DU Zhi-xiong,LAI Xiao-dong,LONG Wen-jin,GAO Liang-liang

1 Rural Development Institute,Chinese Academy of Social Sciences,Beijing 100732,P.R.China

2 Graduate School,University of Chinese Academy of Social Sciences,Beijing 102488,P.R.China

3 College of Economics and Management,China Agricultural University,Beijing 100083,P.R.China

Abstract Family farms are considered the most desirable form of Chinese agriculture. Studies on the risk management of family farms are rare,while the COVID-19 pandemic provides an opportunity to explore how family farms respond to risks.Based on an online survey of 2 324 family crop farms,we examine for the first time the short-term impact (immediate impact or short-term fluctuation,and farms' instantaneous response) and long-term impact (on farms' future or longterm production) of the COVID-19 pandemic on family farms' production and operation in rural China. By using factor analysis and dummy variable regression,we find that the severity of the pandemic,the lockdown of the village,and farmers' knowledge of the pandemic contribute significantly to the short-term impact,but not on the long-term impact.Farmers' characteristics such as gender,age,and education are not related to the short-term impact,but family farms with male owners or owners with high school education or below are more likely to be diversified and large-scale. The number of years the farm has existed for and agricultural insurance affect both short-term and long-term impacts. We suggest that the government needs to pay more attention to stability-enhancing policies,the market environment,vocational training and the agricultural insurance market.

Keywords:family farms,COVID-19,agricultural risk,China

1.Introduction

Compared to other types of agricultural operators,such as smallholder farmers,cooperatives and leading agricultural enterprises,family farms are the most suitable and desirable form of agricultural production and management in China(Du 2018a). First,family farms are larger than traditional smallholder farms,which makes family farms more capable of coping with risks. Second,family farms connect smallholder farmers,cooperatives and leading agricultural enterprises. Unlike cooperatives and leading enterprises,family farmers stay at the forefront of the agricultural supply chain,mainly working on primary agricultural products.Multiple family farms can jointly establish cooperatives,and cooperatives are the main form of agricultural cooperation.Family farms can also get technical services,processing services and marketing services from leading agricultural enterprises. Therefore,family farms have great importance to China's agricultural and rural development1By the end of 2018,nearly 600 thousand family farms were registered with the Ministry of Agriculture and Rural Affairs of China (MARA).Among these registered family farms,83 thousand farms were selected as model family farms at the county level or above. The total operating land area of family farms in China was 162 million mu (10.8 million hectares) in 2018,accounting for 8.01% of the national cultivated area. The total value of agricultural products sold by family farms reached 194.62 billion CNY,with an average of 324 thousand CNY per family farm. Data from MARA:http://www.zcggs.moa.gov.cn/gzdt/202003/t20200327_6340082.htm.

A large and growing body of literature has investigated family farms' behaviors of production and management in China,including functions (Du and Liu 2017; Du 2018b),land use and transfer (Gao 2020),hiring of labor (Gaoet al.2020),green production (Cai and Du 2016; Xiaet al.2019),and the impact of policy (Liuet al.2018). Family farmers face market risks,natural risks,policy risks,social risks,management risks,and technical risks (He 2018; Liuet al.2019). So far,however,there has been little discussion about family farms' risk response and risk behaviors,especially in the context of huge and unexpected disasters.

The COVID-19 pandemic,which started in early 2020,poses huge risks for agricultural production. It provides a natural experiment for evaluating risk responses among family farms. Unlike other kinds of risks,the COVID-19 pandemic does not affect family farms directly. It is the side effects caused by anti-pandemic measures that influence family farms' behaviors. For example,the COVID-19 pandemic may create uncertainty for the global food market,which may lead to fluctuations in food supply and price.

Family farms in China play an important role in achieving China's agricultural policy goals. Among different types of family farms,family crop farms (family farms that focus on planting crops) are expected to ensure the supply of grain,oil,cotton,and other primary agricultural products. Family crop farms are essential to food security in China. Therefore,the study of the risk response and behaviors of family crop farms is of great theoretical and practical significance.

