Impact of climate change on farmers’ crop production in China: a panel Ricardian analysis

Abstract

The impact of climate change on crop production remains a major concern around the world, as cropping is directly linked to global food security. Therefore, accurately estimating the effects on farmers’ crop production is crucial for effective adaptation strategies. We employ the panel Ricardian model to estimate the impact of changes in temperature and precipitation on crop net revenue, using a 17-year panel dataset that includes ~180,000 households in China. Our analysis accounts for market price differences and interactions between temperature and precipitation, and farmers’ adaptation to climate change. We found that temperature and precipitation significantly impact crop net revenue, with farmers’ adaptations partially mitigating these effects. The effects of temperature change on crop net revenue vary among different seasons, with the negative consequences of maximum daily temperature (T_max) concentrated in spring and autumn, and the positive effects of minimum daily temperature (T_min) only being seen in autumn. Additionally, precipitation negatively impacts crops mainly in autumn and winter. Climate change impact is more pronounced for cash crops compared to grain crops and for large farms relative to small farms. The negative effects of spring temperature are concentrated in northern regions, whereas southern regions are more sensitive to temperature and precipitation in autumn. Furthermore, simulations of future climate patterns suggest that the impact of T_max on crop production is decreasing, while the negative effect of increasing T_min is increasing.

Introduction

Agricultural production and food security are key areas that could be significantly impacted by climate change. Cropping is a very important sector that directly provides grain and other agricultural products (e.g., vegetables and fruits) and is the most climate-sensitive agricultural sector (Jagermeyr et al. 2021; Malik et al. 2022). Therefore, determining the impact of climate change on the crop net revenue is critical for enhancing the climate adaptive capacity and farmers’ livelihoods. Despite a dramatic increase in the literature on the impact of climate change (Chen et al. 2013; Huang et al. 2022; Massetti and Mendelsohn, 2011; Wang et al. 2014a, 2014b), how China’s crop production would be affected when considering farmers’ adaptation is still not well-understood.

Cropping is the fundamental industry of agriculture and faces multiple challenges in providing food and nutritional security (Reddy et al. 2022). China’s grain production reached 670 million tons, while cotton, oilseeds, and sugar output reached 5.91, 35.86, and 120 million tons in 2020, respectively. China’s food security is an important part of global food security. Figure 1 illustrates China’s significant contribution to global agricultural outputs from 2014 to 2019 across various crop categories, including soybeans, vegetables, fruits, corn, rice, wheat, and grains. Notably, vegetables consistently account for over 50% during this period. The shares of global production for corn, rice, and wheat are approximately 25%, 30%, and 20%, respectively, thereby underscoring China’s pivotal role in global grain production and its importance for food security.

Fig. 1: China’s share of global crop production.

According to data from the Food and Agriculture Organization (FAO), China’s share of global crop production is substantial. China’s grain output accounts for 20% of the world’s total output, while fruit output approaches 30% and vegetable output exceeds 50%.

Despite China’s significant position in global crop production, the country’s crop production is highly exposed and more sensitive to climate change. We utilized temperature and precipitation data from the past 50 years—from 1970 to 2020—for all counties in China and performed a regression analysis of the annual mean temperature and total precipitation for each county in the respective years. The results show that climate change has led to a more significant increase in temperature in the north compared to the south over the past 50 years. Specifically, the regression coefficient indicates that the temperature rise is more pronounced in the northern regions. Regarding precipitation, the northern and central regions have experienced a decrease, while the south has seen a significant increase. This water–heat mismatch is highly susceptible to adverse effects on food supply efficiency (Carter et al. 2018; Wang et al. 2014a, 2014b), suggesting that climate change increasingly impacts farmers’ production across China. Therefore, determining the impact of climate change on China’s crop production is of great significance.

Literature review

Agriculture is the sector most significantly affected by climate change, and a growing body of literature highlights the impact of climate change on agriculture (Burke et al. 2015; Kotz et al. 2021, 2022). Accurately identifying the impacts of climate change is the first step, as it provides the necessary foundation and prerequisites for subsequent mitigation efforts. Two challenges in identifying the impacts of climate change are generally acknowledged. One is that recent consequences of climate change have exhibited new characteristics, particularly an increased frequency of extreme weather events, which may require longer-term research data to support the stability of conclusions. However, as observed data continues to accumulate, the difficulty of this challenge is gradually diminishing (Burke and Emerick, 2016). The second challenge is that the impacts of climate change may encompass effects beyond climate change itself (Nguyen and Scrimgeour, 2021; Wang et al. 2014a, 2014b). A widely accepted fact is that during the process of climate change, humans consciously or unconsciously engage in various adaptive behaviors, thereby modulating the impacts of climate change. Therefore, identifying the effects of climate change itself from those that consider adaptations is crucial for the formulation and adjustment of climate policies.

