Gendered implications for climate change adaptation among farmers in Madagascar

ABSTRACT

Climate change is impacting farmers worldwide; none more so than the millions of smallholder farmers who rely entirely on rain-fed agriculture. Adapting agricultural practices is an important strategy to mitigate climate change’s harmful impacts to food security and livelihoods. Researchers have increasingly sought to understand the factors influencing farmer decision-making on adaptation practices but application of important theories in low-income country contexts is lacking. Therefore, we apply the Protection Motivation Theory (PMT) to examine farmer intention to adopt agricultural practices in response to climate change based on a survey of 328 smallholder farmers in rural Madagascar. Using structural equation modeling (SEM) to describe the relationship between social connectedness (e.g., interactions with others about farming-related issues), PMT constructs (threat and coping appraisal), and intention to adapt agricultural practices to climate change, we find that coping appraisal is a stronger predictor of intention to adapt than threat appraisal. Further, SEM results show a significant negative relationship between social connectedness and threat appraisal – farmers with greater social connectedness perceive climate change to be less of a threat. Additionally, as female farmers are significantly less socially connected than male farmers, this study provides further evidence of the gendered vulnerabilities to climate change among smallholders. This research has implications for the design of policies and interventions that consider farmer social networks, critical for reducing threat and building coping capacity, in supporting adaptation to climate change among vulnerable farmer populations.

KEYWORDS:

• Smallholders
• climate change adaptation
• social connectedness
• gendered vulnerability
• Protection Motivation Theory (PMT)
• Structural Equation Modeling (SEM)
• Madagascar

1. Introduction

Across the world, farmers are being negatively impacted by changing climatic conditions (Karki et al., 2020a). This is especially true for the nearly 600 million smallholder farmers who rely on family labour for their food and livelihoods (Cornish, 1998; Lowder et al., 2016), many of whom depend on rain-fed agriculture (Fierros-González & López-Feldman, 2021). Thus, adaptation to climate change, defined as actions taken ‘to prepare for and adjust to both the current and projected impacts of climate change’ (EPA, 2022), has emerged as an important strategy to mitigate its harmful impacts (Bedeke, 2023). While research on the topic is growing, more studies are needed in regions most vulnerable to climate change (van Valkengoed & Steg, 2019).

Traditional adaptation measures among smallholders include changing crop types/varieties (Karki et al., 2020b), altering planting timing (Ali & Erenstein, 2017; Shikuku et al., 2017; Simelton et al., 2013), and implementing rainwater harvesting techniques (Gandure et al., 2013). However, farmers face many adaptation constraints, defined by Klein et al. (2014) as ‘a factor or process that makes adaptation planning and implementation more difficult’ (p. 906). The implications of inaction are potentially devastating, as they can lead to increased food insecurity, disrupted livelihoods, and heightened poverty, particularly among women (FAO, n.d.). Consequently, researchers aim to understand the reasons for the persistent climate change adaptation deficit.

Barriers to climate change adaptation include poverty (Deressa et al., 2008; Etana et al., 2020), food insecurity (Carranza & Niles, 2019; Shikuku et al., 2017), farmer health (Hogan et al., 2011), access to credit, markets, and information (Deressa et al., 2008; Ringler, 2010), institutional barriers (Rodríguez-Cruz et al., 2021), as well as cultural factors (Adger et al., 2013; Karki et al., 2020a), such as attachment to traditional customs. Gender also influences climate change adaptation outcomes (Bessah et al., 2021; Carranza & Niles, 2019; Macgregor, 2010; Ravera et al., 2016), as female farmers face specific obstacles due to unequal access to resources, such as land, credit, and extension services (FAO, 2011). There are also gendered differences in the use of assistance to cope with climate change, and which adaptation strategies and types of support are preferred (Assan et al., 2018; Codjoe et al., 2012). Given the multitude of gender differentiated impacts of climate change (Codjoe et al., 2012; FAO, n.d.), more research is urgently needed to understand the elements that support climate change adaptation among women smallholders (Harvey et al., 2018).

1.1. Farmer recognition of climate change as a precursor to adaptation

Agricultural adaptation to climate change has been described as a two-step process: recognizing that changes are occurring (Adger et al., 2009) and/or having a personal experience with extreme weather events (Dessai et al., 2004; Li et al., 2017); followed by taking action to implement adaptation measures (de Matos Carlos et al., 2020; Deressa et al., 2008). Therefore, a farmer’s recognition of climate change, defined as changes in weather over time (IPCC, 2007), is an important precursor to adaptation (e.g., Fierros-González & López-Feldman, 2021; Makuvaro et al., 2018; Simelton et al., 2013). However, societal gender norms often assign women to more ‘climate-sensitive’ activities such as food, fuel and water provisioning, which may result in differential perceptions of climate change between women and men (Codjoe et al., 2012; Hitayezu et al., 2017).

Furthermore, in the absence of climate and weather information services, farmers often rely on personal observations of temperature and precipitation patterns (Salerno et al., 2019). Studies in Global North and South contexts have examined farmers’ recognition of climate change to understand accuracy and its role in decision-making. There is ample evidence that farmers, especially those without irrigation, are acutely and accurately aware of changes (e.g., Fierros-González & López-Feldman, 2021; Nguyen et al., 2021; Roco et al., 2015). For example, Soubry et al. (2020) found that 87% of Global South papers reported farmers’ perceptions of climate change aligning with historical climate records.

1.2. Social connectedness and access to information

It is widely recognized that access to information, such as through advisory services and group membership (e.g., farmers’ associations), supports farmers’ decisions regarding adopting agricultural practices (Ali & Erenstein, 2017; Dang et al., 2019; Nguyen et al., 2021), and can shape climate change adaptation outcomes. For example, training support, what Kashem (1987) calls ‘[farmer] contacts with information sources’ (p.128), can be an important predictor of farmer adoption of agricultural innovations (e.g., Genius et al., 2014; Meijer et al., 2015). As Ricker-Gilbert et al. (2008) found among farmers learning integrated pest management (IPM) techniques in Bangladesh, visits from extension agents and farmer field schools were instrumental to farmer adoption of IPM.

Furthermore, social networks, such as farmer cooperatives, which allow farmers to interact and develop trusted relationships, are recognized as vital for supporting farmers’ capacity to adapt to climate change (Harvey et al., 2018; Hogan et al., 2011; Lubell et al., 2014; Niles & Hammond Wagner, 2019). However, female farmers generally have less access to information and agricultural extension services, as well as lower participation in associations (Jost et al., 2016; Roncoli, 2006). Enhancing social connectedness can be a valuable strategy to promote adaptation. Indeed, Kafeety et al. (2020) recommend interventions ‘rooted in social connection’ be used to support behavioural adaptation to extreme heat events.

