Examining Racial Disparities in
Criminal Case Outcomes among
Indigent Defendants in San Francisco
SUMMARY REPORT
Emily Owens, PhD
(Univ. of California, Irvine School of Social Ecology)
Erin M. Kerrison, PhD
(Univ. of California, Berkeley School of Social Welfare)
Bernardo Santos Da Silveira, PhD
(Wash. Univ. in St. Louis Olin Business School)
May 2017
!
!
Contents
Introduction ................................................................................................................................... 1
Study Overview .............................................................................................................................. 1
Summary of Key Findings .............................................................................................................. 2
Overall Case Outcomes by Race .................................................................................................... 3
Factors Contributing to Racial Disparities in Criminal Case Outcomes ....................................... 4
Criminal History, Poverty, and Police Activity by Neighborhood ............................................. 5
Pre- and Post-filing Case Decisions. ......................................................................................... 6
Case Adjudication: How Charges Evolve and Are Bargained ................................................... 9
Case Adjudication: Convictions and Sentences ....................................................................... 16
Length of Time to Case Resolution. ........................................................................................ 20
Sentencing/Length of Incarceration ........................................................................................ 22
Public Defender Resource Constraints .................................................................................... 24
Conclusions and Questions for Policy Makers ............................................................................ 25!
!
1
!
Introduction
People of color are overrepresented in California correctional facilities. According to a recent report from the
Public Policy Institute of California, approximately 4.4% of the Black male population of California is incarcerated
in a California prison.
1
Black men in California are incarcerated at 100 times the rate of Asian men, ten times the
rate of White men, and five times the rate of Latino men. It is important for criminal justice practitioners,
policymakers, and scholars to understand these disparities and their causes. Potential explanations include
variations in socioeconomic status, access to employment and education opportunities, patterns in policing, and
differences in charging and sentencing decisions made by prosecutors and judges.
Most studies of racial disparities in the justice system have focused on final case outcomes, such as conviction,
incarceration, and sentence length. While important, these data points do not provide sufficient insight into the
many points in the criminal justice process where cases against Black, White, and Latinx defendants could diverge.
To fill this knowledge gap, the Quattrone Center for the Fair Administration of Justice (“Quattrone Center”), in
collaboration with the San Francisco Public Defender (“Public Defender”), reviewed the charging and case
adjudication process for Public Defender clients in San Francisco, so that differences in the processing of and final
outcomes for Black, White, and Latinx defendants could be seen, and to explain the source of any differences that
exist.
Study Overview
We reviewed 10,753 complete case records, consisting of cases between 2011 and 2014, from the San Francisco
Public Defender’s Office. These data were stored in the Public Defender’s GIDEON case management system,
which draws from data maintained by the San Francisco County Superior Court’s larger case management system
database. Unlike previous studies that rely solely on arrest and conviction data, these records cover the entire
pretrial process, providing a richer portrait of the experiences of defendants in the criminal justice system.
These data can help policymakers and stakeholders understand whether racial disparities exist in the outcomes of
San Francisco criminal cases, including cases resolved by plea bargains, and how bargaining affects disparities in
other areas of the criminal justice system, such as corrections.
2
Where disparities were seen, we sought to
understand them and to evaluate what changes could be made to ensure that similarly situated individuals receive
equal and race-neutral treatment in the criminal justice system. Such information could assist the Public Defender,
the San Francisco District Attorney, the San Francisco Police Department, and other criminal justice stakeholders
to ensure equitable treatment of all San Franciscans.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
1
Grattet, R. and Hayes, J., “Just the Facts: California’s Changing Prison Population,” April 2015, accessed May 1, 2017 at
http://www.ppic.org/main/publication_show.asp?i=702.
2
Just under 59% of these cases resulted in a conviction, and the clear majority of all convictions - 91% - involved at least one guilty plea.
2
!
Summary of Key Findings
Our analysis revealed that Black, White and Latinx indigent defendants in San Francisco have substantially
different experiences during the criminal adjudication process. However, disparities by race/ethnicity could largely
be explained by factors determined prior to the initiation of plea negotiations. In particular:
1. The raw data reveal Black/White and Latinx/White disparities across several metrics related to case
processing and outcomes.
a. Black defendants are held in pretrial custody longer than Whites. Black defendants are held in pretrial
custody for an average of 30 days, 62% longer than Whites.
b. Cases involving Black defendants take longer to resolve. It takes an average of 90 days to process a
case for a Black defendant, but only 77.5 days to process a case for White defendants, a delay of 14%.
c. Defendants of color are convicted of more serious crimes than White defendants. Black defendants
are convicted of 60% more felony charges than White defendants, and 10% fewer misdemeanors.
Latinx defendants are convicted of a similar number of felonies to Whites, but 10% more
misdemeanors.
d. Defendants of color receive longer sentences than White defendants. Custodial sentences received
by Black defendants are, on average, 28% longer than those received by White defendants. While
Latinx defendants receive comparable custodial sentences to White defendants, they receive probation
sentences that are 55% longer than those received by White defendants.
2. Even though these disparities are occurring within the plea bargaining system, plea bargaining itself
appears to neither contribute to the disparate outcomes, nor to reduce the disparities. Bargaining
decisions by public defenders and prosecutors did not appear to increase the disparities that were inherited
from the arrest process. There was no disparity seen in either the number of charges added by the DA’s
Office to the booking charges, or the proportion of charges to which individuals plead guilty (across charge
type and severity). At the same time, the more severe initial bookings tended to follow Black defendants
through the process, resulting in a higher rate of felony convictions and longer sentences on average.