Several studies have evaluated the impacts of climate change on various crops across different regions, particularly focusing on staple food crops (Chen and Chen, 2018; Jagermeyr et al. 2021; Lobell et al. 2008; Malik et al. 2022; Tao et al. 2006; Yao et al. 2007; Zhao et al. 2017). However, these studies have not sufficiently considered the socioeconomic factors and behavior of producers. The Ricardian model has significantly contributed to analyzing the impact of climate change on agricultural profitability, while considering adaptation scenarios, and has been applied to many countries and regions worldwide (Mendelsohn and Massetti, 2017), including Africa (Kurukulasuriya et al. 2006), Western Europe (Van Passel et al. 2017), and Italy (Bozzola et al. 2018). Liu et al. (2004) first applied the Ricardian model to China and found that global warming will increase average agricultural net revenue at the county level. Wang et al. (2009) also used this model to evaluate the impact of climate change on crop net revenue based on cross-sectional farm-level data, finding that global warming adversely affects rainfed farms but benefits irrigated farms. However, Chen et al. (2013) found a positive impact of temperature and a negative impact of precipitation on the crop net revenue and believed that climate change may create potential benefits for China’s agriculture. The discrepancies in existing research conclusions may be due to the problem of missing values in the cross-section Ricardian model (Burke and Emerick, 2016), which may not be a valid reference for current policy-making on climate change. Therefore, re-examining the impact of changes in temperature and precipitation on China’s entire crop net revenue using panel data and the Ricardian model at the household level is necessary, and we believe that this study can lead to more reliable and convincing results by improving the model settings and data quality.

Our study makes three primary contributions. First, we utilize the Hsiao two-step method within a panel Ricardian model to estimate the impact of maximum, average, and minimum temperatures (T_max, T_ave, and T_min) and accumulated precipitation on crop net revenue. This approach effectively addresses the issue of omitted variables inherent in cross-sectional models and allows for a more accurate analysis of multi-period net returns in the context of farmers’ inter-period adaptation behavior, which is an aspect previously unexplored. Additionally, we use household-level panel data covering 17 years, which is most representative and authoritative of China’s cropping situation. Our analysis of T_max and T_min is more valuable and comprehensive because crops are more sensitive to boundary temperatures, and our model allows for differences in market price feedback for crops in different regions, making our estimates more realistic (Haile et al. 2016). We also consider the interaction effect between temperature and precipitation, as shown in Table 1.

Table 1 Hypothesis testing for homogenous market shocks across regions and climate interaction.

Second, we analyze the heterogeneity of climate change impacts on crop net revenue. Agricultural sectors differ not only in sensitivity but also in their adaptation strategies (Chatzopoulos and Lippert, 2015). While previous studies predominantly focused on grain crops, our research expands the analysis to include diverse crops, including cash crops and grain crops and examines the responses across various farm scales, such as large- versus small-scale farms. We also analyze the impact of climate change on different regions of China. This analysis is significant as it reveals how distinct agricultural practices and conditions influence vulnerability and adaptive capacities, ultimately providing a robust foundation for targeted policy recommendations aimed at enhancing the resilience of the agricultural sector.

Third, we use future temperature and precipitation data under the moderate emissions scenario to predict the seasonal distributions of T_max, T_min, and T_ave on the net revenue of crop production in 2030, 2050, and 2100. This predictive modeling is of great value in mitigating the adverse effects of climate change and developing effective measures.

Methods and data

Methods

Ricardian model

The Ricardian model assumes that farmers make rational decisions to maximize their profits based on a set of exogenous conditions that they cannot change. In China, farmers typically cultivate multiple crops throughout the year on a single plot of land. To calculate the total net revenue per mu (a unit of area commonly used in China, 1 mu ≈ 667 m²) in one year, the average revenue per mu is subtracted from the average cost per mu. According to the original idea of the Ricardian model, these costs include those of seeds and seedlings, organic fertilizers, chemical fertilizers, pesticides, plastic sheeting, water and electricity for irrigation, animal labor, machinery, land leasing, and labor. Specifically, labor costs include the conversion of self-employed labor and the expenses for hired workers. The formula for maximizing farmers’ profits can be expressed as follows:

$${\rm {M{ax}}\; \pi }=\sum _{i}{p}_{i}{Q}_{i}\left({K}_{i},{E}_{i}\right)-T{C}_{i}({Q}_{i},W,E)$$

(1)

where π is the net crop revenue for each unit of land, p_i is the price of crop i, Q_i represents the production function of crop i, K_i is a vector of agricultural inputs other than land, and E_i is a vector of exogenous environmental factors such as climate change and geographic conditions. The Ricardian model assumes that the net revenue of crop (π*) is a function of exogenous factors only, giving the profit-maximizing farmers’ agricultural production decision. We can divide these exogenous variables into three categories: X_it that changes over time and region, T_t that only changes over time, U_i that does not change over time, and climate variable C_it.

