Climate belief, accuracy of climatic expectations, and pro-environmental action

ABSTRACT

This paper extends the literature on climate change by asking ‘Who believes in climate change?’, ‘Whose expectations regarding future climate coincide with scientific projections?’, and ‘Do beliefs and holding ‘correct’ expectations influence taking concerted action?’ We find that 88% of survey respondents representing New Zealand’s primary sector believe in climate change but that only 34% have expectations regarding future temperatures that coincide with scientific projections, and only 40% have expectations regarding future drought that coincide with scientific projections. We show that men, less educated respondents, those with long family histories of farming, and those who put greater trust in social media are more sceptical of climate change. Among respondents who believe in climate change, these same factors are also associated with lower likelihood that their expectations coincide with scientific projections of future climate scenarios. Furthermore, women and more educated people are less likely to under-estimate and more likely to over-estimate both temperature change and the incidence of drought relative to scientific projections. Finally, we show that previous pro-environmental action and future intentions to undertake pro-environmental action are positively associated with belief in climate change, but that having expectations that match scientific projections does not additionally affect uptake of pro-environmental action.

KEYWORDS:

• Actions
• agriculture
• beliefs
• climate change
• survey of rural decision makers

JEL CODES:

• D83
• C83
• Q15
• Q54

1. Introduction

Effective implementation of adaptation and mitigation actions to address climate change requires an understanding of future climate (Howden et al., 2007; Nguyen et al., 2022; Niles & Mueller, 2016). Personal experiences with climate change and climate-related weather patterns shape individuals’ expectations (Broomell et al., 2015; Spence et al., 2011), as do personal characteristics such as cognitive factors (Myers et al., 2013), habits including news consumption (Bolsen & Shapiro, 2018; Diehl et al., 2021), and preferences such as political affiliation (Jenkins-Smith et al., 2020). Within this growing literature, relatively few studies have documented a causal link between climate belief and concerted pro-environmental action, although Nguyen et al. (2022) use an earlier wave of the same survey employed in this study to demonstrate that farmers who anticipate greater intensity and higher frequency of drought in the future are more likely to undertake on-farm mitigation and adaptation actions.¹

The extent to which climate beliefs reflect underlying climatic trends is also poorly understood. Indeed, we are not aware of any studies that assess the concurrence of future climate expectations with scientific projections in the agricultural sector, where accurate expectations beneficially impact production decisions such as the optimal time of planting, usage of inputs, herd and flock management, and risk management (Guido et al., 2021; Nissan et al., 2019).

Future implications of climate change and changing weather patterns are especially critical in New Zealand, which has approximately 23,000 commercial sheep and beef farms and 11,000 dairy farms (Statistics New Zealand, 2017). These farms hold 25.7 million sheep, 4.0 million beef cattle, and 6.2 million dairy cattle (Beef+Lamb New Zealand, 2022). In 2022, wool exports were valued at $438 million, meat exports were valued at $9.8 billion NZD, and dairy exports were valued at $19.8 billion NZD (Beef+Lamb New Zealand, 2022). The primary sector was responsible for 11% of total GDP, 14% of total employment, and 52.2 billion NZD in export revenue in 2022 (Ministry for the Primary Industries, 2022).

While changing climate has clear implications for farmers, farming practices also have clear implications for changing climate. Indeed, New Zealand’s agricultural sector is responsible for just under half of total gross greenhouse gas emissions (Ministry for the Environment, 2022) compared to 17% globally (FAO, 2020).²

In its 2021 Nationally Determined Contribution to the Paris Agreement, New Zealand committed to reducing national net greenhouse gas emissions by 50% below gross 2005 levels by 2030 (Ministry for the Environment, 2023). The commitment followed the 2019 Zero Carbon Amendment Act (New Zealand Parliament, 2019), which made New Zealand one of the first countries to bind climate commitments into law. To facilitate reductions in the agricultural sector, the New Zealand government, industry bodies, and the indigenous Māori people established the He Waka Eke Noa – Primary Sector Climate Action Partnership.³

In mid-2021, the government also released the ‘Good Farm Planning Principles: Towards Integrated Farm Planning’ guide. This guide provides advice on managing greenhouse gas emissions, fresh water, biosecurity, soils, and other policy priorities (Ministry for Primary Industries, 2021). Other policies changes at this time focused on extreme weather events, including the drought affecting many parts of the country from summer 2020 to autumn 2021. This drought severely affected stock feed and exacerbated water shortages in the food and fibre sector, and the government provided 2.8 million NZD in grant support to affected farmers (OECD, 2022).

The effect of drought on livestock in temperate climates has been subjected to recent scrutiny (e.g. Blauhut et al., 2021), with evidence showing that drought reduces feed and fodder availability, animal productivity, and animal welfare (Salmoral et al., 2020). In the New Zealand context, Tait et al. (2005) describe typical responses to drought on dairy farms, including reducing milking from twice to once per day, drying off poorer performing cows, and reducing stocking rates through selective culling. These practices reduce yields, and hence Pourzand et al. (2020) find that drought positively impacts revenue and operating profits on dairy farms due to temporary price increases⁴ while potentially negatively affecting balance sheets over the long term. Drought may also positively affect revenue on sheep and beef farms as farmers sell off stock, but Pourzand (2023) demonstrates that gross income and profit on sheep and beef farms declined after drought in most New Zealand regions. Pourzand et al. (2020) further show that drought in successive years increases financial strain on farmers.

