as many are today, and as the medieval legal system developed, “a system was created in which the defendant was required to find a surety who would provide a pledge to guarantee both the appearance of the accused in court and payment . . . upon conviction” (Schnacke, Jones, & Brooker, 2010, p. 3). Typically, the surety required the accused to pay in advance the amount to be paid if convicted. In this manner if the accused did flee, the surety would pay the grieved with proceeds received from the now absent accused.

Pretrial release in modern court systems operates under the same basic principle but is more complex. The amount paid for bail is no longer connected directly to the expected financial penalty upon the finding of guilt. Today, bail not only remains a financial incentive to appear at some future court date but can also be a hurdle for defendants considered a risk to public safety. The monetary amount and the formal arrangement of bail vary across arrestees. A relatively few defendants charged with minor offenses, who are judged not to be a flight risk, and are deemed not a threat to society are released with no financial payment. This is known as a personal bond or a personal recognizance bond, and is simply the promise to appear on their respective court date. Most first-time defendants who are charged with relatively minor offenses and are judged not to be a threat to society are typically required to pay relatively low amounts for pretrial release. Repeat offenders charged with the same minor offense will face a higher monetary hurdle, and defendants with criminal histories who are charged with serious offenses face even greater amounts for release.

Funds for bail generally come from three major sources: Defendants may use personal funds to post bail with cash. Using personal funds for bail guarantees a direct connection between a financial loss and failure to appear. They may make arrangements with friends and relatives who provide the funds for bail. The connection between financial loss and failure to appear is nearly as direct if the defendant borrows from friends and relatives. Failure to appear could severely damage or even sever those relationships. Third, the defendant may use personal funds or funds from friends and relatives to pay a bail bond company. The connection between failure to appear and financial loss is weakest when the defendant uses a bail bond company. Many defendants, as in medieval England, avail themselves to a surety bond company to secure release.

Bail bond companies typically require¹ a nonrefundable payment of at least 10% of the bond. Ayres and Waldfogel (1994, p. 990) argue that the nonrefundable payment is a “sunk cost,” and this weakens the connection between what the defendant pays and the incentive to appear in court. From the perspective of the defendant, appearing or not appearing does not affect the nonrefundable character of the bond fee. Other incentives to appear remain, of course, such as complicating the current legal charges and incurring additional financial penalties for not appearing. The bond company has the incentive to get the defendant to court to avoid paying the court the bail amount. The larger the amount, the greater the incentive. This modern system ensures that defendants not among those relatively few who are released on personal bond or who do not have sufficient personal resources, helpful friends and family, or the resources for a bail bond will await their day in court in jail.

The policy of pretrial release in exchange for payment is so enmeshed in the misdemeanor court systems across the country that courts routinely employ bail schedules that codify the monetary requirements for pretrial release. The schedules are typically based upon the seriousness of the alleged crime and a defendant’s criminal history. The implicit assertion is that the monetary amount for bail is a function exclusively of the legal attributes of the case and the criminal history of the defendant. Personal characteristics of the defendant and access to resources are irrelevant, but any defendant who has access to resources for the required bond can expect to attain pretrial release. The system works reasonably well when it generally detains violent repeat offenders until their court date, and defendants accused of serious misdemeanor offenses and who also have extensive criminal histories face higher bail amounts and a reduced likelihood of being released.

The system does not work as well when defendants charged with relatively minor offenses who pose no risk to society are detained simply because they lack the resources to make bail. This can result in the disruption of lives, families, and employment. In addition, the pubic bears a cost of detention in the form of room and board for defendants who could just as easily live at home, rather than taking up space in the local jail, while waiting for their court date.

The system also does not work as designed if the courts consider more than the crime criminal history of the defendant and the flight risk. This occurs if court officials, when assigning bail, consider any of the demographic characteristics of the defendants and assess differential bail amounts for the same crime and history between men and women, across racial and ethnic groups, and across age categories.

This article is comprised of several sections: It first reviews research that indicates extralegal factors contribute to bail decisions related to a criminal case. The next section describes the pretrial process of a large urban misdemeanor court system and tracks defendants through the pretrial process. Bail is set early in the process and can be paid anywhere along the way for pretrial release. It then follows the sample of 3,250 misdemeanor defendants and displays the distribution of defendants detained versus those who are released. The article then reports the empirical results. These include comparison of mean bail amounts across socioeconomic characteristics and the ordinary least squares (OLS) estimates of the influence of legal, extralegal factors and the ability to retain private representation on the amount for bail. After we estimate the influence of numerous factors on the amount of bail to be paid, we use logit analysis to estimate the impact of some of the same factors, as well as the amount of bail, on the probability of pretrial release. We present point estimates of making bail for defendants from several different demographic and legal representation categories. The article concludes with policy recommendations to mitigate some of the adverse unintended consequences of the pretrial release procedure.

Because bail is procured through monetary payment, at the very least we expect to find the unintended consequence of a significant link between a defendant being released and financial wherewithal. This analysis contributes to the burgeoning research exploring the consequences of extralegal determinants of bail in a wealth-based pretrial release policy.

If indeed the magistrates who set bail strictly adhere to the bail schedule, we expect that the only two factors that explain variation in bail amounts are the charge, or charges, against the defendant and the criminal history of the defendant. Demuth (2003) points out that these two factors are the “best predictors of legal decisions” (p. 886). Strict adherence also means that no other factors contribute to the decision.

Literature Review

It is well established that pretrial release is a function of race, ethnicity, and gender. Multiple studies have found that numerous extralegal factors are considered in determining the bail amount (Ayres & Waldfogel, 1994; Freiburger & Hilinski, 2010; Katz & Spohn, 1995; Sacks & Ackerman, 2012; Schlesinger, 2005; Turner & Johnson, 2007) and who ultimately makes bail (Demuth, 2003; Free, 2004; Freiburger, Marcum, & Pierce 2010; Maxwell & Davis 1999; Williams, 2017). Minority men generally face higher monetary hurdles for bail and fail to make bail at much higher rates than their White counterparts. Women make bail at higher rates than men. Relatively high bail amounts most adversely affect lower income defendants. Rabuy and Kopf (2016) report that detained defendants have median incomes not only less than the population in general but also less than the incarcerated population.

Being unable to pay bail and subsequently being detained have been found to have a cascading effect on case outcomes, severity of sentence, and future economic opportunities for the defendant. Heaton, Mayson, and Stevenson (2016) find not making bail has accumulatively negative effects on the outcome. Lee (2016) concludes that detention significantly increases the odds of conviction. The evidence from these authors suggests that innocence or guilt is a function of the pretrial process, and the policies that undergird it, and not simply the evidence of the alleged crime.

