1 Introduction

Climate change poses a severe threat to the natural and human systems. It has also affected the economic well-being across the globe (IPCC 2018). However, the adverse impacts of changing climate are more common in developing countries due to their low level of adaptive capacities and poor infrastructure (Pachauri et al. 2014). Pakistan has experienced some noticeable climate shocks i.e. floods, droughts, climate-induced pests, and diseases, particularly over the past 40 years. Pakistan has contributed less than one percent greenhouse gas (GHGs) emissions to global emissions. But unfortunately, it is included in the list of the most affected countries due to the adverse impacts of climate change (Kreft et al. 2016). Geographical profile, low adaptive infrastructure, and over-dependence of the population on climate-sensitive agriculture are the key contributor to the high level of climate change vulnerability in the country (Ali et al. 2017).

Climate change through increasing temperature and extreme events is expected to pose a serious threat to the sustainability of the agriculture sector in Pakistan through affecting the yields of cereals and other major cash crops. Likewise, projected changes in the water availability due to changing climate could also largely affect the productivity of irrigated agriculture that is solely dependent on the river system. Moreover, projected changes in climate indicate the significant variations in rainfall patterns and an increase in temperature 2–3 °C predicted by 2050 in Pakistan (Imran et al. 2018). Anser et al. (2020) stated the increase in rice–wheat crops yield in Punjab ranged from 14 to 15.9%, 13.6% to 14.5%, using the Agricultural Production Systems Simulator (APSIM) for RCP 4.5 and RCP 8.5, but the losses were ranged from 28.4 to 32.2%, 30.6 to 33.6%, respectively, due to changing climate. A similar pattern was also observed using the Decision Support System for Agrotechnology Transfer (DSSAT). Studies such as (Arshad et al. 2018; Amin et al. 2018) have shown that yields of the major crops have been significantly affected due to the changes in rainfall pattern, a rise in temperature, frequent occurrences of floods, and droughts in Pakistan from 1995 to 2017.

Given the increasing vulnerability of agricultural systems to climate change, adaptation and adjustments in current agricultural practices are required to reduce the adverse effects of changing climate. For instance, Food Agriculture Organization (FAO) has introduced the concept of climate-smart agriculture (CSA) under which innovative and sustainable agricultural practices are proposed to adopt at the farm level (FAO 2013). Compared to conventional agriculture system, CSA is a set of practices to adopt that increases the resilience of current farming systems and improves the resource use efficiency at the farm level along with reducing GHGs emissions. CSA emphasizes on the existing knowledge of sustainable agricultural development to identify viable options and necessary enabling activities. This approach utilizes ecosystem services to support productivity, adaptation, and mitigation. It provides tools for assessing different practices and technologies in relation to their effects on food security objectives under the site-specific effects of climate change. Examples include practices such as improved pest management, landscape approaches, water, and nutrient management, integrating trees into agricultural systems, reduced or zero tillage and use of improved crop varieties (Lipper et al. 2014; Challinor et al. 2014; Lunduka et al. 2019). Literature such as (Brandt et al. 2017) suggested that adoption of CSA measures consists of socioeconomic characteristics, institutional support, and financial resources such as resource endowments, access to extension services, and information for the use of technologies, that explain the differences for adoption in agriculture. This is particularly true when the CSA practices adopted for sustainable agriculture.

A growing body of literature highlights the importance of the adoption of CSA practices across the globe. It stresses the need for a gradual shift from a technology-oriented approach to a more systems-oriented approach to understanding the complexity of farming systems. It may include political dimensions, market infrastructure, other institutional aspects, and the interactions between public government and private sectors that shape the context in which farming takes place (Totin et al. 2018; Kaczan et al. 2013; Thornton et al. 2018). Similarly, some studies also show that applications of CSA practices not only raise the resource use efficiency, crop productivity and farm income but also reduce the GHS’s emissions (Dinesh et al. 2015; Ugochukwu and Phillips 2018; Rosenstock et al. 2019)

Agriculture is the key sector of the economy of Pakistan that provides livelihood to two-third of the population with 42% employment of the total labour force. Further, it contributes around a 19% share to the gross domestic product (GDP). On the one hand, over the past 2 decades, the agriculture sector is not performing well as the agricultural growth remains at low and agricultural yields are declining. On the other hand, the agriculture sector faces serious challenges of meeting the rising food demands for a rapidly growing population at a rate of around 2% (Sardar et al. 2016). Moreover, most of the farmers in Pakistan still engaged in conventional agricultural practices. Therefore, low performance per acre yield is attributed to the lack of skills and innovations, market failures, climatic shocks, and resource constraints. According to the literature, e.g. Imran et al. (2018, 2019), the main reason for the low adoption rate of implementation of CSA practices and technologies particularly for small landholding and marginalized farmers is the limited access to the financial resources and institutional services (e.g., easy access to the credit market and agricultural extension services, advisory services on climate change, availability of resources).

