Climate change adaptation: factors influencing Chinese smallholder farmers’ perceived self-efficacy and adaptation intent
Understanding how individuals perceive their ability to adapt to climate change is critical to understanding adaptation decision-making. Drawing on a survey of 483 smallholder farmer households in the Loess Plateau region of China, we examine the factors that shape smallholder farmer perceptions of their ability to adapt to climate change and their stated intent to do so. We apply a proportional odds ordered logistic regression model to identify the role that determinants of adaptive capacity play in shaping smallholders’ perceived self-efficacy and adaptation intent. Our study provides further evidence that self-efficacy beliefs are a strong, positive predictor of adaptation intent. Our study suggests that human capital, information and technology, material resources and infrastructure, wealth and financial capital, and institutions and entitlements all play an important role in shaping smallholder perceived self-efficacy, while state-society dependencies may reduce smallholder perceived self-efficacy. In addition, our study suggests that perceiving climate change risks and impacts do not necessarily lead to an intention to adapt. Overall, our findings highlight the importance of incorporating both the objective determinants of smallholders’ adaptive capacity and their subjective perceptions of these objective determinants into future climate change adaptation programs and policies in order to facilitate adaptive actions. Identifying factors that cause individuals to have a low estimation of their adaptive ability may allow planned adaptation interventions to address these perceived limitations and encourage adaptive behavior.
KeywordsAdaptive capacity Climate change Climate perception China Smallholder farmers
Smallholder farmers in the Global South are among the most vulnerable groups to climate change (Morton 2007). “Double exposure” to global environmental change and globalization (Leichenko and O’Brien 2008), and low levels of adaptive capacity caused by on-the-ground social conditions, such as inequality, poverty, and poor planning, (Ribot 2009) will likely exacerbate many of the difficulties smallholders already face. As such, researchers have increasingly investigated the role that adaptation can play in mitigating the impacts of climate change on smallholder lives and livelihoods. Recent scholarly calls to shift vulnerability research toward the determinants of adaptive capacity have highlighted that many of the barriers to successful adaptation are the outcomes of social processes (Jones and Boyd 2011). This has led to much recent research theorizing and empirical examination of the social determinants of adaptive capacity at various levels, from individuals to households to countries.
In order to provide conceptual clarity, we first discuss several key definitions. Multiple competing definitions of the term adaptation exist (Burnham and Ma 2015). For the purposes of this study, we define adaptation to climate change as actions in a given social–ecological system taken by an individual, community, or institution “in response to actual and expected impacts of climate change in the context of interacting non-climatic changes.” (Moser and Ekstrom 2010). We define adaptive capacity as the ability of an individual, community, or institution to prepare for and/or adapt to stresses ex ante or react to stresses ex post (Smit and Pilifosova 2001; Brooks and Adger 2005; Engle 2011). Adaptive capacity is linked to the concept of vulnerability (Adger and Vincent 2005). The vulnerability of a system is the product of both its exposure and sensitivity to biophysical and social stressors, perturbations, and processes, and in part arises from its capacity to adapt (Turner et al. 2003). Researchers have argued that understanding the constraints to the adaptive capacity of an individual, community, or institution can help remove barriers that prevent them from engaging in autonomous adaptation, facilitate the development of planned adaptation projects, and reduce the vulnerability of the individual, community or institution (Adger 2003; Smit and Pilifosova 2001). Following Klein et al. (2014), we define an adaptation constraint as “a factor or process that makes adaptation planning and implementation more difficult” (p. 906). Adaptation constraints are synonymous with terms “barrier” and “obstacle” that are frequently used in the adaptation literature (Klein et al. 2014).
Constraints to adaptive capacity can be both objective and subjective (Grothmann and Patt 2005). The objective determinants of adaptive capacity include factors such as financial capital and accesses to technology, which have been well established and have received much attention in the literature (Burnham and Ma 2015; Harmer and Rahman 2014). The subjective determinants of adaptive capacity are mostly related to how individuals and communities perceive the process of adaptation, which has been shown to influence adaptation decision-making (Wolf et al. 2013), and is receiving increasing attention within the scholarly literature (e.g., Blennow and Persson 2009; Frank et al. 2011; Grothmann and Patt 2005; Kuruppu and Liverman 2011; Patt and Schröter 2008). In particular, the work of Grothmann and Patt (2005) highlighted that individuals with a low estimation of their adaptive capacity may be more vulnerable to climate change than those with a high estimation of their adaptive capacity because the low estimation decreases the likelihood that they will engage in adaptive behavior, thereby increasing the likelihood that climate change will negatively impact their livelihood.
In this study, we continue the work of examining individual climate change adaptation decision-making with a focus on the role that identified determinants of adaptive capacity play in shaping Chinese smallholder farmers’ perceptions of their ability to adapt to climate change, as well as their stated intent to do so. We begin by briefly reviewing the literature on the determinants of adaptive capacity, paying particular attention to the role they play in influencing the subjective dimensions of adaptive capacity. We then provide a contextual overview of our study site, the Loess Plateau region of China. This is followed by a description of the methods we used to collect and analyze data, the results of our study, and a discussion of their significance.
