TOUGH ON CRIME (ON THE STATE'S DIME): HOW
VIOLENT CRIME DOES NOT DRIVE CALIFORNIA
COUNTIES’ INCARCERATION RATES—AND WHY IT
SHOULD
W. David Ball*

Abstract
California’s prisons are dangerously and unconstitutionally
overcrowded; as a result of the Supreme Court’s recent decision in Plata
v. Schwarzenegger, the state must act to reduce its prison population or
face court-ordered prisoner releases. The state’s plans to reduce
overcrowding are centered around what it calls criminal justice
“realignment”, whereby California will send a portion of the state prison
population to county facilities. The plan faces opposition from county
officials, who see it as pushing the state’s problem on to the counties.
But what if state prison overcrowding is really a county problem?
I argue that state prison overcrowding is due in large part to county
decisions about how to deal with crime. Using data from 2000-2009, I
will show that California’s counties use state prison resources at
dramatically different rates, and, moreover, that the counties which use
state prisons the most have below-average crime rates.
The contribution the Article makes, then, is twofold. First, it
reinforces that incarceration in state prisons is one policy choice among
many, not an inexorable reaction to violent crime. Counties can and do
make different choices about how to respond to violent crime, including
the extent to which they use prison. Second, the Article demonstrates why
localities are crucial—and critically underexamined—contributors to state
prison populations. Decisions are made at local levels about prosecution,
investigation, plea bargaining, and sentencing, and these decisions are
made by officials who are either elected locally (such as DA’s, judges, and
sheriffs) or appointed locally (police and probation officers). Local
*

Acknowledgements. Dan Ho. Byrd Ball. SCJC Executive Sessions participants.
Santa Clara. Oregon. Research Assistants: Vincent Ang, (Nik Warrior), Eugene Lee.
Naomi Levy.
Ian McAllister-Nevins. Tim Coxon.
Bob Weisberg.
Karthick
Ramakrishnan. Debbie Mukamal. Joan Petersilia.

Electronic copy available at: http://ssrn.com/abstract=1871427

2

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

policies and policymakers affect the state’s corrections budget, even
though the state has no say in designing or implementing these policies.
State officials must take these local differences into account, and create
incentives for counties to behave differently.
The problem is that it is difficult to distinguish between justifiable,
crime-driven incarceration and optional, policy-driven incarceration. I
propose a new metric for distinguishing between these two types of
incarceration, one which defines justified incarceration in terms of violent
crime. This would allow the state to manage local usage of state prison
resources without either penalizing crime-ridden areas or rewarding
prison-happy ones.
This Article is the first of two articles dealing with the state/county
prison relationship. While this Article quantifies the ways in which the
extent of local prison admissions is not necessarily a function of the violent
crime rate, a second Article will examine whether, given these differences,
it makes sense for the state to subsidize county commitments to prison.

Contents
INTRODUCTION....................................................................................................... 3
A. THE COVERAGE MODEL ................................................................................................ 9
B. WHY COVERAGE MATTERS ........................................................................................12
I. SOURCES, LIMITATIONS, AND METHODS OF THE STUDY.................. 14
A. SOURCES ........................................................................................................................15
B. LIMITATIONS OF THE STUDY ......................................................................................20
C. METHODS .......................................................................................................................24
II. VIOLENT CRIME RATES AND NEW FELON ADMISSION RATES ...... 27
A. THE STATE ....................................................................................................................27
B. VIOLENT CRIME AND NFA IN THE FOUR STATE SEGMENTS ................................30
1.
High Use Counties: Dominated by the Subsidized.............................32
2.
High Subsidy Revisited: The Rich Four and the Poor Four ...........33
3.
Low Use Counties: The Convergence of Low Coverage and Low
Subsidy 35
4.
Low Coverage and Subsidy Divided by Income: The High Five
and the Low Six .......................................................................................................................36
5.
Los Angeles..........................................................................................................37
6.
Middle Use Counties ........................................................................................38
III. ALTERNATIVE EXPLANATIONS............................................................... 38
A. ARREST DATA ...............................................................................................................39

Electronic copy available at: http://ssrn.com/abstract=1871427

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

3

B. LOCAL DISPOSITIONS ...................................................................................................41
1.
Jail............................................................................................................................42
2.
Probation .............................................................................................................44
C. LOCAL RESOURCES .......................................................................................................45
D. POLITICS ........................................................................................................................49
E. REVERSE CAUSALITY: IS LOW CRIME THE PRODUCT OF A HIGH NFA? .............52
IV. FISCAL IMPLICATIONS................................................................................ 53
A. SUBSIDY BY SEGMENT .................................................................................................53
B. RECALCULATING STATE COVERAGE RATES BY SEGMENT .....................................58
V. POLICY IMPLICATIONS................................................................................. 60
A. REALIGNMENT, PRISONER RELEASE.........................................................................60
B. DE-SUBSIDIZING PRISON, RE-SUBSIDIZING PROBATION ......................................62
C. STATE POPULATION CONTROL AND DETERMINATE SENTENCING......................63
***.............................................................................................................................. 64

INTRODUCTION
California’s prisons are dangerously and unconstitutionally
overcrowded.1 The state must find a way to cut its prison population by
tens of thousands of prisoners or it will be forced to release prisoners by
the federal courts.2 The state has long conceded that the conditions in its
prisons violate the Eighth Amendment’s prohibition on cruel and unusual
punishment,3 but it has struggled to find ways to sufficiently reduce
overcrowding.4 Earlier this year, the state passed AB 109, a bill which
radically reconfigures the relationship between local governments and the
state prison system.5 AB 109, Criminal Justice Alignment, will, once it is
funded,6 shift many parts of the state prison system from the state level to

1

Brown v. Plata, 563 U.S. ___, slip op. at *4-8 (2011).
Id. at *2 (“[A]bsent compliance through new construction, out-of-state transfers, or
other means … the State will be required to release some number of prisoners before
their full sentences have been served.”).
3
Id. at *9.
4
Id. I note that the state reduced its prison population by 9,000 during the pendency
of its appeal to the Supreme Court. Id. at *3.
5
Because the bill changes so many individual statutes, I have cited to the Legislative
Counsel’s Digest, available at http://www.leginfo.ca.gov/pub/11-12/bill/asm/ab_01010150/ab_109_bill_20110404_chaptered.html.
6
The plan is currently in limbo, as Republicans and Democrats continue to fight
over the state’s budget deficit. See Don Thompson, California Law to Shift Inmates
Hinges on Elusive Funds, Associated Press Apr. 4, 2011, available at
http://www.mercurynews.com/news/ci_17775548?nclick_check=1.
2

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

4

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

the county level.7 Local reaction to the plan has been mixed. Localities
want more control, but they do not want to foot the bill.8 Some members
of the California assembly opposed to the plan see the overcrowding
problem as a failure of state leadership, and fear that realignment will
result in threats to public safety.9
But what if state prison overcrowding is really a county problem, and
the state is simply returning the problem to the counties? Local officials,
not state officials, control the inflow into prison, through decisions about
which crimes to investigate, whom to arrest, and whom to prosecute.
Juries are empanelled locally, and the judges who preside over the
proceedings are elected locally. The only thing statewide about the prison
system is that the state pays for it.10 Zimring and Hawkins famously
referred to this as “the corrections free lunch” in their 1991 book, The
Scale of Imprisonment.11
7

The default punishment for felonies is now 16 months or 2-3 years in county jail;
before AB 109, the default punishment was the 16 months or 2-3 years in state prison.
Id. The bill will also transfer the state’s parole system to the counties.
8
See, e.g., Curt Hagman, Governor’s Plan: Early Release Disguised as
Realignment, San Bernardino County Sun, May 7, 2011. (Author, a California
Assemblyman, agrees that localities can do a better job than the state but argues that it
will cost his county (San Bernardino) money.) See also Thompson, supra note 6 (citing
California State Sheriff’s Association spokesman as saying the program is a “potential
disaster” without guaranteed funding).
9
See, e.g., Shannon Grove, Taxpayers and Prisons, The Daily Independent, June 8,
2011 (Author is a California Assemblywoman).
10
In this Article, I am specifically using the word prison to mean the state prison
system. This is not the only carceral option available, of course. Counties have jails,
where they sentence offenders, process arrestees, and hold those who can’t make bail
until trial.
11
Franklin E. Zimring and Gordon Hawkins, The Scale of Imprisonment 211
(1979). In California, county revenues pay for public protection, which includes judicial
expenditures (including trial courts, clerks, the District Attorney, and the Public
Defender), police and sheriffs, and detention and corrections (adult and youth detention,
probation). Some counties receive block grants from the state through a number of
different programs, most prominently the Local Public Safety Fund (LPSF) and the Local
Safety and Protection Account (LPSA). The LPSF is funded through a ½ cent sales tax.
Cal.Const. Art. 13, § 35. Funds are distributed based on counties’ share of total state
taxable sales. Cal.Gov.Code § 30052 (West 2011). The LPSA is funded through the
vehicle license fund and, in turn, directs most of its funds to particular programs dealing
with juvenile justice, law enforcement, and juvenile probation. Cal. State Ass’n of
Counties, Local Public Safety Funding Summary 2 (May 2009), available at:
www.counties.org/images/.../CSAC-CSSA-CPOC%20FAQ_May%2018.pdf. Both the
juvenile justice program and the law enforcement program make their disbursements

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

5

As the state seeks to manage its prison population, then, it must
account for the potential policy distortions the prison subsidy creates. The
difficulty is in distinguishing between incarceration that is, in some sense,
justified by crime problems, and that which is the result of policy choices
localities make about how to deal with that crime.
While several studies have explored the relationship between
incarceration and crime, most have focused on the state and national
level.12 No study has focused on the ways in which county governments
contribute to overpopulation in the adult prison system. An unpublished
paper by Tuosto and Peckenpaugh suggested that policy differences might
explain the differences in county commitments to the state Department of
Juvenile Justice.13 A recent study looked at sentencing models in rural and
urban areas of Nevada.14 The ACLU has also looked at California county
based on county population; the juvenile probation program allocations are fixed by
statute. California Legislative Analyst's Office, Judicial and Criminal Justice 2008-09
Analysis,
d-21-d-26,
available
at:
http://www.lao.ca.gov/analysis_2008/crim_justice/crimjust_anl08.pdf. 37 counties also
receive funds of equal amounts through the Small and Rural Sheriffs’ Grants.
Cal.Gov.Code § 30070 (West 2011).
I note that none of these disbursements is made on the basis of demonstrated
financial need, nor are they made on the basis of a county’s level of crime. One
complicating point: County revenues themselves come in large part from the state
(29.03%) and federal (17.30%) government, meaning that the division between state and
county (and federal government and county) is complex. California State Controller,
2008-09 Counties Annual Report, iii, available at http://www.sco.ca.gov/Files-ARDLocal/LocRep/counties_reports_0809counties.pdf.
12
Michael Tonry, in his 2004 survey of the existing research, considered several
possible explanations for why the U.S. as a nation incarcerates at such a high rate relative
to other countries, concluding that the high crime explanation “has virtually no validity.”
Michael Tonry, Thinking About Crime 27. (2004). Bruce Western comprehensively
analyzed the commonly-provided causes of incarceration, ranging from politics to state
sentencing, but he focused primarily on the state level as well. Bruce Western,
Punishment and Inequality in America (2006). Western’s compelling examination of
crime and incarceration surveys research involving cities and neighborhoods, but his
analysis does not focus on sub-state political units as political, policy-making entities. Id.
at 36. His own comparison of murder and incarceration rates compares states to one
another. Id. at 49. His analysis of politics, state penal laws, and the role of discretion in
sentencing are all focused on the state level. Id. at 59-66.
13
On file with Author.
14
Victoria Springer et. al, Felony Sentencing in Rural and Urban Courts: Comparing
Formal Legal and Substantive Political Models in the West, available at
http://ssrn.com/abstract=1441593.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

6

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

variations in the imposition of the death penalty.15 The California state
Offender Information Services Branch broke down the population of
second and third strikers by county and strike offense, but did so only for
a single year and only for strike offenses.16 Twenty years after Zimring
and Hawkins wrote that the correctional free lunch required “empirical
and theoretical work which will both complicate and enrich the public
choice model with special reference to decisions about imprisonment,”17
few studies have been produced. This Article and the one to follow will
try to fill that gap.
California is a massive state, with roughly one tenth of the country’s
population. Its prison population is nearly the same size as the population
in the federal system. Los Angeles County alone has a population greater
than all but eight states. Eight counties besides Los Angeles have more
than a million people,18 a population larger than that of the smallest states.
California is, therefore, a good place to start the analysis of the counties’
role in state prison overpopulation: the scale of California’s prisons—as
well as the scale of its overcrowding—are of national import.
California can be thought of not only as a single state, but also as a
collection of 58 counties. Counties are significant political entities in their
own right, distinct from the state. Counties are run by their residents:
there is no statewide politicking in local elections for Sheriff, or District
Attorney, or county council, or judge. A California voter in one county
has no say in how another county makes its criminal justice decisions.
The pair of Alameda and San Bernardino Counties presents perhaps
the starkest example of how these decisions can affect counties’ use of
state prison resources. A ten-year average of county data (2000-09)

15

Romy Ganschow, Death by Geography: A County by County Analysis of the Road
to
Execution
in
California,
2008,
available
at
http://www.aclunc.org/issues/criminal_justice/death_penalty/death_by_geography_a_cou
nty_by_county_analysis_of_the_road_to_execution.shtml.
16
Department of Corrections and Rehabilitation, Offender Information Services
Branch, Second and Third Striker Felons in the Adult Institution Population, June 30,
2009,
available
at
http://www.cdcr.ca.gov/Reports_Research/Offender_Information_Services_Branch/Quart
erly/Strike1/STRIKE1d0806.pdf.
17
Zimring and Hawkins, supra note X, at 215.
18
In alphabetical order: Alameda, Contra Costa, Orange, Riverside, Sacramento,
San Bernardino, San Diego, and Santa Clara.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

7

shown on the chart below indicates that both counties have similarly-sized
populations, similar amounts of reported violent crime (criminal homicide,
rape, robbery, and aggravated assault), similar amounts of reported
property crime (burglary, motor vehicle theft, and larceny-theft over
$400), and similar amounts of all reported “Part I” crime (all of the above
crimes plus larceny-theft under $400 and arson).19 Overall crime rates are
nearly identical: Alameda is a little more violent and San Bernardino is a
little worse for property crime. Both counties are part of the same state,
governed by the same penal code and state judicial system, yet ten-year
averages of prison usage for that time show two radically different
outcomes: San Bernardino’s prison population was more than twice as
high, on average, as Alameda’s, and it sent an average of more than three
times as many “new felons” to prison each year.
Table 1: Crime Comparison Between San Bernardino and Alameda
Counties, Average Yearly Values 2000-2009
San Bernardino

Alameda

Ratio of San
Bernardino to
Alameda

Total Population

1,923,360

1,506,740

1.28

Reported Violent
Crime

9,956.6

10,629

.94

38,762

36,072

1.07

72,454

74,194

.98

11,441

4,555

2.51

3,792

1,088

3.49

Reported
Property Crime
All Reported
Part I Crime
Yearly Prison
Population
Yearly New
19

The Uniform Crime Reporting Program divides crimes into Part I and Part II.
Part I crimes include criminal homicide, forcible rape, aggravated assault, burglary
(breaking and entering), larceny-theft not of a motor vehicle, motor vehicle theft, and
arson.
U.S. Dept. of Justice, FBI, UCR Offense Definitions, available at
http://www.ucrdatatool.gov/offenses.cfm. These offenses were chosen “because they are
serious crimes, they occur with regularity in all areas of the country, and they are likely
to be reported to police.” Id.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