However, the literature on the adoption of CSA in Pakistan is limited and still growing. Therefore, more evidence is required to justify the importance of CSA practices enhancing the farmers’ adaptive capacity to mitigate the adverse impact of changing climate. Against the above background, this study is designed to assess the adoption of CSA practices, and the intensity (either applied single or full package of CSA practices), its determinants, and conditional impact on crop yield and farm income.

The rest of the paper is organized as follows. In Sect. 2, we develop the conceptual framework of the study. Section 3 outlines the study area, survey design, and methodology followed to answer the research questions of the study. Section 4 presents the results and discussions, while Sect. 5 describes the conclusion and implications of the study findings.

2 Conceptual framework

The conceptual framework of this study consists of three main components. First is climate change and its vulnerability. Second is the adaptation process based on CSA practices, while the third is the farmers’ income. The application of this conceptual framework is based on a top-down approach (see Fig. 1). Straight lines with the positive and negative signs are used to identify its positive and negative impacts, respectively, that one component has contributed to the other components of the framework. The climate vulnerability is included various factors such as natural and social indicators that focus on the sensitivity and the adaptive capacity of the farmers. A farm household is vulnerable if his livelihood is more sensitive to the changes in climate-related risks and at the same time he has limited adaptive capacity due to poor assets endowment. So, climate vulnerability can be reduced if farmers adopt timely and effectively to climate change. Farmers’ intentions for the adoption of CSA practices depend on understanding and accurate knowledge of climate change and vulnerability. Socioeconomic and institutional determinants such as knowledge, skills, experience, and institutional support may affect the farmers’ understanding and perceptions about weather variability. Higher farmers’ endowment enables them more to adopt because the adaptation is required resources and affordability. In this study, we consider a rational farm household that adopts CSA measures to mitigate the agricultural losses from climate change. The adoption benefits of CSA measures may have direct implications for improving the farmers’ income and reduced the adverse impacts of weather variability by timely management of their crops accordingly. When a farmer decided to adopt CSA practices, it not only reduces the climatic risk but also improves the crop productivity and farm income by increasing the supply of the crops. So, the adoption of CSA practices works for double-sided as shown in Fig. 1. In contrast, if farmers prefer no-adoption pathway or to adopt conventional farming than CSA, so it may have increased the risks by rising the climate change vulnerability and may have adverse impacts on the farm income through reducing per hectare (ha) crop yields. We treated three major crops in our study. So adverse impacts of climate change on crop productivity may have direct implications for farmer’s income due to limiting the crop yields. However, farmers can minimize climate-induced damages by the timely adoption of CSA practices. It might be possible that the adaptation may not help to reduce the potential impact at the farm level because of climatic shocks, but it might have a positive impact on farm income due to improved crop productivity. Therefore, the adoption of CSA practices is better than no-adaptation. The application of this approach started with climate change and its impacts at the farm level and ended with the implications of CSA practices on crop productivity and farmers’ income.

Fig. 1
figure 1

Conceptual framework of the study presenting the relation among climate change, its impacts, CSA practices and farmers’ income

Full size image

3 Materials and methods

3.1 Study area

This study was conducted in Punjab province because Punjab is the most populated and the second largest province in terms of area. It contributes to 74% of the total agricultural production and 56% of the total cultivated land area. Punjab province is categorized into four agro-ecological zones (AEZs) based on different cropping patterns, distinct geography, and climate variability by the Pakistan Agricultural Research Council (PARC) (Adnan et al. 2017). This study has focused on three districts from three AEZs and excluded the Thal region due to budget constraints. All three selected AEZs have different attributes of environment, geography, and socio-economic conditions. Our first study district, Jhang, is located in between Jhelum and Chenab rivers. It falls in central mixed cropping sub-zone and lies in the irrigated plains AEZs. The main crops are wheat, rice, and cotton (GOPP 2018a). The second study district is Rahim Yar Khan. It partly falls in the alluvium plain, and some part lies in the irrigated plains AEZ (Cotton subzone and Cholistan sub-zone). The major crops are wheat, cotton, and sugarcane in this district (GOPP 2018c). The third district, Sialkot, partly located in the Barani (rain-fed) AEZ and some part into irrigated plains AEZ. In Sialkot, wheat and rice are the dominant crops (GOPP 2018b).