Based on an online survey of family farms in February 2020,we examine the short-and long-term impacts of the COVID-19 pandemic on family farms in China. The “shortterm impact” refers to the immediate shocks caused by the anti-pandemic measures and the immediate responses of family farms. This includes impacts on daily production and operation,spring tillage and planting,agricultural inputs,sales,and costs,as well as behavioral responses such as whether emergency measures taken. The “l(fā)ong-term impact” refers to future or long-term production plans or ideas,including planting structure adjustment,land-scale adjustment,and willingness to participate in various agricultural insurance programs. We find that mechanisms for short-and long-term effects are different. We suggest that the government needs to provide vocational education and training for family farm owners,create a policy and market environment that suits the long-term and stable operation of family farms,and improve the agricultural insurance market.

The novelty of this paper is twofold. First,we expand research on family farms by examining both the short-and long-term impacts of the COVID-19 pandemic. As far as we know,this is the first investigation on the impact of the COVID-19 pandemic on family farms in China. Second,since China has the largest number of family farms of any country and family farm is the core agricultural production and management entity,this paper contributes to the research on the COVID-19 pandemic globally.

The remaining parts of the paper proceed as follows:the second section introduces data and methods; the third section first describes the short-and long-term impacts of the COVID-19 pandemic on family farms,then presents the common factors and their scores,and the results of regressions; the fourth section discusses the determinants of the short-and long-term impacts of COVID-19 on family farms; and the last section concludes with policy implications.

2.Data and methods

2.1.Data

The data used in the paper are from a nationwide online survey conducted between February 9th and 13th,2020.The online survey was initiated and designed by the Family Farms Monitoring and Research Team (FFMRT) at the Rural Development Institute,Chinese Academy of Social Science2The team was set up in 2013 to undertake monitoring and research on family farms as entrusted by the formely Ministry of Agriculture(MOA).Professor Du Zhixiong from the Rural Development Institute,Chinese Academy of Social Sciences is the team leader. The team is working closely with MOA/MARA and has jointly published the annual development report on family farms since 2014..The survey aimed to examine the short-and long-term impacts of the COVID-19 pandemic on family farms. The questionnaires were publicized and distributedviaWeChat by members of FFMRT and governmental officials who are in charge of family farms affairs. Due to the limitations of the online survey,we were not able to select samples randomly. A total of 9 527 family farms were involved in the online survey,covering 29 provinces,municipalities and autonomous regions in mainland China3Although we would like to include samples from all regions in mainland China,there were no respondents from Tibet or Xinjiang.. According to MOARA,there were about 800 thousand family farms in the monitoring system in 2019. Therefore,more than 1%of total registered family farms were involved in our survey.The high coverage of our survey (in terms of locations and number of cases) provides an excellent dataset with rich information on family farms,and also helps to alleviate the issue of non-random sample selection.

We exclude cases with missing values and those that are outliners in terms of key variables. After this,we are left with 6 704 family farms that focus on planting (2 324),animal husbandry (315),or both (4 065). Given the importance of family crop farms and the different mechanisms of operation among different types of family farms,this paper restricts the sample to the family crop farms,which gives us 2 324 cases for the analysis.

2.2.Sample characteristics

We show the characteristics of family crop farm owners in Table 1 and the operation characteristics of family crop farm owners in Table 2.

According to Table 1,(1) the majority owners of family crop farms were male,accounting for 82.23%; (2) in terms of age,more than 70% of family farm owners are over 40 years old,while less than 30% are under 40; (3) most of the owners' are educated at the high school level or below.Only 25.39% of family farm owners have a college education or above; (4) most family farm owners are familiar with the COVID-19 pandemic; less than 1% of farmers report that they know the COVID-19 pandemic very little; (5) in terms of the duration of family crop farms,more than 80% of family crop farms last over two years,and 17.05% of family crop farms have been operated for two years or less.

