Survey objectives and hypotheses
Our survey was designed with the following objectives: (1) assess the perceptions of nursery professionals on various aspects of the topic of invasive plant introductions via the horticulture trade; (2) determine levels of participation in voluntary preventive measures; (3) examine the extent of relationships between perceptions, business characteristics, and participation in voluntary preventive measures; and (4) investigate which incentives and obstacles emerge as most important to nursery professionals.
For the third objective, we tested several specific hypotheses pertaining to factors predicting participation in preventive measures. First, we expected that nursery professionals who perceive invasive plants as an important environmental problem would be more likely to engage in preventive behaviors. Second, we expected that nursery professionals who perceive the horticulture trade to be responsible for invasive plant introductions would be more likely to participate in preventive measures. Finally, we anticipated that participation rates would correlate with several business characteristics, including public visibility, business size, and involvement in trade associations.
Studies of other voluntary environmental programs in industry have highlighted relationships between several business characteristics and participation rates (Videras and Alberini 2000; Khanna 2001; Alberini and Segerson 2002; Anton et al. 2004). Based on patterns observed in other industries, we hypothesized that respondents from businesses that are less visible to the consumer public (wholesale, or grower nurseries) would be less likely to participate in preventive measures than retail, or non-grower counterparts, and that respondents from larger businesses would be more likely to participate in preventive measures than those from smaller businesses. We also hypothesized that respondents with greater reported involvement in trade associations would be more likely to participate in preventive measures.
Study population and data collection
Our study population consisted of San Francisco Bay Area wholesale and retail nurseries. Wildlands in this geographic region are valued for their high endemic plant diversity and have become highly invaded by non-native plants. Many of the plant species invading these wildlands are still sold commercially and are thus available for further dispersal via regional nurseries (Connick and Gerel 2005; J.W. Burt, pers obs).
We assembled the population of potential survey respondents by performing a keyword search using the AT&T (formerly SBC) and Bellsouth Internet directories for wholesale and retail horticultural nurseries in the nine San Francisco Bay Area counties (Alameda, Contra Costa, Marin, Napa, San Francisco, San Mateo, Santa Clara, Solano, and Sonoma). We excluded highly specialized nurseries (e.g., nurseries selling only succulents, roses, etc.), generating a list of over 400 nurseries.
The survey was conducted in March 2005. We called nurseries in random order from the generated list until we had completed at least 50 surveys. To reach our goal of 50 respondents, we called a total of 207 entries. Of these 207 entries, 85 were removed from our sample population (leaving 122) because they were either out of business, had an incorrect number listed, did not answer the phone, did not speak English, or were highly specialized. Of the remaining 122 nurseries carrying a general selection of outdoor plants, 48 additional businesses were excluded (leaving 74) because we were unable to reach a suitable participant (an owner, manager, or employee in charge of plant purchasing). Finally, of the 74 potential participants who we successfully contacted and gave the opportunity to take the survey, 54 respondents took the survey, for a response rate of 73%.
There is potential for some bias in our survey results because respondents who were more difficult to reach were somewhat less likely to be surveyed. To offset this potential bias, we conducted survey calls during slow periods for business (e.g., early morning) and made great efforts to schedule appointments convenient for potential respondents. We were thus able to include many respondents regardless of their workload. Potential respondents were asked only to participate in a survey of nursery professionals sponsored by a group at UC Davis and were not otherwise informed of the content or purpose of the survey before taking the survey. Thus, the response rate was not biased by the topic of the survey.
The telephone survey consisted of 25 multi-part, closed-end questions with opportunity for further comment afterward. The survey was designed to minimize response bias, with survey topics progressing from general to specific as the survey proceeded. For example, information on the St Louis Voluntary Codes of Conduct was introduced only after respondents had answered questions about their potential and actual engagement in preventive behaviors. The complete survey is contained in the Appendix. Specific questions used in statistical analyses are described in detail below.
