
POLITICS, IMPRISONMENT AND RACE
An adult black man is seven times more likely than his white counterpart to reside behind bars. Paradoxically, the largest disparities are found in political
domains controlled by liberals  the leaders in the struggle for
racial justice. By examining how criminal behavior is distributed within
the races, the paradox is resolved showing it to be an unintended
consequence of liberal benevolence and goodwill.
"One
should not increase, beyond what is necessary, the number of entities
required to explain anything."
 William of Ockham
(12851349)
Prodigy's
Sociology Colloquium lecture, April 2006
It
was uncommonly kind of Uncle Steve
to
invite me here this
afternoon to participate in your Sociology colloquium
Series. I am delighted to be here, to meet so many of you, and to join
more of you later for dinner.
Skimming through the titles of past
colloquia, I found
that
certain words or their cognates occurred with conspicuous
frequency, among them: race, politics and
prison.
My lecture today concerns each of them  more specifically their
relationship to one another.
We all know
that African Americans are
imprisoned disproportionately to their numbers in the general
population. According to the last decennial census a black man was 7.4
times more likely than his white counterpart to be incarcerated. In the
language I'll use today, we would say that the disparity or
incarceration ratio was 7.4. Statebystate, the figures varied widely from 3.1 to 29.3. But
contrary to expectation, the
highest disparity ratios turned up mostly in politically progressive states,
while the smallest ratios were mostly found in conservative
states. Though the numbers change a bit
from year to year, this racialpolitical pattern of imprisonment
endures. One of the questions I will answer today is, why?
Many here have
devoted their professional
lives
to eliminating racial disparities in prison and of course elsewhere.
And most of the rest of us are philosophical allies sympathetic to
this ideal. Consequently, it must be disconcerting to find the greatest
blacktowhite imprisonment ratios in your own ideological backyards.
But be assured this is not the result of unconscious,
repressed racism,
but rather the innocent product of your goodwill  an accidental consequence of liberal philosophy applied to criminal
justice.
I'll begin with the
data used in this analysis. It was
assembled by Human Rights Watch, one of the more prominent organizations dedicated to eliminating human abuse. HRW
gathered incarceration statistics by race on adult males 18 to 64 from
Census 2000 tabulations. NonHispanic whites are included as a distinct
racial category. Throughout
this lecture, then, the term white is to be taken to mean nonHispanic
white. Appendix II
of the handouts lists incarceration rates
by
race and state.
My lecture today is about the politics of race and
incarceration, how it varies from state to state, and how it influences
the color of the prison population. To gauge political
sentiment, I use the gold standard of political barometers: the
"liberal quotient" or LQ designed and evaluated by the venerable liberal
political organization Americans for Democratic Action. ADA
assigns an LQ to every federal lawmaker. It is simply the percentage of
votes cast
by the legislator in support of the ADA position on 20 key issues. The selfdescribed liberal activist
organization calls a score of
100 "a perfect liberal quotient." ADA designates those achieving this score "heroes." After serving four years in the
Senate, Hillary Clinton finally attained
hero status in 2005. She was one of 22 senators in the heroes club.
Legislators at the other end of the political spectrum are dubbed "zeros." The zeros club is more select. Only five senators
made it in 2005. Though outnumbered by conservatives, liberals still managed
more than four times as many heroes as conservative did zeros. We can only
surmise that liberals are the more passionate political group, tending
in their ideology toward political extremes  all the more reason to wonder why
the greatest disparity ratios are found in liberal bailiwicks. That aside, I
gauged a state's political outlook by its lawmakers' average LQ.
Social critic Steve
Sailer observed
in 2001 that conservative states tend to incarcerate whites at high
per capita rates. Figure 1 confirms Sailer's observation. It shows, for
adult men, the relationship between a state's white
incarceration rate and its average LQ. The relationship is
strong and inverse (R
= 0.56).
Figure
1. The
percentage
of white adult males incarcerated in each state vs. the state's
average LQ. High white incarceration rates are found mostly in
conservative states, low rates
mostly in liberal states. The relationship is strong and
linear. The regression line is shown dashed.
(R = 0.56.)
We seek a measure of a state's conservative leanings. Conservativeleaning
states tend to incarcerate all groups at high rates. But we cannot use a state's
incarceration rate to measure conservatism because the total rate is
strongly dependent on racial and ethnic mix. State's with high
proportions of blacks, for example, will have high incarceration rates
simply because the black rate of incarceration is so high. A
singlegroup incarceration rate does not suffer from this difficulty. It
can work well to gauge conservatism. We chose the white incarceration
rate as a proxy for conservative sentiment.
Figure
2 illustrates the disparity
paradox. The white incarceration rate is the independent
variable. Notice that high
disparity ratios are
found mostly in progressive states, and low ratios mostly in
conservative states. For adult males, the correlation
between the disparity ratio and the white incarceration rate
is strong and inverse (R =
0.54).
Aside from the
strong inverse correlation between white incarceration and the
disparity ratio, the most compelling feature of Figure 2 is the extraordinarily
