the pennsylvania state university more than just money
Post on 16-May-2022
1 Views
Preview:
TRANSCRIPT
The Pennsylvania State University
The Graduate School
College of the Liberal Arts
MORE THAN JUST MONEY: HOW THE BAIL PROCESS CREATES UNINTENDED
RACIAL DISPARITY
A Thesis in
Criminology
by
Lily S. Hanrath
© 2017 Lily S. Hanrath
Submitted in Partial Fulfillment
of the Requirements
for the Degree of
Master of Arts
December 2017
ii
The thesis of Lily S. Hanrath was reviewed and approved* by the following:
Jeff Ulmer
Thesis Adviser
Professor of Sociology and Criminology
Assistant Head, Department of Sociology and Criminology
Darrell Steffensmeier
Professor of Sociology and Criminology
Eric Baumer
Professor of Sociology and Criminology
Head, Department of Sociology and Criminology
*Signatures are on file in the Graduate School
iii
Abstract
Racial disparity in court processes is an important and heavily researched area of
criminology. However, comparatively little attention has been given to race differences that may
occur in the pre-trial process. Looking at racial disparity in bail actions is important as all
subsequent processes may be influenced by these decisions. Previous studies on bail have found
racial differences in detainment but a less strong relationship between bail amount and race. This
may be because of a lack of specificity, in that most studies do not differentiate between
multiple, potentially distinctive bail outcomes. This study adds to the literature by considering
monetary and unsecure bail types separately in models looking at racial disparity between black,
white, and Hispanic defendants in bail type and amount. Employing a focal concerns framework,
it is hypothesized that racial disparity may emerge in bail decisions because race is used as a
perceptual shorthand for blameworthiness and dangerousness. Using a unique dataset collected
from four Pennsylvania counties, this study attempts to understand the interaction between race,
legal factors, and extra-legal factors in predicting the likelihood of different bail outcomes.
Findings suggest that racial disparity between white defendants and black and Hispanic
defendants is present in bail type decisions but not in bail amount. This relationship is further
complicated by the availability of legal information such as prior record.
iv
TABLE OF CONTENTS
List of Figures ............................................................................................................................................... v
List of Tables ............................................................................................................................................... vi
Acknowledgments ....................................................................................................................................... vii
Introduction ................................................................................................................................................... 1
The Bail Process in Pennsylvania ................................................................................................................. 3
Literature Review .......................................................................................................................................... 6
Studies on detainment ............................................................................................................................... 8
Studies on Bail Amount .......................................................................................................................... 11
Consideration of Bail Type ..................................................................................................................... 12
Gaps in Research ..................................................................................................................................... 13
Focal Concerns Perspective ........................................................................................................................ 14
Hypotheses .................................................................................................................................................. 18
Data and Methods ....................................................................................................................................... 20
Datasets ................................................................................................................................................... 20
Sampling ................................................................................................................................................. 21
Coding ..................................................................................................................................................... 22
Measures ................................................................................................................................................. 24
Results ......................................................................................................................................................... 27
Descriptives............................................................................................................................................. 27
Multivariate Analysis .............................................................................................................................. 30
Discussion ................................................................................................................................................... 35
Limitations .................................................................................................................................................. 38
Conclusion and Future Directions............................................................................................................... 39
References ................................................................................................................................................... 42
Appendix: Figures and Tables .................................................................................................................... 45
v
List of Figures
Figure 1: The Bail Process………………………………………………………………………45
vi
List of Tables
Table 1: Frequencies of Bail Types .............................................................................................. 46
Table 2: Bail Type by Race .......................................................................................................... 46
Table 3: Bail Type by Control Variables ...................................................................................... 46
Table 4: Bail Amounts by Bail Type (In $1000s)......................................................................... 47
Table 5: Logistic Model of Bail Type ........................................................................................... 48
Table 6: Logistic Model of Bail Type: Drug Dealing (N=501) ................................................... 48
Table 7: Logistic Model of Bail Type: Violent Crime (N=327) ................................................... 49
Table 8: Logistic Model of Bail Type: Property Crime (N=1,035) .............................................. 49
Table 9: Logistic Model of Bail Type: PRS and Race Interactions* ............................................ 50
Table 10: Logistic Model of Bail Type: No Prior Record (N=1596) .......................................... 50
Table 11: Logistic Model of Bail Type: With Prior Record (N=1,371) ....................................... 51
Table 12: Regression Model of Bail Amount: All Bail Types ..................................................... 51
Table 13: Regression Model of Bail Amount: Monetary Bail (N=2,099) .................................... 52
Table 14: Regression Model of Bail Amount: Unsecure Bail (N=868) ....................................... 52
Table 15: Correlation Matrix ...………………………………………………………………… 53
Table 16: Literature Review of all studies published since 2000 ................................................. 54
Table 17: Dependent Variable Coding ......................................................................................... 55
Table 18: Sampling Frame (Percentage of each group sampled) ................................................. 56
vii
Acknowledgments
I would like to thank my thesis advisor, Jeff Ulmer for helping me plan my thesis as well as
helping with data collection and editing, connecting me to appropriate experts, and keeping me
calm through minor crises.
I would also like to thank David Johnson of the Penn State demography department for helping
me figure out the best method for sampling and guiding me through how to sample correctly.
I am grateful to both Eric Baumer and Darrell Steffensmeier for taking the time to be on my
thesis committee and offering helpful critiques for improving my paper.
Finally, I would like to acknowledge friends and family, in particular Brandy Parker and Sarah
Fry, for offering help along the way in analysis and writing.
1
Introduction
The manifest goal of the American court system is to ensure fair and equal treatment for
defendants; yet it has been shown repeatedly that there are differences in treatment between races
that are unrelated to legal factors (Spohn, 2015). The criminal justice system is a complex set of
decisions with potential risks of unwarranted disparity in every stage from arrest to parole. While
there have been studies on racial disparity in each stage of the court process, differential
outcomes by race in pretrial processes have been understudied (Baumer, 2013, Ulmer 2012).
Many studies on racial disparity in sentencing actually do not control for bail decisions;
therefore, missing the potential influence it may have on outcomes (Baumer, 2013; Ulmer,
2012). Beyond underestimating race effects in research, early stages of the court process may
have a marked effect on later court processes. There is evidence that pretrial detention has a
significant impact on conviction, incarceration, and sentence length (Oleson et, al., 2014;
Williams, 2003; Sacks & Ackerman, 2014). If there are racial differences in bail outcomes, this
disparity will be reflected in subsequent processes.
Along with the possible influence of the bail process on later outcomes, bail decisions
can have severe immediate consequences. Defendants detained pending trial account for a
significant percentage of jail populations. The Bureau of Justice Statistics estimated that pretrial
inmates accounted for 62.8% of jail detainees in America in 2014. Considering the mass
incarceration problem in the United States, combined with the potential possibility that factors
such as race may influence the likelihood of detainment, , these findings and figures are
concerning. Defendants who are unable to make bail may be forced to take out high-interest
loans they cannot afford to pay back or stay in jail and miss work or lose their job by
consequence.
2
United States law explicitly states that bail should be decided on based on community
safety and flight risk and not the perceived blameworthiness of the defendant. However, many
state courts use “bail schedules” which provide one-size-fits-all bail amounts based on charges
(Feeley 1979). The use of these schedules does not directly get at the manifest goals of
community safety and ensuring appearance but, instead, assesses blameworthiness by matching
up crimes to amounts of money. While these schedules may decrease discretion which, in theory,
would lead to less racial disparity, the implication bail decisions being completely related to the
severity of the crime goes against the supposed reason the bail process exists in the first place.
Pennsylvania law does not allow bail decisions to be made based on risk assessments
alone but, instead requires that each decision is overseen by a judge. This allows there to be more
consideration of individual defendant circumstances, such as income, employment, and children
rather than decisions being based on the crime alone. However, this additional discretion of the
judge may also lead to more extralegal disparity. Furthermore, judges are required to make these
decisions within 48 hours of arrest so they often have limited information on cases and limited
oversight that would allow for checks and balances. Judges may use perceptual shorthands
related to race because of this lack of case data, time, and oversight.
The focal concerns perspective posits that court actors have specific considerations in
making their decisions that frame their reasoning. Previous studies on bail hearings suggest that
the focal concerns of blameworthiness, dangerousness, and practical constraints may be
applicable to judges’ bail decisions (Demuth & Steffensmeier, 2004; Freiburger, Marcum &
Pierce, 2010). Judges use both legal and extralegal information that is available to them to assess
how blameworthy and dangerousness defendants are along with what practical constraints the
3
defendants and/or the court may have in that particular case. Judges may use race as an indicator
of these focal concerns, especially when information or time is sparse.
Using a unique dataset collected from four Pennsylvania counties, this article will discuss
the nuances of the bail process in Pennsylvania and how it could be understood in the context of
the focal concerns perspective. In particular, this study focuses on differences in bail outcomes
for white, black, and Hispanic defendants. First I will discuss how the bail process works in the
state of Pennsylvania. I will then discuss previous research on pretrial detention on bail. Third I
will discuss the focal concerns theory and how it may relate to judge decision making during bail
hearings. I will then lay out my hypotheses and explain my methods for data collection and
analysis. Results will be shown and discussed as they relate to previous literature, the focal
concerns lens, and my hypotheses. Finally, I will end with some limitations to my study as well
as possible conclusions and future directions necessary to understand the bail system.
The Bail Process in Pennsylvania Bail could be considered the second step in the criminal justice process, after arrest. The
fact that bail is so near the beginning makes it especially relevant to study because all subsequent
steps build upon bail and are affected by bail outcomes. Federal law requires initial arraignments
to be held “as soon as is reasonably feasible, but in no event later than 48 hours after
arrest” ("County of Riverside v. McLaughlin (89-1817), 500 U.S. 44 (1991)".) . The law in
Pennsylvania specifically specifies that bail must be set in this time period as well (PA §57.B.
§5750-b). In this short time frame, defendants are unlikely to have a lawyer and judges may not
even have access to a defendant’s police report or criminal history. Furthermore, magistrate
judges are not required to have a law degree in the state of Pennsylvania and non-lawyers are
only required to have 140 hours of training (Boehm, 2013).
4
In terms of bail decisions, the bail reform act of 1984 states that the least restrictive bail
should be enacted that will “reasonably assure the appearance of the person as required” and will
not lead the defendant to “endanger the safety of any other person or the community” (United
States Code, Title 18, Sections 3141-3150). Furthermore, it is explicitly stated in the law that
financial conditions should not be imposed so as to result in the pretrial detention of the person.
Notably, the law only states that appearance and community safety should be considered and not
that bail should be used as a form of punishment. This means that the severity of the crime, the
perceived culpability of the defendant, or the defendant’s admittance or denial of guilt should not
explicitly guide a judge’s decision. Furthermore, the eighth amendment posits that “excessive
bail shall not be required” (U.S. Const. amend. VIII). This amendment should, theoretically,
ensure that bail is not burdensome to defendants- making it an ad hoc punishment. However, it is
difficult to assess up front whether punishment is a consideration of a judge and it is unclear
what constitutes excessive bail. Many states use tables known as bail schedules that match up
crime types with suggested bail amounts that arrestees are automatically given without going
before a judge. Despite bail schedules directly linking bail to crime severity, it is argued this is
not punishment because severity is a predictor of flight risk and risk to the community.
