Skip navigation

Criminology - Historical Contingencies Explaining Mass Incarceration 2015

Download original document:
Brief thumbnail
This text is machine-read, and may contain errors. Check the original document to verify accuracy.
HISTORICAL CONTINGENCIES AND THE EVOLVING
IMPORTANCE OF RACE, VIOLENT CRIME, AND
REGION IN EXPLAINING MASS INCARCERATION IN
THE UNITED STATES∗
MICHAEL C. CAMPBELL, MATT VOGEL,
and JOSHUA WILLIAMS
Department of Criminology and Criminal Justice, University
of Missouri—St. Louis
KEYWORDS: incarceration, crime policy, penal policy, race, crime politics
This article combines insights from historical research and quantitative analyses that
have attempted to explain changes in incarceration rates in the United States. We use
state-level decennial data from 1970 to 2010 (N = 250) to test whether recent theoretical
models derived from historical research that emphasize the importance of specific historical periods in shaping the relative importance of certain social and political factors
explain imprisonment. Also drawing on historical work, we examine how these key
determinants differed in Sunbelt states, that is, the states stretching across the nation’s
South from the Atlantic coast to the Pacific, from the rest of the nation. Our findings
suggest that the relative contributions of violent crime, minority composition, political
ideology, and partisanship to imprisonment vary over time. We also extend our analysis beyond mass incarceration’s rise to analyze how factors associated with prison
expansion can explain its stabilization and contraction in the early twenty-first century.
Our findings suggest that most of the factors that best explained state incarceration
rates in the prison boom era lost power once imprisonment stabilized and declined.
We find considerable support for the importance of historical contingencies in shaping
state-level imprisonment trends, and our findings highlight the enduring importance of
race in explaining incarceration.
The last four decades featured transformative shifts in crime control strategies in
the United States that ultimately generated a 450 percent increase in the nation’s
incarceration rate, one of the most striking products of the “get tough on crime”
era. Departing sharply from the nation’s history and the trajectories of all other
advanced democracies, U.S. states constructed the world’s largest network of prisons and
now incarcerate a greater proportion of their citizens than any other nation (Walmsley,
∗

C

Additional supporting information can be found in the listing for this article in the Wiley Online
Library at http://onlinelibrary.wiley.com/doi/10.1111/crim.2015.53.issue-2/issuetoc.
The authors would like to thank Lee Ann Slocum, Heather Schoenfeld, Janet Lauritsen, Rick
Rosenfeld, the editor, and the anonymous reviewers for comments on an earlier draft of this article.
Direct correspondence to Michael Campbell, University of Missouri—St. Louis, Department of
Criminology and Criminal Justice, 324 Lucas Hall, One University Blvd., St. Louis, MO 63104
(e-mail: campbellmi@umsl.edu).

2015 American Society of Criminology

CRIMINOLOGY

Volume 53

doi: 10.1111/1745-9125.12065

Number 2

180–203

2015

180

EXPLAINING INCARCERATION IN THE UNITED STATES

181

2011). This dramatic increase in imprisonment was not driven by a centralized nationallevel strategy for dealing with crime and was not based on a coherent body of empirical
knowledge demonstrating that prisons improved public safety. Instead, it was the product of layers of legislative decisions, primarily enacted at the state level, to charge and
imprison more offenders, increase sentences, limit prison releases, and expand incarceral
capacity. These decisions combined to create a new era in which prisons became primary
weapons in the nation’s war on crime.
This impressive expansion of the state’s power to apprehend, charge, convict, and imprison such a large proportion of its citizens unfolded against the backdrop of volatile
socioeconomic changes that included significant swings in violent crime, ongoing racial
tensions in the wake of civil rights reforms, and partisan dynamics that reshaped party
allegiances in key states, regions, and constituencies. Higher incarceration rates were not
merely the product of increasing crime—states built prisons even as crime declined, and
some continued to build them long after the sharpest decrease in crime in American history. Our temporal distance from the prison boom era and mass incarceration’s remarkable resilience in American criminal justice now provide some valuable historical context in seeking to understand the constellation of forces associated with incarceration’s
rise and relative stability. Because the policy decisions that shape prison populations are
largely (although not entirely) a function of state governments, the United States provides
50 independent cases for analysis. This decentralized federal structure has generated great
variation in how state governments responded to crime and considerable variation in the
timing of those changes, but the overall trend is undeniable—every state’s incarceration
rate increased by at least 150 percent from 1970 to 2000, and the median state increase
was 390 percent (Bureau of Justice Statistics, 2012). Recent trends in declining incarceration rates in some states highlight the heterogeneity of state penal trajectories; we use
this variation to examine relationships among key social, economic, and political factors
and incarceration.
Scholars sought to explain incarceration’s growth by using quantitative analytic strategies that analyzed state-level factors and their relationship to incarceration rates. This
research identified higher crime rates, racial composition, ideology, religious fundamentalism, fiscal resources, and partisanship as important factors in explaining state-level incarceration rates (Greenberg and West, 2001; Jacobs and Carmichael, 2001; Smith, 2004;
Spelman, 2009). By analyzing state-level changes, this body of work suggested that violent
crime was an important factor in explaining prison expansion, but so too were other social
and political factors linked to race, partisanship, and ideology. Another largely independent strand of historical research emerged that studied similar questions by examining the
forces that shaped legal and policy changes within specific states. This historical scholarship pointed to the importance of political culture, state structure, interest group activities, the role of federal courts, and partisan dynamics in explaining penal change (Barker,
2009; Campbell, 2011; Gilmore, 2007; Lynch, 2010; Page, 2011; Schoenfeld, 2010). Integrating findings from these historical studies into a “new political sociology of punishment,” Campbell and Schoenfeld (2013) have argued that the relative importance of
specific variables changes over time and must be understood as period specific. These
lines of research helped identify key forces that shaped incarceration’s rise and have done
much to explain how these mechanisms operated within certain historical conditions. This

182

CAMPBELL, VOGEL, & WILLIAMS

article integrates insights from the literature to highlight the importance of considering
temporal variation in explaining the forces that drove mass imprisonment.
Our analysis aims to integrate insights from the quantitative literature with findings
derived from historical research and the theoretical implications they have inspired. Importantly, we use decennial data that extend to 1970 (reliable annual data do not exist
prior to 1977), a time point that predates the national trend in higher imprisonment and
anchors our analysis in the earliest period of penal change outlined in historical research.
Our aims are threefold. First, we work to fill an important gap in the literature by integrating scholarship from the quantitative and historical realms to analyze whether key
mechanisms, such as crime, race, and partisanship, operate as Campbell and Schoenfeld
(2013) suggested when measured against all 50 states across time. Second, we extend our
analysis beyond previous decennial analyses to examine the post-prison boom era to determine whether predictors from previous studies hold up once rates of violent crime declined and incarceration rates stabilized and even declined in some states. Third, we use
a different definition of region that has emerged from historical research—Sunbelt versus non-Sunbelt states—to analyze whether the factors that explain incarceration differed
across these regions.

THEORY
Our primary goal is to synthesize findings from statistical analyses that have established
strong evidence and valuable insights into the factors associated with incarceration rates
across all 50 states over time, with findings from the growing body of historical case studies that have provided valuable depth and rich complexity to our understanding of how
and why lawmakers chose to imprison more people. Both strands of scholarship have
generally oriented around a relatively consistent constellation of factors—crime and socioeconomic change, race and crime policy, and political dynamics. Given the immense
complexity in analyzing so many factors across states and over time, findings have often varied considerably based on the methods employed, data selection, and overall research design.1 In this study, we focus our discussion on statistical analyses that have
addressed the long-term state-level changes in incarceration that are central to Campbell and Schoenfeld’s (2013) account—violent crime, race, political culture and ideology,
and partisanship. This is not to suggest that factors such as cultural shifts, local governing decisions, criminal justice system processes, judicial decision making, and others are
not important. Unfortunately, reliable state-level data that might accurately measure the
importance of these factors across a broad historical time frame do not exist.
CRIME AND PUBLIC OPINION
The increase in crime rates that began in the 1960s and peaked in the early 1990s coincides well with the nationwide increase in imprisonment rates, and several scholars view
crime as central to understanding mass incarceration. Wilson (1975) explained higher

1.

This account focuses as specifically as possible on research findings derived from state-level analyses that include data beginning in 1970 that is central to Campbell and Schoenfeld’s (2013) account:
1970–2010. For a thorough account of the extensive literature on politics and incarceration, see
Jacobs and Jackson (2010).