A plausible reason to expect that magistrates include factors other than the elements of the bail schedule when they set the bail amount is found in the work of social psychologists (Greenwald & Banaji, 1995; Greenwald, McGhee, & Schwartz, 1998; Hahn, Judd, Hirsh, & Blair, 2014; Hofmann, Gawronski, Gschwendner, Le, & Schmitt, 2005) and criminal justice scholars (Jolls & Sunstein, 2006; Levinson, Young, & Rudman, 2012; Ogletree, Smith, & Wald, 2012; Tonry, 2010). Their research uses the concepts of implicit attitude, implicit stereotypes, and implicit bias, and indicates that criminal justice professionals from the officers on the street to judges sentencing the convicted make decisions based on attitudes and perceptions not perceived by the decision maker.

Implicit bias is one of a number of mental processes that functions without the individual being cognizant of it. It is inaccessible by a person even via introspection and can contradict explicitly expressed attitudes. One can claim to be race or gender neutral (explicitly expressed) but be unaware that she is implicitly biased in favor of one gender or the other. Implicit biases, derived from implicit attitudes and implicit stereotypes, can be found in the context of race (Levinson et al., 2012; Smith & Levinson, 2012) and ethnicity, gender, economic class (Demuth & Steffensmeier, 2004; Smith & Levinson, 2012).

While we do not directly test for the presence of implicit bias, we do conjecture that if bail amounts are not set by elements exclusively in the bail schedule then it is reasonable to expect it has an effect. Greenwald and Krieger (2006) argue that testing directly for implicit bias is unnecessary. In the context of racial prejudice in probation recommendations, they contend that after controlling for nonrace factors that it is quite probable that implicit bias is at play. They base their assertion on three observations: the pervasiveness of implicit bias, evidence of the effect of implicit bias on decision makers, and “findings that implicit bias plays a causal role in discrimination” (Greenwald & Krieger 2006, p. 966). Finding factors not in the bail schedule that explain the variation of bail amounts will be interpreted as the influence of implicit bias. We stop short of stating that implicit bias is the only possible reason but do contend that implicit bias is a very likely contributor. Magistrates have the discretion to adhere strictly to the bail schedule in setting bail or to impose greater monetary amounts than set by the schedule. It is here where implicit bias may affect the decision of the magistrate to unknowingly consider demographic characteristics of the defendant. A magistrate whose implicit racial attitude is neutral will be more inclined to assign bail according to the schedule. A magistrate whose implicit racial attitude is that racial or ethnic minorities are more dangerous will likely impose greater hurdles for bail.

The Process From Arrest to Court Date

The following describes the process of how a misdemeanor arrestee becomes a misdemeanor defendant, and how and when the defendant may, or may not, be released on bail in Harris County, Texas (Odonnell v. Harris County (US TXSD case No. 4:16-cv-1414 05/19/16)).

Arrests can occur without a warrant. Upon arrest, the arresting officer contacts an assistant district attorney (DA) via phone who makes a charging decision. If the decision is no charge, then the defendant is released without further action. If charges are filed, initial bail is set as stipulated by the schedule, and if the defendant can pay the initial bail amount the person is released on bond without being booked. If the defendant cannot pay the bail amount, then the defendant is booked into jail.

If the arrest is with a warrant, the charging decision is made in advance of issuing the warrant, and bail amount is stated on the warrant document. As in the case of the warrantless arrest, a defendant who can pay the bail amount is released. Defendants who are unable to pay the bail amount are detained in Harris County Jail (HCJ).

For those detained defendants, the probable cause hearing, commonly known as the bail hearing, follows shortly after being arrested. Ideally, the probable cause hearing occurs within 24 hr. Sometimes circumstances extend the time from arrest to a probable cause hearing beyond 24 hr. An unusually high number of arrests or slowdowns in the booking process may delay the probable cause hearing. These hearings are conducted via a video link between the jail, the location of the defendant, and the county court house where the hearing officer (i.e., magistrate) is located. The probable cause hearing is essentially a bail hearing as it is here that the magistrate reviews the bail to assure that bail is set according to the bail schedule. Odonnell v. Harris County (2016) observes that defense attorneys are not present at probable cause hearings to argue on behalf of the defendant for the reduction of bail. Defendants who are able to pay bail at the probable cause hearing are released. It is during the probable cause hearing that defendants declare a need for legal representation. Defendants who made bail at their arrest do not have a probable cause hearing of this nature. For those defendants, “probable cause determination in her (sic) case will be made at a subsequent court appearance” (Odonnell v. Harris County, 2016, p. 8). These defendants have legal representation at subsequent dates.

The next stage of the process is known as first appearance, and occurs within one or two days of the probable cause hearing. At this stage, the defendant makes his or her first appearance before a county criminal court of law. This is the first opportunity for defendants to confer with an attorney, and typically at this point the reduction of bail amount is not an option (Odonnell v. Harris County, 2016). It occurs before a county criminal court at law within 1 to 2 days of the probable cause hearing. This time period can be stretched if the defendant has a probable cause hearing Friday afternoon. In that situation, the first appearance will likely not be until the following Monday. Defendants who have probable cause hearings Friday evening or over the weekend may wait until Tuesday of the following week for their first appearance (Odonnell v. Harris County, 2016). At any point during this time, a defendant will be released if she is able to pay the stipulated bail amount. For those unable to pay and unable to retain private representation, first appearance is the occasion that the court appoints attorneys or assigns public defenders to defendants (Harris County Criminal Courts at Law, 2012).

This review of the pretrial process starting at arrest, through the Assistant District Attorney (ADA) filing of charges and affixing bail amount according to his or her interpretation of preset bail schedule,² the probable cause hearing before a magistrate, and first appearance at the trial court, reveals that initial bail is usually set according to the bail schedule. During this process, defendants can be released at any juncture when they pay the stipulated bail amount; those who make bail have immediate access to retained counsel. However, those who do not make bail and who are subsequently booked in HCJ and are unable to retain counsel wait until first appearance at trial court for a court-appointed attorney or a public defender. During these proceedings, there is virtually no opportunity for counsel to argue for a reduction in bail or for noncharge-related considerations when determining bail. The bail schedule is the primary bail setting instrument.