According to Table 2,(1) the proportion of family crop farms with a scale of 100 mu (1 mu=1/15 ha) or less,more than 100 mu and less than or equal to 200 mu,more than 200 mu and less than or equal to 400 mu,and more than 400 mu accounted for 27.02,24.05,23.97 and 24.96%of the sample respectively; (2) nearly 39% of family crop farms were engaged in contract farming. Thus,about 1/3 of family crop farms signed an agricultural products sales contract before the COVID-19 pandemic. About 2/3 of family crop farms did not sign one; (3) more than 40% of family crop farms purchased agricultural insurance before the COVID-19 pandemic. Nearly 60% of family crop farms were not covered by agricultural insurance; (4) due to the COVID-19 pandemic,95% of the villages of family crop farms enforced a lockdown.

2.3.Methods

We adopt the principal-component factor analysis (PCF) and the dummy variable regression to carry out the quantitative analysis. Specifically,we use PCF to get common factors and their scores by reducing ten indictors of the short-term impact and three indictors of the long-term impact. The scores of common factors for the short-and long-term impacts are used as the explained variables in the dummy variable regression,in which we examine the effect of the severity of the COVID-19 pandemic,characteristics of farmers,and characteristics of farm management on theshort-and long-term impacts of the COVID-19 pandemic on family farms.

Table 1 The characteristics of family crop farm owners

Table 2 The operation characteristics of family crop farms

Data reduction by factor analysisIndicators for the short-and long-term impacts of the COVID-19 pandemic on family farms are multi-dimensional. These indicators are not independent of each other. It is difficult to calculate and compare the overall short-or long-term impacts of the COVID-19 pandemic on family farms by looking at these indicators directly. According to Cattell (1978) and StataCorp(2019),since some indicators have common characteristics that may be correlated in some respect,we can reduce the number of variables by using the PCF method. Since PCF can help us to find common factors that are more easily interpreted,PCF is widely used as a statistical technique for data reduction (Barrettet al.1974; Bai and Ng 2002;Ingram and Neumann 2006; Pukthuanthong and Roll 2009).From those common factors,we can create a single index that captures the overall short-or long-term impacts. Those indices can be used as explained variables in regressions.We use ten indicators to measure the short-term impact and three indicators to measure the long-term impact,as shown in Table 3.

ModelWe use the following model to explore the determinants of the short-and long-term impact of the COVID-19 pandemic on family crop farms:

whererefers to the impact of the COVID-19 pandemic on the farmi. We letequal the component score ofF1,F2,F3orFwhen examining the short-term impact. We letequal the component score ofM1when considering the long-term impact.

α0,β,λk,φm,andδare the parameters to be estimated.εiis the random error. When the coefficient ofβ,λk,φmorδis greater than zero,the short-term adverse impact on family farms is more serious and family farms are more likely to diversify and scale up in the long-term. When the coefficient ofβ,λk,φmorδis less than zero,it is the opposite.

Miis the incidence rate,the number of COVID-19 cases per 10 000 people,in the province where the farmiis located. Since the online survey was launched on February 9th,2020,we use the number of COVID-19 cases in each province by February 9th,2020. The resident population in each province is taken from theChina Statistic Yearbook 2019(NBSC 2020),which provides statistical data at the end of 2018. The incidence rate represents the severity of the COVID-19 pandemic in each province at the time of the online survey. From the coefficient ofMiand its significance level,we can see how the severity of the pandemic influences short-and long-term impacts on family farms.

Zkiis a set of variables that reflect the characteristics of family farm owners,including owners' knowledge of the COVID-19 pandemic,age,gender,education,and the number of years the farm has existed for.

Wmiis a set of variables related to the characteristics of family farms,including land area and its square,whether the village was in lockdown during the pandemic,whether the farm was engaged in contract farming before the pandemic,and whether the farm purchased agricultural insurance before the pandemic.

Diindicates whether farmiis located in a national key poverty-stricken county.

Potential endogeneityFor the model in eq.(1),there may be potential endogeneity problems from omitted variables.For example,the family farmers' abilities,which are not included in the model,can affect both farm operation and responses to COVID-19. Local economic conditions and government capacities are also related to the extent to which COVID-19 was controlled.

The OLS model in the paper does not have serious endogeneity problems because of the nature of the variables used in the model. The explained variables,the short-and long-term impacts of COVID-19 (F,F1,F2,F3,andM1),reflect judgments and decisions of family farms after the pandemic.Meanwhile,most of the explanatory variables are facts that occurred before the pandemic. The incidence rate,with which this paper is concerned,is an exogenous shock.