8

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

Felon
Admissions
All figures are ten-year averages, 2000-2009.
The results of this comparison on a yearly basis are shown graphically
in Figure 1. I calculated the yearly data as a ratio (San Bernardino to
Alameda). As in the chart above, a ratio of one means the counties have
equal numbers for that particular category, a ratio above one indicates the
degree to which San Bernardino’s numbers exceed Alameda’s, and a ratio
below one indicates the degree to which San Bernardino’s numbers are
lower than Alameda’s. The chart clearly demonstrates that the year-toyear story is no different than that told by the ten-year average. During
all ten years, San Bernardino had at least twice the prison population and
more than twice the number of new felon admissions—sometimes much
more—and it did so without suffering from any more crime than Alameda.
Figure 1: San Bernardino and Alameda Crime Rates and Prison Usage
5.00

4.50

San Bernardino As Multiple of Alameda

4.00

3.50
Total Population

3.00

Violent Crimes Total
Property Crimes Total

2.50

Part I Crimes Total
Prison Population

2.00

New Felon Admissions

1.50

1.00

0.50

0.00
2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Year

These two counties, then, are almost identical in material ways when it
comes to crime, but they are incredibly different when it comes to their
usage of state prison resources. For new felon admissions alone, San
Bernardino costs the state, on average, $93,045,566 more each year than
Alameda; its total prison population costs the state, on average, an extra
$236,761,677 each year. This is not a difference that can be explained by
reference to reported crime rates. The state is paying for San
Bernardino’s decision to treat crime with prison, but Alameda—indeed,

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

9

any California citizen who does not live in San Bernardino—has no say in
electing the people who design San Bernardino’s criminal justice policies.
Why should the state pay for a decision only some of its citizens make,
when residents of other counties make different decisions?
The most persuasive justification for the use of prison is that it is a
response to crime; that is the argument I will primarily address in this
paper. I assume for purposes of my analysis that crime rates are
exogenous. Taking the “prison as a response to crime” argument at its
strongest means assuming that counties do not (or cannot) breed crime
through policy. I do not believe this is necessarily true, but I wish to
cabin the scope of the Article.20 I am also not arguing that prison should
not be used to treat crime; I am simply saying that violent crime rates
alone cannot explain the difference in usage. I specifically focus on
violent crimes because all the dominant justifications for imprisonment—
incapacitation, retribution, and deterrence—consider violent crimes to be
the most worthy of incapacitation, the most deserving of punishment, and
the most serious offenses to be deterred.21
My analysis starts with the proposition that the average of a state as
large as California—and with a single county larger than all but eight
states—smooths over very real differences, much like taking the per capita
average income in a room with Bill Gates would also be misleading.
While I do examine data at the statewide level, the bulk of my analysis
will focus at the county level. This analysis shows that San Bernardino
and Alameda are not anomalous: the state as a whole is divided among
counties which persistently use prison resources at high rates and those
which use prison at low rates. The group of counties with the highest
usage of prison has, as a whole, below-average violent crime rates. They
also have lower property and “Part I” crime rates as well. The argument
that prison usage is driven by violent crime rates has no statistical support.
A. The Coverage Model
In this Article, I propose that violent crime rates should be driving the
state’s willingness to pay for localities’ prison commitments. I divide the
20

For the argument that prison is criminogenic, see, e.g., Sharon Dolovich,
Incarceration American-Style, 3 Harv. L. & Pol’y Rev. 237 (2009).
21
Some observers have argued that drug crimes might best be dealt with outside the
criminal justice system entirely. There are no reported drug crimes, however.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

10

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

state’s counties into four segments, based on the relationship within each
county between reported violent crime and the number of new felons it
sends to prison. To measure this relationship, I define a new variable, the
violent crime coverage rate. Coverage is the amount of new felon
admissions (NFA)22 for a given county in a given year as a percentage of
reported violent crime23 for that county in the same year. Mathematically,
Coveragecountyyear = NFAcountyyear /(Violent Crimecountyyear).
A county with 100 reported violent crimes and 50 NFA would have a
coverage rate of 50%. A county with 100 reported violent crimes and 10
NFA would have a coverage rate of 10%. Higher numbers indicate more
carceral responses: for a given level of violent crime, a county with higher
coverage sends a larger number of offenders to prison. Counties with
lower numbers “cover” their violent crimes with fewer NFA.24 Some
variance might be explained by the types of violent crime—more murders,
for example—and I will explore whether this is the case.25
I focus on NFA, not total prison population, for a number of reasons.
First, I find persuasive Stephen Raphael and Michael Stoll’s model of the
total prison population: they model prison population as a function of
admission rates, release rates, and the prison population the year before.26
22

NFA measures admissions to prison of those convicted of a new crime, and, as
such, is distinct from other parts of the prison population, most notably parolees
returning to prison on either a “technical” parole violation (e.g. failed drug test) or a new
crime (charged as a parole violation instead of, say, a burglary). NFAs describe new
terms for new offenses; they do, of course, include recidivist prisoners who have been
previously incarcerated.
23
Reported violent crimes include homicide, rape, robbery, and aggravated assault.
24
I note initially that coverage rates might be explained by a number of factors:
higher clearance rates (more efficient law enforcement), more aggressive policing
strategies (e.g. broken windows), or something to do with the seriousness of the
particular offenses (e.g., those facts deserving of more serious punishment).
25
My preliminary conclusion is that rates of each type of violent crime are lower in
counties which use a lot of prison resources, and, moreover, that the more serious
crimes, such as homicide, have too few cases to account for much of a difference.
26
Stephen Raphael and Michael Stoll, Why Are So Many Americans in Prison?, in
Raphael and Stoll (eds.), Do Prisons Make Us Safer? The Benefits and Costs of the
Prison Boom 6 (2008). [Note: my pagination refers to the electronic copy available at
http://www.law.berkeley.edu/files/why_are_so_many_americans_in_prison.pdf.]
Raphael and Stoll conclude that the increase in population is not due primarily to
increases in crime, characterizing the rise in incarceration as a policy experiment. Id. at
65.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

11

Though I believe this model to be a good one, it would be difficult to test
empirically. Testing this model with my data would require me to isolate
changes in sentences for particular classes of offenders (which affects time
to release), the momentum effect of a large population, and the length of
time to which new prisoners are sentenced. It would be difficult to tease
out whether a county had a large population in a given year because there
were a sizeable number of people from that county who remained in
prison on long terms or because that county was sending more people to
prison in the first place. Not only is the data difficult to obtain; it is
harder still to determine whether a given sentence is justified or not. It is
difficult enough to determine what constitutes a “real offense”; it is that
much more difficult to determine the “real” sentence length of a given
offense, as the U.S. Sentencing Commission has so often demonstrated.
NFA, instead, simply measures who enters prison from a given
county, not how long they stay there. Its simplicity is not without its
costs, however. It is, of course, possible that Low Use counties are
nevertheless sending non-serious offenders to prison, and that High Use
counties are sending only the most hardened criminals to prison. If that is
the case, the method I have chosen will not account for that. However,
the fact that the violent crime offense mix is no worse in High Use
counties than in Low Use ones might indicate that this is unlikely.
In addition to using coverage in its own right, I also use it to calculate
the prison subsidy a county receives or forgoes. I define the amount of
necessary incarceration as violent crime in a county times the statewide
coverage rate. That is, the state average is the “fair” amount of
incarceration justified by a particular amount of violent crime; anything
above that constitutes a local policy choice that is being subsidized with
state funds. This is obviously a strong choice on my part, but it aligns
with the thrust of my argument: that a county’s deviations from state
policy should not be subsidized by the rest of the state. If a county makes
different choices from the state as a whole, it should bear the cost of those
burdens (and reap the benefits).
The statewide average, then, is a proxy for the amount of incarceration
dictated by violent crime itself, not a county’s unique response to violent
crime. Calculating subsidies in this way more closely ties prison usage to
the justification for that usage, and differentiates between counties which
have to use a lot of prison and those which choose to use a lot of prison.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

12

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

I calculated the subsidy as follows. I used the state’s coverage rate for
a given year and multiplied it by the number of reported violent crimes in
each county that year to determine the “crime justified” NFA. I then took
the actual NFA numbers for a county and subtracted the “crime justified”
number from it to arrive at that county’s NFA surplus (or deficit). I then
multiplied this surplus figure by the per capita prisoner cost to arrive at
subsidy/deficit numbers. Mathematically,
Subsidycountyyear = (GrossNFAcountyyear – (Coveragestateyear * Violent
Crimecountyyear)) * Per Capita Prison Coststateyear
I emphasize that these subsidy figures are not, again, a measure of the
total cost of prison.27 This estimate only calculates the cost for the first
year of imprisonment for new felons. Sentence lengths are undoubtedly
an immense factor in determining the total cost of a county’s use of state
prisons. That is, a county with a below-coverage NFA number might
nevertheless have higher costs because their felons stay in prison longer.
(Of course, it could also be that counties with higher NFA rates also give
longer sentences, in which case the subsidy numbers will be
underweighted relative to the ultimate fiscal cost.) Nevertheless, I believe
that the cost of NFA provides us with a useful entry point to see which
counties benefit from prison subsidies and which counties are taxed by
them.
B. Why Coverage Matters
If the violent crime to NFA relationship is not predictive at the state
level, this raises two obvious questions: what might explain it, and why
does this even matter? As to the first question, I consider a variety of
explanations: other crimes, local law enforcement, politics, per capita
income, and the use and type of in-county dispositions. My exploration of
these subjects is, for space reasons, tentative, but I have posted my dataset
online and encourage others to do more detailed analysis.

27

We can easily get that number by multiplying the total numbers of prisoners from
a given county by that year’s cost per prisoner. That number, however, treats prison as a
thing unto itself. Using coverage to calculate subsidies, however, accounts for the best
reason for incarceration: violent crime. Incarceration at the statewide coverage rate is
justified; anything else is surplus.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

13

As for the significance of the analysis, I see three contributions. First,
there are very real fiscal impacts to counties’ usage of prison, ones which
are not transparent enough in the present system. By controlling for the
influence of violent crime, my estimation of the fiscal impacts is a closer
representation of differences in policy among counties, policy choices
which are subsidized by the unwilling residents of other counties. This
Article is part of a two-part series which examines why states should
subsidize state prisons when local officials decide who is sent there.28
This Article will, I hope, dispel the idea that the level of prison usage in
California is a necessary result of crime.
Using the coverage rate model of prison subsidy, I will show that
some individual counties who make different policy choices—choices not
dictated by the average response to violent crime—cost the state tens of
millions of dollars a year, every year, while others leave tens of millions
of dollars of prison resources on the table. I also explore what would
happen if the entire state incarcerated at the coverage rate of the most
carceral counties. I do this to raise a key question: if one county or set of
counties is entitled to incarcerate at a given rate, why shouldn’t other
counties do so as well? And if the state can only afford to have some
counties incarcerating greater numbers of people per violent crime, which
ones get to do so, and on what basis? Ultimately, I am concerned with
how residents of underincarcerating counties can rein in over-incarcerating
counties in the present system, given that all citizens pay for prison
equally through general state revenues, regardless of how heavily their
counties use prison.
The second point is that state prison problems are not necessarily best
addressed by statewide solutions. As this Article demonstrates, counties
operating under the same set of laws and in the same court system get
widely different results. Statewide solutions—such as changes to statutes,
sentencing commissions, and the like—are almost always proposed as the
means of addressing state prison overpopulation. But, because they fail to
address the differences in local enforcement, they cannot effectively
address the problem. In other words, without a correct diagnosis of the
cause of state prison usage, solutions cannot cure the disease.29
28

See also Why Should States Pay For Prisons, When Local Officials Decide Who
Goes There? Available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1871274.
29
Franklin Zimring, in a recent article, observes that the huge growth in prison
population during the 1970s and 1980s was not accompanied by any significant changes

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

14

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

Third, this analysis has important ramifications for the state’s
implementation of criminal justice realignment. The question of how
much incarceration counties will be expected to deal with inside the county
depends crucially on how California sets the baseline rate of how much
state resources the county is permitted to use. The current plan is to set
the baseline at current levels of prison usage. This would be a mistake, in
my view, because it would make permanent the state subsidies of what
appear to be policy choices. Just because a county has been using state
prisons at a given rate does not mean that it had to. I propose, instead,
that the state base prison usage on reported violent crime rates and the
statewide violent-crime-to-new-felon-admission coverage rate. This would
tie funding to need, rather than funding to use.
***
This Article proceeds in four parts. In Part I, I explain the sources
and methods used for this Article. In Part II, I examine the relationship
between crime and incarceration. In Part III, I examine other possible
explanations for differences in county commitments to state prisons. In
Part IV, I examine the fiscal implications of differences in incarceration
rates, demonstrating that counties which incarcerate at a relatively greater
rate are doing so at great cost to the state: that is, they are tough on crime
on the state’s dime. I conclude by discussing some potential policy
implications this analysis has for the future of California criminal justice
reform.
I. SOURCES, LIMITATIONS, AND METHODS OF THE STUDY
In this Section, I will discuss briefly how I conducted this study. I
begin by describing the data sources used in this Article, all of which are
made available online by the state. I then discuss some limitations with
this study which might explain the results. I then discuss further the ways
in which I subdivided the state on the basis of violent crime coverage rates

in state penal codes. Because of the discretion in the American system, however,
“substantial changes in aggregate punishment policy can take place without any
substantial change in the legislation governing the levels of punishment available or the
choice of punishments in individual cases.” Franklin E. Zimring, The Scale of
Imprisonment in the United States: Twentieth Century Patterns and Twenty-First Century
Prospects, 100 J. Crim. L. & Criminol. 1225, 1232 (2010).

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

15

and the calculated prison subsidy.
A. Sources
The state of California maintains several public databases available on
the internet; it also publishes annual reports on the offender population
incarcerated in the state’s prison. The data I used in this analysis came
from these sources and dates from 2000 to 2009. All data has been
compiled into a single spreadsheet which I have made available online.30 I
will discuss sources for particular data, as well as changes to the data I
made, where necessary to account for things such as the difference
between calendar year and fiscal year reporting.
County population. The California Department of Justice uses
estimates from the State Department of Finance to generate three
potentially useful county population figures, divided by age: Total
Population at Risk, (ages 10-69), Adult Population at Risk (ages 18-69),
and Juvenile Population at Risk (ages 10-17).31 The term “at Risk”
presumably refers to those people who are at greatest risk of becoming
involved with the criminal justice system, either as juveniles or adults. I
have used the Adult Population at Risk (APAR) figures throughout this
Article and have calculated crime, arrest, and new felon admission rates
using raw numbers and dividing by these population figures, normalizing
per 100,000.32 I did so to avoid differences in rates that might stem from
using different population types. The California Department of Finance
30

https://docs.google.com/leaf?id=0B5rP1OL_xn65MzYyYzVhYjMtNGZkNy00NW
E5LWEzNWYtZTQ0YWQwNzRjYjcw&hl=en_US.
31
See, e.g., Cal. Dept. of Justice, Office of the Attorney General, Criminal Justice
Statistics Center, Statistics by City and County, “Population Estimates, 2000, by
County”, Tbl. 27 (2000), available at http://stats.doj.ca.gov/cjsc_stats/prof00/00/27.pdf.
32
The entry page for the Criminal Justice Statistics Center is available at
http://ag.ca.gov/cjsc/datatabs.php. Individual population reports are available at the
following
locations:
http://stats.doj.ca.gov/cjsc_stats/prof00/00/27.pdf
(2000),
http://stats.doj.ca.gov/cjsc_stats/prof01/00/27.pdf
(2001),
http://stats.doj.ca.gov/cjsc_stats/prof02/00/27.pdf
(2002),
http://stats.doj.ca.gov/cjsc_stats/prof03/00/27.pdf
(2003),
http://stats.doj.ca.gov/cjsc_stats/prof04/00/27.pdf
(2004),
http://stats.doj.ca.gov/cjsc_stats/prof05/00/27.pdf
(2005),
http://stats.doj.ca.gov/cjsc_stats/prof06/00/27.pdf
(2006),
http://stats.doj.ca.gov/cjsc_stats/prof07/00/27.pdf
(2007),
http://stats.doj.ca.gov/cjsc_stats/prof08/00/27.pdf
(2008),
http://stats.doj.ca.gov/cjsc_stats/prof09/00/27.pdf (2009).