For several of our analyses, the dependent variable was respondents’ level of participation in seven preventive measures (based on the St Louis Voluntary Codes of Conduct for nursery professionals and listed in Table 1). Our “participation” metric for each respondent consisted of the number of preventive measures (from 0 to 7) in which they reported they “have engaged.”
We employed combined metrics to construct several of our predictor variables in order to take advantage of complementary survey questions and to incorporate nuances between questions. We rated respondents’ perception of invasive plants as an environmental problem (“awareness”) according to their responses to two related survey questions. Respondents scored their agreement with two statements—“invasive plants have a negative impact on native plants and animals” and “invasive plants are an important environmental concern”—on a 5-point Likert scale, with 5 equivalent to “strongly agree” and 1 equivalent to “strongly disagree.” We used the sum of these scores as our “awareness” metric.
Similarly, we rated perceived responsibility of the horticulture trade for invasive plant introductions (“responsibility”), by combining responses to two survey questions. The first assessed agreement with the statement “the horticulture trade plays a role in the introduction of invasive plants” using a 5-point Likert scale as described above. The second question called on respondents to assign responsibility scores (on a scale of 1–5, with 5 = “most responsible”) for prevention of horticultural introductions of invasive plants to each of seven groups (consumers, retailers, wholesalers, growers, policy makers, government agencies, and scientists). For this second question, we calculated the average of responsibility scores assigned to the three horticultural groups (retailers, wholesalers, and growers). We then summed this average horticultural responsibility score with the response to the first responsibility question to create the horticultural “responsibility” metric.
We also used four business characteristics as independent variables. Nursery size was taken directly from a question on the survey in which we asked respondents if they considered their nursery to be small, medium, or large relative to other nurseries in the region. We defined respondents that classified their nurseries as primarily retail or both retail and wholesale as “retail,” while those who classified their nurseries only as wholesale were considered “wholesale.” We classified nurseries that grew any of their own plants as “growers.” We determined involvement in trade associations for each respondent by their level of activity within five trade associations (listed in Appendix, Question 10). For each trade association, respondents were given a point for each of the following attributes: having heard of the organization, being a member, reading the association’s literature, and attending meetings. The total (summed) scores of all five trade associations constituted the “involvement” metric.
We analyzed how awareness of invasive plants, perceived responsibility for invasive plant introductions, and business characteristics relate to participation in preventive measures using a general linear model (PROC GLM, SAS version 8.0, SAS Institute, 1999). Data met parametric normality and homoscedasticity assumptions. The linear model relating awareness, responsibility, and business characteristics to participation in preventive measures was based on the a priori hypotheses described earlier. In order to test the robustness of conclusions derived from that model, we conducted a basic model selection procedure to determine if any interactions between independent variables should be included. No interaction terms were selected for inclusion based on a stepwise procedure utilizing the Schwarz-Bayesian Information Criterion (PROC GLMSELECT, SAS version 9.0, SAS Institute, 2006). Inclusion of the best candidate interaction terms (awareness × responsibility and type × size) did not change the qualitative results of the model so we present the results from the a priori model without any interaction terms.
We ran a parallel analysis using “willingness to participate” as the dependent variable, where “willingness” was scored as the sum of “have engaged” or “would engage” responses for the seven preventative measures. This model did not have significant predictive power (overall model p = 0.41, R
2 = 0.15) and is not discussed further.
We conducted an ANOVA to assess whom respondents indicated as most responsible for preventing invasive plant introductions. Using individual respondents as a class variable, we compared responsibility levels assigned to each of seven groups (retailers, wholesalers, growers, policy makers, scientists, government agencies, and consumers). To test the hypothesis that respondents assigned a different responsibility level to horticultural groups (retailers, wholesalers, and growers) than to non-horticultural groups (policy makers, scientists, government agencies, and consumers) we performed an a priori contrast (Gotelli and Ellison 2004, p 339). After transformations failed to resolve problems of lack of homoscedasticity, we used a variance-weighted ANOVA for these tests (PROC GLM, SAS version 8.0, SAS Institute, 1999).