Pennsylvania law allows for the use a pretrial risk assessment tool that evaluates factors
such as crime severity and type but the rule explicitly states that “a risk assessment tool must not
be the only means of reaching the bail determination” (PA § 234.523 part C(1)). Pennsylvania
procedural code suggests ten factors judges should consider which include:
“(1) the nature of the offense charged and any mitigating or aggravating factors that may bear
upon the likelihood of conviction and possible penalty; (2) the defendant’s employment status
and history, and financial condition; (3) the nature of the defendant’s family
relationships; (4) the length and nature of the defendant’s residence in the community, and any
past residences; (5) the defendant’s age, character, reputation, mental condition, and whether
5
addicted to alcohol or drugs; (6) if the defendant has previously been released on bail, whether
he or she appeared as required and complied with the conditions of the bail bond; (7) whether
the defendant has any record of flight to avoid arrest or prosecution, or of escape or attempted
escape; (8) the defendant’s prior criminal record; (9) any use of false identification; and;
(10) any other factors relevant to whether the defendant will appear as required and comply with
the conditions of the bail bond.”- PA § 234.523 part C (1)
Note that this list of suggestions does not advise what weight judges should put to considering
each of these factors or whether or not the existence of a factor should result in an increase or a
decrease in bail severity. The range and vagueness of these rules allows for a large amount of
discretion by judges in setting bail. In discussions had with three former magistrate judges in
Pennsylvania and one director at the sentence commission, it was emphasized to me that judges
differ greatly in their decision making process and are often influenced by their political
leanings, limits of time, and personal knowledge of the defendant in question. It was noted to me
that judges often times are setting bail in as little as twelve hours after arrest and may have to set
bail for multiple people arrested at the same time, such as in the case of a riot or gang warfare.
These limitations mean that judges may not have police reports or prior record information
available to them and must assess the risk level of the defendant based off an informal interview
with the defendant.
Before deciding on bail amount, judges in Pennsylvania first have several options of bail
type. On either side of the spectrum judges may choose to remand the defendant, meaning to
hold them in jail with no option of release, or to release the defendant on their own recognizance;
meaning defendants are free to go without paying any bail. Defendants may also be released with
some nonmonetary conditions such as a restriction on alcohol use or an order to stay away from
the victim of the crime. Most commonly, however, judges choose between setting a bail amount
that is classified as either monetary or unsecured. Monetary bail is an amount given that the
defendant is required to pay before they are allowed to be released from jail. In contrast,
6
unsecure bail amounts are only paid by defendants if they fail to show up to court or violate a
stipulation of their release. By ignoring this major difference in bail type, previous studies may
be conflating defendants who are paying large amounts of money to be released with defendants
who are essentially paying no money at all. Furthermore, a large proportion of defendants given
monetary bail cannot afford to pay and, instead, sit in jail while, in contrast, those with unsecure
bail are always released. The consequences of this difference in bail type have implications for
subsequent court processes. Studies find that defendants who are jailed pretrial are more likely to
take plea deals (Heaton, Mayson, & Stevenson, 2017), and more likely to get longer sentences on
average (LaFrentz & Spohn, 2006; Spohn, 2009).
Literature Review
Despite the potential for significant effects, few studies have actually delved into pretrial
detention and what factors into its outcomes. While significant results have been found for the
effects of race; results differ greatly across studies. A review of 30 bail and pretrial studies done
by Free (2004) found that 50% of studies concluded minorities had more negative outcomes than
white defendants, 40% found no race effect, and 10% found white defendants to fare worse. This
breakdown shows a striking divide in the literature. However, all the studies analyzed by Free
(2004) were published before 2000 and the majority of data were collected during the 1970s so
they may not be a good representation of the modern day system. Six studies published after
2000 that are not mentioned in Free (2004) all find some negative outcomes for minorities
compared to white defendants (Demuth, 2003; Demuth & Steffensmeier, 2004; Freiburger,
Marcum, & Pierce, 2010; Freiburger & Hilinski 2010; Schlesinger 2005; Wooldredge 2012))1.
However, all of these studies find race does not significantly differ for certain bail outcomes or
1 A literature review table of these six studies can be found in table 16 of the appendix
7
that racial differences are explained by other legal differences between defendants. Also, there is
some evidence that the effect of race depends on the gender and/or age of the defendant (Demuth
& Steffensmeier, 2004; Freiburger, 2010; Wooldredge 2012). In any case, the literature shows
little agreement.
Part of the reason behind the wide array of findings may be the construction of the
dependent variable. Studies on the bail system have looked at a number of outcomes, the most
common being bail amount, and detainment. In order to understand the processual nature of bail,
it is important to consider each aspect of pretrial decision making. Demuth and Steffensmeier
(2004) describe the pretrial release process as having four key decision points. Firstly, the judge
must decide whether to remand the defendant or give them some sort of release option. If the
judge decides not to remand the defendant, they must further decided whether to release the
defendant on their own recognizance or set some sort of monetary bail. Judges who decide on
giving bail must further consider the amount the defendant should be required to pay. Finally,
defendants given bail may decide to either pay that bail and be released or be held in jail until
further hearings. Notably, while the Demuth and Steffensmeier (2004) model is clear and concise
in breaking down the bail process, the third stage only accounts for financial versus nonfinancial
bail. However, bail options are more diverse than this, especially in the case of financial bail
types. Different financial conditions may require defendants to pay all, some, or none of their
bail amount2. Studies that do not account for this instead conflate unsecure bail with either
monetary bail or ROR decisions or, in cases where a percentage is monetary, record the
monetary portion of decisions and dismiss the nonmonetary portions. By grouping these
2 See Figure 1 for a visual representation of the bail decision making process
8
outcomes into mixed categories, researchers may be missing nuances both in how judges may
differentially treat defendants and in the ultimate outcomes faced by defendants.
The focus of this study is racial disparity in judge decision making, so the analysis will
focus on the first three stages: remand versus release, type of release (which includes ROR,
monetary bail, or unsecure bail), and bail amount. In order to understand the field thus far, I
break down past research findings into three types: detainment, bail amount, and bail type. I also
will discuss how the bail process has been shown to correlate to later court outcomes including
plea deals, sentence length, and sentence severity.
Studies on detainment
Detainment is the most common dependent measure in the studies on racial disparity in
the bail system that I reviewed. Thirteen of the studies reviewed by Free (2004) looked at
whether or not defendants were released from jail while five of the six studies done since 2000
considered detainment outcomes (Demuth 2003; Demuth & Steffensmeier 2004; Freiburger,
Marcum, & Pierce 2010, Freiburger & Hilinski 2010, Schlesinger 2005). Looking at who is held
pretrial is important as it adds to the jail population, costs the court system a large amount of
money, and harms the defendant’s ability to take care of their family and maintain employment.
However, as pointed out by Wooldredge (2012), studies on detainment often are flawed in that
they combine court decision making with release outcome (p 45). Detainment can occur due
either to a judge remanding a defendant or due to a defendant not posting bail. Analyzing these
two outcomes together is spurious as it is unclear whether significant findings are due to the
decision making of the judge or the defendant. Since remand is only supposed to occur for
defendants eligible for life or death sentences or when judges can prove the defendant is a
serious risk, it is likely that decisions to remand are much rarer than decisions to release
9
defendants. More often, detainment is likely due to defendant’s not being able to post bail. This
is an important outcome to consider as it may be that bail amounts are excessive which violates
the eighth amendment. However, ultimately, looking at the relationship between race and
detainment may actually be measuring disparity in economic status rather than disparity in court
decision making.
The few studies that have looked specifically at judges’ decisions to remand defendants
have found significant racial disparity. All three of these studies, however, have used data
compiled from the State Court Processing Statistics (SCPS) 1990-2000 dataset. The SCPS
dataset is composed of a sample of felony cases taken from the 75 most populous counties. Since
all recent studies looking at remand have used this dataset, it must be taken into account that
these results are likely to echo each other and do not represent rigorous replication processes
over various areas and conditions. Furthermore, this dataset includes only the 75 largest counties
in America and it may be that processes in large cities vary from midsized county procedures and
decision making. Schlesinger (2005) found that black defendants have odds 25% higher than
whites of being denied bail and Latinos have odds 24% higher than white defendants. When
looking at crimes by type, black defendants have 33% higher odds than white defendants of
being denied bail for violent crimes and 80% higher odds for drug crimes. In contrast, Latino
defendants have 67% higher odds than white defendants for drug crimes but differences are not
significant for any other type of crime. These results imply that racial disparity in decision
making is not uniform and may vary greatly by crime. However, the nonsignificant differences
may also be due to a small n as only 9% of black defendants and 7% of white and Latino
defendants are denied bail. Bail is likely denied at low rate in property and public order related
crimes which may explain nonsignificant findings. Demuth (2003) reached similar conclusions
10
to Schlesinger (2005) that black and Hispanic defendants have increased odds of being denied
bail in comparison to white defendants. He makes note, however, that denial of bail is very
uncommon and occurs mostly in cases where defendants are charged with serious crimes are
have significant criminal histories. It may be that race effects found in these studies are due to
racial differences in prior criminal justice system involvement or in the severity of crimes
committed. Demuth and Steffensmeier (2004) do not find significant differences for white,
black, and Hispanic defendant’s odds of being remanded despite also using the SCPS dataset.
The only major difference in Demuth and Steffensmeier’s model in comparison to the Demuth
(2003) model is that they separate “criminal history” into four variables: prior felony arrest, prior
felony conviction, prior jail incarceration, and prior prison incarceration. It may be that racial
differences in bail being denied are due to differences in criminal histories between defendants.
It has been noted that SCPS has crude criminal history measures due to missingness and
differences across state jurisdictions (Ulmer, 2012). It may be that Demuth and Steffensmeier
(2004) increase the error in their model by over emphasizing criminal history which causes racial
differences to appear insignificant.
Demuth (2003) and Demuth & Steffensmeier (2004) further analyzed racial differences
in defendants who were not remanded but were detained pretrial due to not posting bail. Both
articles found that when controlling for bail amount, black and Hispanic defendants were
significantly less likely to be able to post bail. Since the amount decision made by the judge is
controlled for, this outcome is likely due to economic differences between races rather than
differences in court treatment.
11
Studies on Bail Amount
Studies on racial disparity on bail hearings most commonly focus on differences in bail
amounts. This is logical as bail amount is seemingly a straightforward interval variable and it is
an inadvertent measure of detainment since we can assume that defendants with larger bail
amounts are more likely to stay in jail. However, the assumption that bail amount is an interval
variable should be used with caution as it may be that the difference between $500 and $1000
means something very substantively different than the difference between $20000 and $20500.