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

16

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

estimates the total adult population for each county as of July 1 of each
year.33 I have used total population figures to contrast with Adult
Population at Risk only where noted. These figures do not include
relevant information about population distribution—e.g., degree of
urbanization—that might be relevant drivers of crime and/or carceral
responses, nor do they include figures about racial and/or ethnic
subpopulations within a given county, which might also be relevant.
Department of Finance figures do, however, account for both legal
residents and “unauthorized foreign immigrants.”34
Prison Population by County, New Felon Admissions by County,
and Parole Violators with a New Term by County. The California
Department of Corrections and Rehabilitation publishes annual population
reports on prisoners housed in state prisons. Each year, the state
publishes the total population of prisoners by county of commitment as of
December 31 of that year,35 as well as yearly totals by county for new
33

State of California, Department of Finance, California County Population
Estimates and Components of Change by Year, July 1, 2000-2010. Sacramento,
California,
December
2010,
available
at
http://www.dof.ca.gov/research/demographic/reports/estimates/e-2/2000-10/view.php.
34
Id.
35
The
entry
page
for
these
reports
is
available
at
http://www.cdcr.ca.gov/Reports_Research/Offender_Information_Services_Branch/Annu
al/CalPrisArchive.html. The reports are authored by the Data Analysis Unit, Cal. Dep’t
of Corr. & Rehab. and titled “California Prisoners & Parolees” followed by the year
[hereinafter CDCR Population Report]. 2002 and 2001 reports contained data from the
year prior; reports after 2004 contained data from that year. In 2003, the Data Analysis
Unit combined two years’ worth of data into one report. Specifically, I used the
following tables from the following annual reports: 2000 data came from the 2001 annual
report
at
tbl.
10,
available
at
http://www.cdcr.ca.gov/Reports_Research/Offender_Information_Services_Branch/Annu
al/CalPris/CALPRISd2001.pdf; 2001 data come from the 2002 annual report at tbl. 10,
available
at
http://www.cdcr.ca.gov/Reports_Research/Offender_Information_Services_Branch/Annu
al/CalPris/CALPRISd2002.pdf; 2002 and 2003 data were both contained in a 2003
report, with 2002 data at tbl. 10 (pdf page 34) and 2003 data at tbl. 10 (pdf page 139),
available
at
http://www.cdcr.ca.gov/Reports_Research/Offender_Information_Services_Branch/Annu
al/CalPris/CALPRISd2003.pdf; 2004 data are at tbl. 10 of the 2004 report, available at
http://www.cdcr.ca.gov/Reports_Research/Offender_Information_Services_Branch/Annu
al/CalPris/CALPRISd2004.pdf; 2005 data are at tbl. 10 of the 2005 report, available at
http://www.cdcr.ca.gov/Reports_Research/Offender_Information_Services_Branch/Annu
al/CalPris/CALPRISd2005.pdf; 2006 data are at tbl. 10 of the 2006 report, available at
http://www.cdcr.ca.gov/Reports_Research/Offender_Information_Services_Branch/Annu

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

17

felon admissions and parole revocations.36 I note that this population
figure is taken in a different month (December) than the county population
figures noted above (July), and that prison figures represent actual
headcounts, while county population figures are estimated.
Crime and Arrest Figures; Probation and Jail Figures. I used
Department of Justice published data for reported crimes,37 felony
arrests,38 adult probation caseload,39 and jail population figures.40 As
al/CalPris/CALPRISd2006.pdf; 2007 data are at tbl. 14 of the 2007 report, available at
http://www.cdcr.ca.gov/Reports_Research/Offender_Information_Services_Branch/Annu
al/CalPris/CALPRISd2007.pdf; 2008 data are at tbl. 14 of the 2008 report, available at
http://www.cdcr.ca.gov/Reports_Research/Offender_Information_Services_Branch/Annu
al/CalPris/CALPRISd2008.pdf; 2009 data are at tbl. 14 of the 2009 report, available at
http://www.cdcr.ca.gov/Reports_Research/Offender_Information_Services_Branch/Annu
al/CalPris/CALPRISd2009.pdf.
36
This data is reported in the CDCR Population Reports, supra note 9, at tbl. 5A for
the years 2000 and 2002-2006 and tbl. 15A for the years 2007-2009. 2000 data is in the
2001 report, 2002 data is in the 2002 report, and the 2003 report provides 2003 data in
the second set of tables (tbl. 5A, pdf page 128). Thereafter the data for a given year are
in that year’s report.
2001 data was not given in any of the the annual reports. It was, instead, taken from
Data and Analysis Unit, Dep’t of Corr., Characteristics Of Felon New Admissions And
Parole Violators Returned With A New Term, Calendar Year 2001 at tbl. 11 (Felon New
Admissions) and tbl. 16 (Parole Violators Returned with New Term), available at
http://www.cdcr.ca.gov/Reports_Research/Offender_Information_Services_Branch/Annu
al/ACHAR1/ACHAR1d2001.pdf. The figures from tables 11 and 16 were added to
arrive at total new admissions figures (calculated).
37
The entry page for the Criminal Justice Statistics Center’s county crimes data is
available at http://ag.ca.gov/cjsc/statisticsdatatabs/CrimeCo.php. Individual county data
was taken by following links to each county.
38
The entry page for the Criminal Justice Statistics Center’s county arrests data is
available at http://ag.ca.gov/cjsc/statisticsdatatabs/ArrestCoFel.php. Individual county
data was taken by following links to each county.
39
The entry page for the Criminal Justice Statistics Center’s adult probation data is
available at http://ag.ca.gov/cjsc/statisticsdatatabs/SuperCo.php. Individual county data
was taken by following links to each county. The data is incomplete: Contra Costa,
Merced, Sacramento, Siskiyou, Tulare, and Yolo county did not report separate
misdemeanor population counts. See Criminal Justice Statistics Center, Criminal Justice
Trend Data Footnotes, available at http://stats.doj.ca.gov/cjsc_stats/prof09/footnotes.pdf.
Mariposa County reported -47 people on the misdemeanor probation caseload for 2000,
so I deleted all data from that year; the same is true for San Joaquin County for 2002,
which reported a felony probation caseload of -423. Gaps in the data also crop up
intermittently and are a result of no data being reported; they should not be read as
zeroes.
40
The reports for jail data are located at the same url in note 13, supra. Total
figures might not add up due to projections and rounding of numbers. See Criminal

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

18

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

noted earlier, I have chosen to calculate rates per 100,000 APAR myself,
rather than rely on the state’s rates, to avoid differentials based solely on
different numbers (or definition) of population. I use crime and arrest
figures for two reasons. First, arrest figures can serve as a proxy for how
aggressive and/or effective law enforcement is in a particular locale at the
front-end (through the use of community policing, etc.). (I also examined
county clearance rates to determine how effective a given county was at
solving crimes.41) Second, because there are no reported drug crime
statistics, drug arrests serve as a proxy for drug crimes, albeit an
imperfect one, since they conflate policing resources, strategies, and
priorities with the level of underlying activity.
These data are subject to a number of limitations.42 If multiple crimes
take place, only the most serious is recorded.43 Crime is generally seen to
be underreported: a particular county might have actual crime rates that
are a greater or lesser percentage of reported crimes. The same is true
when an offender is arrested for multiple offenses.44 The state collects
information on dispositions; however, this data is marred by a very large
“other” category and the state cautions that dispositions “data may or may
not be representative at the county level.”45 Accordingly, I have focused
only on county jail and probation figures. Within the jail data, I have
ignored data on Type I facilities, which are used only for detentions of up
to 96 hours, not sentencing; I have, instead, used figures for Type II, III,
and IV facilities,46 because they can be used to sentence offenders. These
Justice Statistics Center, Criminal Justice Trend Data Footnotes, available at
http://stats.doj.ca.gov/cjsc_stats/prof09/footnotes.pdf.
41
The entry page for the Criminal Justice Statistics Center’s county clearance data is
available at county.http://ag.ca.gov/cjsc/statisticsdatatabs/ClearanceCo.php. Individual
county data was taken by following links to each county; I calculated clearance rates
using the number of cleared cases.
42
See, e.g., Criminal Justice Statistics Center, “Data Characteristics and Known
Limitiations,” available at http://stats.doj.ca.gov/cjsc_stats/prof09/limits.pdf. See also
Criminal Justice Statistics Center, Criminal Justice Trend Data Footnotes, available at
http://stats.doj.ca.gov/cjsc_stats/prof09/footnotes.pdf.
43
Id. at 1.
44
Id. at 2.
45
Id.
46
Cal. Code Regs. Title 15 § 1006 defines Type II, III, and IV facilities thus:
“Type II facility” means a local detention facility used for the detention of
persons pending arraignment, during trial, and upon a sentence of commitment.
“Type III facility” means a local detention facility used only for the
detention of convicted and sentenced persons.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

19

figures are taken from actual population reports and are divided between
sentenced and non-sentenced prisoners. Non-sentenced prisoners are
those who are denied bail, unable to make bail, or on some form of
temporary detention.
Cost per Prisoner. I calculated the cost per prisoner by using
corrections budget figures47 and dividing by the prison population. This is
a crude approximation of the cost per prisoner, since there are certain
fixed costs in the state prison system that are not fully realized on a
marginal basis, and because some of the funds go to the Department of
Juvenile Justice. However, this is the same method the Bureau of Justice
Statistics has used in its State Prison Expenditures series.48 Again,
because the state’s fiscal year goes from July to June 30, I averaged two
years together in order to get approximations of calendar year figures,
with the exception of 2000, for which I simply used 2000-01 figures.
“Type IV facility” means a local detention facility or portion thereof
designated for the housing of inmates eligible under Penal Code Section 1208 for
work/education furlough and/or other programs involving inmate access into the
community.
47
These figures come from the Final Budget Summary published by the Department
of Finance for a given year. The entry page for these documents is available at
http://www.dof.ca.gov/budget/historical/. Specifically, my budget figures for particular
years
came
from
the
2000-01
report
at
48,
available
at
http://www.documents.dgs.ca.gov/osp/GovernorsBudget/pdf/2000-01budsum.pdf;
the
2001-02
report
at
41,
available
at
http://www.documents.dgs.ca.gov/osp/GovernorsBudget/pdf/2001-02budsum.pdf;
the
2002-03
report
at
384,
available
at
http://www.documents.dgs.ca.gov/osp/GovernorsBudget/pdf/2002-03budsum.pdf;
the
2003-04
report
at
2,
available
at
http://www.documents.dgs.ca.gov/osp/GovernorsBudget/pdf/2003-04budsum.pdf;
the
2004-05
report
at
6,
available
at
http://www.documents.dgs.ca.gov/osp/GovernorsBudget04/pdf/fbudsum_04.pdf;
the
2005-06
report
at
11,
available
at
http://www.documents.dgs.ca.gov/osp/GovernorsBudget05/pdf/fbudsum_05.pdf;
the
2006-07
report
at
400,
available
at
http://www.documents.dgs.ca.gov/osp/GovernorsBudget/pdf/fbudsum_06.pdf; the 200708
report
at
14,
available
at
http://www.documents.dgs.ca.gov/osp/GovernorsBudget/pdf/fbudsum_07.pdf; the 200809
report
at
18,
available
at
http://www.documents.dgs.ca.gov/osp/GovernorsBudget/pdf/fbudsum_0809.pdf;
the
2009-10
report
at
7,
available
at
http://www.documents.dgs.ca.gov/osp/GovernorsBudget/pdf/fbudsum_09.pdf.
48
See, e.g., James J. Stephan, State Prison Expenditures, 2001, available at
http://bjs.ojp.usdoj.gov/content/pub/pdf/spe01.pdf.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

20

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

My calculations are actually lower than the estimates published by the
state Legislative Analysts’ Office, which estimated that the cost of
incarcerating each prisoner in California in 2008-09 was $47,102.49 My
estimate for the calendar year 2008 was $41,200.05. Because the LAO
has not released estimates for all the years in my survey, however, I
decided to use calculated figures. If anything, this indicates that the
subsidy the state pays to counties which are heavy users of the state prison
system—and the corresponding tax to those who do not use it as heavily—
might be greater than the figures used in this Article.
B. Limitations of the Study
The main difficulty with this study is deciding what proxy to use for
the “fair” rate of prison usage to which a county is rightfully entitled. I
make no normative claim about how a county should use prison, nor have
I found a statistical one.50 There is no consensus on this in California,
academia, or elsewhere. In fact, that is the point of this series of articles:
that given this lack of consensus, residents of a particular county should
not have to pay for the policy choices of residents of another county.
High coverage rates are not necessarily bad, nor are low ones good. My
point is only that if there is no consensus, high rates should not be
subsidized, nor low rates penalized. In other words, while I make no
claims about high usage itself, I do claim that the state’s prison resources
should not be distributed on a first-come, first-served basis.
While using violent crime rates is a crude measure of the need for
prison, I do not believe there is a “real offense” alternative. That is, there
is no way to readily look at a given criminal case or set of criminal cases
49

See California Legislative Analysts’ Office, California’s Annual Costs to
Incarcerate
an
Inmate
in
Prisonhttp://www.lao.ca.gov/laoapp/laomenus/sections/crim_justice/6_cj_inmatecost.asp
x?catid=3. See also California State Auditor, California Department of Corrections and
Rehabilitation: It Fails to Track and Use Data That Would Allow It to More Effectively
Monitor and Manage Its Operations, September 2009 at 26 (estimating an annual cost per
inmate in 2007-08 of $49,300); available at http://www.bsa.ca.gov/pdfs/reports/2009107.1.pdf.
50
In fact, as I have argued elsewhere, I believe that normative questions cannot be
avoided even in a heavily-quantified context. See W. David Ball, Normative Elements of
Parole Risk, 22 Stan. L. & Pol'y Rev. 395, 397 (2011) (questioning whether parole
release is “inherently about risk or inherently about desert, or whether it is irreducibly
about both.”)