EXPLAINING INCARCERATION IN THE UNITED STATES

183

imprisonment as a policy response to popular demands to increase the costs of crime
as offending increased. Garland (2001) suggested that crime’s increase undermined an
already declining faith in rehabilitation as a guiding penal ideology, and more punitive
policies reflected a broader crisis of instability in economically advanced late modern
societies. Others question whether crime rates drive public attitudes, noting that public
opinion has been consistently punitive regardless of changes in offense rates (Zimring
and Johnson, 2006) and that public concern more closely aligns with media and political
attention to crime than with actual offending (Beckett, 1997).
State-level quantitative analyses have consistently found that higher violent offense
rates are a significant predictor of higher incarceration rates. Greenberg and West (2001)
and Jacobs and Carmichael (2001) both analyzed decennial census data from 1970, 1980,
and 1990, and they found higher crime strongly linked to higher imprisonment when controlling for several key socioeconomic factors, such as unemployment, citizen ideology,
and state spending. Spelman’s (2009) econometric analysis of annual data from 1977 to
2005 also found a strong positive link between higher rates of violent offending and spending on prison expansion and higher incarceration.2 This literature strongly suggests that
higher rates of violent offending are positively associated with incarceration rates.
PARTISAN POLITICS AND RACE
A considerable body of research has suggested that higher incarceration rates are
linked to changes in race relations and that aggressive state action against crime serves
as a proxy for racial conflict. Beckett’s (1997) research tracked changes in public opinion
on crime and found that public concern usually followed political and media attention
on its seriousness. By racializing and disproportionately emphasizing the dangerousness
of street crime, politicians drove a wedge between working-class Whites and Black and
minority voters (Alexander, 2009; Beckett, 1997). Alexander (2009) has argued that race
has been a central factor in the legal and political processes that have driven mass incarceration and that America’s criminal justice system helps relegate African Americans to
second-class citizenship. Some scholars have argued that Republicans in particular successfully deployed these strategies and won support from Southern and suburban voters
that had traditionally voted Democratic (Beckett, 1997; Hagan, 2010). Recent historical
case studies of state-level political developments have found considerable support for the
link between Republican Party political strategy and policies that emphasize incarceration, especially in Sunbelt3 states that became increasingly influential in national politics
(Campbell, 2011; Gilmore, 2007; Lynch, 2010; Schoenfeld, 2009).
Quantitative analyses have yielded mixed results regarding the links among racial
threat, Republican Party strength, and imprisonment because of data limitations and differences in model specification. Most analyses of decennial data and some annual analyses found a significant link between racial threat (variously measured as percentage
non-White or percentage Black) and higher incarceration rates (Beckett and Western,
2001; Greenberg and West, 2001; Jacobs and Carmichael, 2001; Smith, 2004). Importantly,

2.
3.

For an analysis that failed to find violent offending significantly associated with incarceration rates,
see Smith (2004).
Sunbelt states include Alabama, Arizona, California, Florida, Georgia, Louisiana, Mississippi,
Nevada, New Mexico, North Carolina, South Carolina, and Texas.

184

CAMPBELL, VOGEL, & WILLIAMS

Greenberg and West’s (2001) findings highlighted temporal variation in the role of race
and ideology—they found the effects of race to be increasing dramatically and that states
with more conservatives and a growing proportion of religious fundamentalists experienced the sharpest increases in incarceration (but see Stucky, Heimer, and Lang, 2005.)
Most studies have also found a significant positive relationship between measures of
Republican Party strength and incarceration (Beckett and Western, 2001; Jacobs and
Carmichael, 2001; Smith, 2004), although Greenberg and West did not (2001). Stucky,
Heimer, and Lang’s (2005) analysis of annual data from 1977 to 1995 further complicated
our understanding of this relationship. They examined partisan strength in the legislature
and district-level electoral competition, and they found that Republican Party strength
was most important in explaining prison admissions when district-level electoral competition and competition over control of the legislature were highest. Their findings and
that of Jacobs and Carmichael (2001) suggested that the association between Republican
Party power and higher incarceration increased over time. Differences between findings
likely reflect methodological choices over how best to model partisan strength and choices
between prison stock and flow, but the preponderance of evidence suggests that race and
partisanship play an important but complex role in explaining state-level imprisonment
and that the explanatory power of racial threat and Republican Party strength increased
over time.
Other time-series analyses have targeted the proximate policy decisions that helped
drive imprisonment’s rise. Stemen and Rengifo (2011) found a negative relationship between determinate sentencing laws and incarceration rates, and they suggested that presumptive sentencing guidelines were associated with lower incarceration rates under certain conditions. Using econometric analytical methods and panel data from 1977 to 2005,
and considering capital outlays and not just incarceration rates as a dependent variable,
Spelman (2009: 34) found “a remarkably simple explanation for what caused the prison
boom of the last 30 years: persistently increasing crime rates, sentencing policies that put
more offenders behind bars and kept them there longer, and sufficient state revenues to
pay for it all.” Spelman’s results are in line with findings from the historical case study
literature that has outlined how changes in sentencing policy and drug laws have contributed to prison overcrowding. Activist federal courts forced many states to increase
prison capacity or release inmates; mostly they chose to expand prisons, which are costly
to build and costlier to operate, which drives higher spending (Schoenfeld, 2010).
HISTORICAL RESEARCH
A growing body of historical case studies also has helped to illuminate how state-level
social and political processes interacted to drive changes in punishment policy in the latter twentieth century, especially the increase of mass incarceration. This scholarship has
emphasized the importance of political dynamics and institutions that tend to favor more
populist and aggressive responses to crime. State institutional structures (Barker, 2009;
Campbell, 2014), interest group activity and activist law enforcement lobbies (Campbell,
2011; Gottschalk, 2006; Page, 2011), federal court activism (Schoenfeld, 2010), fundamental shifts in the nation’s political economy (Gilmore, 2007), and conservative political and penal traditions (Lynch, 2010) all seem to have contributed to more aggressive
crime-control policies and higher incarceration rates. Others have highlighted historical
continuity in American punishment practices, noting that “correctional” institutions in

EXPLAINING INCARCERATION IN THE UNITED STATES

185

the United States have consistently oriented around particular ideas about labor and
race (Goodman, 2012). This research has emphasized the importance of state institutions
and political processes in explaining why lawmakers chose to invest so heavily in prison
expansion.
Campbell and Schoenfeld (2013) analyzed historical case studies of eight states (Arizona, California, Florida, New York, Minnesota, Oregon, Texas, and Washington) and
examined state-level changes in incarceration rates against the shifting national-level socioeconomic and political context of the latter twentieth century. They argued that mass
incarceration’s rise is best understood as unfolding in distinct periods in which the relative
significance of certain factors changes as new realities structure lawmakers’ options and
strategies (Campbell and Schoenfeld, 2013). They suggested that national-level political
and governing processes including financial incentives to states, interest group activity
among law enforcement and victims’ groups, federal court activism, and presidential politics created shifting incentives that helped fuel increasingly politicized and aggressive
crime policies at the state level. Their model suggests that over time, state lawmakers
experimented with new crime policies as crime rates increased and faith in less punitive
responses waned. Lawmakers increasingly embraced aggressive “warlike” responses to
crime and incrementally passed mandatory sentencing legislation, sentencing enhancements, harsh antidrug laws, and bills financing prison expansion that fueled prison growth
and higher incarceration. By establishing three distinct periods—deconstruction, contestation, and reconstruction—they illustrated how the interaction between these state and
national institutions and forces established a growing partisan consensus around a new
penal order that emphasized the “otherness” of criminals and prioritized incarceration as
never before (Campbell and Schoenfeld, 2013).
Deconstruction Period (1960–1975)
Campbell and Schoenfeld (2013) contended that higher rates of violent crime, urban
unrest, and racial tensions helped destabilize entrenched penal regimes and generated a
period of uncertainty about penal policy. Higher rates of violent offending not only led
to more inmates committed to prisons in the short term but also stimulated political and
institutional changes that generated momentum for change in the long run. We would
therefore expect higher rates of violent crime to help explain higher incarceration over a
longer horizon than is sometimes assumed. They argued that as old penal regimes were
destabilized, crime politics emerged as a central issue and that the crime problem was
effectively framed as one of ineffective criminal justice capacity. Crime’s politicization
benefited politicians who framed crime in highly racialized terms of urban and moral
decay and was most successful in states with higher rates of violent crime, larger minority
populations, and histories of harsh penal regimes.
Contested Period (1975–1992)
Although many states’ penal regimes were destabilized by the mid-1970s, no consensus immediately emerged regarding a new crime policy direction. Some Republican candidates and lawmakers balked at the cost of prison expansion, and conservative
Southern Democrats were prone to maintain the status quo in state corrections, which
featured brutal prison systems and operated at minimal cost (Campbell, 2011; Lynch,
2010; Schoenfeld, 2010). But federal court rulings forced states to address the

