Community/Agency Trust and Public Involvement in Resource Planning

Abstract

Abstract We hypothesize and test a positive relationship between the extent to which local community members trust a management agency and their willingness to engage in resource-related public discourse and involvement. We employ a multilevel generalized mixed model to analyze data collected from five different samples of residents living near managed resource areas. Counter to our proposed hypotheses, results suggest individuals' level of dispositional trust, their belief that management shares similar values as them, and their trust in the moral competency of the management agency were all found to be significantly and negatively related to public involvement in resource-related activities. These findings suggest that the central role of building trust among local constituents within many planning frameworks needs to be reconsidered with consideration given to both the needs of individuals who trust an agency and the desires of distrusting individuals who are more likely to become involved in public involvement efforts. Keywords: community/agency relationshipstrustwatershed management Notes Note. The latent factor, public involvement, is the dependent structural variable. Adj. R 2 = 14.41. The Adj. R2 = 10.93 when the latent trust factors are removed from the model. Comparisons of the first and last waves of respondents on gender, education, income, and age showed no differences in the Lake Shelbyville and Navigation Project samples. In the Carlyle Lake sample, first-wave respondents differed from third-wave respondents on the amount of education obtained. Significantly more males were sampled than females, but this difference may be related to property records used to obtain names and addresses of study participants. More males than females are listed as owning property. Surveys were mailed to the person on the property record but the surveys may have been filled out by another adult (spouse) living in the household. We utilized the user-written command GLLAMM in the statistical package Stata 11.0 (see Rabe-Hesketh et al. [Citation2004a, Citation2004b] for information on GLLAMM; also see Zheng and Rabe-Hesketh [Citation2007] and Rabe-Hesketh et al. [Citation2005] for information on the 2PL). Frequently used maximum likelihood estimation of latent parameters using mainstream software programs (e.g., LISREL, AMOS) does not discriminate scale items, or public involvement actions in this case, against their relative degree of difficulty. In the 2PL model (Birnbaum Citation1968), the probability of taking action on item i by person n is modeled as a function of an item parameter, δ i , representing the difficulty of that particular action, and a person parameter, θ n , representing the person's general level of public involvement. The additional slope parameter, λ i , is referred to as a discrimination parameter, and determines how well action item i "discriminates" across varying levels of public involvement, θ n . This probability is expressed as: Goodness of fit in generalized linear latent and mixed models is typically reported through log-likelihoods, Akaike information criterion (AIC), and Bayesian information criterion (BIC) statistics and is used to compare models. We also constructed a multigroup model, specifying distinct sample populations as groups, using maximum likelihood estimation to generate parameter estimates and more conventional model fit statistics. Given the complexity of our hypothesized model, we determined it adequately fit the data given the following fit statistics: χ 2 = 3363.11, df = 1410; χ 2 /df = 2.39; RMSEA (90% CI) = 0.03 (0.03 − 0.04); CFI =.88; TLI =.86.