Institutional Violence Rates & Future Dangerousness

An Integrative Review

The public has been very fearful of inmates that are sentenced to a term in jail, or even the death penalty, whose sentences are commuted to lesser ones or they are released from prison altogether. The stigma of their offenses assumes that these persons will continue to be threats while in confinement or while out on the street.

In Texas the legislature has enacted laws that mandate a jury to find beyond a reasonable doubt that a defendant is a future danger to society before he can be sentenced to death. This “future-dangerousness” question is one of many concerns in the research world.

Research conducted by Sorensen and Cunningham shows that these inmates who are sentenced to death are no more likely to become a threat to other inmates and staff in prison than inmates sentenced to prison for other criminal offenses (2006). This poses the question that if inmate’s rate of institutional violence are low, then how can the State justify satisfying the provision that an inmate will be a future danger when incarcerated for life or placed on death row?

Twenty-one states have enacted legislation that requires a jury to determine future dangerousness in order to sentence a defendant to death by execution.

Prison officials and clinicians have often feared that when you release an inmate back into general prison population due to court action or federal law changes then these inmates pose a risk and constitute a substantial risk to that prison population and security staff (Marquart & Sorensen, 1988). While violence tends to be a patterned behavioral trend for some of the individuals, reoccurrences of serious violence, such as murder, is a rarity, not the norm (Sorensen, Marquart & Bodapati, 1990).

In 1972 the United States Supreme Court held in Furman v. Georgia (1972) that capital punishment was in violation of the United States Constitution (Sorensen, Marquart & Bodapati, 1990). The Court held that the death penalty statutes passed by most state legislatures were unconstitutional, violating the cruel and unusual punishment amendment.

The goal of this paper is to examine the empirical literature dealing with prison violence in the Post-Furman era. The central element here is understanding the base rates of violence and how operational definitions impact the rate of violence. We are also attempting to find the most consistent predictors of violence and our current level of ability to predict such violent outcomes.


We retrieved 16 studies that examine prison misconduct rates of inmates on death row, LWOP (Life without Parole), and life with parole sentences. These study populations varied from state to state. We looked at research that was collected from inmate pools from Texas, Missouri, Florida, Arizona and numerous other states. We are attempting to see if the research is correlated with various measures from each of the present studies (see table 1).

We also want to examine if this collection of empirical studies has similar variables that can be correlated with other studies of prison misconduct and if they do than what are they and how much of an effect do they have. In other words, we are examining the average rates of violence on deep end prisoners and the correlations that are within violent misconduct itself. We are also including comparison samples from death row inmates that were commuted, life without parole inmates and inmates that are eligible for parole.

The variables examined in Table 1 include a list of the relevant literature in inmate violent misconduct, the type of population that was examined and the number of inmates sampled, the rate of prison violence at an annual rate per 1000 inmates, and in Table 2 correlated variables that were significant in measuring and predicting violent behavior.

Two of the studies were populated with post-Furman inmates. Some of these inmates were part of the cohort of inmates sentenced to death and subsequently later commuted to life in prison due to the United States Supreme Court’s declaration that the death penalty violated the U.S. Constitution in 1972 (Sorensen, Marquart & Bodapati, 1990).


Custodial officers, psychiatrists, and prison administrators feared that the release of commuted death row inmates into general populations would be catastrophic. These administrators believed that the former death row inmates were different from other inmates because they represented an increased security threat than that of other types of custodial inmates. To examine the literature on these rates of violence we broke down the population of inmates into comparison groups. These groups were inmates that were sentenced to death or commuted from a death sentence, inmates serving life in prison without parole and inmates serving life in prison that are eligible for parole.

Prison misconduct base rates among deep end prison populations

These base rates serve as a reliable assessment when attempting to predict future behavior of inmates confined in a correctional setting. However the research might have a problem with external validity because researching inmates in a correctional setting does not necessarily generalize to the outside population. Inmates that are confined are restricted in their movement around the facility, thus making it harder to assault an inmate or guard. In the outside world the inmate has no restriction in movement allowing for a greater chance of violent behavior to exist. To illustrate this point, Cunningham, Reidy & Sorensen (2005) compared rates of violent prison misconduct among death, life-without-parole, and parole eligible inmates from a Missouri prison general population. The logical reasoning of this study is to see if death row inmates could be integrated into the prison population without an increase in violent misconduct. Are inmates that are sentenced to life in prison more violent because they have nothing to lose? Cunningham, Reidy & Sorensen concluded that death sentenced and life without parole inmates were less likely to be involved in violent misconduct than their general population counterparts (2005). The results of this research offers questionable applicability to other populations outside the state of Missouri; however, Cunningham et al. reiterates that external validity of this nature may not a problem.