186

CAMPBELL, VOGEL, & WILLIAMS

overcrowding, brutality, and neglect that swamped many corrections systems, and a growing tide of states responded by expanding penal capacity rather than releasing lowlevel offenders (Campbell, 2011; Schoenfeld, 2010). More importantly, crime politics became increasingly salient in state elections in the wake of President Ronald Reagan and
President George H. W. Bush’s successful presidential campaigns, which emphasized aggressive “warlike” approaches to crime in the 1980s. Republican lawmakers proved increasingly willing to abandon their commitment to fiscal conservatism and push for prison
expansion. By the early 1990s, high rates of violent crime and frenetic political activity on crime helped drive profound policy changes that set the stage for further prison
expansion.
Reconstruction Period (1992–2001)
By the final period, Democrats, most notably marked by President Bill Clinton’s support for the Violent Crime Control and Law Enforcement Act of 1994, joined a new
partisan consensus that embraced longer prison sentences for more crimes. New forces,
operating at the federal and state level, including interest group activity, and new federal
incentives for states to adapt harsh sentencing regimes helped drive and sustain higher
imprisonment rates. Even though violent crime rates began to decline, new political and
institutional realities sustained continued prison expansion that by the millennium had
generated mass incarceration. This “reconstruction” of a new penal order looked much
more like the penal regimes that had long characterized Sunbelt states, where incarceration rates had always been higher.
Stabilization and Contraction (2001–2010)
Campbell and Schoenfeld (2013) did not elaborate on developments beyond 2001, but
their theory would suggest that the newly entrenched prisons-first ethos would remain
largely intact unless broad socioeconomic forces or political shocks destabilized it. Federal incentives, limited federal court activism, and powerful local interest groups would
help sustain high incarceration rates despite lower violent crime. States with histories
of intense racial conflict would remain staunch defenders of the new order, and their
national-level electoral strength would limit national-level political opportunities to unsettle high incarceration rates.
Developments in incarceration in the 2000s have been important for several reasons,
although they have not been systematically examined against the theoretical expectations
derived from historical work. Changes in several theoretically significant variables in the
1990s and early 2000s have provided a valuable new historical context for understanding
incarceration. Violent crime dropped precipitously beginning in the 1990s, the nation’s
Hispanic population grew sharply, and for the first time in decades the nation’s overall
incarceration rate stabilized and even declined. This decline reflected a notable shift from
decades of increases, but many state incarceration rates still grew, and most state prison
systems were still operating beyond design capacity in 2010 (National Research Council,
2014). The political dynamics associated with crime policy also changed; the War on Terror and the sharp economic decline of the late 2000s established a new political context
that seems less amenable to crime politics. This variation provides an opportunity to examine whether those same forces that shaped incarceration’s rise are also associated with
its stability and decline.

EXPLAINING INCARCERATION IN THE UNITED STATES

187

CURRENT STUDY
Using state-level decennial data from 1970 to 2010, we build on and go beyond existing quantitative analyses by situating our findings against the context of Campbell and
Schoenfeld’s (2013) theoretical expectations about period effects, integrating key concepts from the historical literature into a theoretically driven quantitative analysis of penal change. Importantly, we also go beyond the temporal scope of previous quantitative
analyses and Campbell and Schoenfeld’s theory by incorporating data from 2000 and 2010
to examine whether the constellation of forces associated with incarceration during the
period of prison expansion changed once crime rates dropped and incarceration growth
stabilized and declined. Our work should provide insights into whether the trends in the
importance of race identified by Greenberg and West (2001) and partisanship suggested
by Stucky, Heimer, and Lang (2005) continue beyond their temporal frameworks into
the post-prison boom era. Our analysis also departs from other works by employing a
revised definition of region that groups Sunbelt states. As Lynch (2010) and others have
argued, Sunbelt states were bellwethers of national trends in incarceration, and they became increasingly important in national politics as their populations expanded. We consider whether the factors associated with imprisonment operate differently in states in
this broad region than in other parts of the United States. Sunbelt states are notable because many have histories of intense racial antagonism, and many had political and penal
cultures that did not embrace rehabilitation and had long histories of aggressive law enforcement regimes that disproportionately targeted minority populations (Lynch, 2010;
Perkinson, 2010; Schoenfeld, 2010).
Figure 1 outlines the empirical relationships for key variables for specific periods as defined by Campbell and Schoenfeld (2013). We expect findings from each of our decennial
measures to capture the product of the historical forces that preceded it. Following their
periodization strategy, we expect our findings for 1970 roughly to reflect the outcome
of the processes they describe for the Destabilization period (ß1960–1975), findings for
1980 and 1990 to capture processes unfolding in their Contestation period (1976–1992),
and for 2000 to explain the outcome of events in their Reconstruction period (1992–2001).
We also project how their explanation would apply to the post-2001 era, although their
work did not directly address this period in depth.
As figure 1 illustrates, the expected influence of some explanatory variables changes
over time, whereas others are more stable. Campbell and Schoenfeld (2013) suggested
that race is an important contextual factor in explaining higher incarceration rates. Thus,
we would expect states with larger proportions of minorities to have higher incarceration
rates in each period. Republican Party strength is less stable and should not be correlated with incarceration in 1970, but it should develop a stronger association in 1980 and
1990, as crime becomes increasingly politicized, before declining by 2000, reflecting the
bipartisan consensus on a prisons-first crime policy agenda that emerged in the 1990s.
Violent crime rates should be associated with higher incarceration in the destabilization
and contestation periods before weakening in the Reconstruction period and beyond.
Campbell and Schoenfeld’s (2013) explanation also emphasized historical differences
between Sunbelt and non-Sunbelt states in explaining incarceration. We examine these
differences by analyzing the importance of key variables in Sunbelt and non-Sunbelt
states across periods. They argued that Sunbelt states with histories of higher incarceration rates, large minority populations, and conservative political cultures were especially

188

CAMPBELL, VOGEL, & WILLIAMS

Figure 1. Temporal Expectations of State-Level Developments for
Violent Crime, Partisanship, and Race (Campbell and
Schoenfeld, 2013)
Destabilization (1970)

Contestation (1980 and 1990)

Reconstruction (2000)

Implications for Post-2001

Race (larger minority
populations) will be
positively correlated with
incarceration rates

Race (larger minority
populations) will be positively
correlated with incarceration
rates

Race (larger minority
populations) will be positively
correlated with incarceration
rates

Race (larger minority
populations) will be positively
correlated with incarceration
rates

Partisanship will not be
strongly correlated with
incarceration rates

Partisanship (Republican
strength) will be correlated
with incarceration rates

Partisanship will not be
strongly correlated with
incarceration rates

Partisanship will not be
strongly correlated with
incarceration rates

Violent crime will be
positively correlated with
incarceration rates

Violent crime will be
positively associated with
incarceration rates

Violent crime will not be
strongly correlated with
incarceration rates

Violent crime will not be
positively associated with
incarceration rates

prone to “law and order” politics and were more likely to turn to prison expansion. Therefore, we would expect Republican Party power to be strongly associated with incarceration in Sunbelt states in the contested period as state-level Republicans successfully used
crime to usurp Democratic control of many Sunbelt states. For non-Sunbelt states with
smaller minority populations and histories of penal moderation, we would not expect a
strong relationship between Republican Party power and incarceration. We also examine
periodicity in these key variables across the two regions.
Importantly, Campbell and Schoenfeld’s (2013) model emphasized the importance of
interactions between national and state-level processes in tilting long-term changes in
crime policy and incarceration at the state level. As noted, they argued that federal court
activism, federal programs channeling funding to state systems, and presidential election
cycles altered local political and institutional contexts in ways that privileged a prisonsfirst crime policy ethos. Unfortunately, no effective measures exist to quantify these factors. Although these limitations make a direct assessment of how national-level forces
shaped specific state trajectories impossible, we can assess the degree to which state-level
trends conformed to national-level trends over time and within specific periods. We perform this analysis and consider its implications against our broader findings in the Discussion section.

METHODS
DATA
Consistent with much prior research on the social and political determinants of incarceration, our analyses use state-level decennial data drawn primarily from the U.S. Census
(e.g., Jacobs and Carmichael, 2001). We depart from prior research by extending the lens
of the analysis to 2010, thereby incorporating data from 2000 and 2010. We rely on decennial data as intercensual data on several of the variables, such as racial composition and

EXPLAINING INCARCERATION IN THE UNITED STATES

189

Table 1. Means and Standard Deviations Across Time and Across States
(N = 250)
Variables

Mean
1970

Mean
1980

Mean
1990

Mean
2000

Mean
2010

Overall
SD

Cross
State
SD

Across
Time
SD

Ln incarceration rate
Ln violent crime
Percentage Black
Political ideology
Republican strength
Ln religious fundamentalism
Ln percentage Hispanic
Determinate sentencing
Unemployment
Economic inequality (GINI)
Tax base (Ln mean income)
Urbanicity

4.22
5.27
8.76
44.48
21.35
−2.51
.52
.00
4.84
.35
8.24
66.00

4.84
5.92
9.13
42.44
19.00
−2.70
.78
.14
6.78
.37
9.16
66.86

5.54
6.03
9.52
48.52
15.98
−2.53
1.00
.20
6.36
.38
9.80
68.15

5.87
5.96
9.90
43.29
28.58
−2.65
1.55
.34
3.90
.42
10.25
71.69

5.93
5.87
10.33
47.42
29.73
−2.77
1.99
.34
8.84
.45
10.56
73.57

.78
.63
9.31
15.61
26.41
1.09
1.16
.40
2.29
.04
.84
14.77

.37
.54
9.33
13.89
13.20
1.09
1.00
.29
1.16
.02
.14
14.40

.69
.34
1.02
7.36
22.94
.15
.60
.28
1.98
.04
.83
3.72

ABBREVIATIONS: Ln = log-transformed; SD = standard deviation.

urbanicity, is not available for earlier decades.4 Our analytic sample consists of data on
50 states measured at five time points, generating a pooled sample of 250 state-years. Descriptive statistics are presented in table 1. A full list of sources for the various measures
is included in the online supporting information.5
Dependent Variable: Incarceration Rate
The dependent variable is the size of the incarcerated population, measured as the
number of prisoners under state jurisdiction per 100,000 population as reported by the
Bureau of Justice Statistics. The emerging historical research has indicated that the effects
of partisanship on the size of the prison population may be delayed, as legislation likely
takes several years between passage and implementation (Campbell, 2011; Schoenfeld,
2010). Therefore, measuring the size of the incarcerated population concurrently with
the Republican partisanship of a state could downwardly bias the empirical association.
To account for this possibility, the outcome variable is measured as the average of the
incarceration rate 2, 3, and 4 years beyond the decennial year (e.g., when partisanship was
measured in 1970, the size of the incarcerated population was measured as the average of
1972, 1973, and 1974).6 We use the overall size of the incarcerated population rather than
new admissions to capture state-level policy and administrative processes. Admissions
might reflect the greater propensity of local governments to send offenders to state prison,
but the actual size of the prison population reflects how state actors invest resources in

4.