1.3. Farmers’ perceptions of climate-related risk and adaptive capacity

Studies have also shown that both perceived adaptive capacity, defined as the ‘extent to which [actors] feel prepared to endure changes and take necessary steps to cope with them’ (Seara et al., 2016), and objective forms of adaptive capacity, such as access to various assets or capital, play crucial roles in agriculture adaptation (Gardezi & Arbuckle, 2019; Shah et al., 2019; Singh et al., 2016). Smallholders, particularly in remote agrarian settings, often have very limited adaptive capacity (Adger et al., 2003; Klein & Nicolls, 1999). However, until fairly recently, most research primarily focused on resource limitations and overlooked the cognitive dimensions of climate change adaptation (Grothmann & Patt, 2005).

Furthermore, while there is a growing body of research looking at farmers’ perceptions of climate-related risk in the Global North (e.g., Arbuckle et al., 2013; Schattman et al., 2016; Takahashi et al., 2016), less work has examined the role of perceived threat from climate change in the Global South. For example, in a systematic review on farmer decision-making related to climate change in the Global South from 2007–2017 (Waldman et al., 2020), the authors find just 11.5% (n = 17) of papers used cognitive approaches to examine adaptation behavior, indicating a clear need for more research among rural agrarian communities in low-income countries.

Therefore, the objective of this study is to fill this gap by applying the Protection Motivation Theory (PMT), a theoretical framework that combines assessment of both threat and adaptive capacity, to better understand climate change adaptation decision-making among male and female smallholder farmers in an extremely vulnerable context. Specifically, after establishing that farmers are experiencing changes in climate, we examine the direct role that threat and adaptive capacity play in forming intentions to adapt agricultural practices (Figure 1). We also examine how climate change perception, social connectedness and prior adaptation of agricultural practices predict intention to adapt practices in the future, as well as how the predictive power of social connectedness is moderated by gender.

Figure 1. Conceptual framework for study.

1.4. Theoretical framework

Protection Motivation Theory (PMT; Rogers, 1983, 1975) attempts to explain the psychological processes behind behaviour change in response to perceived threats. Originally applied to health-related self-protection behaviour (Floyd et al., 2000; Milne et al., 2000), PMT has been used in the context of natural hazards such as floods, wildfires, and earthquakes (e.g., Babcicky & Seebauer, 2019; Bamberg et al., 2017; Grothmann & Reusswig, 2006; Westcott et al., 2017), and pro-environmental behaviour (e.g., Bockarjova & Steg, 2014; Chen et al., 2020; Cismaru et al., 2011; Tapsuwan & Rongrongmuang, 2015) and human migration decisions (Mallick et al., 2022), in response to climate change. PMT is increasingly used in research on farmer adaptation to climate change in both the Global North (Buelow & Cradock-Henry, 2018) and South (e.g., Bagagnan et al., 2019; Feng et al., 2017; Grothmann & Patt, 2005; Le Dang et al., 2014; Luu et al., 2019; Nguyen et al., 2021; Truelove et al., 2015), and researchers have found PMT suitable for understanding factors motivating climate change adaptation behaviour (van Valkengoed & Steg, 2019). In the past decade, quantitative studies on smallholder farmer decision-making processes incorporating a psychological approach have utilized PMT, or its variations, as a theoretical framework (Waldman et al., 2020).

PMT consists of two main constructs – threat appraisal and coping appraisal – that predict intentions and subsequent protective actions (Figure 2). Threat appraisal encompasses perceived risk (e.g., from climate change) and fear of the impacts, while coping appraisal is comprised of response efficacy, the belief that methods available are effective in protecting against the threat, and perceived self-efficacy (or adaptive capacity), the belief that one is capable of taking the actions necessary to reduce the threat (Plotnikoff & Trinh, 2010). Ordered PMT (Tanner et al., 1991) extends PMT by assuming a sequential relationship between threat appraisal, fear reaction, and coping appraisal. This leads to intention setting and protective responses based on levels of perceived threat and coping ability. While the literature often focuses on factors leading to intention-setting, PMT can also address the three stages of adaptation: perception, intention, and adaptation (Abid et al., 2019).

Figure 2. Theoretical framework using PMT applied to climate change adaptation, after Rogers (1983) and adapted from Abid et al. (2019), Babcicky and Seebauer (2019), and Grothmann and Patt (2005). Perception and intention stages (dotted outline) included in this study.

Note: *= not included in this study.

The components of PMT have been found to support farmer decision-making in response to climate change (van Valkengoed & Steg, 2019). When farmers perceive climate change to be a threat (high threat appraisal) and have confidence in their coping abilities (high coping appraisal), they are more likely to intend to adapt their agricultural practices. However, while some studies find that higher levels of concern about changing conditions among farmers correspond to farmer adaptation (e.g., Luu et al., 2019; Woods et al., 2017), others maintain that if perceived threat is not met with sufficiently high coping appraisal, protection motivation may not occur (Babcicky & Seebauer, 2019).

Indeed, while there is debate in the climate change communication literature (e.g., Tannenbaum et al., 2015), research shows that threat-oriented fear appeals (e.g., Tunner et al., 1989), as well as fear-based messaging around climate change, are largely ineffective at influencing attitudes (e.g., Armbruster et al., 2022; Spence & Pidgeon, 2010; Stern, 2012). Furthermore, scholars applying PMT have described how fear reaches a plateau, becoming less effective at motivating behaviour change (Westcott et al., 2017). There is also evidence that fear triggers protective motivation in acute emergencies (e.g., evacuations), but is less effective for crises with slower onsets (Babcicky & Seebauer, 2019). Thus, while PMT suggests that coping appraisal is a positive predictor of protective responses (Milne et al., 2000), threat appraisal (and induced fear) may lead to avoidant maladaptation stemming from fatalism,¹ denial, or wishful thinking (Babcicky & Seebauer, 2019; Grothmann & Patt, 2005).