Further, we focused on the three major crops that had cultivated in combinations of rice–wheat and cotton–wheat crops annually, for two main reasons. First, these are the primary source of food that contribute about 70% of the daily per-capita caloric intake in Pakistan (Khalid et al. 2019). Second, it accounts for, as described earlier, about 74% of the total agricultural production and 56% to the total cultivated land area. This wide-ranging cultivation allows us to study rice–wheat and cotton–wheat growing farmers across different regions to investigate the choice of CSA practices and its associated benefits at the crop yield and on the farm income in Punjab.

3.2 Sampling and data collection

This study used the primary dataset to examine the research objectives of the study. The survey data were collected in three districts between August and September 2018 based on AEZs prepared by the PARC. We interviewed 420 farmers, about 140 farmers from each district of Punjab province using a multi-stage sampling technique. But the sample size was reduced to 417 farm households in this study due to some missing information. This multi-stage sampling procedure consists of the following stages. First, we selected the Punjab province as the main study area. Second, we included three AEZs out of four AEZs in our study based on diverse characteristics of climate, geography, and cropping patterns. Third, we randomly selected three districts from the three AEZs such as Jhang (irrigated plains), Rahim Yar Khan (partly lies in the irrigated plain and some part falls in marginal land) and Sialkot (Barani zone i.e. rain-fed area). Fourth, we randomly choose two sub-districts (tehsils) from each district. Fifth, we randomly selected 5–6 villages from each tehsil. Six, in the last stage, we randomly selected 10–12 farmers from each village based on the list of farmers collected from the agriculture department.

Before starting the interview, informal verbal consent was taken from the farmers. Farmers who were not willing to give the interviews were replaced by other farmers via a random process. A pre-tested structured questionnaire from all the consenting farmers was used for collecting information about households and socio-economic characteristics, CSA practices, the institutional support for adopting CSA technologies, and the perceptions of climate change. Prior to the beginning of the survey, off-field training and in-field training were given to the enumerators about the data collection methods and for the structured questionnaire to improve the quality of the survey. The enumerators were graduated and well trained for collecting the survey data about the objectives of the study. In total, we collected data from 35 villages across the three districts. As the climate of Punjab province witnessed both hot summers and cold winters. Therefore, two growing seasons were considered. Winter (Rabi) season and summer (Kharif) season comprised of November–April and May–October (PMD 2018).

3.3 Analytical framework

3.3.1 Econometric framework and estimation strategy

In this study, we used a micro-econometric structural Ricardian model in which farmers are assumed to use a combination of CSA practices to maximize their expected profits. The ith farmer adopted a combination of CSA practices J, (see Table 1) and earned his expected profit, \(\pi_{ij}^{*}\). Following the literature such as (Thierfelder et al. 2017; Nyasimi et al. 2017; Rosenstock et al. 2016), the decision to the adoption of CSA practices J, where j (j = 1,2,…j) is the vector of a latent variable that is determined from the household indicators, socio-economic characteristics, institutional support, financial conditions (\(X_{i}\)) and unobservable characteristics \((\mu_{ij} )\). Thus, net benefits can be expressed as:

$$\pi_{ij}^{*} = X_{i} \beta + \mu_{ij}$$
(1)
Table 1 Expected relationship of CSA practices and the outcome
Full size table

Following the standard theory of technology adoption, we assume that a farm household wants to maximize the expected utility function by choosing a mix of CSA practices for crop production (Di Falco et al. 2012; Teklewold et al. 2017). Here, the utility function is assumed as state independent. So, representing a farm household will solve this problem by choosing an optimal mix of CSA practices. Let \(A_{\text{h}}\) be an index that denotes the choice of combination of CSA practices at the farm household level, such as

$$A_{\text{h}} = {\text{Prob}} \left( {{\text{CSA}}\;{\text{practices}}} \right) = j\;{\text{iff}}\;\pi_{ij}^{*} > \mathop {(\hbox{max} }\limits_{m \ne j} \pi_{\text{im }}^{*} ) \;{\text{or}} \;\varphi_{ij} < 0\quad {\text{for}}\;\;{\text{all}}\;\;m \ne j$$
(2)

where \(\varphi_{ij} = \mathop {\hbox{max} }\limits_{m \ne j} ( \pi_{\text{im }}^{*} - \pi_{ij }^{*} ) < 0\) (Abidoye et al. 2017).