The judges of the County Criminal Courts at Law of Harris County adopted the misdemeanor bail schedule in 2012 (Harris County Criminal Courts at Law, 2012). Table 1 presents four categories of offenses and the corresponding bail amount: The first two categories are standard Class B and Class A misdemeanors, and the second two stipulate bail amounts for the special offenses of family violence and/or the threat of violence and driving while intoxicated (DWI). Additional bail amounts are called for if the defendant was already on bail, parole, probation, or some other type of supervision at the time of arrest.

Table 1.

Harris County Bail Schedule.

Class of misdemeanor	Stipulated bail (US$)
B, Standard offense First offense Second offense	$500 $500, plus $500 for each prior misdemeanor conviction, plus $1,000 for each prior felony conviction (not to exceed $5,000)
A, Standard offense First offense Second offense	$1,000 $1,000, plus $500 for each prior misdemeanor conviction, plus $1,000 for each prior felony conviction (not to exceed $5,000)
Family violence or threat of violence First offense Second offense	$1,500 $1,500, plus $2,000 for each prior conviction for a violent offense or threat of violence
DWI First offense Second subsequent offense	$500 $2,500, plus $1,000 for each prior conviction not to exceed $5,000
Any offense committed while on bond, community supervision, intervention, or parole Any motion to adjudicate or revoke community supervision	$5,000 $5,000

Note. DWI = driving while intoxicated.

On its surface, one might argue that the process is objective and all misdemeanor defendants are treated equally. Defendants facing the same charge who have comparable criminal histories face the same bail amount. Defendants without a criminal record who are charged with relatively minor misdemeanors face minimal bail amounts. Defendants who face charges of serious offenses face higher bail amounts that reflect the seriousness of the charge. Any defendants who can afford to pay are released. No one is detained if bail is paid. The inequity arises from variation in the defendants’ ability to pay. According to the bail schedule, no consideration is given to the ability to pay bail (Odonnell v. Harris County, 2016). It also does not consider extralegal factors such as age, sex, or race and ethnicity of the defendant.

Multiple Paths to Pretrial Release or Detention

The data source is Harris County, Texas District Attorney Office. Harris County is the most populous county in Texas, 4.5 million, and the third most populous in the country. It is the central county of the Houston, Texas Metropolitan Statistical Area. The data were randomly selected from misdemeanor cases filed from January 1, 2012 through April 30, 2012. During that period, 23,309 misdemeanor arrests were made in Harris County, and from those we generated a random sample of 3,317 misdemeanor defendants. Figure 1 exhibits the various pretrial release paths that the 3,317 defendants take to the resolution of their respective cases. As of April 30, 2015, the end of the 3-year study period, all but 54 cases have been resolved. Of the 3,263 resolved cases, 635 defendants are not booked at HCJ. Thirteen of 635 nonbooked defendants are nonarrest cases. The remaining 622 nonbooked defendant bail group plus 2,628 booked defendants comprise our sample of 3,250 defendants. We track those 3,250 through the pretrial release process from date charged to date of disposition. All 622 nonbooked defendants are released on bail and remain on bail at their court date. Of the 2,628 defendants who are booked at HCJ, 1,295 make bail but 195 experience a bond forfeiture or revocation before their court date. That leaves 1,100 booked defendants on bail at their court date. The number of nonbooked defendants on bail (622) plus the number of booked defendants on bail at their court date (1,100) is 1,722. This accounts for 53% of 3,250 disposed study cases.

Figure 1.

Orange jumpsuit: Misdemeanor random sample.

As previously explained, defendants can secure bail through a number of methods. They may be released on their personal recognizance, PTR (their promise to appear is all that is required), a surety bond that is financed by a bail bond company or by paying cash from personal sources. The entries at the bottom of Figure 1 indicate the distribution of defendants who are booked at Harris County and make bail by the type of bail as well as defendants who are not booked immediately upon arrest and make bail by the type of bail. Those who are booked at the county jail make bail via Cash (59), Surety (1,036), and PTR (200). These amounts include defendants who initially made bail but were detained on their court date. Nonbooked defendants make bail via Cash (45), Surety (549), and PTR (28). The aggregate for the two defendant groups is Cash (104), Surety (1,585), and PTR (228) for the grand total of 1,917 defendants who make bail. Approximately 59% of the 3,250 defendants make bail. Of those who make bail, 83% use bail bond companies.³

Descriptive Data and Empirical Results

Table 2 displays descriptive statistics for variables in the study. Nearly 60% of total misdemeanor defendants in the study sample make some type of bond. The average monetary value of set bond is approximately US$3,350. Slightly less than half of the defendants have prior misdemeanor convictions with an average of 1.46 convictions. In total, more than half of the defendants are charged with offenses more serious than Class B Misdemeanor. This is important because a Class B Misdemeanor typically has a US$500 bail requirement. The typical defendant is a White male, about 30 years old and does not retain a private attorney.

Table 2.

Descriptive Statistics.

Variable	N	M or %	SD
Made bail	3,250	58.98%	0.49
Bond amount	3,248	US$3,348	US$5,267
Prior misdemeanor (yes/no)	3,250	48.83%	0.50
Number of prior misdemeanors	3,241	1.46	2.63
Class A misdemeanor (yes/no)	3,250	33.85%	0.47
Offense against family (yes/no)	3,250	5.94%	0.24
DWI (yes/no)	3,250	14.00%	0.35
Retained private attorney (yes/no)	3,250	41.63%	0.49
Male (yes/no)	3,250	77.17%	0.42
Black (yes/no)	3,250	39.60%	0.49
Hispanic (yes/no)	3,250	26.28%	0.44
Age	3,249	30.75	11.16

Note. DWI = driving while intoxicated.

As part of our preliminary analysis, we also explore the relationships among sex, race, and type of representation. Table 3 displays the chi-square statistics for each combination of the three variables. Women are more likely than men to be White but are less likely to retain private representation. Whites, in general, are more likely than other racial/ethnic groups to hire private counsel. This likely reflects the relative wealth advantage that Whites enjoy over other racial and ethnic groups.

Table 3.

Crosstabs of Extralegal Factors.

Crosstabs	n	χ²
Race by sex	3,250	χ²(1) = 111.83, p = .000
Race by private representation	3,250	χ²(1) = 165.98, p = .000
Sex by private representation	3,250	χ²(1) = 4.53, p = .030

Determinants of Bail Amount

We now turn our attention to the factors that contribute to the amount of bail that a defendant is required to pay. Finding no relationship between demographic traits and the amount of bail would indicate that Harris County court officials set bail in accordance with the bail schedule, without regard to personal characteristics of the defendant. Finding a relationship between the amount of bail and any single trait, or the traits as a group, would imply that officials are influenced by factors not appearing on the schedule. The preliminary tests of the null hypotheses of no relationship are accomplished by conducting a comparison of mean bail amounts across defendants categorized by demographic groups.