5.
6.

An alternative approach would be to interpolate annual data points for non-Census years. One
drawback of this strategy is that it assumes a linear trend in the key predictors between time points,
potentially introducing error for non-decennial years.
Additional supporting information can be found in the listing for this article in the Wiley Online
Library at http://onlinelibrary.wiley.com/doi/10.1111/crim.2015.53.issue-2/issuetoc.
Unfortunately, the most recent incarceration data available from the Bureau of Justice Statistics
are from 2010. Therefore, the dependent variable for the 2010 series is the 2010 incarceration rate.

190

CAMPBELL, VOGEL, & WILLIAMS

the long-term maintenance of prison capacity. The variable has been log transformed to
the base e in the regression models to reduce skew and ensure multivariate normality.
Determinants of Incarceration
Drawing from the extant research on the social and political determinants of incarceration, the key predictors in the analyses include measures of social threat, political partisanship, crime, and sentencing policy. We focus specifically on measures of Republican
Party strength, citizen ideology, racial and ethnic composition, violent crime, and religious
fundamentalism.
Following Jacobs and Carmichael (2001), Republican strength is constructed from two
pieces of information: 1) whether the state had a Republican governor (1 = Republican
governor; 0 = non-Republican governor) and 2) the percentage of Republicans in the
state legislature (combining the upper and lower houses). The measure was created by
multiplying the dummy variable for Republican governor by the percentage of Republicans in the legislature. When the observed decennial year was also an election year,
the measure was calculated using the incoming elected official’s party identification. Any
states with a non-Republican governor, regardless of the composition of the state legislature, were assigned a value of zero. States with a Republican governor were assigned a
value reflecting the percentage of Republicans in the legislature.7,8
Citizen ideology is the mean position on a liberal–conservative scale of the current electorate in a state, constructed by Berry et al. (1998). It is measured by identifying the ideological position of each member of Congress in each year using interest group ratings,
followed by the estimation of citizen ideology in each district using an ideology score
for each district’s incumbent, an estimated score for a challenger to the incumbent, and
election results that are assumed to reflect the ideological division in the electorate. The
citizen ideology scores are then weighted to each candidate’s share of support within a
district, followed by the averaging of the scores of each district to compute a liberalism–
conservatism state score. Scores range between 0 and 100, with lower scores reflecting
more conservative states. This variable is measured at each of the decennial years.
7.

8.

The Nebraska state legislature is unicameral, and lawmakers are not officially affiliated with parties. In years when there was a Democratic governor, Nebraska was assigned a value of 0. In years
where there was a Republican governor, Nebraska was assigned a value of 50 percent. This coding
strategy allowed for Nebraska to be retained in the analyses. Subsequent models excluding Nebraska produced nearly identical results to those reported below, making it unlikely this coding
decision unduly influenced the findings (available upon request).
Although this measure of Republican strength is consistent with prior research in this area, it is
not without limitation. The product term assigns any state with a Democratic governor a value
of 0. The clear implication being that the size of the Republican legislature is irrelevant when a
Democrat is in the governor’s seat. Our inclusion of the measure here reflects 1) our decision to
remain consistent with prior research in this area and 2) a desire to retain degrees of freedom
in the regression models. A more precise measure of Republican strength would necessitate a
dichotomous indicator for whether a state had a Republican governor, a measure of the percent
of Republicans in the state legislature, and a product term between the two. This would allow
for the estimation of the effect of Republican control of the legislature in states with a Democratic
governor. Unfortunately, this measure would necessitate the inclusion of an additional 10 variables
in the model (two additional main effects and eight additional product terms). Only 11.6 percent
of all observations (state years) had both Democratic governors and Republican majority control
in the legislature (29 of 250 state-year observations).

EXPLAINING INCARCERATION IN THE UNITED STATES

191

Racial and ethnic composition is measured through two variables—the percentage nonHispanic Blacks and the percentage of Hispanics in each state. Given the skewed distribution of the percentage Hispanic, this variable is log transformed in the regression
models. Violent crime is measured as the natural log of the average rate of assaults, rapes,
homicides, and robberies per 100,000 population in the decennial year, and the 2 years
preceding the decennial year, as reported in the Federal Bureau of Investigation’s Uniform Crime Reports. For example, the violent crime rate for 1970 is the average rate
across 1968, 1969, and 1970. We chose to use the 3-year average because it seems likely
that higher rates of violence might take time to generate sufficient political momentum to
affect electoral outcomes and policy significantly. This measure captures the potential effects of the immediate context and the potential effects of violence in periods immediately
preceding elections and legislative sessions.
Religious fundamentalism measures the number of adherents to fundamental religious
bodies relative to the state population as recorded by the Association for Religious
Data Archives. Note that this measure presents an improvement over prior work as we
can measure fundamentalism at each of the decennial time points. Prior research incorporated measures of fundamentalism measured at only a single time point—1982 (see
Greenberg and West, 2001; Jacobs and Carmichael, 2001).
Control Variables
The regression models control for several factors likely to influence incarceration and
typically included in state-level analyses. Unemployment is the percentage of the population in the labor market and currently seeking employment as reported by Bureau of
Labor Statistics. Urbanicity is measured as the percentage of the population characterized
as living in an urban area. We use this measure of urbanicity rather than percentage of the
population living in metropolitan statistical areas (MSAs) because of 1) missing data for
three states for the MSA variable in 1970 and 2) changes in census definitions of MSAs
in 2000 and 2010 that make comparisons over time difficult. Economic inequality is measured through the GINI coefficient. Higher values indicate a greater degree of economic
inequality. Tax base is the natural log of resident’s mean income for each state. Determinate sentencing is a dummy variable that measures whether a state allows the discretionary
release of prisoners prior to the end of the judge’s sentence. Note that all values in 1970
equal zero because the first determinant sentencing policies were not implemented until
1976 by California and Maine (Stemen, Rengifo, and Wilson, 2006). The 2010 measure of
determinate sentencing was assumed to be the same as it was in 2002 because no state had
repealed the policy since Connecticut did so in 1990 (Stemen and Rengifo, 2011). Based
on previous research (Lynch, 2010), we developed a dummy indicator for Sunbelt states,
which included Alabama, Arizona, California, Florida, Georgia, Louisiana, Mississippi,
Nevada, New Mexico, North Carolina, South Carolina, and Texas.
ANALYTIC STRATEGY
Fixed-effects models are the standard approach to assess multivariate relationships
among pooled data with more panels than time points (Rabe-Hesketh and Skrondal,
2008). Fixed-effects models correct for interdependence over time by incorporating a
series of dummy variables for each state—in essence, holding constant time-invariant
covariates by treating each state as its own control. An alternative strategy is the