To better address the omitted variables and to obtain aconsistent estimate like a “fixed effect” model,we employ a dummy variable regression and control county dummies in regressions. According to Wooldridge (2016),we should control each farm as a dummy variable in the dummy variable regression. A common risk of putting too many individual dummies into regressions is that software like Stata will not produce the result because of matrix problems.Instead of using dummies at the farm level,here we use dummies at the county level. First,we have 2 324 family farms in 538 counties. On average,each county has 4.3 family farms. Thus,county dummies are a suitable replacement of individual farm dummies in reality. Second,223 counties only have one family farm. This 10% of cases presents no problems for the dummy variable regression.

There are another 83 counties that have only two family farms. Together there are fewer than eight farms in 469 counties,which cover 57% of the sample. The distribution of family farms in counties ensures that models with county dummies can produce approximate “fixed effect” estimation results.

Table 3 Indicators to measure the short-and long-term impacts of the COVID-19 pandemic

3.Results

3.1.Descriptions of the impact of the COVID-19 pandemic on family crop farms

When faced with the unexpected COVID-19 pandemic,family farms instinctively formulated a short-term response strategy to reduce the immediate impact of the pandemic.Besides the preliminary impact,the pandemic had a profound impact on family farms' perceptions and attitudes towards risks,as well as their long-term production and management strategies. In this section,we distinguish the short-and long-term impacts of the COVID-19 pandemic on family crop farms. “Short-term impact” refers to immediate shocks caused by anti-pandemic measures and the immediate responses of family farms. “Long-term impact”refers to how the future or long-term production plans or ideas are affected by the COVID-19 pandemic; these effects may have a more significant long-term impact on China's agriculture and deserve more attention.

The short-term impact of the COVID-19 pandemic on family crop farmsThe short-term impact of the COVID-19 pandemic on family crop farms includes impact on daily production and operation,spring tillage and planting,agricultural inputs,sales income,and emergency responses.

First,family crop farms' daily production,operation,spring tillage and planting have been greatly affected by the COVID-19 pandemic. Table 4 shows that nearly 70% of family crop farms were unable to carry out daily production and operation activities due to the pandemic. Since the pandemic occurred during the spring season,only 13% of family crop farms reported no impacts from the pandemic on spring tillage and spring planting. In contrast,nearly half of family crop farms said the pandemic had a large or huge impact on their spring tillage and spring planting.

Second,the COVID-19 pandemic had a large effect on agricultural inputs. During the pandemic,governments imposed restrictions on the movement of people and vehicles. Most villages implemented traffic control and closed roads in and out of villages. Those anti-pandemic measures affected the availability of agricultural inputs,including agricultural materials,labor and land transfer.According to Table 5,(1) more than one-third of family crop farms were unable to purchase any agricultural materials.Only about one quarter of family crop farms indicated that their purchase of agricultural materials was not affected; (2)although it was not the season for high labor demand,hiring of labor on farms was seriously affected by the COVID-19 pandemic. About 36% of farms could not hire any labor because of the pandemic,whereas only 17% of farms reported that their hiring was not affected by the pandemic;and (3) land transfer was partly affected by the COVID-19 pandemic. In our sample,19% of farms did not have any plan for land transfer. Nearly 70% of farms have plans for land transfer,and their plans are not affected by the COVID-19 pandemic. Nearly 10% of farms would reduce land transfer,while only 1.5% of farms would increase land transfer. Compared with effects on agricultural materials and labor,the demand for land is much less affected by the pandemic. The reason for the small impact of the COVID-19 pandemic on land transfer is that land transfer usually occurs during autumn and was thus largely completed before the pandemic. Demands and purchases of agricultural materials and labor,however,are made during the pandemic.

Third,80% of family crop farms have adopted emergency measures to deal with the pandemic. Table 6 shows that multiple emergency measures may be implemented at the same time. For example,46% of farms made labor arrangements in advance; 44% of farms carried out agricultural operations such as seeding in advance; 22%of the farms purchased seeds,fertilizers,pesticides and other materials in advance; 21% of farms purchased agricultural machinery services in advance; 14.94% of farms sold inventory products in advance or speeded up the sale; 18.71% of farms negotiated land leases in advance.Among different emergency measures,the majority are labor-related. The dominance of labor-related measuresindicates that labor shortage may still be a problem for family farms even though it is now common for machines to have replaced labor in many agricultural tasks.