3. The majority of these disparities seem to be generated by two factors that pre-date the case adjudication
process:
a. People of color receive more serious charges at the initial booking stage, reflecting decisions made by
officers of the San Francisco Police Department; and
3
!
b. People of color have pre-existing racial differences reflected in their criminal record, based on
previous encounters with the criminal justice system in San Francisco County. This criminal history
has a “ripple effect” that impacts plea negotiations for subsequent charges, as police, prosecutors, and
defense attorneys make plea bargain decisions based in part of the individual’s prior criminal history.
Overall Case Outcomes by Race
Black, White and Latinx indigent defendants in San Francisco experience the criminal adjudication process
differently, as shown in Table 1 below.
Table 1. Average Case Outcomes by Race
Notes: * indicates statistically significant difference from White, p < .05
White
Black
Latinx
Booking
% diff.
% diff.
Number of Booked Charges
2.57
2.75*
7%
2.58
0%
Felonies
1.46
1.81*
24%
1.30*
-11%
Misdemeanors
0.96
0.80*
-17%
1.12*
17%
Prosecutor Activity
Number of Added Charges
0.95
0.91
-4%
1.01
6%
Felonies
0.34
0.43*
26%
0.32
-6%
Misdemeanors
0.57
0.46*
-19%
0.63
11%
Case Adjudication
Guilty of any charge
56.7%
60.0%*
6%
59.2%
4%
Number of Convicted Charges
0.695
0.739*
6%
0.721
4%
Felonies
0.186
0.299*
61%
0.178
-4%
Misdemeanors
0.514
0.451*
-12%
0.557*
8%
Sentence Length (in days, if
convicted)
89.3
189.7*
112%
104.5
17%
Method of Resolution
Plead guilty of any charge
53.5%
54.7%
2%
54.2%
1%
Number of Plead Charges
0.647
0.665
3%
0.637
-2%
Case Processing
Days from First to Last Court Event
77.5
90.3*
17%
80.9
4%
Days in Pretrial Custody
18.8
30.4*
62%
20.5
9%
Sample Size
3,831
4,749
2,173
!
4
!
Table 1 reports average outcomes for defendants of different races. These are simple comparisons that do not
account for contributing factors other than race that may explain the observed overall disparities (e.g., criminal
history). In general, White defendants fared better than minorities, although for some important outcomes the
differences between Blacks, Latinx, and Whites were not statistically significant.
Factors Contributing to Racial Disparities in Criminal Case
Outcomes
We have taken two approaches to highlighting racial disparities in San Francisco’s criminal justice system. The
first is to show “raw” or unadjusted overall differences in case outcomes across defendants of different races, as in
Table 1 above. Such comparisons are useful, but can be oversimplified and misleading, as they may not show
legally or socially relevant factors that differ across cases involving defendants of different races. Failing to account
for such differences could lead to an inaccurate view of the role of race in the criminal justice system.
Criminal history is an excellent example. In most jurisdictions, the sentencing scheme is structured to increase the
penalty for criminal conduct if the defendant has prior criminal convictions. In such a system, observations that
one racial group tends to receive longer sentences could be the result of biased treatment, but they could also simply
reflect that the group receiving the longer sentences has more prior convictions, leading to the assignment of
longer sentences.
To properly measure racial disparity, then, one would ideally take two pools of otherwise similar defendants that
differ only in race, and compare outcomes across such groups. Such an ideal comparison is not possible here,
because no two cases are exactly the same. However, we can statistically adjust for a range of legally relevant
contextual factors that might vary across defendants drawn from different racial backgrounds, in an effort to isolate
race from other factors. Disparities that remain after accounting for other legally relevant race-neutral factors
deserve further investigation.
Accordingly, we performed a statistical analysis of the data that accounts for factors other than race that might
explain disparities, and analyzed which characteristics are most important for explaining the existence of racial
disparities.
3
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
3
To examine this, we conducted a decomposition analysis, which calculated the portion of the unadjusted disparity that is explained by
the various contextual factors considered in the analysis. For example, if the results indicated that the unadjusted Black/White disparity in added
felonies was 20% - meaning Black defendants on average had 20% more felonies added to their case by prosecutors than White defendants - and
50% of this disparity can be explained by criminal history, then for Black and White defendants with identical criminal histories (rather than the
more extensive criminal histories among Black defendants that is actually the case in these data), we would expect Black defendants to have only
10% more added felonies than White defendants.
These contextual factors are more likely to be identified in the decomposition analysis as substantial contributors to disparity if they vary
appreciably across minority and White defendants and if, other things being equal, they tend to be more predictive of the outcome in question. It
is also possible with such an analysis for a portion of the raw disparity to remain unexplained, meaning that contextual factors outside of those
considered in the analysis may be driving the observed disparity.
5
!
!
Criminal History, Poverty, and Police Activity by Neighborhood.!Table 2 illustrates the variance across White,
Black, and Latinx defendants of several important factors that could contribute to or help explain the racial
disparities set forth in Table 1 above. In addition to criminal history, whose importance is explained above, we
examined the role of geography, in terms of socioeconomic levels in different neighborhoods that might lead to
different types or levels of criminal behavior, as well as disparities that occur due to decisions made by police
officers in so-called “high crime” versus “low crime” neighborhoods. To understand this, we examined court
records that identified the exact location of each arrest, as well as the defendant’s home address.
Several differences are worth noting:
1. The likelihood that an individual defendant has had previous contact with the criminal justice system
is greater for Black than for White defendants, and greater in turn for White than for Latinx
defendants. Blacks averaged almost twice the number of prior arrests and twice the number of prior
convictions than whites.
2. Poverty rates in the defendant’s neighborhood of residency were higher for Blacks (15%) than for
Latinx (11.5%) or whites (9%).
3. Police activity in the neighborhood of residence (which combines both crime rates and police
presence) and arrest rates were higher for Blacks than for Whites and higher for Whites than for
Latinx.