Similarly, scientific evidence of the link between high temperatures and heat stress in livestock was firmly established in the 1960s (e.g. Johnson et al., 1963), and the New Zealand National Institute of Water and Atmospheric Research (NIWA) reports that New Zealand beef and dairy cattle start experiencing heat stress at temperatures above 25 degrees (NIWA, 2023). Consequences of heat stress among cattle include reduced dry-matter intake, slower weight gain, reductions in fertility, higher mortality, and reductions in the quantity and character of milk (e.g. Crescio et al., 2010; Dunn et al., 2014; Jordan, 2003; Silanikove, 2000; West, 2003; West et al., 2003; and Polsky & von Keyserlingk, 2017). In New Zealand, hot conditions have been causally linked to reductions in milk yield, milk solid yield, fat content, and protein concentrations for the three most common breeds of cows on dairy farms (Bryant et al., 2007).⁵

In this paper, we draw from a large survey of New Zealand farmers, foresters, and growers (hereafter, ‘farmers’) to establish correlates of belief in climate change across New Zealand’s rural sector. Comparing individual expectations of future temperature and drought to high-resolution projections based on the Intergovernmental Panel on Climate Change (IPCC)’s Representative Concentration Pathway (RCP) 4.5, we then identify whose expectations of future climate coincide with scientific projections. Finally, we analyse the extent to which belief in climate change and accuracy of future expectations matter for having taken concerted environmental action in the past and having made plans for such action in the future.

Several results stand out. While 88% of farmers believe in climate change, only 34% have expectations regarding future temperatures that coincide with forecasting and only 40% have expectations regarding future drought that coincide with forecasting. We find that underestimating temperature (49%) and drought (41%) vis-à-vis IPCC models is common, while overestimating temperature and drought is less common (<20%).

In addition, belief matters for action. Specifically, farmers who believe in climate change are more likely to have implemented pro-environmental actions in the past and are more likely to plan to implement pro-environmental actions in the future. This finding holds for a wide variety of pro-environmental actions but is particularly strong for actions intended to reduce greenhouse gas emissions. We also find that the accuracy of farmers’ expectations (relative to scientific projections) with respect to temperature and drought does not affect the uptake of actions over and beyond believing in climate change, indicating that belief in climate change itself is sufficient for action.

The next section of the paper discusses our survey data and the IPCC forecasts of temperature and drought in 2050 at the local level. We then map individual-level, spatially explicit expectations of future temperature and drought onto these projections. Section 3 identifies correlates of belief in climate change and expectations of future temperature and drought that coincide with scientific projections. Section 4 discusses pro-environmental action – whether previously undertaken or planned – and describes the extent to which belief in climate change and holding expectations that coincide with scientific projections impact these actions. Section 5 concludes the paper.

2. Data and descriptive statistics

2.1. Survey data

The biennial Survey of Rural Decision Makers (SRDM) collects detailed information on issues of topical interest to primary industry in New Zealand. The survey was enumerated online.⁶ The questionnaire was reviewed under Manaaki Whenua – Landcare Research’s social ethics process (approval number 2021/46NK) and informed consent was obtained on the first page of the survey. Open from 1 June until 15 August, the 2021 wave of the survey yielded 6,740 complete responses. Respondents represent the entirety of the primary industry as well as lifestyle block owners (i.e. ‘hobby farmers’) in all 66 districts in New Zealand. Because the potential impacts of changing climate are more significant for commercial farmers and because commercial farming is a backbone of the New Zealand economy, we exclude lifestyle block owners from our analysis, reducing our sample to 2,304. Demographics are representative of the primary sector as a whole (Stahlmann-Brown, 2021).

The 2021 wave of the survey consisted of 130 potential questions and 599 potential datapoints for each respondent, although branching and randomisation meant that no respondent saw every question. The questionnaire emphasised land use, land-use change, land-management practices, personal values and motivation, and expectations regarding future climate. Respondent demographics and postal codes were also recorded.

The first question pertaining to future climate asked whether the respondent believed that climate change is real, without asking the underlying cause. The question wording and answer set are shown in the Appendix.

Fewer than 10% of respondents selected ‘unsure’. These responses were dropped from the analysis. Among the remaining sample, 63.0% believed that climate change is real and already affecting New Zealand, 22.0% believed that climate change is real and will affect New Zealand in the future, 3.0% believed that climate change is real but will not affect New Zealand, and 12.0% believed that climate change is not real. See Table 1. For our purposes, belief in climate change is constructed as a dummy variable that is equal to 1 if the respondent selected one of the first three options and 0 otherwise.

Table 1. Descriptive statistics.

