Commercial cash crop production and households’ economic welfare: Evidence from the pulse farmers in rural China

MA Ji-liang ,LI Fan ,ZHANG Hui-jie ,Khan NAWAB

1 Institute of Agricultural Economics and Development,Chinese Academy of Agricultural Sciences,Beijing 100081,P.R.China

2 College of Economics and Management,Huazhong Agricultural University,Wuhan 430070,P.R.China

3 Agricultural Information Institute,Chinese Academy of Agricultural Sciences,Beijing 100081,P.R.China

4 College of Management,Sichuan Agricultural University,Chengdu 611130,P.R.China

5 Macro Agriculture Research Institute,Huazhong Agricultural University,Wuhan 430070,P.R.China

Abstract Whether promoting cash crop production can increase household welfare has long been the focus of the food policy debate. This study first investigated the determinants of household behavior in commercial pulse farming. It then examined how households’ commercial pulse production improves their economic welfare. We used a dataset of 848 households collected from 2018 to 2019 to estimate the determinants of household behavior in commercial pulse farming by the Heckman two-step model. The endogenous treatment regression (ETR) method was employed to examine the impact of commercial pulse farming on household economic welfare. The results showed that factors such as market purchase prices,agricultural technology services,farmers’ access to loans,and government subsidies promoted smallholders’ commercial pulse farming;production costs and perceptions of climate change risks constrained smallholders’ commercial pulse production. Overall,commercial pulse production has increased household farm income but there was a limited impact on household off-farm income. Our findings suggest that policies aiming to increase households’ cash crop production and market access could significantly improve the economic welfare of pulse farmers.

Keywords: pulse farming,cash crop,commercialization,economic welfare,China

1.Introduction

Whether promoting cash crop production can improve household welfare has been central to the food policy debate in many developing countries (Weberet al.1988;Tankari 2017). Several studies have shown that cash crop production can effectively improve households’economic welfare (e.g.,Huanget al.2016). For instance,Christiaensenet al.(2006) studied the coffee farmers in Tanzania and found that when there was a presence of health and drought shocks,coffee farmers,compared with other staple crops farmers,were still economically resilient,indicating a positive effect on farmer’s economic welfare.Similar studies were also found in other developing countries. For instance,Kennedyet al.(1992) found that participation in cash crop schemes increased household income in six African and Southeast Asia countries(Gambia,Guatemala,Kenya,Malawi,the Philippines,and Rwanda). Later,Finnis (2006) showed that farmers who were planting cash crops received much higher economic and social benefits in south India. In Malawi,households selected to grow cash crops also exhibited a significantly higher income than those that did not farm cash crops(Masanjala 2006).

Several acknowledged channels promoting cash crop production could improve households’ economic welfare. First,cash crop production can effectively increase households’ agricultural income (Cuong 2009).Specialization in cash crop production (vs.staple crops)often brings a higher economic return per unit of land they had devoted,including land,water,technologies,and to some extent,labor input. Second,promoting cash crop production contributes to households’ livelihood diversification,which further improves households’ resilience to both economic shocks (e.g.,market price shocks) and other climate-related shocks (e.g.,drought,extreme heat,and low temperatures). For instance,several studies found that crop diversification,such as intercropping and crop rotations,can improve the resilience of the households’agricultural production (Barbieriet al.2017). Third,the benefits from cash crop production also accrue to other noncash crop farmers through the effect on employment since most cash crop production is labor-intensive (Irzet al.2001;Poultonet al.2008). The increase in labor demand for highvalue cash crops might increase the average wage among non-cash crop farmers. Moreover,the introduction of cash cropping opportunities indicates that households can be less cash-constrained and are able to purchase improved inputs for crop production (Govereh and Jayne 2003). Their ability to adopt yield-increasing technologies and agronomic practices is thus enhanced (Li Met al.2020). This cash income ultimately offers opportunities for farmers to invest in and improve the management of their farms,subsequently stimulating agricultural innovation and increasing yields (Li Met al.2020).

