Public opinion about management strategies for a low-profile Species across multiple jurisdictions: whitebark pine in the northern Rockies

1. As public land managers seek to adopt and implement conservation measures aimed at reversing or slowing the negative effects of climate change, they are looking to understand public opinion regarding different management strategies.

2. This study explores drivers of attitudes toward different management strategies (i.e., no management, protection, and restoration) for a low-profile but keystone tree species, the whitebark pine (Pinus albicaulis), in the Greater Yellowstone Ecosystem. Since the whitebark pine species has a range that traverses different federal land designations, we examine whether attitudes toward management strategies differ by jurisdiction (i.e., wilderness or federal lands more generally).

3. We conducted a web and mail survey of residents from Montana, Idaho, and Wyoming, with 1,617 valid responses and a response rate of 16%.

4. We find that active management strategies have substantially higher levels of support than does no management, with relatively little differentiation across protection and restoration activities or across different land designations. We also find that support for management strategies is not influenced by values (political ideology) but is influenced by beliefs (about material vs. post-material environmental orientation, global climate change, and federal spending for public lands) and some measures of experience (e.g., knowledge of threats).

5. This study helps land managers understand that support for active management of the whitebark pine species is considerable and nonpartisan and that beliefs and experience with whitebark pine trees are important for support

Methods
Our study employed a cross-sectional design with a survey methodology to test our hypotheses. We distributed the questionnaire initially to 9,000 randomly selected addresses in Montana, Wyoming, and Idaho, proportional to the population in each state. We made multiple efforts to increase response rates (Dillman, Smyth, & Christian, 2014). Two letters were sent in two-week increments to direct potential respondents to a web version of the survey into which they would enter an authentication code to prevent duplicate entries; a hard copy of the survey with a business reply envelope was sent to non-respondents after another two weeks. We then drew another random sample of 1,000 new addresses, again proportional to state population. In this round, we sent only a paper version of the survey, with no web option. For all 10,000 randomly selected residents, we also randomly assigned an incentive value ($0, $1, or $2), with corresponding response rates of 9.9%, 17.3%, and 21.7%.

We test our hypotheses primarily using Wilcoxon signed-rank tests for matched pairs (Wilcoxon, 1945) and ordered logistic regression analysis. We use the Wilcoxon tests in comparing attitudes across management strategies and land types, as the data for the ordinal variables are matched at the individual respondent level. These tests are appropriate as the hypotheses (H1-H2) deal with comparison of variable distributions rather than association between variables. However, we apply Chi-square tests in a follow-up analysis exploring relationships among the six ordinal management strategy variables in an attempt to clarify the substantive significance of the Wilcoxon signed-rank test findings. We employ ordered logistic regression to account for the ordinal nature of the dependent variables (Long & Freese, 2014) in testing the remainder of the hypotheses, some of which involve continuous independent variables. Ordered logistic regression permits the calculation of post-estimation statistics to assess the marginal influence of one variable on the other. In order to facilitate interpretation of the regression results, we calculate changes in predicted probabilities for the dependent variables taking on particular values as the independent variables change values. Predicted probabilities are a common way to demonstrate marginal effects with ordinal dependent variables, as the regression coefficients can be difficult to interpret otherwise. 

Data were cleaned and variables were recoded and relabeled in STATA 14.