	Obs.	Mean	Std. dev.	Min	Max
Belief
CC real, already affecting	1,673	0.63		0	1
CC real, will affect 10y	1,673	0.13		0	1
CC real, will affect >10y	1,673	0.09		0	1
CC real, will not affect	1,673	0.03		0	1
CC not real	1,673	0.12		0	1
CC believer	1,673	0.85	0.36	0	1
Drought expectations	1,673
Decrease a lot		0.000
Decrease a little		0.003
No change		0.33
Increase slightly		0.39
Increase a lot		0.28
Heat waves	1,275
Decrease a lot		0.000
Decrease slightly		0.004
No change		0.44
Increase slightly		0.35
Increase a lot		0.21
Age	1,673	59.76	11.86	22	96
Male	1,673	0.76	0.43	0	1
Generations of family farming	1,673	3.35	1.76	1	7
Education	1,673
Secondary or less		0.26		0	1
Professional training		0.32		0	1
University		0.41		0	1
Experimenting	1,673	4.56	2.78	0	10
Values	1,673	6.25	2.29	0	10
Trust in social media	1,673	0.04	0.19	0	1
Area	1,673	429.58	1,532.59	0.1	45,000
Pred: Temperature increase	1,673	0.94	0.11	0.608	1.37
Pred: PED increase	1,673	72.28	20.83	8.87	127.28
Exp vs projections: Temp	1,275
Under		0.57		0	1
Coinciding		0.30		0	1
Over		0.12		0	1
Exp vs projections: Drought	1,450
Under		0.47		0	1
Coinciding		0.37		0	1
Over		0.16		0	1
Actions
Greenhouse Gas Past	1,668	0.28	0.45	0	1
Greenhouse Gas Future	1,662	0.47	0.50	0	1
Biodiversity Past	1,668	0.47	0.50	0	1
Biodiversity Future	1,662	0.56	0.50	0	1
Waterways Past	1,668	0.73	0.44	0	1
Waterways Future	1,662	0.73	0.44	0	1
Soil Erosion Past	1,668	0.62	0.49	0	1
Soil Erosion Future	1,662	0.62	0.49	0	1
Biosecurity Past	1,668	0.72	0.45	0	1
Biosecurity Future	1,662	0.74	0.44	0	1

Notes: Difference is defined as expectation – prediction.

Respondents who believe in climate change and who believe that climate change is either already affecting New Zealand or will affect New Zealand in the future were additionally asked questions about expectations regarding the intensity of drought and heat waves in the future; specifically they were asked to indicate whether they believed that the intensity would decrease a lot, decrease somewhat, stay the same, increase somewhat, or increase a lot. The question wording and answer set is shown in the Appendix.

Dropping the 3.7% of observations for temperature and 2.6% of observations for drought who responded ‘unsure’, we find that 0.3% expect draught to decrease somewhat, 33% expect there to be no change, 39% expect drought to increase somewhat, and 28% expect drought to increase a lot. Similarly, 0.4% expect heat waves to decrease somewhat, 44% expect there to be no change, 35% expect heat waves to increase somewhat, and 21% expect heat waves to increase a lot.

The now extensive literature on climate change perceptions has delivered largely consistent conclusions that older people, men, and less educated individuals are more sceptical of climate change and its anthropogenic causes (e.g. Lewis et al., 2019; McCright et al., 2016; Poortinga et al., 2011), including in New Zealand (Milfont et al., 2015; Swerdloff et al., 2023). Studies that have focused on farmers have further identified that generations of family farming history – an indicator of intergenerational experience and knowledge (Booth et al., 2020; Fiske, 2016; Stahlmann-Brown & Walsh, 2022) – and size of the property (Davidson et al., 2019) also influence climate belief and climate-rated behaviour. Hence, we include these demographic factors in our analysis. Years of age and gender are measured as expected. Education is a categorical variable indicating whether the respondent has secondary education or less, professional training in the form of a certificate or diploma, or higher education in the form of a bachelor’s, master’s, doctoral degree, or postgraduate diploma. Intergenerational farm experience is taken from the question ‘How many generations has your family been involved in farming, forestry, and growing food in Aotearoa New Zealand?’. The average respondent is male (76%), is 60 years old, is a third-generation farmer, and has a university degree (41%). The average farm size is 430 hectares.

Personal values also strongly inform expectations about future climate (Hornsey et al., 2016; Poortinga et al., 2011; Whitmarsh & Capstick, 2018). Hence, we include risk tolerance and social norms as additional covariates. Risk tolerance, which is negatively associated with climate adaptation (e.g. Sara et al., 2016) is measured via the extent to which respondents agree with the statement ‘I/we prefer leaving experimenting with new ideas to someone else’, recorded on a scale from 0 (strongly disagree) to 10 (strongly agree). Social norms, which are positively associated with climate belief and adaptation (Cialdini & Jacobson, 2021), are measured via the extent to which respondents agree with the statement ‘My/our values are similar to those of other commercial operators in the industry’, recorded on a scale from 0 (strongly disagree) to 10 (strongly agree). We find that the average respondent is risk-neutral and that more respondents are risk-loving than risk-averse (44% vs. 19%). The average score for social norms is 6.3.

Finally, while social media has contributed to increased polarisation regarding climate change and its causes (Williams et al., 2015), a meta-analysis of surveys conducted in 20 countries shows that higher trust social media platforms (e.g. YouTube, Facebook, and Twitter) as a news source is associated with lower climate scepticism (Diehl et al., 2019). Thus, we also include a dummy variable to indicate whether survey respondents consider social media to be among the most important sources of information on topics (including climate change) related to their industry. In our sample, 4% of respondents consider Facebook or YouTube among the most important sources of information related to their industry.