Table 4 displays the results of the comparison of means of bail amount by sex, race/ethnicity, and age. We find that variation in the bail amount exists across these demographic traits. The average bail amount for men (US$3,672) is 63% greater than that for women (US$2,255) and is statistically significant at p value = .00. The mean bail amount for Whites, Blacks, and Hispanics is US$2,882, US$3,705, and US$3,356, respectively. The mean bail amount for Whites is significantly less than the mean bail amount for Blacks (p value = .00) and Hispanics (p value = .03). The mean bail difference between Blacks and Hispanics is significant at p value = .07. Older defendants pay higher bail. For men and women, significant increases occur as the age of the defendant increases from below 20 to 45 years of age, and then levels off for those above 45.

Table 4.

Mean Bail Amount by Demographics.

Category	M (US$)
Sex
All men (n = 2,506)	3,672***
All women (742)	2,255
Race/ethnicity
Whites (n = 1,042)	2,882***
Blacks (n = 1,287)	3,705
Hispanic (n = 854)	3,356** and *
Age
Men
Below 20 (n = 377)	2,391
Twenties (n = 979)	3,345***
30-45 (n = 812)	4,358***
Above 45 (n = 338)	4,397***
Women
Below 20 (n = 77)	1,214
Twenties (n = 327)	2,000**
30-45 (n = 247)	2,789***
Above 45 (n = 91)	2,603

Note. Categories do not sum to 3,250 due to missing data.

*p < .10. **p <.05. ***p < .01.

The limitation of these comparisons of means tests is that they do not control for other factors that may affect the differences. For example, the differences shown in Table 3 may be partially attributable to the difference in the severity of the alleged crime and the difference of the criminal histories. Table 5 contains the summary of results of chi-square tests that test for relationships between the items on the bail schedule and sex, race, and private representation, respectively. All four bail schedule items are related to the sex of the defendant at a .05 level of significance. For example, women are more likely to be charged with Class B misdemeanor than a Class A in comparison with men. This helps understand the lower bail amounts for women relative to men. The hypotheses of no relationships are rejected between race and the charge of DWI/driving under the influence (DUI) and between race and criminal history but cannot be rejected between race and seriousness of charge or between race and a family offense charge. One could argue that Blacks have relatively high bail as a reflection of their more extensive criminal history. With regard to private representation, defendants charged with Class A misdemeanors are no more likely to use private counsel than defendants charged with Class B misdemeanors. The same holds for defendants charged with a family violence offense. Defendants charged with DWI/DUI are more likely to use private representation than their non-DWI/DUI counterparts, and defendants who have criminal histories are more likely to retain private counsel than first-time offenders. The comparison of means and the crosstabs tests suffers from not controlling for other factors while estimating one of a number of relationships.

Table 5.

Crosstabs of Bail Schedule Elements by Sex, Race, and Legal Representation.

Crosstabs	n	χ²
Sex by seriousness of charge	3,250	χ²(1) = 11.948, p = .001
Sex by family offense	3,250	χ²(1) = 13.872, p = .000
Sex by DWI/DUI	3,250	χ²(1) = 5.733, p = .017
Sex by misdemeanor priors	3,250	χ²(1) = 35.604, p = .000
Race by seriousness of charge	3,250	χ²(3) = 5.554, p = .135
Race by family offense	3,250	χ²(3) = 2.017, p = .569
Race by DWI/DUI	3,250	χ²(3) = 182.65, p = .000
Race by priors	3,250	χ²(3) = 109.418, p = .000
Private representation by seriousness of charge	3,250	χ²(1) = .005, p = .940
Private representation by family offense	3,250	χ²(1) = .003, p = .958
Private representation by DWI/DUI offense	3,250	χ²(1) = 173.875, p = .000
Private representation by misdemeanor priors	3,250	χ²(1) = 142.026, p = .000

Note. DWI = driving while intoxicated; DUI = driving under the influence.

Controlling for such affects is of course easily accomplished by regressing the amount of bail on the number of variables under discussion in an OLS equation. An additional advantage of linear equations is the capability to compare the explanatory power of an equation including a group of variables with the explanatory power of an equation excluding the group of variables such as demographic characteristics. Comparing the goodness of fit between two equations that differ only by the omission of a group of variables provides insight into the effect of the variables as a group.

We estimate two models that explain the bail amount required for defendants. Model 1 in Table 6 contains only variables that reflect elements of Harris County bail schedule. Class B misdemeanor charges are distinguished from Class A misdemeanor charges in the schedule with greater bail required for Class A misdemeanors. To capture the effect of a more serious crime, the variable classAmisd has a value of 1 for defendants charged with a Class A misdemeanor and 0 for defendants charged with a Class B misdemeanor. The bail schedule differentiates violence against family members and DWI from the standard Class A or Class B misdemeanors. Class A misdemeanors that include charges of violence against family members warrant a higher bail than a standard Class A misdemeanor. The variable family is a dummy variable that has the value of 1 if the charge is for violence against a family member and 0 otherwise. Likewise, the charges of drinking while driving and DUI are differentiated from the standard Class A misdemeanor, and have correspondingly higher bail amounts. The variable dwi takes the value of 1 if the defendant is charged with DWI or DUI, and 0 for any other type of charge. Regardless of the alleged offense, the schedule calls for higher bail amounts for those defendants who have prior convictions. The variable priors indicates criminal history. If the defendant has been convicted of any type of crime in the past, then priors takes the value of 1. It has a value of 0 for defendants who have no prior convictions.

Table 6.

OLS Results of Natural Log of Bond Amount.

Variable	Model 1: Restricted model	Model 2: Unrestricted model
Variable	Estimated coefficient (SE)	Estimated coefficient (SE)
ClassAmisd	0.4465*** (0.0325)	0.4420*** (0.0317)
Family	0.4968*** (0.0637)	0.4691*** (0.0625)
dwi	−0.1771*** (0.0390)	−0.1598*** (0.0409)
Priors	1.1261*** (0.0290)	1.006*** (0.0301)
Privateatt		−0.2408*** (0.0301)
Black		0.0558* (0.0340
Hispanic		0.0170(0.0376)
Female		−0.2490*** (0.0345)
Age		0.0076*** (0.0013)
Constant	6.791*** (0.0245)	6.5051*** (0.0544)
	N = 3,237	N = 3,237
	F(4,3232) = 513.20	F(9,3227) = 257.34
	Prob > F = .000	Prob > F = .000
	R² = .3884	R² = .4178
	Adjusted R² = .3877	Adjusted R² = .4162

Note. OLS = ordinary least squares; DWI = driving while intoxicated.

