Crime is widely recognized as a central issue of public concern in American society. As a result, elected leaders spend significant political capital cultivating policies to address the crime problem. Quite often the proposed solutions offered by elected leaders boil down to escalation of punishment through longer prison sentences and harsher penal conditions. Presumably, if implemented, these proposals would arrest the crime rate and reduce mass incarceration over time. During the past two decades, however, an interesting paradox has developed surrounding crime and the political responses to crime. American states have witnessed a significant drop in violent crimes such as murder, rape, assault, and robbery [1,2]. Yet their total imprisonment rates have continued to rise. This development creates an interesting theoretical puzzle for social scientists.

A simple look at recent crime trends alongside aggregate prison population data tells the story succinctly. Table 1 shows the overall decreasing crime trend in the American states during the first term of George W. Bush administration. On average, assaults per 100,000 population decreased by 5 percent from 271.5 in 2001 to 257.4 in 2004. Similarly, robberies dropped 8.5 percent from 110 per 100,000 in 2001 to 100.6 per 100,000 in 2004. At the same time, incidence of the most feared predatory crimes, murder and rape, remained flat resulting in a net decrease of 5.5 percent in violent crime overall (Table 1). A more detailed look at the trend for various crimes is presented in Figure 1.

Table 1: Change in Violent Crime and Incarceration Rate Per 100,000 Population in the American States, 2001-2004.

Figure 1: Data source: U.S. Department of Justice, Bureau of Justice Statistics

What is particularly interesting is the apparent paradox that has emerged: As the overall violent crime rate continued to drop during that period, the total prison population witnessed an increase of approximately 4 percent, from 380 prisoners per 100,000 populations in 2001 to 394 per 100,000 populations in 2004, thus continuing the prison boom that started in the 1960s. Addressing this inconsistency is the centerpiece of this article. The growing calls to reduce prison overcrowding, including the U.S. Supreme Court’s opinion in Brown v. Plata [3] addressing severe adverse impact on inmates’ constitutional rights, make this paradox an important issue of public concern. Tackling this inconsistency would give us a window into the effectiveness of political accounts of the crime problem. Furthermore, it would help political leaders make more informed choices that can improve criminal justice policy and reduce prison overcrowding without compromising public safety. Although aggregate prison population has been growing steadily over the last several decades and a wealth of literature has been produced to account for it [4,5] our interest in this study lies with recent years, specifically 2001 to 2004. We focus on this time period because of availability of appropriate data and because this was the period when American states witnessed a significant drop in violent crime rates and a continued rise in imprisonment rates.

In this article, we propose a theory arguing that gubernatorial rhetoric is a strong predictor of mass incarceration boom. Our theory relies on the agenda-setting power of the governor and hierarchical power relations between the governor and state employees who implement criminal justice policy. Collectively, these make up the governor’s institutional power. We find strong support for our theory. In addition, we compare the explanatory capacity of our rhetorical model with other theories of punishment to develop a more comprehensive explanation of mass incarceration in the American states.

Theories of Mass Incarceration

Social scientists have proffered various theories to explain increases in the rate of mass incarceration. Among these are the instrumental perspective [6-10] and conflict-based theories that focus on the structure of political and social arrangements [11-14].

Instrumentalism

The instrumental framework is a mechanical theory that postulates a direct and proportional correspondence between the crime rate and the swelling of American state prisons. Under the instrumental perspective, an increase in criminal activity causes tremendous public fear and anxiety. To assuage these fears, policymakers respond by passing increasingly punitive measures such as three-strikes and other forms of punishment enhancement legislation that leads to prison population growth. Indeed, one proponent of this approach, Yair Listokin [7] maintains that under the instrumental framework, “incarceration rates should move in direct proportion to the crime rate” p. 184. Thus in its purest sense, an elasticity of one is implied by the instrumental theory of mass incarceration. A one percent change in the crime rate should lead directly to a one percent change in the prison population. Despite this predicted linearity, studies investigating the connection between crime rates and mass incarceration report inconsistent findings, showing either very anemic effects of crime rate on incarceration [6,15,16] or no effects at all.[4,9,10,17,18]

Some researchers point to methodological problems as the cause of these inconsistent findings. For example, Listokin [7] blames these inconsistent findings on failure of previous researchers to account for endogeniety bias. Endogeniety bias is a problem that occurs because of omitted confounder variables and simultaneity between the predictor and outcome variables. Incapacitation is one consequence of incarceration that can lead to endogeniety problems. Addressing this issue indirectly, Wilson [19] asserts that “when criminals are deprived of their liberty, as by imprisonment (or banishment or very tight control in the community), their ability to commit crimes against citizens is ended.” Some scholars have empirically validated Wilson’s assertion, showing that the number of prisoners limits the amount of crime [20]. As a result, treating the crime rate as an exogenous variable in a regression of crime rate on imprisonment rate (without any correction for endogeniety) is inappropriate and can cause incorrect estimates and wrong conclusions.

Minimizing endogeneity involves using lagged crime rate, lagged incarceration rate, or instrumental variables as regressors when modeling incarceration. In addressing whether more crime means more prisoners, Listokin [7] corrected for endogeniety by including a one-period lag of the crime rate in the model and by using abortion rate as an instrumental variable replacing the crime rate. Listokin [7] reported a strong positive relationship between crime and prison admission rate, thus supporting instrumentalism. Listokin’s analysis is useful for our understanding of incarceration; however, it raises key concerns. For example, Listokin [7] indicated that “the year effects are all positive” including, apparently 1994 to 1997, years in which the violent crime rate dropped measurably. It would seem reasonable that such a crime drop should register some contemporaneous decline in the prison admission rate for some (if not all of these years). Second, the reliance on the prison admission rate (with arbitrarily assigned weights for violent and property crimes) as the dependent variable rather than the prison population rate does not adequately capture the fact that violent crime is punished with much longer incarceration spells than property crime. Finally, criminal punishment is inherently a political issue and the lack of any political control variables in Listokin’s analysis raises credible doubts about model specification.