Furthermore, it may not be that larger amounts are always correlated with likelihood of
detainment as judges are encouraged to take into account the income of the defendant and bail
amount may vary depending on SES. These factors may account for the generally weak
correlations found between bail amount and race in the literature. In fact, all five studies that
have considered bail amount since 2000 have found no significant differences between black and
white defendants when taking into account the legal factors of the cases (Demuth 2003, Demuth
& Steffensmeier 2004, Freiburger et al. 2010, Schlesinger 2005, Wooldredge 2012). Studies that
look at outcomes for Hispanic defendants did find that they received bail amounts seven to eight
percent higher than similar white defendants but all three of these models were done using the
same SCPS dataset (Demuth 2003, Demuth & Steffensmeier 2004, Schlesinger 2005). One
possible explanation for the weak differences in bail amounts found between races may be that
researchers are conflating monetary and unsecured bail. Hypothetically, a black defendant may
get $5000 monetary bail and a white defendant might get $5000 unsecure bail which would
appear the same in bail amount regressions despite being substantively different in reality. Even
within the monetary bail category, black and Hispanic defendants may be getting higher amounts
than white defendants but those results may be overshadowed by white defendants who are given
high unsecure bail.
12
Consideration of Bail Type
A few studies have considered differences in bail type such as ROR vs financial release
(Demuth 2003, Demuth & Steffensmeier 2004, Schlesinger 2005 Freiburger et al. 2010,
Wooldredge 2012) but no modern studies have taken into account different financial bail types.
Free (2004) does make mention of three such studies that look at more detailed bail outcomes
including Mann (1984), Nagel (1983) and Albonetti et al. (1989). Mann (1984) is a qualitative
study of a small sample of women felons in Atlanta. Although Mann’s discussion of bail is brief,
it is meaningful in that she considers monetary and nonmonetary bail. She finds no significant
effects of bail type; however, this may be because of the use of the simplistic chi-square method
as well as the small sample. It may also be that the effects of race differ for women than they do
for men. Nagel (1983) discusses the need to consider cash alternatives when looking at bail
amount noting that “a cash alternative is most distinctive for its unpredictability” (510). Nagel
notes that the options for bail beyond cash or ROR are numerous and vary greatly between
judges and suggests that future studies should explore whether there are patterns to non-
monetary bail decisions or if they are simply a reflection of individual judge discretion. Most
notably, Albonetti et al. (1989) considers 9 bail outcomes in there paper on pretrial judicial
decision making and race: personal recognizance, unsecured bond, unsecured bond plus
supervision, bail contingent on a 10% cash deposit, bail contingent on a 10% cash deposit plus
supervision, bail with collateral or collateral plus supervision, bail contingent on a surety bond,
bail contingent on a surety bond plus supervision, and bail denied (67). The thoroughness of this
variable reflects both the complexity of bail and emphasizes the oversight made by literature
since the article was published in 1989. The analysis treats these nine possible bail outcomes as
an ordinal variable measuring “bail severity” with recognizance being “1” and denied bail being
“9” which may discount key nuanced variations between groups. Findings do not indicate
13
significant differences between black and white defendants for white defendants on the surface;
however, Albonetti et al. (1989) do find that education and income influence the severity of bail
differently for black than white defendants with a stronger negative correlation occurring for
whites. The study also indicates severity of statutes and defendant dangerousness relates to more
severe bail outcomes for white defendants than for black defendants. This result may indicate
that there is disparity towards white defendants for bail or it may be that black defendants are
given more severe bail whether or not they have high grade charges or indicators of
dangerousness. In general, studies that have included bail outcomes have generally found weak
effects of race but it is difficult to say if this is due to an actual null finding or because of poor
data, ineffective analytical techniques, or inaccurate variable formation. Furthermore, the bail
system has changed since the 1980s and results may not reflect the modern processes and
outcomes.
Gaps in Research
While research on bail has analyzed diverse outcomes, few studies look at the process
holistically. Studies differ in most in their conceptualization of the dependent variable, with
many forced to choose how to operationalize bail based on data availability and convenience.
The few studies that have been able to look at multiple pretrial outcomes have found nuanced
and often competing effects of detainment, bail amount, and bail type (Demuth 2003, Demuth &
Steffensmeier 2004, Albonetti et al. 1989). Understanding of differences in monetary bail types
is all but missing from research despite early discussion of the possible importance of separating
the different combinations of monetary sanctions a defendant may be given (Nagel 1983,
Albonetti et al. 1989). This study adds to the literature by considering multiple bail outcomes and
differentiating between bail types so to more fully understand the nuanced nature of judge
decision making in the pretrial process
14
More complete treatment of race/ethnicity is often lacking in the literature thus far. Most
studies only consider differences in black and white defendants and do not account for
Hispanics. Studies that look at black versus white outcomes may be coding Hispanic defendants
as white, while studies that look at nonwhite versus white outcomes are mixing together black
and Hispanic defendants. This means that such studies are not strictly comparable. Furthermore,
meaningful differences may exist between Hispanic, black, and white defendants, and these may
be conflated. Demuth (2003) and Demuth and Steffensmeier (2004) found meaningful
differences between white, black, and Hispanic defendants, which indicates that it is important to
consider these three groups separately.
Finally, much of the research on bail outcomes has not been well grounded theoretically.
In order to move the field forward it is necessary to not just look at how bail outcomes differ by
race but to also consider why these differences may be occurring. By grounding research on bail
in theory researchers will be better able to create conceptually meaningful models and will be
better equipped to paint a broader picture on the overall process. Based on parallels to other court
processes such as prosecution, plea bargaining, and sentencing, I believe that the focal concerns
perspective has the potential to add a strong frame to our formation of the bail process. Previous
research on bail has supported this framework through either the use of variables measuring the
key concepts of focal concerns or through suggestions on its future use in research (Albonetti et
al. 1989; Freiburger, Marcum, & Pierce 2009; Demuth & Steffensmeier 2004).
Focal Concerns Perspective
The focal concerns perspective theorizes that court actors are motivated by a set of
general goals that frame their decision making. In the original application; the focal concerns of
sentencing were categorized under blameworthiness of the defendant, danger to the community,
15
and practical constraints (Steffensmeier, Ulmer, & Kramer 1998). Court decision makers apply
their own interpretations to what constitutes these concerns and weigh factors of a given case
based on these guiding orientations. While focal concerns are subjective by nature, court
community norms and legal guidelines shape court actor interpretations.
Despite the existence of legal guidelines, judges often do not have the time or resources
to learn all the particulars of a case that are relevant; this is known as bounded rationality
(Albonetti 1991). Also, court actors may not have access to all the information on a particular
case. For instance, judges in bail hearings may be missing police reports on the defendant and
must depend on limited information from asking questions of the defendant or officers instead.
Court actors handle this bounded rationality or limited information by using perceptual
shorthands to make their decisions. Perceptual shorthands are stereotypes of cases based on past
experiences and the norms of the court communities. For example, judges might interpret
characteristics like gender and race of a defendant in a drug case as indicators that the defendant
is “typical.” So, while legal factors such as past offenses and crime types are commonly used to
assess blameworthiness and dangerousness (Spohn 2000); extralegal factors like race gender and
age also come into play (Spohn 2009).
Judges also take into consideration practical constraints when making their decisions.
Practical constraints refer to organizational, monetary, or other constraints and consequences
both to the court and to the defendant. Applied to bail decisions, for example, jail space and cost
of housing could be considered when deciding whether or not to remand a defendant. Practical
constraints on the defendant, such as having children to look after or a job, can also be used as
justifications either for or against release. Sentencing research has shown judges sometimes
further magnify the effects of extralegal factors by justifying that practical constraints, such as
16
being imprisoned, will have a lesser impact on certain defendants (Steffensmeier et al. 1998;
Kramer & Ulmer, 2009). In the case of bail, for example, a judge may be more willing to remand
a black male because the judge assumes that black males are less likely to have legal
employment than white males. Judges do not only use practical constraints to justify harsher
decisions, they also take into account practical constraints when considering leniency. For
example, some studies on sentencing have found that judges are less likely to incarcerate females
because of concern that women have family responsibilities as mothers (Daly, 1987;
Steffensmeier, 1980). Along with practical constraints on the defendants, practical constraints
such as jail capacity and budget affect decision making.
Focal Concerns theory has been applied mostly to look at prosecutorial and judicial
decision making in sentencing (see review by Ulmer 2012). These studies have generally found
that a combination of legal and extralegal case characteristics, including race are used by judges
to assess the dangerousness and blameworthiness of the defendants as well as any practical
constraints.
Few studies have applied Focal Concerns to processes other than sentencing. It may be
that court actors involved in other practices have a different set of concerns or weigh concerns
differently than do actors making sentencing decisions. Huebner & Bynum’s (2006) study of
parole decisions, for instance, noted that concerns discussed by parole boards when deciding
outcomes for sex offenders were not always synonymous with the concerns during sentencing.
For example, they found that community protection seemed to be of a larger concern than the
blameworthiness of the defendant. Furthermore, older defendants were assessed to be more
dangerous than younger ones, in contrast to findings in sentencing research. These differences in
findings are important because they emphasize how focal concerns theory is interpretive.
17
Varying factors implicate the focal concerns to court actors based on situational and goal
contexts. While some of the correlations found in sentencing crossed over to parole correlations,
they did not all apply or apply in the same way. It is likely that indicators of focal concerns for
court actors making pretrial detention sentence decisions will be differential from indicators in
other court processes.
Freiburger, Marcum, and Pierce (2009) suggest that research on pretrial decisions show
patterns that implicate focal concerns as a main mechanism for decisions. In particular,
Freiburger and colleagues note that the influence of prior convictions and offense serious support
the theory that judges consider factors related to dangerousness and blameworthiness. Demuth
and Steffensmeier (2004) emphasize how focal concerns would predict black and Hispanic
defendants to have more severe pretrial outcomes. They note that research on court actors finds
that black and Hispanic defendants are often stereotyped to be “violent-prone, threatening,
disrespectful of authority, and more criminal in their lifestyles” (Demuth & Steffensmeier 2004,
p 226). They further theorize that judges perceive Hispanic and black defendants to be a higher
flight risk because of the assumption they “have fewer community ties and may also be illegal
immigrants” (p 226). From my review of law as well as discussing the process with judges, I
believe that the largest considerations are community safety and flight risk. Community safety is
assessed by looking at how likely the defendant is to harm others while out on bail which marks
it as a measure of dangerousness. Flight risk is important to judges because of a combination of
dangerousness if the defendant goes on the lam and court costs if the defendant skips a hearing:
which is related to practical constraints. It may be that when constrained by limited information
and time, judges use race as a perceptual shorthand where black and Hispanic defendants are
perceived as more dangerous and blameworthy, and more likely to be costly to the court.