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

21

and determine which ones should result in a prison sentence and which
ones shouldn’t. There are a number of complicating factors. The first is
plea bargaining. Charged offenses are an inaccurate measure of the real
offense because a DA might overcharge for strategic reasons, in order to
posture during plea or charge bargaining. Offenses might also be
undercharged as the result of such bargaining. The second complicating
factor is evidentiary. The strength of an individual case has as much to do
with evidentiary concerns as with the heinousness of the underlying
conduct. A case with bad facts might nevertheless get a lower sentence
due to a lack of witnesses or a lack of high-quality witnesses (for example,
witnesses who can be impeached due to prior criminal offenses). There
might also be evidence excluded due to police violations of the Fourth
Amendment, or confessions invalidated due to violations of the Fifth
Amendment. A third issue has to do with what the defendant might be
able to offer in a different case. Individuals with valuable testimony to
offer can exchange that testimony for reduced sentences even if they’re
caught red-handed. This, again, has nothing to do with the real offense
conduct at issue. Finally, isolating aggravating sentencing factors such as
prior offenses, use of a particular weapon, proximity to school (in the case
of drug dealing), etc., would be far too complex. I considered using
“wobblers”—California crimes which can be charged as felonies or
misdemeanors—but could not control for the above variables. If there
were a way to determine whether an offense should have been charged as
a felony or misdemeanor, one could obviously see how it was, in fact,
charged and determine over- or under-use of prison accordingly. But
asking how a wobbler should have been charged is, in fact, the question
we can’t answer. The point of this study is not to question the decisions
of individual DA’s, judges, or juries in individual cases, but to start to
explore the systematic differences that might explain why California
counties use prison at different rates.
To understand how counties use the state prison system, one could also
look at county NFA by offense category (e.g., San Bernardino sent X new
admissions to prison in 2006 on drug offenses).51 The state,
51

The state, does, however, publish commitment offense data for the prison
population as a whole. See, e.g., CDCR Population Reports, supra note X, at tbl. 9 for
the years 2000-2006 and tbl. 8 for the years 2007-2009. In 2009, for example, 55.5% of
men in the CDCR were imprisoned for crimes against persons (including homicide,
robbery, assault, kidnapping, and sex offenses including not only rape but other sex
offenses), 19.5% for property offenses, 17.0% for drug offenses, and 8.1% for other
crimes (including arson and escape).

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

22

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

unfortunately, does not make public its commitment offense data broken
down by county I have chosen to approximate the effect of crime through
the use of coverage rates, but a longitudinal study breaking down county
sentences per offense per year would be extremely useful.52 One could
look at the reported crimes for a given offense type and then see how
many of those offenses were actually covered by NFA. This would show
precisely what effects particular types of crime have on NFA. That said,
because all types of crime are lower in what I call “High Use” counties,
even without this data I conclude that there is no real issue about whether
crime is driving incarceration. The only outstanding issues are the precise
magnitude of the prison use, and the particular mechanisms by which it
takes place.
I used counties as the subdivision of the state primarily because there
are several county-wide elected officials instrumental in criminal justice.
County citizens elect sheriffs, DA’s, and judges, counties administer
parole, cities within counties elect the mayors who appoint police chiefs,
and juries are drawn from within counties. Perhaps a better way of
putting it is that California citizens outside their own counties have no say
in selecting another county’s sheriffs, judges, DA’s, or juries. Counties
are thus responsible for the overwhelming proportion of law enforcement
within their borders, the charges that are filed, the trials that take place,
and the jails or probation departments to which offenders might be sent.
California also publishes its crime data by county.
Nevertheless, I concede that different parts of counties can be different
from one another, and might have more in common with parts of
neighboring counties than they do with parts of their own counties.53
Counties can be a mix of rural and urban, for example, and this might
bear on the way crime manifests itself. Cities within counties also drive
their own policies, primarily through municipal police departments. Some
counties might have transient populations, or be victimized by criminals
52

I have advocated before on behalf of collecting and disseminating criminal justice
data in California. See E Pluribus Unum: Data and Operations Integration in the
California Criminal Justice System, 21 Stan. L. & Pol’y Rev. 277, 309 (2010)
(“Ultimately, giving citizens comprehensive, detailed information about the policies and
practices of criminal justice agencies can promote more well-informed decisions,
transparency about existing practices, and better civic discussions about the purpose of
criminal law.”).
53
For a fuller discussion, see id. at 294 (discussing shortcomings with the county as
the base unit for criminal justice).

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

23

who reside in neighboring counties. Even within a given county agency,
different parts of the county might have different approaches. Different
offices of a county DA might have different “going rates” for a given
crime, for example, particularly in a county as large as Los Angeles.
I look at rates, not numbers, for a variety of reasons. The primary
reason is the high degree of collinearity between population size and gross
amounts of violent crime and new felon admissions.54 That is, bigger
counties have more crime and more NFAs as a result of having more
people. Population size has nothing to do with NFA rate, however:
county population is not a reliable predictor of NFA rates normalized to
100,000 residents.
Comparing rates within a given year has the additional advantage of
isolating for year-to-year statutory and regulatory changes. Statutes—
albeit not their execution—are uniform across the state for every given
year, but they change from year to year. This study looks laterally from
county to county in a given year, not within a county across time. Yearto-year NFA rates, for example, would have to account for changes in the
penal code during the period studied. Proposition 36, for example, was
passed in 2000 and went into effect in 2001, and allowed for first- and
second-time nonviolent drug offenders to be diverted into treatment in lieu
of incarceration.55 This likely had some year-to-year effect on drug NFA.
I note that my conclusions are only as good as the reported data. I
take no position on how accurate the data is, and I note that the state has
expressed skepticism about particular counties’ data in particular years.56
I am unaware, however, of any systematic bias in the data. I note further
that the bias would have to operate for a particular county in a particular
direction on a multi-year basis in order to skew the results. That is,
54

Running a linear regression with gross (non-normalized) amounts of NFA as the
dependent variable and gross (non-normalized) amounts of county population, violent
crime, and property crime as the independent variables, the tolerance levels are between
.035 and .105, meaning that 89.5 percent or more of the variance of each predictor can
be explained by the other predictors. The variance inflation factors (VIFs) for each are
also high, ranging from 9.521 to 28.338. VIFs above 2 are considered problematic.
55
See California Proposition 36, http://www.prop36.org/.
56
See, e.g., Criminal Justice Statistics Center, “Data Characteristics and Known
Limitations,” available at http://stats.doj.ca.gov/cjsc_stats/prof09/limits.pdf. See also
Criminal Justice Statistics Center, Criminal Justice Trend Data Footnotes, available at
http://stats.doj.ca.gov/cjsc_stats/prof09/footnotes.pdf.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

24

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

Alameda would have to over report crime for ten years and/or San
Bernardino underreport it for the same duration, for example, in order to
skew my results.
Finally, there are the obvious limitations of statistical analysis itself
(and of my abilities). There is more than one way to analyze data, and
several tools to do so. My goal in this Article is to dispel the idea that
NFA are the necessary result of crime rates. While I believe that the data
provide potential insights, lack of a statistically significant correlation does
not mean that there is in fact no correlation given chance variability. The
analysis may also be altered by omitted variables.
C. Methods
This part explains the methods I used to subdivide California into four
groups on the basis of violent crime coverage and the calculated prison
subsidy: High Use counties, Low Use counties, Los Angeles County, and
Middle Use counties. The terms “high use,” “low use,” and “middle
use” are, of course, relative, given that there is no consensus on the “fair”
level of incarceration, and thus these counties might have more accurately
been described as “Higher Use” counties. For my purposes, “High Use”
meant a county that appeared in the top quartile more than 7 times in 10
years in either coverage rate or subsidy; “Low Use” meant a county that
appeared the same number of times in the bottom quartile of these
measurements.
Coverage, again, is the ratio of new felon admissions (NFA) to
reported violent crime, expressed as a percentage. I calculated the yearly
state average coverage rate for each of the ten years of the study (200009). I then calculated yearly coverage rates for each of California’s 58
counties. I expressed the county coverage rate as a percentage of the state
coverage rate, which gave me a relative measure of how much a given
county’s coverage exceeded or undercut the state rate for that given year.
Mathematically, the formula was
Relative Coverageyear = County Coverageyear /(State Coverageyear).
This had the benefit of controlling for year-to-year statewide differences in
crime rate, pinpointing which counties were relatively more carceral, not
which years were. I divided the results into quartiles. The top quartile

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

25

contained county coverage rates that were almost twice as large as that of
the state coverage rate for that year (199.75%); that is, in those years,
these counties sent almost twice as many people to prison per reported
violent crime as the state as a whole. Two counties appeared in the top
quartile all ten years: Kings and Sutter. Eight more appeared at least
seven times: Glenn and Trinity (eight years) and Butte, Colusa, Inyo,
Lake, Lassen, and Shasta Counties (seven). In the bottom quartile, six
counties had coverage rates less than or equal to 88.29% of the state
coverage rate in all ten years: Alameda, Contra Costa, San Francisco, San
Joaquin, and Santa Cruz. Eight more were in the bottom quartile at least
seven times: Marin (nine), Imperial (eight), and Alpine, Nevada,
Sacramento, San Benito, Sonoma, and Stanislaus (seven). I used all ten
years of data for any county listed above, even those with some yearly
data not in the top quartile. I did so because the purpose of this study is to
discover whether there is something inherent in these particular counties,
not to explore what might have happened in anomalous years.57
I then divided the state based on a calculated prison subsidy. The
ultimate focus of this Article is on the use of state prison resources.
Because small counties with high coverage rates nevertheless consume
very little of the state’s ten billion dollar prison budget, this measure
accounted for gross numbers of each county’s reported violent crime.
As with coverage, I looked at counties who appeared in the top or
bottom quartile more than seven times. Counties appearing in the top
subsidy quartile all ten years were Butte, Kern, Kings, Orange, Riverside,
San Bernardino, and Santa Clara; Fresno and Shasta were in the top
quartile nine times; Placer and Santa Barbara eight; Sutter seven.
Counties in the bottom quartile all ten years were Alameda, Contra Costa,
Sacramento, San Francisco, San Joaquin, and Santa Cruz; Imperial, Los
Angeles, and Marin were in it nine years; and Nevada, San Diego,
Sonoma, and Stanislaus were in it seven years. I included data from all
ten years for each county in the top and bottom group; San Diego actually
appeared twice in the top quartile for subsidies, which shows how these
figures are sensitive to small changes in coverage for counties with large
populations.58

57

A complete list of all counties is in Appendix A. I have mapped these counties in
Appendix B.
58
My estimates are below that of the LAO’s, so I may have understated the costs.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

26

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

Initial analysis revealed that both coverage and subsidy top and bottom
quartiles exhibited similar responsiveness to the key variables I will
analyze. I grouped them together in what I call the Low Use and High
Use groups respectively. High use counties, in other words, contain
counties with high coverage, high subsidies, or both. Low use counties
contain counties with low coverage, low (negative) subsidies (taxes), or
both. I will discuss general observations about these groups in the
following Section.
Because Los Angeles is such a large county, I decided to calculate
results for the Low Use group without it, even though Los Angeles had a
negative subsidy in nine of the ten years of the study. This also means
that the populations of the High and Low Use groups are relatively
similar—and relatively similar to that of Los Angeles—and thus that
contrasts between them can be more readily observed.59
This leaves twenty five other counties, with a combined average
population of 4.5 million, who did not appear more than seven times in
either the top or bottom quartile. While the bulk of my analysis will focus
on the other three segments of California, I note that this group is more
heterogeneous than would indicate. For example, Merced and Yolo are
both members of this group, have almost identical NFA numbers, yet
Yolo has much less violent crime (and property and Part I crime) than
Merced, giving it a much higher coverage rate. Several counties appeared
in the top quartile more than four times: Modoc and Yuba (six); Plumas,
Sierra, and Siskiyou (5); and Amador, Calaveras, and Tuolomne (4).
Only one county appeared more than four times in the bottom quartile:
Monterey (5).
Table 2: Demographics of the Four State Segments, Average
Yearly Values 2000-2009
High Use
Population
(millions)
APAR

Low Use

Los
Angeles

Middle
Use

State Total

11.74

10.17

10.07

4.53

36.51

7.62

6.81

6.55

3.00

23.98

59

Average total population from 2000-09 (adult population at risk in parentheses) for
Los Angeles was 10.1 million (6.6 million), Low Use was 10.2 million (6.8 million), and
High Use was 11.7 million (7.6 million).

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

27

Population
(millions)
55,079
37,023
54,187
17,612
164,000
Prison
Population
18
14
1
25
58
Number
of
Counties
Bold: highest value; Italics: lowest value. All figures based on ten
year averages.
All data was prepared in Excel. I then used SPSS to draw histograms
and scatterplots, using linear regression models and linear fit lines.
Syntax for my SPSS work has been posted to the folder with the rest of
my data.
II. VIOLENT CRIME RATES AND NEW FELON ADMISSION RATES
On a statewide level, although violent crime rates and NFA rates are
correlated, violent crime does not sufficiently explain why counties have
such disparate NFA rates. Why do counties respond to violent crime so
differently? Throughout this section, I will use the coverage variable as
my proxy variable for a county’s carceral response to violent crime.
I begin with a discussion of the statewide numbers, then examine High
Use counties, Low Use counties, Los Angeles County, and the rest of the
state.
A. The State
In this section, I will first demonstrate that some counties
systematically use prison at different rates. I will then look at whether
violent crime explains this differential usage at the statewide level.
First, counties send people to prison at different rates, even without
correcting for crime. Figure 2 plots NFA rates, normalized to 100,000
APAR. The chart looks at all 58 counties for all ten years of data, and
counts the number of instances counties reported a particular NFA rate.
Figure 2: Frequency of NFA Rates/100K APAR, 2000-2009

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

28

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

The shape of the histogram is relatively normal, although high NFA
counties skew the distribution right. If counties were in these ranges an
equal amount of time, the distribution would be normal as well. But, as
noted supra, certain counties appear consistently in the top and bottom
quartiles. Some counties consistently send people to prison at greater
rates than others.
But NFA only tells part of the story. NFA looks normal when
compared to population. NFA as a function of violent crime presents a
more chaotic picture. Figure 3 plots NFA rates and rates of reported
violent crime for all 58 counties and all 10 years.
Figure 3: Violent Crime to NFA (Rates, 100,000 APAR), 2000-2009

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

29

Although the relationship of the Violent Crime Rate per 100,000
APAR is statistically significant at the 1 percent level, it is not a
significant statistic. The amount of variance it explains is minute
(r^2=.032, which means changes in Violent Crime rates explain 3.2
percent of the variance in NFA rates), and the standard error is relatively
large (root mean squared error (RMSE) = 98.50139). What does this
mean? The scatterplot data shows that, although a linear fit line can be
drawn, the data do not cluster around it and the relationship is barely
above zero. In other words, if we were to use violent crime rates to
predict NFA rates at the statewide level, the amount it would predict
would be very small, and there is a chance that it would not be able to
predict even that small relationship with accuracy.
Just looking at NFA rates, the data are normal. Looking at NFA and
violent crime, it looks more chaotic. Segmenting the state will help
clarify whether there are patterns in the data, to see whether violent crime
might affect NFA in different segments of the state.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

30

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11
B. Violent Crime and NFA in the Four State Segments

Crime rates do not, themselves, explain why some segments have
higher NFA and higher total prison populations than others. High Use
counties have below-average crime, and Low Use counties have aboveaverage crime.
I will look at criminal justice statistics for each of the four segments
(High Use, Low Use, Los Angeles, and Middle Use) to see what, besides
levels of state prison usage, distinguishes them, in hopes of shedding light
on why each segment uses state prison resources at such different rates.
The analysis here will largely be descriptive, not predictive.
Table 3: Crime Rates and Prison Usage, Average Yearly Values, 20002009
High Use

Low Use

Los
Angeles

Middle
Use

State
Average

622.67
835.94
1,128.27
609.13
819.70
Violent
Crime
2,618.73
3,134.31
2,780.05
2,296.28
2,768.84
Property
Crime
5243.54
6404.46
5494.90
4881.71
5,596.56
Part I
Crime
223.57
122.04
211.87
167.99
184.58
NFA
35.90%
14.60%
18.78%
27.58%
22.52%
VC
Coverage
Rate
723.13
543.63
826.68
586.48
683.36
Total
Prison
Population
Bold: highest value; Italics: lowest value. All figures except VC
Coverage Rate are calculated per 100,000 APAR. State averages include
Los Angeles County.