Table 2. Group Differences in Contextual Factors
Notes: * indicates statistically significant difference from White, p < .05.
Incident and arrest rates are measured per 1000 residents.
White
Black
Latinx
Defendant Characteristics
% diff.
% diff.
Transient
29.5%
18.8%*
-36%
14.0%*
-53%
Female
15.9%
19.0%*
19%
16.4%*
3%
Age at Arrest
36.27
36.86*
2%
33.51*
-8%
# Previous Arrests
7.85
13.08*
67%
4.88*
-38%
# Previous Convictions
1.59
2.97*
87%
1.13*
-29%
Neighborhood of Residence
% Adults w/ Limited English
3.5%
3.9%*
11%
5.4%*
54%
% Adults w/ Some College
69.1%
60.2%*
-13%
61.9%*
-10%
% Families in Poverty
8.9%
15.3%*
72%
11.5%*
29%
Police Incident Rate
7,391
9,738
32%
5,749
-22%
Warranted Arrest Rate
383
506
32%
299
-22%
Gang-Related Incident Rate
179
234
31%
145
-19%
6
!
Neighborhood of Arrest
Same as Home
13.4%
12.9%*
-4%
14.6%*
9%
% Black
7.0%
12.3%*
76%
8.1%*
16%
% Hispanic
15.9%
18.2%*
14%
22.5%*
42%
% of Housing Units Not Owner-
Occupied
74.6%
75.3%*
1%
70.9%*
-5%
Police Incident Rate (per 1,000
pop.)
82,176
111,466*
36%
58,503*
-29%
Sample Size
3,831
4,749
2,173
!
!
Pre- and Post-filing Case Decisions.!We also examined pre- and post-filing phases of the case adjudication
process to understand their impact on the overall disparities shown in Table 1 above. We examined many
interactions during the case adjudication process where similarly situated defendants could receive different
treatment from the criminal justice system. Specifically, we analyzed the decisions of booking officers, prosecutors,
public defenders, judges, and probation officers during pre- and post-filing phases
“Pre-filing outcomes” are decisions made by booking officers and prosecutors, often before a client is assigned to the
Public Defender’s Office. These initial decisions on what to charge establish the foundation of the criminal
proceedings going forward and influence the defendant’s bargaining position during the adjudication phase. Pre-
filing outcomes include:
• The total number of charges for which one is booked into a San Francisco jail;
• The number of felony and/or misdemeanor charges for which one is initially booked;
• The total severity
4
of the charges for which one is booked, including:
o “Top” charge (i.e., most serious offense, as defined by the District Attorney’s severity scale);
o Total number of charges;
o Total severity of all charges; and
• The number, type, and severity of charges that are added to the initial booking by the District
Attorney’s Office.
“Post-filing outcomes” include determinations of guilt or innocence for whatever number of charges has been
brought. They reflect the ability of defendants, and/or the willingness of prosecutors, to modify the initial charges
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
4
This severity score is based on the California Attorney General’s ranking of criminal charges, which can be found here:
https://oag.ca.gov/law/code-tables
7
!
based on individual defendant characteristics or circumstances.
Figures 3 and 4 below display Black/White and Latinx/White disparities across four pre-filing outcome case
measures: total booked charges, booked felonies, booked misdemeanors, and case severity. The “case severity”
measure combines all booked charges into a single summary measure that considers both the number and
seriousness of booked charges. For example, being booked for robbery is more serious than being booked for
loitering, and being booked for three similarly serious counts is worse than being booked for one.
The blue bars in the chart show the raw, or unadjusted disparity, while the orange bars show the measured disparity
after statistically controlling for the contextual factors noted above using regression analysis. In other words, the
orange bars show the expected difference in pre-filing outcomes for a Black or Latinx arrestee as compared to an individual
who is similar in age, gender, residential and arrest neighborhood characteristics, and prior criminal history – but is White.
Figure 3 shows that Blacks in our dataset are booked for 7% more crimes than Whites on a raw or unadjusted
basis, while they are booked for 8% more crimes than Whites with similar age, gender, criminal history, and other
characteristics.
Figure 3: Black/White Disparities in Pre-filing Outcomes
Note: * denotes a statistically significant Black/White difference
!
!
The blue bars show the raw, or unadjusted disparities between Blacks and Whites: Black defendants were booked
on average for more charges overall than White defendants, including more felonies. They were booked for fewer
misdemeanors than White defendants (suggesting greater severity in charging on average, even controlling for
contextual factors). Black arrestees faced initial cases that were about 50% more severe than White arrestees in
terms of number and severity of charges.
5
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
5
Total severity on the California Attorney General’s scale, the severity scale used in this analysis, roughly correlates to the length of a typical
sentence. Thus, a 50% increase in total severity score can be thought of as roughly equivalent to a 50% increase in length of a typical sentence.
7%*
24%*
-17%*
48%*
8%*
22%*
-11%!*
46%*
-30% -10% 10% 30% 50% 70%
Total! booked! charges
Total! booked! felonies
Total! booked! misdemeanors
Severity!of!booked!offenses
Raw!disparity Disparity!after!controlling!for!contextual!factors
8
!
The orange bars show that for Black arrestees, controlling for contextual factors does little to diminish the observed
disparity. The Black/White differences in booked charges cannot be explained by factors such as age,
homelessness or poverty, or crime rates in the neighborhoods in which Black citizens reside or routinely
encounter police, though there may be unobserved, legally relevant factors other than bias (e.g., actual criminal
conduct, or how particular individuals interact with officers) that are unaccounted for in the analysis and explain
the observed disparities.