18
However, the focal concerns perspective emphasizes how Judges use all available
information to make their decisions and it may be that judges with more information are less
likely to be influenced by the race of the defendant. Demuth and Steffensmeier (2004) found that
racial differences between female defendants were less pronounced than for male defendants. In
fact, for some outcomes such as bail amount, there were no racial differences for female
defendants. It may be that the perception of females as less dangerous and blameworthy
overrides the perception that Hispanic and black defendants are more dangerousness and
blameworthy. Knowledge related to legal factors also may influence how much judges ultimately
take race into account. Wooldredge (2012) found that bond amounts and the likelihood of getting
ROR rather than bail were significantly different for white and black defendants but these
differences became not significant when the model accounted for prior record, felony counts,
number of counts, and whether or not the defendant was under the influence of drugs during the
crime. Judges who have more complete case details or more details on the criminal history of a
defendant may take these legal factors into account more strongly than race when making their
decision. If this is the case, racial disparity will be lower or nonsignificant when comparing cases
with the most complete legal information.
Hypotheses
Based on Albonetti et al. (1989) the severity of bail types can be ordered from most
severe to least as follows: Remand, Monetary Bail, Unsecure Bail, Released on Recognizance.
Going off prior research as well as the theory that judges will use race as an indicator of
dangerousness blameworthiness, and flight risk; I believe that when choosing between two
outcomes, judges will be more likely to err to the more severe one for black and Hispanic
defendants. Thus my first three hypotheses:
19
H1: Black and Hispanic defendants will be more likely to be remanded rather than get
monetary bail in comparison white defendants
H2: Black and Hispanic defendants will be more likely to be given monetary bail as
opposed to unsecure bail compared to white defendants
H3: Black and Hispanic defendants will be more likely to be given unsecure bail than
ROR in comparison to white defendants
The focal concerns perspective theorizes that racial differences occur because judges use
race as perceptual shorthand when legal factors are absent. Assuming this is accurate, there
should be less racial disparity when judges have more legal information.
H4: Legal factors such as prior record and crime severity will moderate the effect of race
on bail type
While prior research has found no significant difference in bail amount between black
and white defendants when controlling for legal factors, I believe this may be due to the
conflation of monetary and unsecure bail. If white defendants are more likely to be given
unsecure bail, then high unsecure bail amounts may hide the racial differences within monetary
bail. However, it may also be that bail amount is more explicitly linked to crime type than bail
type because judges use going rates from bail schedules to calculate amount.
H5a: Black and Hispanic defendants will be given higher average bail amounts than
white defendants when bail is separated into monetary and unsecure bail
H5b: Black and Hispanic defendants will not have significantly different bail amounts
than white defendants when bail is separated into monetary and unsecure bail
20
Data and Methods
Datasets
Two data sources were utilized for this study. First, cases were sampled from the
Pennsylvania Commission on Sentencing (PCS) dataset. The PCS dataset consists of convicted
offenders in the State of Pennsylvania from 1986 to 2011. Identification numbers needed to
randomly sample dockets were procured from the PCS as well as information on race, gender,
age, and prior record. Second, I collected data on bail as well as initial charges from common
pleas court dockets accessed through the Unified Judicial System of Pennsylvania web portal
(https://ujsportal.pacourts.us/DocketSheets/CP.aspx). While these dockets are public access,
there is no available aggregate dataset and a full name or identifier is required to open a given
sheet. I therefore used offense tracking numbers (OTNs) acquired from the PCS as my sampling
frame. OTNs are a unique number given by the Pennsylvania courts at the time of the first
arraignment. This form of identification was used over alternative measures (such as docket
numbers or state identification numbers) for three reasons; it allowed me to differentiate by court
case as well as individual so as to not to have multiple dockets per number, it was given at the
time of arraignment which was my period of interest, and it is the most complete unique
identifier in the PCS dataset. Due to data transfer issues, both the PCS dataset and Pennsylvania
court dockets before 2004 may have more misinformation and missingness. Therefore, I sampled
from cases from 2004-2011.
While ideally this study would be able to look at offenders who were not convicted as
well as those convicted, this was not feasible within the scope of this study. Currently, there is no
past research or theory that would lead me to expect significant differences in how non-convicted
and convicted defendants would be treated in their bail hearings. However, convicted defendants
21
may be more likely to have more serious charges which may mean that more severe bail types
like monetary bail or remand may be overrepresented in my sample. Also, if there is racial or
ethnic selection bias in conviction, this study may, therefore, underestimate race/ethnic effects on
bail: making my estimates conservative. However, there is ample variation in the bail types,
amounts, and race/ethnicity in the dataset so if there is bias in conviction, it does not fully
remove all differences. Secondly, information was collected on original charges and that data
was used rather than the final charges of which defendants were convicted. This means that the
charge information properly reflects the knowledge of the charges that one would expect judges
to have during the bail hearing.
Sampling
Since the focus of my study is racial differences between outcomes rather than across
outcomes, I sampled in the way that garnered me comparable groups of white, black, and
Hispanic offenders. I choose to sample from the four counties with the largest number of
Hispanic offenders, excluding Philadelphia and Allegheny (Pittsburg) county as they were too
large for me to code an adequate percentage of cases. The four counties sampled are Lancaster,
Lebanon, Lehigh, and Northampton, which are all suburbs of Philadelphia. This is due to
Pennsylvania being in general majority white; counties that have larger black and Hispanic
populations tend to be either cities or close suburbs of cities. While having four counties within a
small area of Pennsylvania is not ideal, these counties do represent areas where racial disparity is
most likely to occur since there are larger minority populations in comparison to more rural
counties. While my findings cannot be extrapolated to Pennsylvania as a whole, using four
counties within one state allows me to look more at a broader sample of cases than previous
studies that sampled from a single county. I sampled ten percent of each county’s convicted
22
defendants with one third of the sample being white, one third black, and one third Hispanic3. All
together I sampled 894 cases from Lancaster, 516 from Lebanon, 990 from Lehigh, and 816 from
Northampton for a total of 3,216 individual cases.
Coding
All information on bail was collected by me personally from the online Pennsylvania
court documents. By coding this data alone, I remove the possibility of intercoder differences. To
account for possible mistakes, a small selection of court cases were examined more than once to
ensure accuracy; especially in cases where bail outcomes were particularly complicated or
unusual, or in cases where there was a higher degree of missingness for certain variables.
Pennsylvania court dockets are stored as pdfs in text rather than numerical format so data on bail,
charges, and judges had to be recorded by hand into a separate datasheet. Each OTN was typed
into a number field on the Unified Judicial System of Pennsylvania Web Portal which brings up
a link to a docket sheet and a court summary. On the first or second page of each docket sheet
there are two boxes entitled “bail information” and “charges” (see Figure 1) from which I coded
all given information. The bail information box begins with the “bail action” which refers to each
bail hearing or change to bail that occurs for the defendant. Many defendants are actually given
more than one bail as further arraignments can occur if the defendant or their lawyer disputes the
first charge. In fact, in Pennsylvania, all defendants have a right to request a reassessment of
their bail without needing to give a reading. Despite this right, only about one third of defendants
are given more than one bail action; possibly because defendants are unaware of this right and
3 While sampling of race was disproportional, no total population was sampled and the highest percentage sampled was 41% of Hispanic defendants in Lancaster (see table 18). Consultations with multiple methodologists both in and out of Criminology leads me to the conclusion that the percent sampled will not cause problematic issues with representation considering my purpose of conducting a comparative study. Due to constraints on time to code, it was not possible to sample large enough percentages of cases to have enough racial diversity in a randomized sample.
23
public lawyers do not have the time to attend extra hearings. It is important to note the existence
of multiple bail actions, however, as this is not noted in previous studies: making it unclear
whether researchers chose to consider the first or last bail that was set for a defendant. Being that
defendants who are given secondary bail actions are not randomly selected as well as the
possibility of more information for the judges and lawyers, all bail actions are not comparable.
For this study, I chose to consider only the first bail action which would have to have been made
within 12 hours of arrest. Following the bail action and date of that action, the information box
provides the bail type, percentage, and amount. It should be noted that, while some states
commonly only require 10% of bail, my data only had a percentage given in 6.5% of cases.
Generally, states that use percentages are stipulating that, essentially, 10% of bail is monetary
while the other 90% is unsecured, that is, paid if future court dates are missed. The amount
variable in my analysis was adjusted to reflect the actual amount paid if bail was both monetary
and a percentage was provided.
Along with bail information, I also collected information on initial charges (see figure 2).
It is common for defendants to have charges dropped or to be sentenced for a different set of
charges than they were initially given (source here). It is important to account for the initial
charges since the judge presiding over the bail hearing has no ability to know of which charges
the defendants will be convicted and which charges may be dropped or decreased. I recorded the
number of total charges, the number of each grade given (S, M, M1, M2, M3, F, F1, F2, F3), and
the statutes and statute descriptions for the highest grade charges. Since the documents were pdfs
and I had to type in each statute and description in to my data by hand, I was unable to collect all
charges that defendants had. However, informal consultation with Pennsylvania judges and other
court actors led me to the conclusion that magistrate judges generally only consider the top
24
charge when making bail decisions due to time pressure and the inconsistency of how police
charge. By only looking at top charges, I am cutting down on background noise that is likely not
very consequential to a judge’s decision making.
After this collection was finished, I went back through data points with missing
information and attempted to fill in blanks using information from lower court dockets that may
have been transferred to criminal court. The final sample was 3,121, after dropping missing cases
and cases where the bail dates where either before arrest or after conviction indicating the docket
was from a different case, giving me a missing cases rate of 3%.
Measures
Dependent Variable
The dependent variables in this study are bail amount, and the type of bail action
defendants were initially given. There are four main options judges have for bail type which
include remand, release on recognizance (ROR), monetary bail, or unsecured bail. While
originally I expected to be able to compare all four possible bail outcomes the data shows that,
interestingly, very few defendants are remanded or released without some sort of bail (see table
1). In fact, of the 46 defendants who were remanded, the most common charges were homicide
(14 defendants) or skipping bail previously (9 defendants). Since Pennsylvania law states that
remand should only occur if the defendant is up for a life or death sentence or if there is a clear
and specific justification it may be that judges avoid remanding defendants unless absolutely
necessary and, instead, give extremely high bail amounts that are impossible to pay. In fact, there
are a few cases of defendants charged with homicide who were not remanded but were given
monetary bail over one million dollars which they did not pay. Since homicides are clearly
outliers regardless of race, the cases with homicide charges were dropped from the final
analyses. Perhaps more surprisingly, only 86 defendants were released without any type of bail
25
or stipulation. This implies that either judges believe most defendants to be some sort of danger
to the community or a flight risk or, more ominously, that judges believe defendants deserve
some sort of punishment despite the law stipulating bail should not be used as such. Since so few
defendants were detained or released on recognizance, the multivariate models compare only
defendants given monetary or unsecure bail.4
Independent Variables
In order to measure racial differences I created three dummy variables for black,
Hispanic, and white defendants.