Theoretical background and methodology
There is broad agreement within the literature on factors that determine adaptive capacity, though no definitive typology exists (Barnett et al. 2013). Smit and Pilifosova (2001) argued that for communities, regions, or countries, the “determinants of adaptive capacity relate to the economic, social, institutional, and technological conditions that facilitate or constrain the development and deployment of adaptive measures.” Adger et al. (2007) listed five general categories of barriers to adaptation: financial, technological, cognitive, cultural, and institutional. Communication and information (Moser and Ekstrom 2010); values, beliefs, and norms (Jones and Boyd 2011; Moser and Ekstrom 2010); and physical and ecological factors (Jones and Boyd 2011) have also been identified as factors influencing adaptive capacity. Further, researchers have shown that social capital can contribute to and constrain adaptive capacity (Adger 2003; Pelling and High 2005). Specifically, Adger (2003) argued that adaptive capacity is the product of access to resources, how those resources are distributed within and between groups, and the institutions that govern the resources. Identified social barriers to adaptation include knowledge, emotions, and cultural factors such as place attachment and identity (Adger et al. 2013; Barnett et al. 2013). At the smallholder household level, recent representative research has empirically demonstrated that the ability of smallholder households to adapt is determined by factors such as access to crop insurance (Panda et al. 2013), the availability of credit (Bryan et al. 2013; Hisali et al. 2011; Mertz et al. 2010; Tambo and Abdoulaye 2013), local government and market-based institutions (Wang et al. 2013), property ownership (Below et al. 2012), and access to technical information about agricultural management and climate change through agricultural extension services (Bryan et al. 2013; Deressa et al. 2010; Young et al. 2009). Commonly reported constraints to adaptive capacity include lack of land (Barbier et al. 2009; Piya et al. 2012; Tucker et al. 2010) and lack of human capital (Young et al. 2009).
Grothmann and Patt (2005) have argued that previous research has not fully taken into account the cognitive factors that impede individual adaptive capacity. Specifically, while it is generally acknowledged that perceived adaptive capacity is an important component of an individual’s decision to adapt to climate change, little research has been done to examine the factors that play a role in shaping it (Grothmann and Patt 2005; Kuruppu and Liverman 2011). To address this gap, Grothmann and Patt (2005) developed a model of private proactive adaptation to climate change (MPPACC), and posited that subjective determinants of adaptive capacity related to an individual’s cognitive processes are at least as important as objective determinants of adaptive capacity in determining a person’s ability to adapt.
The model identified two “bottlenecks” in an individual’s adaptation decision-making process. The first, “risk appraisal,” consists of two components: (1) a person’s determination of the probability that they will be “exposed to the threat” and (2) a person’s determination of how much harm the threat will do to the things they value. The second bottleneck, termed “adaptation appraisal,” is an individual’s assessment of the positive and negative consequences that would result from taking an action, as well as their ability to perform the action. Adaptation appraisal only occurs if the individual’s appraisal of the risk posed by climate change exceeds a minimum threshold. The adaptation appraisal process has three components through which a person determines: (1) whether an adaptive action will succeed in protecting them from the threat (i.e., “perceived adaptive efficacy”); (2) whether they have the ability to carry out the adaptation (i.e., “perceived self-efficacy”); and, (3) the costs of taking the action (i.e., “perceived adaptation costs”). According to Grothmann and Patt (2005), perceived self-efficacy, in part, determines a person’s perceived adaptive capacity.
Drawing on the MPPACC, Blennow and Persson (2009) found a significant association between Swedish forest landowners who had not adapted to climate change and those who lacked an understanding of how to adapt or did not believe in the efficacy of particular adaptive strategies. Frank et al. (2011) investigated the role social identity plays in shaping perceived self-efficacy among coffee farmers in Chiapas, Mexico. They found that how an individual viewed herself in terms of social group membership influenced her perceptions of the risk posed by climate change and her adaptive capacity. Kuruppu and Liverman (2011) examined the role of perceived self-efficacy in determining the formation of intention to adapt to climate change-induced water stress in the central Pacific islands of Kiribati. They found that high levels of perceived self-efficacy were an important driver of adaptation intent and that a person’s belief in their own self-efficacy may depend more on past experience with water stress than a detailed understanding of climate impacts. Jones and Boyd (2011) found that discrimination against Dalit and Humli populations in Nepal and India restricted the availability and type of job opportunities available to them, resulting in low perceived self-efficacy and limiting their ability to cope with climate stress. Finally, Lo (2013) demonstrated that perceptions of social norms played a mediating role between risk perception and the adaptive action of purchasing flood insurance in Australia, and better explained flood insurance purchasing behavior than did risk perception alone.
Determinants of adaptive capacity, adapted from Eakin and Lemos (2006)
Categories of determinants
Specific factors/variables within each category
Knowledge (scientific, “local,” technical, political), education level, health, individual risk perception, labor
Material and non-material resources (i.e., information, technology, material resources, and infrastructure)
Communication networks, technology transfer and data exchange, innovation capacity, early warning systems, technological relevance, transport, water infrastructure, buildings, sanitation, environmental quality
Wealth and financial capital
Income and wealth distribution, marginalization, accessibility and availability of financial instruments (insurance, credit), fiscal incentives for risk management
Institutions and entitlements
Informal and formal rules for resource conservation, risk management, regional planning, participation, property rights, and risk sharing mechanisms
We next apply our typology to determine which factors influence the adaptation intent of smallholder farmers. Previous research on the relationship between attitudes and behavior positions behavioral intention as playing an important role in predicting actual behavior (Ajzen 1991; Fishbein and Ajzen 1975; Triandis 1977). However, as Grothmann and Patt (2005) noted, intention does not always lead to realized behavior. Recent research on attitude–behavior relations has shown that intentions can only be expressed as behavior when a person is able to control the behavior (Webb and Sheeran 2006). Thus, not all stated adaptation intentions result in actual adaptive actions, and one of the barriers that prevents realization of a linear relationship between the two is a lack of adaptive capacity, including factors such as access to resources, technical skill, and social support as these affect a person’s ability to control their behavior (Grothmann and Patt 2005; Liska 1984). While a gap between intended and actual behavior exists, intentions are considered a good predictor of future behavior in unstable or changing conditions such as those brought about by climate change (Kuruppu and Liverman 2011; Webb and Sheeran 2006). As Kuruppu and Liverman (2011) noted, including adaptation intention in the MPPACC provides insight into people’s commitment to adapt to climate change. A stated intent demonstrates how much effort a person is willing to exert in order to achieve their desired outcome (Ajzen 1991; Webb and Sheeran 2005), thus indicating how motivated they are to adapt (Ajzen 1991). And motivation has been shown to be a critical determinant of adaptive action (Frank et al. 2011). Thus, we examine the role of determinants of adaptive capacity in shaping smallholders’ adaptation intent in order to provide insight into the cognitive processes that motivate or impede adaptive action and how those processes are shaped by the external world. This task is important as planned adaptations will have to take into account the cognitive barriers to adaptive action in order to succeed (Patt and Schröter 2008).