$${\pi }_{{it}}^{* }={X}_{{it}}\beta +{U}_{i}\gamma +{C}_{{it}}\alpha +{T}_{t}\theta +{u}_{{it}}$$

(2)

Hsiao two-step method

The value of the Ricardian model lies in its assessment of the impact of long-term climate change (typically a 30-year average). If we take years of panel data directly using the above equation for fixed-effect regression, climate variability will likely be insufficient to obtain accurate α. Therefore, this paper refers to the Hsiao two-step method proposed by Massetti and Mendelsohn (2011) to obtain a robust estimate of α, which makes the fixed-effect estimation of time-varying variables robust at the household level for ignoring time-invariant variable (Blanc and Schlenker, 2017). The details of the Hsiao two-step method for the panel Ricardian model are as follows.

First, we used a fixed-effect model to regress the net crop revenue to all time-varying variables, excluding climate variables:

$${\pi }_{{it}}^{* }={X}_{{it}}\beta +{\varepsilon }_{{it}}$$

(3)

where ε_it is the error term. X_it—X represents a matrix of variables that change over time and individuals. Specifically, these variables include irrigation, age and educational level of the main workforce, the number of plots of land, whether the household has a village cadre, the number of laborers, and the size of the farm.

We then use the residual term obtained from estimating Eq. (3) and calculate its mean in the time dimension. Subsequently, we regress the time-mean residuals to climate and other time-invariant control variables:

$${\bar{\pi }}_{i}^{* }-{\bar{X}}_{i}\hat{\beta }={U}_{i}\gamma +{C}_{i}\alpha +{\bar{u}}_{i}$$

(4)

As differences in soil properties and climate in different regions can lead to systematic differences in crop selection and planting yields, such systematic differences may lead to the estimation bias of the α in the above equation. Panel Ricardian models control these potential influences by using two-way fixed effects.

Additionally, the Ricardian model implicitly assumes long-term adaptation in crop selection, that is, farmers will choose different crops according to different climatic conditions. However, different agricultural markets are likely to react differently to price changes (Nguyen and Scrimgeour, 2021). If heterogeneous price change effects cannot be addressed, the α estimated by the above equation may still be biased. A common practice to introduce heterogeneous price feedback is to include an interaction of the region and year dummy variables in the regression model.

In our research, a crucial assumption is that farmers make decisions based on annual profits. This assumption aligns with the actual decision-making practices of farmers and has been supported by multiple studies (Nguyen and Scrimgeour, 2021; Wang et al. 2009). Therefore, despite the possibility of farmers cultivating multiple crops with varying growth periods and the differential impact of climate change on different crops, from an individual farmer’s decision-making perspective, the average impact on multiple crops over the course of a year remains a significant factor driving farmers to adjust their planting decisions.

Identification strategy

To predict the impact of climate change, we take the logarithm of net crop revenue as the dependent variable. The quadratic terms of temperature and precipitation and the interaction of the two are introduced into the model. After estimating the first step of the Hsiao two-step method, the resulting time-mean residuals are regressed on climate and time-invariant variables:

$${\bar{\pi }}_{i}^{* }-{\bar{X}}_{i}\hat{\beta }=\,{\alpha }_{1}{T}_{i}+{\alpha }_{2}{T}_{i}^{2}+{\alpha }_{3}{P}_{i}+{\alpha }_{4}{P}_{i}^{2}+{\alpha }_{5}{T}_{i}{P}_{i}+\beta {C}_{i}+{\bar{u}}_{i}$$

(5)

where T_i stands for temperature and P_i for precipitation. C_i represents all time-invariant variables, including elevation, topography, distance to the main highway, year- and province-fixed effects, and year times province interactions. ${\bar{u}}_{i}$ represents the residual term that is not climate-dependent. The marginal impact of different seasonal temperatures and precipitation on crop net revenue is calculated as follows:

$$\frac{\partial {\bar{\pi }}_{i}^{* }}{\partial {T}_{i}}={\alpha }_{1}+2{\alpha }_{2}{T}_{i}+{\alpha }_{5}{P}_{i}$$

(6)

$$\frac{\partial {\bar{\pi }}_{i}^{* }}{\partial {P}_{i}}={\alpha }_{3}+2{\alpha }_{4}{P}_{i}+{\alpha }_{5}{T}_{i}$$

(7)

Since the dependent variable is logarithmic, the marginal effect is estimated using Eqs. (6) and (7) are interpreted as a change in one unit of the corresponding climate variable resulting in a percentage change in agricultural revenue.