192

CAMPBELL, VOGEL, & WILLIAMS

random-effects model. Unlike fixed-effects models, random-effects models require fewer
parameter estimates and allow for the inclusion of time-invariant covariates. Although
both approaches have their relative strengths and weaknesses, the default practice in the
empirical literature is the fixed-effects model (for a detailed discussion of the trade-off
between modeling strategy, see Rabe-Hesketh and Skrondal, 2008).
The empirical analyses unfold in three steps. First, following prior work in this area
(e.g., Greenberg and West, 2001), we estimate a two-way fixed-effects model to assess the
influence of the predictor variables on the size of the incarcerated population while holding constant time-invariant state-level predictors and taking into account historical contingencies in the size of the incarcerated population.9 The second set of analyses assess the
importance of historical contingencies (i.e., periodicity) in the key predictor variables by
incorporating a series of product terms between racial composition, violent crime, Republican Party strength, political ideology, and the year-specific dummies into the baseline regression model.10 Finally, given the historical contingencies in explaining the prominent
role Sunbelt states played in shaping national trends, we assess regional variation in the
key predictors as well as the period effects by estimating separate fixed-effects regression models for Sunbelt and non-Sunbelt states. We examine differences in the predictive
power of the key independent variables by computing a series of Clogg tests to assess
differences in regression coefficients across regions (Paternoster et al., 1998).
Supplemental Regression Models
Campbell and Schoenfeld (2013) suggested that the most important factors driving incarceration rates differed across historical periods. They argued that violent crime was
important in previous periods when crime became politicized, but it mattered less once
a growing consensus emerged around a prisons-first ethos. Importantly, these period effects occurred in different states at different times. Drawing from these arguments, we
perform a series of supplemental analyses to determine whether subjective factors (e.g.,
race and partisanship) driving the incarceration rate interact with objective factors (i.e.,
violent crime) in explaining incarceration trends within states over time. In other words,
these analyses allow us to examine whether partisanship, racial composition, political ideology, and religious fundamentalism have a stronger effect on the size of the incarcerated
population when and where violent crime rates are low. To accomplish this, we estimate
a series of two-way fixed-effects models with product terms for violent crime rate × percentage Black, violent crime rate × political ideology, violent crime rate × religious fundamentalism, and violent crime rate × Republican strength.
Finally, following the recommendation of McDowall and Loftin (2009), we examined
whether incarceration rates in all 50 states followed a national trend during the timeframe under study, or whether meaningful state trends emerged. We accomplished this
in two ways. First, we estimated a simple fixed-effects model with only dummy variables
for the states and then a second model incorporating the year dummies. The difference
in the squared partial correlation coefficient between the first and the second model indicates the amount of variation in incarceration that can be explained by the period dummy
9. Two-way fixed-effects models incorporate dummy variables for both state and time periods.
10. Supplemental analyses (available upon request) indicate that the associations between the remaining predictor variables and the size of the incarcerated population did not vary over time.

EXPLAINING INCARCERATION IN THE UNITED STATES

193

variables, which can be interpreted as the relative contribution of a national trend. We
next examined the proportion of states that followed the national trend during this time
period. To accomplish this, we generated four change scores by subtracting the previous
decennial period from each time period included in the analyses. These reflect withinstate changes in incarceration rates from 1970 to 1980, 1980 to 1990, 1990 to 2000, and
2000 to 2010. At the national level, each of these difference scores was positive, underscoring the well-documented upward trend in U.S. incarceration rates across the last five
decades. We then examined whether all states experienced this same upward trend by
comparing state trajectories with the overall national trajectory.

RESULTS
Table 2 presents the two-way fixed-effects models examining periodicity in the key predictors of incarceration from 1970 to 2010. Model 1 presents the logged incarceration rate
regressed on the key predictor variables and the dummy variables for 1980, 1990, 2000,
and 2010 (with 1970 as the referent). Coefficients for the log-transformed variables can
be interpreted as the expected percentage of change in the size of the incarcerated population for a 1 percent increase in the predictor variable over time. Coefficients for the
nontransformed variables in the model can be interpreted as the expected proportional
change in the size of the incarcerated population for a one-unit increase in the predictor
variable over time. Multiplying these coefficients by 100 yields the expected percentage
change in the size of the incarcerated population given a one-unit increase in the predictor
variable. The results from this model demonstrate that violent crime is positively associated with the size of the incarcerated population; a 1 percent increase in violent crime is
associated with a .17 percent increase in the size of the incarcerated population. The increasing magnitude of the year-dummies, while not statistically significant, highlights the
sharp increase in state prisoners from 1970 through 2010. None of the other predictors
emerge as statistically significant in the model.
With the exception of violent crime, the lack of significance for the remaining covariates
can be interpreted as potential evidence for the importance of historical contingencies—
theoretical determinants are relevant only in specific historical periods. When the periodization effects of key predictors are not considered, as is the case in the baseline fixedeffects model, we might expect the effects of these predictors to be washed out. To evaluate empirically the periodization hypothesis, a series of product terms between the year
dummies and the key predictor variables were introduced to the baseline fixed-effects
model (model 2). In these models, the coefficients for percentage Black, partisanship, and
violent crime can be interpreted as the expected effect of each predictor on the size of the
incarcerated population in 1970. The product terms reflects the strength of these predictors in subsequent decennial years relative to 1970. The expected association between the
year-specific coefficients and the size of the incarcerated population can be gleaned by
adding the year-specific coefficient to the baseline coefficient. The interpretation of the
log and nonlogged coefficients remains the same as those presented in model 1.
Three trends are worthy of note. First, similar to Jacobs and Carmichael (2001)
and Greenberg and West (2001), we found conservative citizen ideology was associated with higher incarceration when controlling for periodization in the key three
predictor variables. In 1970, a one-unit increase in the measure of political ideology is associated with a 1 percent decrease in the expected size of the incarcerated

194

CAMPBELL, VOGEL, & WILLIAMS

Table 2. Two-Way Fixed-Effects Regression of Logged Incarceration
Rates on Key Predictors 1970–2010 (N = 250)
Model 1
Variables
Ln violent crime
Percentage Black
Political ideology
Republican strength
Ln religious fundamentalism
Ln percentage Hispanic
Determinate sentencing
Unemployment
Economic inequality (GINI)
Tax base (Ln mean income)
Urbanicity
Year dummy = 1980
Year dummy = 1990
Year dummy = 2000
Year dummy = 2010
Republican strength80
Republican strength90
Republican strength00
Republican strength10
Percentage Black80
Percentage Black90
Percentage Black00
Percentage Black10
Political ideology80
Political ideology90
Political ideology00
Political ideology10
Ln violent crime80
Ln violent crime90
Ln violent crime00
Ln violent crime10
Intercept
R2

b
.168∗
.013
−.003
.001
.057
−.093
−.045
.007
−.143
.230
.005
.325
.883
1.188∗
1.209
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
2.167

Model 2
(SE)

b

(SE)

(.079)
(.018)
(.002)
(.001)
(.138)
(.060)
(.058)
(.016)
(1.360)
(.291)
(.006)
(.284)
(.460)
(.586)
(.696)
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
(2.512)

.065
−.027
−.010∗∗
−.002
.034
−.084
−.061
−.008
−.216
−.088
.006
.276
.713
1.319∗
1.951∗
.003
.004∗
.004∗
.001
.020∗∗
.015∗
.020∗∗
.015∗
.004
.011∗∗
.006
.000
.208∗
.223∗∗
.094
.010
4.679

(.095)
(.020)
(.003)
(.001)
(.142)
(.060)
(.054)
(.016)
(1.559)
(.324)
(.006)
(.388)
(.518)
(.663)
(.791)
(.002)
(.002)
(.002)
(.002)
(.007)
(.007)
(.008)
(.007)
(.003)
(.004)
(.004)
(.004)
(.084)
(.085)
(.103)
(.113)
(2.711)

.921

.940

NOTE: The em-dash (—) indicates entries that are not applicable.
ABBREVIATIONS: b = coefficient; R2 = the overall square; SE = standard error.
∗ p < .05; ∗∗ p < .01 (two-tailed).

population (–.01 × 100). The effect of political ideology on incarceration was no different
in 1980, relative to 1970. In 1990, the effect of political ideology was effectively null (–.010
+ .011). In the subsequent two decades, the negative effect of political ideology on the
size of the incarcerated population was no different than 1970. This indicates a general
trend in which more liberal states have consistently lower incarceration rates. The time
interactions suggest that the association between political ideology and incarceration was
consistent in 1980, 2000, and 2010 (relative to 1970). However, it was indistinguishable
from zero in 1990 (relative to 1970). Second, we detect no association between religious
fundamentalism and the size of the incarcerated population.
Third, the model presents clear evidence of the importance of historical contingencies
in partisanship, minority population size, and violent crime. For instance, the association
between partisanship and the size of the incarcerated population was stronger in 1990 and