The situation for Latinx defendants in San Francisco is somewhat different (Figure 4). While the difference in
total booked charges between Latinx and White defendants was not statistically significant, the makeup of the
charges placed on Latinx defendants was unique. Latinx defendants were booked on fewer felony charges, and
more misdemeanor charges, than White defendants with the same background characteristics. After accounting
for contextual factors, however, Latinx arrestees faced pre-filing charges that were roughly similar in severity to
comparably situated White arrestees.
Figure 4: Latinx/White Disparities in Pre-filing Outcomes
Note: * denotes a statistically significant Black/White difference
!
0%
-11%*
16%*
18%*
-5%
-18%*
13%*
2%
-30% -10% 10% 30% 50% 70%
Total! booked! charges
Total! booked! felonies
Total! booked! misdemeanors
Severity!of!booked!offenses
Raw!disparity Disparity!after!controlling!for!contextual!factors
9
!
Case Adjudication: How Charges Evolve and Are Bargained
The detail available in the Public Defender’s case files enabled us to examine how prosecutors and defense
attorneys actually bargain to reach final case outcomes. First, we looked at plea bargaining in a traditional sense –
whether defendants pled guilty to any charges, and the number of charges to which they pled guilty (or nolo
contendere). The rate at which Black, Latinx, and White defendants pled guilty to any charge was similar, and
we observed no statistically significant differences in the number of charges discharged or dismissed among Black
and Latinx defendants.
Figures 5 and 6 depict disparities between Blacks and Whites, and between Latinx and Whites, respectively, in
the application of prosecutors’ charging discretion. Put differently, they depict the decision of prosecutors to
modify the original charges booked by the police, based on the prosecutor’s review of the case record and what
charges are possible based on the facts alleged. We looked at the probability that a felony would be downgraded
to a misdemeanor, the probability that a misdemeanor would be refiled as a felony, and the number of times the
District Attorney’s office refiled a charge in court documents for any reason.
Felony charges filed against White defendants were more likely to be downgraded (31%) than felony charges filed
against Black (23%) and Latinx (29%) defendants. However, these differences across groups were not statistically
significant after adjusting for contextual factors. Most of the Black/White disparity can be explained by
combining the variation in the criminal history of Black defendants (explaining 26% of the disparity) and the
charges for which they were booked (explaining 48%). The disparity in outcomes for Latinx and White
defendants also appears to be driven largely by booking charges (explaining 70% of the disparity).
Latinx defendants were much less likely to have their misdemeanors upgraded to felony convictions, doing so at
only 2.3 percent of the rate that misdemeanors for White defendants were upgraded to felonies for White
defendants. On the other hand, since felony convictions for Latinx defendants are more likely to raise immigration
or citizenship-related concerns than those confronted by White and Black defendants in San Francisco, it is a
potentially important source of inequality in the justice system. Very little of this difference can be explained using
the study’s control variables; even the variation in booked charges can explain only 21% of the Latinx-White gap.
Again, the blue bars depict the raw or unadjusted disparities shown above in Table 1, while the orange bars depict
disparities that persist after adjusting for contextual factors. For these comparisons, in addition to accounting for
the demographic and neighborhood characteristics mentioned previously, the adjusted comparisons also account
for racial differences that occurred at the booking stage. Thus, the figures compare added charges for two
defendants with similar demographics, criminal histories, etc. and booking charges who differ only in race.
10
!
Figure 5: Black/White Disparities in Prosecutor Charging
Note: * denotes a statistically significant Black/White difference.
!"#
$%#&
!$'#&
$(#&
!$)#&
$$#
!*#
!+#
!*#
!(,#
!%#
!$(#
!,'# !$'# !('# '# ('# $'# ,'#
-../.012345/6
-../.07/89:;/6
-../.0<;6./</3 :946
-../.06 /=/4;>?
@/89:;/60>90<;6./</3:946
A;6./</3:9460>907/89:;/6
B3C0.;6D34;>? E;6D34;>?037>/4019: >4988;:50794019:>/F>G380731>946
!
!
While the raw or unadjusted data shows a disparity in the number and severity of felonies charged against Blacks
versus whites, when we adjust for the various contextual factors, we see no statistically significant differences in
the number or severity of charges added by prosecutors for either Black or Latinx as compared to Whites. This
suggests that the discretion of the booking (police) officer is more impactful than that of the district attorney in
terms of the disparities in the number and seriousness of charges filed. In fact, we found no evidence that district
attorneys file more or fewer charges against Black or Latinx defendants than they file against Whites. While it
does appear that charges added by the DA against Black defendants were more likely to be felonies and less likely
to be misdemeanors; these differences disappeared after accounting for contextual factors (including booking
charges), suggesting that race was not a contributing factor to the decision. Similarly, DAs may have added more
misdemeanors and more severe charges to Latinx defendants after booking, but these differences are not
statistically significant. For both groups, once the differences in criminal background (including type of charges
booked for) were accounted for, the overall disparity was explained.
11
!
Figure 6: Latinx/White Disparities in Prosecutor Charging
Note: * denotes a statistically significant Latinx/White difference
!"
#$"
%%"
&%"
#'"
#($")
%"
#*"
+"
&"
,"
#(*")
#+," #*," #(," #&," #%," ," %," &," (,"
-../.012345/6
-../.07/89:;/6
-../.0<;6./</3:946
-../.06/=/4;>?
@/89:;/60>90A;6./</3:946
A;6./</3:9460 >90@/89:;/6
B3C0.;6D34;>? E;6D34;>?037>/4019:>4988;:50794019:>/F>G380731>946
!
!
!
The additional felonies that are added by the District Attorney’s Office to the cases of Black defendants can be
explained by differences in police booking decisions. There appear to be certain booked charges made by the police
that are more likely to cause an Assistant District Attorney to add further charges. One hypothetical example of
this could be that an aggravated assault in which a gun was displayed might be more likely to have an illegal gun
possession charge added by the DA.