1.5. Madagascar context

Madagascar, the world’s poorest non-conflict country (USAID, 2022), has approximately 80% of its population relying on smallholder agriculture as their primary livelihood (Rakotobe et al., 2016; World Bank, 2018). However, the government provides minimal support,² and much of the island experiences high rates of food insecurity (Harvey et al., 2014). As climate change reduces rice production (Nematchoua et al., 2018), the main staple food of Madagascar, and increases disease and pest risks for crops like cassava (Niang et al., 2014), these rates will worsen. Additionally, with rising sea levels, warming waters, and air temperatures predicted to increase by more than 2.5–3 degrees Celsius in some parts of the island over the next ten years (Nematchoua et al., 2018; Tadross et al., 2008), as well as limited governance capacity to tackle these issues (Weiskopf et al., 2021), Madagascar is highly vulnerable to climate change (Harvey et al., 2014). The country is also extremely prone to tropical cyclones, which are becoming more frequent and intense as a result of climate change (Tadross et al., 2008; Weiskopf et al., 2021). In particular, this study focuses on southeastern Madagascar, characterized by food insecurity and high cyclone risk (Randrianarison et al., 2020).

Within this context of extreme vulnerability, this paper tests the suitability of the Protection Motivation Theory (PMT) as a framework to examine farmers’ intention to adapt agricultural practices in response to climate change. Using a theory-informed latent variable path model, or structural equation model (SEM), the study explores the effects of cognitive processes (threat appraisal and coping appraisal) on farmer intention. No prior studies have attempted to identify how social connectedness, climate change perceptions, and gender link with threat and coping appraisal to predict climate adaptation behaviour decisions from a PMT perspective among a population of highly vulnerable smallholder farmers in Madagascar / the Western Indian Ocean region. While this work focuses on the Madagascar context, the findings have broader implications for island nations and other countries with large numbers of farmers reliant on rainfed agriculture and/or increased vulnerability to cyclones due to climate change (e.g., Puerto Rico, Mozambique).

This research is guided by the following hypotheses (H), with proposed pathways for H1-H6 shown in Figure 3:

H1: Higher climate change perception is significantly associated with higher a) threat and b) coping appraisal.

H2: Greater social connectedness is significantly associated with a) reduced threat and b) higher coping appraisal.

H3: Greater social connectedness is significantly associated with stronger intention to adapt practices in response to observed changes in temperature and/or rainfall.

H4: Past adaptation of farming practices in response to observed changes in temperature and/or rainfall is significantly associated with greater future intention to adapt agricultural practices.

H5: High threat appraisal is significantly associated with reduced intention to adapt practices.

H6: High coping appraisal is significantly associated with greater intention to adapt practices.

H7: Gender moderates the relationship between social connectedness and PMT constructs (threat appraisal and coping appraisal), and therefore, intention to adapt practices.

Figure 3. Model proposal with hypothetical pathways.

2. Methods

2.1. Study site

The sample frame for the study was approximately 750 households in 15 villages surrounding the Manombo Special Reserve protected area on the southeastern coast of Madagascar (Figure 4). Villages were selected from the target population of our partner NGO, Health in Harmony (HIH), considering security concerns due to heightened bandit attacks linked to ongoing drought conditions. Farmers in this region primarily practice small-scale rainfed polyculture, focusing on subsistence farming with rice and cassava as staple crops (Moore et al., 2022). Both men and women have central and ‘complementary’ roles (Dahl, 1999, p. 97) in agricultural production, though roles are often somewhat differentiated. The region faces high levels of poverty and food insecurity (Randrianarison et al., 2020), and is highly cyclone-prone. Limited market access and agricultural extension services, which tend to be male-dominated, further challenge farmers due to the area’s remoteness.

Figure 4. Map of Manombo area [authors].

2.2. Data collection

A cross-sectional study of 328 small-scale rice farmers (64.3% female; 35.7% male) over the age of 18, each representing a separate household, was conducted in February 2021 by a team of five native Malagasy-speaking enumerators. Respondents included both participants in a rice-growing training given by HIH (60% of participants were female), as well as randomly selected non-training participants from separate households in the same villages to reduce self-selection bias. Probability proportional to size (PPS) sampling (Skinner, 2014) was used, and a ‘within-household respondent selection procedure’ was implemented to reduce gender bias. The questionnaire covered topics such as 1) perceived changes in temperature and rainfall in the last five years, 2) perceived threat and coping capacity in response to climate change, and 3) intended and actual changes to agricultural practices, and socio-demographic information. Data was gender disaggregated.³ Paper survey responses were entered into Qualtrics, then analyzed using IBM SPSS Statistics (version 28) and MPlus Diagrammer (version 1.8). Informed consent was obtained from all respondents, and the study received exemption from the University of Vermont’s Institutional Review Board (IRB; study #00001290).

2.3. Structural Equation analysis

Structural Equation Modelling (SEM, Bentler, 1980; Jöreskog, 1978) is a statistical technique that combines factor and regression analysis to allow for more complex path models. It is useful in testing the relationships between the components of PMT (Babcicky & Seebauer, 2019). SEM consists of measurement and structural models, and incorporates both latent and observed variables.

2.3.1. Factor analysis

To identify the measurement model, exploratory and confirmatory factor analyses were conducted to determine the latent constructs, or variables (LVs), for threat and coping appraisal. Supplemental Table 1 shows the results of the exploratory factor analysis⁴ (EFA) conducted using maximum likelihood (ML) estimation in SPSS 28.0. Following the eigenvalues-greater-than-one rule (Kaiser, 1960), two factors were extracted explaining 62.1% of the total variance. Three items reported on a 4-point Likert scale loaded onto a single factor (threat appraisal construct), with standardized factor loadings ranging from 0.40 to 0.90. Two items regarding perceived ability to cope with future cyclones were used as a proxy for perceived ability to cope with changes in temperature and/or rainfall (aka climate change) and reported on a 5-point Likert scale loaded onto another factor (coping appraisal construct), with factor loadings from 0.62 to 0.65. Cyclones are a relevant and appropriate measure of perceived coping ability among this population, as they are highly visible. Factor loadings under 0.4, the generally agreed upon cut-off value (Costello & Osborne, 2005), were suppressed. One item, measured by the statement ‘Farmers like me are likely to be affected by climate change,’ did not load onto either factor. While there is debate on the use of Cronbach’s alpha in the literature (e.g., Tan, 2009), we found alpha coefficients for the two latent constructs to be greater than 0.5, which can be considered acceptable reliability (Taber, 2018).

Following EFA, classical test theory (Jarvis et al., 2003), consisting of confirmatory factor analysis (CFA) and reliability testing, was used to validate the facture structure obtained from EFA. For the confirmatory factor model, we fitted a two-factor logit model using ML estimation with robust standard errors in Mplus Diagrammer version 1.8 (Supplemental Figure 1). All items had satisfactory standardized factor loadings of at least 0.5 and were significant. Given the distribution of the variables, ordinal variables were treated as categorical and model fit indices were given as AIC/BIC (2623.48/2710.72).