The Furman case had a rippling effect in the United States because people were afraid that if death row inmates were to be released into the general population then they were eventually able to be released back into society. As Sorensen and Cunningham have shown, numerous post-Furman inmates have been released back into society but the public fear is ill-informed. Sorensen et al. have shown that inmates released back into society from Furman or other commuting mechanisms show empirically that they are no more violent than any other inmate that would have been released.

These inmates serve as a catalyst of fear bestowed on to society. Society feels that the rate of violent offenses will increase because if the court system fails us and releases the more violent offenders, then these monsters are destined to kill again. Sorensen, Marquart & Bodapati focused their research on inmates in California just after their state Supreme Court declared that the death penalty was unconstitutional (1990). Just a few months later the United States Supreme Court ruled the same in Furman v. Georgia. Sorensen et al. compared Anderson commuted inmates (N=107) with information such as current status, recidivism information, and prior felony convictions (1990).

These inmates had been on death row from the early 1960’s to 1972. According to Sorensen et al. these inmates spent an average of 3 years on death row (1990). Sorensen et al. also concluded that only 29.3% of these Anderson commutees committed a new offense after being released (1990). The logic of that is 70.7% of the inmates did not commit any new law violation after being released. Based on a review of these individual base rates from the research, Sorensen et al. concluded that Anderson commuted inmates were no more violent than those of a regular prison population (1990). The authors have concluded from surveying post-

Furman inmates in California and examining their disciplinary behavior that due to a small percentage of the inmates that committed future crimes the rate was no different than that of current inmates being released from prison back into society (Sorensen, Marquart & Bodapati, 1990).

In 1972, Furman v. Georgia (1972) invalidated over 600 death sentences all over the United States (Marquart & Sorensen, 1988). Marquart & Sorensen (1988) examined the post-release behavior of 47 of these Furman commutees.

Prior to this mass commutation many of the prison administrators felt that integrating these post-death row inmates into the general population would cause a surge in violence within the prison system (Marquart & Sorensen, 1988). Marquart & Sorensen concluded after examining the disciplinary records of these post-Furman inmates that these inmates committed only a few serious violent prison misconduct incidents (1988). This is the antithesis of what the correctional administration originally anticipated would have happened.

Marquart & Sorensen (1989) studied the prison and release behavior of post-Furman inmates.

The questions that are being examined were:

  • what happened to these former death row prisoners,
  • and did these inmates or former inmates commit future crimes that would reinforce prison administrator’s fears that these inmates were always violent?
  • Is this consistent with what the literature is saying?

Marquart & Sorensen utilized various variables such as current status, prior felony convictions, prison disciplinary history, victim information, and whether the crime was committed in the commission of a felony to analyze the rates of violence in prison(1989). See table 1 to compare the variables and operational definitions for each of these studies. This study consisted of examining records from numerous states. The other studies only examined one state at a time to extrapolate future behavior unlike Marquart & Sorensen (1989) which examined numerous states to increase the external validity. Marquart & Sorensen also concluded that 80% of those released to the free society have not, at least officially, committed additional crimes.

Compared to the previous studies outlined in table1, the result is that these studies reiterate a strong conclusion that increased prison violence among death penalty and life without parole inmates when integrated into general population is greatly exaggerated. As a method of comparison we have broken the rates and populations by court sentences. These comparisons are to determine whether there is actually an effect of an inmate’s behavior in comparison to the type of sentence that is imposed. For the purpose of this research we examined the literature on prison rate behavior to include former death row inmates, life without parole inmates and life with parole inmates.