Data

Data sources

Climate data from 1983 to 2019 used in this study was obtained from the China Meteorological Data Sharing Service System (http://cdc.nmic.cn), which provides the most detailed and accurate climate data in China. We matched the meteorological sites with the survey sample counties based on their latitude and longitude information. For counties that did not have matching meteorological stations, we used the nearest-neighbor matching method to match the stations closest to the centroid of the county. Daily meteorological information for all sample counties was collected. To represent the weather conditions of each village, we used the weather information from the county in which the village is located, following the common practice in previous literature. We also clustered the standard error at county levels to ensure the robustness of our estimation results.

Future climate data were generated from the model of NESM in Coupled Model Intercomparison Project phase 6 (CMIP6, https://esgf-node.llnl.gov/projects/cmip6/). Two SSP-RCP scenarios were chosen to project the future net revenue: SSP-245, a “medium mitigation” middle-of-the-road pathway; and SSP-585, a “no mitigation” high-emission pathway.

Microeconomic data was obtained from the National Fixed-Point Survey (NFP) collected by the Rural Economy Research (RCRE) of the Ministry of Agriculture of China in 1986. For data comparison, we used annual data waves between 2003 and 2019, covering nearly 180,000 households in 359 villages across 31 provinces. Representative NFP villages were selected based on region, revenue, cultivation patterns, population, and non-agricultural activities. In each selected village, a random sample of households was included in the survey. In rare cases where entire families were permanently relocated, they were replaced by another family. We used 17 years of panel data at the household level and employed a two-way fixed effect model to control for missing variables.

Descriptive analysis

Following Nguyen and Scrimgeour (2021), the net revenue of cropping is calculated to be the gross revenue less the operating expenditures for seedings, farm manure, fertilizers, plastic sheeting, pesticides, irrigation, animal power, mechanical operation, depreciation, repairs, family and hired labor, and so on. All price data are treated at constant prices for the base year 2003. We have deflated the price indices of agricultural inputs by region. We used the net revenue per mu in our regression analysis.

The maximum daily temperature is the highest temperature that occurs during the day. The daily minimum temperature is the lowest temperature that occurs during the day. The average daily temperature is the average of the temperature at four time points during the day (02:00, 08:00, 14:00, and 20:00 h). Then, based on the data of the daily temperature, the annual T_max, T_min, and T_ave are obtained. The impacts of climate change are integrated. However, which temperature indicator has the most obvious impact on the crop net revenue remains unclear. For example, for winter wheat, in addition to the obvious impact of precipitation, the impact of the minimum temperature in winter is greater than that of the maximum temperature, because the minimum temperature in winter is closer to the growth threshold of winter wheat and more closely related to the disease and pest conditions (Huang et al. 2022). Conversely, for summer-grown crops, summer temperatures may break through crop growth thresholds more easily. We report variable summary statistics in supplementary materials Table 1. The average annual revenue of farmers in China’s planting industry is 4088.5 yuan (572.39 US dollars; 1 USD = 6.96 Chinese yuan), and the average operating land area is 7.6 mu (1 mu = 1/15 ha). As our analysis sample included cash crops and grain crops, the yield was relatively high.

The seasonal temperature changes are reported in Fig. 2. Obvious differences exist in the temperature changes in the four seasons and the characteristics of the changes. Among them, the temperature rises most evidently in spring and summer, but almost no obvious change exists in temperature in winter and autumn. The variation range of the maximum temperature in spring is relatively large, and the change between the average temperature and the minimum temperature is relatively small. However, the variation of the maximum temperature in summer is not as obvious as that of the average and minimum temperatures.

Fig. 2: Changes in temperatures within seasons over two periods.

To examine the temperature changes since the beginning of the 21st century, we selected 2000–2003 as the base period and found that the recent temperature change (2016–2019) has significantly changed compared with the base period. The figure shows the distribution of the maximum temperature, average temperature, and minimum temperature in the two periods within the season.