To test for reliability of the constructs in the measurement model, composite reliability (CR), considered more appropriate than Cronbach’s alpha for SEM-based studies (Cheung et al., 2023), was calculated. CR was 0.84 (threat appraisal) and 0.69 (coping appraisal). According to Hair (2009), CR > 0.7 indicates good reliability. To test for convergent and discriminant validity (Campbell & Fiske, 1959), the Average Variance Extracted (AVE) and Discriminant values (DV) were determined. AVE for both constructs was above 0.5 (Fornell & Larcker, 1981): 0.64 for threat appraisal and 0.54 for coping appraisal. DV was 0.80 (threat appraisal) and 0.74 (coping appraisal). Both DVs were greater than the correlation (0.34) between the two LVs. Thus, the measurement model was found to be acceptable.

2.3.2. Structural modelling

As the dataset contained missing data unrelated to the response values, full information maximum likelihood (FIML) was used, which treats missing data under the MAR (missing at random) assumption (Cham et al., 2017), to estimate the overall SEM with standardized latent factors. The best-fit model was selected by comparing neighbourhood models and using Akaike information criterion (AIC), used to compare non-nested models with categorical variables (Akaike, 1987). Percentage of explained variance in the outcome variables is presented in Supplemental Table 2.

In addition, multi-group latent class analysis and ML estimation with robust standard errors were used to test the effects of gender as a binary moderator variable on the direct and indirect relationships between the exogenous variables, the two mediating variables (threat and coping appraisal) and the outcome variable.

2.3.3. Operationalization of constructs

A description of variables included in the model, their measurements and category of data are provided in Table 1. Three exogenous variables (climate change recognition, social connectedness, and prior adaptation to observed changes in temperature and/or rainfall) were selected based on the relevant literature and other survey instruments, such as the Climate Change, Agriculture and Food Security (CCAFS) 2010–2012 Household Baseline Survey (CCAFS, 2015) and that used by Rodríguez-Cruz and Niles (2021). Climate change recognition was measured by asking respondents about their observations of changes in temperature and/or rainfall in the last five years, as this timescale has been previously used to measure perceptions in climate change among smallholders in developing countries (e.g., CCAFS). Social capital has been documented as impacting farmer agricultural decision outcomes. To measure, a social connectedness variable was calculated by totalling the number of social connections via group membership and various forms of agriculture-related social interactions that a respondent had (e.g., belonging to a farmer’s cooperative, participating in a farmer training programme, having helped a farmer on their farm in the last year, having been visited by an extension agent or NGO worker, etc.; for a full list see Appendix 1). Frequency of social interaction by type was also calculated by asking respondents about who they spoke to regarding agriculture, and how often. The covariate, prior adaptation, which has been shown to influence farmers’ future adaptation decisions (e.g., Etana et al., 2020), was included in the model as a control variable, and measured by reported implementation of agriculture adaptation practices based on observed changes in temperature and/or rainfall in the last five years, which has also been used frequently (e.g., CCAFS).

Table 1. Variable descriptions and measurements.

Variable	Description/question/statement	Measurement scale	Categorization of data
Exogenous variables
Climate change perception	Observations of changes in rainfall and/or temperature in the last five years	1 = Recognized changes; 0 = Did not recognize changes	Binary
Social connectedness	Total number of social connections/interactions that a respondent reported having (for a full list see Appendix 1)	Count (0–15)	Continuous
Prior adaptation	Previous adaptation to farming practices in response to changes in rainfall and/or temperature in the last five years	1 = Yes; 0 = No	Binary
Indicators of latent mediator variables
Threat appraisal
Perceived severity (risk) of climate change to food security	How much risk, if any, do you feel climate changes pose to your food security?	1 = No risk 2 = Low risk 3 = Medium risk 4 = High risk	Ordinal
Perceived severity (risk) of climate change to income/livelihood	How much risk, if any, do you feel climate changes pose to your income/livelihood?
Worry/fear about changes in climate	How worried are you, if at all, about changes in climate?	1 = Not worried 2 = Slightly worried 3 = Worried 4 = Very worried	Ordinal
Coping appraisal
Motivation/willingness to change	“I feel motivated to change my agricultural practices to prepare for future cyclones.”	1 = Strongly disagree 2 = Disagree 3 = Neutral 4 = Agree 5 = Strongly agree	Ordinal
Perceived adaptive capacity	“I feel that I do have the capacity to change my agricultural practices to prepare for future cyclones.”
Outcome variable
Intention to adapt	Composite variable based on responses to Likert-scale questions investigating three agricultural adaptation behaviours	Integer scale ranging from 1–5, with 14 numerical outcomes	Continuous

The outcome variable in the models was intention to adopt climate change adaptation practices (protective response). A composite variable was created based on responses to Likert-scale questions investigating three specific agricultural adaptation behaviours (i. change timing of planting, ii. change types of crops planted, and iii. change the variety of crops currently cultivated) to the question ‘How likely are you to do any of the following in the next five years in response to changes in climate?’ Cronbach’s α (0.692) was used to test for internal consistency of the scale. Polychoric correlation was used to see the relationship between the variables; all were significantly and positively correlated. The average of the three intended strategies⁵ was then used to create the new variable.

2.3.4. Social connectedness analysis

A nuanced analysis of the relationship between social connectedness and perceived threat to climate change was also conducted. Social connectedness levels (below average, average, above average) were determined based on the mean and one standard deviation. A Pearson Chi-Square test examined the effect of gender on social connectedness levels, and an Independent Samples t-test compared average social connectedness between male and female farmers. The frequency of farmer interactions by social connectedness type (e.g., government and NGO workers, other farmers) was also assessed, and Chi-Square tests examined gender differences.

3. Results

Demographics: Most respondents came from households with very low levels of education and assets. The highest level of education per household was 3.7 years (SD 3.07) on average. The mean number of household assets, on a scale from 0 to 30, was 4.8 assets (SD 2.86). Eighty percent (n = 262) of households were male headed, with an average size of 6.2 people (SD 2.64).

Perceptions of climate change and attributions: As expected among a population of farmers predominantly reliant on rainfed agriculture, respondents were highly aware of climate changes (Figure 5a). Many (68.3%, n = 224) reported that temperatures were getting hotter in the last five years. Similarly, 90.9% (n = 289) of respondents noticed changes in rainfall, with the most common response being that rains come later (38.1%, n = 122). Some respondents (24.4%, n = 78) mentioned rains come earlier, while 19.7% (n = 63) found the timing of rains to be less predictable. One farmer expressed frustration with changing rainfall patterns, stating that ‘[rain] does not come when it is needed, it comes when we do not need it.’

