Weihong HAN, Xianfeng YANG, Xinjiletu YANG, Yuqing LIU

Inner Mongolia University of Technology, Hohhot 010051, China

Abstract ENSO-driven extreme weather events such as droughts and floods can cause significant damage to agricultural production and different intensities of ENSO events make uncertain changes to rural residents’ welfare in different regions. To emphasize this uncertainty, the stochastic CGE model is constructed by imbedding a stochastic parameter into the production module to analyze the impacts of ENSO on the welfare of rural residents in various regions of China and the volatility of uncertain ENSO events on the welfare of residents. The role of agricultural technology in improving welfare stability and transfer payments in reducing welfare losses from ENSO are also examined. The results show that weak ENSO events have little effect on the welfare of rural residents while strong ENSO events cause the welfare of rural residents a significant decline, and the largest decrease appears separately in the southwest region and the smallest one in the northeast. The uncertainty of ENSO events seriously affects the stability of the welfare of the residents, with the average fluctuation level of 200% in the change of the rural residents’ welfare in all regions under El Nio. Technically improving the anti-risk ability of agriculture can effectively reduce the fluctuation of residents’ welfare. Besides, if the government increases the transfer payment to the rural resident for disaster relief, the welfare would increase, and the higher the payment, the greater the improvement of the welfare.

Key words ENSO events, Stochastic CGE, Rural residents’ welfare, Uncertainty

1 Introduction

In recent years, extreme weather events frequently occur around the globe and ENSO (El Nio-Southern Oscillation) events are closely related to the occurrence of extreme weather[1-2]. ENSO is a fluctuation of the wind field and sea surface temperature that occurs in the equatorial eastern Pacific. It is characterized by El Nio and La Nia phenomena in the ocean and Southern Oscillation in the atmosphere. These phenomena have different impacts on different regions of the world through complex conduction mechanisms, which are mainly reflected in the abnormal changes in precipitation and temperature in different regions, resulting in frequent extreme weather events and sea level rise[3-4].

Agriculture is one of the sectors sensitive to climate change[5-6]. Abnormal temperature and precipitation caused by ENSO events can directly affect crop yields, further impacting the agricultural economy and threatening food security[7]. In recent decades, the impacts of droughts, floods and precipitation changes on agriculture have received increasing attention under the background of food security[8-10]. However, little attention has been paid to the changes in the welfare of rural residents after the ENSO events caused an impact on agricultural production. In fact, ENSO, as a climate phenomenon, increases the frequency of extreme weather events, resulting in reduced yields of crops that are more sensitive to climate change, further causing supply shortages and higher prices for many agricultural products[11]. Agricultural crop failure and product price fluctuations make rural households whose income relies mainly on agriculture production more vulnerable to the negative impact of extreme weather, resulting in the loss of real welfare[12-14].

The effect of ENSO on crop yield varies with ENSO phases, geographic location and crop type[1,15]. From the global perspective, El Nio may increase the global average soybean yield, but the impacts on maize, rice and wheat yield are uncertain; global average yields for all four crops in La Nia years are below normal[16]. Shuai Jiabingetal.[17]analyzed the impact of ENSO events on maize production in northern and southern China. In El Nio years, yields increased in most northern regions while decreasing in parts of the South; and in La Nia years, yields decreased significantly in the north and northeast while generally increasing in the south. The research of Shuai Jiabingetal.[18]showed that ENSO had significant effects on wheat, rice and maize yields: in southeastern China, the increase or decrease in wheat yield was mainly determined by the amount of annual El Nio (La Nia) precipitation; rice yield in northeast China was mainly related to ENSO-induced temperature changes; the changes in temperature, solar radiation and precipitation caused by ENSO all have significant effects on maize yield in the north. In summary, different crops have different sensitivities to temperature and precipitation, and therefore different crops show different effects from ENSO events. Among the six major food crops in Asia and the Pacific, maize is the most sensitive to ENSO events[19]. Liu Yuanetal.[20]suggested that in the North China Plain, maize production was more vulnerable than wheat production due to the reduction in annual precipitation from El Nio (La Nia). Qian Yonglanetal.[21]assessed the risk of ENSO events on wheat, soybean, rice and maize. The results showed that wheat was the most vulnerable to ENSO events, and rice is at far lower risk than the other three crops in ENSO events, especially in La Nia. In fact, the effect of ENSO on the yield of various crops in different regions is mainly due to anomalous fluctuations in temperature and precipitation.