Results

Main results

Figure 3 presents the marginal effects of long-term, short-term, and inter-temporal changes in temperature on crop net revenue in different seasons. We also report the elasticity of the impact of precipitation on net revenue in different time scales. Figure 3a shows that long-term T_max, T_min, and T_ave have significant effects on net revenue. Our findings that the increases in T_min in winter, spring, and summer negatively impact crop net revenue are consistent with the study of Lobell et al. (2008), who also found that cooler temperatures during the growth period of crops can lead to a decrease in crop productivity. Similarly, the positive effect of T_max increases in winter and summer on net revenue is in line with the research of Schlenker and Roberts (2009), who indicated that warmer conditions can have beneficial effects on crop production in some regions. However, an increase in T_min in autumn leads to an increase in net revenue. Increases in T_max in winter and summer positively affect net revenue, but the opposite is true in spring and fall. The variation in T_ave has a similar direction of impact as the T_max, but the effect is less pronounced. An increase in precipitation in winter and fall adversely affects net revenue, with the adverse effects being significantly greater in the fall than in winter. Conversely, increased precipitation in spring and summer is favorable for crop net revenue. The finding that the negative impact of excessive autumn precipitation on crop net revenue is supported by the findings of Deryng et al. (2014). They found that autumn precipitation can lead to harvest delays and an increased risk of crop diseases, resulting in reduced yields and net revenue. The positive effects of spring and summer precipitation are consistent with the well-established knowledge that adequate moisture during the growing season is essential for crop growth, as stated by Trenberth (2011). Figure 3b shows that short-term temperature increases during summer favor crop net revenue, whereas the opposite effect is observed during the fall. Additionally, changes in precipitation have little impact on net revenue in the short term. The possible reason is that the increase in low temperatures during the winter and spring seasons can lead to a rise in pests and diseases and declining net returns of agriculture. The negative impact of autumn precipitation is particularly pronounced, as autumn is the harvest season for many crops. In this season, crops do not require additional rain, so excessive precipitation can complicate harvesting, resulting in decreased net returns. The above findings indicate clear seasonal differences in the impacts of temperature and precipitation on crop net revenue, while the effect of short-term weather shocks is slightly smaller than that of long-term climate change.

Fig. 3: Impact of climatic factors on crop net revenue.

a We use a panel Ricardian model, assuming that the farmer’s lagged adaptive behavior is reflected in the net revenue for the year. The marginal effects represent the influence of long-term climate change on the crop net revenue. b We use a panel fixed effects model, where minimal adaptation of farmers is assumed. Farmers had no time to respond to the short-term weather shock of that year. c We consider intertemporal (5 years difference) behavioral adjustment of farmers. All regressions were clustered at the provincial level. d We investigate the impact of annual temperature and precipitation on crop net revenue, without controlling for seasonal effects. All models include quadratic terms for temperature and precipitation, as well as interaction terms for temperature and precipitation. To control for the regional heterogeneity of market shocks, we also control for the interaction between province and year. The terrain and distance from the main road at the village level were also controlled for.

Figure 3c, d show the response of intertemporal crop net returns to intertemporal changes in temperature and precipitation after considering farmers’ intertemporal behavioral adaptation to mitigate the adverse effects of a climate shock. With the adjustment of intertemporal behavior (see Fig. 3c), the adverse effect of temperature in autumn changes from negative to positive, with the adverse effect of precipitation being no longer significant. The results are related to farmers’ adaptive behaviors to climate change. Studies have shown that farmers can adjust their planting and harvesting schedules and irrigation practices to cope with climate variability (Morton, 2007). Our finding provides empirical evidence of the effectiveness of farmers’ adaptive behaviors at the intertemporal level. If we do not distinguish between seasons (see Fig. 3d), farmers’ behavior adjustment is also efficient in alleviating the adverse effects of temperature and precipitation. These results suggest that farmers’ adaptive behaviors are effective strategies for both seasonal and annual climate change.

To ensure the robustness of our estimates, we conducted the following robustness tests. First, considering the association between farmers’ adaptive behaviors and climatic change characteristics within regions, we clustered the regression at different levels—specifically at the county, city, and provincial levels. The regression coefficients were insensitive to these different clustering levels, providing evidence for the robustness of our estimates. We only report the results clustered at the county level in the paper. Second, we employed a stepwise regression approach to examine the stability of the estimated coefficients in both short-term and intertemporal regressions by controlling for different variables. The results indicated that after controlling for several household-level characteristics, the coefficients of the core variables became insensitive to subsequently added control variables. Therefore, we believe that the probability of a continued correlation between the core variables and residuals is very low. Finally, we employed different classification criteria, such as using multiple standards to distinguish between large and small farmers, to illustrate the heterogeneity, which ensured the robustness of the main findings. These tests significantly enhance the robustness of estimated results.

Grain crops versus cash crops

Differences in growth characteristics and management practices mean that the effects of climate change on grain crops and cash crops significantly differ (Zhang et al. 2018). To account for these differences, we estimated Eq. (5) separately for different kinds of crops; the results are presented in Fig. 4a, b. In addition to grain crops, such as wheat, rice, corn, soybeans, and potatoes, we also studied cash crops, including cotton, oilseeds, sugar, hemp, tobacco, sericulture, vegetables, and fruits. This is a valuable extension of existing studies which have mainly focused on grain crops (Wang et al. 2009, 2014a, 2014b). Our results indicate that, generally, the effect of temperature on cash crops is significantly greater than on grain crops. Grain crops were found to respond significantly only to T_max changes and were not sensitive to T_min and T_ave. Conversely, winter and summer temperatures were favorable for cash crops, while the opposite was true for spring and fall. Precipitation had a negligible effect on cash crops but impacted grain crops. Our findings suggest that current policies may place too much emphasis on the importance of grain crops, and show that paying more attention to the impact of climate change on cash crops, which are also important agricultural commodities, is necessary.