Study site and methods
Climate and adaptation on the Loess Plateau
The Loess Plateau region of China sits at about 37° north latitude in the middle and upper reaches of the Yellow River, and covers an area of approximately 250,000 square miles across parts of five provinces. It is situated just northwest of the East Asian monsoon (EAM) zone. The climate is a typical continental monsoonal climate, and the advance of the EAM each year causes the region to have distinctive wet and dry seasons (Ding and Chan 2005). The climate in the region ranges from arid/semiarid to sub-humid, and averages about 140 mm of rainfall in the northwest to 800 mm in the southeast. Approximately 70 % of annual precipitation falls between June and September in the form of heavy storms, leaving many crops vulnerable to drought in the early growing season (Li et al. 2012). Between 1970 and 2010, mean annual temperature increased by 0.06 °C per year and annual precipitation decreased by 0.51 mm per year (Zhang et al. 2012). Drought frequency and intensity also increased across much of the region (Zhang et al. 2013).
The Loess Plateau has been identified as one of the most agriculturally vulnerable regions to climate change in China (Wu et al. 2013). Researchers have predicted increases in average annual temperature, drought frequency, and soil erosion, as well as higher levels of precipitation in the winter when it is less useful for agriculture and lower levels of precipitation in the summer when it is crucial. Future climate scenarios project that by the 2080s, annual mean temperature in the Loess Plateau region will increase by as much as 5 °C. Although average annual precipitation is expected to increase by 54 mm to 150 mm (Liu et al. 2011), the evapotranspiration rate is expected to increase by 12 % over its 1961–2009 average (Li et al. 2012). Additionally, the intensity of precipitation events are likely to increase (Li et al. 2010), while runoff from the Yellow River is expected to decrease, leading to water shortages that will be exacerbated by a growing population (Wang and Zhang 2011; Li et al. 2010; Piao et al. 2010). In certain areas, the duration of seasonal snow cover will likely be shortened, with snow packs thawing in advance of spring onset, and runoff possibly being reduced by 20–40 % (Wang and Zhang 2011).
Despite these dire predictions, climatic and ecological changes are not new to the Loess Plateau region (Zhao et al. 2013), and the farming community has long been involved in the process of perceiving and adapting to aridity and rainfall variability to reduce their vulnerability (Li et al. 2013), similar to agricultural producers in many other parts of the world (Thomas et al. 2007). Specifically, farmers on the Loess Plateau have reported that drought conditions and temperatures have increased since the early 1980s and precipitation has decreased in the last decade, and these perceptions correspond with local weather station data (Li et al. 2013; Otswald and Chen 2006). In addition, studies have shown that smallholders attribute different levels of risk to different types of climate change. For instance, farmers on the Loess Plateau have perceived changes in rainfall timing to be more important in determining their livelihood success than changes in the amount of rainfall (Hageback et al. 2005). They have also been shown to perceive changes in spring temperature and growing season precipitation, but not other aspects of the changing climate that are less important to their agricultural livelihoods (Burnham et al. 2016). Compared to the literature on smallholder climate perception, little has been done to investigate smallholder adaptation to perceived climate variability and change in the Loess Plateau region. An exception is Hageback et al. (2005), which provides a preliminary understanding of how and why smallholders adapt to climate change and other stressors. This study shows that smallholders in Ansai County, Shaanxi, have become less vulnerable to climate change over the preceding 20 years because they have been able to diversify their livelihoods and have become less dependent on agriculture. Further, this study shows that the adaptations that farmers have made are more likely to be driven by new economic opportunities, such as new markets for crops, and policy changes than they are to be by climate variability and change. Likewise, Li et al. (2013) demonstrated that smallholders in Ningxia have applied a range of practices to retain or increase soil moisture to maintain agricultural production in the face of climate variability and drought conditions. However, their work also showed that changing socioeconomic conditions were more important determinants of livelihood trajectories than climate variability or change.
To identify interviewees, we consulted with the local agricultural bureau in each county to identify households for whom agriculture was an important component of their livelihoods. We then used a snowball sampling method to recruit additional interviewees (Noy 2008). The interviews were designed to elicit information about general village and farming life; changes smallholders had made to their farm management and livelihood practices over the last 30 years and why; their perceptions of past and future climate change and attendant risk; the major challenges and risks they face; and their social and professional networks and socioeconomic status. A qualitative approach is useful for obtaining data about a little understood topic or area (Didier and Brunson 2004), which can then be used to inform the development of an appropriate set of questions for a broader survey (Tremblay 1957).
In 2012, a survey of 483 smallholder households was completed. The survey questionnaire was designed iteratively, drawing on theoretical insights established in a literature review and findings from the interviews conducted in 2011. The survey was designed to collect information about household socioeconomic characteristics, perceptions of past and future climate change, adaptations made in response to climate change and other livelihood stressors, livelihood challenges and risks from climatic and non-climatic sources, the perceived impacts of future climate change on farming and livelihoods, and factors affecting farm management decision-making.