In China, most rural residents are engaged in agricultural production, which is directly related to the income and consumption of rural residents. That is an important channel through which climate shocks affect residents’ welfare is the ENSO-price linkage[22]. In terms of the income, ENSO-related weather anomalies disrupt agricultural activities and decrease crop yields in agriculture, which can result in significant social and economic costs, higher food prices and lower rural incomes[13,23-24]. At the same time, the decrease in income will force the purchasing power of rural residents to curtail, and the increase in commodity prices caused by the ENSO shock will also limit residents’ consumption to some extent. For example, weather anomalies caused by ENSO will not only bring severe impact to the agricultural production sector, but also disrupt the production of non-agricultural sectors in the short term, triggering price volatility for a large number of commodities[22]. Ramgovindetal.[25]found that ENSO led to an increase in the price of basic foodstuffs, which had an impact on the lives of South African farmers. Cashinetal.[26]also found that the impact of El Nio would increase the prices of energy and non-energy commodities, causing short-term inflation in most countries. Therefore, ENSO events, in a direct or indirect manner, mostly have a negative impact on the living standards of residents and damage the welfare of residents, especially rural residents. Meanwhile, China’s agricultural production is widely distributed, with a complex structure and large total agricultural output. The overall level of economic development and agricultural development varies greatly from region to region, as does the ability to risk defense, and the extent to which rural living standards are affected by climate inevitably varies from region to region. It is important to take regional differences into account when exploring changes in the welfare of Chinese rural residents following the effects of ENSO events on agricultural production.

Continual exposure to unanticipated extreme events is a contributing factor for global crop loss, and the uncertainties associated with extreme weather or the disasters caused by ENSO events, which have occurred frequently in the last 70 years, have serious implications for agriculture[21]. This uncertainty will not only affect the stability of rural residents’ income, but also change people’s estimates of market prices, transmitted to rural residents through the economic system, resulting in changes in the physical value of residents’ welfare. The research points out that the income instability of rural residents is the factor that directly affects the consumption of rural residents[27]. In other words, the uncertainty of residents’ labor income will trigger residents’ precautionary saving tendency, which will limit residents’ consumption[28]. These will have an uncertain impact on the welfare of rural residents. Moreover, the welfare level of rural residents is lower than that of urban residents, and the ability to cope with climate shocks is obviously insufficient[29-30]. Many studies have shown that establishing early warning systems and improving agricultural technology can effectively reduce the risk of agricultural losses and have a positive impact on farmers’ welfare[31-32]. Therefore, we should pay more attention to the fluctuation of this uncertainty on the welfare of rural residents in order to improve the stability of their welfare. Computable General Equilibrium (CGE) models are widely used to solve various climate change mitigation problems at regional, national and international levels[33]. Stochastic CGE models, combining CGE models with stochastic components, are used to analyze uncertainty in climate change and economic impact after stochastic shocks such as droughts[34-36]. For example, Solaymani[13]used an integrated method comprising of a CGE model and a stochastic method, and analyzed the impacts of the changes in rainfall and temperature simultaneously on food safety in Malaysia. Siddigetal.[37]introduced stochastic shocks into a single country dynamic CGE model to explore the economic impact of climate change in Sudan. Therefore, the application of the stochastic CGE method is appropriate for the problem to be analyzed in this article.

The existing literature mainly focuses on the negative impact of ENSO on agricultural production, or the overall economic impact on only one country or region, rather than to compare the changes in the welfare of rural residents in different regions after the ENSO shock to agriculture. Besides, previous studies focused on assessing the impact of a particular ENSO event, while ignoring the fluctuations in economic variables due to the uncertainty of ENSO events in realistic situations. Therefore, based on the standard CGE model, this paper expands and constructs the stochastic CGE model by imbedding an uncertain parameter into the production module. Taking the geographical location and crop type as the criteria for a regional division of China, this paper not only discusses the impacts of different intensities of ENSO events on the welfare of rural residents in different regions of China, as well as the fluctuation level of residents’ welfare caused by the uncertainty of ENSO events, but also investigates the role of agricultural technology in improving welfare stability and transfer payments in reducing welfare losses from ENSO. Through analyzing the impacts of ENSO events on the welfare of rural residents from the perspective of regional differences, it is expected to help the government to develop regional and differentiated agricultural policies to mitigate the negative impacts of climate change on agricultural producers. The stability of rural residents’ welfare in turn solidifies agricultural development and ensures national food security.