EXPLAINING INCARCERATION IN THE UNITED STATES

195

2000 than in 1970. This finding likely indicates that Republican partisanship had a growing association with incarceration through the 1980s and 1990s, a time when Republican
rhetoric on the war on crime had reached a fever pitch. By 2010, the effect of partisanship
on the size of incarcerated population was no different than in 1970.
The pattern for the size of the Black population and violent crime is a bit more perplexing. Once controlling for the violent crime rate, the relative size of the non-Hispanic
Black population is associated more strongly with incarceration relative to 1970; however, the relative size of the product term is reduced by 25 percent from 1980 to 1990
(in relation to 1970) ([.020 –.015] / .020) and then increases again in 2000 (relative to
1970). In 1980, a one-unit increase in percentage Black was associated with a 2 percent
expected increase in the size of the incarcerated population (.020 × 100). In 1990, the
same increase was associated with a 1.5 percent increase in the size of the incarcerated
population. Notably, violent crime was a much stronger determinant of incarceration in
1980 and 1990 (relative to 1970), a period characterized by the precipitous increase in violent crime across the states. For instance, a 1 percent increase in the violent crime rate
was associated with a .208 percent and a .223 percent increase in the size of the incarcerated population in 1990 and 2000, respectively. By the time violent crime peaked in the
early 1990s, the size of the non-Hispanic Black population was closely linked with violent
crime. The attenuation in the coefficient for percentage Black in 1990 could reflect punitive drug policies that disproportionately affected African Americans (Provine, 2007). By
2000, the violent crime rate had declined to 82 percent of its 1990 level, and its association with incarceration was no different than it was in 1970. Importantly, as violent crime
declined in importance, the strength of the coefficient for minority population size once
again gained magnitude relative to its 1970 level. This might suggest that rates of violent
crime are most strongly associated with incarceration when offending is highest, but attitudes and beliefs about African Americans and crime endure even when offense rates
decline.
On the whole, the results of the fixed-effects models underscore clear periodization in
the determinants of incarceration. Consistent with prior research, these results provide
further evidence that minority population size, partisan politics, and violent crime played
a significant role in shaping the American carceral state from the 1970s through the early
years of the millennium. The introduction of the two additional decennial years reveals
a departure from prior research—although these factors help explain the rise of incarceration, they do little to explain its continual increase in the face of declining crime
rates. By 2010, only the association between the size of the non-Hispanic Black population and incarceration is larger than the association for 1970, the pre-prison boom
benchmark.
Table 3 presents the results of the separate fixed-effects regression models for the Sunbelt and non-Sunbelt states. The models take on the general form presented in table 2.
Model 1 presents the logged incarceration rate regressed on the key predictor variables
and the dummy variables for 1980, 1990, 2000, and 2010 (with 1970 as the referent). Two
interesting trends emerge here. First, the size of the Hispanic population is negatively associated with the size of the incarcerated population in the Sunbelt states. For instance, a
1 percent increase in the size of the Hispanic population is associated with a 0.44 percent
decrease in the overall size of the incarcerated population. Although this effect might
seem trivial, this decrease is roughly equivalent to 9 fewer people in prisons per 100,000
people in this region during the entire time period. For means of illustration, a 1 percent

196

CAMPBELL, VOGEL, & WILLIAMS

Table 3. Two-Way Fixed-Effects Regression of Logged Incarceration
Rates on Key Predictors Between Sunbelt and Non-Sunbelt
States, 1970–2010
Non-Sunbelt States (n = 190)
Model 1
Variables
Ln violent crime
Percentage Black
Political ideology
Republican strength
Ln religious fundamentalism
Ln Percentage Hispanic
Determinate sentencing
Unemployment
Economic inequality (GINI)
Tax base (Ln mean income)
Urbanicity
Year dummy = 1980
Year dummy = 1990
Year dummy = 2000
Year dummy = 2010
Republican strength80
Republican strength90
Republican strength00
Republican strength10
Percentage Black80
Percentage Black90
Percentage Black00
Percentage Black10
Political ideology80
Political ideology90
Political ideology00
Political ideology10
Ln violent crime80
Ln violent crime90
Ln violent crime00
Ln violent crime10
Intercept
R2

b
.219∗
.003
.000
.001
−.109
−.004
−.067
.026
.235
.037
.012
.324
1.045∗
1.392∗
1.314
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
2.988

(SE)
(.090)
(.021)
(.003)
(.001)
(.164)
(.074)
(.070)
(.019)
(1.637)
(.336)
(.008)
(.318)
(.521)
(.665)
(.778)
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
(2.831)
.922

Sunbelt States (n = 60)

Model 2
b
.112
−.058
−.010
−.001
.001
.051
−.058
.007
−.720
−.053
.014
.048
.573
1.008
1.511
.002
.003
.004
−.001
.030∗
.027
.024
.024
.005
.010∗
.008
.002
.143
.201
.152
−.027
3.755

(SE)
(.119)
(.029)
(.004)
(.001)
(.167)
(.075)
(.066)
(.020)
(1.909)
(.382)
(.007)
(.509)
(.625)
(.794)
(.929)
(.002)
(.002)
(.002)
(.002)
(.014)
(.014)
(.014)
(.014)
(.004)
(.004)
(.005)
(.005)
(.122)
(.120)
(.149)
(.161)
(3.120)
.942

Model 1
b
−.016
.019
−.006
.000
.412
−.263∗
−.085
−.029
1.854
.082
−.005
.937
1.463
1.904
2.248
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
3.321

(SE)
(.184)
(.033)
(.005)
(.001)
(.320)
(.107)
(.097)
(.030)
(3.267)
(.616)
(.014)
(.621)
(1.026)
(1.323)
(1.622)
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
(5.736)
.961

Model 2
b

(SE)

.007
(.271)
−.019
(.036)
−.006
(.014)
−.008
(.004)
.210
(.636)
−.439∗∗
(.124)
−.161
(.104)
−.043
(.032)
1.597
(3.405)
−1.351
(.851)
.006
(.015)
1.789
(.984)
2.715
(1.477)
3.859
(1.903)
5.331∗
(2.164)
.021∗
(.008)
.009
(.006)
.010
(.006)
.008
(.007)
.019
(.013)
.017
(.014)
.030
(.019)
.026
(.020)
.001
(.017)
.014
(.017)
.018
(.018)
.000
(.014)
.027
(.357)
.163
(.293)
−.612
(.342)
−.329
(.408)
15.046
(7.948)
.986

ABBREVIATIONS: b = coefficient; Ln = log-transformed; R2 = the overall square of correlation between the
predicted and the observed values; SE = standard error.
∗ p < .05; ∗∗ p < .01 (two-tailed).

increase in the size of the Hispanic population in 2000 was associated with roughly 2,700
fewer incarcerated persons in the Sunbelt states.11 The size of the Hispanic population
had no association with the size of the incarcerated population in non-Sunbelt states. The
difference in coefficients is statistically significant (t = 3.38). At first blush, these trends
might seem counterintuitive, as minority size could be associated with higher incarceration rates. But we suspect this reflects the relationship between growth in the size of the
11. This figure was calculated by multiplying the total number of incarcerated persons in Sunbelt states
in 2000 (n = 612,970) by .0044.

EXPLAINING INCARCERATION IN THE UNITED STATES

197

Hispanic population and their associated political influence. It is important to note that
this finding suggests that the rate of growth in incarceration becomes less steep as the Hispanic population size becomes larger, not that incarceration rates were declining in states
with large Hispanic populations.
Second, the time dummies indicate that incarceration rates in the Sunbelt states are significantly higher in 2010 (relative to 1970) compared with non-Sunbelt states (t = 2.16).
This likely reflects the steadfast approach to incarceration in the Sunbelt states in an era
when many non-Sunbelt states were experiencing stability and reductions in incarceration
rates (most notably in the Northeast). Model 2 in table 3 presents the analysis examining
regional differences in the periodicity of the key independent variables. These results suggest that Republican strength was more strongly associated with incarceration in Sunbelt
states in 1980 relative to 1970 (t = −2.36). There is no evidence of periodicity in Republican strength in non-Sunbelt states. This likely reflects the growing success “law and
order” Republican politicians experienced in Sunbelt states as they increasingly emphasized more aggressive crime policies. No other meaningful differences were found in the
predictor variables between Sunbelt and non-Sunbelt states.
SUPPLEMENTAL ANALYSES
Table 4 presents the results of the supplemental regression models examining the interactions between violent crime rates and political ideology, religious fundamentalism,
percentage Black, and Republican strength.12 Coefficients for the product terms can be
interpreted as the relative strength of these subjective factors when and where violent
crime is high. The results of this analysis reveal two important trends. First, the product
term presented in model 2 indicates that the size of the Black population is more strongly
associated with incarceration rates when and where violent crime is high. This finding
underscores the results presented in table 2, highlighting the linkage between the size of
the Black population and violent crime rates—the size of the Black population seems to
be intricately linked with levels of violent crime in explaining higher incarceration rates.
To be clear, this likely reflects public opinion surrounding violence and the criminalization of the Black population. Unlike the results presented in table 2, these models do not
include the time-specific dummy variables; as a result, they allow for the assessment of
interactions between Black population size and violent crime rates not associated with
the decennial periods. In other words, in states where violent crime and Black population
size increase, regardless of when it is was experienced, there is a concurrent increase in the
size of the incarcerated population. It would seem that race and violence combine to generate a powerful impetus for higher incarceration but that other cultural and institutional
forces can moderate that association.
Finally, we assessed whether the incarceration rates in all 50 states followed the consistent upward trend at the national level. We first regressed the logged incarceration rate
on dummy variables for each state. This generated an R-squared value of .22. We next
added the time dummies to the initial model, producing an R-squared value of .69. The
difference in the R-squared values between equations indicates the amount of variance
uniquely explained by national-level forces. In this case, 47 percent ([.69 − .22] × 100) of
12. Note that each of these variables was mean centered prior to the construction of the product term
to reduce collinearity.

