Identifying Factors that Influence Farmer and Rancher Decisions to Adopt and Manage Agroforestry Systems

A comprehensive literature review of North American agroforestry adoption studies and a series of interviews with Minnesota farmers informed the survey design. Seven semi-structured interviews were conducted prior to designing the survey – four with farmers that currently maintain agroforestry practices and three with farmers that have decided to not adopt an agroforestry practice but have implemented another conservation agriculture practice, such as cover crops or no-till. Farmers were asked about: their reasoning for adopting or not adopting agroforestry practices; the barriers to the implementation of agroforestry practices; any factors that may encourage them to adopt agroforestry practices; and the impact that agroforestry practices have had or may have on farm operations, management, and profitability. The answers to these questions, in addition to reported constraints and opportunities from the agroforestry adoption literature review, were used to create a list of constraints to agroforestry adoption and a list of opportunities for agroforestry adoption.

A mailing list of farmer addresses were purchased from Dynata, a private company that maintains mailing lists and databases for surveys and marketing. The mailing list included the names and addresses of 3,000 farmers, 1,500 of which reside in Minnesota, and the other half of which reside in Wisconsin. The sample within each state was proportionately weighted by county, meaning that counties with more farmers received more surveys than counties with fewer farmers. Mail surveys were sent to farmers in three rounds between July 2023 and February 2024. 

The survey packet delivered to each farmer in the mailing list included a cover letter, questionnaire, and pre-paid return envelope. The cover letter described the purpose of the study, the research rights of the participants, and the contact information for the principal investigators. The questionnaire was structured into eight sections. The first six each highlighted one of the six agroforestry practices included in this study. In each of the practice-specific sections, participants were provided with a short description and graphic of each practice and then asked if they currently maintain the practice on their land (yes or no response), the degree to which the list of constraints identified through interviews and the literature review limit their adoption (Likert scale from 1 to 5), and their willingness to accept the opportunities identified through interviews and the review (Likert scale from 1 to 4). In the seventh section, participants were asked to rate their likelihood of adopting each of the six agroforestry practices on a six-point Likert scale from “very unlikely” to “very likely.” The final section gathered farm and socioeconomic demographic information, such as age, acres farmed, and more. 

Data was analyzed in IBM SPSS Statistics to gather descriptive statistics, such as the frequency of responses and mean values of age and acre variables, and conduct binary logistic regressions. The independent variables included in this regression analysis were the degree to which each constraint is limiting, the likelihood of accepting each opportunity, total acres farmed, the proportion of acres farmed that are owned, acres of corn and soybean, acres of pastured cattle, income source, age, gender, and highest education level. The dependent variable being impacted by the independent variables was the likelihood of adoption for each practice. As the likelihood of adoption was measured on a six-point Likert scale, dummy variables were created to allow for binary analysis. In this way, all levels of unlikely (“very unlikely,” “moderately unlikely,” and “slightly unlikely”) were recoded to 0, and all levels of likely (“very likely,” “moderately likely,” and “slightly likely”) were recoded to 1. A binary logistic regression was conducted in SPSS for each practice, measuring the effect of the independent variables on that practice’s binary likelihood of adoption. The results were scanned for the independent variables (i.e., “factors”) that demonstrated a statistically significant effect on the likelihood of adoption. Producers were expected to be likely to adopt agroforestry practices as the constraints become less limiting and the opportunities become more available.

Collinearity, in which independent variables are correlated instead of remaining separate, may misrepresent regression results. To ensure independent variables were not collinear, binary logistic regressions were conducted as linear regressions in SPSS to identify the variance inflation factor (VIF) for each independent variable. A VIF value under five indicates low to moderate correlation and is not concerning, while a VIF value over five demonstrates moderate to substantial correlation and may interfere with regression results. The independent variables with a VIF over five were investigated to identify the other variables with which they were correlated, and some were deleted to ensure all independent variables were under a value of five.