2 Methodology

2.1 Stochastic componentsGenerally, ENSO events are classified as El Nio in the warm phase and La Nia in the cold phase. According to the Nino 3.4 index [an ENSO event can be identified if the absolute value of the 3-month running mean Nio 3.4 index (to one decimal point) reaches or exceeds 0.5 ℃ and lasts for at least 5 months], 33 ENSO events have occurred since 1951, including 19 El Nio events and 14 La Nia events. The probability of El Nio is slightly higher than La Nia. ENSO event intensity levels are classified as weak, moderate, strong and super strong events according to China national standardIdentificationmethodforElNio/LaNiaEvents(GB/T 33666-2017). In this paper, weak events are classified as weak ENSOs, while medium, strong and super strong are uniformly classified as strong ENSOs. The uncertainty of agricultural production based on historical data can be explained by yield variation, which is mainly caused by weather fluctuations[38]. Therefore, to describe the effect of stochastic shocks of ENSO-induced weather on crop yields, an exogenous stochastic componentxGis incorporated. The calculation ofxGuses yield data for seven major crops in China from 1951-2018 to reflect the impact of ENSO events on agricultural production efficiency in a stochastic analysis. The error in crop yields is assumed to be log-normally distributed, and the formula forxGis as follows.

(1)

Table 1 Influence coefficient of various agricultural activities under weak El Nio phenomenon

Table 1 Influence coefficient of various agricultural activities under weak El Nio phenomenon

RIWHMASOCOPERAAGNE0.989 2650.989 8610.990 5640.997 978-0.997 0690.994 6360.995 726 NC0.998 3200.992 1510.991 3490.993 9951.003 3020.991 9180.963 8280.995 229 EC0.999 0310.866 8930.999 2961.004 2611.004 3360.995 6010.984 9570.997 862 SC0.998 955-0.925 5830.985 911-0.996 4681.031 2471.000 014 MYE0.999 6760.995 4470.992 9470.997 2721.012 2351.000 4701.114 8490.997 015 MYA0.997 6430.993 0710.998 3781.001 9061.005 7840.999 7090.999 3500.997 947 SW1.000 1850.990 6960.998 1741.002 8761.031 0071.000 4460.996 2720.999 454 NW0.998 4510.997 8150.996 6220.989 1680.992 6791.031 1330.997 9110.997 880

Note: RI: rice, WH: wheat, MA: maize, SO: soybean, CO: cotton, PE: peanut, RA: rapeseed, AG: other agricultural products and services; the data are calculation results of GAMS program, the same in Table 2-4.

Table 2 Influence coefficient of various agricultural activities under strong El Nio phenomenon

Table 2 Influence coefficient of various agricultural activities under strong El Nio phenomenon

RIWHMASOCOPERAAGNE0.985 6410.986 2370.986 9400.994 355-0.993 4450.991 0120.992 102 NC0.986 5370.980 3680.979 5670.982 2120.991 5190.980 1350.952 0450.983 446 EC0.952 3150.820 1760.952 5800.957 5450.957 6200.948 8850.938 2400.951 145 SC0.964 057-0.890 6850.951 013-0.961 5700.996 3490.965 116 MYE0.958 4160.954 1860.951 6870.956 0120.970 9740.959 2101.073 5880.955 754 MYA0.993 9940.989 4230.994 7300.998 2571.002 1350.996 0610.995 7020.994 299 SW0.988 0640.978 5750.986 0530.990 7561.018 8860.988 3260.984 1520.987 333 NW0.985 5380.984 9010.983 7090.976 2550.979 7651.018 2200.984 9970.984 966

Table 3 Influence coefficient of various agricultural activities under weak La Nia phenomenon

Table 3 Influence coefficient of various agricultural activities under weak La Nia phenomenon

RIWHMASOCOPERAAGNE1.033 7401.005 1311.018 5071.006 918-1.031 8540.972 3461.012 817 NC1.021 2521.020 4161.022 0690.997 3721.016 0901.028 7321.613 3721.011 519 EC1.007 5730.892 4901.011 5871.007 6230.978 7431.026 5071.015 0881.006 004 SC1.007 970-1.174 4751.025 764-1.022 0531.087 2841.003 507 MYE1.024 1871.015 4221.024 7461.016 3791.030 4861.023 8301.174 2061.012 194 MYA1.009 2951.013 4801.024 4671.013 9971.010 4791.046 1211.024 7431.008 418 SW1.006 6521.004 5091.014 0781.025 2030.943 9261.029 1811.031 8591.006 127NW1.022 8601.013 7011.028 9911.038 0061.055 8290.482 9841.053 2051.014 988

Table 4 Influence coefficient of various agricultural activities under strong La Nia phenomenon

Table 4 Influence coefficient of various agricultural activities under strong La Nia phenomenon