198

CAMPBELL, VOGEL, & WILLIAMS

Table 4. Fixed-Effects Regression Examining Interaction Effects
Between Violent Crime Rates and Percentage Black, Political
Ideology, Republican Strength, and Religious Fundamentalism
on Incarceration Rates 1970–2010 (N = 250)
Model 1
Variables
Ln violent crime
Percentage Black
Political ideology
Republican strength
Ln religious
fundamentalism
Ln percentage Hispanic
Determinate sentencing
Unemployment
Economic inequality
(GINI)
Tax base (Ln mean
income)
Urbanicity
Year dummy = 1980
Year dummy = 1990
Year dummy = 2000
Year dummy = 2010
Percentage Black × violent
crime
Ideology × violent crime
Republican strength ×
violent crime
Religious fundamentalism
× violent crime
Intercept
R2

b

(SE)

.168∗
.013
−.003
.001
.057
−.093
−.045
.007
−.143

Model 2
b
.209∗∗

(.079)
(.018) .008
(.002) −.005∗
(.001) .001
(.138) .042
(.060) −.065
(.058) −.043
(.016) .000
(1.360) .257

.229

(.291)

.005
.325
.884
1.187∗
1.209
—

.153

(.006) .004
(.284) .382
(.460) .970∗
(.586) 1.267∗
(.700) 1.351
—
.014∗

(SE)

Model 3
b

(.079) .147
(.073) .010
(.002) −.003
(.001) .000
(.136) .088

b
.168∗

(.081)
(.018) .013
(.002) −.003
(.001) .001
(.142) .057

(.060) −.108 (.061) −.093
(.057) −.045 (.058) −.045
(.016) .007 (.016) .007
(1.348) −.172 (1.359) −.152
(.288)

.313

(.006) .005
(.281) .266
(.454) .767
(.577) 1.050
(.687) 1.053
(.005) —

—
—

—
—

—
—

—
—

.003
—

—

—

—

—

—

2.167

(SE)

Model 4

(.299)

.228

(SE)

Model 5
b
.171∗

(.080)
(.018) .011
(.002) −.003
(.000) .000
(.138) .056

(SE)
(.079)
(.017)
(.002)
(.001)
(.137)

(.060) −.132∗ (.063)
(.058) −.052
(.058)
(.016) .006
(.016)
(1.375) −.323 (1.355)
(.292)

.360

(.297)

(.007) .005
(.289) .327
(.471) .886
(.596) 1.191∗
(.708) 1.211
—
—

(.007)
(.287)
(.464)
(.590)
(.701)
—

.006
.213
.711
.975
.984
—

(.007)
(.289)
(.466)
(.593)
(.702)
—

(.003)
—

—
.000

—
(.001)

—
—

—
—

—

—

—

–.078

(.042)

(2.512) 2.687 (2.481) 1.456 (2.587) 2.180 (2.532) 1.074 (2.565)
.842
.844
.828
.842
.815

ABBREVIATIONS: b = coefficient; Ln = log-transformed; R2 = the overall square of correlation between the
predicted and the observed values; SE = standard error.
∗ p < .05; ∗∗ p < .01 (two-tailed).

the state variation in incarceration rates over time can be attributed to national factors,
indicating a reasonably strong degree of concurrency between national and state-level
trends.
Next, we examined the proportion of states that followed the national trend between each of the periods. As presented in table 5 from 1970 to 1980, only one state
(North Dakota) experienced a decrease in incarceration. From 1980 to 1990, all 50 states
experienced increased incarceration rates. From 1990 to 2000, two states (Alabama and
Delaware) experienced slight declines. Overall, during these three periods, most states
experienced incarceration trends consistent with the national trend. However, from 2000
to 2010, 17 states (32 percent) experienced declines in their incarceration rates, suggesting
a clear departure from the national trend characterizing the United States through most
of the last half-century.

EXPLAINING INCARCERATION IN THE UNITED STATES

199

Table 5. Correspondence Between State and National Incarceration
Trends (N = 250)
Year

Percentage of States Going Against Trend

1970–1980
1980–1990
1990–2000
2000–2010

2 percent
0 percent
6 percent
34 percent

DISCUSSION
Our findings identify important strands of change and continuity in understanding the
historical arch in imprisonment in the United States—race was and remained an important predictor of higher incarceration, but the relative power of other factors changed
over time. Although previous quantitative analyses extending back to 1970 identified a
strong relationship between violent crime and incarceration, our findings suggest that this
relationship became increasingly important as violent crime increased in 1980 and 1990,
but it declined relative to 1970 and was no longer powerful as crime declined by 2000.
Our findings also help address unresolved issues regarding the importance of partisanship
(Greenberg and West, 2001; Jacobs and Carmichael, 2001; Smith, 2004; Spelman, 2009);
Republican Party strength was important in Sunbelt states, but it did not have a significant effect outside of the Sunbelt, and its explanatory power increased until 1990 before
dropping to pre-prison boom (1970) levels by 2000. Unlike crime and partisanship, race
remained important. The relationship between the percentage of the non-Hispanic Black
population and incarceration is the only factor exhibiting a more powerful effect than it
did in 1970.
Our findings on race could reflect the persistence and hardening of attitudes about
African Americans and crime regardless of crime’s decline throughout the latter 1990s,
which was even more pronounced for Blacks than for Whites (Black male homicide rates
declined by 52 percent and Whites by 39 percent between 1991 and 2000) (Cooper and
Smith, 2011). Despite the sharp decline in offending among Blacks, the size of the Black
population seems to have continued to operate as a proxy for real offending. Greenberg
and West (2001) used Skolnick’s (1966) phrase “symbolic assailants” to explain similar
results in their analyses, and the phrase seems just as relevant in the post-prison boom,
lower crime era. This phrase suggests that race was particularly important in sparking,
driving, and sustaining political demands for more incarceration even when violence declined, especially in politically conservative contexts.
This result provides strong support for Alexander’s (2009) argument that mass incarceration must be understood against a long-term view of the nation’s racial history. She
argued that mass incarceration’s rise and persistence reflects a new post–Civil Rights
era of “colorblind” racism where law, policy, and practice are officially race neutral but
operate in clearly discriminatory and oppressive ways. Historical case studies, quantitative analyses, and the starkly disproportionate numbers of African Americans incarcerated in the United States all confirm race’s central role in understanding punishment
in America. Complex processes continue to operate within the nation’s criminal justice
systems and socioeconomic structure that continue to marginalize African Americans
disproportionately.

200

CAMPBELL, VOGEL, & WILLIAMS

Although our decennial data make precise measures of the evolving influence of the
Hispanic population impossible, these results suggest that the percentage of the Hispanic
population is associated with lower incarceration rates in Sunbelt states and that this
association differed significantly across our two regions (with no effect in non-Sunbelt
states). Future research could use annual data to explain more fully the evolving effect
that the growing Hispanic population might play in understanding state-level incarceration rates. States with large Hispanic populations might prove less willing to invest
state resources in prisons. Conversely, the growth in a state’s Hispanic population might
energize political opposition to their presence and power. Much work remains to be
done in understanding how these powerful demographic and social developments altered
political processes over time that affected crime policy, especially in Sunbelt states where
it seemed to matter most.
Our findings regarding trends in partisanship and our use of Sunbelt versus nonSunbelt states might help untangle the sometimes-conflicting findings regarding the importance of partisanship in explaining incarceration (Greenberg and West, 2001; Jacobs
and Carmichael, 2001; Stuckey, Heimer, and Lang, 2005). As Stucky, Heimer, and Lang
(2005) suggested, partisan competition might be more important than the mere presence
of Republican lawmakers in state capitols in explaining when Republican Party strength
was associated with higher levels of imprisonment. Our findings suggest that this effect
might have been especially important in Sunbelt states when incarceration rates first began to increase (captured in our 1980 findings).
These findings fit well with historical research highlighting the importance of Sunbelt
political cultures in explaining mass incarceration. Lynch (2010) suggested that Sunbelt
states share common strands of penal and political histories that have made them more
prone to harsher penal regimes and political tendencies that marginalized and oppressed
minority populations. As Campbell and Schoenfeld (2013) suggested, these factors seem
to have become increasingly manifest in the partisanship of Sunbelt states as Republican
Party candidates were increasingly linked to crime politics. We find that, relative to 1970,
Republican strength was an important predictor in Sunbelt states but not in non-Sunbelt
states in 1980. This finding suggests that it might not simply be partisanship that helps
explains changes in incarceration that unfolded in the 1970s, but a particular brand of
Republican partisanship that was forged in Sunbelt states when the nation’s “law and
order” movement gained momentum that matters most.
These results support key aspects of Campbell and Schoenfeld’s (2013) argument that
the most important factors in explaining mass incarceration’s rise and persistence are period specific. The relative importance of race, violent crime, and partisanship ebbed and
flowed across historical periods. As Campbell and Schoenfeld suggested, objective factors
that should explain incarceration—violent offense rates—no longer do so once a prisonsfirst ethos becomes a reality across most states and the nation by 2000. Partisanship had
a stronger effect when crime was most highly politicized in the increasingly politically important Sunbelt states, but the influence of partisan politics had decreased to pre-prison
boom (1970) levels by the twenty-first century. Also, Campbell and Schoenfeld might
have overestimated the effect of crime’s politicization and its utility for Republicans in
non-Sunbelt states where we found no effect for partisanship. And although race operated in the expected direction that Campbell and Schoenfeld suggested, its importance
seems more pronounced than their account assumed. Whereas they specifically cite race’s
importance in shaping political contexts favorable to harsher policies, our findings seem
to require a more central role for race. Future research testing the implications of their