Political conflict

Beyond the instrumental framework, alternative theories of mass incarceration have been proposed and tested. Political sociologists have offered conflict-based explanations bearing upon the structure of macroeconomic and social arrangements. These theoretical explanations of mass incarceration mirror the social construction of target groups in the articulation, design, and implementation of public policy [21]. Conflict theories emphasize group threat to social and economic arrangements posed by racial minorities such as blacks [22,23] and members of what Hagan and Albonetti [24] labeled the “surplus population.” According to Schneider and Ingram [21], law is a tool of the ruling class and so the social construction of target groups affects policy design, institutional structure, and policy implementation in ways that disadvantages groups that are politically weak and negatively constructed. Escalation of punishment through longer incarceration spells and other punitive measures is one important mechanism for controlling the behavior of these negatively constructed target groups and minimizing their menace to society [25].

Within this vein, political scientists such as Scheingold [26], Wilson [19] and, more recently, Smith [14] and Yates and Fording [27] have emphasized differences in ideological orientation among the major political parties or their standard bearers toward the crime problem as a plausible reason for the rise in mass incarceration. For similar emphasis in sociology [28,29]. In their detailed comparative analysis of incarceration of black and white offenders, for example, Yates and Fording reported a strong connection between state government partisanship and ideology and the percent of state population that is black, with conservative governments tending to incarcerate blacks more than whites as the black population density in the state increases. Racial imbalance in the prison population is an enduring phenomenon that has been documented for both state and federal prison systems as a policy consequence of social control strategies employed by political leaders [30-32]. Racial imbalance in prison is thus a manifestation of the structural explanation of mass incarceration.

Prison population varies widely across the American states, even when normalized by the size of state civilian populations. While many states’ prison populations have continued to grow, some states have actually witnessed a decline in recent years. Part of the reason for the variance in incarceration rate across states is captured by Meier [31] who asserts that the “politics of sin” is driven primarily by legislators’ misinterpretation of public desire for greater punitive policies. Because of strong desire to appear tough on crime and to win over the electorate, policy makers in many states continually champion legislation that tinker with sentencing guidelines and increasingly escalate the severity of punishment, leading to an overall swelling of prisons in the American states.

While instrumental and conflict-based explanations have indeed advanced our understanding of criminal punishment in American society, we think that they overlook a central element in state politics and policies: the governors’ political rhetoric when responding to crime. Opinion polls point to crime being a salient political issue [33,34] one where state governors and legislators stake out policy positions in the hope of claiming credit and advancing their own electoral prospects with voters [35,36]. As a result, gubernatorial rhetoric often centers on crime---its levels, socially destructive and threatening nature and, most importantly, the need for aggressive measures to lessen its incidence and impact on communities. We emphasize here the role of gubernatorial rhetoric as a process instrument contributing to mass incarceration in the states.

Using mass incarceration data from 2001 to 2004, along with data derived from content analysis of state of the state addresses in all 50 states, we uncover evidence indicating a strong link between the crime rhetoric of governors during their state of the state addresses and the continuing increase in state prison populations. This link is profoundly interesting to say the least and it has never before been empirically described and tested in the literature. We test it here for the first time alongside other theories and discuss its implications.

Gubernatorial rhetoric and mass incarceration

State employees in law enforcement, members of parole boards, prosecutors, judges and citizens all have perceptions about the level of crime in their states and what should be done about it. These perceptions are informed by cues and information derived from several sources, including communication from superiors in the workplace, official government reports, mass media, and personal experiences. But most importantly, state employees and citizens receive cues and information about crime and the justice system from the state governor through the annual or biennial state of the state addresses.

Through these addresses, governors communicate a multiplicity of policy visions, including objectives about solving the crime problem. These objectives are typically embedded in messages that are not only absorbed by state workers but affect their orientation and decisional patterns. Gubernatorial rhetoric sends signals about what government officials are supposed to do, identifies which citizens are deserving of protection (hardworking and law-abiding citizens), which are not (sexual predators, murderers, and thugs) and what type of attitudes and decisional patterns are appropriate and would be rewarded in the delivery of safety to citizens. Absorption of the governor’s message about crime and punishment is likely because of the authority and control inherent in being governor.

Across the United States, most state constitutions require the governor to pay a regular visit to the General Assembly at the beginning of each legislative session to deliver a speech about important matters confronting the state and, working with legislators, to develop programs to solve policy problems. Scholars agree that this speech presents the governor with a unique opportunity to make appeals directly to state employees, attentive voters, and the legislature about the administration’s agenda regarding salient policy issues such as crime and public safety [37,38]. The following excerpts are a sampling of statements about crime and punishment from gubernatorial state of the state addresses in recent years. They indicate that governors often employ aggressive political rhetoric to signal the need for punishment escalation:

• Governor Don Siegelman (D-AL):

“For repeat murderers and sexual predators, the guesswork will now be gone. You will serve your full sentence. We will put you away, where the Parole Board can’t let you out. Repeat violent criminals and sexual predators will serve 100 percent of their sentence...no ifs...ands...or buts. We will lock you up and throw away the key. Help me make this the law of Alabama.” (February 6, 2001)

• Governor Jeb Bush (R-FL):

