Determinants of Climate-Smart Adaptation Strategies: Farm-Level Evidence from India

Abstract

The paper attempts to examine the factors that influence climate-smart adaptation (CSA) strategies. The study used binary logit and multivariate probit models to understand the dynamics and factors of agricultural households’ behavioural decisions on CSA strategies. Based on the results of the binary logit model, the study indicated that factors such as access to extension services and training, gender, educational level, land ownership, access to irrigation, access to credit and crop damage level positively and significantly influenced farmers’ decisions to use CSA strategies. Similarly, the results of the multivariate probit model reveal that factors such as educational level, access to extension services and training, and land ownership had significant impacts on the adoption of the majority of CSA strategies. To improve the intensity of CSA strategies, the study recommends expanding training and extension services to farming masses, the expansion of irrigation facilities and weather information at the farm level.

Keywords

climate-smart adaptation agricultural household binary logit model multivariate probit model adaptive capacity

Introduction

According to the Intergovernmental Panel on Climate Change (IPCC), ‘the climatic system is now clearly warming, as evidenced by observable increases in global average air and ocean temperatures, melting of glaciers and snow, and rising global sea level’ (Solomon et al., 2007: 5). Numerous studies demonstrate that climate change has a disproportionately negative impact on agricultural and smallholder farming systems in poor nations, including India (Bandara and Cai, 2014; Kahsay and Hansen, 2016). It jeopardises food security and agricultural development goals – namely, raising farmer income and reducing poverty – in the face of an increase in the frequency of climate extremes, such as flooding and severe drought (Food and Agriculture Organization (FAO), 2014). As a result of their lack of information about the potential CSA options, constrained assets and risk-taking capacity to acquire and employ technologies and financial services, smallholder farmers and pastoralists suffer the brunt of climate change repercussions (FAO, 2014; Moharaj and Rout, 2021; Sahoo and Sridevi, 2021).

Agriculture in India, which accounts for around 16% of total GDP and employs approximately 49% of total employment (Economic and Survey, 2017 -18), is being threatened by the potential consequences of climate change and variability. Kumar and Parikh (2001) estimated that climate change reduced net farm revenue per hectare in India by approximately 9% to 25% using Ricardian approach. Sanghi and Mendelsohn (2008), also utilizing the same Ricardian technique, found that a 2°C temperature increase plus a 7% increase in rainfall result in a nearly 10% decrease in agricultural net revenue. In addition, Sanghi et al. (1998) and Kumar (2011) demonstrated that climate change has a detrimental influence on Indian agriculture. However, in a crop-specific context, Pattanayak and Kumar (2014) reported an annual loss of 4.4 million tonnes of rice between 1969 and 2007. In addition, wheat production is expected to have decreased by approximately 5.2% between 1981 and 2009. According to Kumar et al. (2006), a 19% decline in summer monsoon rainfall affects food grain output by around 18%. In addition, a recent study by Pingali et al. (2019) found an overall output loss of 4% to 8%, which may climb up to 25% by the year 2050. The actual impact of climate change on agriculture varies with the crop, location and farmer’s adaptive capacity to climatic hazards (Vermeulen et al., 2012) and “farmers’ adaptable capacities also drive agricultural productivity” (Aryal et al., 2020; Panda et al., 2013). For example, “farmers in the Hindu Kush Himalayan region, which includes India, are especially vulnerable to climate change due to their increased reliance on agricultural food and livelihood, restricted access to global markets, low crop productivity, and weak infrastructure” (Rasul et al., 2019).

Climate-Smart Agriculture, as defined by the Food and Agriculture Organization (FAO) at the 2010 Hague Conference on ‘Agriculture, Food Security, and Climate Change’, aims to achieve sustainable development goals by addressing three interconnected challenges: (1) sustainably increasing agricultural productivity and farm income; (2) adapting and strengthening farming systems’ resilience to climate change; and (3) reducing greenhouse gas emissions. Increasing marginal and small farmers’ adaptive capacity and making adaptation alternatives available to all types of farmers are key components of developing sustainable solutions to the negative consequences of climate change on agriculture.

Adaptation strategies are defined as changes made to a system’s resilience or vulnerability to observed or projected changes in its climatic condition (IPCC, 2007). Adaptation to climate change has been classified in the literature as either autonomous or planned. At the farm and household level, private players facilitate autonomous adaptations (Stage, 2010), whereas the latter is facilitated by government policy intervention (Stage, 2010). Numerous existing climate-smart adaptation strategies and choices are being implemented in response to the harmful implications of climate change. The following are appropriate adaptation strategies: ‘crop or livelihood diversification’ (Thornton and Herrero, 2014); ‘changing crop types and varieties’ (Fisher et al., 2015); ‘crop rotation, intercropping, composting, and using different fertilisers’ (Nigussie et al., 2018); ‘assured irrigation and water use, watershed management’ (Belay and Beyene, 2013); ‘Improved crop varieties’ (Loria and Bhardwaj, 2016; Raghavendra and Suresh, 2018); ‘drought and flood-resistant varieties’ (Fisher et al., 2015; Loria and Bhardwaj, 2016; Raghavendra and Suresh, 2018) and ‘early-maturing varieties’ (Kibue et al., 2016; Prasad et al., 2014) have all been demonstrated to be the effective and efficient adaptation strategies and strategies by farmers.

Odisha, one of India’s eastern states, is regarded as the capital of natural disasters (see Table 1). The agriculture of Odisha is vulnerable to the unfavourable consequences of climate change due to the state’s climate, which is characterized by high temperatures and humidity. It receives approximately 80% of annual precipitation during the South-West monsoon (June–September) (Mishra and Mishra, 2010). However, rainfall quantity and distribution are insufficient, irregular, and unequal over time and space. In addition, it has been established that just 13 years out of 52 have normal rainfall, giving the state a 75% probability of being hit by a natural calamity of any kind (Natural Disaster Plan for Odisha, 2014). These natural disasters have a significant impact on crop production and productivity. For instance, after a solid agricultural growth rate of 12.30% in 2012–2013, the state experienced a negative growth rate in 2013 to 2014, owing mostly to super-cyclonic storm Phailin and flash flooding (Odisha Economic Survey, 2014–2015).