RIWHMASOCOPERAAGNE1.013 1080.984 4980.997 8740.986 285-1.011 2220.951 7130.992 184 NC0.810 1190.809 2830.810 9350.786 2380.804 9560.817 5991.402 2390.800 386 EC0.964 7170.849 6330.968 7300.964 7670.935 8860.983 6510.972 2320.963 147 SC0.941 476-1.107 9820.959 270-0.955 5591.020 7910.937 013 MYE0.969 7010.960 9370.970 2610.961 8930.976 0010.969 3441.119 7200.957 708 MYA0.996 5751.000 7601.011 7481.001 2770.997 7591.033 4011.012 0230.995 698 SW0.978 6910.976 5470.986 1160.997 2410.915 9651.001 2191.003 8970.978 165 NW0.827 9230.818 7650.834 0550.843 0700.860 8920.288 0480.858 2680.820 052

Then, by adding the stochastic variable obtained by the measurement method afterxGto explain the uncertainty of ENSO events on agricultural production activities[39].

x=xG+∑TG·RV0G

(2)

There are covariate relationships between production activities in a region, andRV0Gis a stochastic variable for each agricultural activity and it is designed to describe the part that cannot be predicted in ENSO events. The changes in the randomness of ENSO events are expressed by controlling the probability characteristics ofRV0G. Initially,RV0Gis assumed to conform to the standard normal distribution.TGis the square root of the variance-covariance matrix of the regional agricultural activity which is obtained by Cholesky decomposition.

2.2 CGE model frameworkBased on the standard CGE model[40], the stochastic CGE model is constructed by imbedding a stochastic parameter into the production module, which mainly includes: production, trade, income and expenditure, welfare, investment and savings, market modules (Fig.1). Agricultural production activities are subject to exogenous shocks from the ENSO event, which will be directly reflected in changes in supply and demand leading to changes in prices of commodities and factors of production. Price changes in turn affect economic activities, such as income, consumption, investment, revenue and international trade[41].

Fig.1 Framework of the stochastic CGE model

The production module, which simulates the quantitative relationship between input and output, is the basis of the generation of social and economic value. As an exogenous shock, the ENSO event directly acts on the production function and then affects the final output of the production link. The module is described by using the two-layer nested constant elasticity of substitution (CES) production function, which takes labor, capital and intermediate inputs as production input, where producers minimize costs as production principle. Only the first layer of the CES production function is given below, and the same applies to the second layer. The specific equation is set as follows:

(3)

(4)

PAa·QAa=PVAa·QVAa+PINTAa·QINTAa

(5)

QINTca=icac, a·QINTAa

(6)

PINTAa=∑cicac, a·PCc

(7)

The income and expenditure module mainly describes the income and expenditure activities of residents, enterprises and the government, in which saving is equal to the balance of income minus consumption. Residents’ income is mainly derived from their labor income, capital income and transfer payments from the government and enterprises. The consumption of residents is determined by the disposable income and the marginal propensity to consume. There are 16 resident categories in the model, differentiated by region (eight regions) and regional affiliation (rural and urban).The income of enterprises mainly includes capital income and government transfer payments, and enterprise expenditure comprises income tax, value-added tax, other taxes and transfer payments to residents. Government revenue comes from various taxes, including value-added tax, income tax for residents and businesses, tariff and other taxes. Government expenditure is mainly government consumption and transfer payments to businesses and residents. Total investment in social production comes from total savings, including savings of residents, enterprises, and government, as well as foreign savings.

In this paper, the consumption of residents is used a measure of welfare variables, the monetary unit to detect changes in residents’ welfare or utility. Two indicators, equivalent variation (EV) and compensatory variation (CV), are used to measure the changes in residents’ welfare from the initial level to the ENSO shock. CV refers to the increase or decrease of consumers’ income in order to keep the original effect level under the new price level; EV refers to how consumers’ income should be adjusted to achieve the current utility level at the original price level. This approach has been applied to empirical welfare research in a variety of fields[42-44]. The welfare module for calculating residents’ welfare:

(8)

(9)

u(QHc, h)=∏c(QHc, h)shrhc, h

(10)

(11)

The trade module reflects the substitution relationship between domestic output and foreign departments in the process of commodity trade. Domestic output is divided into domestic products for domestic sale and export commodities in the form of the CET transformation function, while domestic products for domestic sale and imported commodities form the total domestic demand based on the Armington hypothesis.

After the market reaches equilibrium, the aggregate supply equals the aggregate demand. In addition, the supply of capital and labor is sufficient, and investment and employment are determined by demand, so the Keynesian closure principle is adopted.

2.3 Social accounting matrixThe data basis for CGE model operation is Social Accounting Matrix (SAM), and the specific data sources mainly includeChinaRegionalInput-OutputTable(2017),FinanceYearbookofChina(2018),ChinaStatisticalYearbook(2018),andChinaTaxationYearbook(2018). To create a micro-SAM table, it is necessary to divide the regions, production sectors and residents.