“Public safety, too, has been protected, and convicted criminals will continue to serve at least 85 percent of their sentences. Over the last two years, the violent gun crime rate is down by more than 25 percent, translating into 18 fewer gun assaults each day in this state in 2000 compared to 1998.” (January 22, 2002)

• Governor George Pataki (R-NY):

“Let’s pledge...to make New York the safest of any state in America. To achieve this goal we will need both administrative and legislative actions. First, we will strategically target areas in the State where crime is disproportionately high. Operation IMPACT, or Integrated Municipal Police Anti-Crime Teams, will draw upon all of our State criminal justice resources and consist of over 300 State Police officers. IMPACT Units will be mobilized at the request of local officials and will work with local police and community leaders to combat crime on an unprecedented scale.” (January 7, 2004)

• Governor Jim Gilmore (R-VA):

“Virginia is not for Criminals...We have sent a clear message across the Commonwealth: illegal drugs are not an acceptable part of our society...In its first full year of operation, Virginia Exile has put 95 violent, gun law violators in prison for at least five years. Those convictions, out of 111 cases, produced an 86 percent conviction rate. Virginia Exile’s success has attracted wide attention, from other states, major cities, and in Congress, which approved a national Exile law based on our model.” (January 10, 2001)

As these statements show, governors prefer aggressive rather than soft language when discussing crime and punishment. By what causal process does gubernatorial rhetoric influence mass incarceration? We think the answer lies in the agenda setting role of the governor and in the hierarchical power relations between the governor and state workers who implement various aspects of criminal justice policy.

Agenda setting

The governor is clearly the most important spokesperson and agenda setter in state government. The incumbent plays a prominent role in controlling the budget, crafting state policy, and formulating the overall vision of government based on perceived concerns of citizens [38]. Scholars have described the importance of gubernatorial leadership in agenda setting as “chief legislator” [39-41]. In that sense, the choice of rhetoric allows the governor to frame issues in a way that maximizes impact on legislators as well as policy administrators and rank and file employees. Because crime is a valence issue about which citizens report heightened personal vulnerability, citizens expect their governor to monitor crime levels and implement measures to keep them and their communities safe. This expectation is not unlike expectations placed upon presidents regarding national economic performance [42-45] or upon governors regarding state economic conditions [46,47]. Governors take these citizen expectations seriously because they can be held accountable. This explains why governors speak frequently about crime and employ aggressive rhetoric to frame their responses during their state of the state addresses. Framing is important because it structures expectations [48,49] and establishes a common language—including catchphrases, metaphors, and imagery—which serves as a reference point upon which implementing populations such as police officers, judges, and parole boards can rely when deciding about crime and incarceration policy.

Hierarchical power relations

In our theory, hierarchical power relations between the governor and state workers are important in harnessing the potential effect of gubernatorial rhetoric on incarceration rate under the assumption that as the governor delivers the address, state workers responsible for implementing the governor’s crime fighting initiatives are taking notice. We think that through hierarchical power relations, the governor’s communication of criminal justice objectives becomes embedded in messages that are internalized by state workers and affect their orientation and decision making. This presumption of message internalization by state workers and influence on decision making is validated through the growing power inherent in the role function of the governor as the popularly elected head of state government. In the words of organizational theorists, the governor’s rhetoric is given credence and can generate obedience and responsiveness because of “legitimizing principles” that justify the call for action [50]. These principles are included in the governor’s institutional power—the ability to mobilize and reorganize state resources, manipulate incentive and reward systems to solve important policy problems, hire and fire state workers, etc [51].

Most scholars agree that governors enjoy the power of the bully pulpit [52-54]. Like the attention paid by the national media to the State of the Union addresses delivered by U.S. presidents, local media widely report on governors’ addresses from the State Assembly, followed by analysis and commentary by experts and pundits. While not all citizens comprehend the full details of the governors’ proposals, the broad contours of the appeal and the tone of the governors’ message [what Beckett [18] calls the “interpretive dimension” of the political construction of the crime problem], we contend, is usually well understood. For example, it is not inconceivable that state workers listening to the governor would develop an understanding that the governor seeks to crack down on “violent thugs” who engage in predatory criminal behavior and are deserving of escalated punishment. Similarly, listeners may internalize the message that the governor is developing proposals to collaborate with federal officials to address the potential threat of home grown terrorism and, importantly, is seeking the help of state workers to achieve that goal. Therefore, the state of the state address is a signaling mechanism that communicates the general tone of the governors’ programs and intentions of government involvement in fighting crime, keeping perpetrators locked up, and protecting citizens.

Whether governors’ tough rhetoric is sufficiently persuasive to affect mass incarceration remains an empirical question. We are not aware of any study that has addressed it. What we do know is that state of the state addresses do have direct impact on the policy process. Ira Sharkansky’s [55] classic work on appropriations requests shows that policy priorities enumerated in state of the state addresses were heavily relied upon by agency administrators when making budget requests to the legislature. From an administrator’s perspective, it is a matter of immense strategic advantage to have the governor taking the lead to push an agenda that brings benefits and visibility to the agency. In the area of education policy, Herzik [40] reported that gubernatorial rhetoric on education delivered during state of the state addresses in the 1970s had a reliable influence on the adoption of innovative education policies across several states. At the presidential level, Whitford and Yates [56] present evidence that presidential rhetoric concerning the war on drugs had a significant impact on drug policy enforcement and implementation. Cohen [57] used data on presidential rhetoric in state of the union addresses and reported that increased presidential attention to economic, foreign, and civil rights policies leads to increased public attention to these issues. Similarly, Hill [58,59] used data from state of the union statements, Gallup polls, and various indicators of real world influences and reported that the representational linkage is unidirectional for civil rights policy but bidirectional for foreign and economic policies. Wood, Owen, and Durham [43] found that presidential rhetoric, operating through indirect channels of the news media, exerted substantial impact on subsequent macroeconomic performance. Thus, there is strong empirical evidence at both state and national levels that political rhetoric of governors and presidents affects the content and implementation of government policy---a relationship that exists above and beyond the ideological bent of the chief executive. Based upon the theoretical premise we have presented, we formulate two main hypotheses about gubernatorial rhetoric. First, we hypothesize that gubernatorial rhetoric about getting tough on crime translates directly into an increase in mass incarceration rate.