Table 1.

Chronology of natural calamities in Odisha during the period 2010–2020.

Year	Natural calamities
2010	Flood, Heavy Rain, Drought & unseasonal cyclonic rain
2011	Drought & flood
2012	Drought & flood
2013	Very Severe Cyclonic Storm ‘Phailin’, flood
2014	Very Severe Cyclonic Storm ‘Hudhud’, flood
2015	Drought, flood & heavy rain
2016	Drought, flood & heavy rain
2017	Flood, Heavy Rain, Drought & Pest Attack, Unseasonal Rain
2018	Cyclonic Storm ‘Titli’ and ‘Pethai’ / Drought
2019	Extremely Severe Cyclonic Storm ‘Fani’ and ‘Bulbul’
2020	Severe Cyclone ‘AMPHAN’ and heavy flood

Source: Odisha Economic Survey (2020–2021).

Against these backdrops, there is an urgent need to investigate CSA strategies and understand the factors influencing them in a more context-specific manner by combining multiple socioeconomic, institutional and developmental dimensions. In the Odisha context, there is scanty research notably at the micro-level. The study intends to investigate the following objectives in light of these issues:

(i) to identify and investigate CSA strategies that farm households employ to mitigate climate risks and (ii) to gain an insight into factors or determinants that influence farmers’ adoption of CSA techniques.

The study is organized as follows. The second section discusses the data sources, study area and empirical strategy. The third section delineates the results and discussion, and the fourth section summarizes the conclusion and policy implications.

Data sources, study area and empirical strategy

Data source

The study is based on secondary data and primary data. The secondary data on rainfall and temperature were collected from CMIE and the India water portal. Primary data were obtained during the agricultural year 2019–2020 from a random sample of 170 farm households¹ representing four distinct farmer classifications² in Mandiapalli village, Ganjam district, Odisha, using a semi-structured household interview schedule. The study employed a multi-stage sampling technique. In the first stage, the Ganjam district was chosen from 13 undivided districts based on the highest score on the agricultural vulnerability index. The second stage was the selection of one block, Rangeilunda, from a total of 30 blocks based on its sensitivity to several threats such as flooding, cyclones and drought (see Figure 1). This block was severely impacted in 1999 by a ‘super cyclone’, cyclone ‘Phailin’ in 2013 and was also impacted in 2014 by cyclone ‘Hudhud’. The third stage was the deliberate selection of Mandiapalli village in Rangeilunda block as the location of the study area. In the fourth stage, 170 households from the target group of farm households in Mandiapalli village were randomly selected for the study. The criteria for selecting the target population of farm households in the village is that a household’s primary source of income should be farming to understand better the agricultural impact of climate change and variability on agriculture.

Figure 1.

Study area map.

Study area

Ganjam district of Odisha is geographically situated between the latitude range of 19.4-20.17° North and the longitude range of 84.7-85.12° East. Ganjam district has a mixed agro-climatic zone; a part of it is in the East & South-eastern coastal plain, while another part is in the north Eastern Ghat. The district’s wide soil types are classified as lateritic, deltaic, alluvial, saline, red and mixed red and black (District Census Handbook, 2011). According to the 2011 census, approximately 66.64% of working people are engaged in agriculture, and approximately 71.09% of the population is literate. However, marginal and small categories of holdings collectively account for about 87.79% and about 93.44% of all operational holdings in 2005–2006 and 2010–2011, respectively, but account for only 58.48% and 73% of the total operated area, respectively. Table 2 contains information on the household characteristics of the study area.

Table 2.

Household characteristics of the study area (N = 170).

Household characteristics	Number	Percent
Marginal farmers	59	34.71
Small farmers	44	25.88
Medium farmers	25	14.71
Large farmers	42	24.71
Average age of farmers	37	–
Household size	6.44	–
Average farm experience	9	–
Farmers dependent on only farming	77	45.29
Farmers dependent on both farming and non-farming	93	54.7

Source: Author’s calculation based on household survey.

Empirical strategy

The binary logit model was used to understand the factors affecting farmers’ decision to adopt the climate-smart strategies. The farmers’ decision is a discrete variable (1,0), where one (1) indicates the farmers who adapted climate-smart strategies and zero (0) indicates the farmers not adapting climate-smart strategies. The general form of the binary logit model is as follows (Greene, 2003)

P_{i} (Y_{i} = 1) = \frac{e^{X β}}{1 + e^{X β}},

(1)

where $β$ is the vector of parameters, and $X$ is the matrix of factors affecting climate-smart strategies.

The marginal effect (ME) coefficients of the binary logit model are determined by the formula (Greene, 2003; Trinh et al., 2018)

M E = \frac{\partial Λ (X^{'} β)}{\partial X} = Λ (X^{'} β) [1 - Λ (X^{'} β)] β .

(2)

The diagnostic procedures for collinearity and specification error of the binary logit model have been verified. The variance inflation factor (VIF) has been calculated to check for multicollinearity among the explanatory variables. The rule of thumb for VIF is that if it is more than 5, there is a severe multicollinearity problem

V I F (\hat{β}) = \frac{1}{1 - R_{i}^{2}} .

(3)

Multivariate probit model

The multivariate probit (MVP) model was used to analyse the factors affecting farmers’ decision to adopt each of the CSA strategies to cope with extreme climatic conditions in their agricultural production. However, Different CSA strategies or options offer varying degrees of benefit (Mengistu and Haji, 2015). As a result, individual households frequently express their preferences for various strategies based on their expertise and knowledge. However, according to Deressa et al. (2008) and Gbetibouo and Ringler (2009), the decision to adopt or not to adopt any adaptation strategy is made within the ‘general framework of utility and profit maximization’. Furthermore, a rational farmer will generally employ adaptation strategies only when the net benefits of doing so outweigh the costs of not doing so (Mendelsohn, 2012). Although the net benefit is not directly observable, the economic agents’ behaviour (in this case, the individual farmer) is observable through the decisions they make (Deressa et al., 2008). Assume that $Y_{j}$ and $Y_{k}$ represent the benefits a household receives from two different options, indicated by $U_{j}$ and $U_{k}$ , respectively.

