How perceptions of the institutional environment shape organizational priorities: findings from a survey of police chiefs

Abstract

A long tradition of research has examined the influence of organizational environments on criminal justice agencies. Based on survey data from a sample of local police chiefs, this study explores the effects of the institutional environment on police agency priorities. Specifically, we investigate how the perceived importance of different sectors of the institutional environment influences police agency priorities, as reported by police chiefs. The analyses reveal that certain sectors of the institutional environment exert greater influence on police organizational priorities than others. Moreover, the influence of institutional sectors differs according to the specific type of priority. Our findings reveal that institutional considerations exert more consistent effects on the importance of maintaining relationships with constituents than on maintaining law and order or adopting innovative practices. We draw on institutional theory in explaining the study’s findings.

Keywords:

• Police chief
• institutional environment
• institutional theory

Notes

1. Drawing on institutional theory, several researchers have applied the concept of ‘loose coupling’ to police organizations (see Burruss and Giblin 2014; Crank and Langworthy 1996; Maguire and Katz 2002; Mastrofski, Ritti, and Hoffmaster 1987; Zhao, Lovrich, and Robinson 2001).

2. For purposes of this study, the population of eligible participants included current police chiefs in jurisdictions meeting the TPCLS selection criteria. Among the 14 respondents deemed ineligible for inclusion in the present study, 9 were retired chiefs and 5 were chiefs in major cities when they completed the survey. This left 912 useable responses. Due to leadership transitions in some agencies during the time period covered by the study, more than one chief from the same agency may have participated in the training and completed a survey. As a result, the 912 useable survey responses represent 898 unique agencies.

3. We estimated a measurement model containing only the three latent dependent variables and their indicators. The model fit the data well according to several fit measures (χ² = 279.4, df = 59, p < .000; root mean square error of approximation [RMSEA] = .064; confirmatory fit index [CFI] = .987; Tucker-Lewis Index [TLI] = 0.983; WRMR = 1.08). Model estimation and model fit issues are discussed in more detail later in the paper.

4. The zero-order correlations between latent variables are as follows: maintaining relationships and adopting innovation (r = .68); adopting innovation and law/order (r = .50); law/order and maintaining relationships (r = .71).

5. Most composite measures in the social sciences are based on the assumption that the indicators are effects of the underlying latent variable being measured. The logic underlying factor analysis and various measures of reliability is that items sharing a common cause (the latent variable) should be highly correlated with one another. This approach is known as ‘reflective’ measurement because the items are said to reflect the underlying concept being measured. However, in some cases, the latent variable is thought to be caused by the items rather than the other way around. This approach is known as ‘formative’ measurement because the items are said to form the underlying concept being measured. In such instances, conventional approaches to measurement such as factor analysis and internal consistency tests no longer make sense because they are based on the assumption that the items share a common cause. Here, we assume that assessments of the importance of each individual constituency can be combined to form composite measures of the importance of seven sectors of the environment. We do not report measures of reliability because these constructs are based on formative rather than reflective logic (Bollen 2002; Bollen and Lennox 1991). We rely on additive indices because the specification of formative models in a CFA framework raises a number of complex challenges associated with model identification and multicollinearity (Diamantopoulos, Riefler, and Roth 2008).

6. We tested alternative coding schemes for education, including both binary and ordinal specifications. The ordinal response format provided the best fit.

7. Agencies reporting zero full-time sworn officers employed one or more part-time sworn officers.

8. Though it is standard to report χ² in structural equation models, its diagnostic value as a fit statistic has been questioned because it is often too strict (Bowen and Guo 2012). For the RMSEA, values ranging from .01 to .06 constitute close fit (Browne and Cudeck 1993; Hu and Bentler 1999). For the CFI and the TLI, values of .95 or greater indicate close fit (Hu and Bentler 1999). For WRMR, simulation evidence suggests that values below 1 are indicative of good fit (Yu 2002). Here, the model fits the data well according to CFI, TLI, and RMSEA, but the WRMR is slightly inflated.

9. We computed variance inflation factors (VIFs) for every independent variable to test for collinearity. Only two VIFs exceeded 2 and none exceeded 3, suggesting that collinearity was not problematic in this analysis (see Belsley, Kuh, and Welsch 1980).

10. The regression analysis resulted in 45 coefficients, 21 of which were associated with variables measuring police chiefs’ perceptions of the 7 environmental sectors and 24 of which were associated with the control variables included in the model.