2.2. Climate data

Scientific projections of future temperature and drought are based on RCP 4.5, a trajectory of greenhouse gases adopted by the IPCC (IPCC, 2007; Moss et al., 2010) used in the fifth Assessment Report in 2014. RCP 4.5 is considered a low-moderate scenario in which emissions peak between 2040 and 2045 and decline thereafter (IPCC, 2014). It is considered the most probable baseline scenario under the assumption of no climate policies and exhaustion of non-renewable energy sources (Meinshausen et al., 2011).

Temperature predictions are measured in degrees Celsius. Drought is measured using Potential Evapotranspiration Deficit (PED) (Sheffield et al., 2012). PED is a continuous, non-negative variable based on the water-balance model and defined as potential evapotranspiration less actual evapotranspiration. This measure can be thought of as millimetres (mm) of water needed by vegetation to grow under no water shortage (Ministry for the Environment, 2018; Prudhomme & Williamson, 2013). Because PED uses both water demanded by and water available through the environment, it is a robust measure of drought severity, especially in agricultural production (Ministry for the Environment, 2018).

We mapped a high-resolution raster layer of projections of temperature and PED onto a polygon layer of New Zealand postal codes to generate expected changes within each postcode in New Zealand. Under RCP 4.5, temperatures in the average New Zealand postal code will increase by 0.95 degrees Celsius, with a minimum of 0.61 degrees Celsius and a maximum of 1.37 degrees Celsius (Figure 1, left panel and Table 1). PED will also increase in all postcodes in New Zealand by an average of 73 mm. Again, the range is substantial from 8.87 mm to 127.28 mm (Figure 1, right panel and Table 1).⁷

Figure 1. Histogram of IPCC predictions of temperature and drought at the post code level.

Notes: Kernel density estimates shown in red. Black, vertical line represent cut-off points for the mapping of predictions to expectations.

Given that the mean and median predicted temperature increase is about 1 degree Celsius, we categorise temperatures in postal codes with predicted increases of 0.6-1.0 degrees as ‘increasing somewhat’ and temperatures in postal codes with predicted increases >1.0 degrees as ‘increasing a lot’ to coincide with the survey question.⁸ For drought, we use the 10th percentile (42 mm) as the first cut-off and the median (77 mm) as the second cut-off. That is, predicted changes in PED of 0-42 mm is considered ‘no change’, 42.1-77 mm is considered to ‘increase somewhat’, and >77 mm is considered to ‘increase a lot’. These cut-off points are depicted as black vertical lines in Figure 1. Although our cut-offs are arbitrary, we undertook extensive robustness checks and found that our results are robust to significantly different cut-offs, as shown in Table A.3 in the Appendix.

2.3. Comparing survey respondents’ expectations of future temperature and drought to scientific projections

In Figure 2, we compare the classifications of IPCC projections described above (in black outline) to respondents’ expectations based on survey data (in solid blue) within postal codes. We find that respondents’ expectations of future temperatures coincide with IPCC projection 34% of the time (See Table 1). Respondents’ expectations diverge from IPCC expectations regarding future temperature 64% of the time, most often by underestimating future temperature vis-à-vis IPCC projections. Similarly, respondents’ expectations of future drought coincide with IPCC projections 40% of the time and diverge 60% of the time. In cases of divergence, survey respondents most commonly project less severe drought conditions than the IPCC. The correlation between expectations of drought that coincide with scientific projections and expectations of temperature that coincide with scientific projections is low at ρ = 0.06.

Figure 2. Mapping projections and expectations.

Notes: Figure shows mapping from IPCC projections (blue bar) to farmers’ expectations (grey bar). Left panel shows temperature and right panel shows drought. Observations with ‘unsure’ responses dropped.

Based on these mappings, we construct two variables indicating whether respondents’ expectations of temperature and drought coincide with IPCC projections. These variables are defined as dummies that take a value of 1 if the expectation and projection coincide and a value of 0 otherwise.

3. Correlates of belief in climate change and expectations that coincide with IPCC projections

We estimate the following fixed-effects logit model: (1) $Y_{i,j,k}=\alpha +\beta X_{i,j,k}+{\theta }_j+{\vartheta }_k+{\epsilon }_{i,j,k},$(1) where $Y_{i,j,k}$ is a binary variable indicating either that farmer i at location j in industry k believes in climate change or, in separate regressions, has expectations about temperature or drought that coincide with scientific projections, and (2) $\mathrm{Prob}(Y_{i,j,k})=F(\alpha +\beta X_{i,j,k}+{\theta }_j+{\vartheta }_k).$(2) Here, the vector $X_{i,j,k}$ contains our set of covariates. We consider two sets of control variables: a core version with strictly exogenous variables and an extended version with further controls that we believe to be potentially important variables but that are less clearly exogenous. The core version includes age, age squared, male, education, and generations of family farming history. The extended model additionally includes the log of the land area, values, trust in social media, and risk tolerance. All versions include region fixed effects, ${\theta }_j$ and industry fixed effects, ${\vartheta }_k$.⁹ Standard errors are robust to heteroscedasticity.

Since believing in climate change and having beliefs that coincide with scientific projections are binary variables, we use a logit estimator. However, to assess the variance of beliefs for scientific projections, we use an ordered logit estimator and present the results graphically. Further, as mentioned above, we undertook extensive robustness checks and found that our results are robust to significantly varying our classifications of respondents’ future expectations.