The ENSO phenomenon has a wide range of effects on China, depending on the season and region in which it is present. Generally, El Nio winters are warmer, and summers are prone to flooding in the south and drought in the north. Most of North China is experiencing dries condition, while most of Southeastern and South China has shown positive precipitation anomalies in summer[45-46]. According to the records of natural disasters, severe floods occurred in the middle and lower reaches of the Yangtze River in 1931, 1954 and 1998, which are the years of the strong El Nio. During the La Nia phase, these anomalies reverse signs[47]. China is prone to "colder winters and hotter summers", that is, the temperatures in winter are lower than normal year, and the summer are higher. In terms of precipitation, China suffers a typical "southern drought and northern flooding" anomalous climate. In addition, La Nia phenomenon often causes winter weather anomalies, such as severe snow disasters in some parts of the south in 2008. Therefore, in this paper, considering the differences in climatic factors, China is divided into eight regions: northeast (NE, including Heilongjiang, Liaoning, and Jilin), northern coast (NC, including Beijing, Tianjin, Hebei, and Shandong), eastern coast (EC, including Shanghai, Jiangsu, and Zhejiang), southern coast (SC, including Fujian, Guangdong, and Hainan), middle Yellow River (MYE, including Shanxi, Inner Mongolia, Henan, and Shaanxi), middle Yangtze River (MYA, including Anhui, Jiangxi, Hubei, and Hunan), southwest (SW, including Guangxi, Chongqing, Sichuan, Guizhou, and Yunnan), and northwest (NW, including Tibet, Gansu, Qinghai, Ningxia, and Xinjiang) for analysis.

To highlight the impacts of the ENSO event on the agricultural sector, the crop industry, which has the largest share in China’s agriculture, is broken down into seven major crops. Based on the division of production departments by region, the agricultural sector is subdivided into rice (RI), wheat (WH), maize (MA), soybean (SO), cotton (CO), peanut (PE), rapeseed (RA) and other agricultural products and services (AG), and sectors other than agriculture are combined into industry (IN) and services (SE). Since rural residents are engaged in production activities mainly in agriculture, while urban residents are mainly in industry and services, the residents of each region are divided into urban and rural areas.

The specific data of the seven crops in the micro SAM table are obtained from theChinaRuralStatisticalYearbook(2018), theChinaYearbookofHouseholdSurvey(2018), theAlmanacofChina’sFinanceandBanking(2017), theChinaFoodDevelopmentReport(2018) and the official website of the General Administration of Customs, PRC. After the required data are calculated and sorted out, the SAM table is balanced.

2.4 Scenario settingThis paper focuses on exploring the impact of ENSO shocks of different types and intensities on the welfare of rural residents in different regions and how to reduce this impact. Therefore, the baseline scenario (S0) simulates steady-state rural residential welfare in the absence of the ENSO event and other extreme weather events in 2017, and serves as a baseline for comparison with the impacts of the ENSO event.

Firstly, scenarios 1 (S1) and scenarios 2 (S2) denote the actual changes in the welfare of rural residents after a weak ENSO event and a strong ENSO event have affected agricultural production, respectively. And differences in historical crop yields are used to reflect ENSO variability in this stochastic analysis[37]. Secondly, the ENSO impacts are quite erratic and considered very difficult to make accurate prediction[48]. Scenario 3 (S3) simulates the volatility of ENSO uncertainty on the welfare of rural residents, and validates the extent to which improving risk warning technology would mitigate welfare volatility from ENSO events. Assuming that the logarithm of the random variable is normally distributed, the improvement in the technology of agricultural resistance to risk is expressed by changing the values of its expectation and variance[39]. Finally, ENSO events generally cause a decline in the welfare of rural residents in the previous analysis. In order to protect the welfare of the residents, governments often resort to increasing transfer payments to counteract the negative effects of random shocks like natural disasters on the residents[49]. Scenario 4 (S4) describes that under a strong ENSO event, the government takes measures to increase transfer payments to compensate residents’ losses of agricultural production. The specific parameters of the scenario settings are as follows: the value ofxGis 1; the values ofxGare shown in Table 1 and Table 3, the values ofxGare shown in Table 2 and Table 4; under uncertain ENSO events,xGtakes the same values as S2 andRV0G-N(0,1); uncertain ENSO events after the upgrade of agricultural anti-risk technology,xGtakes the same values as S2 andRV0G-N(0.2, 0.5); government transfers to rural residents are increased by 4%, 8% and 12%, respectively.