Hypothesis 1: Gubernatorial rhetoric during the state of the state address about the need to combat crime translates directly into an increase in mass incarceration rate.

We recognize, however, that gubernatorial rhetoric on crime might exhibit codependent explanations. In particular, the effect of gubernatorial rhetoric could be contingent upon other factors, such as the governor’s institutional power to prioritize policy initiatives. To investigate this possibility, we rely on the composite measure of institutional power of the governor developed by Thad Beyle, [60] which includes component ratings on governor’s tenure, power of appointment, control over state budget, organizational power, and veto power. According to an analysis of policy prioritization communicated through state of the state addresses of 14 states from 1971-1990, Gosling [60] reported that governors with high institutional power are inclined to pursue their policy priorities through the budget. As a result, we would expect institutional power of the governor to detract from the direct influence of gubernatorial rhetoric as an explanation for the prison boom. We test this possibility through an interaction term that involves gubernatorial rhetoric and governor’s institutional power.

Hypothesis 2: The effect of gubernatorial rhetoric on mass incarceration is moderated by the institutional power of the governor.

Our consideration of possible interaction of gubernatorial rhetoric with other factors does not end there. As indicated in the above statement of Governor George Pataki of New York, tough-on-crime rhetoric of governors often accompany a discussion of effort to increase police presence in the streets. Thus, it is plausible that in addition to the main effect of rhetoric on incarceration, this effect is moderated by the number of police officers patrolling the streets and by the governor’s partisanship. We generate interaction effects to explore these other possibilities.

Political theorist Harold Lasswell long ago conceptualized politics as being about “who gets what, when, and how” from government [61]. Conflict is therefore a mainstay of politics. One dimension of conflict-based explanations of incarceration emphasizes differences in ideological orientation toward crime and punishment among political parties. Although Republicans and Democrats both detest crime, their responses to criminal behavior differ tremendously. For instance, analysis of Gallup Poll data by Carroll [62] shows a 20 percentage point gap in the “political divide” in public support for the death penalty between democratic and republican partisans. These differences are rooted in fundamental beliefs about individualism, democracy, and justice. [63]

Republicans generally view crime in terms of individual moral responsibility [11]. They believe that punishment deters crime and that crime should be met with escalating punishment in order to restore order and tranquility in society [19,64]. It is for this reason that during his campaign for president in the late 1960s, Republican Richard M. Nixon vowed to appoint law-and-order-minded justices to the Supreme Court who, Nixon had hope, will reverse the progressive cast of Warren Court decisions in criminal justice. Democrats, on the other hand, emphasize rehabilitation and reform as a major objective of the criminal justice system. They believe crime is fundamentally a structural problem that is associated with poverty, weakened family arrangements, and poor opportunities. Consequently, Democrats are willing to take these individual and contextual circumstances into account when crafting responses to crime. [14,65] Thus, for Democrats, imprisonment is not always the optimal response for criminal behavior. At the very least, they actively explore opportunities to turn citizens away from criminal activity with programs for community empowerment, job training, education, and structured activities for young adults. In light of this discussion, we expect Democratic governors to be less punitive than Republican governors towards incarceration.

Hypothesis 3: The incarceration rate will drop during the administration of Democratic governors relative to that of Republican governors.

An approaching election is a time when politicians find it strategically advantageous to buttress their tough-on-crime credentials. Therefore, we include variables designed to capture electoral pressure that governors and similarly situated state policymakers face. Extensive political science research indicates that elected state officials behave differently, usually more aggressively and more punitively on issues of crime, when they are facing an election [66-69]. We include in our analysis a variable that captures whether or not the governor is facing an election during the year. Furthermore, since presidential elections tended to create a high stimulus informational environment that might lead to greater discussion of crime and punishment, we follow Smith [14] and include a measure for the year in which presidential election is being conducted. For our purposes that year was 2004 when Senator John Kerry tried unsuccessfully to unseat incumbent President George W. Bush. We expect the competitive pressures of both gubernatorial and presidential elections to increase the incarceration rate in the American states.

Hypothesis 4: The pressures of gubernatorial and presidential election year will lead to an increase in the incarceration rate.

At a more macro level, this facet of the political explanation of incarceration speaks to political culture writ-large. The political ideology of the state, conceptualized in terms of the average position of state government officials on a liberal-conservative continuum, fits in well with this emphasis on political culture [70]. State political ideology is based on the ideological scores for the governor and all members of major party delegations in both houses of state legislatures. State political ideology is therefore a holistic measure of the liberalism and conservatism of state government. We expect that a liberal state ideology will be inversely related to mass incarceration.

Hypothesis 5: A liberal state ideology is negatively associated with mass incarceration.

Another aspect of political explanation for mass incarceration relies upon what Schattschneider [71] called the “socialization of conflict”. Conflict theories emphasize resource asymmetry among groups, suggesting unequal treatment of citizens based upon their placement on the socioeconomic ladder [11]. Specifically, conflict theories predict escalation of punishment for groups that, generally speaking, are negatively constructed and thus politically weak such as blacks (because of opportunity constraint and a legacy of discrimination), the poor (because of resource constraint), and the unemployed (because of skills constraint) [21,72]. For conflict theorists, incapacitating members of these groups through imprisonment is a way of controlling their potential threat to economic and social stability. In addressing this theory, Yates and Fording reported a significant increase in black imprisonment rate relative to whites as the size of the black population grows in the state. They further reported that this relationship was especially pronounced when social conservatives controlled state government.