Equation (2) shows that ith farm household will choose a combination of CSA practice 1 over CSA practice j if and only if the chosen CSA practice provides greater expected utility than any other sets of CSA practices. CSA adaptation benefits may include reduced climatic vulnerability and improved crop productivity and farm income. Flowing the study (Brandt et al. 2017), we described the factors that influence the individual’s behaviour for the adoption of multiple CSA practices by employing a multinomial logistic (MNL) regression framework. This model classifies as discrete choice models and has been used in the various studies of econometrics literature such as (Thornton et al. 2018; Lipper et al. 2014). Specifically, this MNL model is employed for modelling the choice behaviour. In our study, the model perceived the outcomes in terms of socio-economic conditions, individual characteristics, and institutional support prospective. The probability for adopting ith farmer for choosing CSA practices j with characteristics \((X_{i} )\) can be stated in the multinomial logistic model (MNL) as follows:

$$P_{ij} = {\text{Pro}} \left( {\varphi_{ij} < 0{\mid }X_{i} } \right) = \frac{{\exp \left( {X_{i} \beta_{j} } \right)}}{{\mathop \sum \nolimits_{m = 1}^{j} \exp \left( {X_{i} \beta_{m} } \right)}}$$
(3)

where \(\beta_{j}\) is a vector for each farm household having choices j = 1,2,…j. In our specification, the outcome of \({\text{CSA}}\;{\text{practices}}_{j}\) consists of base category, ‘no-adoption’ i.e. j = 1 and in the remaining combinations, we applied at least one CSA practice i.e. j = 2,…, 6. A concatenation of all j categories makes the variable of adopting CSA practices for the ith farmer.

In the second stage, we estimated the impact of adopting multiple CSA practices on crop yield and farm income. The empirical approach to estimate the relationship abstracts from the conditional Ricardian framework. The outcome variable is conditional, as the farm household applied specific CSA practice j either in isolation or in combination if the expected difference of utility for adoption versus non-adoption is positive. For addressing these issues, we employed the conditional Ricardian model in our study as it is adapted in the literature such as (Kurukulasuriya et al. 2011; Seo and Mendelsohn 2008) for credible estimates. Therefore, the conditional Ricardian specification for the observed outcome variable in terms of the multiple adopting CSA practices applied by the ith farm household can be summarised as:

$${\text{Outcome}}\;j:\;Y_{ij} = W_{i} \alpha_{j} + \varepsilon_{ij} \quad {\text{if}}\;A_{h} = j,$$
(4)

where \(Y_{ij}\) shows the potential outcome variables (such as crop yield and farm income) of the ith farm household if and only, in response to the adoption of CSA practices j is applied. Let we assume that error terms \(\varepsilon_{ij}\) are distributed with \(E\left( {\varepsilon_{ij} {\mid } X, W} \right) = 0\) and \(\text{var} \left( {\varepsilon_{ij} X, W} \right) = \sigma_{j}^{2}\). As differences in adopters of CSA practices in isolation or in combination and between the non-adopters may result in self-selection bias. So, it may produce a correlation between error terms \(\mu_{ij}\) and \(\varepsilon_{ij}\), and it can generate inconsistent estimates. Therefore, we added selection bias-corrected term \(\left( {\sigma _{{j\mu }} \acute{\eta} _{{ij}} } \right)\) in Eq. (5) following the study (Bourguignon et al. 2007) which is given as:

$${\text{Outcome}}\;j :\;Y_{ij} = W_{i} \alpha_{j} + \sigma_{j\mu } \acute{\eta}_{ij} + \psi_{ij} \quad {\text{if}}\;A_{\text{h}} = j,$$
(5)

where \(\acute{\eta}_{ij}\) is the inverse Mills ratio. We computed it from Eq. (4) by using formula

$$\acute{\eta}_{j} = \mathop \sum \limits_{m \ne j}^{j} \lambda_{j} \left[ {\frac{{\hat{p}_{\text{im}} \ln \left( {\hat{p}_{\text{im}} } \right)}}{{1 - \hat{p}_{\text{im}} }} + \ln \left( {\hat{p}_{ij} } \right)} \right]$$
(6)

while \(\lambda_{j}\) is the correlation coefficient which implies \(E( \mu_{ij} , \varepsilon_{ij} , \psi_{ij} ) = 0\).

Further, there may be another issue for the possible correlation between the plot-invariant unobserved heterogeneity and observed covariates (\(X_{i}\)) at the farm household level. To overcome this problem, we followed the study (Mundlak 1978). We included instrumental variables (e.g. plot disturbance indexFootnote 1 and the average of farm household characteristicsFootnote 2) in Eq. (5) as additional covariates to account for unobserved heterogeneities in the model.