Former and Current Death Row Inmates:

Utilizing death row sentenced inmates as the experimental group Reidy, Cunningham & Sorensen examined the disciplinary records of 39 death row inmates in Indiana (N=39) that were transferred in to general population because of a change in their case status (ie. commutation, court reversals etc.). Reidy et al. concluded that a total of 35.9% of the inmates were involved in violent acts and only 26% were involved in violent acts while still under a sentence of death. They also concluded that only 20% of the inmates were involved in violent acts while sentenced to life in prison (2001).

Only one study that was located offered research to the contrary (see Delisi et al.). Delisi et al. examines a population of inmates in Arizona and determines that violent inmates sentenced to death are violent within the institution. Delisi states that violent inmates are more dangerous than others (Delisi et al., 2003). However, when examining this study with detail, one sees that Delisi et al. expands the parameters of inmates and does not clarify what variables he is measuring. One can study 10 violent felons that have killed prison guards and conclude that these inmates are violent, but how can we transfer that to the entire population of death row inmates? We cannot as asserted by Sorensen and Cunningham.

To illustrate the point of the how low the prison violence rate is Marquart, Ekland-Olsen and Sorensen (1989) stated that, of 92 capital murder offenders in Texas who have been sentenced to death after the jury had stated to the special issue, there was a probability the individual would be a future danger.

However according to Marquart et al. capital juries are not likely to be accurate when determining a predictiveness of future criminal behavior or dangerousness. Marquart, Ekland-Olsen and Sorensen (1989) concluded that an annual prison violence rate of former death row inmates was .016 annual and .026 annual with convicted murders per 1000 inmates. These risk numbers were calculated from an assessment scale that was devised to predict this future behavior.

Cunningham, Sorensen & Reidy (2005) used an experimental scale to examine the assessment of prison violence among maximum security inmates. Cunningham et al. examined inmates that were sentenced to death and life-without-parole (2005).

The authors concluded that utilizing a logistical regression analysis that predictor variables such as age, type, length of sentence, education, prior prison term, prior probated sentence and years served acted as good predictors of future violent behavior with an area under the curve (AUC) of .719 (Cunningham, Sorensen & Reidy (2005)).

Life without Parole (LWOP) Inmates:

Another group of inmates that were examined were those inmates sentenced to only life without parole. The research question is reiterated how violent are these inmates in comparison to other high security inmates? Cunningham and Sorensen (2006) conducted an examination of prison misconduct rates among life-without-parole and other long term high security inmates and found that from 1998 to 2003, the likelihood and pattern of disciplinary infractions and rule violations among LWOP inmates is similar to those of other inmates that were sentenced to long term sentences with the possibility of parole. Sorensen and Wrinkle (1996) also compared 323 life without parole inmates that were convicted of capital murder and 232 inmates that were sentenced to life with parole for murder. Sorensen and Wrinkle (1996) found there was no significant difference between the rates of disciplinary infractions between life without parole (LWOP) inmates and life with parole inmates. This offers the conclusion that the sentence has no effect on predicting future criminal behavior behind bars. Cunningham and Sorenson found in more recent research that the same thesis applies with inmates sentenced in Florida from 1998 to 2002 (2006). Cunningham actually states that this effect is a stabilizing effect that is produced by LWOP inmates rather than a disruptive force in the prison system (2006).

Life with Parole Inmates:

Sorensen & Pilgrim examine the above research questions but use murders with a variety range of sentences as the experimental group. Sorensen and Pilgrim examined the disciplinary records of 10,121 murderers in Texas (2000). These murderers were eligible for parole in Texas. At the time of this study Life without Parole in Texas was not an option for Texas juries. They found that measures such as gang membership, prison history, and prior felony involvement in murder were a significant factor that influenced an inmate’s chances committing violence in prison (2000). Sorensen et al. calculated an inmate violence rate of .024 per 1000 inmates and .084 per 1000 inmates for serious assaults only. The prison violence rate was drastically lower than that of inmates sentenced for non-death property offenses.

Measures of Prediction:

In the criminal justice system it regularly depends on three types of predictions of future behavior (Marquart, Ekland-Olson and Sorensen, 1989). The first is predicting behavior based on past behavior. The second is the behavior of a person with similar characteristics (i.e. drug use, profiles etc.). And according to Marquart et al, the third and most widely used is the clinical judgment of an expert in the psychological field (1989). To illustrate the expert point, Edens et al. examined Texas Court of Criminal Appeals opinions on death penalty cases that an expert testified regarding the defendant being a future danger (2005). Edens et al. identified 155 cases that fit this criterion (2005). Edens et al. concluded from examining the trials of the 155 cases that clinical assessments and clinical risk instruments appear to be highly inaccurate in predicting future violent behavior of prison inmates (2005).