Significance at .10. **Significance at .05. ***Significance at .001

The dependent variable is the log transformation of the dollar amount required to make bail. The regression model is defined as a log-linear model, and the coefficients are interpreted as the percentage change in the dependent variable, given a unit change in the independent variable. The coefficients on the dummy variables estimate, ceteris paribus, the percentage difference in bail between those in a particular group (say, Class A misdemeanor) and those in the comparison group of the dummy (Class B misdemeanor).

As expected, all of the estimated coefficients of Model 1 are statistically significant, and the adjusted R² = .39. This confirms the importance of the elements of the bail schedule in setting bail. As the coefficients estimate a percentage change in the dependent variable given a unit change in the independent variable, we can compare the relative importance of each. The most important element that affects the amount of bail is the defendant’s criminal history (Priors).

Assaulting a family member (Family) has the second greatest effect on the amount of bail required for a misdemeanor defendant. On average, defendants charged with a Class A misdemeanor pay approximately 45% more for bail than do their counterparts who are charged with Class B Misdemeanors. Defendants charged with a DWI or DUI face bail amounts 18% less than non-DWI or non-DUI defendants.

The variables discussed thus far are included on the Harris County bail schedule.⁴ Strictly speaking, policy dictates that these are the only factors to be considered when assigning bail amounts. Model 2 contains all the variables of Model 1 plus legal representation and demographic information about defendants. This allows us to test the hypotheses that magistrates do not consider these factors when setting bail, both as individual variables and as a group.

Defendants who retain private attorneys have access to legal representation much earlier in the process than defendants who are represented by court-appointed attorneys or by public defenders. Those defendants can contact their attorney immediately after being arrested ensuring an advocate for them at every stage of the proceedings. This includes attempting to reduce the amount of bail assigned either initially by the ADA or by the magistrate. Those assigned an attorney typically do not get one until their first court appearance (Odonnell v. Harris County, 2016). This is the first opportunity for their attorney to argue for a reduced bail. For defendants who have private representation, the variable Privateatt takes the value of 1. The Black, Hispanic, Female, and Age operationalize demographic characteristics. Black takes the value of 1 for African American defendants and 0 for non-African American ones. Hispanic likewise takes the value of 1 for Hispanic defendants and 0 otherwise. The variable Female is also a dummy variable equal to 1 for women and 0 for men. To test the influence age of the defendant has on the bail amount, the defendant’s age, in years, is operationalized by Age.

If magistrates assign bail strictly by the bail schedule, the estimated coefficients of the private attorney and demographic measures are not expected to be statistically significant. The dollar amount of bail will not be affected by the defendant’s sex, age, race, or ethnicity. If, however, the estimated coefficients are found to be statistically significant, this could be interpreted as the evidence of implicit bias.

The results of Model 2⁵ in Table 6 clearly indicate that defendants’ traits beyond criminal attributes of the case contribute to the variation of the bail amount. They suggest that magistrates do not strictly follow the bail schedule. The standard tests of significance for the respective coefficients of individual variables indicate that the only null hypothesis not rejected is that Hispanics pay the same amount of bail as Whites. Controlling for the type of crime, criminal history, and the other control variables, Blacks pay an average bail 6% more than Whites. This is significant at the .05 level. Bail for women is 25% less than for men. Age has a relatively small, statistically significant effect of .76% increase in bail for each year.

Defendants who retain a private attorney have typical bail amounts 24% less than those who do not. This is significant at the .05 level. The effect of a private attorney is not unexpected as defendants who are represented by public attorneys or court-appointed attorneys often meet their attorney just before first appearance. Bail amounts have been reviewed and set by prosecution by that point. Private attorneys have numerous opportunities to advocate for their client for a lower bond amount.

The results of the tests of significance of individual coefficients suggest that magistrates and trial judges use extralegal factors when setting bail. Further evidence of this behavior is revealed by comparing the goodness of fit of Model 1 with the goodness of fit of Model 2. Model 1 restricts the demographic and representation estimated coefficients to 0, and has a R² of .3884. Model 2 has no restrictions on the estimated coefficients and has a R² of .4178. We tested the hypothesis of demographic and representation coefficients all equal zero using the respective R² from each model. The test statistic⁶ (Pindyck & Rubinfeld, 1981) of 29.81 is significant at the .001 level. As a group and individually, the demographic and legal representation variables play a significant role in determining the amount of bail a misdemeanor defendant must pay. Overall, the crime and criminal history play dominant roles, and White defendants with the income to hire an attorney pay significantly less than others for pretrial release.

Determinants of Making Bail

This section addresses the probability of a defendant in our sample gaining pretrial release and the factors that influence it. As noted earlier, substantial research supports that a mix of legal and extralegal factors contribute to who is and who is not released on bail. Controlling for legal factors, women are more likely to make bail relative to their male counterparts, and Hispanic and Black males are more likely to be detained relative to their White counterparts. Given that the defendant must pay an assigned amount of bail to be released, the monetary amount of bail is expected to be an important factor. It is de facto the price of freedom while the case is being resolved. An inverse relationship between the amount of bail and the probability of bail is hypothesized. The monetary amount of bail is transformed by the natural log function to simplify the interpretation of the coefficient.

A defendant’s ability to pay the amount is also important. The data do not contain information about the individual’s income or wealth. Although the literature contains examples of using the zip code level data to operationalize the defendant’s income or wealth (Heaton et al., 2016), we contend that measure moves the analysis away from the individual level and introduces the risk of measurement error for the defendant’s income. We use two proxies for wealth that are directly measured for each defendant. Our data contain whether the defendant retains a private attorney or is represented by a court-appointed or public defender. The advantages of retaining private counsel were argued earlier, and here we contend that retaining private counsel is an indicator of wealth or at least access to financial resources to pay for a private attorney. Having the financial resources for private representation implies that the defendant has the resources to pay the bail amount. Admittedly, this is not a direct measure of income but it is more closely connected to the individual defendant than the income measure of the zip code of the defendant’s residence. We hypothesize that private representation will increase the probability of making bail. The second proxy is the number of prior convictions of the defendant. A criminal history reduces the economic opportunities of a defendant. A criminal history makes the defendant ineligible for some occupations and can push the defendant into relatively low paying others. This diminishes the potential of earnings, and with the present alleged offense the defendant with prior convictions may find relatively few monetary resources for bail. It is expected that economic opportunities are diminished by a criminal history which diminish monetary resources and hence the probability of making bail.