large disparity ratio found in the
nation's capitol. The
ultraliberal, sixtypercentblack District of Columbia checks in off
the charts with black male per capita incarceration 29.3
times that of whites.
Figure
2. Blacktowhite
incarceration (disparity) ratios vary inversely with the white
incarceration rate, a proxy for conservatism. R = 0.54. Relative to the regression
line (shown dashed), the District of
Columbia and Hawaii are
prominent outliers.
Resolving the disparity paradox, wherein progressive states have the largest disparity
ratios
is the primary goal of this research. A secondary goal is to determine
how criminality is distributed within white and black populations. The
two goals are intertwined, because to resolve the paradox we need the
distributions. When the dust settles, I suspect
revealing some details of the criminality distributions will be the more important
achievement. But let's get started.
I begin with a
quote
from La
Griffe du lion:
Sometimes
human traits
can be difficult to define, let
alone
measure. Still, we have a sense of their meaning. Take for
example, "ambition." A man is not simply ambitious. He is very
ambitious or moderately ambitious or not ambitious at all.
Such properties vary by degree but resist precise quantitation. We call
them fuzzy variables.
The pattern of behavior
leading to arrest and
imprisonment is,
in the Griffian sense, a fuzzy variable. I call it criminality.
To account for incarceration data, I propose the following threepart model.
1.
Each racial group
has its own characteristic,
geographyinvariant distribution of criminality.
2.
Criminality
distributions are approximately Gaussian (as mandated by the central limit
theorem for properties resulting from many independent factors).
3.
Political
jurisdictions tolerate crime to varying degrees.
They set thresholds of criminality, which when crossed result in
incarceration. Thresholds differ by jurisdiction according to
regional custom.
Figure 3 shows normalized distributions of criminality for
black and white adult males. The distributions are Gaussian in
accordance with the postulates. They have been rendered
accurately in the figure by using parameter values obtained from the
best least squares fit of calculated to observed
disparity ratios. (See Appendix
I for details.) The white criminality distribution
lags behind the black. It is also broader. That is, whites display more
variability. The blackwhite mean difference is 1.32 SD (in units of the white standard
deviation); the (black/white) variance ratio is
0.76.
Figure
3. Distributions
of criminality
for adult black and white males. Thresholds for
incarceration are shown for two states. They differ according to
regional practice. The areas under the distribution curves to the right
of the thresholds are the incarcerated fractions of the
corresponding groups.
Two behavioral thresholds are shown in
Figure 3. They correspond to different standards for incarceration applied in
two different states. The lower threshold could have been set by a
toughoncrime,
nononsense, conservative state like Oklahoma. In Oklahoma, it doesn't
take much to land
behind bars. The upper threshold belongs to a state with a higher
tolerance for criminal behavior, say, New York. In each case, the
incarcerated fraction of a group equals the area under its distribution curve
to the right of the threshold. The ratio of these areas, black
over white, is the disparity
ratio. Figure 4 isolates these areas for the thresholds shown.
Highthreshold New York has the greater disparity ratio. If Figure 4 fails to
convince you that moving an incarceration threshold
to a higher level of criminality increases the disparity ratio, I present a more formal treatment in
Appendix
I.
Figure
4. The ratio of
areas representing the disparity ratio is shown schematically for two
states. The ratio is greater in the more permissive, highthreshold New
York than in the toughoncrime, lowthreshold Oklahoma.
Theory and
observation
In Appendix I
a relation
between the incarceration ratio and the white incarceration
rate is developed from our model. Figure 5 shows three
of a family of curves illustrating the relation. The curves differ in the choice
of parameter values in the criminality distributions. In the neighborhood of very small white
incarceration (i.e., in ultraliberal jurisdictions) the curves can be very
steep, in part accounting for the extraordinary disparity ratio found in the
District of Columbia.
Figure
5. A
family of curves relating
the
disparity ratio to white incarceration rates. The curves shown
were generated by choosing different sets of parameter values in the criminality
distributions.
We can determine the criminality distributions in
the populations of adult black and white males by finding the curve in the
family that, in the least squares
sense, best fits observed incarceration data. However,
any assessment of group criminality should include former as well as present
felons. The incarceration data we have includes only those
currently incarcerated.
The number of living exprisoners is
actually much
greater than the number incarcerated. In 2001, for example, former
prisoners accounted for 77% of all adult U.S. residents who had ever been
incarcerated. Excluding them would seriously distort the analysis. As a
zeroth order approximation, we estimated the number of men who
were ever incarcerated by multiplying the inmate count by 77/23, the
factor that would apply exactly if the 2001 figure applied across the board.
Figure 6
shows the best leastsquares fit of incarceration ratio to the
everincarcerated estimates. The predicted incarceration ratio cuts
through the core of state data. It swings wildly upward when approaching
superliberal territory taking aim at the extreme incarceration ratio
found in DC.
Figure
6. Variation of the disparity ratio with the percentage of white men