3.1. Belief in climate change

Our regression results for belief in climate change are presented in Table 2 columns 1 (core version) and 4 (extended version). We find that the probability of believing in climate change is not affected by age among our sample of commercial farmers. We also find that the probability of believing in climate change is inversely related to the number of generations one’s family has been farming. Further, men are 5.2% less likely to believe in climate change than women. Finally, we find that higher education is positively associated with belief in climate change.

Table 2. Regression results

	(1)	(2)	(3)	(4)	(5)	(6)
	Climate change believer	Coinciding expectations: temperature	Coinciding expectations: drought	Climate change believer	Coinciding expectations: temperature	Coinciding expectations: drought
Age	0.006	0.005	−0.014*	0.005	0.007	−0.017**
	(0.005)	(0.008)	(0.007)	(0.005)	(0.008)	(0.008)
Age^2	−0.000	−0.000	0.000*	−0.000	−0.000	0.000*
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Male	−0.052**	−0.054*	−0.044	−0.068***	−0.050*	−0.049*
	(0.022)	(0.028)	(0.028)	(0.022)	(0.030)	(0.030)
Generations	−0.010**	−0.015**	−0.016**	−0.008*	−0.015**	−0.016**
	(0.005)	(0.007)	(0.007)	(0.005)	(0.007)	(0.007)
Education
Professional Training	0.062***	0.104***	0.010	0.047**	0.093***	0.011
	(0.023)	(0.032)	(0.031)	(0.023)	(0.034)	(0.033)
University	0.105***	0.103***	0.059*	0.087***	0.095***	0.035
	(0.022)	(0.030)	(0.031)	(0.022)	(0.032)	(0.032)
Log(Area)				0.001	−0.002	0.012
				(0.007)	(0.010)	(0.010)
Values				0.007*	−0.006	−0.005
				(0.004)	(0.005)	(0.005)
Trust in Social Media				−0.104***	−0.135*	−0.042
				(0.039)	(0.071)	(0.070)
Experimenting				−0.007**	−0.003	−0.005
				(0.003)	(0.005)	(0.004)
Obs.	1,850	1,362	1,573	1,729	1,279	1,473
Pseudo R^2	0.03	0.04	0.03	0.06	0.04	0.03
Fixed Effects
Region	YES	YES	YES	YES	YES	YES
Industry	NO	NO	NO	YES	YES	YES

Notes: Tables show average marginal effects from logit regressions. Robust standard errors in parenthesis. Constant not shown. Significance levels: *: p < 0.1, **: p < 0.05, ***: p < 0.01.

These results are confirmed in the extended version (column 4), We additionally find that values, trust in social media, and risk preferences matter. Specifically, having similar values to other operators in the industry increases the probability of believing in climate change, although the effect is quantitatively small. Similarly, preferring to leave experimenting to others is negatively associated with belief in climate change. Further, those who rely on Facebook and YouTube as key sources of information are 10.4% less likely to believe in climate change.

3.2. Concordance of expectations and scientific projections

Columns 2 (core version) and 5 (extended version) of Table 2 present the results for having expectations that coincide with scientific projections of future temperatures. We find similar effects of the core control variables on the probability of having expectations of temperature that coincide with scientific projections. Specifically, age has no effect on having temperature expectations that coincide with projections. Men are 5.0% less likely than women to have expectations that accord with scientific projections. Similarly, each generation of family farming experience is associated with a 1.5% reduction in having expectations of future temperature that coincide with scientific projections. Finally, higher education levels are positively associated with having temperature expectations that coincide with projections.

The probability of having expectations about drought that coincide with scientific projections shows different underlying factors compared to temperature expectations (Table 2, columns 3 and 6). In contrast to the previous results, we find that age has a U-shaped relationship with the probability of having concordant drought expectations. The trough occurs at 65 years of age, implying that younger and older respondents are more likely to have expectations of drought that coincide with scientific projections. Being male, as before, reduces the likelihood of having drought expectations that coincide with projections, by 4.9%. Higher education is positively associated with having drought expectations that coincide with scientific projections in the core model but not the extended model. Finally, none of the extended variables associated with having expectations that coincide with scientific projections.

We can also evaluate whether respondents underestimate or overestimate future drought conditions relative to scientific projections using an ordered logit model. These results are presented graphically in Figures 3 and 4.

Figure 3. Determinants of concordance in expectations with IPCC projections of future temperature.

Notes: Results from ordered logit regression with 95% confidence bands. All regressions include industry and region fixed effects. Constant not shown.

Figure 4. Determinants of concordance in expectations with IPCC projections of future drought.

Notes: Results from ordered logit regression with 95% confidence bands. All regressions include industry and region fixed effects. Constant not shown.