Hypothesis 6: States with larger black population will have higher rates of incarceration compared to states with lower population of black residents.

Evidence on the relationship between poverty, labor market conditions and incarceration is less settled. Some researchers have reported that a growth in the unemployment rate is associated with increases in imprisonment rates, even after controlling for crime rates [73-75]. However, using incarceration rates and labor market data from 1977 through 1996, Stucky, Heimer and Lang [76] found no systematic evidence to support a relationship between unemployment rate and prison admission rates. In terms of poverty, Smith [14] reported that poverty was unrelated to mass incarceration. Using more recent data on incarceration, we test the importance of unemployment rate and poverty on incarceration rates. We expect that states with larger proportions of the poor and unemployed citizens will have greater levels of mass incarceration compared to states with fewer poor or unemployed individuals.

Hypothesis 7: States with higher proportions of the poor and unemployed citizens will have higher incarceration rates compared to states with smaller proportions of each of these groups.

Instrumental explanations of the prison boom hold that the prison population will change in direct proportion to changes in the crime rate. This means that since the violent crime rate has dropped 5.5 percent during the period we examined, we should at the very least expect a drop in the incarceration rate. Critics of instrumentalism argue that mass incarceration is a misplaced political response to perceived public fear of crime, not necessarily a response to the crime rate. For critics, the connection between crime and incarceration is either nonexistent or too tenuous in effect to matter [18,31]. If this is true, it would rule out instrumentalism as an explanation for mass incarceration and signal that something else, perhaps more akin to politics, is responsible for the incarceration boom. Indeed, Western [75] noted that the prison boom is purely a “political project” launched mostly by social conservatives to gain the votes of disaffected working class whites. In that sense, the prison boom offers private benefits to politicians in the form of electoral support rather than benefitting the public by calming their fear of crime or reducing the crime rate [18,77]. We test instrumentalism by focusing on the violent crime rate, which includes murder, rape, robbery and assault rates.

Hypothesis 8: There is a direct and proportional relationship between the crime rate and mass incarceration.

Finally, we include in our analysis crime control variables that capture specific characteristics of states that are possibly associated with incarceration. The number of police officers in full time employment (FTE) is important because higher police presence on the streets raises the probability of detecting illegality and apprehending suspects [78]. We expect that higher police FTE will increase state incarceration rates.

Ever since California popularized the “three-strikes-and-you-are-out” version of mandatory minimum sentencing legislation, many other states have enacted similar laws designed to increase the certainty and severity of prison sentences imposed. However, their effects are highly mixed and controversial. Sorrenson and Stemens reported that three-strikes legislation registered no effects on prison intake rate and incarceration rate. However, Nicholson-Crotty found that when reforms auspiciously link sentencing guidelines to correctional expenditures, they help to mitigate the incarceration rate. We adapt from Nicolson-Crotty (2004) a measure reflecting resource-linked sentencing guidelines (coded 1; zero otherwise) in our analysis to test the effect of sentencing reform. We expect states with three-strikes policy to experience an increase in incarceration rate. However, in an era of tremendous fiscal austerity across many states, we expect adoption of resource-linked sentencing guidelines to decrease incarceration rates.

Empirical evidence suggests that religious orientation of government officials affects their decision making patterns in a number of ways. Songer and Tabrizi [79] analyzed state Supreme Court justices to determine whether being evangelical, protestant, Catholic, or Jewish was associated with greater punitiveness in death penalty, gender discrimination, and obscenity cases. They concluded that religiosity was indeed linked to punitiveness. However, the effect was not significant across all religions. Only evangelical and Catholic identifiers showed a proclivity toward greater punitiveness. We test here whether adherence to religion among state residents is related to mass incarceration.

Our primary objective is to explain criminal punishment in the American states, especially how punishment is conditioned by gubernatorial rhetoric and other factors in the political and social environments. Our dependent variable, imprisonment, is measured as the total number of prisoners sentenced to 12 months or longer per 100,000 population.

An important role of politics in punishment, we suggest, can be formalized by considering the rhetoric used by governors to communicate crime and punishment policy in the states. Our data on gubernatorial rhetoric were gathered by Daniel Coffey who shared these with us. These data are based on content analysis of governors’ state of the state addresses from all 50 states. See Coffey 2005 for a detailed description of the data. These data were originally intended to provide a measure of gubernatorial ideology by coding the number of liberal and conservative sentences in the speech along economic and social dimensions. We utilize one aspect of these data that deals with the total number of times sentences related to “law and order” were uttered by governors. We divided the number of law and order sentences by the total number of sentences in the speech to derive a ratio of the entire speech that is dedicated to discussing the crime problem. We think this ratio is a reasonable approximation of the importance governors place on crime-fighting during that year.