Fig. 4: Impact of climatic factors on different crops and different sizes of farms.

a, b Represent the effects of climate factors on cash and grain crops, respectively. c, d Represent the impacts of climate factors on large-scale and small-scale farms, respectively. Grain crops include wheat, rice, corn, soybeans, and potatoes, while cash crops include cotton, oilseeds, sugar, hemp, tobacco, sericulture, vegetables, and fruits. We use very strict classification criteria. If a farmer only grows cash crops and not grain crops, we classify them as a cash crop farmer. Conversely, if a farmer only grows grain crops and not cash crops, we classify the theme as a cash crop farmer. As of 2021, smallholder farmers account for over 98% of the total number of agricultural operators in China. Approximately 210 million smallholder farmers operate on less than 10 acres of cultivated land, where 1 mu is equal to 666.67 m². The definition of a smallholder farmer varies by region, and no clear standard exists for how many acres of operating area constitute a smallholder. For example, smallholder farmers in Hebei Province, China, typically operate on 600–1200 m² of land, whereas in Heilongjiang Province, the per capita arable land area is 8571 m². We primarily report our results using a delineation criterion of 10 mu (6670 m²). However, we also conducted robustness analyses using different criteria, including 5, 20, and 30 mu. The results can be found in supplementary materials.

Large-scale farms versus small-scale farms

The abilities of farmers of different sizes to cope with climate change significantly differ. Smallholders are generally considered more vulnerable to climate change impacts (Donatti et al. 2019; Mishra et al. 2018). To investigate these differences, we divided our sample into two groups—large- and small-scale farmers. Figure 4c, d reports the differential effects of temperature and precipitation on the net revenue of farmers with different scales. Our findings suggest that, generally, the overall impact of T_max and T_ave and precipitation is more pronounced for large farmers than for small farmers. This is because larger farming operations are exposed to higher climate risks. Specifically, we observe that the rising T_max and T_ave are favorable for both small- and large-scale farmers in winter and summer, but the opposite is true in spring and autumn. Moreover, small-scale farmers are more sensitive to T_min than large-scale farmers. Winter precipitation is clearly unfavorable for both small- and large-scale farmers, while spring precipitation increases cropping returns for all farmers. However, increased precipitation in autumn is unfavorable, with this disadvantage being more pronounced for small farmers. Overall, our results suggest that the impacts of climate change on crop production and revenue are unevenly distributed across farmers of different scales. Both small and large households require additional support and resources from the government to adapt to changing climate conditions.

Regional heterogeneity

Both climate change and crop planting patterns significantly differ across regions. Excluding Tibet, the areas are divided into six regions based on climate and economic characteristics (e.g., North China, Northeast China, Northwest China, East China, South China, and Central China). As shown in Fig. 5, the impact of seasonal temperature and precipitation on crop net revenue varies significantly across regions in China. This finding is consistent with Wang et al. (2009), who also found that both temperatures and precipitation have different seasonal impacts on farmers’ production in different regions of China. Specifically, both T_max and T_ave increases in winter and summer positively impact crop returns. Adverse effects of T_max in autumn are pronounced in southern regions, particularly in southern China, while T_min in autumn and T_max in winter have a beneficial effect. Increased precipitation in the fall is most unfavorable to eastern and central China. Overall, the adverse impacts of precipitation are mainly concentrated in winter and autumn. This is consistent with the spatial distribution of harvesting time in China’s cropping production, as autumn is the primary harvesting season, and excessive precipitation during this period can be detrimental to crop harvesting. However, spring precipitation is beneficial to crop net revenue in the north, as it can alleviate the drought in spring that is often experienced in this region. The regional variation results indicate that targeted agricultural policies are necessary to address specific climatic challenges in each region during crop growing and harvesting periods.

Fig. 5: Regional and seasonal distribution of temperature and precipitation effects.

a–f Represent the impact of climate factors on crop production in North China, NorthEast China, NorthWest China, East China, South China, and Central China, respectively. Due to the distinct climate change characteristics of the Tibetan region, Tibet was not included in our sample. To analyze the impact of climate change on crop net revenue in different regions of China, we grouped the provinces into six regions. Specifically, North China comprises Beijing, Tianjin, Hebei, Shanxi, and Inner Mongolia; Northeast China comprises Heilongjiang, Jilin, and Liaoning; East China includes Jiangsu, Shandong, Zhejiang, Anhui, Jiangxi, Fujian; South China comprises Guangdong, Guangxi, Hainan; Central China comprises Henan, Hubei, Hunan, Sichuan, Guizhou, Yunnan, and Chongqing; and the Northwest China region comprises Shaanxi, Gansu, Qinghai, Ningxia, and Xinjiang. Our estimation results indicate that the effects of temperature and precipitation on net returns do not significantly differ in Southwest and South China. Therefore, we grouped the Southwest provinces into South China for our analysis.

