An investigation into the risk perceptions held by maximum security correctional officers

ABSTRACT

Maximum security correctional officers play a crucial role in the establishment of order within their respective institutions of employment, yet they are also exposed to numerous occupational dangers that can threaten their general welfare. When they perceive high levels of injurious risk from workplace hazards, this cannot only jeopardize their job performance but lead to a poorly managed prison institution. Currently though, few studies have explored correctional officer perceptions of workplace dangers and risks, and even fewer have explored the factors that influence officer perceived risk of injury. Questionnaire data from a statewide population of maximum security correctional officers (N = 649) were gathered in order to examine officer perceived risk of injury from workplace dangers, and the antecedents to this judgment. Results illustrated how officers perceived a high degree of injurious risk from their work, and that their risk perceptions were largely a product of psychological features of dangers. Theoretical and policy implications are discussed.

KEYWORDS:

• Correctional officers
• risk perceptions
• psychological predictors
• dangers
• prisons

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. Such biases were noted in Garcia’s (2008) dissertation where she used secondary data from the Prison Social Climate Survey to estimate predictors of correctional officer danger perceptions. Two of the items used to create a composite danger perception index (‘how often have inmates used physical force on staff?’ and ‘have you been physically assaulted by an inmate in the past 6 months?’), were directly related to past events. Risk perceptions represent judgments about the possibility of future events, and as such, these scale items used in her study measured neither a risk nor a danger. Furthermore, in Garcia’s (2008) investigation, no questions asking about the risk of an event from a danger were asked. Once again, to accurately measure a risk perception from a danger, it has been strongly suggested to have respondents evaluate the potential for a future risk from a specified danger, which would more correctly represent the above noted cause-effect scenario. 

2. Another less renowned theory of risk perceptions comes from the socio-cultural perspective and includes the social amplification of risk. For a complete discussion of this theory, see Krimsky and Golding (1992).

3. Recently as of 2013 within the South Carolina Department of Corrections (SCDC), where the current investigation was undertaken, the department instituted a new policy of desegregation of inmates afflicted with any type of infectious or communicable disease (SCDC, 2014). Prior to 2013, inmates diagnosed with potentially life threatening diseases such as Hepatitis, or HIV were quartered in separate wings of maximum security facilities. Today, this is no longer the case as these inmates are housed with the general population. All correctional officers employed within these institutions, as a result, are required to work alongside these inmates, thereby potentially increasing their risk of contracting one of these diseases. The problem though is that through personal communication with SCDC administrative officials, it was determined that officials are not doing their best at informing all officers about which offenders are afflicted with illnesses. As a result, and in direct relation to this study, this could potentially play a role in influencing their risk perceptions of harm from these inmates in particular.

4. There are a total of 28 state-operated correctional institutions within South Carolina, 8 of which are maximum level security. Of these eight, two are female only facilities and the remainder are male only. As of 31 January 2014, there were 7444 inmates housed in state-operated maximum level security facilities, making for a nearly 7:1 ratio of inmates to officers. South Carolina’s maximum security prison facilities house violent offenders who are serving extended sentences of greater than 60 months and who may be exhibiting certain behavioral, mental or cognitive problems that require medical attention. Offenders here are often isolated from one another, have their activities constantly supervised and are enclosed within single-celled structures that are surrounded by 20' high perimeter fences with extensive electronic and physical monitoring (SCDC, 2014).

5. An electronic survey account was purchased using the services of QuestionPro.com. The survey was made available between 22 January 2014 and 22 February 2014. Correctional administrative officials uploaded the survey to computers at all eight maximum security facilities via their intranet service, which permitted officers the opportunity to complete the survey during their shift. Completed surveys were submitted through QuestionPro.com, with no identifying information contained within them.

6. Updated records as of January 2014 regarding the total number of employed officers were provided by the Research and Development team of the South Carolina corrections department.