Table 4 The short-term impacts of the COVID-19 pandemic on production,operation,spring tillage and planting

Table 5 The short-term impacts of the COVID-19 pandemic on agricultural inputs

Table 6 Emergency measures adopted by family crop farms

Fourth,for most of the family farms,sales volume and sales revenue are expected to fall,while costs are expected to rise. The falling sales and rising operating costs have an adverse effect on the income growth of family crop farms. According to Table 7,(1) nearly two-thirds of farms reported a decrease in their sales volume. One-fifth of farms expected to reduce sales volume by 40 percentage points or more. Nearly one-third of farms expected to hold their sales volume. Only 2.54% of farms expected an increase in sales volume; (2) more than 80% of farms reported a decrease in their sales revenue. Nearly one-quarter of farms expected to reduce sales revenue by 40 percentage points or more. And 17% of farms expected to hold their sales revenue. Only 2.79% of farms expected an increase on sale revenue; and (3) over 70% of farms reported that their operating costs would rise by 20-40 percentage points;while 7.5 and 21.5% of farms expected that their cooperating costs would increase by more than 40 percentage points and less than 20 percentage points,respectively.

The long-term impact of the COVID-19 pandemic on family crop farms“Long-term impact” refers to effects on the future or long-term production plans or ideas,including planting structure adjustment,scale adjustment,and willingness to purchase agricultural insurance,as shown in Table 8.

First,47% of family crop farms reported that they would adjust their future planting structure:17% of farms will increase the number of species they plant and 29% will reduce the number of species. As suggested in other studies (Valdiviaet al.1996; Dercon 1996; Zhang and Liu 2016; Zouet al.2019),planting structure adjustment is a traditional method of managing risks. Large-scale agricultural operators like family farms can use planting structure adjustment more flexibly to deal with risks.

Second,24% of farms reported that they would reduce their current scale,while 9% of farms reported that they would increase their current scale. Together,33% of farms wish to adjust their scale to manage risks. The remaining 67% of farms,however,will hold their current scale.

Third,84% of farms said that they are willing to buy agricultural insurance in the future. The high level ofwillingness to purchase agricultural insurance indicates that family crop farms have realized the importance of agricultural insurance after experiencing a pandemic that has generated real risks.

Table 7 The short-term impacts of the COVID-19 pandemic on sales volume,sales revenue and operating costs

Table 8 The long-term impacts of the COVID-19 pandemic on family crop farms

3.2.Common factors and their scores

First,we check whether our data is suitable for factor analysis. The Kaiser-Meyer-Olkin (KMO) test value for the ten indicators of short-term impact is 0.810,which indicates the proportion of variance in those ten indicators may be caused by underlying factors. The The Bartlett's Test of Sphericity Chi-square value for the ten indicators of short-term impact is 3 768 and is statistically significant at 0.001 which suggests that the ten indicators are related and therefore suitable for structure detection. As to the three indicators of long-term impact,the KMO test value is 0.506 and Bartlett's test value is 321 and is statistically significant at 0.001. Both validity tests show that the ten indicators of short-term impact and the three indicators of long-term impact are suitable for factor analysis.

Second,we select three common factors (F1,F2andF3)for the short-term impact and one common factor (M1) for the long-term impact,depending on whether the eigenvalue is larger than 1. Table 9 shows that percentages of explained variance forF1,F2andF3are 21.37,19.11 and 11.84,respectively. The accumulated percentage of explained variance forF1,F2andF3is 52.32. The percentage of explained variance forM1is 45.56.

Third,we use the rotated component matrix to determine what the common factors represent. Table 10 shows that the common factorF1is most highly correlated withX1,X2,X3andX4,which suggests thatF1mainly represents the operational efficiency of family farms. The common factorF2is most highly correlated withX5,X6,X7andX8,which indicates thatF2mainly represents resource allocations of family farms. The common factorF3is most highly correlated withX9andX10,which means thatF3mainly represents risk perceptions and responses of family farms. As to the common factorM1for the long-term impact in Table 11,M1is most highly correlated withD1andD2,which implies thatM1mainly represents adjustments to planting structure and scale in the future.