Simulation results

For climate projections, we use two SSP emission scenarios (SSP245 and SSP585) to analyze the future impact of climate change on farmers’ crop net revenue. To propose adaptation policies, measuring the impact of marginal changes in future climate conditions on future crop net revenue in China is important (Yang et al. 2021). We assume that future technological changes will not cause abrupt changes in crop yields and that such effects may be reduced if future technological advances are considered. To model the impact of climate change on crop net revenue, we calculate the effect of marginal changes in future temperature and precipitation on future net revenue using methodologies established in prior research (Trinh et al. 2018; Wang et al. 2009). Figure 6 illustrates the effect of changes in temperature and precipitation on net profits in agriculture for two scenarios, SSP245 and SSP585, considering both adaptative and non-adaptative impacts. In Fig. 6a, the temperature rise negatively affects net profits in agriculture, with this trend appearing to level off after 2050. It is noteworthy that adaptive behavior markedly reduces this negative impact. In Fig. 6b, the future impact under the SSP585 scenario is mostly negative, with the negative impact increasing linearly. In Fig. 6c future precipitation has a mostly negative impact on net revenue under the SSP245 scenario, despite substantial variability. Mitigating the negative impact of future precipitation is possible through adaptive behavior. Figure 6d shows the effect of precipitation on net profits in agriculture under the SSP585 scenario. Before 2050, the impact is negligible, but as time progresses, the negative impact of precipitation on net profits becomes increasingly unfavorable. Overall, the most important insight is that rising temperatures and precipitation negatively affect agricultural net profits in both SSP245 and SSP585 scenarios, with adaptive behavior playing a crucial role in mitigating these impacts.

Fig. 6: Estimation of the influence of future climate change on the crop net revenue.

a, b Represent the impact of future temperature change on crop net revenue under SSP245 and SSP585 scenarios, respectively. c, d Represent the impact of future precipitation change on crop net revenue under SSP245 and SSP585 scenarios, respectively. The data used here was sourced from NESM3. Specifically, we extracted the daily maximum and minimum temperatures as well as precipitation data in China, under both the SSP245 and SSP585 scenarios. To derive the average temperature, we divided the maximum and minimum temperature phases by two. We computed the variance of future temperature and precipitation in comparison with the 1990–2020 average. Finally, we multiplied this variance by the previously estimated marginal coefficients or elasticities.

Discussion

Cropping plays a vital role in global food security, and the impact of climate change on crop production is a pressing concern for scholars and policymakers. Previous Ricardian analyses conducted in China mainly relied on cross-sectional data, which may have resulted in omitted bias, and neglected the role of adaptation in mitigating the effects of climate change. Our study aims to address these issues by employing the panel Ricardian model to estimate the sensitivity of China’s crop net revenue to climate change. We utilized a 17-year panel dataset and employed the Hsiao two-step method to obtain more accurate results. Furthermore, we improved our model settings by considering the regional heterogeneity of market shocks and the correlation between temperature and precipitation.

Our results indicate that the effects of T_max and T_min on cropping production significantly, which has been overlooked by previous studies (Chen and Chen, 2018; Nguyen and Scrimgeour, 2021; Wang et al. 2009; Yao et al. 2007). Rises in T_max have unfavorable effects in spring and fall, while the increases in T_min have positive effects in fall. These findings suggest that the differential effects of T_max and T_min should be distinguished in promoting agricultural adaptation to climate change in different seasons. The adverse impact of fall precipitation is prominent, as excessive fall precipitation can affect harvest. Crop growth has typical seasonal characteristics and differently reflects the changes in temperature and precipitation in different seasons, encouraging farmers to adopt different measures to adapt to climate change (Ye et al. 2019). We further found that adaptation is effective in mitigating the adverse effects of temperature and precipitation, which is consistent with previous findings (Burke and Emerick, 2016; Chatzopoulos and Lippert, 2015; Kurukulasuriya and Mendelsohn, 2008). Therefore, emphasizing farmers’ adaptation behavior can increase agricultural profitability and ensure food security.