The linear regression models can be specified as

U_{j} = β_{j}^{'} X_{i} + ϵ_{j} .

(4)

U_{k} = β_{k}^{'} X_{i} + ϵ_{k}

(5)

If a household decides to use option j, it follows that the perceived benefit from option j is greater than the benefit from other options (i.e. k), which can be written as

U_{i j} (β_{j}^{'} X_{i} + ϵ_{j}) > U_{i k} (β_{k}^{'} X_{i} + ϵ_{k}), j \neq k,

(6)

where $U_{i j}$ and $U_{i k}$ are the perceived benefits for adaptation options j and k by the ith farmer. $X_{i}$ is the matrix of explanatory variables that influences the choice of the adaptation options. $β_{j}^{'}$ and $β_{k}^{'}$ are the parameters to be estimated. $ϵ_{j}$ and $ϵ_{k}$ are the error terms that are assumed to be independently and identically distributed.

The probability that a household will use option j from a set of adaptation options can be defined as follows

\begin{matrix} P (Y_{i} = 1 / X) = P (U_{i j} > U_{i k} / X) = P (β_{j}^{'} X_{i} + ϵ_{j} - β_{k}^{'} X_{i} - ϵ_{k} > 0 / X) \\ = P ((β_{j}^{'} - β_{k}^{'}) X_{i} + ϵ_{j} - ϵ_{k} > 0 / X) = P (β^{*} X_{i} + ϵ^{*} > 0 / X) = F (β^{*} X_{i}) = Y_{i}^{*}, \end{matrix}

(7)

where $β^{*} = β_{j}^{'} - β_{k}^{'}$ is the vector of unknown parameters that can be interpreted as the net influence of the matrix of explanatory variables on adaptation. $ϵ^{*} = ϵ_{j} - ϵ_{k}$ is the random disturbance term. $F (β^{*} X_{i})$ is the cumulative distribution function of $ϵ^{*}$ , evaluated at $β^{*} X_{i}$ . The exact distribution of F depends on the distribution of the random disturbance term, $ϵ^{*}$ .

Econometric specification of the MVP model

Theoretically, the adaption options are ‘highly interrelated and interdependent’. In other words, these alternatives are correlated, which allows for a correlation of the error terms associated with each option’s regression (Belderbos et al., 2004). Nevertheless, ‘the multivariate probit model is capable of eliminating these correlations’ (Huguenin et al., 2009; Nhemachena and Hassan, 2007). The MVP model is defined as ‘simultaneous models in which the set of explanatory factors influences each of the many options and error terms are allowed to be flexibly correlated’ (Greene, 2003). ‘The MVP model presupposes that the multivariate response is an unobserved latent variable originating from a multivariate normal distribution given the explanatory variables’ (Piya et al., 2013). The MVP model for the ith observation and mth equation has the following formula (Cappellari and Jenkins, 2003; Tocco et al., 2013).

$Y_{i m} = 1$ if $Y_{i m}^{*} > 0$ and 0 otherwise (i = 1,2,. . . .,N; m = 1,2,. . .,M)

Y_{i m}^{*} = X_{i m}^{*} β_{m} + ε_{i m}

(8)

‘N is the number of observations; M is the number of options; X_im is the matrix of explanatory variables; $β_{m}$ is the matrix of parameters; and $ε_{i m}$ is the matrix of error terms’

The dependent variable of the MVP model includes ‘eight specific climate-smart adaptation strategies that assume a value of 1 if farmers apply any specific practice and 0 otherwise. In addition, the MVP allows a flexible correlation structure for the unobservable variables’ (Huguenin et al., 2009). The model was also tested for heteroscedasticity using the robust standard error procedure.

Results and discussion

Farm households’ perception of climate change

Changing rainfall patterns are cited as a significant manifestation of climate change and variability, based on the responses of sampled household heads (about 95.88%). Concerning the temperature pattern, around 97.05% of household heads report an increase in the inter-annual temperature (see Figure 2).

Figure 2.

Farmers’ perception of changing rainfall pattern and increase in inter-annual temperature.

Farmers’ perceptions and views on climate change and variability were shown to be consistent with observed trends in meteorological variables (Kom et al., 2020) The examination of meteorological variables in the Ganjam district reveals a growing tendency in surface temperature (maximum and minimum) and a varying amount of annual rainfall. Annual rainfall anomalies from 1901 to 2015 are depicted in Figure 3. It demonstrates that, on average, rainfall has increased by around 0.065% per year since 1901. However, rising patterns in maximum and lowest temperatures are glaringly obvious (see Figure 4). Between 1901 and 2002, we observed an average increase of about 0.005 C and 0.006 C in maximum and minimum temperatures.

Figure 3.

Trend of annual rainfall in Ganjam district, 1901–1902 to 2015–2016.

Figure 4.

Trend of maximum and minimum temperatures in Ganjam district, 1901–2002.

Impact of climate change on crop-level productivity at the micro-level

The intriguing point is that climatic change and variability have had negative effects on the productivity of all crops across different size groups of farmers (see Figure 5). To bolster these findings, two farmers of the study area are asked an open-ended question about why the agricultural productivity of various crops has tended to diminish over time. The first farmer said, “In the current environment, farmers either engage in non-farm businesses or remain idle at home, as farming is no longer a profitable activity. The causes for this are numerous. On the one hand, the primary causes include changing rainfall patterns caused by the unpredictable South-West monsoon, and the growing incidence of pests. On the contrary, agricultural costs, including fertilizer and pesticide prices, are fast growing. This drop in agricultural productivity caused by climate change is extremely noticeable in our region” (Bilas Rabi, Village Head).

Figure 5.

Percent change in productivity for the year 2019-20 compared to the average of 2016–2017 to 2018–2019 in the study area.