A number of scholars (Demuth & Steffensmeier, 2004; Jones, 2014) have found that gender, race, and ethnicity all contribute positively and negatively to the likelihood of bail. Generally, men of color are less likely to make bail than women and White men. The variables Black, Hispanic, and Female are dummy variables included in the model as controls.

The results of the estimated logit model are shown in Table 7. The first column contains the variable names and model statistics. The second column contains the estimated coefficient of the logit model and its respective standard error. The asterisks indicate the level of significance with the greater number of asterisks reflecting more demanding criteria for significance. The third column is the estimate of the change in the odds of bail, given a change in the respective independent variable.

Table 7.

Estimate Coefficients and Odds Ratio of Logit Equation: Probability of Bail.

Variable	Coefficient(SE)	Exp(B)
Lndbondamt	−0.7018*** (0.0577)	0.4957
Privateatt	3.8086*** (0.1505)	45.0891
Priors	−0.0706*** (0.0247)	0.9318
Black	−0.0277(0.1149)	0.9727
Hispanic	−0.3724*** (0.1339)	0.6890
Female	0.2381** (0.1192)	1.2688
Constant	3.8720*** (1.1716)	121.6024
N = 3,228
Wald χ²(6) = 779.94
Prob > χ² = .000
Pseudo R² = .4145

Significance at .10. **Significance at .05. ***Significance at .001

These estimates indicate, as expected, that the dollar amount of the bail reduces the likelihood of a defendant making bail. The estimated coefficient for lnbondamt, −0.70, is significant at a p value of .00. The value indicates that the log odds of being released on bail decreases as the amount required increases by 1%. Column 3 is another way of interpreting the relationship. The odds of making bail are defined as the ratio of the probability of making bail to the probability of not making bail. These results indicate that as the amount due for bail increases by 1%, the odds of making bail decreases by 0.49. Ceteris paribus, the odds of defendants making bail is 0.49 the odds of their counterparts making bail who face a bail amount 1% smaller. This means that the probability of making bail relative to the probability of not making bail decreases by nearly one half. A calculation of the estimated probabilities is presented in the next section. As expected, we find an inverse relationship between the price of release and the probability of a defendant gaining pretrial release.

Having the resources to retain private representation is the most important factor in attaining pretrial release. The coefficient of the log odds is positive, significant at .001 level, and indicates that being represented by a private attorney more than triples the log odds of bail. More directly, the odds of a defendant making bail who hires private representation is 45 times greater than the odds of making bail of a defendant who does not have private representation. No other variable in the model approaches the implied importance of these results. If arrested, even for a misdemeanor, get a private attorney.

Being able to retain private counsel is one indicator of the importance of wealth in the bail process. Priors is the other variable that reflects access to monetary resources. We find an inverse relationship between the number of prior convictions and the log odds of making bail that is significant at the .001 level. The impact on the odds of making bail does not meet the magnitude of the impact of the dollar amount of bail. With each additional prior conviction, the odds of making bail is 0.93 of the odds of making bail without the additional conviction.

Two of the three race, ethnicity, and gender variables confirm what is found in the literature. The log odds of bail for Hispanics are less than that of Whites, and difference is significant at the .001 level. The odds of Hispanics making bail is only 0.69 the odds of Whites making bail. Women, as found in the literature, have greater log odds of making bail than do men, ceteris paribus. The log odds of bail for women increases 0.24, and the odds of women making bail is 27% greater than that for men.

These results also indicate that the odds of making bail for Blacks is arithmetically less than that for Whites, but the difference is not statistically significant. This contradicts what is found in the literature. The odds of making bail for Blacks is 97% of the odds of Whites making bail.

The model indicates that the odds of making bail are increased for Harris County misdemeanor defendants who face relatively low monetary hurdles for bail, who have the resources to retain private representation, who have no or relatively few prior misdemeanor convictions, and who are not Hispanic. A poor Hispanic male with an extensive criminal history has the odds stacked against him for making bail. Regardless of race, ethnicity, or gender, the most important factor is whether the defendant can afford a private attorney. Wealth matters.

Probability of Bail for Specific Defendant Types

The previous section addresses the marginal effects of changes in variables in terms of log odds and odds of bail. Given the results from the logit model, we use the logit probability function to calculate point estimates of the probability of making bail for specific defendant types. The logit probability function is given as follows:

\begin{array}{l} P r o b (b a i l) = \frac{1}{1 + e^{- Z}}, \\ where Z = 3.8720 - 0.7018 l n b o n d a m t + 3.8086 \\ P r i v a t e a t t - 0.0706 M i s d \Pr i o r s - \\ 0.0277 B l a c k - 0.3724 H i s p a n i c + 0.2381 F e m a l e . \end{array}

Z contains the estimated coefficients from Table 7 in algebraic form. For any particular defendant type, we insert values for the independent variables to calculate Z, and then use Z in Equation 1 to compute a point estimate of making bail for that defendant type. For example, for a first-time offending White male charged with a Class A misdemeanor who retains a private attorney and must pay US$2,315 for bail, the values for the independent variables lnbondamt, Privateatt, MisdPriors, Black, Hispanic, and Female are US$2,315, US$1, US$0, US$0, US$0, and 0, respectively. Given the parameter estimates and the values of the independent variables, Z equals 0.9041. This is the probability of bail for this hypothetical defendant.

Table 8 displays the expected probability of bail for 12 different types of defendants and their type of legal representation. The effect of retaining a private attorney is quite apparent. The highest expected probability of bail for any defendant type who did not have the resources to retain private representation is only .12. If the defendant has the resources to retain private counsel, the probability of making bail never drops below .75.

Table 8.

Estimated Probability of Bail for Selected Defendant Types.