Similarly, we consider state political ideology as measured by Berry et al. [70]. We subtract percentage of conservative identifiers from liberal identifiers so that higher total values correspond to more liberal states. Additionally, a handful of control factors have been shown to trend closely with prison population. We consider those criminal, economic, and political measures here. Our crime control measures include total number of police officers employed full-time and the violent crime rate which reflects murders per 100,000, robberies per 100,000, rapes per 100,000 and assaults per 100,000 populations. For our analysis, we combine all these incidents of violent crime and divide by four to form the violent crime rate, the main independent variable for testing the instrumental theory. Our economic measure is annual civilian unemployment rate. Our political measures are percent of women state legislators, partisan control of the legislature (coded -1=Democrat, 0=Independent, 1=Republican), governor’s party identification (coded similarly), a dichotomous variable for gubernatorial election year, a dichotomous variable for the 2004 presidential election year, and Beyle’s index of governor’s institutional power, which is a composite measure of both hierarchical authority and agenda power. Religion adherence is the proportion of state residents who identify as Christians; three strikes law is a dummy variable indicating whether the state adopted such a law (=1) or not (=0). Finally, if the state adopted a sentencing reform law linking sentencing to correctional resources, that state was coded 1, 0 otherwise (see Nicholson-Crotty [80]. We included a measure of the percentage change in the number of adults citizens paroled in each state per year (i.e., from January 1 to December 31 of that year) as a way of controlling for the impact of parole boards on incarceration.

What accounts for the incarceration boom of recent years in light of declining violent crime trends? Do governors lead on law and order issues via the bully pulpit? We begin our consideration of the incarceration boom by examining the relative ranking of American states in their rate of citizen incarceration. As reported in Figure 2, Louisiana has the highest citizen incarceration rate relative to its population. During the 2001 to 2004 period we examine, more than 700 inmates on average lived behind bars for every 100,000 Louisiana residents. Because this period predates Hurricane Katrina, that imprisonment rate does not reflect successful prosecutions for looting and other crimes committed in the aftermath of Katrina’s devastation of Gulf Coast states. Trailing behind Louisiana in relative incarceration rates are: Texas, Arizona, and Mississippi, respectively. At the opposite end of the spectrum, we find that Maine has the lowest relative imprisonment rate, followed by North Dakota, Minnesota, and New Hampshire, respectively.

Figure 2: Average Prisoners Per 100,000 Population, 2001-2004.

In Figure 3, we present a rank-ordering of states on how much attention governors place on crime and punishment during their state of the state addresses. On average, Nebraska governors devoted the greatest amount of space to the discussion of crime and punishment whereas Rhode Island, Washington state and Massachusetts governors rarely mention crime and punishment during their state of the state addresses. Later in the discussion, we explore reasons that might account for this interesting variance in crime rhetoric.

Figure 3: Crime Rhetoric in State of the State Address.

The distributional data reported in Figure 2 show variance in incarceration between states and across the four-year timeframe we study. They cannot explain why. We begin our exploration of the reasons for this variance by first examining the association between the proportion of gubernatorial rhetoric and incarceration rate in the states. In Table 2, we divide the United States into five geographic regions: New England, Mountain West, Appalachia, Midwest, and South. Within each region, we identify two states that share a border but that experienced different outcomes in incarceration rates (large or small change) during the period studied. We also calculated the level of emphasis that governors from these states placed on crime during their state of the state addresses. We want to assess whether crime rhetoric makes a difference in incarceration rates. Overall, all the states identified in these regions experienced an increase in their prison population during the period we study except Mississippi, which saw a decline of 2.9 percent. More importantly, the analysis shows a positive association between governors’ rhetoric and their states’ prison population, thus providing initial support for our rhetorical theory. States where governors employ more crime rhetoric during their state of the state address experienced a large relative increase in prison population compared to their geographic neighbors where governors used less crime rhetoric.

Table 2: Association between Gubernatorial Rhetoric and Percent Change in Mass Incarceration.

To further explore the reasons for the variation in incarceration rates, we estimated a series of random effects GLS regression models to test our rhetorical explanation of incarceration boom alongside alternative perspectives. We isolate and test the instrumental, rhetorical, and political theories separately culminating in an integrated model that combines all three perspectives. In this way, we can assess the strength and consistency of each approach. From a modeling perspective, the integrated model is the most correct since it accounts for the greatest number of possible confounding factors. Because we have time series cross sectional data (50 states across four years), we anticipated and tested for potential unit effects resulting from cultural and historical differences between the states. This test uncovered evidence of unit effects, meaning that the data are not independently and identically distributed. Consequently, we conducted our analysis via random-effects generalized least squares regression. This method is appropriate because it has the capacity to capture change over time, address potential distributional problems associated with unit variances, and produce unbiased and consistent estimates.

The results are presented in Table 3. The dependent variable is the incarceration rate. Recall that the instrumental model holds that the crime rate is proportionally related to incarceration, i.e., with an elasticity of one [7]. On the surface, it is a highly intuitive theory but one that is not easily tested and thus has generated a great deal of controversy.

Table 3: Testing Three Explanations of Mass Incarceration in the American States Via Random Effects GLS Regression.

We test the theory via two alternate methods. In the first, we created lagged values of both incarceration rate and violent crime rate as a way of addressing endogeniety. We included the unemployment rate, poverty, and percent of state residents who are black as control variables along with a key crime control variable measuring police capacity. The results “appear” to strongly support the basic premise of instrumental theory of mass incarceration. A one percent increase in the violent crime rate leads to a 1.07 percent increase in mass incarceration, holding all other variables constant at their means. If correct, this would amount to an increase of roughly 206 prisoners per 100,000population. However, that result, though statistically significant, is not consistent with the objective reality we seek to explain. As we noted in Table 1, the violent crime rate dropped 5.5 percent during the time we study. Indeed, the drop in crime rate has been steady since 1993, during Bill Clinton’s first term in office. Therefore, consistent with the theory, we predicted a decline in the prison population resulting from the drop in violent crime. Instead the analysis suggests that a decline in violent crime actually leads to an increase in the prison population rate, which is inconsistent with objective reality, even after controlling for the lagged crime rate, lagged prison population rate, unemployment rate and police capacity. We surmise that either the theory is inadequate or that lagging does not completely address endogeniety bias.

