Just don't call it climate change: climate-skeptic farmer adoption of climate-mitigative practices

Research exploring the factors that contribute to practices that emit greenhouse gases has offered some rather pessimistic observations. First, skepticism or denial among the public is notably higher than that expressed by the scientific community. Second, even among the concerned, researchers have documented a disconnect between level of concern and the adoption of environmentally favourable practices, an observation termed the value-action gap [1, 2]. This body of research is focused primarily on general public samples, however, and there are other segments of society that may warrant particular scrutiny. Farmers, for example, make up a tiny proportion of the populations of developed countries, yet their production decisions and related choice of management of practices have significant implications for both emitting and sequestering greenhouse gases in large volumes. Worldwide, food systems are responsible for 19%–29% of global anthropogenic greenhouse gas emissions, although this contribution varies widely by region [3].

We present the results of a survey of farmers in a technology-intensive, high-yield agricultural environment in Western Canada (Alberta). Agriculture, and particularly livestock production, has historically been an important sector to the Canadian Prairies. The province of Alberta, heavily reliant on fossil fuels as well as agriculture, has also been historically quite conservative, embracing neoliberal policy approaches to economic development and environmental regulation. Even in comparison to the rest of North America, climate skepticism is high in Alberta [4]. Today the agricultural sector in Alberta consists of over 40 000 large-scale farms with an average size of over 1230 acres, half of which produce livestock, bringing the provincial beef herd up to 5.4 million [5], accounting for 2.3% of provincial employment in 2017, and 71% of these workers are male [6].

The survey sought information on the adoption of a large suite of recognized climate-mitigative practices, and several control variables informed by previous research. Specifically, we investigate [1] how climate change beliefs vary in this population [2], what types of climate-mitigative practices farmers are adopting and at what rate [3], what are the stated motivations for these practices, and [4] to what extent do demographic characteristics and identity factors such as learning orientation, conservation ethic, and climate change norms, offer effective predictors of adoption. Respondents express low agreement with anthropogenic climate change, yet these beliefs do not correspond with adoption of management practices with climate-mitigative attributes. Farmers are motivated to adopt such practices for other reasons, including economic benefits, as well as expected co-benefits for soil quality and biodiversity. These findings suggest that strategies to encourage farmer adoption of climate-mitigative practices through climate change education campaigns will have limited success, and there may be far more effective strategies that align better with farmer identity and decision matrices.

Climate change and farmer behaviour

Choice of method always comes with certain trade-offs, and our survey-based method is not an exception. While survey methods have limited utility for exploring some of the more complex nuances of certain conceptual categories, such as identity, they are highly useful for testing some well-established theories, and also to identify broad patterns of relationships in populations. Our panel-based survey of 301 livestock and grain producers in Alberta was administered by Kynetec, an agricultural market research firm. Producers were to have made at least $10 000 in annual sales, and be geographically distributed across the province. We targeted a minimum of 100 beef producers; the remainder of the sample consisting of grain producers. Survey items measuring farm characteristics were based on questions from the Census of Agriculture.

Our comprehensive list of agricultural practices associated with climate-mitigative benefits, either in the form of greenhouse gas reduction or carbon sequestration, was based on mitigation research conducted by Smith et al [40]. This initial list was then expanded upon through consultation with colleagues, within Agriculture and Agri-food Canada. We generated a final list of 21 practices. Respondents were then asked which practices they have adopted, had not but would be willing to, or had not and are not willing to. For adopted practices, respondents were asked to indicate the reasons that motivated them from a list that included a range of environmental and economic considerations.

The survey included four items each to measure our three predictive factors (appendix A, table A1). To measure beliefs about climate change, we used a question developed by Arbuckle et al [41], asking respondents to choose one of five statements regarding the existence of climate change, and whether humans are partially or mostly the cause. Items measuring climate change norms were developed based on values-beliefs-norms theory and work by Stern et al [42], regarding the perceived importance of individual, sectoral, and societal contributions to climate change mitigation, and whether climate change poses a serious problem for society. Conservation Orientation was measured using a scale developed specifically for farmers by Maybery et al [26]. Items measuring learning orientation were adapted to the agricultural context from a scale aimed at measuring learning and innovation in firms [43].