Defendant type	Description of defendant type	Probability of making bail
1	White male charged with a Class A misdemeanor who has no prior misdemeanors is represented by a private attorney and pays the mean bail amount of this group (US$2,315).	.9041
2	Black male charged with a Class A misdemeanor who has no prior misdemeanors, represented by private attorney, paid the mean bail amount of this group (US$3,481).	.8732
3	Hispanic male charged with a Class A misdemeanor who has no prior misdemeanors, represented by private attorney, paid the mean bail amount of this group (US$2,105).	.8741
4	White male charged with a Class A misdemeanor who has no prior misdemeanors is represented by a public defender or a court-appointed attorney, and pays the mean bail amount of this group (US$4,058).	.1236
5	Black male charged with a Class A misdemeanor who has no prior misdemeanors is represented by a public defender or a court-appointed attorney, and pays the mean bail amount of this group (US$5,329).	.1018
6	Hispanic male charged with a Class A misdemeanor who has no prior misdemeanors is represented by a public defender or a court-appointed attorney, and pays the mean bail amount of this group (US$4,375).	.0844
7	White male charged with a Class A misdemeanor who has one prior misdemeanor conviction is represented by a private attorney, and pays the mean bail amount of this group (US$4,436).	.8477
8	Black male charged with a Class A misdemeanor who has one prior misdemeanor conviction is represented by a private attorney, and pays the mean bail amount of this group (US$4,931).	.8341
9	Hispanic male charged with a Class A misdemeanor who has one prior misdemeanor conviction is represented by a private attorney, and pays the mean bail amount of this group (US$4,950).	.7803
10	White male charged with a Class A misdemeanor who has one prior misdemeanor conviction is represented by a public defender or a court-appointed attorney, and pays the mean bail amount of this group (US$6,635).	.0851
11	Black male charged with a Class A misdemeanor who has one prior misdemeanor conviction is represented by a public defender or a court-appointed attorney, and pays the mean bail amount of this group (US$6,561).	.0836
12	Hispanic male charged with a Class A misdemeanor who has one prior misdemeanor conviction is represented by a public defender or a court-appointed attorney, and pays the mean bail amount of this group (US$6,403).	.0617

These results indicate the importance of wealth in attaining bail. If one has the wealth to retain private representation, then the probability of making bail, even with one prior misdemeanor conviction, is nearly .80. We contend that those who have the wealth to hire an attorney can also afford the 10% of the bail amount needed to post bond via a bail bondsman necessary for pretrial release. Less than US$700 would be needed for any of the 12 types of defendants as shown in Table 8. Defendant Types 1 and 3 would need slightly more than US$200 to make bail. The ability to hire private counsel leads to the ability to post bail and attain pretrial release.

Emerging research (Lee, 2016; Sacks & Ackerman, 2014) has found evidence of the cascading effect of not making bail in subsequent decisions in the process of resolving a case. Not making bail is the first step in a process that exhibits what Wooldredge, Frank, Goulette, and Travis (2015) refer to as cumulative disadvantage. Earlier adverse decisions affecting the defendant influence later decisions to the detriment of the defendant. What is becoming well established is that the poor are unable to make bail, and that inability significantly influences the eventual outcome of the case.

Conclusion and Policy Recommendations

From the perspective of court officials, little time is available for deliberation and reflection when setting bail amounts. Decisions are frequent and must be made quickly, and people in circumstances as this often unintentionally rely on judgment that is based on implicit attitudes and biases. The process is likely behind why we find evidence that bail amounts are influenced by the implicit bias of the magistrates. More direct exploration of the degree of implicit bias is warranted.

These results reveal the importance of retaining private counsel when establishing the amount of bail and its importance, once the amount is set, in the probability of being released. Determining the amount of bail includes more than the elements of the bail schedule (i.e., seriousness of the offense and prior record). Our findings reveal that defendants who retain private counsel pay a bond average of US$2,558, whereas defendants who are too poor to retain private counsel pay US$3,912; more than US$1,354 than their wealthier counterparts. With respect to making bail, 90% of the 1,353 defendants who retain private counsel make bail. Only 27% of the 1,897 defendants who do not retain private counsel make bail. OLS results indicate that retaining private counsel reduces the bail amount by nearly 25%, and logit results estimate that retaining private counsel increases the odds of making bail by a magnitude of 45. Misdemeanor defendants who can afford private representation have good reason to retain counsel. Those who cannot afford private counsel face higher bail amounts and lower probability of attaining pretrial release. Stated differently, low income and poor defendants face higher bail and lower probability for release than their wealthier counterparts.

It is unrealistic to suggest that defendants be provided the resources to retain private representation. One alternative is to ensure court-appointed attorneys and public defenders greater access to defendants between the probable cause hearing and their first court appearance. Many defendants who do not retain private counsel meet their attorney immediately before their first appearance, in some instances more than 3 days after being the probable cause hearing (Odonnell v. Harris County, 2016). This may be accomplished by expanding the number of attorneys in the Office of the Harris County Public Defender.

Harris County Courts may also consider a policy change that expands the use of personal bonds (PTR). Of the 228 defendants initially released on PTR, 90% remained on bond at the resolution of their case. In comparison, 1,585 defendants paid a bail bond company to gain pretrial release, and approximately 90% of those defendants remained on bond at the disposition of their case. When considering only Class A misdemeanor defendants, to control for seriousness of alleged crime, PTR defendants remain on bail at disposition at a slightly lower rate than nonPTR defendants (79% vs. 88%, respectively). Twenty seven of the 312 DWI defendants awaited trial via PTR, and all 27 were on bail at the disposition of their case. Conversely, 260 of the 285 defendants on pretrial release other than PTR were on bail at disposition. Class A misdemeanor defendants released via PTR are only slightly less likely to have their bond revoked or forfeited and to appear in court on their appointed date. DWI defendants on bail via PTR were more likely to avoid bond revocation or forfeiture and to appear in court on their appointed date.

Admittedly, our results are limited to a particular criminal court system. Scholars conducting research in bail policy have conducted their work using a variety of data sources. Free (2004) reports of 30 studies examining the effect of race on bail decisions, 18 were based on individual cities or counties. Freiburger and Hilinski (2010) examine the impact of race, gender, and age on pretrial decisions based on an urban county in Michigan. Using a single-court system avoids the different pretrial release structures imposed by different states. Nonetheless, future research will benefit by including more court systems that operate under the same state rules and regulations.

Given the limitations, our findings indicate that socioeconomic factors not included in the bail schedule contribute to the amount one must pay to attain pretrial release, and that defendants who can afford private representation are at a distinct advantage over those without the resources to do so. Legal representation comes later in the process to those unable to retain private representation, they face higher bail amounts, and hence they are less likely to make bail. It is improbable that money for pretrial release will end but the court system does have options to reduce the deleterious effects of being accused and being poor.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors thank Themis Research for partial funding for this research.