This leads us to use another method, two-stage residual inclusion (2SRI) technique, to re-estimate the model. The 2SRI method was first proposed by Hausman [81] as a means of directly testing for endogeniety bias in econometric models. The technique is especially appropriate for nonlinear data structures with unit variances such as ours and it has seen wide application among researchers [82]. Under 2SRI we first regress violent crime rate and unemployment rate on prison population on via ordinary least squares method and save the residuals. Then in the second stage, we include these predicted residuals in our random effects GLS model, along with violent crime rate, police capacity, and other predictors. The results are displayed in column three of Table 3. They indicate that including the residual term makes an important difference. The residual term is highly statistically significant. Indeed, once violent crime rate is adequately cleansed of endogeneity through inclusion of the first-stage residuals, it reports the theoretically correct sign that we predicted. Contrary to the predictions of instrumental theory, however, violent crime rate fails to reach statistical significance. Hence instrumentalism (hypothesis 8) is not supported. Instead the proportion of blacks in the state shows a strong impact on prison population growth, supporting the political conflict perspective. Our finding validates the claim raised by Zimring and Hawkins [10] that instrumentalism is a “theory in search of facts”. We conclude that instrumentalism is an inadequate theory for explaining mass incarceration because it cannot explain why a falling violent crime rate does not reduce imprisonment rate. We think the real explanation for the prison population boom lies in the bully pulpit, specifically, the political rhetoric governors use to explain their programs for addressing crime and criminality as part of their overall political agenda.

The most important theoretical insight we hope to contribute to the literature on mass incarceration concerns the value of gubernatorial rhetoric during state of the state addresses as a potential driver of mass imprisonment in the context overall crime drop. We argue that the use of aggressive political rhetoric during state of the state addresses contributes to the rise in mass incarceration but that this effect is moderated by the governor’s institutional power. We find strong support for our theory. In the fourth column of Table 3, we display results for the rhetorical model, which tests this argument directly. The explanatory capacity of the model is moderately good with an overall R² of .24. It suggests that the model is plausible and not simply a product of chance.

Our analysis shows that gubernatorial rhetoric does indeed contribute directly and powerfully to the growth in mass incarceration rates in the American states. On average, governors devote 4.7% of their state of the state address to law and order sentences. This percent is low relative to what one might expect in the high-crime era of the 1970s and 1980s. The analysis indicates a coefficient of 3236.58 on the rhetoric variable. Therefore each year, the direct effect of gubernatorial rhetoric on mass incarceration amounts to an increase of 152 prisoners on average per state, holding other variables constant. However, the full impact of gubernatorial rhetoric is contingent upon the institutional power of the governor (which reflects among other things, ability to veto legislation, control budgets, and appoint agency personnel). The interaction effect is the additional impact of gubernatorial rhetoric for each unit of governor’s institutional power or the additional impact of governor’s institutional power for each unit of gubernatorial rhetoric. For a typical state of the state address delivered by a governor with very low institutional power, gubernatorial rhetoric contributes 45 additional prisoners to the system. When crime rhetoric occupies 38 percent of the address (the maximum value in the data) and the governor is institutionally weak, this adds 365 extra prisoners to the prison system. This suggests that a weak governor who dwells more on crime rhetoric during the state of the state address sets the stage for more individuals to be either arrested, prosecuted, and imprisoned or denied parole altogether. A weak governor is one who is electorally vulnerable and is therefore more likely to use harsh rhetoric and to aggressively implement programs to reduce crime and improve voter support.

Conversely, gubernatorial rhetoric attenuates prison populations as the governor’s institutional power increases. The effects are strong and resilient to alternative specifications even after controlling for partisanship, police capacity, state political ideology, and religious adherence in the state. Overall, this finding reflects the fact that governors, especially powerful governors, have multiple avenues to influence mass incarceration, not just through the force of their own rhetoric. Governors can do so through budget proposals they send to the legislature, through the nature of the relationships they cultivate with agency administrators and staff, or through their ability to reject bills they dislike. Apparently, these potential alternative avenues detract from the force of the powerful governor’s rhetoric in affecting state imprisonment rates.

We explore here a number of other possibilities for the causal process through which rhetoric might influence incarceration. We created an interaction between gubernatorial rhetoric and two other process variables: police capacity and the governor’s partisanship. None of these additional interaction effects reached the conventional significance level of p<.05. However, political and cultural factors concerning a liberal political ideology in the state and a high level of citizen adherence to religion are significant and negatively related to incarceration rates, all else equal. This result is consistent across all the models we estimated.

The political model performs well, explaining 40% of the variance in prison growth overall. Many of the variables in that model are statistically significant and support the strong role of politics in mass incarceration policies of the states. In particular, state political ideology has a strong and direct impact on mass incarceration. Blue states where liberal political ideology typically dominates public discourse such as Massachusetts, New York, California, and New Mexico, are less likely to experience an increase in incarceration rate. A liberal political ideology leads to four fewer inmates in prison during the 2001 to 2004 period.

Among the most interesting and possibly controversial findings is the partisan control of the legislature. Traditionally, Republicans have cultivated the toughest posture on crime control and imprisonment. However, the analysis herein reported shows that Republican controlled legislatures are actually associated with a reduction of about 36 inmates from America’s prisons from 2001-2004. This finding is consistent throughout the models we estimated. We think there is a connection here to the extreme sensitivity of Republican policy makers to high budget deficits and their growing effort to reduce taxes and cut spending on many social and economic programs in the states, including corrections. Indeed, Austin [83] reports that in recent years, several Republican controlled legislatures (e.g., Michigan and Indiana) have either increased parole options or eliminated mandatory sentencing for nonviolent offenders as an attempt to cut their deficits. Some states (e.g., Kansas) have even linked sentencing reforms to correctional expenditures, which Nicholson-Crotty (2003, 406) concluded are associated with reduced prison population rates. Our analysis indicates that under Republican controlled legislatures, the average percentage change per year in the number of adult inmates paroled is 5.09% compared to 4.07% under Democratic control. We find a similar trend for governors. Under Republican governors, parole of adult prisoners increased an average of 7% per year compared to 3.5% under Democratic governors. Clearly, change is afoot in the attitude of Republicans toward mass incarceration.