3.3.2 Estimation of adoption impacts

To examine the unbiased average effects of adopting CSA practices in terms of productivity and income gains, different methods are proposed and used in the literature including propensity score matching, switching regression, and difference in difference approaches. For instance, numerous studies such as (Ho et al. 2011; Makate et al. 2019) have used a two-stage least square (2SLS) estimation technique to control the endogeneity problem and to estimate the adaptation impact on crop yield and farm income. This method is normally preferred in the studies when the investigators have no control over the adoption of different combinations of practices, and the outcome is conditionally dependent on these assignments of the practices. Following the study (Carter and Milon 2005), the average treatment effect is calculated by comparing the expected crop yield and farm income of the adopters and the nonadopters, i.e. counterfactual outcome, that we computed from Eq. (7) and (8), respectively.

Farmers with the decision to adopt CSA practices:

$$E( Y_{ij} {\mid }A_{\text{h}} = j) = W_{i} \alpha_{j} + \sigma_{j}$$
(7)

Farmers decided not to adopt (counterfactual):

$$E\left( { Y_{i1} {\mid }A_{\text{h}} = j} \right) = W_{i} \alpha_{1} + \sigma_{1} \acute{\eta}_{ij} .$$
(8)

The average treatment effect on the farmers for adopting CSA practices j = 1 with j = 2,….j is defined as the difference between Eqs. (7) and (8).

The average treatment effect on treated (ATT):

$${\text{ATT}} = E\left( {Y_{ij} {\mid } A_{\text{h}} = j} \right) - E\left( {Y_{i1} A_{\text{h}} = j} \right)\left[ {\alpha_{j} - \alpha_{1} } \right] + \acute{\eta}_{ij} \left[ {\sigma_{j} - \sigma_{1} } \right]$$
(9)

We estimated Eq. (9) with two-stage least square to give unbiased parameters after controlling the selection bias and endogeneity problem from randomly chosen farm households who adopted CSA practices in isolation or in combination to mitigate the adverse impact of changing climate. The expected relationship of CSA practices with the outcome is given in Table 1.

4 Results and discussions

4.1 Descriptive statistics

Table 2 outlines the summary statistics and a description of the variables used in the study. The differences in the farm households’ characteristics of adopters and non-adopters indicate the importance of how to understand the determinants of CSA practices in the context of the local adaptation to climate change. For convenience, we calculated the statistically significant differences in the mean values of the adopted and non-adopted farmers’ characteristics. Table 2 shows that adopters had more education and better financial resources as compared to non-adopters. Literature shows that more educated and experienced farmers have a better knowledge of ongoing changes in the climate than less educated and less experienced farmers. Therefore, they adopted more. Likewise, farmers who had larger farmland areas and more access to the institutional services, better market information, and more knowledge about the weather condition may act as positive determinants for the adoption of CSA practices. Table 2 shows that specifically, these characteristics were found more likely in adopters than non-adopters. This implies that less education, limited institutional access, and less knowledge about the changing climate are the main hurdles at the farm level that may restrict the farmers from the adoption of CSA practices. These findings are in line with studies, e.g. Mwongera et al. (2017) and Campbell et al. (2014).

Table 2 Descriptive statistics of sample farmers used in the study
Full size table

From Table 2, it confirms that farmers who adopted CSA practices were intended to earn more in term of crop yields and farm income. The average yield was reported by the rice–wheat adopters about 8449 kg per ha compared to nonadopters’ yield of 7319 kg per ha in study districts. Likewise, the cotton–wheat productivity of adopters was about 1136 kg per ha higher compared to non-adopted farmers. The average farm income of rice–wheat adopters was more US$298 per ha than non-adopters. Similarly, in cotton–wheat adopters were earned more US$488 per ha as compared to nonadopters.

4.2 Adoption of climate-smart agriculture (CSA) practices

Study findings show that overall, about half of the farmers adopted one or more combinations of CSA practices to cope with climate change and weather shocks (see Fig. 2). The major adoption of CSA practices includes changing cropping dates, zero or minimum tillage, water management measures, improved crop varieties, and nutrient management options.