Fourth,we use the component score coefficient matrix(as shown in Tables 10 and 11) to calculate the component scores for each common factor. The scores of individual common factors and overall common factors are calculated as follows:

Table 9 Explained variance after rotating

Table 10 The rotated component matrix and component score coefficient matrix of the common factors for short-term impact (F1,F2 and F3)

For the short-term impact,the higher the values of the ten indicators,the more adverse the effects of the COVID-19 pandemic on family farms. Accordingly,the higher the values of the component scores,the more adverse the effects of the COVID-19 pandemic on family farms. For example,the higher the value of the common score ofF1,the more adverse the effects of the COVID-19 pandemic on family farms' operational efficiencies,such as sales volume,sales revenue and costs. By aggregating the common scores ofF1,F2andF3with weights of explained variance,we can get an overall common score (F). Similarly,the higher the value ofF,the more adverse the effects of the COVID-19 pandemic on family farms.

For the long-term impact,the higher the values of the three indicators,the more likely family farms are to be diversified,large-scale,and have insurance. Since the common factorM1is more related toD1andD2,the higher the value ofM1,the more diversified and large-scale family farms are.

The scores and distributions ofF,F1,F2,F3,andM1are shown in Table 12.

First,the incidence rate,lockdown of the village,knowledge of COVID-19 and contract farming are positively correlated with short-term impact,but there is a negative correlation between agricultural insurance and short-term impact. According to Table 12,(1) the higher the value of the incidence rate,the larger the scores ofFandF1; (2) the scores ofF,F1,F2andF3for family farms with “l(fā)ockdown of the village” are larger than those without “l(fā)ockdown of the village”. This means that lockdown of the village has a large negative impact on family farms in the short-term; (3)the scores ofF,F1,F2andF3for those who report a lot of knowledge of COVID-19 are larger than the scores of those with little or very little knowledge of COVID-19; (4) the score ofFfor family farms with contract farming is larger than the score for family farms without contract farming; and (5) the scores ofF,F1andF2for the family farms with agricultural insurance are smaller than the scores for family farms without agricultural insurance.

Second,incidence rate and education level are negatively correlated with long-term impact,but having a male owner,contract farming and agricultural insurance are positively correlated with long-term impact. According to Table 12,(1)the higher the value of the incidence rate,the smaller the score forM1. This means that family farms that experienced a more severe pandemic tended to be less diversifiedand large-scale; (2) compared to owners with high school education or below,theM1score for owners with college education or above is smaller; (3) theM1score for family farms with male owners is larger than it is for family farms with female owners; and (4) compared to family farms without contract farming or agricultural insurance,theM1score for family farms with contract farming or agricultural insurance is larger.

Table 11 The rotated component matrix and component score coefficient matrix of the common factors for long-term impact (M1)

Table 12 The mean scores of common factors (F1,F2 and F3) and overall common factors (F and M1)

3.3.Regression results

Scores of common factors (F1,F2andF3) and overall common factors (FandM1) are used as the explained variables in the dummy variable regression. Descriptive statistics of explaining variables are shown in Table 13.

We estimate the model of the overall short-term impact(F) in column 1 in Table 14 and the model of the overall long impact (M1) in column 2. We also estimate models of the three common factors of the short-term impact (F1,F2andF3) in columns 3,4 and 5,respectively. TheF-statistics of the five models (row 17) are all statistically significant at the level of 0.0000 (row 18),which indicates good fits for the regressions. Values ofR2for five models (row 16) are nearly 0.3 or above.

4.Discussion

4.1.The determinants of the short-term impact of COVID-19 on family farms

From columns 1,3,4 and 5 in Table 14,we have the following main results related to the short-term impact of the COVID-19 pandemic on family farms.