However,there are also reasons to question the positive effect of cash crop production on households’economic welfare (Orr 2000;Masanjala 2006). For example,some recent studies have shown that cash crop production has failed to increase household economic welfare,particularly the poorest ones,in some developing countries,due to high entry barriers (Kumaet al.2019).They found that promoting cash crop production made almost no improvement in the poorest households’ living standards,and these poorest are often neglected or inhibited from joining these cash crop productions (Dinget al.2020). In China,previous studies on cash crop production mainly focused on two aspects. One stream of literature focused on the concepts and theoretical basis of farmers’ cash crop production choices and their operational mechanisms (Liuet al.2021). Another stream focused on how cash crop production affects households’ labor allocation,subsequent household migration decisions,and other non-economic outcomes,such as environmental and ecological consequences(Suet al.2016;Liet al.2018). Although it has been well documented that cash crop production can be an effective approach to improving households’ economic welfare (e.g.,Leiet al.2019;Dinget al.2020),most of these observations are correlations. There was rather limited evidence regarding its causal relationship. It is still unclear to what extent and under which conditions cash crop production can achieve an ideal outcome at the micro-household level (Masanjala 2006;Jones and Gibbon 2011).

Moreover,farmers’ decisions on cash crop (vs.staple crop) production have been increasingly affected by the perceived risks caused by climatic change(Mtambanengweet al.2012;Ajuanget al.2016). In the context of pulse production in China,farmers’ perception has been enhanced due to recurrent climate variability(e.g.,excessive rainfall and floods in different regions),which directly affects farmers’ expected pulse outputs and economic benefits (Fosu-Mensahet al.2012). In response to such an enhanced climate-related risk perception and the observed adverse effect on pulse production,farmers might consider alternatives and more resilient crops to cope with such potential negative impacts (Quigginet al.2010). As posited by Asrat and Simane (2018) and Ojo and Baiyegunhi (2021),adaptation to climate change involves a multi-step process in which a strong perception (or strongly perceived changes of the climatic conditions) has to be built,and subsequently,a proper field response might be initiated to those changes.There were several studies examining such relationships(Maddison 2007;Hassan and Nhemachena 2008;Idrisaet al.2012;Mtambanengweet al.2012;Ajuanget al.2016;Huonget al.2019),and it has been concluded that the households’ climate change adaptation behaviors are directly associated with their perceptions (Fahadet al.2020). Nevertheless,there is limited study on the combined implication of climate change perception and its impact on commercial farming (Fierros-Gonzalez and Lopez-Feldman 2021) and economic welfare (Ojo and Baiyegunhi 2021).

The present study investigated the effect of commercial pulse farming on the economic welfare of rural households in China,considering the farmers’ perceptions of the impact of climate change (as an important explanatory variable). We explored the mechanism through which commercial pulse production could boost household economic welfare and to what extent it improves their economic welfare. First,we assessed the determinants of rural households’ commercial cash crop farming behaviors(both in binary and continuous measurements). Second,we estimated the effect of commercial pulse farming on a household’s economic welfare. We measured a household’s economic welfare by using the index of income (including total household income,agricultural income,and off-farm income).

The key contribution of this study is,firstly,to provide a quantitative assessment of the promotion of commercial crop production. This can be an effective way to improve the economic welfare of Chinese rural households. We focus on pulse production,the cash crops with distinctive local features being the main income sources of local farmers. However,the same economic findings can also be extended to other cash crops,such as rubber farming in South China (Minet al.2019). Second,scrutinizing the literature,a substantial share of early studies about cash crop productions are either based on case studies or correlation studies. Our study quantitively examined the causal relationship between commercial pulse farming and households’ economic welfare. Third,we further disentangled the internal mechanism of how cash crop production improves household income and quantitively examined to what extent and under which conditions commercial pulse farming can lead to a significant increase in households’ income,agricultural income,and off-farm income.