3 Results and discussion

3.1 Impacts of El Nio on the welfare of rural residents

Comparing Fig.2 and Fig.3, in general, El Nio phenomenon damages rural residents’ welfare, and different intensity of El Nio phenomenon has different influences on rural residents’ welfare in different regions. As the price elasticity of demand for agricultural products is less than 1, residents’ income will increase in the case of crop failure. Although residents’ income has been temporarily improved, the overall market price level of products has risen, which exceeds the degree of income increase, resulting in the decline of the actual income level of rural residents. Therefore, it has a negative effect on the welfare of rural residents. (i) In terms of regional differences, it is the southwest that experiences the largest decline in rural residents’ welfare, followed by the middle reaches of the Yellow River and the Yangtze River, and the northeast the smallest, both under weak and strong El Nio. According to statistics, among the 31 ENSO events that occurred during 1960-2016, the probability of drought and flood disasters in the same year or the next year in the southwest was as high as 93.55%[50]. This indicates that the southwest region shows a stronger sensitivity to ENSO events, so the agriculture and residents’ welfare in this region are more vulnerable to the effects of ENSO compared to other regions. The middle reaches of the Yellow River and the middle reaches of the Yangtze River have a large share of agriculture due to their flat terrain, moderate climate and abundant water resources, which make them suitable for agriculture. Additionally, ENSO events tend to trigger droughts in the Yellow River basin and extreme precipitation in the Yangtze River basin[51], and the occurrence of extreme weather increases the risk of agricultural production losses. These make the agriculture in these two regions more negatively affected by ENSO events and the welfare of the rural residents decreases more. (ii) From the point of view of the intensity of El Nio, the negative impact of strong El Nio on the welfare of rural residents in each region is greater, 5-8 times greater than that of weak El Nio. Under the strong El Nio, the welfare decline of rural residents in the southern coastal and southwestern areas is about eight times greater than under the weak El Nio, and the dropped in the northern coastal areas is the least 5.6 times. At higher El Nio intensities, agricultural yields fall more and the real income levels of rural residents decrease more. The economic shock of a higher intensity ENSO event causes the price level of the product market to rise more and therefore the welfare of the residents to fall more.

Fig.2 Welfare changes of rural residents in a weak El Nio phenomenon

3.2 Impacts of La Nia on the welfare of rural residentsAs shown in Fig.4 and Fig.5, the La Nia phase of ENSO also has negative consequences for the welfare of rural residents, and different intensity of La Nia has different effects on residents’ welfare in various regions. (i) From the regional perspective, it is consistent with the El Nio phenomenon. Under La Nia, the decrease in rural residents’ welfare is the largest in the southwest region, followed by the middle reaches of the Yellow River and the middle reaches of the Yangtze River, and the smallest in the northeast. The reason for this result is the same as the El Nio phenomenon. (ii) The effect of weak La Nia on the welfare of rural residents is not significant. The northeast region, with the smallest impact, sees welfare change close to zero, while the southwest, with the largest impact, sees a decline of only a fifth of that seen in a weak El Nio. The reason is that the frequency and intensity of La Nia occurring trend is lower than that of El Nio under the global warming, and the weak La Nia does not show obvious characteristics of "colder winters and hotter summers", so it has little negative impact on agricultural sector. (iii) However, under the strong La Nia phase, the decrease degree of the welfare of rural residents in various regions is tens or even hundreds of times that of weak La Nia, and exceeds the decline degree of residents’ welfare under strong El Nio. One reason may be that La Nia generally lasts longer than El Nio, and therefore its impact on agriculture is more pronounced. Another reason may be that La Nia generally occurs in the year following El Nio, and the lagging effects of El Nio are reflected in the La Nia.

Fig.3 Welfare changes of rural residents in a strong El Nio phenomenon

Fig.4 Welfare changes of rural residents in a weak La Nia phenomenon

3.3 Uncertain impact of ENSO events on the welfare of rural residentsScenario 3 is to conduct a stochastic analysis on the uncertainty of ENSO events, to find the fluctuation level of rural residents’ welfare changes when the intensity of the ENSO event is unknown. As can be seen from the results of Scenario 1 and Scenario 2, no matter in El Nio or La Nia phenomena, the changing trend of equivalent and compensatory variations in welfare is roughly the same. Therefore, only the equivalent variation of rural residents’ welfare is explained below.