!
!
!
12
!
Figure 7. What Affects the Black/White Disparity in Charges Added by Prosecutor?
!
!
Recall that in Table 1, we showed that prosecutors add 26% more felonies to cases with Black defendants than to
cases with White defendants, and they add 23% fewer misdemeanors to cases with Black defendants. Figures 7
and 8 report which of various contextual factors best explain these charge disparities, with blue bars showing added
felonies and orange bars showing misdemeanors. A value above 0% shows that the contextual factor reduces the
minority/White disparity, while a negative value shows an increased disparity.
0%
4%
0%
-3%
18%
10%
1%
130%
1%
2%
3%
1%
1%
-1%
14%
26%
77%
2%
-20% 0% 20% 40% 60% 80% 100%120%140%
Individual!Demographics
Day,!Month,!and!Week!of!Arrest
Police!Activity!at!Home
Police!Activity!where! Arrested
Charac teristcs!of!Home
Chacteristics!of!Arrest!Location
Criminal!Record
Booked!charges
Attorney!load
Added!Misdemeanors Added!Felonies
13
!
Figure 8. What Affects the Latinx/White Disparity in Charges Added by Prosecutor?
!"
#"
$"
%&"
%'"
'("
%("
)&"
$"
'*"
%''"
*"
%+"
%'*"
%!&"
*$"
+("
%*"
%*$" %&$" $" &$" *$" ($" ,$"
-./010/23456789:;3<=0>?
[email protected]=A53./5D77E59F5G;;7?C
H940> 75G>C010C@53C5I987
H940>75G>C010C@5J=7;75G;;7?C7/
K=3;3>C7;0?C>?59F5I987
K=3>C7;0?C0>? 59F5G;;7?C5L9>3C09.
K;080.345M7>9;/
N99E7/5>=3;:7?
GCC9;.7@5493/
G//7/580?/7873.9;? G//7/5 F749.07?
!
!
In this decomposition analysis, booking decisions accounted for 130% of the observed raw Black/White disparity
in added felonies, more than enough to explain the entire discrepancy.
6
Both criminal history and booking charges
play a role in explaining raw differences in added charge severity, with criminal history accounting for 26% of the
Black/White disparity and 40% of the Latinx/White disparity, and booking charges accounting for 18% of the
Black/White disparity and 39% of the Latinx/White disparity. However, for both groups, a substantial fraction
of the disparity in added charge severity remains unexplained.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
6
In other words, if Blacks were booked for the same crimes as Whites, and all other factors were equal, the Blacks would have fewer additional
felonies added by the prosecutor than Whites by a factor of 30%.
14
!
We examined the evolution of all charges against an individual over the course of the adjudication process (i.e.,
from initial booking through conviction), including:
• The seriousness of the charges for which the client was convicted;
• The seriousness of the charges that were dismissed or discharged;
• The number of charges downgraded from felonies to misdemeanors (or upgraded from misdemeanors to
felonies) during the adjudication of the case; and
• The number of charges dismissed in exchange for a guilty plea to another charge.
Figure 9 illustrates the factors that affect the difference between the charges that exist at the outset of the case,
and the charges that ultimately exist at conviction.
Figure 9. What Affects Black/White Differences in Charge Evolution?
!"#
!$#
!$#
%#
&#
$$#
'(#
&)(#
*#
!"#
+#
)#
)#
"#
!"#
$*#
",#
&#
!$)# )# $)# ")# *)# ,)# &))# &$)#
-./010/23456789:;3<=0>?
63@A5B9.C=A53./5D77E59F5G;;7?C
H940>75G>C010C@53C5I987
H940>75G>C010C@5J=7;75G;;7?C7/
K=3;3>C7;0?C>?59F5I987
K=3>C7;0?C0>?59F5G;;7?C5L9 >3C09.
K;080.345M7>9;/
N99E7/5>=3;:7?
GCC9;.7@5493/
B0?/7873.9;?5C95F749.07? O749.07?5C9580?/7873.9;?
!
!
Here again, when we control for the disparity between blacks and whites in booked charges at the time of arrest,
we see that the disparity among charges at the time of booking is substantial enough to remove the raw disparities
completely for misdemeanors, and to remove roughly half (48%) of the disparity in charge evolution.
15
!
In contrast to the situation with Blacks, however, it appears that the evolution of charges in cases involving Latinx
defendants may act in the defendants’ favor. For Latinx defendants, the contributing factors are similar but
differently weighted, as seen in Figure 10. Booking charges continue to be the largest factor explaining the
disparities between Latinx and White defendants in the evolution of charges. Controlling for booking charges
accounts for 70% of the disparity between Latinx and whites in terms of their booked misdemeanor charges, and
22% of the disparity in the evolution of felony charges during the adjudication period. Surprisingly, though, we
see that the defendants’ criminal history adds to the disparity in misdemeanors by 24%. Remember that a negative
result in this chart means that the Latinx defendants, whose charges are more likely to be misdemeanors, are
increasingly evolving from felony charges to misdemeanor charges as their cases evolve. Thus, it appears that police
and prosecutors are more likely to agree to a misdemeanor charge for Latinx than whites.
Figure 10. What Affects Latinx/White Differences in Charge Evolution?
!"#
!$#
%#
!&#
$#
!'(#
'%#
))#
!)#
'$#
'"#
'#
')#
'$#
!'&#
!)*#
&(#
'%#
!*(# !)(# (# )(# *(# "(# +(#
,-./0/.123456789:2;</=>
52?@4A8-B<@42-.4C66D48E4F::6> B
G83/=64F=B/0/B?42B4H876
G83/=64F=B/0/B?4I<6:64F::6>B6.