The survey was conducted in three counties in the Loess Plateau region: Mizhi and Yangling, in Shaanxi, and Hongsipu (county-level municipality) in Ningxia. Graduate students from Northwest Agriculture and Forestry University (NWAFU) in Yangling, Shaanxi, were trained and employed as survey enumerators. The survey was pretested in two villages near NWAFU, and subsequent refinements were made. Within each county, stratified random sampling was used to select eight villages in which household surveys were conducted. The eight villages were stratified according to their distance from the county population center to account for socioeconomic differences, with seven villages categorized as far (>20 km), eight villages categorized as middle (>10 km and ≤20 km), and nine villages categorized as near (≤10 km). Within each village, simple random sampling was used to select 20 households and the primary agricultural decision-maker was surveyed face-to-face in each household by an enumerator.
To assess the role of determinants of adaptive capacity in shaping smallholders’ perceived self-efficacy and adaptation intent, we constructed two empirical models, one for each dependent variable. One dependent variable is EFFICACY, which represents smallholder perceived self-efficacy. The other is INTENT, which represents smallholder adaptation intent (Supplementary Table 1). To measure EFFICACY, we asked respondents to indicate their ability to make changes to their farming practices or livelihoods to prevent damage caused by climate change without assistance from agricultural professionals or the government. The respondents were able to choose from five options: (1) not possible, (2) very difficult, (3) somewhat difficult, (4) somewhat easy, and (5) very easy. Because only five respondents indicated that adapting to climate change would be very easy, those responses were combined with option four in our models. To measure INTENT, we asked respondents to indicate the likelihood that they would make changes to their farming practices or livelihoods to prevent damage caused by climate change without assistance from agricultural professionals or the government based on their current understanding of the situation. The respondents were able to choose from four options: (1) very unlikely, (2) somewhat unlikely, (3) somewhat likely, and (4) very likely. The independent variables in our empirical models were drawn from the determinants of adaptive capacity identified by Eakin and Lemos (2006). As discussed above, we were able to collect data on variables in four adapted categories of objective determinants (Table 1). A detailed description of each independent variable is provided in Supplementary Table 1. In the INTENT model, we included EFFICACY as an independent variable to examine the role of smallholders’ perception of their ability to adapt in determining their adaptation intent.
A key of assumption of the proportional odds model is that logit coefficients are equal across logit equations for the different cut points, implying that the effect of each independent variable on the log odds of the dependent variable is the same regardless of which levels of the dependent variable are being compared. This enables the reporting of a single coefficient for each independent variable, similar to a binary logit model (Fullerton 2009). We used a Wald test to determine whether our data violate this assumption (Brant 1990; Long and Freese 2006). The overall equality of the coefficients (i.e., omnibus test) and each individual variable were tested. No statistically significant test statistics were returned at the 0.05 level, indicating that the assumption was not violated. Pairwise correlations were calculated for all independent variables in the models to check for multicollinearity. All pairwise correlations fell below 0.6; thus, all independent variables were retained. The variance inflation factor (VIF) was also calculated for both empirical models. The VIF for the full EFFICACY model was 1.70, while the VIF for the full INTENT model was 1.71. Both are below the commonly used heuristic that a VIF of 10 may indicate a problem of multicollinearity.
Principal component analysis
Three sets of survey items in the smallholder household survey questionnaire measured various aspects of smallholder perceptions of climatic and non-climatic sources of risk. The first set included 15 items measuring perceived risks and challenges from non-climatic sources to livelihoods. The second set included 16 items measuring perceived risks and challenges from climatic sources to livelihoods. The third set included 10 items measuring perceived impacts of climate change on farming and livelihoods. High correlations among several of the survey items within each set indicated that the data are not one-dimensional, warranting a data reduction procedure. Principal component analysis (PCA) is a statistical technique that can be used to reduce a large number of correlated variables into a smaller number of uncorrelated, composite variables called principal components with a minimal loss of information. The results of a PCA are usually discussed in terms of PC loadings, which represent the correlations between the survey items and the PCs, and are used to define and name the PCs. An absolute PC loading of 0.50 or greater indicates a strong association among survey items used to generate that PC, and in this study we used a cutoff point of 0.55. We also calculated Cronbach’s alpha for each PC to measure the internal consistency of survey items to determine the reliability of the PC (Cronbach 1951). Generally speaking, values of 0.70 or higher indicate sufficient scale reliability (Nunnally 1978), and PCs with a Cronbach’s alpha value lower than 0.70 should be interpreted with caution. To determine which PCs to retain, Kaiser (1958) stated that all PCs with an eigenvalue of one or greater should be retained.