First,the more severe the pandemic,the greater the short-term impact on family farms. For example,the figure in row 1,column 1 is positive and statistically significant,which means the incidence rate of COVID-19 is positivelyrelated with the overall short-term impactF. Specifically,the incidence rate has adverse effects on family farms' efficiency(figure in row 1,column 3) and resource allocations (figure in row 1,column 4).

Table 13 Descriptive statistics of variables

Table 14 The determinants of the short-and long-term impacts of COVID-19 on family farms

Second,the lockdown of the village significantly increased the short-term impact of the COVID-19 pandemic on family farms,as shown by the figure in row 2,column 1.Since the lockdown of the village is mainly affecting transportation,the coefficient in row 2,column 4,is much bigger than other coefficients in column 1. The lockdown of the village makes it hard to buy agricultural materials,hire laborers and perform daily operations.

Third,the knowledge of the COVID-19 increased the short-term impact of the COVID-19 pandemic on family farms,as shown by the figure in row 3,column 1. This may be because the more farm owners understand the COVID-19 pandemic,it indicates that the short-term behavior of their farm production and operation is more affected by the pandemic and more timely emergency measures are taken.

Fourth,family farm owners' personal characteristics have no significant effect on the overall short-term impact of the COVID-19 pandemic on family farms. The coefficients of age,gender and education for the model ofFin column 1 are not statistically significant. Female owners,however,received more shocks to resource allocations than male owners,as shown by the figure in row 4,column 4.

Fifth,the scale and the duration of the farm have no significant effect on the short-term impact of the COVID-19 pandemic on family farms. All the coefficients related to scale and duration in columns 1,3,4,and 5 are not statistically significant.

Sixth,contract farming has mixed effects on the shortterm impact of the COVID-19 pandemic on family farms.Compared to farms without contract farming,farms with contract farming have a larger overall short-term impact(row 10,column 1),greater response to the pandemic (row 10,column 5),and fewer shocks to resource allocations(row 10,column 4). The advantages of resource allocation for farms with contract farming may come from two sides.One is that farms with contract farming may have more reserves of agricultural materials with which to fulfill the contract. Another is that partners of contract farmers,such as cooperatives and leading agricultural enterprises,are more likely to be able to provide agricultural materials to farms with contracts. Leading agricultural enterprises and cooperatives play a role in ensuring the supply of important production materials to family farms. Therefore,compared to farms without contract farming,farms with contract farming do not have to directly face the market risks of agricultural materials and other production factors.

Seventh,agricultural insurance can reduce the shortterm impact of the COVID-19 pandemic on family farms.Farms with agricultural insurance before the pandemic have lower overall short-term impact (row 11,column 1),and less adverse effects on efficiency (row 11,column 3) and resource allocations (row 11,column 4). As to the response to the pandemic (row 11,column 5),however,the immediate reaction of farms with agricultural insurance is stronger than farms without agricultural insurance.

Eighth,family farms in poor counties are not less affected by the COVID-19 pandemic in the short-term than farms in other regions. Before and after the pandemic,poor counties received a certain degree of special treatment in terms of industrial policies. Although the coefficient of poor county is negative,this effect is not statistically significant (row 12,column 1). Farms in poor counties have less adverse effects on efficiency (row 12,column 3). This may be because many areas have used “consumption poverty alleviation” and other methods to treat agricultural products in poor counties with special treatment.

4.2.The determinants of the long-term impact of COVID-19 on family farms

From column 2 in Table 14,we have the following main results related to the long-term impact of the COVID-19 pandemic on family farms.

First,the severity of the pandemic,the lockdown of the village,and knowledge of the pandemic are not significantly related to the long-term impact of the pandemic on family farms. This suggests that long-term behaviors like adjustment of planting structure and scale are not affected by the pandemic,after controlling other variables.

Second,farm owners' personal characteristics such as gender and education have significant effects on the long-term impact of the COVID-19 pandemic on family farms. Family farms with male owners tend to be more diversified and large-scale than those with female owners.Compared to owners with high school education or below,owners with college education or above are less likely to be diversified and large-scale. Owners with higher education may have made arrangements in terms of diversification and scale during their daily operations. They do not have to wait for huge risks like the COVID-19 pandemic to make adjustments.