In brief, we find that underestimating temperature (relative to projections) is the most common outcome across all levels of all variables in our model. The probability of underestimating temperatures is higher for men than for women. Further, more respondents had temperature expectations that coincided with projections than had expectations that overestimated temperatures (by about 15 percentage points). The probability of underestimating temperature is by far the largest among 40-year-olds (about 80%) and decreases over age. Again, having expectations that coincide with scientific projections is more likely compared to overestimating temperatures; however, the gap is smaller at approximately 10 percentage points. For all three education categories, underestimating temperatures is most likely, having expectations that coincide with scientific projections is second most likely, and overestimating temperatures is least likely. The probability of underestimating temperatures decreases with education, with the probability falling from about 60% to about 40%. For generations of family farming history, underestimating temperatures is the most likely outcome, and we find little variation over the levels of generations. The last panel indicates that respondents in the forestry and horticulture industries are most likely to have temperature expectations that coincide with scientific projections. In contrast, respondents in the arable, diary, sheep and beef, and other livestock industries are most likely to underestimate future temperatures relative to scientific projections.

Women are most likely to have expectations of drought that concord with science while men are more likely to underestimate drought. For age, we find that there is no statistically significant difference between holding expectations that coincide with science and underestimating drought. Underestimating drought relative to scientific projections is most common across the number of generations one’s family has been involved in farming. Higher education is positively associated with having expectations of drought that coincide with scientific projections. Finally, we find that dairy, sheep and beef, and other livestock farmers are most likely to underestimate drought. We note that for dairy farmers at least, this finding may be associated with evidence that post-drought milk prices may rise (e.g. Pourzand et al., 2020). For the remaining industries (forestry, horticulture, and arable), we find no statistically significant difference between having expectations that concord with science and underestimating drought vis-à-vis scientific projections.

4. Pro-Environmental action

Next, we analyse the extent to which belief in climate change and holding expectations of future temperature and drought that coincide with scientific projections impacts the past and intended uptake of pro-environmental actions, including actions that are directly related to climate change (namely, reducing greenhouse gases) and actions that are less clearly related to climate change (namely, increasing native biodiversity on the property; improving the health of waterways on the property; reducing soil erosion; and managing biosecurity on the property). For each action, respondents could select ‘not much of a focus’, ‘minor focus’, ‘moderate focus’, or ‘major focus’. The question is shown in full in the Appendix.

We consider past and intended actions separately because past actions have already been implemented whereas future plans may not actually come to fruition, a reflection of adaptive capacity (Yletyinen et al., 2024). For each outcome variable, we construct a dummy variable which equals 1 if the respondent answered that an action has been (or will be) a moderate or major focus and 0 otherwise.

In our sample, 28% of farmers have focused on measures to reduce greenhouse gas emissions in the past and 47% stated that this will be a focus for the future (Table 1). For the broader measure of environmental actions, we find that 47% have put actions in place to increase biodiversity and 56% plan to do so. Actions to improve the health of waterways have been implemented by 73% of respondents and the same share plans to do so. Actions to reduce soil erosion have been implemented by 62% of respondents and 62% of respondents plan to do so in the future. Finally, 72% of respondents have put actions into place to manage biosecurity and 74% say that they will do so in the future.

Table A.5 in the Appendix presents analogous statistics by industry. Across industries, we find higher intentions for pro-environmental actions in the future compared to already implemented actions. Starting with already implemented actions, reducing greenhouse gas emissions has been highest in the forestry and dairy sectors. For improving native biodiversity, the sheep and beef and forestry sectors have seen the highest uptake. For improving the health of waterways, the dairy and sheep and beef sectors lead the statistics. The arable sector and the dairy sector have the largest uptake of actions aimed at reducing soil erosion. Actions to manage biosecurity are most common in the diary and arable sectors. In the future, intentions to reduce greenhouse gas emissions are highest in the dairy and arable sectors. Intentions to improve biodiversity are highest in the sheep and beef and forestry sectors. Intentions to improve waterways are highest in the sheep and beef and dairy sectors. Intentions aimed at reducing soil erosion are highest in the arable and sheep and beef sectors. Finally, intentions to improve biosecurity are highest in the dairy and the arable sectors.

The estimated impact of belief in climate change on past actions and future intentions are presented in Table 3.¹⁰ We consider both past actions (first row) and future intentions (second row). For parsimony, we only present the estimated effects of belief in climate change in Table 3, but full results are reported in the Appendix (Table A.4)

Table 3. Regression results.

	(1)	(2)	(3)	(4)	(5)
	Reducing greenhouse gases	Increasing native biodiversity	Improving health of waterways	Reducing soil erosion	Managing biosecurity
Past 2 years	0.169***	0.113***	0.077***	0.072**	0.040
	(0.036)	(0.035)	(0.029)	(0.033)	(0.030)
Next 2 years	0.297***	0.178***	0.104***	0.101***	0.074***
	(0.035)	(0.032)	(0.028)	(0.032)	(0.028)

Notes: Tables show average marginal effects from logit regressions. Controls not shown (see Table A.4 in the Appendix). All regressions include industry and region fixed effects. Robust standard errors in parenthesis. Constant not shown. Significance levels: *: p < 0.1, **: p < 0.05, ***: p < 0.01.

Column (1) focuses on reducing greenhouse emissions. We find that belief in climate change is positively associated with both having already implemented actions to reduce greenhouse and having future plans to do so. The effect is larger for forward-looking plans compared to backward-looking implementation: respondents who believe in climate change are 17% more likely to have previously undertaken action to reduce greenhouse gases and 30% more likely to have plans to do so.

This pattern holds for other pro-environmental actions: Column (2) considers native biodiversity actions, column (3) the health of waterways, column (4) the reduction of soil erosion, and column (5) managing biodiversity. Only for previously undertaken action for biosecurity does belief in climate change not matter.