Figure 5. Climate perceptions: a) observation of changes in temperature and rainfall over the last 5 years, b) attribution of changes observed.

A chi-square test of independence was performed to evaluate the relationship between gender and climate change recognition. There was no significant difference in perception of changes in temperature (X2 = 0.28, 1, p = .596) or rainfall (X2 = 0.22, 1, p = .643) between men and women (Table 2).

Table 2. Overall and gender-disaggregated mean statistics and standard deviations for climate change recognition, social connectedness, prior adaptation, threat and coping appraisal variables, and intention (or not) to adapt agricultural practices. Test statistics were obtained using appropriate tests (see below table) depending on distribution of variables.

		Overall (n = 328)		Male (n = 117)		Female (n = 211)
Variable	Description	Mean	SD	Mean	SD	Mean	SD	Test statistic
Climate change perception	Observed changes in temperature	0.90	0.30	0.89	0.32	0.91	0.29	0.28^a (n.s.)
	Observed changes in rainfall	0.92	0.28	0.93	0.26	0.91	0.28	0.22^a (n.s.)
Social connectedness	Total number of social connections/interactions	2.99	1.42	3.38	1.58	2.78	1.27	3.81^b***
Prior adaptation	Made changes in agricultural practices in the last five years.	0.15	0.36	0.19	0.39	.13	.34	1.45^a (n.s.)
Threat appraisal	Worried about changes in climate	3.60	0.79	3.50	0.90	3.66	0.72	2.47^c (n.s.)
	Feel that climate change poses a high risk to food security	3.91	0.40	3.86	0.46	3.94	0.36	4.47^c*
	Feel that climate change poses a high risk to livelihoods	3.85	0.52	3.78	0.65	3.89	0.43	1.69^c (n.s.)
Coping appraisal	Agreement with statement “I feel motivated to change my agricultural practices to prepare for future cyclones.”	4.10	1.0	4.11	1.03	4.10	0.99	0.07^c (n.s.)
	Agreement with statement “I feel that I do have the capacity to change my agricultural practices to prepare for future cyclones.”	2.99	1.33	3.00	1.32	3.00	1.34	0.00^c (n.s.)
Intention to adapt	Likelihood of making changes to agricultural practices in the next five years	4.16	0.86	4.03	0.91	4.22	0.82	−1.92^b (n.s.)

Note: * = p < .05; *** = p < .001; n.s. = not significant

Statistical test used: ^a. Chi-square test, ^b. Independent samples t-test, c. Kruskal-Wallis H test

However, while awareness of changing climatic conditions was high, the attribution of these changes to anthropogenic activities was low (Figure 5b). Nearly half of respondents (46.5%; n = 152) did not know the causes of the changes. Only 15.7% (n = 52) attributed them to human activities, while 11.5% (n = 38) attributed changes to natural causes. Interestingly, 10.3% (n = 34) believed that changes were caused by zanahary (God), a belief also documented among Indigenous farmers in Bolivia (Boillat & Berkes, 2013). Similar supernatural attributions have been observed among the Vezo in southwestern Madagascar, where local fisherfolk sometimes blame adverse weather on the improper burial of a deceased mermaid (Muttenzer, 2020). In addition, of those that responded: ‘Other’ (16%; n = 54), open-ended responses attributed climate change to actions of foreigners and scientists (18.5%; n = 10) and the government (7.4%, n = 4), as well as to forest loss (11.1%; n = 6).

Threat appraisal (Fear and perceived risk from climate change): Most farmers (87.5%; n = 287) expressed worry about perceived changes in climate, with 68% (n = 223) strongly agreeing with the statement, ‘Farmers like me are likely to be negatively affected by climate change.’ Respondents also reported high levels of perceived threat, feeling that climate change poses high risk to both food security (93.9%, n = 307) and livelihoods (90.2%, n = 295). Gender had no significant effect on worry about climate change (X2 = 2.47, 1, p = .116) or perceived risk of climate change to livelihoods (X2 = 1.69, 1, p = .194) based on results of Kruskal–Wallis tests. However, women’s perceived risk to food security was significantly higher than men’s (X2 = 4.47, 1, p = .037) (Table 2), possibly due to women’s traditional roles in food provisioning and preparation at home.

Coping appraisal (Self-efficacy): Around three-quarters (76%; n = 248) of respondents expressed motivation to adapt their agricultural practices in order to mitigate damage from future cyclones. However, only 38.7% (n = 127) felt they had the capacity to make these changes. There were no significant gender differences in willingness to change practices (X2 = 1.09, 4, p = .895) or perceived adaptive capacity (X2 = 0.17, 4, p = .997) (Table 2).

Actual and intended adaptation measures in response to climate change: Despite farmers’ awareness and concern about climate change, few had made changes to their agricultural practices in the last five years (Table 2). However, a majority expressed intention to make changes in the next five years: 80.5% (n = 264) planned to change the timing of planting, while 87.2% (n = 286) and 87.5% (n = 287) intended to change crop varieties and types, respectively. Notably, there were no significant gender differences in prior adoption or intended adoption of adaptation measures (Table 2).

Social connectedness: In general, results showed farmers to have low social connectedness with an average score of 2.99 (SD = 1.42) on a scale of 0 to 15 possible social interactions. Male farmers had significantly higher social connectedness, scoring an average of 3.38 (SD = 1.58), while female farmers scored an average of 2.78 (SD = 1.27), [t (324) = 3.81, p < 0.001] (Table 2). Female farmers were also significantly more likely to have below-average connectedness compared to male farmers, X2= 12.83, 2, p = .002.

In terms of types of social interactions, farmers reported very infrequent to no interaction with government workers and Madagascar National Parks (MNP) staff, and only slightly more frequent interactions with NGO workers, while interactions between farmers were more common (Figure 6a). Nearly 30% of farmers reported interacting with other farmers on at least a monthly basis, unsurprising given that farming is the predominant livelihood in these communities. In addition, over half of respondents participated in the recent rice-growing training (60.7%, n = 199) and 15.5% (n = 51) belonged to a farmer’s cooperative. Additionally, when farmers were asked who they typically consulted with about farming-related decisions, only 4.6% (n = 7) cited consulting with agricultural extension/NGO workers, while 57.1% (n = 88) had consulted elders in the community and 6.5% (n = 10) had sought out the advice of astrologers on propitious planting times, etc.

Figure 6. a) Frequency of farmer interactions with others, b) Frequency of interaction types by sex based on 64.3% female and 35.7% male respondents.