ever incarcerated in fifty states and the
District of Columbia. The predicted variation is shown by the curve.
In sum, we see that: 1)
incarceration rates by state correlate strongly and inversely with the
ADA "liberal quotient"; 2) variation of disparity ratios by
state is well
described by a model in which racial groups have characteristic,
geographyinvariant, Gaussian distributions of criminality; 3) applied to
incarceration data, the model yields criminality distributions for adult
black and white males, finding the black distribution narrower (variance ratio = 0.76), and
displaced toward higher
criminality values by 1.32 SD; the model predicts the
disparity paradox, showing it to be an unintended result of setting
high incarceration thresholds.
Thank you.
<
polite applause
>
I see Uncle Steve has signaled that we have time for a few
questions. Yes ma'am?
First
Professor: Are you
saying
that liberals are the cause of the disparity ratio?
Prodigy
: No. The
disparity
ratio is greater than unity because blacks commit more crime per capita than
whites. Liberals simply tweak the
ratios upward by setting high incarceration thresholds in jurisdictions
they control. The tweaking is sufficient to produce an
observable trend in which the largest disparity ratios are found in progressive
states. Isn't it ironic that the high disparity ratios found in
progressive states are the result of the tolerance and goodwill
we've learned to associate with progressive government?
Second
Professor: The
American
Anthropological Association in its statement on race declares:
"The
'racial' worldview was invented to
assign some
groups to
perpetual low status, while others were permitted access to
privilege, power, and wealth. The tragedy in the United States has
been that the policies and practices stemming from this
worldview succeeded all too well in constructing unequal populations
among Europeans, Native Americans, and peoples of
African descent."
How does this square with your analysis?
Prodigy:
I have simply
described one consequence of the policies and practices stemming from
the
'racial'
worldview. As such, I
hope anthropologists will welcome my findings. Sociologists too!
Third
Professor:
I notice
in Figure 6 there is
considerable scatter of the observed disparity ratios about the curve of predicted ratios.
Why is that?
Prodigy:
The curve in
Figure 6
describes how disparity ratios would vary with white incarceration
rates if
the world were
constructed exactly like my model. But the
mapping of the model onto the real world is not perfect. Also, it would help if the estimates of
everincarcerated percentages were replaced with actual observations.
Nevertheless, the substantial agreement
we do achieve indicates that the model incorporates the most salient features of
the real world that bear upon the problem.
Fourth
Professor: Can your
analysis teach us how to reduce the disparity ratio?
Prodigy:
It
can, but you probably won't like what it tells us. I think we could all
agree that the very best way to reduce
the disparity would be by modifying the criminality distributions so that blacks
and whites would commit crime at the same rate. But criminality
distributions have stubbornly resisted change, leaving us only the incarceration thresholds to play with.
To lower the disparity ratio in
your state, simply lower the incarceration threshold imprisoning more
bad guys of all stripes. That is the price you will have to pay to
reduce the disparity ratio.
Sixth
Professor: At the
January
Conference of the Society for Personality and Social Psychology,
several presenters found
correlations between political philosophy and racial bias. One study
found that conservatives had stronger selfadmitted and implicit biases
against blacks than liberals did. Do you care
to comment?
Prodigy:
It does not
surprise me. Of course, some might say that one man's bias is another's perspicacity.
Uncle
Steve:
Ahem. We can
continue this over dinner. Thank you, Prodigy, for a stimulating talk, and for a refreshingly novel approach
to a sociological problem undertaken honestly and with
nobility
of purpose. Speaking
for sociologists, we do appreciate your explanation of a sociological
bibelot that has puzzled us for years.
<
enthusiastic
applause
>
APPENDIX I. The
details
Both black and white criminality distributions are
assumed Gaussian. That is, their distributions looks like this:
where
μ
and σ are
the mean and standard deviation, respectively.
A simplification can be
obtained by choosing the unit of
criminality to be one standard deviation of the white distribution.
Also, because we are interested only in the relative positions of the
black and white distributions on the criminality axis, we can and do
set the mean white criminality μ_{W}_{ }=
0. The white
criminality distribution then looks like:
and the black distribution like:
in which Δ
is the mean difference, μ_{B} 
μ_{W}.
The primes are included to remind us that we used the
transformations, x' = x / σ_{W}
and
dx' =
dx / σ_{W}
to arrive at (3). Note that with our choice of units, the black
distribution depends on the dimensionless ratio
of standard
deviations, σ_{B} / σ_{W}.
The
dependence of the disparity ratio on the white incarceration rate,
plotted in Figure 6, is
obtained as follows.
The fraction of whiteadult males ever imprisoned is given by:
where Λ is
the criminality threshold
measured from the zero of the white distribution. The inverse function, Λ(f_{W}),
is obtained numerically. The everincarcerated fraction of blacks, given by
may then be evaluated for particular
values
of Δ and σ_{B}
/ σ_{W}
, and the disparity ratio, f_{B}
/ f_{W}
, computed.
The quantities, Δ
and (σ_{B}
_{
}
/ σ_{W})
were adjusted to give the best least squares fit of the calculated disparity ratios, f_{B}
/ f_{W}
, to the observed
ratios.
APPENDIX
II. Percent of adult men (18  64) behind bars*