J<2:2=B6:/>B=>48E4H876
J<2=B6:/>B/=>48E4F::6>B4K8 =2B/8-
J:/7/-234L6=8:.
M88D6.4=<2:96>
FBB8:-6?4382.
A/>.6762-8:>4B84N638-/6> N638-/6>4B84A/>.6762-8:>
!
16
!
Case Adjudication: Convictions and Sentences
Because criminal cases in San Francisco are primarily resolved by plea bargain rather than bench or jury trials, the
study also examined the number of charges to which defendants pled guilty (or nolo contendere). Previous studies
have simply compared cases where there is, or is not, a plea bargain;
7
this focus ignores the substantial variation in
how many and which types of plea deals are made.
8
Our research tracked each individual client of the San
Francisco Public Defender from initial booking through case disposition, and accounted for each defendant’s local
criminal history, enabling the researchers to consider several pieces of information available to prosecutors,
defenders, and judges when they make their decisions. As a result, we can more precisely identify disparities that
might arise from the menu of charges for which someone is booked, and their full criminal history in San Francisco
County.
Figure 11. Black/White Disparities in Case Adjudication
Note: * denotes a statistically significant Black/White difference.
!
!
!
!
!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
7
See, e.g., e.g. Bushway, S. D., Redlich, A. D. and Norris, R. J. (2014), An Explicit Test of Plea Bargaining in the “Shadow of the Trial”.
Criminology, 52: 723–754
8
For example, the Bureau of Justice Statistic’s State Court Processing Statistics only includes information on the most serious charge filed.
!
6%*
6%*
60%*
-12%*
49%*
28%*
40%*
-5%
-29%*
3%
2%
12%
-1%
7%
-1%
-5%
13%
8%
-30% -10% 10% 30% 50% 70%
Conviction!rate
Number!of!convicted!charges
Convicted!felony!charges
Convicted!misdemeanor!charges
Convicted!charge!severity
Sentence!(days),!all!defendants
Sentence!(days)!convicted!only
Probation!(days),!all!defendants
Probation!(days),!convicted!only
Raw!disparity Disparity!after!controlling!for!contextual!factors
17
!
Figure 12: Latinx/White Disparities in Case Adjudication
Note: * denotes a statistically significant Latinx/White difference.
!
!
!
Tables 13 and 14 evaluate racial disparities in convictions. In general, Black defendants are convicted of more
charges than White defendants. For Black defendants, prior contact with the criminal justice system has a ripple
effect that is seen in the severity of punishment for future contact. More specifically, differences in the number of
times that Black defendants were previously arrested, convicted, and incarcerated explain almost all of the
difference in conviction rates among Black and White defendants.
The fact that booking charges have such a substantial impact (see Figure 14 below) suggests that Latinx defendants
are being booked for charges for which a conviction tends to be more certain (e.g., littering, which requires a
simple observation, vs. assault with intent to injury, which requires a proof of the defendant’s state of mind).
Differences in education, employment, and facility with the English language also explain a small amount of the
disparity in conviction rates for Black and Latinx defendants, compared to White defendants. When the study
looked at how many different charges people are convicted of, booking charges appeared to drive convictions for
Latinx defendants, as distinguished from Black defendants, where the driver appears to be previous convictions.
4%
4%
-4%
8%*
27%*
-15%*
-27%*
55%*
68%*
0%
0%
2%
1%
9%
-4%
-12%
7%
23%*
-30% -10% 10% 30% 50% 70%
Conviction!rate
Number!of!convicted!charges
Convicted!felony!charges
Convicted!misdemeanor!charges
Convicted!charge!severity
Sentence!(days),!all!defendants
Sentence!(days)!convicted!only
Probation!(days),!all!defendants
Probation!(days),!convicted!only
Raw!disparity Disparity!after!controlling!for!contextual!factors
18
!
The unadjusted comparisons reveal that Black defendants were convicted of more felonies and fewer
misdemeanors than White defendants, and were convicted of more serious charges overall than White defendants.
Latinx defendants were convicted of more misdemeanors, and more serious charges overall, than White
defendants. All of these disparities can be explained by differences in demographics, criminal history, booking
decisions, and public defender caseloads.
Figure 13. What Affects Black/White Differences in Convictions?
!"#
$#
%#
!&#
'#
!&#
((#
)"#
!&#
&$#
(#
%#
!(#
*#
(#
&#
*$#
!*#
!$%#!&%# %# &%# $%# (%# )%# +%# "%# *%# '%#
,-./0/.123456789:2;</=>
52?@4A8-B<@42-.4C66D48E4F::6>B
G83/= 64F=B/0/B?42B4H876
G83/=64F=B/0/B?4I<6:64F::6>B6.
J<2:2=B6:/>B=>48E4H876
J<2=B6:/>B/=> 48E4F::6>B4K8=2B/8-
J:/7/-234L6=8:.
M88D6.4=<2:96>
FBB8:-6?4382.
J8-0/=B6.4A/>.6762-8:> J8-0/=B6.4N638-/6>
!
!
!
!
!
19
!
Figure 14. What Affects Latinx/White Differences in Convictions?
!"#$
!%#$
!%$
!#$
!%&$
!'$
!%#$
%(#$
($
!%$
!('$
($
!%)$
!)&$
&#$
%&&$
%"%$
($
!%**$ !(*$ *$ (*$ %**$ %(*$ "**$
+,-./.-01234567891:;.<=
41>?3@7,A;?31,-3B55C37D3E995=A
F72.< 53E<A./.A>31A3G765
F72.<53E<A./.A>3H;5953E995=A5-
I;191<A59.=A<=37D3G765
I;1<A59.=A.<= 37D3E995=A3J7<1A.7,
I9.6.,123K5<79-
L77C5-3<;1985=
EAA79,5>3271-
I7,/.<A5-3@.=-5651,79= I7,/.<A5-3M527,.5=
!