2.Pulse farming and its commercial production in China

China has a long history of farming and consuming pulses (Guo 2014). There are more than 20 types of legume crops ranging from mung and adzuki to common broad beans and peas. Pulses are characterized by strong adaptability and rich nutrition. Most of these pulses have a short growth period,wide adaptability,and symbiotic nitrogen-fixation properties (Maet al.2022a,b;Wanget al.2022). They are also suitable crops for intercropping with cereals,potatoes,maize,and other staple crops. Farmers often plant pulses to utilize the fallow land and to reduce disaster damage (Chenget al.2009).

The proportion of pulse planting in China’s grain crop planting area was approximately 0.49% by 2020. The yield of pulses has increased from 1 398.9 kg ha-1in 2001 to 1 899 kg ha-1in 2020 (Fig.1). Although pulse farming accounts for a small portion of China’s total crop planting area and approximately 0.5% of the total output,the demand for pulses has increased noticeably. By 2020,per capita pulse consumption was approximately 6 kg per person,three times higher than in 2010.

Pulses are important commercial agricultural products(Table 1). It has become an effective measure in rural poverty alleviation. Since 2015,several national central committee No.1 documents have encouraged structural adjustment and optimization by promoting pulse farming.Different pulse varieties,such as the Baicheng mung bean,Dali broad bean,and Cochran red kidney beans,are all important local agricultural products for market and self-consumption. For instance,the fresh broad bean industry in Dali has developed rapidly in recent years. In 2019,the annual planting area for broad beans increased to 300 000 mu (1 mu=0.67 ha). Although smallholders were the dominant pulse farming group,commercialization has grown fast. Pulses have become the main source of revenue for many locals,with a total output value exceeding 1.2 billion CNY per year.According to the mode of intensive management and large-scale development,the Baicheng mung bean has been actively developed and has obvious advantages compared to mung beans produced in other places in China.

Table 1 The price of pulses and the main food crops in China1)

Our study was conducted in two main pulse farming regions in China: Yunnan and Jilin provinces. We focused on two types of pulses -mung beans in Baicheng City and broad beans in Dali Bai Autonomous Prefecture. These two regions have a significant role in China’s pulse production and market development.For instance,the mung bean is a certified geographical indicated (GI) product with an average annual planting area of 1.2 million mu in Baicheng. The annual mung bean production in Baicheng reaches 100 thousand tons,accounting for 11% of the total national production.Approximately half of the Baicheng mung bean crop was exported,representing more than 30% of its total exports (The People’s Government of Baicheng City 2015). Dali is the largest producer of broad beans.Broad beans are the primary pulse crop in the country,representing 16.5% of the national planting area. Both regions represent typical farming systems in China.Yunnan has a typical smallholder farming system,and the development of the pulse industry as an economic development strategy has significant implications for a large area of the southwestern region of China,while Baicheng is a typical medium-to-large scale farming system. The development of its pulse industry can impact its economic welfare significantly,and it also reflects a different agricultural development strategy with a large-scale farming system. Studying these two regions can yield significant implications for rural agricultural development strategies by promoting the pulse industry.

3.Research design

3.1.Sampling and data collection

The data were collected through large-scale field interviews. Instead of using an equal-weight random sampling strategy,we used a cluster sampling method to sample the study cities/counties and villages within these two regions (United Nations Statistics Division 2005).First,the surveys were conducted in five cities/counties in Baicheng and four cities/counties in Dali. The samples cover 32 townships with a total of 66 villages. Second,approximately 15 to 35 households were randomly surveyed in each selected village,including both commercial and subsistence farmers. The study involved a two-round household survey,which was conducted in October 2018 and 2019 as pulses were fully harvested in August and September. Among the 938 questionnaires distributed,848 valid responses were obtained (the response rate was above 90%). The distribution of households reflects the geographic concentration of pulse production,with wide cultivation of broad beans in Dali and high concentrations of mung beans in Baicheng(mainly in Tiaonan and Tongyu) (Table 2).