High Use counties are not the group with the highest violent crime,
property crime, or total Part I crime rates. All three rates, in fact, are
below the state average. What’s more, both the Low Use counties and
Los Angeles have higher Violent, Property, and Part I crime rates while
maintaining lower NFA rates. Low Use NFA rates are nearly half those

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

31

of High Use counties, even though each measure of crime is more than
twenty percent higher. The Middle Use counties have the lowest crime
rates in all three categories, but still use prison at substantially higher rates
than the Low Use counties.
The chart also demonstrates the importance of choosing what to
measure. Los Angeles has a significant violent crime problem, but
property and Part I crime rates are at the state average. Los Angeles
incarcerates at an NFA rate lower than that of the High Use counties, but
its total prison population rate is the highest of the four segments. For my
purposes, because Los Angeles has a significant violent crime problem, its
coverage rate is half of High Use counties. But on alternative measures of
crime, such as property and Part I, Los Angeles is at the state average, so
its above-average NFA rate expressed in terms of Property Crime
Coverage or Part I Coverage would be unjustified.
It is also not the case that NFA differences are explained by the type of
violent crime a county experiences. As seen in Table 4, rates of all four
categories of violent crime are below the state average in High Use
counties. More importantly, the numbers of more serious crimes are not
large enough to drive differences in NFA. There simply just aren’t that
many rapes and homicides to account for the difference, even if High Use
counties had a 100 percent clearance rate on those crimes.
Table 4: Average Yearly NFA and Violent Crime Rates, by
Offense, 2000-2009
High Use
NFA
Homicide

Low Use

Los
Angeles

Middle
Use

State
Average

223.57

122.04

211.87

167.99

184.58

6.81

8.66

14.91

6.54

9.51

37.41
42.36
37.85
41.34
39.43
Forcible
Rape
173.83
293.87
423.43
139.60
271.84
Robbery
404.62
491.05
652.07
421.65
498.92
Aggravated
Assault
Bold: highest value; Italics: lowest value. All figures except VC
Coverage Rate are calculated per 100,000 APAR. State averages include
Los Angeles County.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

32

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

I will now discuss each of the segments of the state in greater detail,
describing how they are different and what impact each has on the overall
state prison population.
1. High Use Counties: Dominated by the Subsidized
The High Use counties are made up of three more or less equal
numbers of counties: those in the top quartile of coverage, those in the top
quartile of subsidy, and those who were in both. However, though the
numbers of counties are similar, their populations are not. The large
counties in the subsidy group are the overwhelming source of this
segment’s population, and will get the majority of the analysis.
Table 5: High Use Counties, Average Yearly Values, 2000-2009
High
Coverage

High
Subsidy

Both

High Use
Total

High Use
as % of
State Total

.18
10.94
.62
11.74
32.16%
Population
(millions)
6
8
4
18
31.03%
Number of
Counties
1,085
49,391
4,603
55,079
33.60%
Prison
Population
271.68
215.75
344.10
223.57
38.46%
NFA
626.19
598.84
622.67
24.12%
497.50
Violent
Crime
54.61%
34.45%
57.46%
35.90%
N/A
Coverage
Rate
3,882.92
5,274.60
5,109.53
5,243.54
29.75%
Part I
Crime
Rate
Bold: highest value; Italics: lowest value. NFA,Violent Crime, and
Part I Crime Rate are calculated per 100,000 APAR off 10 year averages.
State averages include Los Angeles County.

Counties in the “high coverage” and “both” groups are generally too
small to make much of a difference statewide. High coverage counties in
particular are not populated enough to make much of an impact on the

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

33

state’s prison population or on its bottom line. The counties with both
high coverage and high subsidies are also small, but they incarcerate at
such high coverage rates that they nevertheless manage to make it into the
top quartile of subsidy. NFA rates for the “both” group are nearly twice
that of the state average (344.1 versus 184.58), even though violent crime
rates are just three-quarters of the state rate (598.84 versus 819.70). This
yields a coverage rate more than twice that of the state average (57.46%
versus 22.52%). These counties are so far out of step with the rest of the
state that despite having just over 620,000 people, their average group
subsidy is almost thirty million dollars.60
The “subsidy” group is relatively tame by comparison, incarcerating at
a coverage rate only fifty percent more than the state average. In fact,
looking at NFA rates alone (184.58 for the state, 215.75 for the subsidy
group), the subsidy group doesn’t appear to be so unusual. But these
NFA figures are higher despite the fact that the justifications for prison—
crime rates—are below the state average in all three major categories:
violent crime, property crime (not pictured), and Part I crime. Again, this
underscores the fundamental difference between looking at prison usage
alone—i.e., NFA rates—and tying that usage to its justification. Looking
at rates based on population alone can, in some cases, obscure the fact that
a county lacks a crime-based justification for the level of incarceration it
uses.
2. High Subsidy Revisited: The Rich Four and the Poor Four
The high subsidy counties can be further divided on the basis of
income. They divide neatly into two groups of four counties, both with
roughly the same population. The “Rich Four” counties are Orange,
Placer, Santa Barbara, and Santa Clara. Three of these counties reported
incomes above the state per capita average each of the ten years in the
study, and one of them (Santa Barbara) was above the state average seven
times (missing by less than $617 the other three times). The “Poor Four”
counties are Fresno, Kern, Riverside, and San Bernardino. Each of these
counties reported incomes below the state per capita average for all ten
years, and none got any closer than $8,000 below the state average in any
of those years.
Table 6: The Rich Four and the Poor Four, Average Yearly
60

See Table 17 and accompanying text, infra.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

34

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11
Values, 2000-2009
Rich Four
5.52

Poor Four

State Total

5.42

36.51
Population
(millions)
3.68
3.40
23.98
APAR
Population
(millions)
164,000
17,280
32,111
Prison
Population
149.31
287.46
184.58
NFA
442.79
824.17
819.70
Violent
Crime
Rate
1,887.55
3,447.97
2,768.84
Property
Crime
Rate
3,967.13
6,686.01
5,596.56
Part I
Crime
Rate
33.72%
34.88%
22.52%
Coverage
Bold: highest value; Italics: lowest value. All figures based on ten
year averages.

This chart again reveals how coverage changes the analysis. The rich
and Poor Four have dramatically different NFA rates, but because they
also dramatically different violent crime rates, their coverage rates are
very similar. If one looked only at NFA rates per 100,000 APAR, the
Rich Four appear to use very little prison, with an NFA around nineteen
percent below the state average. The problem is that the Rich Four’s
violent crime rate is approximately 46% below the state average. The
Rich Four incarcerate less than the state, but not as much as their crime
rate indicates. Over-use is relative, and using less can be using more if
your crime rate is low.
The Poor Four, on the other hand, have violent crime rates slightly
above the state average, but their NFA rate is more than 50% greater than
the state’s NFA. They are justified in incarcerating at a higher rate, but
not to the extent that they are. Again, looking at NFA rates themselves
obscures the fact that violent crime is not driving these rates.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

35

The Rich Four and the Poor Four are a drain on the rest of the state.
To the extent that these counties are being subsidized for prison usage that
cannot be explained by violent crime, the Rich Four cannot justify their
subsidy on the basis of need. It would be difficult for them to argue that
that they are due a larger share of the state prison budget either because
they cannot afford it (they can) or because crime demands that they do so
(they are relatively safe). The Poor Four do not have the same resources
as the Rich Four, but they consume many more prison resources than the
Rich Four, and more than the state coverage rate would indicate. To the
extent the state needs to focus on overcrowding, however, these are the
counties that incarcerate at a high rate and in large numbers.
3. Low Use Counties: The Convergence of Low Coverage and Low
Subsidy
Low Use counties are clustered in the “both” category of both low
coverage and low subsidy. The eleven counties with most of the
population also have the lowest coverage rates, which means they have the
lowest subsidy. I use the word subsidy only for consistency—based on
violent crime rates, all members of the Low Use group are negatively
subsidized. That is, they pay a very substantial prison tax.
Table 7: Low Use Counties, Average Yearly Values, 2000-2009
Low
Coverage
Population
(millions)
Number of
Counties
Prison
Population
NFA Rate
Violent
Crime Rate
Coverage
Rate
Part I

Low
Both
Subsidy
(minus LA)

Low Use
Total

Low Use
as % of
State Total

.06

3.03

7.08

10.17

27.85%

2

1

11

14

24.14%

1,262

12,713

24,183

37,023

22.59%

124.53
648.35

140.93
671.93

113.89
908.08

122.04
835.94

18.77%
28.96%

19.21%

20.97%

12.54%

14.60%

N/A

4,340.26

5,051.86

7,003.53

6404.46

32.49%

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

36

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

Crime Rate
Bold: highest value; Italics: lowest value. NFA,Violent Crime, and
Part I Crime Rate are calculated per 100,000 APAR off 10 year averages.
State averages include Los Angeles County.
As stated earlier, Los Angeles was in the subsidy bottom quartile but
is being excluded for other reasons, leaving this group with only San
Diego in the subsidy category. San Diego has violent and Part I crime
rates well below the state average, and a coverage rate around 1.5%
below the state average. Because it is a large county, however, small
changes in coverage result in large changes to the calculated subsidy. In
fact, San Diego appeared in the top quartile for subsidy twice, but because
it was in the top quartile seven years, I included it in the Low Use list.
The low coverage counties, Alpine and San Benito, are too small to
deserve much comment.
The rest of the counties in the group are relatively large. The “both”
counties have a coverage rate a third that of the High Use counties. These
counties have violent crime and Part I crime rates well above the state
average, with an NFA just two-thirds of the state average. In these
counties, then, consisting of twenty percent of the state’s population,
higher crime rates are associated with lower prison use.
4. Low Coverage and Subsidy Divided by Income: The High Five and
the Low Six
These counties can also be divided into relatively equal populations on
the basis of income, but they do not divide as neatly. Including only the
counties below the state per capita income level in all ten years (Imperial,
Sacramento, San Joaquin, Stanislaus) would have resulted in too unequal a
division of population, so I added two counties with the next lowest
incomes (Nevada and Sonoma). The richer 5 counties are Alameda,
Contra Costa, Marin, San Francisco, and Santa Cruz.
Table 8: Dividing Low Coverage, Low Subsidy Counties, Average
Yearly Values, 2000-2009
Low Six
Population
(millions)

3.23

High Five
3.85

State Total
36.51

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

37

2.09
2.64
23.98
APAR
Population
(millions)
14,797
9,386
164,000
Prison
Population
158.09
78.96
184.58
NFA
961.07
866.2
819.70
Violent
Crime
Rate
3744.92
2952.36
2,768.84
Property
Crime
Rate
7,521.39
6,594.29
5,596.56
Part I
Crime
Rate
16.45%
9.12%
22.52%
Coverage
Bold: highest value; Italics: lowest value. All figures based on ten
year averages.

Table 8 summarizes the differences between the two groups. I note,
however, that the distribution of crime among these counties does not
track income group. Both sets of crime rates are above the state average,
and they are more or less equally distributed on either side: Marin (rich)
and Nevada (poor) have violent crime rates in the 300’s, Alameda, San
Francisco (rich), Sacramento, San Joaquin, and Stanislaus (poor) have
violent crime rates above 1000, and Contra Costa, Santa Cruz (rich),
Imperial, and Sonoma (poor) are in the 500 and 600’s. Coverage rates are
generally lower in the high income areas, however, as are NFA.
5. Los Angeles
Los Angeles County is atypically large, accounting for slightly less
than a third of the state’s population and about a third of its prison
population, but its prison usage is not atypically high when its high violent
crime rate is factored into account. On a per capita basis, LA’s NFA rate
is higher than the state average. However, its violent crime rate is almost
fifty percent greater than the state average. The coverage variable
expresses this relationship more simply: LA’s coverage rate is less than
the state average, and about half that of the High Use counties. LA does
have below average property and Part I crime rates, however; an analysis

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

38

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

that did not center on violent crime might conclude that LA’s prison usage
is not negatively subsidized.
Figure 4: Los Angeles County and the Rest of the State, 2000-2009
Ratio of Los Angeles County Values to Rest of California (1 Denotes Equal Values)
0.80

Los Angeles As Multiple of Rest of State

0.70

0.60

0.50

Total Population
Violent Crimes Total
Property Crimes Total

0.40

Part I Crimes Total
Prison Population
New Felon Admissions

0.30

0.20

0.10

0.00
2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Year

The above chart summarizes LA’s relationship to the rest of the state
graphically; LA comes in generally at about 40-50% of the rest of the
state numbers, except for violent crime in the early part of the past
decade.
6. Middle Use Counties
The population of these counties is small to medium range, ranging
from tiny Sierra County to relatively populous San Mateo and Ventura
counties. Yearly coverage rates bounce around, reaching lows of about
1/3 the state coverage rate and highs several times the state rate. Annual
NFA rates range from less than 100 to more than 400 in particular years.
These counties, though, were ones that might have particular years—or
even several years—of High or Low Use that nevertheless did not exhibit
the kind of consistency (seven of ten years) required for inclusion into
either group.
III. ALTERNATIVE EXPLANATIONS

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

39

In this section, I discuss what else besides violent crime could be
causing these disparities. Because I have already looked at property crime
and Part I crime (the general crime rate), and because other types of crime
(notably sex offenses) are so rare that they could not account for the
disparity, I will briefly examine drug offenses. I then look at law
enforcement, using general arrest data as a crude proxy for how active a
force is, to try to explore whether high coverage is simply a matter of
more effective law enforcement. I next look to in-county dispositions—
jail and probation—to see if differential usage of these resources explains
differences in prison usage. I next look at local resources—using per
capita income as a proxy—as a means of exploring whether counties rely
on prison because they do not have the money to do anything else. I
examine the role of politics by analyzing voter registration numbers, to see
if party politics or levels of participation might explain what’s different
about different segments of the state.
From time to time, I will discuss state segments as they bear on the
variables in question. These factors will not operate similarly across
counties—California is a huge, diverse place. The principal statistical
inquiry was, of course, whether violent crimes explain differences in
prison usage. This part simply tries to shed some light on what might
explain differences in usage, although it should be seen only as a very
preliminary investigation.
A. Arrest Data
As we have seen, differences in property crime and Part I crime rates
inadequately explain differences in NFA rates. In this section, I look at
other types of crimes—notably drug crimes—for a possible explanation.
As stated earlier, drug crimes themselves are not reported. Therefore I
will use drug arrests as a very crude measurement of actual drug crimes.
There are obvious problems with this method, because arrests are never a
complete—or accurate—measure of any criminal activity, but drugs are
such a big part of the prison system, I believe it’s better to attempt an
explanation than to leave it undiscussed. This analysis, however, should
be taken even more provisionally than the rest of the Article.
Arrest data might also be used as a proxy for law enforcement activity
and effectiveness, or for differences in policing strategies. One might
associate higher arrest rates with broken windows style policing, or

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

40

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

perhaps lower rates with a less active (or more cautious?) force. Without
getting into the merits of different policing strategies, this section analyzes
whether policing inputs could explain differences in NFA.
Table 9: Arrest Data, Average Yearly Values, 2000-2009
High Use
NFA
Total Offense
Arrests
Violent
Offense
Arrests
Coverage of
Reported
Violent
Crimes
Violent
Crime
Clearance
Rate
Property
Offense
Arrests
Property
Coverage
Sex Offense
Arrests
Drug Arrests