!
Decisions made at booking explain almost half (46%) of the Black-White disparity in the number of felony
convictions that Black defendants faced. Criminal history also plays an important role, explaining a third of the
disparity. Thus, roughly 20% of the increased number of felony convictions against Blacks remains unexplained
or is explained by other factors.
Differences in booking charges are also the primary explanation for why Black defendants were convicted of fewer
misdemeanors, and why Latinx defendants were convicted of more misdemeanors. To put these differences in
perspective, note that, on average, White defendants in our data set were convicted of 0.19 felony charges on
average, while Black defendants were convicted of 0.30 felony charges, a roughly 60% increase. Based on these
estimates, if White defendants were booked for the same offenses as similarly situated Black defendants, shared
their criminal history, and otherwise were identical on average to Black defendants in contextual factors other than
20
!
race, White defendants would on average be convicted of 0.28, rather than 0.19, felonies, reducing the disparity
with Blacks to 7%. Latinx defendants were convicted of 0.56 misdemeanors, which is 0.01 more misdemeanors
than would be expected among White defendants with the same criminal records, booking charges, and other
contextual factors as the Latinx defendants (other than ethnicity). Thus, while the unadjusted differences across
racial groups are large, once pre-adjudication contextual factors are adjusted for, the racial gaps become smaller
and in most cases no longer statistically significant.!
!
Length of Time to Case Resolution.!How cases are processed, and in particular whether defendants are
released on bail, has a direct influence on outcomes. Longer cases can benefit defendants, as evidence and witness
cooperation deteriorate over time, making it harder for the state to prove their case. If clients are in custody,
however, there is a direct cost to this extra time, particularly for indigent defendants charged with low-level crimes.
In addition to the physical and emotional toll of incarceration, many defendants operate with little or no economic
safety net, and even brief periods of incarceration can have widespread collateral consequences including loss of
employment, loss of housing, loss of custody and/or child support, and loss of other public benefits. In some
instances, even the time burden of appearing at court to handle their cases may disrupt work or other obligations
for indigent individuals not in custody, causing them to plead guilty to charges simply to have them resolved and
in the past.
We evaluated the time taken to process defendants of different races in the San Francisco County criminal justice
system, including:
• Days passed between arrest and adjudication;
• Days a client was in custody;
• Number of times charges were refiled; and
• Court events
9
that took place.
White, Black and Latinx defendants respectively spent 19, 30, and 21 calendar days detained over the course of
their case. That means Black defendants were in custody for 11.6 additional days relative to White defendants,
which is statistically and substantively significant (Table 1). This disparity falls by 7 days to 4 days after adjusting
for contextual factors, but those remaining four days are still statistically meaningful (Figure 15). Black/White
disparities in days in custody may be explained in large part by criminal record (accounting for 25% of the disparity)
and booking charges (accounting for 42% of the disparity).
These data suggest that the main driver of the increased length of time to resolution of cases involving Black
defendants is their (on average) more extensive criminal history.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
9
A “court event” as used in this paper means a hearing or other procedure that caused the defendant or the defendant’s counsel to appear
in court.
21
!
Figure 15: Black/White Disparities in Case Processing
Note: * denotes a statistically significant Black/White difference.
!
16%
62%
11%
7%
21%
3%
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Days/to/Resolution
Days/in/pretrial/custody
Number/of/hearings
Disparity/ after/controlling/for/contextual/factors Raw/disparity
!
Figure 16: Latinx/White Disparities in Case Processing
Note: * denotes a statistically significant Latinx/White difference.
4%
9%
-1%
3%
15%
-4%
-5% 0% 5% 10% 15% 20%
Days.to.Resolution
Days.in. pretrial.custody
Number.of.hearings
Disparity.after.con trolling.for. contextual.facto rs Raw.disparity
!
!
!
22
!
An additional measurement that reflects the complexity of the case is the number of court events associated with
that case. Black defendants had a statistically significant 1.7 additional court events relative to White defendants
(Table 1). As was the case for pretrial custody days, this disparity appears to be driven by criminal history
(explaining 24% of the disparity) and charges filed at booking (explaining 45% of the disparity).
For Latinx defendants, there were no statistically significant differences in days to resolution or custody days
relative to White defendants. Latinx defendants did have roughly 10% fewer hearings than White defendants, a
statistically significant difference. The measured gap in hearings remains virtually unchanged after accounting for
the contextual factors in the model, so this disparity remains largely unexplained. One speculated possibility is
that the need to accommodate the language needs of some Latinx defendants led to different patterns of
scheduling of hearings.
!
Sentencing/Length of Incarceration.!!For those who were convicted, sentence length (in days) was measured.
Without adjusting for contextual factors (but limiting the influence of outlier sentences), Across all defendants
(i.e., those convicted of crimes and those who ultimately were not), Blacks received sentences that were on average
27.9% longer than Whites, and Latinx defendants received sentences that were 15% shorter than White
defendants. Among the subset of Black defendants that were convicted of crimes, sentences for Black defendants
were 40% longer than those of White defendants, while sentences for Latinx defendants were 27% shorter than
for White defendants.
Again, however, as shown in Figure 15 below, these unadjusted disparities almost completely disappear when we
account for contextual factors. The main source of the disparities in length of incarceration is criminal history
and, in particular, previous incarcerations, which account for 70-90% of the raw Black/White disparity and 40-
50% of the Latinx/White disparity. Booking decisions remain an important secondary explanation for the
observed Black-White and Latinx-White disparities.
While Latinx defendants receive shorter terms of incarceration than White defendants, they receive longer
sentences of probation. When comparing Latinx defendants who were convicted to their White counterparts,
Latinx defendants received probation sentences that were 23.9% longer, for reasons that could not be identified.