Fig.5 Welfare changes of rural residents in a strong La Nia phenomenon

Table 5 and Table 6 analyze the expectation, standard deviation and coefficient of variation of residents’ welfare changes under El Nio and La Nia with uncertain intensity. The coefficient of variation of rural residents’ welfare changes is larger, which means a greater fluctuation of residents’ welfare and higher welfare loss risk, and a worse stability of welfare. As shown in Table 5, the average fluctuation level of rural residents’ welfare change reaches 200% under the uncertain El Nio phenomenon, and the fluctuation level is roughly the same in all regions, among which the southern coastal area fluctuates the most. By contrast (Table 6), the impact of uncertain La Nia phenomenon on rural residents’ welfare fluctuations reaches 80%. The fluctuation of residents’ welfare in the middle reaches of the Yangtze River is most affected by the uncertainty of La Nia, and the northwest is the smallest. It can be seen that uncertain El Nio seriously affects the stability of the welfare of the residents, which will cause unpredictable welfare losses. Compared with El Nio, the coefficient of variation of La Nia is much smaller, and the risk of welfare loss is relatively small.

The above results indicate the uncertainty of ENSO events has a strong effect on the stability of the welfare of rural residents. The following set is to improve the anti-risk ability of agriculture technically to test whether it will reduce the economic fluctuation of ENSO events. As shown in Table 7 and Table 8, the fluctuation level of rural residents’ welfare caused by uncertainty El Nio and La Nia decrease by 30%-70% and 20%-40%, respectively, after technological improvements, and the extent of decline are suppressed. In general, improving the anti-risk ability of agriculture technically can effectively reduce the fluctuation level of residents’ welfare caused by the uncertainty of ENSO events, and has a positive effect on inhibiting the decline of rural residents’ welfare.

Table 5 Uncertainty of El Nio affects the fluctuation level of the equivalent variation welfare of rural residents

Table 5 Uncertainty of El Nio affects the fluctuation level of the equivalent variation welfare of rural residents

ItemRegionNENCECSCMYEMYASWNWExpectation-118 558.0-355 824.0-325 612.0-364 710.0-427 044.0-450 661.0-761 002.0-190 276.0Standard deviation250 598.0691 425.0666 277.0788 567.0844 599.0945 237.01 624 010.0377 447.0Coefficient of variation∥%-211.4-194.3-204.6-216.2-197.8-209.7-213.4-198.4

Table 6 Uncertainty of La Nia affects the fluctuation level of the equivalent variation welfare of rural residents

Table 6 Uncertainty of La Nia affects the fluctuation level of the equivalent variation welfare of rural residents

ItemRegionNENCECSCMYEMYASWNWExpectation-313 098.0-973 278.0-870 652.0-965 093.0-1 176 006.0-1 148 118.0-2 061 301.0-523 478.0Standard deviation276 654.0766 139.0732 776.0842 682.0910 761.01 017 270.01 720 784.0403 524.0Coefficient of variation∥%-88.4-78.7-84.2-87.3-77.4-88.6-83.5-77.1

Table 7 Uncertainty of El Nio affects the fluctuation level of the equivalent variation welfare of rural residents after technology updated

Table 7 Uncertainty of El Nio affects the fluctuation level of the equivalent variation welfare of rural residents after technology updated

ItemRegionNENCECSCMYEMYASWNWExpectation-21 931.0-85 935.0-72 619.0-76 769.0-104 118.0-98 834.0-167 174.0-46 795.0Standard deviation31 144.0132 125.0126 006.0150 359.0164 169.0181 301.0309 754.075 570.0Coefficient of variation∥%-142.0-153.7-173.5-195.9-157.7-183.4-185.3-161.5

Table 8 Uncertainty of La Nia affects the fluctuation level of the equivalent variation welfare of rural residents after technology updated

Table 8 Uncertainty of La Nia affects the fluctuation level of the equivalent variation welfare of rural residents after technology updated

ItemRegionNENCECSCMYEMYASWNWExpectation-217 732.0-661 212.0-600 953.0-680 214.0-805 951.0-808 475.0-1 445 232.0-359 770.0Standard deviation137 180.0363 048.0330 424.0358 870.0464 451.0403 715.0733 530.0346 463.0Coefficient of variation∥%-63.0-54.9-55.0-52.8-57.6-49.9-50.8-96.3

3.4 Impacts of government transfer payment on the welfare of rural residentsIn Scenario 4, to prevent the living standards of rural residents from being severely affected from ENSO, it is assumed that the government increases transfer payments to rural residents for disaster relief. As illustrated in Fig.6, when a strong El Nio occurs, the transfer payment of rural residents will increase, which can clearly show the relative gap in the change of the welfare in different regions. The welfare of rural residents in some regions would change from negative to positive. The government’s transfer payment has played a positive role in improving the welfare of rural residents, and the higher the payment, the greater the improvement of the welfare. The growth rate of the welfare in the southwest is the fastest, from the lowest when the transfer payment rate increases by 4%, to the second when the government transfer payment increases by 12%, which shows that the government transfer payment has an obvious inhibiting effect on the decline of the welfare of rural residents in the southwest in El Nio years.