Table 2 Sampled distribution in Dali Bai Autonomous Prefecture and Baicheng City

The field surveys were administered in four blocks of information. The first block recorded the demographic characteristics and family features of the household head,including age,educational level,and family size. The second block collected information on the farm conditions,including total farm size,the size of farmland used for pulses,and the percentage of farmland equipped with irrigation facilities. The third block interviewed farmers regarding production costs,sales status,and loans and subsidies. The fourth block asked about perceptions ofclimate change-associated risks,including attitudes and beliefs about climate change.

3.2.Farmer’s pulse farming purposes and their land uses

To investigate how households’ pulse production affects their economic welfare,two specific challenges need to be addressed before we conduct the empirical analysis.First,farmers plant cash crops for various purposes. A substantial share of farmers plant pulses for commercial purposes (as discussed in Section 2),while still,numerous farmers plant pulses for their consumption(thus,subsistence pulse farmers). Pulses are cash crops with a high market value but are also commonly consumed by rural farmers. It is often difficult to directly determine whether a farmer is a commercial pulse grower or a subsistence pulse farmer. In reality,farmers can quickly adjust their production purpose and cropping structural changes.

Second,what variable can be used as the best proxy to capture the different purposes of pulse farming?The terms ‘subsistence’ and ‘commercial’ have various definitions (Glover and Jones 2019). Some research uses land allocation (reflecting the change of cropping structures) or farm size to evaluate commercial or subsistence farming (Chenet al.2015;Suet al.2016). In this case,land as the primary input can be an important indicator to imply if a household has a commercial purpose or is a subsistence farmer. Some papers used the percent of total production sold as the proxy indicator of commercial farming (Foster 1988). According to Foster(1988),farmers who sell less than 10% of their products could be defined as subsistence farmers. Henceforth,our study assumed that farm size and the sold-out ratio could be close proxies (or identifiers).

Farmers planting a pulse crop larger than 1 mu were categorized as commercial farmers in the present study.1Some studies used the sold-out ratio as a primary identifier of whether household farming is for commercial purposes or selfconsumption (Foster 1988). In these studies,the widely used criteria are whether households sell more than 10% of their crops to the market.We used 1 mu as the threshold because,in the field survey,farmers usually indicated that they would plant less than 1 mu if they decided to plant a pulse for their consumption (as subsistence farmers). Further,to test the robustness of commercial and subsistence farming,we further strictly defined commercial farmers as having a pulse crop larger than 2 mu,while strictly self-sufficient farmers were defined as planting a pulse crop of fewer than 0.5 mu. The results are presented in Appendix A,and the results are rather consistent with the main results.

3.3.Defining farmers’ economic welfare

When investigating the effect of commercial pulse planting on farmers’ economic welfare,the selection of outcomes may include many different aspects. Economic welfare can be measured through a variety of factors,and in fact,there is still limited agreement on what should be the definition of households’ economic welfare in the existing literature (Tankari 2017). The proposed household economic welfare includes households’ income,level of literacy,life expectancy,and living standards. Most literature on economic welfare is on the social macrolevel,and these concepts are mainly focusing on producer surplus,consumer surplus,infra-marginal rents,and socio-economic rents (Nordhaus and Tobin 1972;Jensenet al.2019). However,some early works had made some progress in measuring households’ economic welfare.For instance,Masanjala (2006) used two categories of variables to measure households’ economic welfare,which includes households’ income (both total income and family off-farm income) and households’ food security(measured with households’ total food purchases and total food consumption). Later,a study conducted by Carlettoet al.(2013) also used food security as the main indicator of household economic welfare. Some other studies chose even more indirect measurements. For instance,Woodet al.(2013) used children’s height and weightz-score to measure households’ economic welfare,and Mmbandoet al.(2015) and Tankari (2017) used households’ food expenditure as an indirect indicator.