EXPLAINING INCARCERATION IN THE UNITED STATES

201

model will hopefully better explain the socioeconomic, political, and institutional mechanisms that facilitate the tight link between Blacks and incarceration in the United States.
Although our findings for 2010 must be taken with some caution, we believe that they
suggest that the factors that explained incarceration’s rise do no better job of explaining state patterns in 2010 than they did in 1970, prior to the prison boom era. These
findings reflect new social and political patterns that have begun to destabilize the once
entrenched prisons-first ethos and the policy responses that supported it (i.e., the war on
drugs, mandatory sentencing, and truth-in-sentencing). The sharp and prolonged drop
in violent offense rates that began in the early 1990s, and the shifting federal focus to
the War on Terror might have generated windows of opportunity for reform. The results
from our supplemental analyses show that a growing number of states departed from the
overall upward national trend from 2000 to 2010, which might signal a new period of
state-level penal policy experimentation emphasizing alternatives to prison.
As Campbell and Schoenfeld (2013) suggested in explaining incarceration’s rise, interactions between state and federal policy developments were important in generating
momentum that supported prison expansion. Our findings from national versus statelevel trends suggest some states have again deviated from national-level developments.
This could suggest that state-level policy experimentation, this time toward decarceration, might again be ushering in a new period of destabilization. State-level reforms legalizing marijuana have gained momentum, and states like New York have drastically decreased their prison populations by diverting low-level offenders. U.S. Attorney General
Eric Holder’s recent announcement that he was encouraging federal prosecutors to leave
smaller drug offenses to local criminal justice authorities seems to follow these trends that
have already begun in the states. Federal courts have forced California to reduce its prison
population drastically because the state has failed to provide inmates with basic medical
care. Importantly, Republican lawmakers have not raised highly charged objections to
reform in some states, and some conservative interest groups have advocated for reform,
citing concerns about costs (Blakeslee, 2014). Whether this is a temporary departure or
an historical shift remains to be seen, but new historical realities pose serious challenges
to the once-entrenched prisons-first ethos that emerged in the latter twentieth century.

REFERENCES
Alexander, Michelle. 2009. The New Jim Crow: Mass Incarceration in the Age of Colorblindness. New York: The New Press.
Barker, Vanessa. 2009. The Politics of Imprisonment: How the Democratic Process Shapes
the Way America Punishes Offenders. New York: Oxford University Press.
Beckett, Katherine. 1997. Making Crime Pay: Law and Order in Contemporary American
Politics: Studies in Crime and Public Policy, eds. Michael Tonry and Norval Morris.
Oxford, U.K.: Oxford University Press.
Beckett, Katherine, and Bruce Western. 2001. Governing social marginality: Welfare,
incarceration, and the transformation of state policy. In Mass Imprisonment: Social
Causes and Consequences, ed. David Garland. London, U.K.: Sage.
Berry, William D., Evan J. Ringquist, Richard C. Fording, and Russell L. Hanson. 1998.
Measuring citizen and government ideology in the American States, 1960–93. American Journal of Political Science 42:327–48.
Blakeslee, Nate. 2014. Why fewer prisons are better for Texas’s economy. Texas Monthly.
March 14.

202

CAMPBELL, VOGEL, & WILLIAMS

Bureau of Justice Statistics. 2012. Prisoners in 2010, Series NCJ 236096 and earlier reports.
Washington, DC: Bureau of Justice Statistics.
Campbell, Michael C. 2011. Politics, prisons, and law enforcement: An examination of the
emergence of “law and order” politics in Texas. Law & Society Review 45:631–65.
Campbell, Michael C. 2014. The emergence of penal extremism in California: A dynamic
view of institutional structures and political processes. Law & Society Review 48:377–
409.
Campbell, Michael C., and Heather Schoenfeld. 2013. The transformation of America’s
penal order: A historicized political sociology of punishment. American Journal of
Sociology 119:1375–423.
Cooper, Alexia, and Erica Smith. 2011. Homicide Trends in the United States,
1980–2008. Washington, DC: Bureau of Justice Statistics. http://www.bjs.
gov/index.cfm?ty=pbdetail&iid=2221.
Garland, David. 2001. The Culture of Control: Crime and Social Order in Contemporary
Society. Chicago, IL: University of Chicago Press.
Gilmore, Ruth Wilson. 2007. Golden Gulag: Prisons, Surplus, Crisis, and Opposition in
Globalizing California. Berkeley: University of California Press.
Goodman, Phillip. 2012. Another second chance. Social Problems 59:437–58.
Gottschalk, Marie. 2006. The Prison and the Gallows: The Politics of Mass Incarceration
in America. Cambridge, U.K.: Cambridge University Press.
Greenberg, David F., and Valerie West. 2001. State prison populations and their growth,
1971–1991. Criminology 39:615–53.
Hagan, John. 2010. Who Are the Criminals? The Politics of Crime Policy From the Age of
Roosevelt to the Age of Reagan. Princeton, NJ: Princeton University Press.
Jacobs, David, and Jason T. Carmichael. 2001. The politics of punishment across time and
space: A pooled time-series analysis of imprisonment rates. Social Forces 80:61–91.
Jacobs, David, and Aubrey L. Jackson. 2010. On the politics of imprisonments: A review
of systematic findings. Annual Review of Law and Social Science 6:129–49.
Lynch, Mona. 2010. Sunbelt Justice: Arizona and the Transformation of American Punishment. Critical Perspectives on Crime and Law. Stanford, CA: Stanford Law Books.
McDowall, David, and Colin Loftin. 2009. Do U.S. city crime rates follow a national
trend? The influence of nationwide conditions on local crime patterns. Journal of
Quantitative Criminology25:307–24.
National Research Council. 2014. The Growth of Incarceration in the United States: Exploring Causes and Consequences. Washington, DC: The National Academies Press.
Page, Joshua. 2011. The “Toughest Beat:” Politics, Punishment, and the Prison Officers
Union in California. New York: Oxford University Press.
Paternoster, Raymond, Robert Brame, Paul Mazerolle, and Alex Piquero. 1998. Using the correct statistical test for the equality of regression coefficients. Criminology
36:859–66.
Perkinson, Robert. 2010. Texas Tough: The Rise of America’s Prison Empire. New York:
Metropolitan Books.
Provine, Doris Marie. 2007. Unequal Under Law: Race in the War on Drugs. Chicago, IL:
University of Chicago Press.
Rabe-Hesketh, Sophia, and Anders Skrondal. 2008. Multilevel and Longitudinal Modeling Using Stata, 2nd ed. College Station, TX: Stata Press.

EXPLAINING INCARCERATION IN THE UNITED STATES

203

Schoenfeld, Heather. 2009. The Politics of Prison Growth: From Chain Gangs to Work
Release Centers and Supermax Prisons, Florida, 1955–2000. Evanston, IL: Department
of Sociology, Northwestern University.
Schoenfeld, Heather. 2010. Mass incarceration and the paradox of prison conditions litigation. Law & Society Review 44:731–68.
Skolnick, Jerome H. 1966. Justice Without Trial: Law Enforcement in Democratic Society.
New York: Wiley.
Smith, Kevin. 2004. The politics of punishment: Evaluating political explanations of incarceration rates. Journal of Politics 66:925–38.
Spelman, William. 2009. Crime, cash, and limited options: Explaining the prison boom.
Criminology & Public Policy 8:29–77.
Stemen, Don, and Andres F. Rengifo. 2011. Policies and imprisonment: The impact of
structured sentencing and determinate sentencing on state incarceration rates, 1978–
2004. Justice Quarterly 28:174–201.
Stemen, Don, Andres F. Rengifo, and James Wilson. 2006. Of Fragmentation and Ferment: The Impact of State Sentencing Policies on Incarceration Rates, 1975–2002.
Washington, DC: U.S. Department of Justice.
Stucky, Thomas D., Karen Heimer, and Joseph B. Lang. 2005. Partisan politics, electoral competition and imprisonment: An analysis of states over time. Criminology 43:
211–48.
Violent Crime Control and Law Enforcement Act, 42 U.S.C. § 1414 (1994).
Walmsley, Roy. 2011. World Prison Population List, 9th ed. London, U.K.: International
Centre for Prison Studies.
Wilson, James Q. 1975. Thinking About Crime. New York: Basic Books.
Zimring, Frank, and David T. Johnson. 2006. Public opinion and the governance of punishment in democratic political systems. Annals: American Association of Political and
Social Science 605:311–37.
Michael C. Campbell is an assistant professor in the Department of Criminology and
Criminal Justice at the University of Missouri–St. Louis. His research examines the socioeconomic, political, and historical forces that shape law and policy, with a particular
focus on incarceration in the United States.
Matt Vogel is an assistant professor in the Department of Criminology and Criminal
Justice at the University of Missouri–St. Louis. His research examines the social ecology
of adolescent behavior and the contribution of population dynamics to crime.
Joshua H. Williams is a PhD student at the University of Missouri–St. Louis. His current research interests include pretrial incarceration, court processing, and politics and
punishment.

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article at
the publisher’s web site:
Table S.1. Sources of Data