Control Variables
From the docket data I was able to create variables controlling for the severity of the
charges against each defendant. I control for the current grade of the crime (M, M3, M2, M1, F,
F3, F2, and F1) which should indicate both dangerousness and blameworthiness. My grade
variable is operationalized as a continuous variable ranging from one to eight with “M” being
one and “F1” being eight. Other iterations of this scale where tested with no substantive
differences seen in outcomes. After careful consideration and discussion with Pennsylvania
judges, I decided not to control for the number of charges given. My data showed me that more
charges occurred when crimes were of a low grade, in particular forms of fraud such as check
fraud or welfare fraud were the most likely to have a large number of charges. Including a
variable accounting for number of charges may cancel out other measures of charge severity as it
appears to be inversely related to the seriousness of the crime. As mentioned previously,
informal interviews with judges confirmed my belief as I was told that, during bail hearings;
judges generally considered only the top charges and did not take number of charges into
account. Along with grade, I made binary variables indicating whether the top charge was a
4 See table 17 for more information on how the dependent variables were coded.
26
violent crime, property crime, or drug dealing charge. Previous studies on bail have indicated
that these categories might have a differential effect on bail since they may be correlated with
seriousness and blameworthiness (Source). By using binary variables categorizing crime type, I
am better able to compare like cases rather than assuming that all types of crimes are evaluated
in the same way. Finally, county location was controlled for with Lehigh as the reference county
due to it being the largest.
The PCS also provided a variety of important control variables including age, gender, and
Prior Record Score (PRS). PRS is used by Pennsylvania to measure the severity of prior
convictions. Using the PRS as a control variable allows me to account for previous convictions.
Prior record scores range for 0-5 plus two score indicators RFEL for a repeat felony (coded as 6)
and RVOC for a repeat violent crime (coded as 7).The SCPS dataset has been most commonly
used in prior studies of state bail outcomes and it has been noted that a weakness of SCPS is the
difficulty to create a uniform and detailed measure of prior history (Ulmer, 2012). However,
prior record scores are constructed for each individual defendant after conviction so, while
judges may have information on prior record, judges considering bail would not have PRS itself.
Still, PCS acts as a relatively good proxy variable for prior convictions about which judges likely
have some knowledge. While it is unclear to what degree judges deciding bail are aware of and
are influenced by prior record, previous convictions have been shown to be consistently the
strongest predictor of future arrests and convictions so it is an important to account for the
existence and severity of prior convictions as they may have an influence on bail decisions as
well (Ulmer, 2012).
27
Results
Descriptives
The bivariate statistics regarding the bail type breakdown of the sample clearly show that
most defendants get some sort of bail rather than remand or ROR (see table 1). In fact, about
68% of defendants are given monetary bail and 28% are given unsecure bail which accounts for
96% of all cases. The majority of previous studies on the bail system have looked at in-out
decisions; relating to whether or not defendants stayed in jail (Shook & Goodkind, 2009;
Freiburger & Hilinski, 2010; McIntyre & Baradaran, 2013). It is unclear if researchers have
defined in-out decisions on whether or not judges remanded defendants or by whether or not
defendant’s stayed in jail. If the former, studies may be comparing a very small proportion of
defendants to the majority. If the latter, studies are conflating defendants who are remanded with
defendants who are given monetary bail and either cannot or choose not to pay. This is an issue
because defendants who are poorer are more likely not to be able to post bail and SES may be
correlated with race. Therefore, it may appear that black and Hispanic defendants are perceived
as a higher threat by judges when, in actuality, it is the economic system that is creating bias
rather than the court system.
When bail decisions are broken down further by race; it appears that black and Hispanic
defendants are more likely to get monetary bail than white defendants, while white defendants
are more likely to get unsecured bail than minority defendants (see table 2). Black defendants are
around 3% more likely than the average to get monetary bail and 3% less likely to get unsecured
bail while Hispanics are about 6% more likely to get monetary and 6% less likely of getting
unsecured. In comparison, white defendants are around 9% less likely on average of getting
monetary bail and 9% more likely of getting unsecure bail. The matching differences appear to
imply that judges are deciding between monetary and unsecured bail in these decisions rather
28
than between remand and monetary bail or unsecured and ROR. Since other bail types show
similar likelihoods for all three races, it appears that disparity is most likely to occur in cases
where judges are choosing between money-based bail types.
Turning to the control variables, patterns are evident within variables as to who is given
unsecure and who is given monetary bail. The percentage of defendants who are given monetary
bail increases as the grade of their charges go up. The one exception is if F grade charges are
highest which has a higher percentage than F3s. This may because F grades represent a mixture
of F3, F2, and F1 charges and are, therefore, representing higher crime severity than F3 charges.
This pattern implies that judges do take into account crime severity as a factor that implies
dangerousness and blameworthiness and are more likely to give more severe crimes monetary
bail. Similarly, the percent of defendants given monetary bail goes up steadily as prior record
score goes up. The largest jump between scores is between no prior record and a score of one-
meaning that having a prior record may increase a defendant’s chance of getting monetary bail
by as much as 9.68%. This has interesting implications as it may mean that defendants are not
only assessed by their current charges, but by past crimes. Since Pennsylvania has no statute of
limitations on prior record scores, some defendants may not have even committed a crime for
many years yet may still be given monetary bail as opposed to unsecure based on their past.
Turning to gender, males have a 73.99% likelihood of being given monetary bail in
comparison to unsecure bail while females have only a 55.68% likelihood. This may be because
males are considered more dangerous and blameworthy than females. However, it may also be
that males commit more severe crimes or are more likely to have prior records. Interestingly, the
age variable does not increase steadily but, instead has a curve that peaks between the ages of 20
and 40 before decreasing. It may be that judges are more lenient of teen defendants because their
29
immaturity makes them seem less blameworthy for their actions and more lenient of older
defendants because they are seen as less dangerous. It may also be that the types of crime
committed may vary by age group.
Over 90% of cases involving drug dealing charges or violent crime charges are given
monetary bail in comparison to unsecure bail. Under focal concerns, these types of crime may be
seen as indicators of dangerousness as well as blameworthiness of these defendants. It should be
noted that only 73 of the drug dealing charges are to white defendants in comparison to 227
black and 204 Hispanic defendants (supplementary table reference here). It may be that the high
percentage of drug dealing defendants given monetary bail is due to race as opposed to the
perception of the crime itself. In contrast to drug dealing and violent crime charges, around 68%
of defendants who have committed a property crime are given monetary bail which is slightly
less than the around 71% given monetary bail rather than unsecure bail in the entire sample.
Since property crimes often do not directly endanger any person’s life, they may be perceived as
less reprehensible and less likely to indicate a danger to the community.
Finally, there appears to be a correlation between county size and the percent given
monetary bail. As county size gets bigger, the percentage of defendants given monetary bail also
increases. This could be due to a number of factors. It may be that more serious crimes are
committed in more urban areas or there are more repeat offenders in those areas. However, this
trend may also be due to practical constraints such as jail space and budget that occur when
courts have to handle larger populations.
Looking at how bail amount differs between monetary and unsecure bail adds
complications to our understanding of the pretrial process. Firstly, monetary bail has a much
wider array of bail amounts/much greater variation in amounts than does unsecure bail (Table 4).
30
The largest unsecure bail given is $200,000 while, in contrast, the largest monetary bail is
$2,000,000. This makes bail amounts somewhat difficult to compare. Furthermore, the average
bail amount for monetary bail is around 32,770 while it is just 7,700 for unsecure bail. This,
however, is not the most accurate measure since there are clearly large outliers within both bail
types. Both bail types have a mode of $5,000 which accounts for 18.3% of cases where bail was
given. In fact, there appears to be “going rates” in terms of amounts that are most commonly
given: $1,000 (5.95%), 2,500 (6.27%), 5,000 (18.3%), 10,000 (13%), 25,000 (10.37%), and
50,000 (7.73%). That is, bail amounts tend to cluster around these levels. Other than those six
values, bail is very diverse with a standard deviation of 44,000 for monetary bail and 13,650 for
unsecure bail.
Multivariate Analysis
Echoing the bivariate analysis pointing to inequality between races, the multivariate model
shows clear differences between white, black, and Hispanic defendants (table 5). While legal
factors like grade and prior record score significantly predict decisions for monetary bail over
unsecure bail, race is also a strong predictor. Black defendants have 39% higher odds than white
defendants of being given monetary bail rather than unsecure bail and Hispanic defendants’ odds
are 82% higher than whites.
I reran the regression using black as the reference group (table not shown) and found that
Hispanic defendants have about 31% higher odds of getting monetary bail than black defendants.
While this difference is significant (p=.02), it is a smaller difference than that between Hispanics
and white defendants. This is important to note since it may be some previous studies comparing
black and white defendants coded Hispanic defendants as white. If Hispanic offenders generally
31
have outcomes more similar to black offenders; coding Hispanics as white would reduce the
differences between white and black defendants which may explain nonsignificant findings.
Along with the significance of race, other variables in the logistic regression support the
focal concerns perspective. Defendants with higher grades and higher prior records, both
indicators of dangerousness and blameworthiness, are significantly more likely to be given
monetary bail compared to unsecure bail. Men have greater odds than women of getting
monetary bail as do younger defendants which may be because they are seen as more of a threat
to the community. The largest odds ratio is for drug dealing charges which shows that drug
dealers have 240% higher odds than non-drug dealers of being given monetary bail. In
comparison, property crime charges lead to a 29% decrease in the odds of getting monetary bail
in comparison to unsecure. This is logical under the focal concerns perspective as property
crimes are generally non-violent so defendants may be seen as less dangerous and blameworthy
than other crimes. While violent crime charges do not have a significant effect, this may be due
to the comparatively small number of violent crime charges (N=327). Still, it can be noted that
the odds ratio is in the expected direction given that violent defendants are more likely to be seen
as dangerous. Lehigh was used as the comparison county in the analysis due to it being the
largest county with the most convicted defendants. All three other counties give less monetary
bail than does Lehigh. This may be because larger counties have more practical constraints such
as jail space and budget issues than do the smaller counties.
Judges may evaluate bail outcomes differently depending on the type of case with which
they are dealing. Accordingly, I ran logistic regressions of bail type within the groups of drug
dealing crimes, violent crimes, and property crimes. Interestingly, the models’ race effects
differed depending on the type of crime category. Black defendants have 310% higher odds of
32
getting monetary bail compared to unsecure than do white defendants when looking at only drug
dealing crimes (table 6) while Hispanic defendants have 187% higher odds. Unfortunately,
Pennsylvania statutes do not specify what drug or how much is being dealt in drug dealing
charges so this effect may be due to different practices that correlate with racial differences.
However, it may also be that black and Hispanic drug dealers are perceived as a higher threat and
more blameworthy than white defendants. In the full model, Hispanics differed from whites
more than blacks did yet within drug dealing black defendants have much higher odds compared
to white defendants.