In China,when farmers plant cash crops for commercial purposes,it might significantly affect their households’ total income and agricultural income,their time and labor allocations,and ultimately food expenditures. Taking these into account,this study used household income as the indicator to measure household economic welfare. It includes the households’ agricultural income (both cash crop income and other crop production income),off-farm income,and total income.

4.Empirical strategies

The empirical analyses were conducted in two steps to examine the driving factors of households’ pulse commercial growing behavior and further evaluate whether commercial (vs.subsistence) pulse farming will significantly improve households’ economic welfare. First,we employed the Heckman two-step model to examine the driving factor of households’ commercial pulse farming behavior with the consideration of the potential presence of the sample selection bias. Second,we used the endogenous treatment regression (ETR) model to examine how and to what extent commercial pulse farming affects households’ economic welfare outcomes.

4.1.Heckman’s two-step estimation of farmer’s pulse planting choices

Assuming that households make decisions in a sequential approach. They first decide whether to plant pulse for self-consumption (as a subsistence farmer) or for commercial purposes (to sell to market for economic benefit). Once such a decision has been made,households then decide how much of their land would be used for pulse production (vs.other staple crops). In this case,we observed households’ actual pulse farm size with no hint of self-consumption or commercial purposes. Following the definition of commercial pulse farmer in Section 3.2,we first estimated the factors that affect the probability of a farmer being a defined commercial (dij=1) or subsistence pulse farmer (dij=0) with a potential concern about the potential sample selection bias (Heckman 1976). Thus,we estimated the following specification using Probit estimates (as the first step):

where in eq.(1),Zijis a vector of predetermined households’ characteristics,which are believed to affect households’ choices of being a commercial or subsistence pulse farmer. We particularly added the households’perceived climate change and if they had received an agricultural subsidy. Both covariates were expected to have a positive correlation with the household being a commercial pulse farmer. First,several studies have shown that receiving an agricultural subsidy can improve farmers’ credit access,particularly when the subsidies are calculated according to the size of the farm. Thus,the larger farm receives large subsidies,which leads to a larger pulse farming size for the market. Second,when farmers perceive climate change,increasing pulse farming size can effectively counter the negative impact on staple crops or other crops to avoid climatic risks (Barbieriet al.2017). Therefore,once the above specification is estimated,we plugged the estimated coefficients into the following eq.(2) to calculate the inverse mills ratio (mit):

The inverse mills ratio represents the ratio between the probability density and cumulative distribution functions,which captures the potential sample selection bias of each observation. Secondly,we run the following specification by adding the inverse mills ratio (mit) as an additional covariate.

whereYitis households’ actual pulse farming size in yeart,Zitis a vector of explanatory variables,which might overlap with the vectorZijas shown in eq.(1). The second stage of analysis ensured that at least one of these variables differed from the variables considered in the first-stage estimation to avoid correlation (Wooldridge 2015).μitis a random disturbance vector. If the parameterβ2is statistically significant in the secondstep estimation,it indicates that there is a substantial magnitude of sample selection bias (Ochoaet al.2019).

4.2.Endogenous treatment regression model

We examined how commercial pulse farming affects households’ economic welfare. Instead of using the pulse farming size as a continuous explanatory variable,we used the binary variable to define the commercial pulse farmer and subsistence pulse farmer and subsequently used the ETR model to perform the analysis. This is believed to be an effective and unbiased estimation for two reasons. First,the pulse cropping size was used as a proxy to capture the natural differences between the commercial pulse farmers and subsistence pulse farmers. Comparisons among commercial pulse farmers with different pulse farm sizes might play rather a minor impact since commercial pulse farmers could adjust their farm size in the face of increased market demand.Second,the ETR can fully adjust the potential sample selection bias when the binary variable is used to conduct the estimation. The use of ETR can produce consistent estimates by removing the bias originating from both observed and unobserved factors (Hübler 2016;Li Met al.2020). It also provides a robustness check when the criteria of pulse farming size were adjusted to 2 mu(commercial farmer) and 0.5 mu (subsistence farmer) as the robustness check.