Utilizing Marquart et al and Edens et al psychological tests are also utilized in prediction of future dangerousness however modifications to these measures are needed. These modifications include developing variables to be associated with other populations (other states etc.) to better generalize the entire population.
Risk assessments are the crux of predicting future violent behavior in high security inmates. Cunningham & Sorensen (2006) examined the RASP-Potsi (Risk Assessment Scale for Prison) risk assessment scale to assess prison violence. The authors examined the RASP scale with 14,088 inmate disciplinary records. Cunningham et al. utilized variables such as age, education, prison confinement, offender type and sentence type (2006).

They concluded that the scale was successful in predicting prison violence with an AUC of .645 to .707 (2006).
Utilizing the research we conclude that the RASP predictor variables have a relative strong association factor in predicting future behavior. The RASP instrument must utilize various variables and correlates to make its predictions. We examine these variables and correlates in the research reviewed.

Examination of Variables and Correlates (Table 2)

Table 2 discusses the types of correlates that are used within the scope of our research. For this research we confined the correlates that were most prevalent to this body of research. These correlates include age, education, race, gender and prior criminal history. The research points to other variables to explain this behavior, however we have summarized in table 2 some of the main correlates used within the current literature. The table shows an inverse relationship is consistent with age and prison violence. The exception to Cunningham and Sorensen’s work is research conducted by Delisi and Munoz. Delisi’s concluded that death row inmates were more violent than other inmates in the prison system. Cunningham and Sorensen concluded that Delisi and Munoz’s work was infested with methodological flaws. Going back to table 2 we see that the standard statistical method used within our retrieved literature is regression analysis. However, some of the studies included non-parametric tests to show an association between the various correlates.

Cunningham & Sorensen (2006) analyzed disciplinary records of 136 incarcerated capital murder offenders in Texas in the early stages of their sentences (between 6-40 months). Examining the disciplinary records of these inmates at the Texas Department of Criminal Justice-Institutional Division Cunningham et al. found, consistent with prior studies, that variables such as age, prior prison history and assaultive misconduct were consistent in predicting behavior to a AUC of .715-.766 (2006). This predictive model has been needed to develop a “common sense” tool to determine if future dangerousness is a viable test for executing an inmate.

To further show how different variables can explain an association in prison violence, Sorensen & Cunningham (2006) reviewed 1,659 disciplinary records of convicted murderers sentenced to the Texas prison system between February 2001 and November 2003. Sorensen & Cunningham found that characteristics of inmates in regards to their age, coupled with more serious murder convictions, were associated with a higher incident of prison assaults (2006) Utilizing a logistic regression analysis with a determining baseline the authors concluded that the choice of measures used in researching prison violence is crucial in making true and accurate predictions.

To further illustrate this point Sorensen & Cunningham looked at the characteristics of 2003 disciplinary data on 51,527 inmates from Florida that were convicted of murder and lesser-included offenses of murder (2006).

The authors examined the rates and correlates of prison misconduct and violence. These correlates are distinguishable as a research baseline to compare the rates of violent misconduct of these Florida inmates. See table 2 to examine the correlates of the various studies examined in this paper. A negative binominal regression was used as a predictive method and revealed that convicted murderers in Florida were 20%-30% less likely to engage in potentially violent misconduct than other non-violent offenders such as property offenders (Sorensen & Cunningham, 2006). However, according to Sorensen & Cunningham, inmates serving sentences for assault or attempted murder were not associated with a lower rate of prison violent behavior.

The inverse of this is true, according to Sorensen and Cunningham (2006) that inmates who serve sentences for violent convictions were associated with a lower rate of prison violence (2006).


The current study analyzed 16 studies that involved inmate disciplinary records and predictive models. We also included an analysis of various correlates that corresponded with each study. We included in table 1 a chart of each study and its corresponding population, state, measure, operational definition and variables. We observe many similarities of each study including the various correlates observed (table 2).