We also tested whether holding expectations regarding future temperature and drought that coincide with scientific projections have an additional effect on actions. To do so, we ran a Heckman selection model in which the first stage modelled belief in climate change according to the full model used above (cf. Table 2, column 4). The second stage then estimated the effect of holding expectations that coincide with scientific projections on actions in the sample of climate change believers. Table 4 presents our results. We find that only two out of 20 estimates of expectations are significant at the 5% level. That is, having expectations that coincide with scientific projections of future temperature and drought does not make farmers more likely to have undertaken action or to have plans to undertake action. The only significant effect is found for past and future improvements of the health of waterways, which is positively affected by having temperature expectations in line with scientific projections (see column 4 in Table 4). Importantly, this leads us to the conclusion that belief in climate change alone is sufficient to affect actions and that having expectations that coincide with scientific predictions has no additional impact on actions.

Table 4. Regression results.

	(1)	(2)	(3)	(4)	(5)
	Reducing greenhouse gases	Increasing native biodiversity	Improving health of waterways	Reducing soil erosion	Managing biosecurity
Past 2 Years
Coinciding expect:	0.054	0.009	0.067**	−0.007	−0.023
temperature	(0.036)	(0.033)	(0.030)	(0.031)	(0.024)
Coinciding expect:	0.046	0.028	0.019	0.027	0.035
drought	(0.029)	(0.032)	(0.029)	(0.030)	(0.027)
Obs.	1,234	1,234	1,234	1,234	1,234
Log Pseudolikelihood	−1,159.08	−1,216.06	−1,076.23	−1,154.80	−1,118.34
Next 2 Years
Coinciding expect:	0.025	−0.018	0.046**	−0.016	−0.022
temperature	(0.035)	(0.026)	(0.023)	(0.030)	(0.022)
Coinciding expect:	0.006	0.036	0.012	0.001	0.028
drought	(0.030)	(0.025)	(0.022)	(0.029)	(0.022)
Obs.	1,230	1,230	1,230	1,230	1,230
Log Pseudolikelihood	−1,216.36	−1,176.70	−1,061.27	−1,150.43	−1,086.94

Notes: Tables show average marginal effects from Heckman-probit regressions, which includes all controls from the model used in Table 2 and region and industry fixed effects. Robust standard errors in parenthesis. Constant not shown. Significance levels: *: p < 0.1, **: p < 0.05, ***: p < 0.01.

Finally, Tables A.6 to A.10 in the Appendix present the same analysis while restricting the sample to New Zealand’s three largest primary industries, i.e. sheep and beef, dairy, and horticulture. While the results for the sheep and beef sector are in line with the results presented in Table 3, we only find a significant effect of climate change beliefs on greenhouse gas emissions (past and planned) for the dairy sector. For horticulture, we only find a significant relationship between climate change beliefs and increasing native biodiversity in the future and improving the health of waterways in the future. For the Heckman selection model, we find no significant effects of expectations coinciding with scientific projections for the sheep and beef sector and only one significant effect for the diary sector: namely, expectations coinciding with scientific projections for drought lead to more planned actions to increase native biodiversity (for horticulture, the Heckman selection model failed to converge). The caveat of this exploratory analysis is that the sample sizes become small, making it hard to achieve statistical significance even if coefficients are large.

5. Conclusion

Given the primacy of primary industry for the New Zealand economy (Beef+Lamb New Zealand, 2022) and the potential for drought and heat waves to negatively impact farm production over time (e.g. Polsky & von Keyserlingk, 2017; West, 2003), it is critical to understand farmers’ adaptation behaviours and intentions. We begin this study by analysing farmers’ beliefs about climate change and whether their expectations regarding future temperatures and drought coincide with scientific projections. Our results show that 88% of farmers believe in climate change but only 34% (40%) have expectations regarding future temperature (drought) that coincide with IPCC projections under RCP 4.5. Farmers tend to underestimate future drought and temperature relative to scientific projections.

We document substantial differences in the correlates of belief in climate change across different groups of respondents. Our results support the literature that shows that men are more sceptical of climate change and that scepticism decreases with education (Borick & Rabe, 2010; Hornsey et al., 2016; Shao & Goidel, 2016; Whitmarsh, 2011). In addition, those whose families have long histories of farming are less likely to believe in climate change, in line with findings that long family histories of facing challenges posed by nature instil a sense of controllability and manageability (Menapace et al., 2015) and dampen risk perceptions (Weber et al., 2002). We also show that higher trust in social media is associated with lower belief in climate change, which is consistent with the notion that disengagement with climate change is in part caused by distrust in political actors (Bickerstaff et al., 2004; Poortinga & Pidgeon, 2006). Finally, pro-social belief (specifically, believing that one’s own values are similar to peers’ values) and willingness to embrace risk are associated with higher belief in climate change.