Description of survey items measuring smallholder perceptions of risk and challenge to livelihood from non-climatic sources
Survey items: perceived risk and challenge from non-climatic sourcesa
Rotated principal component loadingb
High cost of farming inputs
Low income from farming
Unpredictable crop markets
Lack of available markets to sell crops
Low or unfair crop prices from crop buyers
Low market prices
Pests and diseases
Not enough water to irrigate sufficiently
Lack of mechanized tools
Lack of access to technical help
Not enough labor to farm properly
Personal illness or injury
Land being taken away (i.e., lack of tenure security)
Polluted irrigation water
Description of survey items measuring smallholder perceptions of risk and challenge to livelihood from climatic sources
Survey items: perceived risk and challenge from climatic sourcesa
Rotated principal component loadingb
Decreased length of rainy season
Less rain during rainy season
Increased rainfall intensity during rainy season
Increased spring temperature
Increased summer temperature
Increased winter temperature
Increased morning/afternoon temperature difference during growing season
Change in the timing of 24 solar terms
Spring arriving late
Longer time for frozen ground to thaw in the spring
Increased spring frosts
Increased spring cold spells
Description of survey items measuring smallholder perceptions of impacts of climate change on farming and livelihood
Survey items: perceived impact of climate change on the respondent’s farma
Rotated principal component loadingb
My crops will need more water
The timing of when I need to irrigate my crops will change
Less water will be available for irrigation
The amount of time during which I can harvest my crops will become shorter
The growing season will become shorter
The amount of land that can be farmed in my village will be reduced
Climate change will negatively affect the profitability of my farm in the future
Farming will no longer be profitable in my village
My crop yields will decrease
Pest invasions will increase
Profile of smallholder respondents and their farms
Household and farm characteristics of smallholder survey respondents
Household characteristics by county
(n = 160)
(n = 162)
(n = 161)
Gender (% men)
Years of schooling
No. of household members providing farm labor
% employed solely a market-oriented strategy
% employed solely a subsistence strategy
% employed a mixed subsistence and market-oriented strategy
Landholding size (mua)
No. of parcels
Years of agricultural experience
% with access to irrigation water
% as member of a farmer association
No. of extension visits last year
No. of times that land was reallocated in the last 10 years
% perceiving land reallocation will occur in the next 10 years
% of households with at least one member engaged in off-farm work
Annual household income (Yuanb)
% of income from agriculture
% of income from general off-farm work
% of income from local labor or business
% of income from migration
Generally speaking, respondents in each of the three counties were dependent on agriculture to both provide income and supply family food needs, though to varying degrees. In Mizhi, 58 % of respondents identified themselves as employing a diversified subsistence/market strategy, and only 8 % farmed solely for market purposes. In Yangling, 51 % of respondents employed a mixed strategy while 31 % farmed solely for market purposes. In Hongsipu, 25 % employed a mixed strategy while 63 % farmed solely for market purposes. Maize was a major crop across the three counties, with 83 % of respondents growing it. In response to locally led initiatives to alleviate poverty, respondents in Yangling and Hongsipu have started to slowly move into kiwi (6 %) and goji (11 %) production. Aggregated across the three counties, 33 % of respondents grew vegetables or fruit as their primary cash crops, while 64 % of respondents raised livestock, primarily chickens for personal use and pigs for market purposes. In Hongsipu, respondents also raised sheep for sale in times of financial emergency.
Access to irrigation water was differentially distributed across the three counties. In Hongsipu, 100 % of respondent households had irrigated land, while 79 % in Yangling and 41 % in Mizhi did. Of the respondents with irrigated land, 48 % reported using an irrigation schedule. Nearly all households had access to and used some basic technologies, such as pesticides (94 %), fertilizer (91 %), and store-bought seeds (90 %). More expensive technologies such as plastic mulching (37 %), greenhouses (7 %), and various types of water saving irrigation (e.g., drip irrigation) (7 %) were less prevalent.
Respondents were asked to describe their ability to make changes to their farming practices or livelihood to prevent damage caused by climate change without information or assistance from agricultural professionals or the government (i.e., perceived self-efficacy). Of the 483 respondents, 45 % described adaptation as not possible, 33 % very difficult, 16 % somewhat difficult, and 6 % somewhat or very easy. Respondents were then asked to describe their ability to make changes with information or assistance from agricultural professionals or the government. In response, 2 % of respondents described it as not possible, 2 % very difficult, 15 % somewhat difficult, and 81 % somewhat or very easy.
The nested proportional odds ordered logistic regression model assessing factors that influence smallholders’ perceived self-efficacy to adapt to climate change was statistically significant (p < 0.01) across each of the four blocks of the full model (Supplementary Table 2). Several variables were consistently statistically significant at the 5 % level across the nested models. Of the variables measuring the effect of human capital on perceived self-efficacy, having completed middle school or higher, the presence of more household members who provided agricultural labor, income from a local off-farm source, and previous adaptation to rainfall-related climatic changes all had positive significant relationships with perceived self-efficacy.
The three PCs (PRCP_RISK, PRDCT_RISK, FROST_RISK) measuring respondents’ perceived risk posed to livelihood from climatic sources had little to no statistically significant relationship with respondents’ perceived self-efficacy. The same was true for each of the three PCs (WTRSSN_IMPACT, PRFT_IMPACT, YLD_IMPACT) measuring perceived impacts of climate change on farming and livelihood. In contrast, each of the four PCs measuring respondents’ perceived risk posed to livelihood from non-climatic sources had a statistically significant negative relationship with respondents’ perceived self-efficacy in at least one of the four nested models. The largest negative effect was that of perceived market risk on perceived self-efficacy. Specifically, in the full model (i.e., model 4), for a one unit of change in MRKT_RISK, the odds of a respondent indicating the highest level of perceived self-efficacy versus the three lower levels was 0.54 when holding other variables constant.
Building upon model 1, we added four variables to measure the effect of material and non-material resources (i.e., information and technology, material resources and infrastructure) on perceived self-efficacy. The resulting model 2 did not have a statistically better fit than model 1. We added five additional variables to measure the effect of wealth and financial capital on perceived self-efficacy. The resulting model 3 had a statistically better fit than models 1 and 2 (likelihood-ratio test, p < 0.01), and all the added variables were statistically significantly associated with perceived self-efficacy. Specifically, having a higher annual household income was positively associated with reporting higher self-efficacy. For example, for respondents with an annual household income of more than 30,000 Yuan, the odds of them reporting higher self-efficacy was 3.17 times higher than the odds for respondents with an annual household income of <10,000 Yuan. Those who indicated market access as an important factor for making farm management decisions were also more likely to perceive a higher level of self-efficacy. In the full model (i.e., model 4), we added seven variables to measure the effect of institutions and entitlements on perceived self-efficacy. While the model was a statistically better fit than the previous three models (likelihood-ratio test, p < 0.01), the only variable that had a statistically significant association with self-efficacy was HONGSIPU. This result indicates that for respondents from Hongsipu, the odds of them perceiving higher self-efficacy was 0.39 times lower than the odds for respondents from Yangling. In addition, two previously added variables became statistically significant in the full model. The use of an irrigation schedule was positively associated with perceived self-efficacy, while using plastic mulching (usually for maize production) was negatively associated with perceived self-efficacy.