! !
23
!
Figure 17. What Affects Black/White Differences in Sentence Length?
!
!
!
! !
-17%
2%
0%
0%
2%
7%
86%
24%
2%
-16%
4%
0%
0%
1%
8%
78%
39%
1%
-40.% -20.% 0.% 20.% 40.% 60.% 80.% 100.%
Individual!Demographics
Day,!Month,!and!Week!of!Arrest
Police!Activity!at!Home
Police!Activity!where!Arrested
Characteristcs!of!Home
Chacteristics!of!Arrest!Location
Criminal!Record
Booked!charges
Attorney!load
Sentence!if!Convicted Se ntence
24
!
Figure 18. What Affects Latinx/White Differences in Sentence Length?
!
!
!
Public Defender Resource Constraints
Finally, we sought to consider possible constraints on the Public Defender’s office, since different cases unfolding
simultaneously compete for the focus of each individual public defender. When charges are modified by the
prosecutor, it is generally due to a negotiation with defense counsel, and the amount of time an individual attorney
has with the defendant and case file can impact the attorney’s ability to learn about the client’s specific situation
and thus to advocate on their behalf. We measured the number of times the client’s attorney representation
changed, meaning that a different attorney was handling a court event for the client. We also calculated the average
number of court events for other cases that each defendant’s primary attorney was responsible for during the weeks
10%
-8%
0%
1%
2%
3%
52%
15%
-2%
6%
-11%
0%
1%
4%
1%
40%
20%
-2%
-20% -10% 0% 10% 20% 30% 40% 50% 60%
Individual!Demographics
Day,!Month,!and!Week!of!Arrest
Police!Activity!at!Home
Police!Activity!where!Arrested
Characteristcs!of!Home
Chacteristics!of!Arrest!Location
Criminal!Record
Booked!charges
Attorney!load
Sentence!if!Convicted Se ntence
25
!
that court events for the defendant’s case took place. The average number of other court events for public defenders
was 26, and was slightly lower (25) for black defendants than white defendants (27), a difference that is statistically
significant. This is consistent with the fact that black defendants are more likely to be facing felony charges, and
our understanding is that the public defender’s office makes efforts to assign fewer cases to attorneys handling
felonies.
Ultimately, caseload differences across public defenders were not a major explanation of racial disparities in case
outcomes, accounting for only 5% (or less) of the unadjusted disparity for all of the prosecutor activity outcomes
listed above in Table 1 and Figures 7-8. This suggests that increasing the number of public defenders representing
this group of defendants is not likely to resolve the different outcomes seen among similarly situated Black, Latinx,
and White defendants.!
Conclusions and Questions for Policy Makers
Disparities in the criminal justice system have an impact that extends beyond the four corners of a criminal charge
or conviction. They create and perpetuate inequalities in poverty, family formation, education, and child
development. Understanding why Black and Latinx defendants experience disproportionately worse criminal
justice outcomes can help policy makers and practitioners mitigate the disparities: by focusing on specific
contributing factors associated with race-based negative outcomes, we reduce the likelihood that race is a cause of
disparate treatment in our system of justice.
Our analysis of several years of cases from the San Francisco Public Defender’s Office suggests that “equal justice
for all” may be elusive in San Francisco for people of color. We observed systematic differences in outcomes for
Black, Latinx, and White defendants across almost all metrics evaluated.
The main factor explaining these disparate outcomes appears to be racially disparate booking charges imposed by
the police, which remain in the system through the downstream case adjudication process managed by prosecutors,
defense attorneys, and judges. Moreover, the influence of these booking decisions is actually larger than what is
shown by our figures, because today’s booking decisions become tomorrow’s criminal record, and a defendant’s
criminal history was the second most important contributing factor in both the length of time a defendant would
spend in custody during the adjudication process, and the length of sentence for those convicted of crimes.
Booking decisions influence downstream decisions made by district attorneys, public defenders, and judges.
District attorneys and public defenders are making what appear to be race-neutral decisions in response to the
charges brought to them by the police – but police bring more severe charges against Blacks and Latinx relative
to Whites, and that then persists throughout the case adjudication process.
26
!
If we desire a criminal justice system in which similarly situated defendants experience similar outcomes, it may
not be sufficient for defense attorneys, prosecutors, and judges to be merely race-blind participants themselves.
Given the important role they play as checks and balances on other parts of the system, it may be necessary for
these parties to actively mitigate unwarranted racial disparities that occur in earlier stages of the process. Our
analysis suggests that to date, the actions of prosecutors, public defenders, and judges do not actively increase
disparities – but neither have they undone disparities attributable to upstream booking decisions.
Booking decisions can be thought of as police responses to alleged criminal behavior committed by a defendant
with specific characteristics. Our data do not permit the perfect separation of these criteria for independent
analysis, and additional research is needed to ensure the utility of further reforms. It is possible that there are
legally relevant factors outside of those accounted for in the present study – most importantly, the actual criminal
behavior observed relative to the specific charges that are filed – that affect racial disparities in charging at the
booking stage. Future studies that examine police behavior and attitudes – dashboard camera media, incident
reports, officer statements, and witness testimony, for example – could shed light on this important issue.
To the extent that the Office of the Public Defender and the District Attorney have a shared goal of reducing
unwarranted racial disparities, careful scrutiny of booked charges is needed. Moreover, policies that can mitigate
the adverse downstream consequences (from the perspective of the defendant) of a prior criminal record—such as
use of actuarial risk assessment tools rather than prior record as a proxy for risk in bail setting, more flexible
sentencing, or improved access to expungement services—may also serve to reduce disparities.
Conviction
Review
Units:
A National
Perspective
4