Low Use

Los
Angeles

Middle
Use

State
Average

223.57

122.04

211.87

167.99

184.58

1,802.26

1,864.45

1,858.76

1,730.58

1,826.38

458.77

452.07

502.76

461.47

469.22

73.68%

54.08%

44.56%

75.76%

57.24%

47.36%

41.69%

44.81%

50.46%

45.05%

472.42

481.17

473.75

418.01

468.46

18.04%

15.35%

17.04%

18.20%

16.92%

36.06

28.27

26.93

37.00

31.47

532.34

565.52

592.97

473.87

551.01

354.15
260.34
220.65
301.69
284.46
Dangerous
Drugs
Arrests
69.33
63.19
80.52
69.94
70.72
Weapons
Arrests
Bold: highest value; Italics: lowest value. All figures except
percentages are calculated per 100,000 APAR. State averages include Los
Angeles County.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

41

Arrest data reveal almost no significant differences across the four
segments for total arrests, property arrests, sex offenses, drug arrests, and
weapons arrests. In addition, sex offenses are too small to make much of
a difference in NFA rates. However, two areas which might merit closer
study are dangerous drug arrests per 100,000 APAR and the ratio of
violent arrests to violent crimes (the violent arrest coverage rate). Both
are much higher in High Use counties than in Los Angeles County or the
Low Use counties. Higher Dangerous Drug arrests might suggest that the
severity, if not the number, of drug crimes might be worse in High Use
areas. The high ratio of violent crime arrests to violent crime might
suggest that violent crimes are policed more aggressively in High Use
counties, leading to more prosecutions and more prison time. High Use
clearance rates are, in fact, higher than in Los Angeles or the Low Use
counties, but the difference between High and Low clearance rates is not
nearly as large as the difference between High and Low Arrest Coverage.
Even taking this as true, however, and assuming that High Use
counties devote more energy and resources towards fighting crime—and
do so more effectively—it is still the case that responding to violent crime
aggressively is a policy response to violent crime, not a function of it.
Accordingly, this policy, as with all other good policies, is subject to the
key question: why should the state pay for it? If it is good policy, 61 after
all, the county should happily make the investment itself. It is the one
who made the choice to deal with crime in this fashion. The issue is not
whether the policies in question are good or bad. The issue is why the
state should pay for something it has no control over, a policy that benefits
a readily identifiable subset of the population who, in fact, drew up and
implemented the policy. Even if we were to think the state should be
subsidizing these kinds of choices, there remains another question: why
subsidize these counties and not others, and these policies and not others?
Or is the state willing and able to subsidize all counties who wish to make
this choice?
B. Local Dispositions
Do offenders go elsewhere in the system if they do not go to prison?
Or does a Low Use county just use fewer carceral resources across the
board—both state and local? One might expect that, on a zero sum view
61

If it is, in fact, a policy and not either random or inadvertent.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

42

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

of offender management, lower use of prison would result in higher use of
jail and probation. That is not the case, however. High use counties use
jails at higher rates than Low Use counties, suggesting that High Use
counties are simply more punitive, using incarceration at a higher rate
irrespective of whether the county or state pays for it. As for probation,
there is almost no difference between Low Use and High Use counties
along any of the dimensions examined, a surprising result which might be
the result of how the probation data are reported.
1. Jail
The issue of local jails and their ability to absorb offenders from state
prisons was given an excellent, comprehensive analysis by Mike Males in
his recent paper, Can California County Jails Absorb Low-Level State
Prisoners?62 Males looked at county jail capacities and county offender
mixes to estimate whether county jails could absorb the numbers of lowlevel offenders most likely to be returned to them under the pending
realignment plan, concluding that county jails “can provide beds for only
around 38% of the 15,400 low-level, non-strike property and drug
convicts now held in state prisons.”63
Males’s study, unfortunately, only has data from one year (2009), so I
was unable to incorporate his findings fully. I use, instead, figures about
jail numbers and jail budgets. I also look at percentages of jail residents
who are sentenced and not sentenced. Non-sentenced residents can be
those too dangerous to be released before trial, those unable to post bail,
or those awaiting processing. Because Department of Justice expenditure
data64 is based on a fiscal year that goes from July 1 to June 30,65 I
averaged adjacent years to calculate an estimated yearly total. That is,
figures for 2000 are the average of 1999-2000 and 2000-2001. There are
no police expenditures for Alpine County; the sheriff provides the county

62

(March 2011,) available at
http://www.cjcj.org/files/Can_California_County_Jails_Absorb_LowLevel_State_Prisoners.pdf.
63
Id. at 4.
64
The entry page for the Criminal Justice Statistics Center’s county crimes data is
available at http://ag.ca.gov/cjsc/statisticsdatatabs/ExpenCo.php. Individual county data
was taken by following links to each county.
65
See, e.g., Criminal Justice Statistics Center, Criminal Justice Trend Data
Footnotes at 4, available at http://stats.doj.ca.gov/cjsc_stats/prof09/footnotes.pdf.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

43

with all of its law enforcement.66 These figures extend only to 2007.
Table 10: Jail Statistics, Average Yearly Values, 2000-2009
High Use

Low Use

Los
Angeles

Middle
Use

State Total

323.30
282.49
353.17
363.96
Jail
Population
121.14
115.31
92.54
137.33
Sentenced
242.83
207.99
189.95
215.87
Non
Sentenced
35.67%
32.76%
38.89%
33.28%
%
Sentenced
14.99%
16.44%
10.54%
17.64%
% County
CJ Budget
Spent on
Jail
Bold: highest value; Italics: lowest value. All figures except Budget
are calculated per 100,000 APAR; Budget figures through 2007 only.
County criminal justice budget is the sum of probation, jail, and law
enforcement budgets.

328.80
113.70
215.11
34.58%
14.18%

The jail numbers do not support the theory that Low Use counties are
sentencing their offenders to jail rather than prison. Jail use is higher in
both High and Middle Use counties than in Los Angeles and the Low Use
counties. This tends to support the theory that High Use counties use
more of all forms of incarceration, not just those subsidized by the state.
These differences, however, are not nearly as stark as those involving
NFA. What these population figures do not account for, however, is how
crowded jails are, and whether these populations are near the jail’s
capacity. Males did not adopt my violent crime coverage methodology,
nor did he group counties by prison use. However, looking at his list of
counties with insufficient space to absorb low-level state prisoners, we see
that all of the Rich Four and 3 of the four Poor Four (Kern, Riverside,
and San Bernardino) are rated as having insufficient unused jail capacity to
absorb returning prisoners.67 On the Low Use side, focusing only on the
combined low coverage/low subsidy group, only those counties with
incomes below the average state per capita income in all four years of the
66
67

Email on file with author.
Id. at 3.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

44

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

study (Imperial, Sacramento, San Joaquin, and Stanislaus) have
insufficient jail space. The other seven counties have sufficient jail space.
Finally, I note that almost 2/3 of jail populations are non-sentenced,
which is in line with the national average.68 U.S. Attorney General Eric
Holder recently remarked that "Almost all of these individuals could be
released and supervised in their communities—and allowed to pursue or
maintain employment, and participate in educational opportunities and
their normal family lives—without risk of endangering their fellow citizens
or fleeing from justice."69 The problem is that many non-sentenced
offenders cannot make bail; Holder suggested, instead, that they be
released on their own recognizance. The numbers suggest that at least a
preliminary exploration of this alternative is warranted.
2. Probation
Counties use probation in dramatically different ways, and an entire
Article could be devoted to the ways in which statewide statistics obscure
real local trends. Statewide figures on total probation caseloads indicate
that use statewide has not changed—but several counties within the period
of study have moved dramatically in non-random ways, expanding in
some counties and contracting in others.70 Moreover, some counties
dramatically changed the way they use probation. To cite just a few
examples, in Riverside County, total caseload almost doubled from 2000
to 2009, and new admissions more than doubled. In Santa Clara, new
admissions (both total and felony only) almost doubled, but total caseload
decreased around forty percent. In Orange, total caseload was also almost
cut in half, but new admissions for felons stayed roughly the same.
Probation might be one area in which county policies show real year-toyear variations, and it is certainly deserving of a much closer analysis than
I give it here.
Table 11: Probation Use by Segment, Average Yearly Values, 2000-2009
High Use

Low Use

Los

Middle

68

State Total

Eric Tucker, Holder: Petty Offenders Should Await Trial at Home, San Jose
Mercury
News,
June
1,
2011,
available
at
http://www.mercurynews.com/natbreakingnews/ci_18184095?nclick_check=1.
69
Id.
70
Note that some probation data is missing. See supra at note 14.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

45

Angeles
Use
937.23
2,245.17
1,411.40
1,461.59
1,444.00
Probation
Caseload
$6,933
$7,074
$7,137
$7,355
$7,082
Probation
Budget
76.51%
79.31%
58.90%
75.66%
% Felony
86.68%
363.77
1,070.32
750.69
904.66
809.95
Probation
New
Admissions
73.99%
78.40%
88.93%
60.09%
74.84%
% Felony
12.34%
12.10%
9.38%
13.46%
11.40%
% County CJ
Budget Spent
on Probation
Bold: highest value; Italics: lowest value. All figures except
percentage figures are calculated per 100,000 APAR. All calculations
made on ten-year averages except budget figures, which are through 2007
only. County criminal justice budget is the sum of probation, jail, and law
enforcement budgets.
These numbers are, frankly, surprising. Some of the data is not
complete, and probation data are limited to “original grants of probation
and do not include subsequent grants of probation to those already under
supervised probation in the same county.”71 It is unclear, though, how the
results obtained could be fully explained by this. I am reluctant to draw
any conclusions of my own from Table 11, but will instead point out areas
that require explanation. Probation budgets are almost identical, Low Use
and High Use counties have similar caseloads and felony populations. Los
Angeles has fewer total probation cases and dramatically lower new
admissions, suggesting perhaps that probation in Los Angeles County is
longer-term than in High and Low Use counties. I am unsure whether
there is a quality versus quantity story to be told here, or why both
probation and jail use is higher in High Use counties. This might also be
one area where individual counties behave so differently within segments
that patterns are not readily discernible.
C. Local Resources
I measured local resources by looking at per capita income: both per
71

Criminal Justice Statistics Center, “Data Characteristics and Known Limitiations,”
3, available at http://stats.doj.ca.gov/cjsc_stats/prof09/limits.pdf.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

46

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

capita income itself and the difference from the state per capita income. I
note first that I chose not to look at gross population size of a county as a
measure of resources. The relationship of Total Population to NFA rates
is not statistically significant at the 1 percent level (p = .089), the amount
that Total Population explains in NFA rates is small (r^2=.005, which
means changes in Total Population explain less than .5 percent of the
variance in NFA rates), and the standard error is relatively large (root
mean squared error (RMSE) = 63.35824. RMSE is a guide to how
closely the data fit the trend line).
My figures for per capita income come from the Bureau of Economic
Analysis.72 These figures do not account for income inequality within a
given county, which might be relevant in explaining crime and responses
to crime, particularly where property crimes are concerned. I take per
capita income as a measure of resources independent of criminal justice
budgets. I note also that state criminal justice funding is not necessarily
related to per capita income, where poorer counties get more resources.73
Some funding comes from a county’s share of state sales tax revenues,74 so
we might even expect more criminal justice resources in wealthier
counties.
Table 12: Per Capita Income by Segment, Average Yearly Values,
2000-2009
High Use
Mean Per
Capita
Income
Max
Min
Standard
Deviation

Low Use

Los
Angeles

Middle
Use

State Total

$36,893

$42,611

$36,198

$38,490

$38,304

$60,038
$16,920
9,013

$93,263
$18,973
15,103

$42,195
$29,865
4,313

$72,576
$18,542
9,615

$43,853
$33,398
3,968

72

U.S. Dept’t of Commerce, Bureau of Economic Analysis, Local Area Personal
Income, available at http://www.bea.gov/regional/reis/default.cfm?selTable=CA13&section=2. (Selected “Per Capita Personal Income,” “California” and the years
2000-2009.)
73
See footnotes 2-9, supra and accompanying text. As a reminder, state funding is
also not related to crime rates, either. Id.
74
See supra note 3.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)
-$1,614
$4,176
-$2,195
$64
Mean
Difference
From State
Average
$49,410
$29,336
$22,161
-$1,528
Max
Difference
-$16,498
-$18,104
-$17,864
-$3,533
Min
Difference
14,437
8,831
8,374
523
Standard
Deviation
Figures based on county per capita income numbers and were
weighted based on county population.

47
N/A

N/A
N/A
N/A

Generally, Low Use counties have higher per capita incomes,
approximately six thousand dollars higher than High Use counties and Los
Angeles.75 All income figures across all segments, however, had a great
deal of variation—and the richest group, Low Use, had the highest
coefficient of variation. The gap between the highest and lowest reported
county incomes for all three segments besides Los Angeles was at least
$40,000, and these same segments had reports of incomes more than
$15,000 below the state per capita income level in a given year and
incomes more than $20,000 above it. Income merits further study; a
project which divided the state into income segments might reveal further
insights about the relationship between income levels and prison usage.
Table 13: Per Capita Income of High Use Counties, Average
Yearly Values, 2000-2009
High
Coverage

Mean Per
Capita
Income

$27,089

High
Subsidy

$37,567

Rich Four

$47,484

Poor Four

$27,481

75

High
Coverage
and
Subsidy
$27,872

These numbers were calculated in order to account for county population size. I
took per capita income in a given county for a given year, then multiplied that number by
the county’s population that year. I added these figures for a given segment of the state,
then divided by total population for that segment. Figures were not adjusted for
inflation.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

48

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

$40,721
$60,038
$60,038
$31,111
Max
$18,021
$21,517
$33,307
$21,517
Min
-$945
-$11,118
-$11,319
$9,057
Mean
Difference
From State
Average
$22,161
-$8,123
-$1,674
$22,161
Max
Difference
-$13,828
-$13,828
-$17,864
-$617
Min
Difference
Figures based on county per capita income numbers and were
weighted by county population.

$34,432
$16,920
-$10,593

-$6,699
-$17,455

In Table 13, we see that there is a sharp divide between the Rich Four
and the Poor Four. Three of the Rich Four counties were above the
average state per capita income every year in the study; Santa Barbara was
below it during only three years, and even then missed it by no more than
$617. The Poor Four, however, were at least $8,000 below the average
state per capita income level every single year. The best a poor county
did relative to the state average was still more than $7,000 less than the
worst a rich county did—and almost thirty thousand dollars less than the
highest Rich Four figure. The mean difference between the two groups
was approximately $20,000 a year. The Rich Four are, in fact, the only
above-average income group of High Use counties—neither high coverage
nor high coverage/high subsidy counties ever broke above the state
average per capita income level for even a single year. I should note,
again, that the Rich Four have large total populations, with around fifty
percent of the High Use segment’s population. I also note that none of
these figures accounts for income differences within a county; counties
undoubtedly have richer and poorer areas.
Table 14: Per Capita Income of Low Use Counties, Average Yearly
Values, 2000-2009
Low Six
Mean Per
Capita
Income
Max

High Five

Low Use Without San
Francisco and Marin

$33,086

$52,295

$39,800

$47,813

$93,263

$58,228

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

49

$18,973
$39,013
$18,973
Min
-$5,395
$13,900
$1,360
Mean
Difference
From State
Average
$4,573
$49,410
$14,375
Max
Difference
-$14,425
$4,880
-14,425
Min
Difference
Figures based on county per capita income numbers and were
weighted by county population.