Respondents were asked to indicate the likelihood that they would make changes to their farming practices or livelihood to prevent damage caused by climate change based on their current understanding and situation (i.e., adaptation intent). Of the 483 survey respondents, 32 % indicated it was very unlikely they would adapt, 42 % somewhat unlikely, 24 % somewhat likely, and 3 % very likely. Respondents were also asked about their likelihood to participate in a program run by the government or local university that would help them make changes to their farming practices or livelihood to prevent damage caused by climate change. The majority of respondents (97 %) indicated they were somewhat or very likely to participate in such a program.
The nested proportional odds ordered logistic regression model assessing factors that influence smallholders’ adaptation intent was statistically significant (p < 0.01) across each of the four blocks of the full model (Supplementary Table 3). Several variables were consistently statistically significant at the 5 % level across the nested models. Of the variables measuring the effect of human capital on adaptation intent, having completed middle school or higher, having previously adapted to rainfall-related climatic changes, and the amount of perceived risk to livelihood from market sources were all positively associated with adaptation intent. As opposed to the EFFICACY models, several PCs that measured respondents’ perceived risk from climatic sources and perceived impacts of climate change had statistically significant associations with adaptation intent. Specifically, respondents who perceived a higher amount of risk to their livelihood from changes in precipitation-related events, a higher amount of risk to their livelihood from increased unpredictability of climate events, and a higher level of impact of climate change on the profitability of agriculture reported a lower likelihood of make changes to their farming practices or livelihood to adapt to climate change. Related to this, respondents who stated that the climate had been changing over the last 30 years reported a lower adaptation intent than those who stated that the climate had not been changing. Across the nested models, the largest positive effect was that of perceived self-efficacy on adaptation intent. Specifically, in the full model (i.e., model 8), for respondents with the highest level of perceived self-efficacy (i.e., somewhat or very easy), the odds of them indicating an higher level of adaptation intent was 21.29 times higher than the odds for respondents with the lowest level of perceived self-efficacy (i.e., not possible).
Model 6, which included four variables measuring the effect of material and non-material resources, had a statistically better fit than model 5, which only included variables measuring human capital (likelihood-ratio test, p < 0.01). Specifically, having a greenhouse increased the odds of a respondent indicating a higher level of adaptation intent. Building upon model 6, we added variables to measure the effect of wealth and financial capital on adaptation intent. The resulting model 7 did not have a statistically better fit than models 5 and 6. Finally, we included variables measuring institutions and entitlements in the full model (i.e., model 8). While the full model had a significantly better fit than the previous three models (likelihood-ratio test, p < 0.01), the only variable that had a statistically significant association with adaptation intent was DIST_FAR. This result indicates that for respondents who lived more than 20 km away from a local population center, the odds of them indicating a higher adaptation intent was 0.42 times lower than the odds for respondents who lived within 10 km of a local population center.
The profile of smallholder farmers who participated in our survey broadly agrees with previous findings from studies conducted in the Loess Plateau region of China (Burnham et al. 2016; Hageback et al. 2005). Our results generally confirm the findings of earlier research examining the cognitive dimensions of climate change adaptation. In particular, our results provide strong empirical support for the hypothesis that an individual’s perception of their self-efficacy plays an important role in determining intention to adapt (Grothmann and Patt 2005).
In terms of factors influencing perceived self-efficacy and adaptation intent, our results suggest an association between higher incomes and higher levels of perceived self-efficacy. Previous research has offered some possible explanations. One is that households with more assets may be more likely to undertake adaptation activities because they have capital reserves that they can draw on to pay for the activities (Lemos et al. 2013). Another is that higher incomes may enable households to take on risks associated with adaptation activities without jeopardizing the short- or long-term welfare of household members (Panda et al. 2013). It is worth noting that our results suggest no association between income level and adaptation intent while holding other variables, including perceived self-efficacy, constant. One possible explanation is that households with higher incomes in our study were dependent on off-farm jobs more than they were dependent on agriculture, thus making them less likely to adapt their agricultural livelihood to climate change.
Our study also suggests that smallholders with experience adapting to precipitation-related changes in the last 30 years were more likely to report a higher level of perceived self-efficacy and adaptation intent. This result is consistent with what Kuruppu and Liverman (2011) found in Kiribati, suggesting that individual perceived self-efficacy may be more influenced by past experiences dealing with climatic stressors than an understanding of likely climate change impacts.
In addition to past experiences, our results suggest that the use of technology may contribute to higher levels of perceived self-efficacy and adaptation intent by mediating the impacts of climate change on smallholder production. Specifically, we found that smallholders who used an irrigation schedule were more likely to report higher levels of perceived self-efficacy and those who used greenhouses were more likely to report higher levels of adaptation intent. Greenhouses protect crops from climate variability because they allow temperature and water availability to be manipulated in a climate-controlled setting. Similar to greenhouses, irrigation schedules allow smallholders to reduce the threat of water scarcity, a major climatic constraint to crop production in the Loess Plateau region, particular in the early growing season. Interestingly, the use of plastic mulching had a negative relationship with perceived self-efficacy. In the Loess Plateau region, plastic mulching is mostly used in maize production. It enables smallholders to mitigate the negative impacts of early growing season water shortages. In our study, plastic mulching was primarily used in rainfed systems. Thus, while it does enable a certain degree of control over the climate, smallholders may also view it as being all that can be done to deal with aridity short of irrigation. As the MPPACC predicts, smallholders may view plastic mulching as having a low adaptation efficacy, and, as a result, they would have lower perceived self-efficacy levels.