According to research conducted by Sorensen and Cunningham former death row inmates are no more likely to commit violent acts than life with outparole and their life with parole counterparts (Cunningham, Reidy & Sorensen, 2005).

As cited in table 1 this research brings together a list of somewhat consistent violent misconduct rates of inmates serving in all levels of custody. This shows that the courts have not looked closely at how much of a danger inmates really are when a defendant is sentenced to death. Many states have made it mandatory for juries to make a finding that an inmate is a future danger from prior history and assessments from psychological experts. Sorensen and Cunningham have concluded that this method is not consistent with how violent these inmates really are in prison. When you compare the violent misconduct rates from the overall literature you find that juries could be deciding the fates of inmates with less than accurate information.

One difference that we find with the literature is that for each study a different numerical population of inmates is examined. Some of the studies include a large sample of inmates and others only examine a few. The question of external validity must be in the forefront.

Does examining 3000 inmates really constitute a sample of the entire population of inmates in the United States?

Future research needs to include an analysis of the female prison populations. Is a woman that is sentenced to death likely to commit violent acts in relation to her own gender? The above research mainly focuses on males due to the increased numbers however a test of gender would be applicable. This literature is very consistent with its conclusions stating that inmates from death row are no more violent than their prison counterparts.


  • Cunningham, M.D. & Sorensen, J.R (2006) Nothing to Lose? A comparative examination of prison misconduct rates among LWOP and other long term high security inmates. Criminal Justice and Behavior, 33, 683-705
  • Cunningham, M.D. & Sorensen, J.R. (2006) Actuarial models for assessing prison violence risk: Revisions and Extensions Risk assessment scale for prison (RASP). Assessment. To be published 2006
  • Cunningham, Sorensen & Reidy (2005) An actual model for assessment of prison violence risk among maximum security inmates. Assessment. 12, 40-49
  • Cunningham, M.D., Reidy T.J. & Sorensen, J.R. (2005) Is Death Row Obsolete? A decade of mainstreaming Death Sentenced Inmates in Missouri. Behavioral Sciences and the Law. 23, 307-320
  • Delisi & Munoz (2003) Future Dangerousness Revisited. Criminal Justice Policy Review. 14, 287-305
  • Edens, J.F., Buffington-Vollum, J.K. Keilen, A., Roskamp, P., & Anthony, C (2005) Predictions of future dangerousness in capital murder trials: Is it time to “Disinvent the wheel?”. Law and Human Behavior. 29, 55-86
  • Marquart, J.W. & Sorensen, J.R. (1988) Institutional and post release behavior of
    Furman-commuted inmates in Texas. Criminology. 26, 677-694
  • Marquart, J.W. Ekland-Olson & Sorensen (1989) Gazing into the Crystal Ball: Can jurors accurately predict dangerousness in capital cases? Law & Sociery Review, 23, 449- 468
    Marquart, J.W. & Sorensen, J.R. (1989) A national study of the Furman-commuted inmates: assessing the threat to society from capital offenders. Loyola Law Review, 23, 5-28
  • Reidy, Cunningham & Sorensen (2001) From death to life: Prison behavior and former death row inmates of Indiana. Criminal Justice and Behavior, 28, 67-82
  • Sorensen, J.R & Wrinkle, R.D. (1996) No Hope For Parole: Disciplinary Infractions among LWOP and Death-sentenced inmates. Criminal Justice and Behavior, 23, 542-552
  • Sorensen, J.R. & Pilgrim, R.L. (2000) An Actual Risk Assessment of Violence Posed by Capital Murder Defendants. Journal of Criminal Law and Criminology, 90, 1251- 1270
  • Sorensen & Cunningham (2006) Conviction Offense and Prison violence: A comparative study of murders and Other offenders (manuscript under review-a)
  • Sorensen & Cunningham (2006) Operationalizing risk: The influence of Measurement choice on the prevalence and correlates Of prison violence among incarcerated murders. (under review-b)
  • Sorensen & Cunningham (2006) Capital Offenders in Texas Prisons: Rates, correlates, and actual analysis of violent misconduct. (under review-c)
  • Sorensen, J.R., Marquart J.W. & Bodapati (1990) Research Note: Two Decades after People v. Anderson. Loyola law Review 24, 45-56


Leave a Reply

Your email address will not be published. Required fields are marked *