We also analyzed the responses to climate change across different production entities, crops, and regions, which is crucial for enhancing the effectiveness of policies. The effects of changes in temperature and precipitation significantly differ between cash and grain crops, which has not been distinguished in existing studies (Piao et al. 2010; Yao et al. 2007). Temperature affects cash crops more than grain crops, while grain crops are more sensitive to precipitation than cash crops. Compared with cash crops, grain supply is the core of ensuring China’s food security. The central government has invested heavily in the construction of farmland irrigation facilities, which can ensure that grain production can be irrigated in time in the face of high temperatures and effectively mitigate its adverse effects (Li, 2012; Liu et al. 2013). However, for cash crops, the government has invested little in the construction of irrigation facilities, which leads to the obvious impact of high temperatures. We also examined the differences in the response of large-scale and small-scale farmers to climate change, considering the relationship between the scale of agricultural operations and resilience to climate change. Our study found that the impact of climate change on large-scale farmers is more pronounced than on small-scale farmers, because large-scale farmers may be at greater risk of economic losses from climate change, and their adaptation costs will be higher than those of small-scale farmers. Although China’s crop production is dominated by small farmers (Piao et al. 2010), land transfer will increase the scale of operation and create many large farmers in the future, making climate risk more concentrated and amplified. Therefore, while the government is concerned about the impact of climate change on crop production of small-scale farmers, it should also pay more attention to large-scale farmers. Additionally, the effects of temperature and precipitation on crop net revenue are regionally heterogeneous. We divided different regions into greater detail than previous studies (Wang et al. 2014a, 2014b), indicating that policies and technologies to address the adverse impact of climate change should be more accurate in different regions.

Projections of future climate change impacts show that the effects of higher temperatures gradually diminish when adaptation is considered. Many forecasts of future impacts of climate change have shown pessimistic conclusions (Agnolucci et al. 2021; Iizumi et al. 2021). However, some studies have also pointed out that there are regional differences in the uncertainty of wheat yield forecasts in future climates (Wang et al. 2021), and the chance that crops, such as European soybeans, will be self-sufficient in the future is high (Guilpart et al. 2022). This is likely because the main purpose of current adaptation measures is to mitigate the adverse effects of high temperatures (Chen and Chen, 2018), including the research and development of crop varieties, such as heat-, drought-, and pest-resistant crops, as well as the construction of agricultural infrastructure and the promotion of water-saving agriculture. In the future, we should focus on promoting these measures to increase agricultural resilience to climate change.

Conclusion

Sustainable crop production is the key to global food security. In recent years, the impact of climate change on crop production has become a hot issue in China and even around the world. Accurately estimating the impact of climate change on farmers’ crop production is the foundation for effective adaptation in agriculture. Based on panel data from ~180,000 households in China over the past 17 years, the panel Ricardian model and Hsiao two-step method were employed to more comprehensively estimate the impact of climate change (temperature and precipitation changes) on net crop income. Our conclusions are as follows.

First, the effects of temperature and precipitation changes on crop net revenue vary among different seasons. Therefore, the seasonal change of climate factors in the process of agricultural response to climate change must be considered. Policymakers and agricultural producers should fully raise awareness of the differentiated impact of seasonal changes in climate factors (temperature and precipitation) on agricultural production. In addition, the agricultural sector should develop different adaptive measures based on seasonal changes in temperature and precipitation to enhance the ability of agriculture to cope with climate change.

Second, farmers’ adaptive behaviors can reduce the negative impact of temperature and precipitation changes on crop net revenue. That means that promoting farmers’ adaptive behavior is the top priority to ensure the livelihoods of rural households under climate change. Therefore, the government should promote the dissemination of information on the agricultural sector’s response to climate change by organizing specialized training, distributing materials, and utilizing online media, such as radio and television, thus enhancing farmers’ awareness and ability to cope with climate change. Additionally, the government can reduce the application cost of adaptation measures through policy subsidies and technical practical guidance. Consequently, farmers can be encouraged to adopt climate change adaptation measures, such as drought-tolerant crops, water-saving irrigation, and changing planting structures, to ensure crop production under climate change.

Third, the effects of temperature and precipitation changes on crop net revenue particularly vary across different crop types and farm sizes. Specifically, temperature and precipitation changes have greater impacts on the net income of large-scale households and farmers who cultivate cash crops. However, the current agricultural policies in China tend to support small-scale households and farmers of grain crops, which limits the ability of large-scale households and farmers of cash crops to cope with climate change to some extent. Therefore, the government should increase subsidies and support (such as agricultural insurance policies and investment in farmland infrastructure) for large-scale households and farmers planting cash crops to ensure the livelihood of small-scale households and enhance their adaptive ability to deal with climate change.

Although our analysis significantly contributes to existing research, there are opportunities for further research to advance understanding. We assume the constant state of agricultural technology to simulate the effects of future climate change. Consequently, our estimates of the impacts of climate change do not include technological changes in future crops or farming techniques. Moreover, as with most Ricardian analyses, our study implicitly assumes that farmers will adapt to a changing climate through various approaches and measures. Future research could investigate these specific approaches and measures, as well as the response of China’s agricultural system to climate change, such as changes in land allocation, which will help to better understand climate change impacts and their impact on policy.