Note: * = p <.05; ** = p <.01.

When examining the frequency of farmer interactions by gender, male and female farmer social connectedness is clearly different (Figure 6b). For example, men were significantly more likely to have vertical ties with government (p < .001), MNP (p < .001), and NGO staff (p < .001), while women reported significantly less frequent interactions with other farmers (p < .05), government (p < .05) and MNP workers (p < .001). Women were also significantly less likely to consult with elders in the community about agricultural decisions than men, X2 = 4.57, 1, p = .032. However, there were no statistically significant differences between men and women in terms of their membership in a farmer’s cooperative (X2 = 0.92, 1, p = .337) or their participation in the HIH agricultural training (X2 = 0.13, 1, p = .723). Thus, group membership and trainings emerge as important ways that both male and female farmers can engage with other farmers about agricultural decisions and to share information.

3.1. Social connectedness reduces perceived threat of climate change

Figure 7 shows how various levels of farmer social connectedness (below average, average, above average) impact farmers’ appraisal of climate change threat. Having average and above average social connectedness was associated with reduced perceived threat. Greater social connectedness was also associated with reduced fear (worry) that farmers had about the threat of climate change. We also identified significant relationships between social connectedness level and perceived risk of climate change to income (livelihoods) (X2= 18.57, 6, p = .005). However, social connectedness and perceived risk of climate change to food security (X2 = 8.52, 6, p = .202) or worry about climate change (X2= 10.84, 6, p = .093) did not have significant associations.

Figure 7. Perceived risk of climate change to a) income (livelihoods) and food security, and b) degree of worry about climate change, by level of social connectedness (below average, average, above average).

3.2. Structural Equation Modeling (SEM) results

In exploring the two core constructs of the PMT (threat and coping appraisal), we find multiple predictors and relationships to intention to adopt practices. A diagram of the path analysis and measurement model from the overall and multi-group SEMs are shown in Figure 8, and unstandardized model estimates, standard errors, and significance levels are presented in Table 3.

Figure 8. Path diagrams for a) overall and b), c) multi-group models (men and women) with standardized estimates. Circles indicate latent variables; squares indicate observed variables. Solid lines represent statistically significant relationships (p < .05). Dotted lines indicate non-significant pathways

Table 3. Hypothesis testing results from SEM for overall sample and with multi-group (men vs. women) moderation effects. Effect sizes in table are unstandardized as recommended by Ockey and Choi (2015). Comparative model fit statistics (AIC/BIC) for each model below table.

		Overall			Group 1 Men (n = 117)			Group 2 Women (n = 211)
Hypothesis	Relationship	Estimate	SE	p=	Estimate	SE	p=	Estimate	SE	p=
H1	Climate change recognition →
	a) Threat	1.155 (a)	0.33	.000***	1.124	0.67	.094	1.283	0.34	.000***
	b) Coping	1.143 (b)	0.43	.007**	1.398	0.50	.005*	1.081	0.40	.006**
H2	Social connectedness →
	a) Threat	−0.190 (a)	0.07	0.07	−0.190	0.09	.028*	−0.127	0.10	.222
	b) Coping	−0.007 (b)	0.07	.883	−0.065	0.06	.311	0.043	0.07	.544
H3	Social connectedness → Intention to adapt	0.035	0.04	.339	0.161	0.07	.017*	−0.044	0.04	.304
H4	Prior adaptation → Intention to adapt	0.255	0.12	.030*	0.353	0.20	.074	0.225	0.14	.105
Protection Motivation Theory
H5	Threat → Intention to adapt	0.042	0.08	.619	0.098	0.22	.659	0.006	0.07	.936
H6	Coping → Intention to adapt	0.265	0.09	.002**	0.121	0.18	.504	0.329	0.09	.000***

Note: * = p < .05; ** = p < .01; *** = p < .001

Overall model AIC/BIC: 3083/3205 Multi-group model AIC/BIC: 3470/3629

In support of H1, we find that climate change recognition is a significant, positive predictor of both coping (b = 0.170, p = .006) and threat appraisal (b = 0.167, p = .000). However, while social connectedness does not significantly predict coping appraisal (H2b), it is a significant, negative predictor of threat appraisal (b = -0.258, p = .002) (H2a). Therefore, we find that being more socially connected is critical to lowering perceived threat and, in turn, reducing maladaptive response pathways.

Our results also support our fourth hypothesis (H4). Previous adoption of adaptation measures is a significant, positive predictor of intention to adapt (b = 0.110, p = .030). As predicted, those that have adopted agricultural adaptation practices in the past are more likely to intend to do so in the future.

In terms of PMT’s validity, coping appraisal was found to be a stronger predictor of intention to adapt than threat appraisal. Coping appraisal is a significant, positive predictor of the ‘Intention to adapt’ outcome variable (b = 0.322, p = .001) (H6), while threat appraisal is not (b = 0.052, p = .621) (H5). This finding is supported by the work of Babcicky and Seebauer (2019), who found coping appraisal to predict protective behaviour, while threat appraisal displayed ‘a non-protective route to non-protective responses.’

The results of the multi-group SEM with gender as a moderator partially support H7. Coping appraisal was found to be a significant predictor of intention to adapt agricultural practices among women (b = 0.403, p < .001), but not men (b = 0.149, p = .504). We also find that that higher social connectedness is significantly associated with reduced threat appraisal in men (b = -0.284, p = .017), with no significant effect for women (b = -0.158, p = .213). Social connectedness is also a significant and positive predictor of intention to adapt agricultural practices among men (b = 0.305, p = .010) (and not women; b = -0.069, p = .312). Furthermore, social connectedness and intent to adapt were not found to be significantly correlated for the whole sample (r = .068, p = .220) or for women (r = -.014, p = .844), but it was for the men-only sample (r = .227, p = .014).

4. Discussion

4.1. Climate change perception

The study supports our hypothesis (H1) that climate change awareness is an important precursor for intention to adapt agricultural practices – results of the SEM show climate change recognition positively predicts both threat and coping appraisal constructs. In addition, despite low social connectedness on average, smallholder farmers in southeastern Madagascar are ‘climate-informed,’⁶ aware of regional climatic changes such as increased temperature and more unpredictable rainfall. These findings align with other studies on farmers’ perceptions of climate change (e.g., Fisher et al., 2015; Karki et al., 2020a; Nguyen et al., 2021), as well as reports by climate scientists (Tadross et al., 2008). However, it is noteworthy that few respondents attribute these changes to anthropogenic activity, which is consistent with findings from other Indigenous farming communities in Brazil (Funatsu et al., 2019) and Peru (Altea, 2020), for example.