Assuming that households who were planting pulse for either commercial or self-consumption purposes are indicated bydit,and the outcomes regarding households’economic welfare are indicated asYit,the following basic model was developed:

where var(εit)=σ2,cov(εit,μij)=σμε,and corr(εit,uij)=ρμε. Ifρμεis statistically significant,it indicates a significant endogeneity due to sample selection bias (Maet al.2018). We fit this constrained model with the maximum likelihood estimator and reported the ETR estimated results in in Section 5.3 regarding households’ incomes.

5.Results and discussion

5.1.Some descriptive statistics

Before reporting the empirical estimation results,we present descriptive statistics regarding sampled households’ pulse production and their demographic and family characteristics. This study first conducted a simple two-sample comparison between the commercial and subsistence pulse farming households to shed some basic insight into their economic welfare outcomes,land allocation (between pulse farming and other crops),and demographic characteristics. The results showed that these two types of farmers exhibited some distinct features. Table 3 reports the details. Commercial farmers had a significantly larger pulse planting size relative to subsistence farmers,and their total farm size also showed a significant difference.

Regarding the households’ economic welfare outcomes,commercial pulse farming households had a relatively higher total income than subsistence farmers.Given the significant difference in farm size,it can be naturally expected that subsistence farmers might have a higher non-agricultural income (due to labor allocations differences) compared to commercial farmers. This is particularly relevant given the increasing heterogeneous livelihood strategies observed in the rural community,and specialization in farming (vs.off-farm employment)is intensifying (Abdullahet al.2019). It was also found that commercial farmers were slightly younger than subsistence farmers (with an average difference of two to three years),but there were no significant differences in their education levels.

Comparing farmers’ pulse market prices and production costs,we found that commercial pulse farmers sold at a relatively higher price than subsistence farmers.The production costs among commercial farmers were significantly lower than those of subsistence farmers,and there were almost 100-CNY differences per mu for pulse production (Table 3). Our intuitive understanding is that the heterogeneous reactions of these two types of farmers to the market are due to different pulses and agricultural inputs. Commercial pulse farmers might have a better market access channel,produce higher quality products,and subsequently sell at a relatively higher price. In contrast,the subsistence farmers might primarily sell in the local market,which might face poor purchasing power and a large quantity of pulse supply.We also noticed that commercial pulse farmers used a substantially higher level of mechanization services and received a substantially larger amount of agricultural loans and subsidies for pulse production. These differences are all statistically significant (at a 1% significance level).

Table 3 Comparisons between commercial and subsistence pulse farmers

Moreover,farmers’ perceptions of climate change riskson pulse production were heterogeneous. Fig.2 shows farmers’ perceptions of climate change risk on pulse production. More than 60% of commercial pulse farmers perceived a significantly negative impact of climate change on their pulse production over the previous five years,while the same was observed for only 20% of subsistence farmers. It is expected that farmers’ perceptions of climate change on pulse production will have a significant impact on their pulse production decisions.

5.2.Determinants of commercial pulse farming

The first stage of the Heckman model (Column 3,Table 4) identified the determinants of the probability of a farmer engaging in commercial pulse farming.The results indicated that farmers’ total farm size(Row 1),percentage of irrigated farmland (Row 2),agricultural production subsidies (Row 10),and climate change perceptions (Row 11) influenced the probability of the decision to engage in commercial pulse farming. Specifically,among the statistically significant coefficients,households’ perceived climate change and the percentage of irrigated farmland had the largest influence on commercial pulse planting behavior. Subsequently,we found that if households received an agricultural subsidy,their total farm size significantly positively affected their commercial pulse farming behavior. These findings are consistent with previous studies (e.g.,Li Wet al.2020). Particularly,the coefficients of households’ perceived benefits and losses due to climate change were significantly positive.This finding suggested that the higher the perception that climate change might benefit crops,the higher the probability that farmers expand the pulse farming areas;the perceived loss due to climate change could substantially reduce the pulse farming size. This relationship is much more significant.