7. Representatives from the Research and Development wing of the South Carolina corrections department supplied demographic information for all maximum security correctional officers employed throughout the state during the data collection period. Of the 1076 officers employed, 645 were male (59%), 430 were female (39%), 484 were White (45%), 592 were Non-white (55%), 753 possessed between 1 and 10 years of job-related experience (70%) and 538 were between the ages of 18 and 35 (50%). When compared to these same sample demographics obtained in this study, these figures demonstrate that our sample is moderately reflective of the target population.

8. Tests for normality revealed how the criterion measure failed to approximate normality. Hair et al. (2010) though explained that for studies with sample sizes larger than 200, bias associated with non-normality ‘effectively diminishes’ (p. 77).

9. Kirkland was designated as the reference category as this institution processes all South Carolina offenders who are sentenced to prison terms, which includes over 15,000 annually. Kirkland houses the most violent and mentally troubled offenders within the state, and includes psychiatric and intermediate care units for male offenders. It has been described as the most dangerous of South Carolina’s maximum security facilities and serves as an adequate comparison to the remaining 7 (SCDC, 2014).

10. Originally it was presumed, given the clustered nature of the data with officers nested within 8 prisons, that multi-level modeling or at the very least cluster robust standard errors would have to be estimated. However, after visualizing relatively equal mean risk perception ratings across every institution (see Table 1) and following a non-statistically significant one-way ANOVA summary between officer risk perceptions and institution of employment F (2.43, p > .10), multi-level modeling was deemed inappropriate for these data. Cluster robust standard errors could have also been estimated, yet Angrist and Pischke (2009) cautioned that this procedure should only be adopted when the number of clusters is both greater than 40 and the total number of independent variables analyzed. Given only 8 surveyed prisons in the current study, more robust and alternative methods that can accommodate small cluster sizes include the pairs cluster bootstrapped t-statistic, the wild cluster bootstrapped t-statistic and the cluster adjusted t-statistic standard errors as these each produce more reliable null hypothesis test statistics and standard errors (Horowitz, 1997). However, even these procedures require that the number of clusters be greater than the number of explanatory variables, and in our analyses we have 15 variables to 8 clusters. With these warnings in mind, and to account for heteroscedasticity following statistically significant Levene’s tests (Hair et al., 2010), robust standard errors only were estimated (Horowitz, 1997). Extreme values analysis using the bacon command in STATA 13.0 was also performed, where inspection of histograms revealed no outliers.

11. Regarding multi-collinearity, across all examined variables, Variance Inflation Factors (VIFs) ranged from a low of 1.05 to a high of 1.87, while Tolerance levels never declined below 0.53. According to Hair et al. (2010), these values generally signal few concerns over collinearity.

12. Anonymous reviewers of this manuscript suggested disaggregating the comprehension, dread and risk perception scales and individually, in accompaniment of the remaining work attitude and demographic covariates, regressing each risk perception for each danger on these individual measures. Therefore, each individual risk perception measure for all seven dangers were regressed on the individual psychological variables (e.g. voluntary, control, knowledge, chronic-catastrophic, calm, fatal and anxious), as well as all others explored in this investigation. Precisely similar results reported in Table 4 were uncovered, where with the exception of role ambiguity, no other demographic or work attitude variables surfaced as statistically significant. Furthermore, and in directions consistent with the composite comprehension and dread scales, measures used to create these multi-item variables significantly predicted perceived risk for each of the seven individual dangers. For space conservation purposes, these results are not displayed in table format, but are available upon request following correspondence with the lead author.

13. Anonymous reviewers of this manuscript noted the considerable variance in response rates between surveyed institutions. To better address generalization issues, separate regression model equations including all examined predictors were estimated separately for all eight prisons. No substantively significantly different results were uncovered, as role ambiguity as well as the comprehension and dread variables were the only predictors to display statistically significant associations with the perceived risk of injury outcome across all individual prisons. Regardless of institution then, aggregated results hold across all studied prisons. Once again for space conservation purposes, these results are not reported in table format, but are available upon request.