Fig.6 Impacts of government transfer payment on the welfare of rural residents under a strong El Nio phenomenon

The government increases transfer payments to the rural resident for disaster relief during a strong La Nia, and the subsidy is the same as during an El Nio. Fig.7 shows that government transfer payments affect improving the welfare of rural residents, but the effect is weaker than that of El Nio. When government transfer payments are increased by 4%, the welfare of rural residents remained negative in all regions. When it increases to 12%, the welfare in all regions (except the southwest region) changes from negative to positive. The change of welfare in the middle reaches of the Yangtze River and the southwest region is still more sensitive to government transfer payment. In the previous analysis, a strong La Nia differs from a strong El Nio in that a strong La Nia has a greater negative impact on the welfare of rural residents, and therefore a smaller increase in the welfare of rural residents with the same level of subsidies.

Fig.7 Impacts of government transfer payment on the welfare of rural residents under a strong La Nia phenomenon

3.5 Model testThe exogenous parameters in the CGE model are derived with reference to previous studies[52-53], and sensitivity analysis is performed for all exogenous parameters using Gaussian Quadrature Method. Taking the coefficient of variation of the output of each agricultural sector as an example (Table 9), the coefficients of variation of all variables in the model are less than 10%, indicating the stability of the model built. Besides, the model passes the consistency test and the homogeneity test. Therefore, the model is suitable for simulating the economic performance of all regions of the country.

Table 9 Sensitivity test results

4 Conclusions and policy recommendations

4.1 Conclusions(i) In China, ENSO events damage rural residents’ welfare, and ENSO events of different intensities have different impacts on rural residents’ welfare in various regions. Under the ENSO event, the areas with the largest and smallest decreases in the welfare of rural residents are the southwest region and northeast region, respectively. The effect of the weak El Nio phenomenon on residents’ welfare has little regional variation. In weak La Nia conditions, welfare changes are not significant. When a strong El Nio occurs, the decline level of rural residents’ welfare is further amplified, especially in the southern coastal and southwestern regions. However, under a strong La Nia, the welfare loss of rural residents in each area is more serious than that under strong El Nio. (ii) The welfare of rural residents is seriously fluctuation due to uncertain ENSO events. The average fluctuation level of rural residents’ welfare change reaches 200% under the uncertain El Nio phenomenon, and the fluctuation level is roughly the same in all regions. Compared with El Nio, the coefficient of variation of La Nia is much smaller, but it still has strong fluctuation. Improving the ability of disaster early warning and risk defense can effectively reduce the negative impact of ENSO events and decrease the damage to the stability of welfare caused by ENSO events. (iii) Financial subsidy measures are conducive to protecting the welfare of rural residents. The government increases the transfer payment to the rural resident for disaster relief, the welfare loss would decrease, and the higher the payment, the greater the improvement of the welfare. To ensure welfare levels, under a strong La Nia, the government should increase the proportion of transfer payments relative to a strong El Nio. The welfare of rural residents in the southwest is more sensitive to the influence of government transfer payment.

4.2 Policy recommendations(i) Increasing agricultural capital investment can boost the development of modern agriculture in sensitive areas of ENSO events, especially in the southwest, middle reaches of Yangtze River and middle reaches of Yellow River. It can be inferred that the level of agricultural development in the southwest is relatively slow and agricultural development could be in urgent need of policy support from government funds. Agriculture in the middle reaches of the Yangtze and Yellow rivers are vulnerable to flooding or drought from ENSO, so agricultural capital should be more invested in crop loss risk prevention, such as crop insurance schemes. (ii) Strengthening the links between meteorological services and agricultural production sectors and improving agricultural early warning systems should be necessary to reduce the impact of ENSO uncertainties. The study of the mechanism of ENSO events can theoretically improve the accuracy of predicting and assessing the intensity, duration, and type of ENSO events. Further, based on the relevant information, agricultural producers would change the type of crops grown in advance to cope with the occurrence of adverse weather, reducing the uncertain impact of ENSO events on the welfare of the residents. (iii) Constructing short- and long-term intervention mechanisms could address the new challenges posed by climate change. In the short term, it is recommended to provide efficient intervention, focusing on financial subsidies for seeds, fertilizers,etc., to cope with sudden meteorological disasters. In the long term, investments to improve agricultural production infrastructure can be suggested, and awareness and training programs for adaptation to climate change are necessary for agricultural producers.