4.2. Social connectedness

This study also emphasizes the role of social embeddedness (Granovetter, 1985), or the extent to which one’s behaviour is shaped by social relations, in smallholder farmer decision-making. Specifically, being more socially connected is a significant predictor of reduced threat appraisal (H2a) (i.e. farmers with greater social connectedness perceive climate change as less threatening). Thus, socially connected farmers feel better supported and less inhibited to take action. Similar findings have been reported in studies on agricultural technology adoption in Malawi (Kansanga et al., 2020) and China (Zheng et al., 2022). Yet, Manombo area farmers exhibited low social connectedness on average and expressed high levels of concern about climate change. Consequently, in highly vulnerable contexts, low social connectedness and high threat appraisal may hinder climate change adaptation behaviour.

Furthermore, because farmers report more frequent interaction with other farmers/neighbors, and consulting with village elders rather than agricultural extension/NGO workers, it is important to consider the horizontal and vertical structure of Malagasy society as it relates to the transfer of information and knowledge spillover. For example, as Dahl (1999) points out, the assumption that pilot farmers will be emulated by neighbouring farmers stems from Western norms which may not be reciprocated in this context. During fieldwork in Madagascar, Dahl observed that successful farmers were often met with suspicion, jealousy, or general disapproval from the community. However, those farmers willing to actively share their ‘know-how,’ Dahl writes, were able to restore fihavanana [horizontal solidarity], the social bedrock of Malagasy communities, and be considered as ray aman-dreny [parents] (p. 96). Thus, as only those willing to share their knowledge are transformed into respected elders and trusted to provide agricultural advice, interventions must carefully decipher whom to enlist as knowledge holders.

Moreover, our research provides further evidence that female farmers have lower social connections compared to their male counterparts. Additionally, our findings demonstrate that female farmers have less connectivity with perceived authority figures, such as government and parks officials, than compared to male farmers. However, group membership (e.g., farmer cooperatives, women’s associations) serves as a crucial source of connection for female farmers. Based on these findings, we recommend that climate change adaptation policies and interventions specifically target women and other vulnerable groups that have traditionally been less socially connected.

Furthermore, while it is important to follow cultural protocols related to communication with village elders and local authorities, who are often men, it should not be assumed that information will effectively reach women through these channels. Instead, messaging should be directed through channels where women are more likely to communicate, such as women’s associations, and delivered at times and locations most suitable for them.

Following gender-responsive agricultural development best practices, such as recruiting and training more female extension workers (Witinok-Huber et al., 2021), ensuring equitable access to training programmes, tailoring advice to crops that women tend to focus on, and providing specific training for women on new tools will also help close the gender gap in agriculture and address the climate change adaptation deficit. Successful examples include the BRAC programme in Uganda, which increased technology adoption among female farmers by establishing a network of female Model Farmers and community-based agricultural agents (Pan et al., 2018). In Madagascar, a Conservation International-led project on climate change adaptation among smallholders produced gender-sensitive training modules and an information exchange platform on climate risk reduction options as part of their Gender Action Plan (Conservation International, 2022). In Mali, training women service providers on the use of the RiceAdvice app led to adoption of new rice-growing technologies by over 20,000 women and youth (AfricaRice, 2020).

4.3. Coping appraisal is a stronger predictor than threat appraisal

This study supports the suitability of PMT in predicting smallholder intention to adapt agricultural practices to climate change. Consistent with previous research (Burnham & Ma, 2017; van Valkengoed & Steg, 2019), our results show that coping appraisal, or perceived adaptive capacity, is an important determinant of adaptation intent. Specifically, we find PMT’s ‘Coping appraisal’ construct to be a stronger predictor of ‘Intention to adapt’ than the ‘Threat appraisal’ construct. Meta-analytic reviews of PMT among health-related studies (Floyd et al., 2000; Milne et al., 2000), as well as studies applying PMT to smallholder decision-making in the Global South (e.g., Truelove et al., 2015), have also found the coping appraisal construct to have greater predictive validity than the threat appraisal construct.

However, unlike studies linking higher concern to greater adaptation likelihood (e.g., Woods et al., 2017), our study suggests that threat appraisal alone does not lead to adaptation intentions. This finding is supported by recent research applying PMT to flood mitigation behaviour (Babcicky & Seebauer, 2019). While some studies highlight risk perception as a motivator for intention and adaptation behaviour among farmers (e.g., Azadi et al., 2019; Feng et al., 2017), other studies, particularly among resource-limited farmers and other vulnerable populations, find that concern does not always translate into behaviour change (Etana et al., 2020; Rodríguez-Cruz & Niles, 2021; Tucker et al., 2010).

Rather, as high threat appraisal can lead to maladaptive behaviours such as fatalism, risk-aversion, and denial, as well as a status quo bias, farmers may be ‘paralyzed’ by fear, thereby hindering the protection motivation that fear is assumed to evoke (Plotnikoff & Trinh, 2010). Indeed, fear of change and vulnerability has been found to inhibit farmers’ adaptation in other Global South contexts (Bagagnan et al., 2019; Luu et al., 2019), and research emphasizes the importance of promoting hope, not fear, in achieving desired outcomes related to climate change policy (Nabi et al., 2018).

4.4. Intention setting does not necessarily lead to adaptive behaviour

Despite research pointing to the importance of ‘intention strength’ in behavioural change (Conner & Norman, 2022), we find that farmers had strong intentions but low rates of prior adaptation. This intention-behaviour gap, or failure to translate intentions into action, is well-documented in various decision-making studies, from organic food purchases (Frank & Brock, 2018) to physical activity goal-setting (Rhodes & de Bruijn, 2013), and more recently among farmers (e.g., Niles et al., 2016; Rodríguez-Cruz et al., 2021). For example, farmers in resource-poor settings, such as Madagascar, may have the intention to change their practices, but they often lack the capacity to do so, as has been demonstrated among farmers in other parts of sub-Saharan Africa (e.g., Bryan et al., 2009; Deressa et al., 2008; Fisher et al., 2015).

5. Limitations

The strengths of this study are its large sample size relative to the population and rigorous sampling design, though we also acknowledge several limitations. Some of the multi-item scales used in this analysis have fairly low reliability. Additionally, several psychometric variables are measured by single items. Our survey also omitted questions related to response efficacy, a component of PMT’s coping appraisal construct found to be a strong predictor of adaptation behaviour (van Valkengoed & Steg, 2019).