The model for violent crime does not show as strong effects (table 7). Black defendants
have a higher odds ratio compared to white while Hispanics have a lower ratio but neither ratio is
significant so these should be interpreted with caution. Grade and gender are the only significant
predictors with higher grades being more likely to get monetary bail and men being more likely.
Looking at the descriptives, 91% of violent crime defendants are given monetary bail compared
to about 9% given unsecure. If a crime is violent, judges may automatically classify defendants
as dangerous and not resort to using other perceptual shorthands such as race to make their
decision.
The property crime logistic model (table 8) is perhaps the most interesting as it has a
larger N than do the other crimes (N=1,035) and there is a relatively even split between the races
in comparison to within the other crime types. Similar to the full model, Hispanics have 75%
higher odds of monetary over unsecure bail in comparison to white defendants. However,
interestingly Black defendants do not have significant differences from white defendants. It is
unclear based on the data I have on why this may be the case and perhaps property crimes should
be studies in more depth in future studies on bail.
33
It may be that judges resort to assessing defendants on factors such as race and gender
only in the absence of legal indicators. While every case has crime types and grades, only about
45% of defendants in my sample have a prior record. Perhaps effects of factors like race might
be influenced by whether or not the judge has information from prior cases. Accordingly, I ran a
logistic regression model that employed race-PRS interactions (table 9). Interestingly, there is a
significant interaction between PRS and Hispanic defendants which implies that judges may
account for race differently depending on the prior record of the defendant. Black defendants, in
contrast seem to not be treated differently depending on whether or not they have a prior record.
Running the model for only defendants with no prior record, the odds of monetary bail
for Hispanic defendants is 127% higher than for white defendants (table 10). However, looking
at a model including only cases with prior records the odds ratio for Hispanic defendants is non-
significant (table 11). It appears that the racial disparity for Hispanic defendants depends upon
the existence of a prior record. Hispanic defendants without a prior record are more likely than
similar white defendants to receive monetary bail as opposed to unsecure bail but a prior record
tends to “equalize” Hispanics and whites. In contrast, the odds ratio for black defendants is
similar both with and without a prior record. This implies that disparity for black defendants is
consistent despite other legal information while disparity for Hispanic defendants depends on
other factors.
Models looking at bail amount show different results in comparison to models on bail
types. For this model I ran a regression model with bail amount coded in thousands as the
dependent variable (table 12). An initial model that combines monetary and unsecure bail
outcomes shows significant correlations with legal indicators such as grade, PRS, and crime
type. Bail amount rises for defendants with higher grades and higher prior record scores which
34
indicates judges are taking into account dangerousness and blameworthiness in their decisions.
Drug dealing charges and violent crime charges also raise bail amounts while property crimes
lower bail. This is also in line with the theory that judges give higher bails to more severe crimes
since they imply dangerousness, Race and age are notably not significant in this model which
implies that extralegal factors are taken into less consideration in bail amount than in bail type.
Perhaps this is due to going rates within court communities that set suggested amounts for
certain crimes that differ infrequently. It may also be that there is more oversight on bail amount
than on bail type and judges are more accountable for their decisions. Gender also is significant
in this model but it may be because gender is confounded with crime type and severity.
It can be argued that since defendants with unsecure bail are not required to pay the given
amount that their cases are meaningfully different than monetary bail cases and monetary models
should be separated by type. Tables 13 and 14 show how the bail amount regression changes
when only considering monetary or unsecure bail. In these models, the correlation of amount
with PRS becomes non-significant which may mean that PRS was confounded initially with
grade. This makes sense if it is assumed that judges use bail schedules based on crime grade and
type to decide on bail amount. It may also be that judges use bail amounts to account for the
current crime and then decide on bail type based on other characteristics such as race or criminal
history. The monetary bail model shows a significant correlation with sex while the unsecure bail
model does not. This supports my earlier conjecture that sex is confounded with severity as
crime severity ranges less for those given unsecure bail than among those given monetary bail.
Likewise, violent crime charges are shown to have a significant effect on monetary bail and not
unsecure likely due to the small number of violent crime cases that are given unsecure bail.
35
Discussion To understand racial disparity in the court system, it is important to study each court
process to see how they interact and build on each other. Analysis of the bail system shows that
pretrial processes are much more complex and important than may have been previously
assumed. This study aimed to understand the bail process in a more holistic manner than
previous studies by taking into account multiple types of bail outcomes and Hispanic defendants
along with white and black defendants. My first hypothesis that minority defendants would be
more likely to be remanded rather than given monetary bail was rendered moot due to the low
number of people who were remanded. It seems that judges tend to remand defendants in only
the most serious of cases or for defendants who have a past history of skipping court
appearances. Similarly, my third hypothesis that black and Hispanic defendants would be given
unsecure bail rather than ROR in comparison to whites also had few cases to compare. These
findings reinforce the importance of considering bail carefully as it is the most likely outcome
for all defendants. My hypothesis that racial disparity would be found when comparing the
likelihood of monetary bail to unsecure bail was strongly supported. The multivariate analysis
from this study shows clear significant differences between outcomes for white defendants and
outcomes for black and Hispanic defendants. Black defendants have odds 39% higher of getting
monetary bail as opposed to unsecure in comparison to white defendants, and Hispanics’ odds
are 82% higher. While all defendants are most likely to be given some amount of bail, white
defendants are much less likely to have to actually pay that amount. This finding supports the
focal concerns perspective that race may be used as an indicator of blameworthiness and
dangerousness, particularly in absence of other legal information.
Racial disparity in bail type interacts further with crime type and criminal record.
Looking at racial differences in bail for defendants who committed violent crimes, there is not
36
significant disparity and, instead, only crime grade and gender are significant. For drug crimes,
black defendants have especially high odds of monetary bail-310% (odds ratio 4.10) compared to
whites. This may be because black defendants are seen as “typical” defendants in drug cases and
their race reinforces the belief of the judge that black defendants are more dangerousness and
blameworthy. Note, however, that I was not able to differentiate between different types of drug
cases. It may simply be that black defendants are more likely to have more severe drug charges
brought to court in my sample. Turning to property crimes, Hispanics are significantly more
likely to be given monetary bail rather than unsecure when compared to whites but there are no
significant differences between black and white defendants. This may once again be because
Hispanics fit judges’ view of the “typical” property criminal or it may be due to differences in
the types of cases. Future research should look into what judges consider the typical offender for
different types of crime and how this may interact with bail decisions.
My fourth hypothesis stated that legal factors may moderate the relationship between race
and bail outcomes as has been found in previous studies. Indeed, when considering interactions
between prior record and race the disparity between white defendants and black and Hispanic
defendants differs. Black defendants have greater odds of monetary bail than whites with or
without a prior record, but differences between white and Hispanic defendants greatly decrease
within the group with prior records. Based on the margins, it seems that prior record greatly
increases the chances of a white defendant being given monetary bail and increases further the
chances of a black defendant being given monetary bail, but does not have a marked effect on
Hispanics odds of this outcome. Instead, white-Hispanic disparity in bail is concentrated among
those without prior records. Focal concerns perspective might predict that this discrepancy
occurs because in the absence of legally relevant information, such as prior record, Hispanic
37
ethnicity informs a perceptual shorthand of dangerousness and risk similar to the evaluation of
risk a judge would come to based on the existence of a prior record. A potential reason for this is
that Hispanic defendants who have not “proven” themselves in previous cases may be thought to
be a higher flight risk due to assumed connections to Mexico. This would mean that, ironically,
Hispanics with prior records could be viewed more favorably by judges. In contrast, no such
perceptual shorthand exists for white or black defendants meaning that the existence of a prior
record would result in a harsher perception of dangerousness and blameworthiness.
My final hypothesis concerned racial differences in bail amount when splitting bail into
monetary and unsecure bail. While previous studies have found null effects of race on bail
amount, I felt that this may be due to the confounding of differing bail types. In line with
previous research, I did not find significant racial differences in bail amount for monetary or
unsecure bail. Instead, it seems that grade and charge type are the strongest predictors of bail
amount. This may be because judges have more complete guidelines on bail amount than they do
for bail type which decreases their discretion and heightens pressures to conform to court
community standards.
The focal concerns perspective suggests that judge may have primary factors they are
interested in and they use the information at their disposal to assess those concerns. In particular,
judges consider blameworthiness, dangerousness, and practical constraints in decisions on
sentencing (Ulmer 1997). The results from analysis on bail type support the focal concerns
perspective with indicators of blameworthiness such as higher grades and prior record and
indicators of dangerousness such as younger age and being male relating to more severe bail
outcomes. The pattern of the predictions by these legitimate factors bolster the theory that race
38
may be used as a perceptual shorthand by judges for blameworthiness, dangerousness, and
practical constraints.
Limitations
While this study found strong results; there are some limitations to be considered. Due to
data restraints, it was necessary in this study to only use information on convicted offenders. It
may be that non-convicted offenders have certain characteristics that make them different and,
therefore, I cannot extrapolate my results to that population. Furthermore, if bail outcomes have
an effect on sentencing outcomes, it may be that this sample over-represents those with more
severe bail outcomes. For example, Heaton, Mayson, and Stevenson, 2017 found that defendants
detained pretrial are more likely to take plea deals. Considering black and Hispanic defendants in
my sample were more likely to be detained, due to higher bail amounts and potentially worse
economic conditions, the sample of black and Hispanic convicted defendants may not parallel
the sample of white convicted defendants. However, if this is the case, it would be reasonable to
assume it is actually the less severe black and Hispanic cases that are overrepresented (since
similar white defendants may have not been convicted) which means that racial disparity
estimates may actually be conservative estimates. Also, descriptives on bail types and the bail
amount models replicate past studies that were able to use a sample of all defendants (Demuth &
Steffensmeir 2004, Demuth 2003, Wooldredge 2012, Schlesinger 2005).
Secondly, since the dataset in this study was sample by race it cannot be extrapolated to
the counties the actors were pulled from or any other broader geographic population. Therefore,
the results in this study should be considered as racial comparisons rather than geographical
representations. However, the data does show clearly that there are differences between races in
these counties. Finally, certain key variables may explain the relationship between race and bail
39
outcomes. Wooldredge (2008) found that certain economic variables such as education and
financial support accounted for a large portion of the racial differences found in his models. It
may be that disparity between racial groups is due to economic differences rather than
differences in race. However, even if the relationship can be explained by spuriousness with
SES, it is still important that races differ alongside these economic differences and that economic
variables may influence judge decision making.
Conclusion and Future Directions
The bail system has powerful influence both on court processes and defendants and their
families. High bails can lead defendants to have to take out costly loans or else stay in jail.
Furthermore, pretrial detention may change the structure of future court processes such as plea
bargaining and sentencing hearings. It is important for our system to ensure bail hearings are
both fair and consistent. Under current guidelines, judges have a lot of discretion on pretrial
decisions with little oversight and few guidelines or training. Policy makers should consider the
implications of the potential disparity this lax system may cause and work to create measures to
better assess the necessity of bail for the manifest goals of community safety and flight risk. If a
clear set of rules is not in place, it seems likely that latent goals such as punishment may come
into play during decision making.