Fig. 2
figure 2

Adopters and non-adopters (%) in Punjab province, Pakistan

Full size image

The results depict the variations in the adoption of CSA practices across the study districts (see Fig. 3). It may be pointing that these variations were due to the differences in the climatic conditions and socioeconomic profiles in the region. Modifying cropping dates measure was predominantly adopted by farmers in Rahim Yar Khan where about half of the farmers reported changing sowing dates due to changes in climate. The same pattern is followed by farmers in Jhang (39%) and Sialkot (38%) districts. Farmers adopted this CSA practice may be due to rising the temperature at the sowing period and water availability in regions, which could ultimately affect the sowing dates in irrigation regions particularly Rahim Yar Khan and Jhang districts. The second most adopted CSA practice was zero or minimum tillage which is mainly adopted by farmers in Jhang (34%) followed by farmers in Sialkot (22%) and Rahim Yar Khan (20%). It is an essential CSA practice for improving soil fertility and conserving soil moisture through fixing of biological nutrients. However, due to the lack of information and awareness, this adoption was very limited in irrigated areas of Punjab. Similarly, nutrient management measure includes soil management practices that are used by maximum 28% of the farmers in Sialkot and this adoption was further dropped 16% to 14% by the farmers in Jhang and Rahim Yar Khan, respectively. Water management measure includes water conservation technologies that were mainly adopted by farmers in Sialkot followed by Rahim Yar Khan and Jhang districts. The average adoption rate of water management was reported only about 21%, which was very low, particularly in the context of the water challenges currently faced by the Punjab province. One of the reasons for this low adoption of water management practice may be the lack of knowledge about the water management options and its use at the farm level. Improved crop varieties measure was absorbed the least adopted CSA practice adopted by the farmers in all study districts. Low adoption of this CSA practice shows the scope for major intervention is required at the farm level where farmers need to be educated about the importance of improving crop varieties to manage upcoming challenges of climate change and to meet growing food demands in the country.

Fig. 3
figure 3

CSA practices adopted by the farm household (%) in all three study districts

Full size image

Further, we tried to figure out the intensity of adopting CSA practices at the farm level and found that most of the farmers adopted either one or two measures, whereas farmers adopting more than two measures were very few (Table 3).

Table 3 CSA intensity adopted by farmers (%) in the three districts of Punjab province, Pakistan
Full size table

Afterward, we compared the crop yields and farm income in relation to the intensity of CSA practices adopted. It also shows the same results as described earlier. For instance, similar trends for adopters and non-adopters are also observed in Fig. 4.

Fig. 4
figure 4

CSA intensity impact on crop yield and farm income

Full size image

4.3 Determinants of CSA practices adoption

In this section, we examined the determinants that influence the farmers’ decision for the adoption of CSA practices and the intensity for mitigating the impact of climate change. Results from multinomial logistic regression (MNL) are shown in Table 4. MNL regressions were estimated for the entire study sample who adopted CSA practices in isolation or in combination conditional to the base category. The dependent variable consists of six options including ‘no adoption’ as a base category. Other categories vary from adopting at least one CSA practice to a maximum of five measures. Before going to the detailed analysis, one of the important assumptions for MNL regression i.e. ‘independence of irrelevant alternatives’ was tested. The Hausman specification test confirmed with insignificant p values at 5% level that coefficients of MNL are independent of additional alternatives. Results of coefficients and marginal effects are shown in Tables 4 and 5, respectively. Here it is necessary to mention that the results presented in Table 4 are the coefficients in odd ratios and only describe the direction of the relationship between dependent and independent variables, whereas the results in Table 5 show the marginal effects of regressors on dependent variables and are discussed in detail. The findings show that the adoption and the intensity of CSA practices were the interdependence between the complementarity and substitutability of the adoption decisions of the farmers. Overall, the adoption of CSA practices and the intensity determinants findings are consistent with the literature Kpadonou et al. (2017) and Sardar et al. (2019).

Table 4 Multinomial logit regression estimates: determinants of single to multiple adoption of climate-smart agriculture (CSA) practices in the study area
Full size table
Table 5 Marginal probability effects for single to multiple adoption of climate-smart agriculture (CSA) practices in the study area
Full size table

Findings confirmed that the adoption of CSA practices and the intensity depend on the farmers’ socioeconomic characteristics, institutional support to the farmers and the farm level vulnerability to the changing climate. Table 5 shows that the farm household characteristics and institutional factors were the major determinants that enabled the farmers to take a decision for the adoption of CSA practices i.e. the farmer who decided to adopt at least one CSA measure. The significant and positive coefficient signs of farm household characteristics such as farming experience (Exp), education (Edu), and land size (Area) signify the importance of education, skills, knowledge, and the information to mitigate the impact of climate vulnerability. Likewise, institutional factors such as weather information (Weather_info) and subsidies on CSA technologies (Sub_tech) were also the main determinants that encouraged the farmers to adopt.