Third,the scale of the farm is not significantly related to the long-term impact of the COVID-19 pandemic,but the duration of the farm is. Compared to farms with short duration,farms with a long duration may have more experience and are more mature in terms of dealing with risks. Old farms may implement measures related to risk during daily operations,so they do not have to revise their long-term strategies as dramatically as young farms.

Fourth,contract farming has a positive effect on the long-term impact of the COVID-19 pandemic on family farms. Compared to farms without contract farming,farms with contract farming face the constraints of the contract;that is,they must fulfill the contract at a given time. This constraint has become a burden in case of risks. In order to better balance the pros and cons of the contract,farms with contract farming may have to make adjustments to long-term strategies. Thus,the long-term impact of the COVID-19 pandemic is larger for farms with contract farming.

Fifth,agricultural insurance increases the long-term impact of the COVID-19 pandemic on family farms. Farms with agricultural insurance before the pandemic may be more sensitive to risks and more willing to adjust long-term strategies.

5.Conclusion and policy implications

The COVID-19 pandemic and anti-pandemic measures have changed the supply and demand side of the agricultural industry in China. However,there has been very little empirical study of the impact of the COVID-19 pandemic on family farms. Using an online survey that includes 2 324 family farms,this paper examines for the first time the shortand long-term impacts of the COVID-19 pandemic on family crop farms in China.

Descriptive analysis shows that the impacts of the COVID-19 pandemic on family farms is multi-dimensional.In terms of the short-term impact of the COVID-19 pandemic,farms' daily production,operation,spring tillage and planting,agricultural inputs are greatly affected. Most farms have adopted emergency measures to deal with the pandemic.Most farms' sales volume and sales revenue are expected to fall,whereas their costs are expected to rise. As to the long-term impact of the COVID-19 pandemic on family farms,nearly half of the farms are going to adjust planting structure and one-third of farms are going to change the scale of the farm. More than 80% of farms intend to buy agricultural insurance in the future.

Regressions show that determinants of the shortand long-term impacts of the COVID-19 pandemic on family farms are different. As a specific and real risk,the COVID-19 pandemic affected family farms mainly through anti-pandemic measures. Those measures,however,are temporary and only affect farms in short-term ways. In the long run,those measures have no significant effect on farms.What matters in the long run are human capital related factors,such as education and experience. Those factors exist regardless of the pandemic and determine the longterm path of farm operations. That is why regressions show that personal characteristics are not significantly related to the short-term impact,but affect the long-term impact of the COVID-19 pandemic on family farms.

As to contract farming,it is a double-edged sword. Under normal circumstances,contract farming brings a stable market and reduces risks of sale. When it comes to risks like the COVID-19 pandemic,however,the breach of contract is like a sharp ax that may hit the head of the family farm at any time. Therefore,contract farming is both positive and statistically significant to the short-and long-term impacts of the COVID-19 pandemic on family farms.

Agricultural insurance reduces the short-term impact of the COVID-19 pandemic on family farms,but increases the long-term impact. Family farm owners who had agricultural insurance before the pandemic are a special group of people who are more sensitive to risks. They are more active in both their short-term and long-term responses to the pandemic.

Based on the above results,we suggest policies should focus on three aspects. First,the government needs to continue to provide vocational education and training for family farm owners. By improving the human capital of family farm owners,vocational education and training will increase family farm owners' capacities to handle risks. Second,the government needs to create a policy and market environment that supports the long-term,stable operation of family farms.Stable operators with long duration are capable of dealing with risks. Third,the government needs to improve the agricultural insurance market. The agricultural insurance market should be more open to different agents.

Acknowledgements

This study received supports from the Cultural Celebrities of “Four Batches” Talents Project,China,the National Social Science Fund of China (17BJY010,17CJY032 and 18CJY032) and the National Natural Science Foundation of China (71803045). We thank Dr.Zhang Zongyi,Dr.Xiao Weidong,Dr.Zhu Sizhu,Dr.Cai Yingping,and other members of the Family Farms Monitoring and Research Team of Rural Development Institute,Chinese Academy of Social Sciences for their great work on questionnaire design and data collection and cleaning. We also would like to express our gratitude to all the officials and family farm owners for their generous and time-consuming involvement in the survey.