Future research needs to be done that looks at bail type and amount in other areas. Free
(2004) noted that most studies (18/30) on pretrial detention and bail have only looked at a single
city or county. Subsequent research should look at larger and more diverse areas as well as use
multilevel analysis to parse out the effects of individual-level variation in judges and county-
level variation in court communities Furthermore, it is vital that studies in the future account for
Hispanic defendants as well as white and black defendants. Hispanics are the largest minority in
40
the United States currently yet are often overlooked in court research due to data constraints.
Considering the results that implicate blameworthiness may come into play when judges
consider bail, it would be advantageous to further delve into judges’ focal concerns and goals
when making bail decisions: perhaps in a qualitative study that employs interviews.
Findings in this study support the focal concerns perspective that would theorize Hispanic
and black defendants would be given harsher bail outcomes than white defendants due to
perceptions that they are more dangerousness and blameworthy. In particular, this study shows
that different conditions influence how judges consider race. Legal factors such as grade, crime
type, and prior record were found to override the use of race as a perceptual shorthand in
decisions on bail amount. Furthermore, prior record specified the relationship between race and
bail type with disparity between black and white defendants being similar with or without a prior
record, but disparity between white and Hispanic defendants actually becoming non-significant
when a prior record existed. Future research on why prior records create this distinction may be
especially valuable to further connect judge decision making in bail hearings to the focal
concerns perspective.
Finally, little research has been done on the effectiveness of the bail system as a whole. If
bail is not effective in decreasing community threat and ensuring court appearances, it may be
that the focal concerns of judges are ultimately inconsequential. Nothing is known on how the
defendants themselves frame monetary or unsecure bail amounts when out before their court
appearances. It may be that defendants view both types of bail the same as amounts they would
have to pay if they violated the terms of their bail or missed a court appearance. Inversely, it may
be that monetary bail is seen as a “sunk cost” that, thus, has little to no influence on defendant
behavior and unsecure bail may be considered an empty threat if appearance is necessary for the
41
court system to collect these debts. Ultimately, the actual effect of bail as a deterrent is under
researched and it is difficult to reach any conclusions on how the bail process actually influences
defendant behavior beyond whether or not defendants are detained. This study acts as an
example of the potential extralegal factors, race in particular, that come into play in judge
decision making during the pretrial process and the complexity that goes into those decisions. It
is clear that, at the very least, bail is an instrumental process within overall criminal justice
procedure and researchers would be wise to spend more time focusing on the ins and outs of bail
and how disparity may occur.
42
References
Albonetti, C. 1991. “An Integration of Theories to Explain Judicial Discretion.” Social Problems
38, 247-266.
Albonetti, C., Hauser, R. M., Hagen J., & Nagel, I. H. 1989. “Criminal Justice Decision Making
as a Stratification Process: The Role of Race and Stratification Resources in Pretrial Release.”
Journal of Quantitative Criminology 5(2), 57-82.
Baumer, E. P. 2013. “Reassessing and Redirecting Research on Race and Sentencing.” Justice
Quarterly : JQ, 30(2), 231.
Boehn, E. 2013. “In PA, it’s Easier to be a Judge than a Cosmetologist.” Watchdog.org url:
https://www.watchdog.org/pennsylvania/in-pa-it-s-easier-to-be-a-judge-
than/article_010a976c-bf20-5f4a-a396-57d840144bde.html
Daly, K. 1989. “Neither Conflict Nor Labeling Nor Paternalism Will Suffice: Intersections of
Race, Ethnicity, Gender, and Family in Criminal Court Decisions.” Crime & Delinquency
35(1), 136-168.
Demuth, S. 2003. “Racial and Ethnic Differences in Pretrial Release Decision and Outcomes: A
Comparison of Hispanic, Black and White Felony Arrestees.” Criminology, 41(3), 873-
908.
Demuth, S., & Steffensmeier, D. 2004. “The Impact of Gender and Race-Ethnicity in the Pretrial
Release Process.” Social Problems, 1(2), 222-242.
Eisenstein, J., Flemming, R., & Nardulli, P. 1988. The Contours of Justice Communities and
Their Courts. Boston: Little, Brown.
Free, M. D. 2004. “Bail and Pretrial Release Decisions.” 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(3), 318-334.
Freiburger, T. L., Marcum, C. D., & Pierce, M. B. 2009. “The Impact of Race on the Pretrial
Decision.” American Journal of Criminal Justice, 35, 76-86
Heaton, P., Mayson, S. & Stevenson, M. 2017. “The Downstream Consequences of
Misdemeanor Pretrial Detention.” Stanford Law Review, 69(3), 711-794.
Huebner, B. M., & Bynum, T. S. 2006. “An Analysis of Parole Decision Making Using a Sample
of Sex Offenders: A Focal Concerns Perspective.” Criminology, 44(4), 961-991.
43
Kramer, J. H., & Ulmer, T. J. 2009. Sentencing guidelines lessons from Pennsylvania. Boulder,
Colorado: Lynne Rienner.
LaFrentz, C. D. & Spohn, C. 2006. “Who is Punished More Harshly in Federal Courts? The
Interaction of Race/Ethnicity, Gender, Age, and Employment Status in the Sentencing of
Drug Offenders.” Journal of Justice, Research, and Policy, 8(2), 25-56.
Mann, C. 1984. “Race and Sentencing of Female Felons: A Field Study.” International Study of
Women’s Studies, 7, 160-172.
Minton, T., & Zeng, Z. 2015. Jail inmates at midyear 2014 Bureau of Justice Statistics
Nagel, I. H. 1983. “The Legal/Extra-Legal Controversy: Judicial Decisions in Pretrial Release.”
Law & Society Review, 17(3), 481-516.
Oleson, J. C., Lowenkamp, C. T., Cadigan, T. P., VanNostrand, M., & Wooldredge, J. 2014.
“The Effect of Pretrial Detention on Sentencing in Two Federal Districts.” Justice
Quarterly, 33(6), 1-20.
Sacks, M., & Ackerman, A. R. 2014. “Bail and Sentencing: Does Pretrial Detention Lead to
Harsher Punishment?” Criminal Justice Policy Review, 25(1), 59-77.
Sacks, M., Sainato, V. A., & Ackerman, A. R. 2015. “Sentenced to Pretrial Detention: A study of
Bail Decisions and Outcomes.” American Journal of Criminal Justice, 40(3), 661-681.
Schlesinger, T. 2005. “Racial and Ethnic Disparity in Pretrial Criminal Processing.” Justice
Quarterly, 22(2), 170-192.
Spohn, C. 2000. “Offender Race and Case Outcomes: Do Crime Seriousness and Strength of
Evidence Matter?” NCJRS
Spohn, C. 2009. “Race, Sex, and Pretrial Detention in Federal Court: Indirect Effects and
Cumulative Disadvantage.” University of Kansas Law Review, 57(4), 879.
Spohn, C. 2015. “Race, Crime, and Punishment in the Twentieth and Twenty-First Centuries.”
Crime and Justice, 44(1), 49-97.
Steffensmeier, D. 1980. “Assessing the Impact of the Women’s Movement on Sex-Based
Differences in the Handling of Adult Criminal Defendants.” Crime & Delinquency 26(3),
344-357.
Steffensmeier, D., Ulmer, J., & Kramer, J. 1998. “The Interaction of Race, Gender, and Age in
Criminal Sentencing: The Punishment Cost of Being Young, Black, and Male.”
Criminology, 36(4), 763-797.
44
Ulmer, J. T. 1997. Social Worlds of Sentencing: Court Communities Under Sentencing
Guidelines. Albany: State University of New York Press.
Ulmer, J. T. 2012. “Recent Developments and New Directions in Sentencing Research.” Justice
Quarterly, 29(1), 1-40.
Williams, M. R. 2003. “The Effect of Pretrial Detention on Imprisonment Decisions.” Criminal
Justice Review, 28(2), 299-316.