Results outline that the farmers’ adoption of CSA intensity varies according to the socio-economic and institutional level characteristics that farmers contain. Results show that farmers’ characteristics such as household size (HH size) are negatively associated with the adoption of at least one CSA practice or more packages of CSA practices. The negative association of HH size may imply that the probability of adopting any CSA practice is likely to decrease with increasing the family members to support due to low per capita income, because low income reduces the opportunities for investing in the adoption of CSA practices.

The positive association of farming experience (Exp) turns to be negative when farmers adopted at least four combinations or a full package of CSA practices. The result implies that the probability of adopting CSA practices decreased with increasing farmer’s experience may be due to risk aversion of CSA practices likely to adopt by older farmers, while education (Edu) is positive and significantly associated with the different packages for the adoption of CSA practices. It may be pointed out that knowledge regarding the importance of adopting CSA practices increases the demands for adaptation measures. Therefore, higher education and more knowledge are likely to help for understanding the importance and implementing the sets of CSA practices for getting the best pair out of them. Moreover, land size (Area) shows a positive association with the adoption of different sets of CSA practices. Evidence suggested that if farmers have a larger land cultivated area, it intensifies to invest more in the use of CSA measures. The large landholding farmers have more financial and institutional access that enables them to adopt more flexible CSA practices than small landholding farmers and earn extra returns on the adoption of CSA practices with larger farm size due to economies of scale.

The institutional indicators such as access to extension institutions (Ext_acc), credit services (Credit_acc), weather forecasting information (Weather_info), market information (Market_info), and subsidies on CSA technologies (Sub_tech) are positively and significantly correlated with the adoption of different sets of CSA practices. Evidence indicated that institutional services could play their constructive role by providing training, workshops, and awareness for the use of CSA measures, because institutions are the important instruments that provide information to the farmers to deal with the low fertility, dryland soils, and about soil conservation. It also enhances the knowledge about the rainwater conservation for optimizing use to cope with water scarcity in the region. The study suggested that structural measures should be taken to improve the poor farmers’ resource base specifically for those farmers who were the least adopters, because higher endowment can enable them to upgrade the scale of adoption rate and the intensity of CSA measure. Therefore, knowledge, skills, and awareness towards intensive use of CSA practices are very important particularly to shape farmers’ decisions for adopting the intensity of CSA measures. The study findings confirm that the farmers who adopted all five CSA measures. They were well endowed in productive resources than the farmers who adopted only one measure. It may be suggested that the financial capacity of the farmers such as credit access and extension services are very important drivers for the farmers’ choices to determine the joint and complementary adoption of CSA measures. These results are concurrence in the line of the literature for example Totin et al. (2018) and Sain et al. (2017).

The results find that financial resources and institutional support chiefly matters for the adoption and the intensity of CSA measures. Significant and positive association and increasing probability coefficient value with adopting different sets of CSA practices confirm the importance of propensity to adopt them. This is consistent with existing literature, e.g. (Makate et al. 2019), which illustrated the importance of institutional support and financial resources for adaptation measures to mitigate the adverse impact of climate change. Lastly, plot-level disturbance index (Climatic_shocks) is associated positively and significantly to the adoption of a different combination of CSA practices. This positive correlation may imply that farmers have learned from their experiences regarding climate change. This indicates that farmers tried to minimize the impact of climate change-related hazards by improving the adaptation measures in the context of ongoing exposure (Sardar et al. 2020).

4.4 Conditional effects of adopting CSA practices on crop productivity and farm income

Table 6 shows the results of two-stage least square (2SLS) estimations to examine the impact of the adoption of CSA practices and the intensity on crop productivity and farm income levels. Here, we compared the impacts of different sets of CSA practices adoption with base category i.e. ‘no adoption’ in all cases. Generally, results reveal that the adoption of CSA practices and the intensity tends to a positive and significant impact on crop productivity and farm income. These findings are generally consistent with the studies (Fentie and Beyene 2019; Pagliacci et al. 2019; Mwongera et al. 2019). Table 6 shows that adopting at least two CSA practices has a positive and significant impact to increase in the crop yield and farm income with the coefficients of 0.244, 0.354 for cotton–wheat crops, and 0.264, 0.376 for rice–wheat crops, respectively. Likewise, when farmers adopted a full package of CSA practices, the absolute positive and significant differential impact was observed for the crop yield and farm income levels with the coefficients of 0.281, 0.378 for cotton–wheat crops, and 0.368, 0.402 for rice–wheat crops, respectively. The coefficient values illustrate that the farmers who adopted two to all five measures of CSA produced 0.037, 0.104 points more cotton–wheat, and rice–wheat yields, respectively, than non-adopters. Further, adoption of CSA measures and the intensity generate 0.024, 0.026 points more crop income when farmers adopted from any two measures to all five measures. Mainly, the positive difference between the adoption of any two to all five CSA practices confirms that the higher intensity of CSA adoption was positively correlated with higher productivity and more farm income. Here, crop yield and farm income were in log-transformed. To see the impact in percentage form, we transformedFootnote 3 crop yield and farm income coefficients into a percentage.