Notes

Author Biographies

Rod V. Hissong is an associate professor of Public Affairs in the College of Architecture, Planning and Public Affairs at the University of Texas at Arlington. His research centers on the economics of public policy including criminal justice policy, focused primarily on the role of economic, social and political factors play in the process.

Gerald Wheeler has over thirty years of research in the area of pretrial services. He researches and publishes on the role and continuing reform of pretrial services. He is the founder and Director of the non-profit Themis Research that focuses on criminal justice within the court system.

References

Ayres

Waldfogel

(1994). A market test for race discrimination in bail setting. Stanford Law Review, 46, 987-1047.

Demuth

(2003). Racial and ethnic differences in pretrial release decisions and outcomes: A comparison of Hispanic, Black, and White felony arrestees. Criminology, 41, 873-907.

Demuth

Steffensmeier

(2004). The impact of gender and race-ethnicity in the pretrial release process. Social Problems, 51, 222-242.

Free

(2004). Bail and pretrial release decisions: An assessment of racial threat perspective. Journal of Ethnicity in Criminal Justice, 2(4), 23-44.

Freiburger

T. L.

Hilinski

C. M.

(2010). The impact of race, gender, and age on the pretrial decision. Criminal Justice Review, 35, 318-334.

Freiburger

T. L.

Marcum

C. D.

Pierce

(2010). The impact of race on the pretrial decision. American Journal of Criminal Justice, 35, 76-86. doi:10.1007/s12103-009-9069-z.

Greenwald

A. G.

Banaji

M. R.

(1995). Implicit social cognition: Attitudes, self-esteem, and stereotypes. Psychological Review, 102, 4-27.

Greenwald

A. G.

Krieger

L. H.

(2006). Implicit bias: Scientific foundations. California Law Review, 94, 945-968.

Greenwald

A. G.

McGhee

D. E.

Schwartz

J. L. K.

(1998). Measuring individual differences in implicit cognition: The Implicit Association Test. Journal of Personality and Social Psychology, 74, 1464-1480.

10.

Hahn

Judd

C. M.

Hirsh

H. K.

Blair

I. V.

(2014). Awareness of implicit attitudes. Journal of Experimental Psychology: General, 143, 1369-1392.

11.

Harris County Criminal Courts at Law. (2012). Rules of court. Retrieved from http://www.ccl.hctx.net/criminal/Rules%20of%20Court.pdf

12.

Heaton

P. S.

Mayson

S. G.

Stevenson

(2016). The downstream consequences of misdemeanor pretrial detention. Stanford Law Review, 69, 711-794.

13.

Hofmann

Gawronski

Gschwendner

Schmitt

(2005). A meta-analysis on the correlation between the Implicit Association Test and explicit self-report measures. Personality and Social Psychology Bulletin, 31, 1369-1385.

14.

Jolls

Sunstein

C. R.

(2006). The law of implicit bias. California Law Review, 94, 945-996.

15.

Jones

(2014). Confronting race in the criminal justice system. Criminal Justice, 27, 12-17.

16.

Katz

Spohn

(1995). The effect of race and gender on bail outcomes: A test of an interactive model. American Journal of Criminal Justice, 19, 161-184.

17.

Lee

J. G.

(2016). To detain or not to detain? Using propensity scores to examine the relationship between pretrial detention and conviction. Criminal Justice Policy Review. Retrieved from http://journals.sagepub.com.ezproxy.uta.edu/doi/full/10.1177/0887403416668016

18.

Levinson

J. D.

Young

D. M.

Rudman

L. A.

(2012). Implicit racial bias: A social science overview. In Levinson

J. D.

Smith

R. J.

(Eds.), Implicit racial bias across the law (pp. 9-24). New York, NY: Cambridge University Press.

19.

Maxwell

S. R.

Davis

J. S.

(1999). The salience of race and gender in pretrial release decisions: A comparison across multiple jurisdictions. Criminal Justice Policy Review, 10, 491-502.

20.

Odonnell v. Harris County. (2016). United States District Court for the Southern District of Texas Houston Division (Case No. 4:16-cv-1414).

21.

Ogletree

Smith

R. J.

Wald

(2012). Criminal law—Coloring punishment: Implicit social cognition and criminal justice. In Levinson

J. D.

Smith

R. J.

(Eds.), Implicit racial bias across the law (pp. 45-60). New York, NY: Cambridge University Press.

22.

Pindyck

R. S.

Rubinfeld

D. L.

(1981). Econometric models and economic forecasts (2nd ed.). McGraw-Hill.

23.

Rabuy

Kopf

(2016). Detaining the poor. Northampton, MA: Prison Policy Initiative. Retrieved from Available from http://prisonpolicy.org

24.

Sacks

Ackerman

A. R.

(2012). Pretrial detention and guilty pleas: If they can’t afford bail they must be guilty. Criminal Justice Studies, 25, 265-278.

25.

Sacks

Ackerman

A. R.

(2014). Bail and sentencing: Does pretrial detention lead to harsher punishment? Criminal Justice Policy Review, 25, 59-77.

26.

Schlesinger

(2005). Racial and ethnic disparity in pretrial criminal processing. Justice Quarterly, 22, 170-192.

27.

Schnacke

Jones

Brooker

(2010). The history of bail and pretrial release. Pretrial Justice Institute. Retrieved from https://cdpsdocs.state.co.us/ccjj/Committees/BailSub/Handouts/HistoryofBail-Pre-TrialRelease-PJI_2010.pdf

28.

Smith

R. J.

Levinson

J. D.

(2012). The impact of implicit racial bias on the exercise of prosecutorial discretion. Seattle University Law Review, 35, 795-826.

29.

Tonry

(2010). The social, psychological and political causes of racial disparities in the American criminal justice system. Crime & Justice, 39, 273-312.

30.

Turner

K. B.

Johnson

J. B.

(2007). The relationship between type of attorney and bail amount set for Hispanic defendants. Hispanic Journal of Behavioral Sciences, 29, 384-400.

31.

Williams

(2017). The effect of attorney type on bail decisions. Criminal Justice Policy Review, 28, 3-17. doi:10.1177/0887403414562603.

32.

Wooldredge

Frank

Goulette

Travis

III . (2015). Is the impact of cumulative on sentencing greater for Black defendants? Criminology and Public Policy, 14, 187-223.