Our study provides further evidence of the role of local institutions in mediating perceived self-efficacy, as shown in previous research (Agrawal and Perrin 2009; Wang et al. 2013). Specifically, respondents from Hongsipu tended to have a lower level of perceived self-efficacy. This may be partly related to the history of Hongsipu and the role of government in shaping smallholder livelihoods in the area. Hongsipu was created in 1998 as a dam construction project relocation site for households from nearby marginal hillsides where government services and irrigation infrastructure were lacking. Many respondents raised livestock in their former villages and did not grow crops. To promote agricultural production, the government sent extension agents to live in Hongsipu for a minimum of 1 year to teach villagers how to farm. Many villagers also participated in government-sponsored workshops four to six times a year to learn agricultural skills, such as pruning. The government set up a canal system to deliver irrigation water from the Yellow River to all the villages included in our study. Furthermore, local and regional governments have assisted in the development of goji berry as a cash crop for the area. Thus, the government as a whole has had a heavy hand in developing agriculture as a viable livelihood pursuit in Hongsipu. As a result, smallholders have become dependent on state involvement. It is not surprising that such dependency would dampen how smallholders view their own ability to make changes to their farming practices and livelihoods to adapt to climate change.
In our study, availability of farm labor and performing local wage work both positively affect perceived self-efficacy, while households with members migrating for work on a permanent or semi-permanent basis did not have higher perceived self-efficacy. We suggest that this is because the relationship between off-farm work and perceived self-efficacy is mediated through the impact of off-farm work on farm labor availability. Even though local wage work and labor migration may provide the same level of income, they have different impacts on the availability of farm labor. Local wage workers are able to provide farm labor during peak times such as harvest season, while migrant workers may not be able to. Thus, holding income and availability of farm labor constant, local wage work contributes to smallholders’ perceived self-efficacy, but labor migration does not. Similarly, our results indicate a negative relationship between perceived risk from personal illness or injury and perceived self-efficacy, which we believe is also mediated through the impact of personal illness or injury on the availability of farm labor. Other studies have also found that farmers perceived the risk of illness and injury to be similar to or greater than the risk posed by climate change (Thomas et al. 2007).
In much of the literature that examines individual response to natural hazards or climate change, researchers assume that higher levels of perceived risk should be positively correlated with higher levels of adaptation intent (Lo 2013). In the MPPACC, the influence of perceived risk is mediated by perceived adaptive capacity, which is partly determined by perceived self-efficacy. Our study provides empirical evidence of this contention. We found that perceiving the climate has changed over the last 30 years, as well as perceiving high levels of risk to agriculture from changes in precipitation-related events and from increased unpredictability of climate events, did not affect smallholders’ perceived self-efficacy, but lowered their adaptation intent. One possible explanation for this relationship between higher perceived climate-related risks and lower adaptation intent is that smallholders may view the risk of adapting as greater than the risk of not adapting because they are concerned about other non-climatic risks (Patt and Schröter 2008). Specifically, our study shows that smallholders generally perceive risk to livelihoods from market sources. Market-related factors are frequently reported by smallholders to pose greater risk to their livelihoods than climate-related factors (Burnham et al. 2015; Frank et al. 2011; Gandure et al. 2013; Tucker et al. 2010). In the Loess Plateau region of China, agricultural strategies for helping smallholders adapt to climate change generally require changes to current cropping or irrigation practices, which will likely increase their engagement with markets, such as finding markets for new crops and selling more crops from increased production. Thus, smallholders may view these adaptation strategies as sources of greater or new risk to their livelihoods, decreasing their adaptation intent.
In the MPPACC, reliance on public adaptation is one factor that decreases an individual’s appraisal of the risk that climate change poses to them, as well as the likelihood they would engage in adaptive behavior on their own. While we did not directly measure the effect of public adaptation on smallholder risk perception, our results add a new dimension to thinking about the role public adaptation plays in the MPPACC. Throughout the Loess Plateau region, local and regional governments, as well as agricultural research centers and universities, have been heavily involved in building irrigation infrastructure and disseminating irrigation technologies, developing and distributing drought-resistant hybrid maize seeds, and offering subsidies that enable smallholders to purchase pesticides, fertilizers, and plastic mulching, among other things. Thus, it is not surprising that in our study smallholders reported low levels of perceived self-efficacy and adaptation intent without information or assistance from agricultural professionals or the government. Conversely, they reported a much greater ability to adapt with information or assistance from agricultural professionals or the government, as well as an overwhelming likelihood to participate in a program run by the government or local university to help them adapt. Indeed, during our interviews we asked smallholders what could be done to adapt to climate change and the most common response was that it is impossible for them to adapt on their own and that it is the government’s responsibility to lead adaptation efforts.
Further evidence of the relationship between public adaptation and perceived self-efficacy is demonstrated by the result that smallholders from Hongsipu tended to have a lower level of perceived self-efficacy. As previously discussed, the government has been heavily involved in the establishment and development of Hongsipu and agricultural livelihoods in the area. Thus, smallholders seem to have developed a reliance on state intervention and are less confident in their own ability to adapt without government help. Similar results have been reported previously (Crate 2011; Eakin and Bojorquez-Tapia 2008; Lemos et al. 2013; Saldaña-Zorrilla 2008), noting that households’ long-term capacities may be eroded when state-sponsored welfare programs create “dependencies between state and society.” Such dependencies can create rigidities that may decrease the ability of smallholders to adapt to climatic or other stressors on their own (Lemos et al. 2013). Our study further demonstrates the role that state-society dependencies play in determining smallholders’ perceived self-efficacy, which will shape their adaptation intent.