Low Use counties, again, do not divide as easily as High Use ones.
This chart looks only at the group of eleven counties with both low
coverage and low subsidies, and excludes Alpine, San Benito, and San
Diego (as well as Los Angeles, as stated earlier). There are four Low Use
counties which never had incomes above the state per capita average
during any year of the study, but an even division of this segment by
population adds two counties with above-average incomes. The mean
difference between the two groups is nearly $20,000, but this segment is
made up mostly of average counties with two outliers: Marin County and
San Francisco. Recalculating the mean per capita income of the segment
without Marin County and San Francisco gives a mean per capita income
of $39,800, approximately $1500 higher than the state average for this
period. While this number is still above the state average, and still above
that of the other three segments, it is lower than the mean income of the
Rich Four.
D. Politics
I looked at voter registration numbers for my political analysis. Voter
registration data came from the California Secretary of State.76 I used the
76

The entry page for the Voter Registration and Participation Statistics is available at
http://www.sos.ca.gov/elections/elections_u.htm. Specifically, I used these particular
reports: Report of Registration as of February 7, 2000, available at
http://www.sos.ca.gov/elections/ror/ror-pages/29day-presprim-00/county.pdf; Report of
Registration
as
of
February
10,
2001,
available
at
http://www.sos.ca.gov/elections/ror/ror-pages/ror-odd-year-01/county.pdf
[NOTE:
Sierra County is reported as having more than 100 percent of its population registered to
vote]; Report of Registration as of
February 4, 2002, available at
http://www.sos.ca.gov/elections/ror/ror-pages/29day-prim-02/county.xls; February 10,

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

50

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

date closest to February for years with multiple reports; this is because
odd-numbered years only have a single registration report, which comes
out in February. I collected percentage data on total registration,
Democratic and Republican registration, and declined to state (as a proxy
for swing voters). I calculated third party registration by taking these
three numbers and subtracting them from 100; this procedure, admittedly,
amalgamates third parties of very different political stripes and should be
read as a measure of anti-two-party sentiment rather than, say, a measure
of Green or Libertarian sentiments. I then calculated the political valence
of a county by subtracting the percentage of Republicans from the
percentage of Democrats, yielding positive numbers for Democratic
majorities and negative numbers for Republican majorities.
I used registration data, rather than actual voting patterns, for a
number of reasons. First, I was wary of including data from actual races
out of the concern that individual candidates and/or issues might shift
turnout one way or another. Second, the data are less readily available.
Registration figures might be seen as a general measure of civic
engagement, and a baseline for individual attitudes, although I
acknowledge that there are a variety of opinions expressed on crime within
parties, and that party affiliation is in no way a guarantee of left/right
tendencies or particular attitudes about crime.77

2003 - Report of Registration, available at http://www.sos.ca.gov/elections/ror/rorpages/ror-odd-year-03/county.xls; February 17, 2004 - Report of Registration, available
at
http://www.sos.ca.gov/elections/ror/ror-pages/15day-presprim-04/county.xls;
February
10,
2005
Report
of
Registration,
available
at
http://www.sos.ca.gov/elections/ror/ror-pages/ror-odd-year-05/county.xls; January 3,
2006 - Report of Registration, available at http://www.sos.ca.gov/elections/ror/rorpages/154day-prim-06/county.xls; February 10, 2007 - Report of Registration, available
at http://www.sos.ca.gov/elections/ror/ror-pages/ror-odd-year-07/county.xls; January 22,
2008 - Report of Registration, available at http://www.sos.ca.gov/elections/ror/rorpages/15day-presprim-08/county.xls; February 10, 2009 Report of Registration, available
at http://www.sos.ca.gov/elections/ror/ror-pages/ror-odd-year-09/county.xls.
77
Of course, it sometimes does indicate something useful. The AB 109 vote in the
California assembly, for example, was almost entirely on party lines, with all but one
Democrat voting yes, and no Republican voting yes (one member was absent or
abstained).
See Vote Information, AB 109 Assembly Bill, available at
http://www.leginfo.ca.gov/pub/11-12/bill/asm/ab_01010150/ab_109_vote_20110317_0532PM_asm_floor.html. Party affiliations were obtained
at the official party websites for the California Assembly. The Democratic site is
available at http://asmdc.org/members/democratic-members, the Republican site is
available at http://republican.assembly.ca.gov/?p=members.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

51

I also calculated my figures without correcting for population. I did so
because I wanted to evaluate the party identity of a county’s political
leadership. In other words, this method simulates the electoral college
model, where all that matters is who finishes first, not the popular vote
model, where the margin of victory also matters.78 My state figures are
calculated means for the group of 580 counties.79
Table 15: Voter Registration by Segment, Average Yearly Values,
2000-2009
High Use

Low Use

Los
Angeles

Middle
Use

70.73%
73.95%
73.88%
70.97%
%
Registration
51.51%
39.27%
36.79%
46.24%
%
Democrats
43.62%
26.88%
31.16%
39.63%
%
Republicans
-6.83%
-0.36%
15.08%
24.63%
Democrats
Minus
Republicans
14.66%
17.18%
17.04%
15.48%
Decline to
State
4.93%
5.42%
4.57%
5.63%
Third Party
Bold: highest value; Italics: lowest value. All figures are not
corrected for population; they are means of the group of counties for
2000-09.

State Total
72.88%
40.39%
38.60%

1.79%

15.66%

All segments showed high rates of voter registration, with a bit
more registration in Low Use counties. High Use counties had more
registered Republicans than other segments of the state, as well as greater
numbers of Republicans versus Democrats. This might suggest that
higher coverage is more associated with Republican politics. I should
78

Consider this thought experiment. If Los Angeles were 99% Democratic and
every other county were 51% Republican, popular (population-adjusted) registration
numbers would indicate a heavy advantage for Democrats, even though county policies
would be under the direction of Republicans in 57 counties.
79
Actual state numbers are slightly more Democratic: 70.70% overall, 43.97%
Democratic, 34.35% Republican, 9.62% Party Differential, 16.90% Decline to State,
4.79% Third Party.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

5.34%

52

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

caution, however, that my analysis is not comprehensive enough to
support more than a tentative observation. Two of the Rich Four counties
are Democratic, for example, and Nevada, which has an extremely low
percentage of its population in prison, is Republican.
Intra-county distribution might affect policy, in that a county might
have Democratic cities within counties, or particular seats on the county
council. Slates for county officials specify not just party, but person, and
individual differences on criminal justice might account for some of the
observed results. Finally, individual county council seats are drawn
within counties and might heighten the effects of how Democrats and
Republicans are distributed within the county.80
E. Reverse Causality: Is Low Crime the Product of a High NFA?
In this Part, I consider whether I have been analyzing the problem
backwards. I have analyzed whether prison is a product of crime.
Perhaps, though, crime is a product of prison. That would mean the low
crime rates associated with high NFA are an indicator that prison works.
Under this theory, because offenders in High Use counties are subject to
swift and certain punishment, this means both that there are fewer of them
left to offend (incapacitation) and that any remaining offenders are less
likely to risk prison (deterrence). I will not attempt to determine whether
changes in prison are, in fact, the cause of changes in crime, not the other
way around, but I note that this is the subject of vociferous—and
voluminous—academic debate.81
I will, instead, frame the problems in terms of the central question of
this paper, a question which provides a transition to the next section of the
Paper, Fiscal Implications. Even if one were to assume that the causation
80

For an evaluation of the role of party politics in sentencing commissions, see
Rachel Barkow and Kathleen O'Neil, "Delegating Punitive Power: The Political
Economy of Sentencing Commission and Guideline Formation," 84 Texas Law Review
1973 (2006).
81
See, e.g., William Spelman, Specifying the Relationship Between Crime and
Prisons, 24 J. Quant. Criminol 149 (2008) (surveying several quantitative studies and
finding that, “[d]espite many years of study, the effect of prisons on crime remains a
controversial question.”). For a more accessible introduction to this controversy, see
The Pew Center on the States, The Impact of Incarceration on Crime: Two National
Experts
Weigh
In,
April
2008,
available
at
http://www.pewcenteronthestates.org/uploadedFiles/Crime%20Incarceration%20QA.pdf.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

53

in fact runs from prison to crime, then why should the state pay for it?
The choice is made in the county and the benefits go to the county. If the
policies are, in fact, effective, then the counties should be happy to pay
for it. If the state pays for it because it believes the policy is worth
subsidizing, which counties should it pay for? Can it afford to subsidize
High Use rates for all counties? Should it subsidize just prison, or should
it subsidize other policy choices as well? I discuss these and other issues
in the following section, which discusses the state’s role in funding
prisons.
IV. FISCAL IMPLICATIONS
This Part examines the fiscal ramifications of the state prison
subsidy. Given that the state pays for prison, and that counties use prisons
at different rates, what is the net prison subsidy (or tax) for counties? I
have adverted to the idea of subsidy, without mentioning the numbers.
This Part details exactly what those numbers are. In addition to exploring
what is, I also calculate what might be if the state emulated the High Use
counties or the Low Use ones. I calculate statewide figures by adjusting
the state’s coverage rate to each segment’s coverage rate. I also calculate
what would happen if a single segment moved to another segment’s
coverage rates. There we see that if just Los Angeles County moved to a
High Use coverage rate, for example, the fiscal impacts would be
substantial.
A.

Subsidy by Segment

The following table calculates prison subsidies in the manner described
earlier.82 I multiplied the coverage rate by the number of violent crimes in
a segment to come up with the “fair” or “justified” NFA number. I then
subtracted this number from actual NFA and multiplied the result by per
capita prison costs to arrive at the subsidy (or tax). I also calculated the
subsidy on the basis of property crime coverage and Part I coverage, to
see if an NFA rate not justified on the basis of violent crime might be
justified on some other measure of crime.
Table 16: Prison Subsidy by Segment, Average Yearly Values, 2000-2009
High Use
82

Low Use

Los

Middle

See infra at 11-12.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

State Total

54

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11
Angeles

Use

NFA
Coverage

17,028
35.90%

8,311
14.60%

13,888
18.78%

5,045
27.58%

44,272
22.52%

Violent Crime

47,427

56,929

73,956

18,293

196,604

10,734
12,982
16,406
4,150
NFA if at State
Coverage Rate
6,294
-4,671
-2,518
895
Excess NFA
$28.97
Average yearly
$210.05 -$166.30
-$72.73
Subsidy
(millions)
$12.92
Highest
$68.78
$5.23
$.97
Individual
Yearly County
Subsidy
(Millions)
-$5.79
Lowest
-$.32
-$85.90
-$145.04
Individual
Yearly County
Subsidy
(Millions)
$14.67
Property
$122.10 -$201.64
$64.86
Coverage
Subsidy
(millions)
$6.41
Part I coverage
$126.40 -$208.52
$75.70
Subsidy
(millions)
Bold: highest value; Italics: lowest value. All figures are gross
numbers. High and Low Subsidy figures refer to individual counties
within the respective groups.

California pays an immense amount of money to subsidize the prison
usage of High Use counties that is not justified by violent crime, an
average of $210 million a year. Individual counties used huge sums of
state resources: San Bernardino’s prison use was subsidized an average of
$51 million a year, with a high of almost $69 million in 2006. These
figures, again, only calculate the cost of the first year of imprisonment of
NFA for that particular year, and only for the number of NFA exceeding

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

N/A
N/A
N/A

N/A

N/A

N/A

N/A

TOUGH ON CRIME (ON THE STATE’S DIME)

55

that justified by the statewide coverage rate. During the ten years of the
study, only one of the eighteen High Use counties had a negative subsidy.
Fresno had a single year (2000) in which its prison usage was not
subsidized (-$320,000). Fresno’s excess prison usage cost an average of
more than $15 million a year between 2000 and 2009.
Low Use counties left millions of dollars of prison resources on the
table. If they had incarcerated at the statewide coverage rate, they would
have used, on average, an extra $166 million in prison resources a year.
The difference between the cost of High Use deviations from the state
average and Low Use deviations is more than $375 million a year, a
tremendous transfer of resources from one-third of the state to another.
Individual Low Use counties forewent huge amounts of crime-justified
prison resources. Alameda County used an average of $48 million dollars
less than its justified amount, with a high (or low) of -$85 million in 2008.
(Again, this estimate only includes the first year’s cost of a new felon
admission.) Estimates of Low Use counties as a segment are somewhat
dampened by the inclusion of San Diego, which was in the top quartile for
subsidies for two years (though its average annual subsidy was -$8.5
million).
Los Angeles County was also on the losing end of the prison subsidy,
averaging a -$72 million subsidy for the ten years of the study. Los
Angeles spent the first five years of the past decade in the -$100 million
range, hitting a peak of -$145 million in 2003 before dropping to -$96
million in 2004. The rest of the decade saw the Los Angeles subsidy
numbers increase as Los Angeles’s coverage rates increased, a product
both of decreasing violent crime and increased NFA. Los Angeles had a
positive net subsidy of $970,000 in 2009.
The Middle Use counties were subsidized overall, and I note again the
heterogeneity of the group. More than half of the Middle Use counties
were subsidized in 9 or more years of the study.83
Table 16 also calculates subsidies according to alternative coverage
rates. If prison is justified for more than just violent crime, are the
subsidy numbers different? The answer depends on which segment of the
83

Amador, Del Norte, Mariposa, San Luis Obispo, Siskiyou, and Tulare were
subsidized in 9 of the 10 years; Humboldt, Madera, Tuolomne, Ventura, Yolo and Yuba
were subsidized all ten years.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

56

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

state one looks at. High Use counties look a little less high use when
coverage is calculated using either reported property crimes or reported
Part I crimes. Their subsidy drops to a yearly average of about $122
million, a little more than $70 million less than the yearly violent crime
subsidy. Low Use counties, however, see their prison resource shortfall
grow, dropping to below $200 million. These numbers can be explained
by reference to the relatively high property and Part I crime rates in both
High and Low Use counties. High property and Part I crime justifies
more of the High Use counties’ NFA and increases the amount of prison
resources left unused by the Low Use counties.
Perhaps the most interesting result of recalculating coverage by
property and Part I crime, though, is that Los Angeles goes from being a
net donor to a net recipient of unjustified prison resources. Remember,
Los Angeles County’s NFA rate is high on a straight per capita basis—it is
low only when adjusted for its high violent crime rate. Because Los
Angeles does not suffer from relatively high property and Part I crime,
however, its high NFA rate is no longer justified when adjusted for these
types of crime, and its property and Part I coverage rates are, accordingly,
higher than the state average. Once again, the measure of subsidy is
ultimately a normative question: what prison admissions are justified, and
on what basis?
Table 17: Prison Subsidies for High Use Counties, Average Yearly
Values, 2000-2009
High
Coverage

Average
Raw NFA
Numbers
Per Year
Coverage
Violent
Crime
Raw
Numbers
NFA if at

High
Subsidy

Rich Four

Poor Four

High
Coverage
and
Subsidy

329

15,284

5,491

9,793

1,415

54.61%
603

34.45%
44,361

33.72%

34.88%

16,284

28,077

57.46%
2,463

138

10,030

3,674

6,357

566

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

57

State
Coverage
Rate
191
5,254
1,817
3,436
Excess
NFA
Average
$6.36
$175.37
$60.85
$115.52
yearly
Subsidy
(millions)
Highest
$3.84
$68.78
$55.39
$68.78
Individual
Yearly
Subsidy
(millions)
Lowest
$.01
-$.32
$1.16
-$.32
Individual
Yearly
Subsidy
(millions)
Property
$5.81
$91.29
$28.27
$63.02
Coverage
Subsidy
(millions)
Part I
$5.95
$92.84
$17.37
$75.48
coverage
Subsidy
(millions)
Bold: highest value; Italics: lowest value. All figures are gross (nonnormalized) numbers. High and Low Subsidy values are for individual
counties within the respective groups. Figures might not add due to
rounding.
Table 17 takes a closer look at just the subsidized counties. The
Poor Four dominate here, sending, in an average year, 3,436 excess new
felons (those sent above the number calculated at the state coverage rate).
These prisoners cost an average of $115 million in just the first year of
their incarceration, and the Poor Four incur this cost every year. The
Rich Four, however, also cost the state large sums of money on the NFA
they send above the state coverage rate. Two rich counties in particular
receive large subsidies: Santa Clara and Orange, which received eight

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

849
$28.32

$13.71

$1.25

$25.00

$24.71

58

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

digit subsidies each year, an average of more than $16 million for Santa
Clara and $36 million for Orange. The state pays for these extra
prisoners, even though the citizens in the counties who sent them make,
on average, no less than $4,000 more than the average Californian.
B. Recalculating State Coverage Rates by Segment
What would happen if other segments of the state began acting like
one another? I consider a variety of scenarios. First, I calculate what
would happen if the state coverage rate were replaced with the coverage
rate of each of the four segments. Even though the resulting figures
include only the cost of the first year of each new felon’s sentence, the
results would be dramatic, ranging from an additional cost of $890 million
to a cut of more than half a billion dollars. Second, I calculate what
would happen if only individual segments of the state changed their
coverage rates. This analysis shows that changing just parts of California
could have profound fiscal impacts.
Table 18: Subsidy Recalculated with Changed Statewide Coverage
Rate, by Segment, Average Yearly Values, 2000-2009
High Use
Coverage

35.90%

Low Use
14.60%

Los
Angeles
18.78%

Middle
Use
27.58%

State Total
22.52%

70,590
28,703
36,919
54,221
44,272
State NFA if
at Segment
Coverage Rate
26,318
-15,569
-7,353
9,949
N/A
Excess NFA
-$526.58
-$248.72
$336.51
N/A
Change in
$890.14
Cost (millions)
Bold: highest value; Italics: lowest value. All figures are gross
numbers. High and Low Subsidy values are for individual counties within
the respective groups.