nonHispanic
White

Black

Ratio
B/W




Alabama

1.055

6.084

5.77




Alaska

0.745

4.233

5.68




Arizona

1.607

9.478

5.90




Arkansas

1.307

7.428

5.68




California

1.312

9.304

7.09




Colorado

1.161

10.729

9.24




Connecticut

0.542

8.985

16.58




Delaware

1.046

8.05

7.70




D.
C.

0.083

2.432

29.30




Florida

1.503

9.117

6.07




Georgia

1.485

6.775

4.56




Hawaii

0.363

1.12

3.09




Idaho

1.441

5.085

3.53




Illinois

0.621

7.575

12.20




Indiana

1.083

8.288

7.65




Iowa

0.905

10.987

12.14




Kansas

1.184

11.036

9.32




Kentucky

1.328

10.343

7.79




Louisiana

1.187

8.415

7.09




Maine

0.636

4.547

7.15




Maryland

0.82

5.649

6.89




Massachusetts

0.61

5.682

9.31




Michigan

1.043

7.596

7.28




Minnesota

0.578

8.459

14.63




Mississippi

0.993

6.01

6.05




Missouri

1.178

7.739

6.57




Montana

1.022

6.72

6.58




Nebraska

0.669

7.067

10.56




Nevada

1.646

8.907

5.41




New
Hampshire

0.71

6.164

8.68




New
Jersey

0.528

8.104

15.35




New
Mexico

0.828

8.085

9.76




New
York

0.534

6.377

11.94




North
Carolina

0.732

5.297

7.24




North
Dakota

0.494

2.939

5.95




Ohio

0.981

8.857

9.03




Oklahoma

1.951

12.164

6.23




Oregon

1.404

10.119

7.21




Pennsylvania

0.845

10.482

12.40




Rhode
Island

0.598

8.392

14.03




South
Carolina

1.095

6.578

6.01




South
Dakota

1.303

13.919

10.68




Tennessee

1.138

6.712

5.90




Texas

1.873

11.036

5.89




Utah

1.003

8.061

8.04




Vermont

0.555

5.026

9.06




Virginia

1.194

8.707

7.29




Washington

1.073

6.929

6.46




West
Virginia

1.004

15.154

15.09




Wisconsin

0.955

12.993

13.61




Wyoming

2.028

14.037

6.92




National

1.072

7.923

7.39








*Based
on 2000 census data collected by Human Rights Watch