Our study contributes to understanding the cognitive dimensions of adaptation decision-making as conceptualized in the MPPACC (Grothmann and Patt 2005). We systematically examined how smallholders’ perceptions of various factors that may enhance or hinder adaptation shape their perceived self-efficacy and adaptation intent. Our study provides further evidence that self-efficacy beliefs are a strong, positive predictor of adaptation intent, as suggested by recent research (Grothmann and Patt 2005; Kuruppu and Liverman 2011). Further, our study indicates that higher income and past adaptation experiences contribute to higher perceived self-efficacy. Farm labor availability also contributes to perceived self-efficacy and mediates the effect of off-farm work as a livelihood diversification strategy on smallholders’ ability to adapt. Our study challenges the notion that perceiving risks to agriculture and livelihoods from climate change leads to intention to adapt. Our results show that perceiving climate change risks and impacts on agriculture were either negatively associated with or had little effect on adaptation intent. Finally, our study adds a new dimension to understanding the role that local institutions and public adaptation play in determining perceived self-efficacy and adaptation intent. Specifically, our study indicates that state-society dependencies, developed through government-led livelihood interventions and agricultural development projects, may reduce smallholders’ perceived self-efficacy. Overall, our findings highlight the importance of incorporating both objective and subjective determinants of smallholder adaptive capacity into future climate change adaptation programs and policies in order to facilitate adaptation actions.
6.47 Chinese Yuan is equal to 1 U.S. dollar.
This research was partially funded by Northwest Agriculture and Forestry University (NWAFU) through the 111 project of Chinese Ministry of Education (No. B12007). The authors are grateful to Drs. Pute Wu, Delan Zhu, Youke Wang, Xining Zhao, Xiping Liu, Yubao Wang from NWAFU for their support for this research. The authors also thank Chunyan Qi, Mengying Sun, and several other NWAFU undergraduate and graduate students for their assistance during fieldwork. Finally, we would like to thank our anonymous reviewers for their helpful comments, which significantly strengthened the paper.
Compliance with ethical standards
Conflict of interest
The authors declare that they have no conflict of interest.
- Adger WN, Agrawala S, Mirza MMQ et al (2007) Assessment of adaptation practices, options, constraints and capacity. Climate change 2007: impacts, adaptation and vulnerability. Contribution of working group II to the fourth assessment report of the intergovernmental panel on climate change. Cambridge University Press, Cambridge. http://www.ipcc.ch/pdf/assessment-report/ar4/wg2/ar4-wg2-chapter17.pdf. Accessed 29 May 2015
- Agrawal A, Perrin N (2009) Climate adaptation, local institutions and rural livelihoods. In: Adger WN, Lorenzone I, O’Brien KL (eds) Adapting to climate change: thresholds, values, governance. Cambridge University Press, Cambridge, pp 350–367Google Scholar
- Barnett J, Waters E, Pendergast S, Puleston A (2013) Barriers to adaptation to sea-level rise: the legal, institutional and cultural barriers to adaptation to sea-level rise in Australia. National Climate Change Adaptation Research Facility, Gold CoastGoogle Scholar
- Brooks N, Adger WN (2005) Assessing and enhancing adaptive capacity. In: Lim B, Spanger-Siegfried E (eds) Adaptation policy framework for climate change: developing strategies, policies, and measures. Cambridge University Press, Cambridge, pp 165–182Google Scholar
- Fishbein M, Ajzen I (1975) Belief, attitude, intention and behavior: an introduction to theory and research. Addison Wesley, ReadingGoogle Scholar
- Klein RJT, Midgley GF, Preston BL, Alam M, Berkhout FGH, Dow K, Shaw MR (2014) Adaptation opportunities, constraints, and limits. In: Field CB, Barros VR, Dokken DJ, Mach KJ, Mastrandrea MD, Bilir TE, Chatterjee M, Ebi KL, Estrada YO, Genova RC, Girma B, Kissel ES, Levy AN, MacCracken S, Mastrandrea PR, White LL (eds) Climate change 2014: impacts, adaptation, and vulnerability. part a: global and sectoral aspects. Contribution of working group II to the fifth assessment report of the intergovernmental panel on climate change. Cambridge University Press, Cambridge, pp 899–943Google Scholar
- Leichenko R, O’Brien K (2008) Environmental change and globalization: double exposures. Oxford University Press, OxfordGoogle Scholar
- Lemos MC, Agrawal A, Eakin H, Nelson DR, Engle NL, Johns O (2013) Building adaptive capacity to climate change in less developed countries. In: Asrar G, Hurrell J (eds) Climate science for serving society. Springer, Dordrecht, pp 437–457Google Scholar
- Long JS (1997) Regression models for categorical and limited dependent variables. Sage Publications, Thousand OaksGoogle Scholar
- Long JS, Freese J (2006) Regression models for categorical dependent variables using Stata. Stata Press Books, College StationGoogle Scholar
- Mertz O, Mbow C, Maiga A, Diallo D, Reenberg A, Diouf A, Barbier B et al (2010) Climate factors play a limited role for past adaptation strategies in West Africa. Ecol Soc 15(4):1–15. http://www.ecologyandsociety.org/vol15/iss4/art25/ Google Scholar
- Nunnally JC (1978) Psychometric theory, 2nd edn. McGraw-Hill, New YorkGoogle Scholar
- Ribot J (2009) Vulnerability does not fall from the sky: towards multiscale, pro-poor climate policy. In: Mearns R, Norton A (eds) The social dimensions of climate change: equity and vulnerability in a warming world. World Bank Publications, Washington, pp 47–74Google Scholar
- Smit B, Pilifosova O (2001) Adaptation to climate change in the context of sustainable development and equity. In: IPCC Working Group II (eds) Climate change 2001: impacts, adaptation and vulnerability. Earth Scan, Geneva, pp 879–906Google Scholar
- Triandis HC (1977) Interpersonal behavior. Brooks/Cole Publishing Company, MontereyGoogle Scholar