The second farmer expressed his opinion:

“Agriculture in the rain-fed area has been subject to the vagaries of monsoon rainfall as a result of the rapid shift in climatic conditions. Every alternate year, the area braces for catastrophic climate occurrences such as cyclones followed by floods and droughts. We faced severe agricultural damage in 2013 as a result of super-cyclone Failin, followed by flooding, and partial damage in 2014 as a result of cyclone Hudhud. During the Kharif season of 2016-17 and 2017-18, the paddy crop became flat due to continued severe rains. Climate change’s impact on cultivated land and agricultural production in our area cannot be overstated.” (Subash Rao)

Factors influencing CSA strategies

We examined the influence of each variable on farmers’ decisions to adopt CSA strategies by using binary logit and multivariate probit model. Table 3 contains a description of the variables used in both binary logit and multivariate probit models. The results of the binary logit model show that Prob > 2 indicates that the models are statistically significant at the 1% level. The value of Pseudo R² in the model indicates that access to extension and training, perception, access to irrigation, the household head’s gender and educational level, access to credit, damage level, and farm size contributed to approximately 79.1% of the probability that farm households would adopt climate change adaptation strategies. Access to extension assistance and training, access to irrigation, gender, educational level, access to credit, the extent of agricultural damage and the size of the farm are all significant variables (Table 4).

Table 3.

Description of variables included in the model.

Variables	Description of variables	Adopters		Non-adopters
Variables	Description of variables	M	SD	M	SD
Dependent variables
Adaptation in logit model	Dummy = 1 if the farmers adopt climate-smart strategy and 0, otherwise	0.65	0.48	0.352941	0.58
Adaptation in multivariate probit model	Dummy = 1 if farmers are using different varieties of seeds and 0, otherwise	0.24	0.34	0.764706	0.32
	Dummy = 1 if farmers are using early-maturing varieties of paddy and 0, otherwise	0.33	0.45	0.670588	0.23
	Dummy = 1 if farmers are planting dates of crops and 0, otherwise	0.20	0.54	0.8	0.49
	Dummy = 1 if farmers are using improved irrigation facilities and 0, otherwise	0.39	0.36	0.605882	0.42
	Dummy = 1 if farmers are adopting crop diversification and 0, otherwise	0.26	0.39	0.735294	0.44
	Dummy = 1 if farmers are adopting agroforestry and 0, otherwise	0.32	0.32	0.676471	0.42
	Dummy = 1 if farmers are adopting livelihood diversification and 0, otherwise	0.47	0.42	0.529412	0.26
	Dummy = 1 if farmers are taking decisions to do migration and 0, otherwise	0.09	0.46	0.905882	0.28
Explanatory variables
Perception	Dummy = 1 if the household perceives the negative effect of climate change; 0 otherwise	0.71	0.56	0.294118	0.29
Farm experience	Years of farming experiences	14.00	27	4	25
Gender	Dummy = 1 for the male-headed household and 0 for the female-headed household.	0.88	0.45	0.117647	0.47
Education	Dummy = 1 if the household head is educated, 0 otherwise.	0.22	0.48	0.776471	0.32
Land ownership	Dummy = 1 if farmers own land, 0 otherwise	0.46	0.46	0.541176	0.29
Access to Irrigation	Dummy = 1 if farmers have access to irrigation, 0 otherwise	0.35	0.42	0.647059	0.49
Income level	Continuous variable (1 US dollar = Rs. 75)	$426	1200	$226	760
Access to Credit	Dummy = 1 if farmers have access to credit, 0 otherwise	0.24	0.34	0.764706	0.52
Access to climate information	Dummy = 1 if farmers have access to climate information through media, 0 otherwise	0.71	0.44	0.294118	0.51
Farm size	Dummy = 1 for the medium and large farmers and 0 for the marginal and small farmers.	0.39	0.53	0.605882	0.322
Access to Extension advice and training	Dummy = 1 if a household has access and 0, otherwise	0.41	0.56	0.588235	0.33
Access to the local market	Dummy = 1 if farmers have access to market, 0 otherwise	0.65	0.58	0.352941	0.47
Farmer-to-farmer extension	Dummy = 1 if farmers have access to farmer-to-farmer extension, 0 otherwise	0.45	0.47	0.552941	0.55
Damage level	Continuous variable (1 US dollar = Rs. 75)	$302	140	$240	270

Table 4.

Maximum likelihood estimates and the marginal effects of binary logistic regression.

Variables	Estimates		Marginal effect
Variables	Coefficients	Standard error	Coefficients	Standard Error
Access to extension and training	0.719**	0.742	0.177***	0.178
Perception	6.342	1.477	0.915	0.054
Access to irrigation	7.340**	1.59	0.930**	0.046
Gender	1.549*	0.979	0.330*	0.162
Education level	3.665***	0.942	0.717***	0.11
Access to credit	2.506*	0.938	0.475*	0.114
Damage level	0.000104**	0.000037	0.000026**	0.00001
Farm size	–3.720***	0.991	–0.666***	0.11
Constant	–9.765***	2.291
No. of observations	170
Log pseudo-likelihood	–20.253
Prob > χ²	0.000***
Pseudo R²	0.791

Robust standard errors in parentheses. Dependent variable is the decision to adopt the climate-smart adaptation strategies.

p < 0.10. **p < 0.05. ***p < 0.01.

Access to extension services and training has a positive and significant impact at the 1% level. It implies that farm households with access to extension services and training are more likely to conduct climate-smart agriculture. Furthermore, in terms of the marginal effect, the probability of farm households having access to extension services and training is quite high at 17.7% as compared to those that do not have access.

The gender variable has a positive and significant coefficient indicating that male-headed farm households are probably more adaptive to CSA strategies. The marginal coefficient of 0.33 (about 33%) suggests that male-headed households are more likely than female-headed households to adhere to CSA strategies. Similarly, the educational level of farm household heads is positively and significantly impacting the log of odd ratios in favour of adapting CSA practices, meaning that households with educated heads significantly increase their likelihood of adopting CSA strategies by about 0.72% as compared to the households with least education.