Wooldredge, J. 2012. “Distinguishing Race Effects on Pre-Trial Release and Sentencing
Decisions.” Justice Quarterly, 29(1), 41-7
45
Appendix
Figure 1: The Bail Process
Arrest
Bail
Hearing
Remand Monetary
Bail
Unsecured
Bail ROR
Bail Amount
Set
Bail Not
Paid
Bail Amount
Set
Bail Paid
Released
from Jail Stay in Jail
46
Tables
Table 1: Frequencies of Bail Types
Bail Type Frequency Percentage
Remanded 32 1.03
ROR 86 2.76
Monetary Bail 2,100 67.48
Unsecured Bail 869 27.92
Other (non-monetary conditions) 20 0.65
Missing 5 0.16
Total 3,107 100.00
Table 3: Bail Type by Control Variables
Unsecure Bail Monetary Bail Total N
29.27% 70.73% 2,969
Grade
M 59.93% 40.07% 292
M3 57.24% 42.76% 145
M2 40.64% 59.36% 438
M1 38.91% 61.09% 478
F 13.11% 86.89% 412
F3 22.83% 77.17% 587
F2 13.09% 86.91% 298
F1 6.27% 93.73% 319
Prior Record Score
0 35.96% 64.04% 1,596
1 26.28% 73.72% 468
2 21.91% 78.09% 324
3 19.10% 80.90% 178
4 17.12% 82.88% 146
Table 2: Bail Type by Race
Bail Type White Black Hispanic Total
Remanded .78% 1.06% 1.25% 1.03%
ROR 2.80% 2.70% 2.80% 2.76%
Monetary 58.71% 70.20% 73.55% 67.48%
Unsecured 36.67% 25.27% 21.81% 27.92%
Other .86% .58% .49% .65%
Missing .19% .19% .10% .16%
Total 1,039 1,037 1,036 3,112
47
5 16.89% 83.11% 225
7 12.50% 87.50% 32
Sex
Male 26.01% 73.99% 2,441
Female 44.32% 55.68% 528
Age
18-20 32.27% 67.73% 626
21-30 26.03% 73.97% 1,141
31-40 26.32% 73.68% 684
41-50 35.58% 64.42% 385
51+ 39.69% 60.31% 131
Crime Type
Drug Dealing 9.96% 90.04% 502
Violent Crime 8.87% 91.13% 327
Property Crime 32.27% 67.73% 1,035
County
Lehigh 20.43% 79.57% 930
Lancaster 28.44% 71.56% 851
Northampton 31.11% 68.89% 704
Lebanon 45.04% 54.96% 484
Table 4: Bail Amounts by Bail Type (In $1000s)
Monetary Bail Unsecure Bail Total # of Cases
Mean 32.77 7.70
Standard Deviation 44 13.65
0-999 4.26% .86% 55
1,000-2,499 23.82% 4.48% 301
2,500-4,999 17.28% 5.95% 276
5,000-9,999 28.77% 19.57% 661
10,000-24,999 17.38% 25.14% 679
25,000-49,999 5.29% 20.05% 467
>49,000 3.11% 23.95% 530
Total # of Cases 2,100 869 2,969
48
Table 5: Logistic Model of Bail Type
(Unsecure compared to Monetary) (N=2,967)
Odds Ratio Std. Err
Hispanic 1.82 *** .20
Black 1.39 ** .15
Grade 1.47 *** .04
PRS 1.34 *** .05
Sex .61 *** .07
Age .99 ** .00
drug dealing Charge 2.40 *** .43
Violent Crime Charge 1.46
.36
Property Crime Charge .71 * .08
Lancaster .70 ** .09
Lebanon .40 *** .06
Northampton .74 * .10
Constant .56 ** .11
*p<.05 **p<.01 ***p<.001 †p<.1
Table 6: Logistic Model of Bail Type: Drug Dealing (N=501)
Odds Ratio Std. Err
Hispanic 2.87 ** 1.12
Black 4.10 *** 1.66
Grade 2.58 * 1.06
PRS 1.29 † .17
Sex .63
.28
Age .99
.02
Lancaster .80
.32
Lebanon .43 † .19
Northampton 2.22 1.33
Constant .03
.06
*p<.05 **p<.01 ***p<.001 †p<.1
49
Table 7: Logistic Model of Bail Type: Violent Crime (N=327)
Odds Ratio Std. Err
Hispanic .74
.36
Black 2.10
1.28
Grade 1.93 ** .38
PRS 1.09
.19
Sex .28 ** .13
Age .98
.02
Lancaster .79
.48
Lebanon .79
.0658
Northampton .38 t .22
Constant .28
.47
*p<.05 **p<.01 ***p<.001 †p<.1
Table 8: Logistic Model of Bail Type: Property Crime
(N=1,035)
Odds Ratio Std. Err
Hispanic 1.75 ** .31
Black 1.12
.20
Grade 1.51 *** .07
PRS 1.34 *** .08
Sex .53 *** .09
Age .98 * .01
Lancaster .92
.18
Lebanon .32 *** .08
Northampton .70 † .15
Constant .42 * .15
*p<.05 **p<.01 ***p<.001 †p<.1
50
Table 9: Logistic Model of Bail Type: PRS and Race
Interactions*
Odds Ratio Std. Err
Hispanic 2.29 *** .33
Black 1.42 * .21
Binary PRS 2.79 *** .43
HispanicXPRS .58 * .13
BlackXPRS 1.04 .23
Constant .47 *** .10
*p<.05 **p<.01 ***p<.001 †p<.1
*Control variables in model but not shown: each interaction ran
separately, Race and PRS variables show odds ratios when the
other is not in the model
Table 10: Logistic Model of Bail Type: No Prior Record
(N=1596)
Odds Ratio Std. Err
Hispanic 2.27 *** .33
Black 1.44 * .21
Grade 1.50 *** .06
Sex .55 *** .08
Age .99 † .01
drug dealing Charge 2.31 *** .49
Violent Crime Charge 1.53
.44
Property Crime Charge .75 † .11
Lancaster .63 ** .10
Lebanon .54 ** .10
Northampton .84
.15
Constant .39 *** .10
*p<.05 **p<.01 ***p<.001 †p<.1
51
Table 11: Logistic Model of Bail Type: With Prior Record
(N=1,371)
Odds Ratio Std. Err
Hispanic 1.33
.24
Black 1.49 * .26
Grade 1.45 *** .07
Sex .68 † .14
Age .98 * .01
drug dealing Charge 3.15 ** 1.10
Violent Crime Charge 1.28
.58
Property Crime Charge .65 * .12
Lancaster .75
.17
Lebanon .28 *** .06
Northampton .56 ** .11
Constant 1.77 † .60
*p<.05 **p<.01 ***p<.001 †p<.1
Table 12: Regression Model of Bail Amount: All Bail Types
Beta Std. Err
Hispanic .03 † 1.70
Black .03
1.72
Grade .29 *** .44
PRS .06 ** 45
Sex -.06 *** 1.84
Age .01
.07
drug dealing Charge .19 *** 2.20
Violent Crime Charge .14 *** 2.92
Property Crime Charge -.07 *** 1.79
Lancaster .02
1.89
Lebanon -.05 ** 2.14
Northampton .06 ** 1.88
Constant . * 3.17
*p<.05 **p<.01 ***p<.001 †p<.1
52
Table 13: Regression Model of Bail Amount: Monetary Bail
(N=2,099)
Beta Std. Err
Hispanic .04
2.12
Black .03
2.17
Grade .30 *** .57
PRS .03
.54
Sex -.07 *** 2.50
Age .00
.09
drug dealing Charge .23 *** 2.48
Violent Crime Charge .17 *** 3.32
Property Crime Charge -.08 ** 2.27
Lancaster .04
2.26
Lebanon -.04 † 2.78
Northampton -.004
2.30
Constant . ** 4.29
*p<.05 **p<.01 ***p<.001 †p<.1
Table 14: Regression Model of Bail Amount: Unsecure Bail
(N=868)
Beta Std. Err
Hispanic -.02
1.04
Black -.06 † .99
Grade .33 *** .28
PRS .01
32
Sex -.02
.97
Age .04
.04
drug dealing Charge .24 *** 1.94
Violent Crime Charge .01
2.58
Property Crime Charge -.11 ** 1.01
Lancaster -.07 † 1.22
Lebanon -.15 *** 1.22
Northampton -.03
1.21
Constant .
1.71
*p<.05 **p<.01 ***p<.001 †p<.1
53
Table 15: Correlation Matrix
1 2 3 4 5 6 7 8 9 10 11 12 13
1. Bail Type
2. Bail Amount 0.29 ***
3. Hispanic 0.10 *** 0.05 **
4. Black 0.04 * 0.04 * -0.50 ***
5. Grade 0.35 *** 0.38 *** 0.04 * 0.00
6. PRS 0.16 *** 0.06 *** -0.03†
0.12 *** 0.00
7. Sex -0.15 *** -0.12 *** -0.07 *** -0.01 -0.04 * -0.17 ***
8. Age -0.05 * -0.04 * -0.06 ** 0.01 -0.05 ** 0.27 *** 0.05 *
9. Drug Dealing 0.19 *** 0.24 *** 0.07 *** 0.11 *** 0.14 *** 0.02 -0.09 *** -0.09 ***
10. Violent 0.16 *** 0.27 *** 0.02 0.00 0.47 *** -0.03†
-0.03†
-0.02 -0.13 ***
11. Property -0.05 ** -0.11 *** -0.01 -0.08 *** 0.20 *** -0.01 0.13 *** 0.04 * -0.31 *** -0.21 ***
12. Lancaster 0.01 0.04 * 0.00 0.00 0.16 *** -0.27 *** 0.01 -0.02 0.03 0.03†
0.11 ***
13. Lebanon -0.15 *** -0.12 *** 0.01 0.00 -0.15 *** 0.03†
0.06 ** -0.01 -0.02 -0.04 * -0.07 *** -0.28 ***
14. Northampton-0.02 0.02 0.00 0.00 -0.11 *** 0.00 -0.01 0.00 -0.07 *** -0.01 -0.04 * -0.36 *** -0.25 ***
Correlation Matrix
*p<.05 **p<.01 ***p<.001 †p<.1
54
Table 16: Literature Review of all studies published since 2000
Article Dataset Races
Examined
Dependent
Variables
Findings
Demuth
(2003)
SCPS
1990,
1992,
1994, and
1996
Black,
White,
and
Hispanic
Bail
Amount,
ROR vs Bail,
Detained,
Denied Bail,
Didn’t make
Bail
Detainment, denial of bail and not
making bail more likely for black and
Hispanic defs. Black and white defs
more likely to get ROR than Hispanic
defs. Hispanics had significantly
higher bail amounts to whites but
whites did not differ from blacks.
Demuth &
Steffensmeier
(2004)
SCPS
1990,
1992,
1994, and
1996
Black,
White,
and
Hispanic
Bail
Amount,
ROR vs Bail,
Detained,
Denied Bail,
Didn’t make
Bail
Black and Hispanic defs more likely
to be denied bail and to be detained.
Hispanics more likely to receive bail
not ROR and to have a higher bail
amount in comparison to white defs
but blacks not different than whites.
Freiburger &
Hilinski
(2010)
One urban
Michigan
county in
2006
Black and
White
Detained Significant race differences found but
became insignificant when accounting
for economic variables
Freiburger,
Marcum,, &
Pierce (2010)
One mid-
sized PA
county in
2000-2003
Black and
White
Bail
Amount,
ROR vs Bail,
Detained
No race differences found for bail
amount or detainment outcomes.
Black defs were 80% less likely to
receive ROR than white defs
Schlesinger
(2005)
SCPS
1990,
1992,
1994,
1996,
1998, and
2000
Black,
White,
and
Hispanic
Bail
Amount,
ROR vs Bail,
Detained,
Denied Bail,
Didn’t make
Bail
Black and Latino defs more likely to
be denied bail, and to be detained
pretrial and less likely to get ROR and
to make bail than white defs. Latinos
had a significantly higher bail amount
than whites but blacks did not differ.
Wooldredge
(2012)
One urban
Ohio
county in
2005
Black and
White
Bail
Amount,
ROR vs. Bail
Black defs significantly more likely to
have a higher bail amount and get bail
rather than ROR but the effect went
away when controlling for legal
factors (prior record, felony counts
etc.). Interaction effects of race were
found for black males within the ages
of 18-29 much more significantly
likely to get bail and a higher bond.
55
Table 17: Dependent Variable Coding
Dependent Variable Definition Code
Bail Type The type of financial
or nonfinancial bail
decision a judge
decides on
Released and
Recognizance (ROR)
Defendant was
released from jail with
no bail required
N/A
Unsecured Bail Defendant was given a
bail amount, but only
has to pay it if he/she
misses a future hearing
0
Monetary Bail Defendant was given a
bail amount he/she
must pay in order to be
released from jail
1
Remanded Defendant was not
given bail and is not
allowed to leave jail
N/A
Bail Amount The dollar amount
given for unsecured
and monetary bail
types
0-999 bail amount between 0
and 999 dollars
0
1,000-2,499 bail amount between
1,000 and 2,499
dollars
1,000
2,500-4,999 bail amount between
2,500 and 4,999
dollars
2,500
5,000-9,999 bail amount between
5,000 and 9,999
dollars
5,000
10,000-24,999 bail amount between
10,000 and 24,999
dollars
10,000
25,000-49,999 bail amount between
25,000 and 49,999
dollars
25,000
>49,000 bail amount over
49,000 dollars
(lowest value in this
category is 50,000)
50,000
56
Table 18: Sampling Frame (Percentage of each group sampled)
Percent of
White
Population
Percent of
Black
Population
Percent of
Hispanic
Population
Total
Number
of Cases
Total
Percentage
of
Population
Lancaster 4.68 16.24 41 894 10
Lebanon 4.56 27.49 22.18 517 10
Lehigh 6.57 14.21 12.87 991 10
Northampton 5.5 16.24 17.78 817 10
top related