We also examined the relationship between households’ demographic characteristics and their probability of being commercial pulse farmers. It was found that family characteristics such as family size,age of household head,and education level had no significant influence on farmers’ commercial production decisions.Surprisingly,households’ previous year’s pulse sales price had limited influence,which might be counterintuitive since commercial pulse farmers are expected to be highly sensitive to the market prices. Nevertheless,the insignificant results might reflect the reality that commercial pulse farming is rather a long-term household livelihood strategy determined by their long-term objectives rather than short-term market price changes.

The second stage of the Heckman model (Columns 1 and 2,Table 4) provided estimations regarding the factors that influence the actual pulse farming size. Inother words,how do different factors affect households’decisions regarding land allocation (for pulse farming)?First,we identified if there is a presence of the sample selection bias. As shown in Table 3,the inverse mills ratio (Lambda,Row 12),which integrated the probability of a farmer becoming a commercial pulse farmer,was statistically significant even at the 1% significance level.This result indicated a significant sample selection bias.The negative coefficient of the inverse mills ratio indicated that the factors associated with subsistence pulse farming households might be underestimated (Woodridge 2010).Thus,it is appropriate to employ the Heckman two-step model instead of separate regressions to control for sample selection bias problems (Ochoaet al.2019).

Table 4 Results of the Heckman two-step model with Probit regression

Therefore,in the second stage,once the estimated was added,the households’ previous year’s pulse sales price significantly and positively affected households’ pulse farming size. Basically,a 1% increase in the previous pulse sales price led to a 0.103% increase in pulse farming size. There was also a significant and negative effect of production costs on households’ pulse farming size.Whether a household receives a loan for pulse production was a crucial driver of commercial pulse production.Agricultural technology service showed a positive influence on households’ commercial pulse farm size,and a 1%technology service improvement could increase the commercial pulse farming allocation by 0.167%.

The family population had a significantly negative correlation with households’ commercial pulse production. This is probably because a larger family endows with higher levels of labor and subsequently more off-farm income. Studies by Ma and Abdulai(2017) and Ochoaet al.(2019) found similar patterns in other regions in China with other crops. Finally,the larger the farm size endowed by pulse farmers,the higher the probability that the households would increase their commercial pulse farming size. An increase of 1% in initial farm size resulted in a 0.001% increase in pulse farming size. However,the portion allocated to pulse production declined when a household had more irrigated plots,with a 1% increase in irrigated land leading to a 0.266% decrease in pulse farming allocation (at the 1% significance level),because households prefer to grow maize or other staple crops on irrigated plots while cultivating pulses on non-irrigated areas.

5.3. Effect of commercial pulse production on farmer’s economic welfare

This section shows the estimated ETR effect of commercial pulse farming on households’ economic welfare,including total income and agricultural income.First,as shown in Table 5,the statistically significant coefficient ofρμεindicated a significant selection bias due to potential unobservable covariates (Ma and Abdulai 2017). Failure to consider this selection bias would underestimate the true effects of commercial pulse production on households’ economic welfare outcomes.

With the ETR model,Table 5 shows a significant positive estimated effect of being a commercial pulse farmer on household total income (Model 1),non-farm income (Model 2),and agricultural income (Model 3).The three income sources enhanced the understanding of the mechanism through which being a commercial pulse farmer affects households’ income. The results showed that the coefficients of the cultivation variable in the ETR model were positive and statistically significant,suggesting that being a commercial pulse farmer increased total household income. Specifically,a 1%increase in pulse farming allocation could lead to a 0.328% increase in total household income. This finding is consistent with several previous studies from other regions (Masanjala 2006;Li Wet al.2020),which found that cash crop production could significantly increase household income. In addition,agricultural income is the main contributor to households’ total income(Table 3). Model 3 (Columns 5 and 6,Table 5) shows that commercial pulse production could significantly increase households’ agricultural income.