One thing is immediately apparent from Table 18: the state cannot
afford for all counties to act like High Use counties. If the state
incarcerated at the High Use coverage rate, it would cost an additional
$890 million each year for just the first year of new felons’ sentences.
The state would also have to find room in its already overcrowded prisons

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

59

to house an additional 26,318 incoming prisoners each year. The state
could, however, shed an average of more than 15,000 inmates if it
adopted Low Use coverage rates statewide. In doing so, it would save
more than $500 million in the cost of the first year of new felons’
sentences.
Table 19: Subsidy with Changed Segment Coverage Rate, by Segment,
Average Yearly Values, 2000-2009
High Use
Segment
Changes to
High Use
Coverage Rate
Segment
Changes to
Low Use
Coverage Rate
Segment
Changes to
Los Angeles
Coverage Rate
Segment
Changes to
Middle Use
Coverage Rate
Segment
Changes to
State Average
Coverage Rate

Low Use

Los
Angeles

Poor Four

Rich Four

N/A

$410.23

$420.40

N/A

N/A

-$341.75

N/A

-$104.52

-$192.59

-$105.31

-$274.72

$80.46

N/A

-$152.90

-$82.29

-$133.55

$249.92

$220.14

-$69.33

-$33.82

-$214.73

$152.48

$93.56

-$117.38

-$61.69

Even if the state were not to change as a whole, just changing a
segment of the state—or just the Rich and Poor Four—could have
significant impacts on prison space and prison budgets. If the High Use
counties changed their coverage rates to the state average, the state would
immediately save an average of more than $214 million a year (plus the
amounts from future sentence years of foregone NFA). If the Poor Four
alone changed to the state average, the state would save $117 million. In
fact, if the Poor Four adopted the coverage rate of any of the other

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

60

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

segments (besides the Rich Four), the state would save millions of dollars
and thousands of prison beds. Alternatively, if the Low Use counties—or
just Los Angeles County—begin to emulate the High Use counties, the
state is in for even higher prison budgets, more than $400 million to cover
the first year of new felon sentences alone.
With this diagnosis, what can California do to change coverage rates
and prison usage, or at least to account for them? The next section
sketches out some answers to that question.
V. POLICY IMPLICATIONS
California faces many challenges relating to its overcrowded prisons.
Once we understand that California’s counties are different when it comes
to prison use, what are the policy implications? What would happen if
California’s policymakers understood that counties are different—and that
a county’s use of prison might be the result of policy choices, not
responses to crime? What effect would it have on policies to promote
prison population reductions? I examine three possibilities: realignment,
probation subsidies, and sentencing.
A. Realignment, Prisoner Release
California must cut its prison population by approximately 37,000
inmates84 within the next two years85 or federal courts will order it to
release prisoners. Recently, the California Assembly passed legislation to
“realign” criminal justice, shifting more responsibilities from the state to
counties.86 The program is in limbo while it awaits funding, but it is
clearly the direction the state is headed in. As the state moves to redefine
its relationship to the counties, the county analysis in this Article might be
useful in blunting the criticism that the state is pushing its problem onto
the counties. With High Use counties, the state is simply returning the
problem to those counties. The state has thusfar given no indication that it
will attempt to tailor realignment to individual counties, but ideally, it
would tailor its responses to High and Low Use counties, and demand
more of the former than the latter.
84

Brown v. Plata, 563 U.S. ___, slip op. at *3 (2011).
Id. at 45. The Supreme Court did, however, strongly hint that the three-judge
panel should extend the timeline if the state requests it. Id. at 46-47.
86
See supra notes 5-9 and accompanying text.
85

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

61

A second way this analysis might help is in the implementation of
realignment, particularly when it comes to setting benchmarks of current
versus desired prison usage. As I have stated, the prison usage per
capita—whether total prison population or NFA—is too crude a measure.
Tying prison population to current usage would merely lock in the existing
subsidy, rewarding (in perpetuity) counties which choose prison—and not
other options—as a response to crime. In some ways, in fact, tying
benchmarks for new reforms to existing prison usage is ironic: it treats
overcrowding by rewarding those counties most responsible for it. Yet a
bill passed in May to reimburse counties for building local jail facilities
would “give funding preference to counties that committed the largest
percentage of inmates to state custody in relation to the total population of
CDCR in 2010.”87 Using per capita prison usage does not eliminate the
prison subsidy, it merely shifts it to another part of the ledger.
The state should, instead, tie realignment benchmarks to the violent
crime coverage rate. This would allow for flexibility in letting counties
imprison greater numbers in response to local outbreaks of reported
violent crime, while tying state subsidies for prison usage to its most
persuasive justification: crime. Violent crimes are readily reported, and
because higher crime rates are political poison, counties have disincentives
to game them. It is unlikely that localities would risk the political
discontent from rising crime rates in order to reserve more prison
resources for themselves.
Finally, one thing that has gotten lost in the realignment
discussion—and in this Article—is the relative size of the county and state
in criminal justice. Prison subsidies figures are sizeable, but they are
dwarfed by local criminal justice budgets. I added statewide budget
figures for local law enforcement (sheriffs and police), jail, and probation
to get an approximation of the amount of money spent locally on criminal
justice—though these figures in particular do not include the budget for the
county’s chief law enforcement official, the District Attorney. I then

87

The bill, known as AB 94, was passed in May 2011 but will only go into effect
once the state funds it. The full text and history of the bill is not yet on the major
research services; however, the state’s electronic version of the bill is available at
http://www.leginfo.ca.gov/pub/11-12/bill/asm/ab_00510100/ab_94_bill_20110510_chaptered.html. The quoted text comes from Govt. Code §
15820.917 (b) (2011).

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

62

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

added a county’s imputed gross prison budget (total prison population
times per capita prisoner cost) to these budget figures. The result gives a
total measure of county criminal justice costs. Prisons are only one
quarter of this total amount. Counties have, on average, three times the
criminal justice resources available in-county that the state spends on its
behalf for imprisonment. Prisoners in state facilities are not the largest
part of county criminal justice. They never have been. I say this only to
give the financial concerns about realignment their proper context.
B. De-Subsidizing Prison, Re-Subsidizing Probation
The state could create two incentive mechanisms to encourage
High Use counties to lower their coverage rates—and to encourage Low
Use counties not to raise theirs. The first would be to decrease the
relative cost of in-county dispositions. The second would be to increase
the cost of prison usage.
Lowering the cost of in-county dispositions means expanding
financial support for diversion programs (such as those aimed at drug
abusers or the mentally ill), jail construction, and probation. As noted
earlier, jail bed numbers can increase without new construction if counties
relied less on bail and released more of the arraigned on their own
recognizance.88 The state could encourage this—or mandate it—through,
inter alia, changes to statutes or the uniform bail schedule, by subsidizing
the bail bond market, or subsidizing electronic monitoring. The state
could also subsidize probation, as, indeed, it did until the mid 1970’s.89
The state would need to ensure that subsidies kept pace with actual costs
to the county, and it could build political will by framing the costs in
terms of money saved on inmate costs. Any program must tie funding to
measurable outcomes, to ensure that the programs actually reduce the
strain on the state’s prisons. Otherwise, the state will be spending money
without saving it.
The second option, charging counties for surplus prison usage, is
more policy neutral. Probation subsidies might encourage an uptake in the
gross numbers of people in the criminal justice system (or at least make it
more affordable). Charging for prison usage is more narrowly targeted at
88

See discussion supra at note 64 and accompanying text.
See, e.g., Kara Dansky, Understanding California Sentencing, 43 U.S.F. L. Rev.
45, 63 (2008).
89

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

63

reducing unjustified use. California actually used capitation fees in its
successful drive to decrease the state’s youth prison population.90 The
state charged counties per a rate schedule inverted with the seriousness of
offense: the state charged counties a lower day rate to house more serious
offenders and a higher day rate to house less serious offenders.91 The
capitation rate policy has not been tried with adult prison populations,
however.92
C. State Population Control And Determinate Sentencing
Although I have stated that prison overpopulation is largely a
county problem, and, accordingly, that statewide solutions generally miss
the mark, I nevertheless have one recommendation for sentencing reform.
The difference is that my suggestion is not on the charging side, but on the
release side. The state should explore the reintroduction of indeterminate
sentences—those terminating in a discretionary parole release decision—on
a wider basis as a means of prison population control. In an indeterminate
system, the state can release prisoners to parole at times of crowding;
determinate sentences means the state has no such leeway. In some ways,
then, indeterminate sentencing systems allow the state to push back on
county decisions by controlling release decisions. In determinate systems,
it can’t.
California moved to determinate sentencing in 1975. Before then,
the state could control when to release an offender, even though it never
controlled who was sent there. Now the state doesn’t have any control.
The only thing that is a variable is who goes to prison and under what
charge they bargained for, both of which are determined long before the
state has custody. There is a large amount of discretion with inputs to the
prison system—all of it at the county level or below—and none on the state
side with release. 93
90

See, e.g., Little Hoover Comm’n, Juvenile Justice Reform: Realigning
Responsibilites 4 (2008), available at http://www.lhc.ca.gov/studies/192/report192.pdf.
91
Id.
92
See, e.g., Zimring and Hawkins, supra note X, at 212 (describing a policy of
“surcharging units of local government for additional offenders referred to state prisons”
but noting that “we know of no American jurisdiction where t his has been seriously
proposed or considered.”).
93
See id. at 211. “Eliminating or reducing the power of parole boards over the
release of prisoners removed a significant means of controlling prison population from
that level of government responsible for the cost of the prison system.”

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

64

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11

Of course, I am well aware of the problems with some forms of
indeterminate sentencing, as I have demonstrated elsewhere.94 I would not
support the introduction of fully discretionary, unguided, haphazard
indeterminate sentencing. Instead, the state should go one of two ways:
setting statewide standards on risk and enforce them system-wide, or to
acknowledge the role of community differences and break up the state
system entirely. I have already written about the former point;95 my next
Article takes on the latter.96
***
California is one state; it is also 58 counties. When it comes to
criminal justice and the state prison population, localities are where the
action is. County criminal justice budgets are much larger than prison
budgets, county officials make most of the key decisions, and county
responses to crime—not crime itself—drive new felon admission rates.
Alameda and San Bernardino are very similar when it comes to criminal
justice except in their usage of prison. It is hard to understand why the
tax revenues from Alameda’s residents should go towards paying for San
Bernardino’s choices. I am not suggesting that the case cannot be made; I
am, however, saying that on the basis of crime, the case has not been
made.97
I want to emphasize, again, that this study is subject to several
limitations. Measuring prison usage in terms of violent crime is a choice I
made in designing the study, not a result of it. I have no smoking gun
evidence that prison usage is a policy choice; I have only evidence that
higher prison usage is not the result of higher crime. Ultimately, the
conclusion of this study is that counties are different. The difficult
question that remains is which of those differences the state should
subsidize, if any.
94

See, e.g., See W. David Ball, Normative Elements of Parole Risk, 22 Stan. L. &
Pol'y Rev. 395 (2011) (describing California’s current parole release system as “less a
form of parole release than parole retention.”). See also W. David Ball, Heinous,
Atrocious, and Cruel: Apprendi, Indeterminate Sentencing, and the Meaning of
Punishment, 109 Col. L. Rev. 893 (2009).
95
See Normative Elements of Parole Risk, supra note 88.
96
See supra note 25.
97
Perhaps Alameda receives a greater share of other state resources that evens out
with San Bernardino’s greater share of prison resources.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)

65

Glossary:
APAR—Adult Population at Risk. The subset of a county
population between the ages of 18 and 69.
Coverage—NFA as a percentage of violent crime. This is a proxy
variable for the degree to which a county responds to crime with
incarceration.
High Use—Counties with annual coverage rates and calculated
subsidy rates in the top quartile for at least 7 of the 10 years of the study.
High Five—The subset of low coverage/low subsidy counties with
relatively high per capita incomes: Alameda, Contra Costa, Marin, San
Francisco, and Santa Cruz
Low Six—The subset of low coverage/low subsidy counties with
relatively low per capita incomes: Imperial, Nevada, Sacramento, San
Joaquin, Sonoma, and Stanislaus.
Low Use—Counties with annual coverage rates and calculated
subsidy rates in the bottom quartile for at least 7 of the 10 years of the
study.
NFA—new felon admissions, prisoners entering prison upon
conviction or plea of a new felony charge. Distinguished from other
entrants to the prison system, such as those who have had their parole
revoked or parolees admitted with a new term (as a result of a new crime).
Poor Four—The four high-subsidy counties with below-average per
capita incomes: Fresno, Kern, Riverside, and San Bernardino.
Rich Four—The four high-subsidy counties with above-average per
capita incomes: Orange, Placer, Santa Barbara, and Santa Clara.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

66

TOUGH ON CRIME (ON THE STATE’S DIME) DRAFT, 7/11
Appendix A: List of County Segments

High Use
High Coverage

High Subsidy

Both

Fresno
Kern
Orange
Placer
Riverside
San Bernardino
Santa Barbara
Santa Clara

Butte
Kings
Shasta
Sutter

Low Use
Low Coverage

Low Subsidy

Both

Alpine
San Benito

(Los Angeles)
San Diego

Alameda
Contra Costa
Imperial
Marin
Nevada
Sacramento
San Francisco
San Joaquin
Santa Cruz
Sonoma
Stanislaus

Colusa
Glenn
Inyo
Lake
Lassen
Trinity

Middle Use
Amador, Calaveras, Del Norte, El Dorado, Humboldt, Madera,
Mariposa, Mendocino, Merced, Modoc, Mono, Monterey, Napa, Plumas,
San Luis Obispo, San Mateo, Sierra, Siskiyou, Solano, Tehama, Tulare,
Tuolumne, Ventura, Yolo, Yuba.

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

TOUGH ON CRIME (ON THE STATE’S DIME)
Appendix B: Map of County Segments

Low Use counties
Middle Use Counties
High Use counties
Los Angeles County

DRAFT—PLEASE DO NOT CITE WITHOUT PERMISSION

67