Table 6 Conditional marginal effects of adopting CSA practices on crop yield and farm income
Full size table

Results outline that the impact of adopting a full set of CSA practices on crop yield is significantly higher 32% and 44% kg/ha for adopters compared to non-adopters for cotton–wheat and rice–wheat crops, respectively. Similarly, adopters of a full package of CSA practices had also earned higher farm income 45% and 48% US$ per ha than non-adopted farmers for cotton–wheat and rice–wheat crops, respectively. It is empirically confirmed that the adoption of CSA had a higher impact on crop productivity and farm income of the farmers. Also, the higher productivity and crop income may lead to an increase in some extra income to the total income of the farmers which could indirectly increase the farmers’ standard of living and also improve the local food security (Abate et al. 2014; Lobell et al. 2014). Overall, we can conclude that the adoption of CSA measures and the intensity have a significant positive impact on crop productivity and farm income levels. Evidence suggested that the farmers who decided to adopt and adopted any one or all five practices achieved higher crop yields and farm income than the farmers who did not adopt or adopted fewer CSA measures. It may be pointed out that farm-level CSA adaptation provides substantial benefits to the farm households and the rural society through increasing the income level and by improving the supply of the crops. The study suggested that government and non-governmental institutions can play their constructive role in addressing the farmers’ resource constraints through improving the adaptive capacity and the awareness regarding adaptation measures. Moreover, agricultural policies should be designed in the regional context, and focus should be given to the on-the-ground research. All the suggested measures may lead to improve the farmer’s frequency for the adoption of CSA practices and the intensity to cope with climate change. It might be enabling the farmers to rise in crop income and to reduce the poverty also.

For the reliability of the analyses, different diagnostic tests were used to ensure the adequacy of the results. Table 7 demonstrates the diagnostic indicators of the model. As this model contains the endogeneity problem, therefore instruments were used to solve the endogeneity problem. The insignificant p value of Hansen J statistic confirms the validity of all instruments used in the study to eliminate the endogeneity issue.

Table 7 Results of diagnostic tests of adopting CSA practices’ impact on crop yield and farm income models
Full size table

Further, significant coefficient values of inverse mills ratio (IMR) show the elimination of selection bias misspecification in the model.

5 Conclusion and policy implications

Adoption of CSA practices and the intensity (application of single to a full package of CSA practices), determinants and their implementation to evaluate its differential impact on the crop yield and farm income were assessed in this study. The key five CSA practices were identified. Adopting to improved crop varieties was the least adopted measure reported in the study area. Obtained results revealed that farmers who implemented single to a full package of CSA practices achieved satisfactory yields and farm income in the less favoured agroecological areas of Punjab. However, impacts were not absorbed uniform across the different implications of CSA intensity adoption in terms of gain in crop yields and farm income. But higher resource base endowed farmers were enabled more to adopt and gained higher benefits than the rest of the farmers. Findings from this study suggested that effective and upscaling adoption-rate of CSA practices may demand extra financial resources and institutional support to the farmers. Additionally, subsidies should be provided on the CSA supporting technologies and should ensure the availability of certified quality seeds on the priority basis in the area where adoption rates are very few or where the farmers are still following conventional agricultural practices. The major contribution of the present study was to provide a finer knowledge to design the effective policy in the local context addressing determinants of CSA practices according to the needs of the vulnerable farmers. Results from this study are expected to help the policymakers in taking necessary measures in adapting the agricultural sector of Pakistan to the changing climate. The implications of this study can be made in any other region for a similar study of CSA adaptation to improve resource use efficiency, crop yields, and farm income.

Moreover, this study has some limitations. As we have used cross-sectional farm level household data, we could not be able to capture the dynamics of adaptation of CSA practices over time. Also, we limited our study to a maximum of five CSA adaptation practices options in the local context with two combinations of yearly cultivated cotton–wheat crops and rice–wheat crops at the farm level.