In contrast to previous studies, we also describe the correlates of having ‘correct’ expectations of future temperature and drought, i.e. of holding beliefs about future climatic conditions that coincide with scientific projections. Conditional on believing in climate change, men’s expectations of both future temperature and drought deviate from scientific projections more than women’s expectations. The same holds for less educated respondents vis-à-vis more educated respondents. We also find that the number of generations one’s family has been farming is negatively associated with having expectations that coincide with the scientific consensus for temperature, but not for drought. Similarly, a high reliance on social media such as YouTube and Facebook for information related to farming is negatively correlated with having expectations that coincide with the scientific consensus for temperature, but not for drought. The direction of the difference also depends on these factors. For example, women and more educated people are less likely to underestimate and more likely to overestimate both temperature change and the incidence of drought relative to scientific projections. The accuracy of climatic expectations also varies by industry, which may stem from short-term income boosts in livestock industries due to price shocks and livestock sales (Pourzand, 2023; Pourzand et al., 2020).

We further show that belief in climate change is positively associated with pro-environmental action – both previously undertaken and intended for the future, consistent with established findings (e.g. Blennow & Persson, 2009; Mitter et al., 2019). The effects of belief in climate change are amplified for climate-friendly actions such as taking measures to reduce greenhouse gases. Perhaps surprisingly, the concordance of climate expectations with scientific projections is not strongly associated with greater uptake of past action or higher intentions to undertake action in the future.

Between the 2019 Zero Carbon Amendment Act, the 2021 Nationally Determined Contribution to the Paris Agreement, the rising prominence of the He Waka Eke Noa – Primary Sector Climate Action Partnership, the 2021 release of the ‘Good Farm Planning Principles: Towards Integrated Farm Planning’ guide, and the 15-month drought that ended in April 2021, climate policy was firmly established in the rural discourse at the time of the survey. The fact that belief in climate change itself is sufficient for action is a positive result in a country where the vast majority of farmers believe in climate change but fewer than 40% have an understanding of likely future climactic patterns in their local areas that concords with scientific projections.

Of course, intention to adapt and actual adaptive behaviour do not necessarily coincide, and intentions are worthless without follow-through. According to the Model of Private Proactive Adaption to Climate Change, people first appraise both risks and their own adaptive capacities before undertaking adaptation actions (Grothmann & Patt, 2005). Adaptive responses are made if perceptions of both the risk and perceived adaptive capacity are high; however, behavioural intentions are often overestimated, and intentions are not carried out if there is a lack of objective adaptive capacity. Maladaptive responses may also result: For example, planting exotic forestry to reduce greenhouse gas emissions poses a long-term risk to biodiversity (McGlone & Walker, 2011).

To help farmers manage the risks associated with climate change and increase adoption of pro-environmental actions, policy makers and industry must understand the factors underlying whether and how farmers perceive the threat. They must also understand how these perceptions inform mitigation strategies, whether previously undertaken or planned for the future. Our findings bolster arguments for the segmented approach to information dissemination and farmer support for mitigation activities advocated by Arbuckle et al. (2013). In particular, policy makers and industry should consider how demographic characteristics, education, risk preferences and values, information sources, and farm-specific needs shape climate beliefs – and thus willingness to act – when designing educational campaigns about climate change. Climate information campaigns targeted at farmers have already proved to increase mitigation in the European context. For example, Alif et al. (2024) show that cattle-rearing farmers in Slovenia who participated in targeted workshops had higher belief in climate change, more positive attitudes toward on-farm climate mitigation, and greater perceived behavioural control. These attitudinal changes resulted in a 19% higher intention to perform mitigation than their peers. Similar workshops may pay dividends in New Zealand, where primary industry is notable for its contributions to both the national economy and our greenhouse gas emissions.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Other studies on the relationship between beliefs and climate mitigation and adaptation include Burgeno and Joslyn (2020) and Guido et al. (2021), who study short-term forecasting of weather. See also the survey by Mase and Prokopy (2014) for studies on forecasting.

2 The high share of greenhouse gas emissions originating from agriculture is due both to the economic prevalence of farming – particularly dairy, sheep, and beef cattle – and the high share of renewables in the electricity mix.

3 ‘He Waka Eke Noa’ can be translated as “we’re all in this together”. Their programme of work emphasised extension and advisory services aimed at measuring, managing, and reducing on-farm emissions and research into mitigation technologies such as methane inhibitors and a methane vaccine. In late 2021, He Waka Eke Noa released outlined alternative models for agricultural emissions pricing options, including a farm-level levy, a processor-level levy, and inclusion in the New Zealand Emissions Trading Scheme, which is widely seen as onerous.

4 Pourzand et al. (2020) further show that these increases disappear when controlling for milk prices.

5 Modelling undertaken in the US in 2003 estimated that heat stress would produce economic losses of 1.7­–2.4 billion USD annually, with most of the losses accruing to the dairy and beef industries (St-Pierre et al., 2003).

6 One criticism levied against online surveying is lack of accessibility, particularly for rural populations. However, New Zealand’s Internet penetration rate stood at 94.9% at the start of 2022 (Datareportal, 2022). The survey was also optimised for mobile devices to increase accessibility for those without high-speed Internet access in their homes.

7 Figure A.1 in the Appendix shows a scatter plot between temperature and drought, suggesting a positive correlation between the two variables.

8 The mean modelled increase in temperature is 0.95 degrees Celsius. The median is 0.96 degrees Celsius.

9 See Tables A.1 and A.2 in the Appendix for details on regions and industries.

10 Our results are confirmed when we use a principal component analysis which combines greenhouse gas emissions and the other, more broad, environmental outcomes (results available upon request).