The positive and significant coefficients of access to irrigation indicate that farm households with access to irrigation facilities are more adaptive to CSA practices than households without access. Similarly, the households with a higher level of agricultural damage have more probability of adopting CSA practices than households with the least damage level due to climate change. This is consistent with Trinh et al. (2018). Furthermore, when farm size is considered, the probability of medium and large farmers adopting CSA strategies is around 66.6% lower than that of marginal and small farms.

In Table 5, the linear predictor (Hat) is significant at the 1% level; however, the squared linear predictor (Hat-square) is insignificant, indicating that the model is free of specification errors. In addition, all explanatory variables have a VIF (Variance Inflating Factor) <5 and a tolerance value >0.20, indicating that the binary logit model does not suffer from multicollinearity (see Table 6).

Table 5.

Diagnostic check for specification error.

Independent variable	Coefficient	Adaptation
Independent variable	Coefficient	Z statistic	p value
Constant	–.2013364	–0.44	0.662
Hat	1.079743***	4.28	0.000
Hat–square	.0535295	1.08	0.279
No. of observations = 140
Log likelihood = –19.7873
Prob > χ² = 0.000
Pseudo R² = 0.7961

Source: Authors’ estimation.

***

represent significance at the 1% level.

Table 6.

Collinear statistics of independent variables.

Variable	Variance inflation factor (VIF)	Tolerance value (TV)
Access to extension and training	3.80	0.2629
Perception	3.23	0.31
Access to irrigation	3.70	0.2701
Gender	3.33	0.3007
Education level	2.02	0.4945
Access to credit	1.41	0.7081
Damage level	3.50	0.2858
Farm size	1.77	0.5643

Source: Authors’ estimation.

The MVP model was used to examine the factors that influence farmers’ decision to implement CSA strategies. The dependent variable of the MVP model assumes eight distinct CSA strategies (using different varieties of seeds, early maturing varieties of paddy, changing planting dates of crops, improving irrigation facilities, crop diversification, agroforestry, livelihood diversification, and migration). It assumes one if farmers follow a particular method and zero if they do not. Prob > 2 values indicate that the overall relationship between farmers’ probability of applying specific CSA strategies and explanatory variables is significant at the 1% level. In addition, as illustrated in Figure 6, the majority of farmers in the study area reported shifting crop planting dates and improving irrigation infrastructure in order to adapt to and mitigate the negative agricultural impacts of climate change and variability. In addition, as indicated by the extremely significant correlation coefficients, there exist complementarities and interdependencies among the various adaptation strategies adopted by farmers (see Table 7). It establishes the validity of the MVP model.

Figure 6.

Percentage distribution of farmers adopting CSA strategies in the study area.

Table 7.

Pair-wise correlation among eight specific CSA strategies in the study area.

Dependent variables	Y₁	Y₂	Y₃	Y₄	Y₅	Y₆	Y₇	Y₈
Y₁	1.000
Y₂	0.709*** (0.001)	1.000
Y₃	0.507(0.501)	0.520(0.630)	1.000
Y₄	0.521*** (0.006)	0.670*** (0.009)	0.041(0.310)	1.000
Y₅	0.700*** (0.001)	0.430*** (0.006)	0.503(0.399)	0.856*** (0.003)	1.000
Y₆	0.350*** (0.010)	0.740*** (0.007)	0.562(0.181)	0.652*** (0.001)	0.770*** (0.008)	1.000
Y₇	0.530*** (0.000)	0.640*** (0.009)	0.520(0.498)	0.640*** (0.007)	0.652*** (0.007)	0.650*** (0.000)	1.000
Y₈	0.231*** (0.000)	0.235*** (0.003)	-0.145(0.604)	0.430*** (0.003)	0.505*** (0.004)	0.453*** (0.000)	0.420*** (0.000)	1.000

Figures in parentheses represent p value. CSA = climate-smart adaptation.

***

Represents the significance at 1% level.

Farm households’ perceptions of negative agricultural impacts due to climate change have a positive and significant effect on their probability of adopting improved irrigation facilities, crop diversification, livelihood diversification, and migration. Household heads with greater farming experience have more probability of adopting different varieties of seeds, crop diversification, and agroforestry than those with less farming experience. More interestingly, the least experienced farmers tend to have more probability of adopting livelihood diversification and migration, but the coefficient is not significant.

There are positive and significant relations between the household head’s gender and the probability of adopting various varieties of seeds. Also, in terms of the positive and significant coefficient associated with household heads’ educational level, educated households are more likely to implement all CSA strategies except livelihood diversification and migration. Similarly, households with land ownership are more likely to respond to the effects of climate change through the use of different varieties of seeds, early-maturing varieties of paddy, improving irrigation facilities, and agroforestry.

With access to irrigation, households are more likely to respond to the agricultural risk associated with climate change and unpredictability through early-maturing paddy varieties, crop diversification, and agroforestry. In addition, households with higher income levels are less likely to adjust to adverse climate change effects through early-maturing paddy varieties and livelihood diversification. Similarly, households with access to credit are more likely to adapt through the adoption of different varieties of seeds, improved irrigation facilities, and livelihood diversification. However, farm size, on the other hand, has a large negative effect on the probability of adaptation. More precisely, marginal and small farmers have a greater chance of adjusting through strategies such as the use of different varieties of seeds and crop diversification.

Households with access to extension services and training have a greater chance of adapting to the adverse effects of climate change through the use of diverse seed varieties, early maturing paddy varieties, shifting crop planting dates, and crop diversification. Furthermore, the proxy of social capital, the farmer-to-farmer extension has a positive and significant impact on the probability of adapting to climatic risks through different varieties of seeds and crop diversification (Table 8).

Table 8.

Maximum likelihood estimates of multivariate probit model.