Furthermore, while our results support our hypothesis (H4) that past adaptation of farming practices influences future intention to adapt, we were unable to examine the predictive power of intention on actual adaptation due to concurrent measurement. Future research should use longitudinal data to understand the predictive value of intention on adoption (and disadoption) of adaptation measures over time, as well as to study the adaptation learning process (Lamichhane et al., 2022).

In terms of limitations of the PMT framework, it does not consider variables such as social norms (Atta-Aidoo et al., 2022; van Valkengoed & Steg, 2019) or cultural dimensions (Adger et al., 2013) that influence smallholder farmer adoption of climate-resilient practices. Future expansions could integrate other behavioural theories that include normative beliefs, sociocultural perspectives and relationships (e.g., trust, cooperation), and individual personality variables, such as negative affectivity (van Valkengoed & Steg, 2019). For example, optimistic farmers in India were more likely to exhibit adaptation behaviour than those who were pessimistic and had fatalistic outlooks (Singh et al., 2016).

6. Conclusion

Using a SEM approach to test PMT for predicting intention to adopt climate change adaptation practices among smallholder farmers, this study highlights the importance of considering the psychological aspects that lead to behavioural adaptation, specifically coping appraisal. However, while our study significantly explains adoption intention, we posit that high threat appraisal and low social connectedness, particularly among women smallholder farmers, may lead to non-protective behaviour or avoidant maladaptation (e.g., risk aversion, fatalism, and wishful thinking) rather than desired agricultural adaptation outcomes, despite high intentions to adapt.

Given high climate change perception among smallholders (Soubry et al., 2020), and as threat alone appears to be non-motivational (Tunner et al., 1989), solely focusing on increasing awareness through fear-based approaches may not effectively drive climate change adaptation. Instead, interventions should prioritize enhancing adaptive capacity and addressing context-specific risks and uncertainties. Considering that social connectedness plays a critical role in reducing threat appraisal and recognizing that women are often less socially connected, future research on climate change adaptation should explore gender-specific access to social networks across contexts (Carranza & Niles, 2019; Macgregor, 2010; Ravera et al., 2016). Lastly, efforts should be directed towards strengthening existing social safety nets and communication channels, such as farmers’ and women’s associations, to better equip less socially connected and historically marginalized groups in coping with the precarity of climate change.

Acknowledgments

We are extremely grateful to the local communities, the survey team led by Kimmerling Razafindrina, as well as support from Health in Harmony, in Madagascar. We also wish to thank Kame Westerman and Luis Rodríguez-Cruz for their insightful suggestions which helped to strengthen the final manuscript, and Tim Trueur for his pivotal role in securing funding.

Disclosure statement

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

Additional information

Funding

This study was funded by a catalyst grant from the Gund Institute for the Environment at the University of Vermont, as well as through the Bridge Sparks Fund.

Notes on contributors

Maya Moore

Dr. Maya Moore is currently a postdoctoral fellow at Columbia University's Climate School, receiving her PhD in Food Systems from the University of Vermont in Fall 2023. Her current research focuses on the gender dimensions of access, use and impact of climate information services.

Meredith T. Niles

Dr. Meredith T. Niles is a professor in the Department of Nutrition and Food Sciences and the Food Systems Program at the University of Vermont. Her research primarily focuses on the impact of climate change, disasters, and other crises like pandemics on food security and health outcomes, as well as the drivers and barriers for farmers to adopt more sustainable management practices for climate change, water, and health outcomes.

Notes

1 Fatalism, or the belief that it is futile to attempt changing things that are predetermined (a ‘why bother?’ attitude), effectively the opposite of self-efficacy, is especially prevalent in the growing body of research on adaptation to climate change (e.g. Etana et al., 2020; Feng et al., 2017). However, fatalistic outlooks do not necessarily precure farmers from attempting to improve their future through actions such as prayer and other ritualistic methods (Roncoli et al., 2002).

2 Madagascar spends far less than most other countries on health and education. In 2014, less than 3% of Madagascar’s GDP was spent on education; about 4-5% on Total Health Expenditure (UNICEF, 2014).

3 Respondents were marked as either lelahy (male; man) or vehivavy (female; woman). There is no distinction in the Malagasy language between sex and gender.

4 A Kaiser-Mayer-Olkin test (KMO’s test; Kaiser, 1974; KMO = 0.56) and Bartlett's (1954) test of sphericity (p < .001) justified the application of EFA. Chi-square approximate = 163.77; DF = 10; p < .001

5 The strategies i) to leave farming and ii) migrate elsewhere to find work were excluded from the analysis as they are not actual farming practices and were found to not be viable strategies for this population based on low level responses.

6 According to the definition given by Schattman et al. (2021), ‘climate-informed’ farmers are those who ‘possess knowledge of climate change and related impacts’ (p. 766).

Appendix 1

Table

Item	Question/statement	Measurement scale
Agricultural training participant	Did you participate in the SRI rice-growing training provided by Health in Harmony (HIH)?	1= Yes 0= No
Group membership_Farmer association/cooperative	Do you belong to any of the following organizations: Farmer association/cooperative?	1= Yes 0= No
Group membership_ Forest Protection	Do you belong to any of the following organizations: Forest Protection?	1= Yes 0= No
Group membership _Tree nursery/tree planting	Do you belong to any of the following organizations: Tree nursery/tree planting?	1= Yes 0= No
Group membership_ Fishing	Do you belong to any of the following organizations: Fishing?	1= Yes 0= No
Group membership_ Women’s group (handicrafts)	Do you belong to any of the following organizations: Women’s group (handicrafts)?	1= Yes 0= No
Group membership_ Women’s group (other activity)	Do you belong to any of the following organizations: Women’s group (other activity)?	1= Yes 0= No
Group membership _Village Savings & Loan	Do you belong to any of the following organizations: Village Savings & Loan?	1= Yes 0= No
Group membership _ Water catchment management	Do you belong to any of the following organizations: Water catchment management?	1= Yes 0= No
Group membership _ Irrigation committee	Do you belong to any of the following organizations: Irrigation committee?	1= Yes 0= No
Group membership_ FRAM (parent-teacher association)	Do you belong to any of the following organizations: FRAM?	1= Yes 0= No
Group membership_Youth association	Do you belong to any of the following organizations: Youth association?	1= Yes 0= No
Help other farmers	Do you do any of the following with other farmers: Help other farmers (not from your household) with their crop production labour – unpaid?	1= Did last year 0= Did not do in last year
Receive help from other farmers	Have other farmers ever helped you with your crop production labour – unpaid?
Extension	Have you consulted with agricultural extension agents about your agricultural practices?	1= Yes 0= No