A probit model using adoption as the binary dependent variable was employed to identify the probability of adoption for each practice while controlling for other factors, with the dependent variable representing each individual adoption decision. The equation below represents the adoption decision for farmer f evaluating action i. The utility associated with adoption is represented by ${U}_{fi},$ and ${A}_{fi}$ represents the observed adoption outcome, where ${A}_{fi}$ = 1 if adoption occurs, and ${A}_{fi}$ = 0 if it does not. If ${U}_{fi}$ is positive, the farmer is better off by adopting the action and ${A}_{fi}$ = 1, while ${A}_{fi}$ = 0 if utility is lost by adopting the action. X_f contains producer and operation characteristics meant to control for the social, personal, and farm-level factors affecting adoption, while $\beta$ represents parameters relating ${X}_{f}$ to ${U}_{fi}.$ X_i contains observable characteristics of the recommended practices meant to indirectly measure practice-related factors affecting the adoption decision and $\alpha$ is a vector of coefficients. The error term, e_fi, is normally distributed with a mean of 0, and u_f represents farm and producer specific effects not accounted for in X_f.

$$\begin{eqnarray*}&&{U}_{fi}=\left(\beta ^{\prime} {X}_{f}+\alpha ^{\prime} {X}_{i}\right)+{e}_{fi}+{u}_{f}\end{eqnarray*}$$

$$\begin{eqnarray*}&&{\rm{w}}{\rm{h}}{\rm{e}}{\rm{r}}{\rm{e}}\,{A}_{fi}=1\,{\rm{if}}\,{U}_{fi}\gt 0,\,{\rm{a}}{\rm{n}}{\rm{d}}\,{A}_{fi}=0\,{\rm{otherwise}}{\rm{.}}\end{eqnarray*}$$

Due to the presence of the unobserved error term, a standard binary choice model will be biased since one may not assume that the overall model error (${e}_{fi}+{u}_{f}$) is independently distributed. Thus, a probit model with robust cluster-corrected standard errors was estimated. This model employs a modified version of the Huber–White 'sandwich estimator' [44, 45], allowing for cluster-correlated variances and providing robust standard errors. This method allows for the error of each observation for a farmer to be correlated, but also implies that each cluster itself is independent [46].

A second step involved the estimation of ordered probit models for five mitigative practices associated with different types of energy and land use-based practices, all of which present significant capital-investment barriers to adoption, while also yielding the highest greenhouse gas savings potential across crop and livestock farm types (bioenergy, wetland restoration, solar power, cover crop, covered manure facility), with the dependent variables representing an ordered likelihood (definite no, no but willing, yes) of adoption of the selected practices.

Our survey sample consists of producers in Alberta's large-scale, commercial beef and grain sector (appendix B, table B1), with an average farm size of 3599 acres and median of 2000 acres. Two-thirds of the sample are primarily grain producers, and the remaining third primarily livestock producers. 89% of respondents are male, the average age is over 54 years, and respondents have managed a farm for 29 years on average. 25% of the sample has a university degree, and the vast majority of respondents receive at least some family income from off-farm. Over half of the respondents, finally, have completed an Environmental Farm Plan (EFP), administered by the Agricultural Research and Extension Council of Alberta, on behalf of the Government of Alberta. The EFP is a voluntary, whole farm, self-assessment tool that assists producers in identifying on-farm environmental risks and develop mitigation plans for which governmental programme funding is available. Our sample represents larger farms than the provincial average of 1200 acres, according to the last agricultural census administered by Statistics Canada (https://statcan.gc.ca/eng/ca2016). Other features of our sample that depart from the provincial average include a slightly higher proportion of males and larger number of farmers who have adopted the EFP (provincial average is 24%).

Climate change beliefs

Adoption of mitigative practices

Motivations for adoption

Predictors of adoption behaviour

Several notable findings emerge from this study. First, a sizable proportion of farmers have already adopted climate-mitigative practices. They are motivated to do so for many reasons, but neither agreement with climate science, nor an ascription of responsibility for mitigation, are among the motivators, a finding that departs from several other studies identifying climate change beliefs as a strong predictor of adoption [9, 29, 41] (beliefs that climate change is human caused have been found to be relatively less significant in studies of adaptation behaviours [47]). Rather, farmers recognize numerous co-benefits of mitigative practices, such as soil quality improvement, and wildlife and biodiversity enhancement. To the extent that numerous (but not all) mitigative practices also offer co-benefits the contribute to conservation, efficiency, and importantly, climate change adaptation, this finding is encouraging. Second, no single model of farmer behaviour has consistently strong predictive value for all decisions. Rather, the influence of both instrumental factors (e.g. size of farm, source of income) and identity factors (e.g. learning orientation) on farmer decision making is practice-specific. Finally, agreeing that climate change is human caused and viewing mitigation to be an important responsibility operate independently of each other, suggesting that agreement with the science of climate change is not a necessary pre-requisite for supporting mitigation strategies.