Consistent with the analysis of Yates and Fording [27], we find that the proportion of black residents in the state proves is a powerful indicator of the growth in incarceration rate. A percentage increase in the black population leads to 649 more inmates in state prisons across the American states. The explanation for this finding is complicated. To consider this finding a clear manifestation of racism would be an uncritical and myopic view. Although racism still exists in our criminal justice system (Murakawa 2008), its form is more symbolic and structural than overt. To some extent, the finding reflects the lack of opportunity structure faced by many black citizens, especially in urban centers. Some authors have argued that due to this scant opportunity structure for blacks, crime has become synonymous with being black and this is manifested in the high arrest and prosecution rates of blacks relative to members of other races [30]. Both poverty and labor force participation rates fail to reach acceptable statistical significance levels in the analysis. The final component of the political model is the variable accounting for the proportion of women legislators. We expected a higher proportion of women in legislatures to correspond to a reduction in imprisonment rate because of traditional preference of women for rehabilitation over punishment escalation that is often favored by men. However, the variable failed to attain statistical significance.

The most complete model we estimated is reported in the last column and it subjects the multiplicity of theories of mass incarceration to the most rigorous statistical testing. We implemented a two-stage residual inclusion to address endogeniety associated with the crime rate and mass incarceration. The results give credence to much of the findings we have already discussed. The model explains 67% of the variance in incarceration between states and 48% overall. Most importantly, we find that gubernatorial rhetoric has a strong and highly significant impact on mass incarceration, even after controlling for 20 other independent variables. The model authenticates the contingency basis of gubernatorial rhetoric, showing that as the governor’s institutional power weakens, gubernatorial rhetoric leads to more individuals being incarcerated. However, as the governor’s institutional power grows, the impact of the governor’s rhetoric on mass incarceration becomes attenuated.

The integrated model in Table 3 also shows support for political explanation of incarceration. Both the effect of state political ideology and partisan control of the legislature are significant. However, as before, the effect of partisan control is contrary to our expectation, possibly reflecting continuing efforts in state legislatures to trim corrections expenditures and accelerate the exit of nonviolent offenders from state prisons in an effort to address budget deficits. The effect of black residents is moderately statistically significant (p<.o8), providing further support for the political conflict explanation. This confirms the unfortunate connection between race and crime, which results in blacks being viewed as a threat to structural arrangements in society. The other structural variables: unemployment rate and poverty failed to matter in explaining mass incarceration.

The findings herein reported give us a strong basis for joining other researchers who conclude that the instrumental model lacks empirical support [11,14]. The violent crime rate fails to explain the rise in mass imprisonment that American states have witness during the 2001 to 2004 period. Under state sentencing guidelines, individuals convicted of violent crimes are usually met with severe and length prison sentences. Therefore, if a relationship exists at all between the crime rate and incarceration rate, that relationship should be most pronounced in the area of violent crime. Our analysis of murder, rape, assault and robbery offenses indicate that something other than these violent crimes is causing the mass incarceration boom in the American states. Beyond traditional indicators of politics such as state political ideology, ideological leanings of the state legislature, and race, we think that the type of rhetoric that governors employ to discuss law and order policies makes state governors unwitting accessories and promoters of the growth in mass incarceration. The causal evidence we have mustered lends credence to this theory. We think that the rhetoric employed by governors in their state of the state addresses serves to amplify the punitive orientation and decision making of parole boards, police officers, prosecutors, judges, and prison officials toward offenders and inmates. At the end of the day, we think that future examinations of the perplexing divergence of crime statistics and mass imprisonment should account for the effect of gubernatorial rhetoric. In Table 4, we disaggregated the violent crime rate to show that variation among state governors in their use of political rhetoric is explained by the murder rate alone. Thus even though murders constitute just a small proportion of violent crimes, they form the basis of governors’ statements on law and order during their state of the state addresses.

Table 4: Explaining Differences in the Use of Aggressive Crime Rhetoric Among American State Governors.

This study was motivated by a puzzle: increasing rates of mass incarceration in the wake of decreasing rates of violent crime. We have attempted to shed some light on this puzzle by proposing and testing an overlooked explanation: governors’ ability to control crime policy and affect prison rates from the bully pulpit through the use of aggressive political rhetoric. Indeed, our analysis carves out a strong role for gubernatorial rhetoric in the swelling of state prisons nationwide. Specifically, we have demonstrated that, even when controlling for factors traditionally believed to explain incarceration rates, including crime rate, state political ideology, unemployment rate, race, and police capacity, the type and amount of rhetoric employed by governors matters significantly. When governors advance tough-on-crime messages as part of their policy agenda, states respond in kind by incarcerating more of their citizens. The analysis shows an amazing level of responsiveness to the governor’s crime rhetoric by those who must implement crime policy. Ours is the first empirically grounded study to provide evidence that gubernatorial rhetoric about crime does indeed add to mass incarceration. The analysis presented here thus moves us toward a deeper understanding of the relationship between political rhetoric and imprisonment in the American states by demonstrating the diffuse power of governors’ statements about law and order even at a time when overall violent crime is declining.