However,engaging in commercial pulse farming had a negative but insignificant effect on households’ offfarm income (Columns 3 and 4,Model 2,Table 5). This might be caused by the trade-off between commercial production and households’ off-farm employment. This is consistent with Li Wet al.(2020),who found that when rural households allocate more time to cultivate labor-intensive cash crops,such as pulses,less time will be available for other off-farm employment,which subsequently results in lower off-farm income.

Table 5 Effect of commercial pulse production on farmers’ incomes

Moreover,the percentage of farmland equipped with irrigation facilities was positively associated with households’ agricultural income and total income. The estimated coefficients were significant at 1 and 10%,respectively. These results echo the findings of Fukase and Martin (2016),who argued that farmers with a better endowment (irrigation facilities) could earn a higher income. However,the ratio of irrigated land had a negative and insignificant relationship with households’non-farm income. Similarly,household farm size was positively associated with total household and agricultural income;however,there was no significant relationship with non-farm income.

6.Conclusion and policy implication

The existing study investigated whether the promotion of cash crops in two different regions of China could improve the economic welfare of households. This study focused on promoting commercial pulse farming in rural households. This study used a sample of 848 farmer households in two typical farming systems in Baicheng and Dali to analyze (1) the determinants of farmers’commercial pulse production behavior and (2) the impact of commercial pulse planting on farmers’ economic welfare with a focus on household income (including total household income,farm income,and non-farm income). The Heckman two-step model was employed to correct for potential sample selection bias and identify the determinants,and then an ETR model was used to estimate the impact of commercial pulse production on household economic welfare.

The empirical results showed that the households’pulse sales prices significantly stimulated households’decision-making regarding land allocation. When households could access the market and sell pulses at relatively higher prices,they were motivated to allocate more farmland for commercial pulse farming.This result indicated that households’ agriculture intensifies their farming to increase the actual value of their farming output per unit of land. On the other hand,households’ pulse production costs and potential losses due to climate change significantly inhibited their pulse farming expansion. Extreme weather caused by climate change could significantly increase their planting risks. Moreover,receiving a loan for pulse production and accessing agricultural technology services could positively affect households’ commercial pulse production.This is consistent with the findings in Malawi and other developing countries. Our study also identified a clear pattern of how commercial cash crop production affects households’ economic welfare. Specifically,commercial pulse production increased both total household income and agricultural income. These results also revealed that cash crop cultivation could significantly increase farm income in China’s low-income regions.

The findings of the study inform several important policy implications. First,commercial pulse production can upsurge the economic well-being of households,but future climate change may increase pressure on commercial agriculture. The increased climate change risks and their related perceived significant investment losses may inhibit households’ actual behavior from increasing the size of pulse cultivation. The extent to which the policy can mitigate this negative impact or at least release households’ perceived risk should be an essential future research direction. Second,commercial pulse production (relative to major crop production) is not a short-term market response but a long-term livelihood strategy for households. Commercial households can quickly adjust the size of pulse farming;however,this adjustment is only from the perspective of land as a single input. Ensuring households have access to more efficient input markets,including capital,technology,and other related technologies,may provide commercial pulse farming households with additional resilience to market changes. In response,we may expect to achieve better household economic welfare outcomes. Therefore,from a policy perspective,more research on market-resilient commercial production systems should be conducted to achieve these long-term goals.

Acknowledgements

This work was supported by the China Agriculture Research System of MOF and MARA (CARS-08-G21)and the National Natural Science Foundation of China(71904190).

Declaration of competing interest

The authors declare that they have no conflict of interest.

Appendixassociated with this paper is available on http://www.ChinaAgriSci.com/V2/En/appendix.htm