Explanatory variables	Climate-smart adaptation strategies
Explanatory variables	Different varieties of seeds	Early maturing varieties of paddy	Changing planting dates of crops	Improved irrigation facilities	Crop Diversification	Agroforestry	Livelihood Diversification	Migration
Perception	0.338(0.282)	0.220(0.247)	0.399(0.289)	1.828*** (0.270)	2.551*** (0.383)	0.294(0.471)	8.141*** (0.743)	3.708*** (0.624)
Farm experience	1.489*** (0.294)	0.021(0.258)	0.075(0.274)	0.450(0.305)	0.897** (0.427)	1.791** (0.881)	–0.241(0.440)	–0.493(0.659)
Gender	1.151*** (0.321)	0.410(0.261)	0.343(0.262)	0.125(0.291)	0.105(0.430)	0.736(0.630)	0.486(0.453)	0.445(0.415)
Education	1.366*** (0.288)	1.480*** (0.232)	1.870*** (0.244)	1.254*** (0.240)	1.679*** (0.598)	3.201*** (0.755)	4.786*** (0.464)	1.109(0.648)
Land ownership	2.013** (0.999)	0.164** (0.251)	0.007(0.267)	0.166* (0.309)	0.592(0.436)	0.864* (0.504)	0.120(0.474)	–0.493(0.593)
Access to irrigation	0.465(0.298)	0.187** (0.221)	–0.249(0.231)	1.935(0.669)	1.217** (0.429)	5.068* (1.246)	–0.336(0.363)	3.193(0.753)
Income level	0.00004(0.00002)	0.000004* (0.00001)	–0.000009(0.00001)	0.00002(0.00002)	–0.000008(0.00003)	–0.00006(0.00003)	0.0000001* (0.00002)	0.00003(0.00003)
Access to credit	0.107* (0.292)	0.510(0.236)	–0.045(0.230)	0.222* (0.263)	–0.281(0.387)	0.234(0.526)	0.057* (0.362)	1.251(0.414)
Access to climatic information via media	1.282(0.405)	0.081(0.346)	0.170(0.373)	0.257(0.429)	0.207(0.674)	1.120(0.779)	0.245(0.513)	0.488(0.445)
Farm size	–1.139*** (0.310)	0.084(0.254)	–0.251(0.256)	0.115(0.306)	–1.158*** (0.336)	0.345(0.474)	–0.029(0.434)	–0.763(0.750)
Access to extension advice and training	0.303** (0.286)	0.640** (0.252)	0.208** (0.231)	0.249(0.301)	3.963*** (0.684)	–0.137(0.498)	0.087(0.437)	0.789(0.497)
Access to the local market	1.390(0.823)	0.802(0.493)	–0.341(0.319)	1.360(0.401)	0.410(0.330)	0.397(0.358)	0.526(0.296)	0.333(0.337)
Farmer-to-farmer extension	1.648** (0.800)	0.674(0.811)	–0.285(0.573)	–0.126(0.498)	1.500** (0.628)	0.695(0.484)	0.727(0.561)	0.087(0.514)
Constant	–4.417*** (0.678)	–1.158(0.492)	–0.733(0.542)	–2.300*** (0.707)	–4.605*** (0.970)	–4.664*** (1.667)	–6.451*** (0.675)	–5.756*** (1.046)
No. of observations	170	170	170	170	170	170	170	170
Log pseudo-likelihood	–49.680	–90.448	–78.484	–62.843	–26.406	–14.674	–22.781	–11.966
Prob > χ²	0.000***	0.000***	0.000***	0.000***	0.000***	0.000***	0.000***	0.000***
Pseudo R²	.568	.226	0.329	0.463	0.774	0.852	0.805	0.884

Robust standard errors in parentheses. Dependent variable is the decision to adopt a specific climate-smart adaptation strategy.

p < 0.10. **p < 0.05. ***p < 0.01,

Discussion

The study’s findings indicate that significant CSA strategies adopted by farmers in the study area include the use of diverse seed varieties, early maturing paddy varieties, changing crop planning dates, improving irrigation facilities, crop diversification, agroforestry, livelihood diversification and migration. Among these strategies, farmers most frequently adjust crop planting dates and improve irrigation facilities in response to climatic risks. In addition, approximately 18% of total farmers (N = 170) do not use any CSA strategy to mitigate the harshness of changing climatic circumstances.

The significant findings of the binary logit model are that access to extension services and training, the gender and educational level of the household head, access to irrigation, access to credit, and the level of crop damage all influence farmers’ decisions to adopt CSA strategies positively and significantly. However, the farm size of the household had a negative and significant effect on this adaptation decision.

In the face of changing climatic conditions, households with access to extension services and agricultural training can employ CSA measures. This is consistent with the findings of Musafiri et al. (2022) and Trinh et al. (2018). In addition, households led by educated members are more likely to undertake CSA. This is consistent with the findings of Musafiri et al. (2022) and Mogak et al. (2021). The most likely reason is that educated households are aware of the challenges posed by climate change to farming systems and use their knowledge to implement adaptation techniques.

The positive coefficient of the gender variable influencing farmers’ intention to embrace CSA techniques could be explained by the evidence that male-headed households are more abled in resourcing CSA strategies than female-headed households (Mogaka et al., 2021). In addition, the probable explanations for the positive and significant coefficients of access to irrigation or credit and damage level are as follows: Farmers with access to irrigation or credit can withstand shocks to the agricultural sector caused by climate risks. Similarly, farmers that experience significant crop damage are more receptive to CSA practices.

The negative coefficient of farm size in the binary logit model could be explained by the fact that small and marginal farming systems are predominantly subsistence-oriented. As a result of this, marginal and small farmers seek to mitigate agricultural risk through the adoption of CSA techniques. This finding is consistent with Mogaka et al. (2021).

Several important findings from the multivariate probit (MVP) model indicate that households having access to extension advice and training, educational level, land ownership, access to irrigation and credit and farming experience are often more likely to use most CSA measures to mitigate climatic risk to agriculture. Climate and weather information accessibility enables farmers to manage climate risks through CSA strategies. This accords with Musafiri et al. (2022). In addition, the availability and accessibility of irrigation and credit facilities facilitate farmers’ adoption of CSA measures, particularly in rain-fed areas (Trinh et al., 2018).