What may be happening here? For one thing, empirical research is always limited by the number of belief and practice measures included. In agriculture, major proposed strategies to induce mitigation have focused on collective, government-led responses including legislative mandates to reduce emissions, taxes, and cap and trade regimes, and much empirical research has been devoted to evaluating farmer support for such initiatives. But focusing on willingness to support a set of policies linked to an explicit climate change political mandate opens up a host of conflating issues, and perhaps does not provide an accurate picture of farm-level decision making, leading to a miscalculation of the potential contribution of farmers to mitigation. Attempts to garner support for such political mandates may trigger cognitive dissonance, leading to rejection of those mandates. Perhaps what we have uncovered suggests that value-action gaps do not necessarily prevail in all circumstances. To the contrary, our findings suggest the potential for action-value gaps as well—in other words, evidence for adoption of practices with climate-mitigative benefits despite the absence of climate change beliefs, or mitigative norms. More precisely, the depiction of the values and actions of relevance often ascribed to by academics and policymakers may be too narrow.

Relatedly, findings of low agreement that climate change is human caused are usually followed by calls for educating farmers about climate change, based on the expectation that doing so is necessary to secure higher levels of support. However, many studies find that education and awareness-raising campaigns rarely affect attitudes and/or behaviours [48, 49]. Continuing to presume that the only way climate-mitigative actions can be motivated is through changing beliefs about climate change implies a dim future, given what we know about how beliefs are formed, and the stability of those beliefs [50]. Trust also factors into acceptance of new information, as would be implied by identity theory. Belief in anthropogenic climate change is strongly associated with environmental groups, in which few farmers ascribe trust [12]. On the other hand, certain aspects of farmer identity may have even more relevance than the aversion of some farmers to an environmentalist identity. In short, a disproportionately large segment of the farmer population in western countries is older, white, male, and politically conservative—precisely that population segment found elsewhere to have the strongest affinity to climate skepticism [51].

Our findings, however, emphasize the extent to which neither such narrowly described identities, nor singular beliefs, are sufficient predictors of practice, lending credence to those scholars who emphasize the multifarious and dynamic nature of identity [31]. Farmers do tend to be older, conservative white males, but in many other respects bear very little resemblance to conservative white males outside the farming population, and thus express several identities, as well as knowledge and experience, that are brought to bear on farming practices. While such insights pose challenges to researchers, it also presents opportunities for applications: if farmers express the tendency to adopt strong climate change norms despite their disagreement that climate change is human caused, or in many cases adopt mitigative practices for other reasons entirely, this opens up an entirely new range of avenues for encouraging policy support. Farmers' beliefs about climate change are immaterial if they are adopting climate-mitigative practices for other reasons. Rather than stymie response, thus, an action-value gap suggests ambiguity or multiple interpretations, which can cause actors to converge in their responses but for entirely different reasons [52–54]. Recognizing this, policy making and messaging may purposefully include strategic ambiguity—'the deliberate use of ambiguity in strategic communication in order to create a 'space' in which multiple interpretations by stakeholders are enabled'—to facilitate collaboration [55].

Given our nonprobability sample, the generalizability of our findings to larger populations may be limited. On the one hand, as noted earlier, Alberta describes a unique socio-political context shaped by high levels of economic dependence on fossil fuels extraction and, perhaps relatedly, high levels of climate skepticism. On the other hand, our sample of large, industrialized operations is representative of modern industrial farming practiced in several other regions around the globe, and we are confident that this study has generated the potential for theoretical generalizability with application not only to other agricultural regions, but to other occupations and industries as well. For example, within Alberta, professional engineers and geoscientists working for oil and gas companies have also been quite skeptical of anthropogenic climate change [56]. To bypass this skepticism, the researchers recommended that a 'risk frame' had the discursive potential to bridge economic, environmental, and anti-/pro-regulation framings, given its ambiguity and interpretive plasticity to these respondents.