Conclusion and policy implications

The study sought to highlight CSA options and the factors that influence farmers’ decisions to adopt them. The study identified eight distinct CSA options, including the use of diverse seed varieties, early maturing rice varieties, shifting crop planting dates, increasing irrigation capacity, crop diversification, agroforestry, livelihood diversification and migration. The results of the binary logit model indicate that access to extension services and agricultural training, the educational level and gender of the household head, the land ownership, access to irrigation and credit facilities and crop damage level influenced farmers’ decision to adopt CSA strategies positively and significantly. However, the results of the multivariate probit model indicate that access to extension advice and training, educational level, land ownership, access to irrigation, access to credit and farming experience were significant factors influencing their decisions to adopt some CSA strategies.

The policy implications of the study are as follows: firstly, extension services and training should be expanded to include climate change adaptation training. However, training sessions should be basic and understandable to farmers with less education and expertise. Secondly, the government should implement public policies and initiatives aimed at increasing farmers’ education levels, expanding access to credit and expanding irrigation coverage. Thirdly, the study informs policymakers about the importance of facilitating weather-related information at the firm level, enabling farmers to implement ex-ante adaptation strategies to mitigate agricultural risk associated with weather extremes and to strengthen the overall resilience of the farming communities.

Footnotes

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors are highly thankful to the Institute of Eminence (IoE), Banaras Hindu University, for granting funds for the study. The project’s title is “Impact, vulnerability, and adaptation to climate change: Farm-level evidence in Odisha, India” (IoE-6031).

ORCID iD

Dibakar Sahoo

Notes

Author biographies

Dr. Dibakar Sahoo is an assistant professor at the Department of Economics, Banaras Hindu University. His research areas of interest are agricultural Economics, climate change adaptation and mitigation, and development economics.

Mr. Prasanta Moharaj is a Ph.D scholar at the Department of Sociology, University of Hyderabad. His research areas of interest are Environmental sociology, climate change and agriculture, and sustainable development.

References

Aryal

Sapkota

Khurana

, et al. (2020) Climate change and agriculture in South Asia: adaptation options in smallholder production systems. Environment, Development and Sustainability 22(6): 5045–5075.

Bandara

Cai

(2014) The impact of climate change on food crop productivity, food prices and food security in South Asia. Economic Analysis and Policy 44(4): 451–465.

Belay

Beyene

(2013) Small-scale irrigation and household income linkage: evidence from Deder district, Ethiopia. African Journal of Agricultural Research 8(34): 4441–4451.

Belderbos

Carree

Diederen

, et al. (2004) Heterogeneity in R&D cooperation strategies. International Journal of Industrial Organization 22(8–9): 1237–1263.

Cappellari

Jenkins

(2003) Multivariate probit regression using simulated maximum likelihood. The Stata Journal: Promoting Communications on Statistics and Stata 3(3): 278–294.

Deressa

Hassan

Alemu

, et al. (2008) Analyzing the Determinants of Farmers’ Choice of Adaptation Methods and Perceptions of Climate Change in the Nile Basin of Ethiopia. Washington, DC: International Food Policy Research Institute. Available at: https://www.ifpri.org/publication/analyzing-determinants-farmers-choice-adaptation-methods-and-perceptions-climate-change (Accessed 1 May 2022).

District Census Handbook (2011) Office of Registrar General & Census Commissioner. New Delhi, India: Ministry of Home Affairs, Government of India.

Economic Survey (2017–2018) Ministry of Finance. Government of India. https://mofapp.nic.in/economicsurvey/economicsurvey/index.html

Food and Agriculture Organization (FAO) (2014) Climate-smart Agriculture and Resource Tenure in Sub-Saharan Africa: A Conceptual framework. Rome: FAO.

10.

Fisher

Abate

Lunduka

, et al. (2015) Drought tolerant maize for farmer adaptation to drought in sub-Saharan Africa: determinants of adoption in eastern and southern Africa. Climatic Change 133(2): 283–299.

11.

Gbetibouo

Ringler

(2009) Mapping South African farming Sector Vulnerability to Climate Change and Variability: A Subnational Assessment. Washington, DC: International Food Policy Research Institute (IFPRI) and Centre for Environmental Economics and Policy in Africa (CEEPA). Available at: https://www.ifpri.org/publication/mapping-south-african-farming-sector-vulnerability-climate-change-and-variability (Accessed 1 May 2022).

12.

Greene

(2003) Econometric Analysis. Chennai, India: Pearson Education India.

13.

Huguenin

Pelgrin

Holly

(2009) Estimation of Multivariate Probit Models by Exact Maximum Likelihood. Institute of Health Economics and Management (IEMS) Working Paper. Available at: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.516.9282&rep=rep1&type=pdf (Accessed 1 May 2022).

14.

IPCC (2007) Climate Change 2007: The Physical Science Basis. Contribution of Working Group 1 to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (ed. Solomon, MD Qin, Z Manning , et al.). Cambridge; New York: Cambridge University Press.

15.

Kahsay

Hansen

(2016) The effect of climate change and adaptation policy on agricultural production in Eastern Africa. Ecological Economics 121: 54–64.

16.

Kibue

Liu

Zheng

, et al. (2016) Farmers’ perceptions of climate variability and factors influencing adaptation: evidence from Anhui and Jiangsu, China. Environmental Management 57(5): 976–986.

17.

Kom

Nethengwe

Mpandeli

, et al. (2020) Climate change grounded on empirical evidence as compared with the perceptions of smallholder farmers in Vhembe District, South Africa. Journal of Asian and African Studies 55(5): 683–698.

18.

Kumar

KSK

(2011) Climate sensitivity of Indian agriculture: do spatial effects matter? Cambridge Journal of Regions, Economy and Society 4(2): 221–235.

19.

Kumar

KSK

Parikh

(2001) Indian agriculture and climate sensitivity. Global Environmental Change 11: 147–154.

20.

Kumar

Rajagopalan

Hoerling

Bates

Cane

(2006) Unraveling the mystery of Indian monsoon failure during El Niño. Science, 314(5796): 115–119.

21.

Loria

Bhardwaj

(2016) Farmers’ response and adaptation strategies to climate change in low-hills of Himachal Pradesh in India. Nature Environment & Pollution Technology 15(3): 895–901.

22.

Mendelsohn

(2012) The economics of climate change adaptation in developing countries. Climate Change Economics 3(2): 1–21.

23.

Mengistu

Haji

(2015) Factors affecting the choices of coping strategies for climate extremes: the case of Yabello district, Borana zone, Oromia National Regional State, Ethiopia. Science Research 3(4): 129–136.

24.

Mishra

(2010) Vulnerability and adaptation analysis in flood affected areas of Orissa. Social Change 40(2): 175–193.

25.

Mogaka

Bett

Karanja Ng’ang’a

(2021) Socioeconomic factors influencing the choice of climate-smart soil practices among farmers in western Kenya. Journal of Agriculture and Food Research 5: 100168.

26.

Moharaj

Rout

(2021) Changing climatic conditions and agricultural livelihoods: an impact study in Jagatsinghpur District, Odisha. Social Change 51(2): 160–179.

27.

Musafiri

Kiboi

Macharia

, et al. (2022) Smallholders’ adaptation to climate change in Western Kenya: considering socioeconomic, institutional and biophysical determinants. Environmental Challenges 7: 100489.

28.

Natural Disaster Plan for Odisha (2014) Government of Odisha. https://www.osdma.org/#gsc.tab=0

29.

Nhemachena

Hassan

(2007) Micro-level Analysis of Farmers’ Adaptation to Climate Change in Southern Africa. Washington, DC: International Food Policy Research Institute. Available at: https://www.ifpri.org/publication/micro-level-analysis-farmers’-adaptation-climate-change-southern-africa (Accessed 1 May 2022).

30.

Nigussie

van der Werf

Zhu

, et al. (2018) Evaluation of climate change adaptation alternatives for smallholder farmers in the Upper Blue-Nile Basin. Ecological Economics 151: 142–150.

31.

Odisha Economic Survey (2014–2015) Planning and Convergence Department, Directorate of Economics and Statistics, Government of Odisha. https://pc.odisha.gov.in/sites/default/files/2020-03/Economic_Survey_2014-15.pdf

32.

Odisha Economic Survey (2020–2021) Planning and Convergence Department, Directorate of Economics and Statistics, Government of Odisha. https://pc.odisha.gov.in/sites/default/files/2021-02/Odisha%20Economic%20Survey%202020-21.pdf

33.

Panda

Sharma

Ninan

, et al. (2013) Adaptive capacity contributing to improved agricultural productivity at the household level: empirical findings highlighting the importance of crop insurance. Global Environmental Change 23(4): 782–790.

34.

Pattanayak

Kumar

KSK

(2014) Weather sensitivity of rice yield: evidence from India. Climate Change Economics 5(4): 145001.

35.

Pingali

Aiyar

Abraham

, et al. (2019) Managing climate change risks in food systems. In: P Pingali, A Aiyar, M Abraham, A Rahman (eds) Transforming Food Systems for a Rising India. Cham: Palgrave Macmillan, pp.241–275.

36.

Piya

Maharjan

Joshi

(2013) Determinants of adaptation practices to climate change by Chepang households in the rural Mid-Hills of Nepal. Regional Environmental Change 13(2): 437–447.

37.

Prasad

Pandey

Newaj

, et al. (2014) Risk and vulnerability due to climate change and adaptation initiatives for agricultural resilience in Panna district of Madhya Pradesh, central India. Range Management and Agroforestry 35(1): 157–162.

38.

Raghavendra

Suresh

(2018) Risk management in rainfed agriculture: an analysis of strategies and adaptations followed by farmers in Madhya Pradesh. The Indian Journal of Agricultural Sciences 88(6): 895–901.

39.

Rasul

Saboor

Tiwari

, et al. (2019) Food and nutrition security in the Hindu Kush Himalaya: unique challenges and niche opportunities. In: P Wester, A Mishra, A Mukherji, AB Shrestha (eds) The Hindu Kush Himalaya Assessment. Cham: Springer, pp.301–338.

40.

Sahoo

Sridevi

(2021) Social vulnerability and adaptation to climate change: evidence from vulnerable farmers’ groups in Odisha, India. Agricultural Economics Research Review 34: 347–520.

41.

Sanghi

Mendelsohn

(2008) The impacts of global warming on farmers in Brazil and India. Global Environmental Change 18(4): 655–665.

42.

Sanghi

Mendelsohn

Dinar

(1998) The climate sensitivity of Indian Agriculture. In Dinar

(ed.) Measuring the Impact of Climate Change on Indian agriculture. Washington, DC: The World Bank.

43.

Solomon

Manning

Marquis

, et al. (2007) Climate Change 2007-The Physical Science Basis: Working Group I Contribution to the Fourth Assessment Report of the IPCC. Vol. 4. Cambridge: Cambridge University Press.

44.

Stage

(2010) Economic valuation of climate change adaptation in developing countries. Annals of the New York Academy of Sciences 1185(1): 150–163.

45.

Thornton

Herrero

(2014) Climate change adaptation in mixed crop–livestock systems in developing countries. Global Food Security 3(2): 99–107.

46.

Tocco

Bailey

Davidova

(2013) The Reallocation of Agricultural Labor Across Sectors: An Empirical Strategy for Micro Data. Factor Markets Coordination. Brussels: Centre for European Policy Studies.

47.

Trinh

Rañola

Jr Camacho

, et al. (2018) Determinants of farmers’ adaptation to climate change in agricultural production in the central region of Vietnam. Land Use Policy 70: 224–231.

48.

Vermeulen

Aggarwal

Ainslie

, et al. (2012) Options for support to agriculture and food security under climate change. Environmental Science & Policy 15(1): 136–144.