key: cord-1001970-39g57ye3
authors: Long, Wenjin; Zeng, Junxia; Sun, Tongquan
title: Who Lost Most Wages and Household Income during the COVID‐19 Pandemic in Poor Rural China?
date: 2021-11-23
journal: China & World Economy
DOI: 10.1111/cwe.12396
sha: a069ed24382ba9100f33145ab4633748a14e7608
doc_id: 1001970
cord_uid: 39g57ye3

China managed to eliminate all extreme poverty in rural areas in 2020. Poor households, however, may risk falling back into poverty due to the COVID‐19. This paper examines the impacts of the pandemic on wages and household incomes among different groups in poor areas of rural China. Using a unique dataset from five poverty‐stricken counties, we found that the pandemic has had large negative effects on wage income for migrant workers and workers in manufacturing, the private sector, and small enterprises. Compared with households relying on wage income, households relying on small businesses have suffered much more from the pandemic, whereas households depending on farming or transfer payments have been less affected. Although poor and ethnic minority households lost significant amounts of wage income due to the pandemic, they did not lose more household income than nonpoor and nonminority households. We conclude that support from the government has kept vulnerable households from suffering more than other households from the effects of COVID‐19. Our findings suggest that the government can play a strong role in alleviating the negative impacts of the COVID‐19.

much of the past progress made in alleviating poverty, as well as progress in improving health , food and nutrition security 2021) , and education (Engzell et al., 2021 ) over the past decade. As a result of the incomes lost during the pandemic (Egger et al., 2021; , millions of people globally could fall into extreme poverty (Decerf et al., 2021; Laborde et al., 2021) .

Although China succeeded in containing the spread of the COVID-19 pandemic within the country, thanks to a combination of bold and relatively early public health and epidemic control measures, China faces extensive socio-economic challenges in its recovery. This is due to both the strict preventive measures implemented in the fi rst quarter of 2020 (Tian, 2021) , as well as wider shocks to the economy and their impacts on households (Rozelle et al., 2020) and enterprises (Dai et al., 2021) . Despite the limited direct health impact of the COVID-19 pandemic in rural areas, challenges to the alleviation of poverty are rising. China lifted 770 million out of extreme poverty over the last four decades and eliminated extreme poverty by the end of 2020 based on current poverty standards (State Council Information Office of the People's Republic of China, 2021), but more effort will be needed to prevent the low-income population from sliding back into poverty. Vulnerable groups who were at risk before COVID-19 are now at greater risk than ever. Those who were not considered vulnerable before the COVID-19 pandemic may become vulnerable (Lancet, 2020) . The global recession resulting from the pandemic could further harm the poor by squeezing fiscal resources for poverty alleviation and reducing employment opportunities for the poor. In China, the government is especially concerned about poor households in poor rural areas. Determining how to mitigate the impact of the COVID-19 pandemic, not just on economic growth but on the wellbeing of vulnerable groups is therefore of great importance to the Chinese government. It is important to identify which groups experienced the most significant losses in wages and household income during the COVID-19 pandemic in order to help vulnerable groups to mitigate the negative impact of COVID-19.

Studies have shown that the COVID-19 pandemic has different impacts on various groups within the population (ILO, 2020; Bargain and Aminjonov, 2021) . The poor, ethnic minorities, and other vulnerable groups were believed to be more negatively affected by the COVID-19 pandemic (Qian and Fan, 2020; Kumar et al., 2021; Nie et al., 2021) . Given the importance of strong public policies to alleviate the negative effects of the COVID-19

pandemic (Gentilini et al., 2020; Song and Zhou, 2020) , this argument may be challenged when a government plays a strong role in supporting the vulnerable and poor. Moreover, many studies on the impact of COVID-19 on employment or income rely on internet surveys Qian and Fan, 2020) , phone surveys (Luo et al., 2020;  ©2021 Institute of World Economics and Politics, Chinese Academy of Social Sciences Who Lost Most Wages and Household Incomes? 97 Dai et al., 2021; , simulations (Zhang et al., 2020) , or field data with small samples (Zhang and Hu, 2021) , due to data unavailability and diffi culties in obtaining data. Thus, studies using fi rst-hand, face-to-face survey data with large samples are rare. Although several studies have examined the impact of the COVID-19 pandemic on employment in China (Che et al., 2020; Luo et al., 2020) , the economic impacts of the COVID-19 pandemic on poor and vulnerable households living in poor rural areas are not well understood due to a lack of detailed and reliable household survey data.

This paper explores the differences in the impact of the COVID-19 pandemic on wages and household incomes among different groups in poor areas of rural China. Although the poor and ethnic minority households lost a large amount of wage income due to the COVID-19 pandemic, they did not lose more household income than nonpoor and nonminority households. We argue that it is the support from the government that keeps vulnerable households from suffering more than other households from COVID-19. We suggest the government should provide more support for the vulnerable and poor groups during a crisis like the COVID-19.

This paper has two original aspects. First, it uses a unique dataset from fi ve povertystricken counties where the governments implement many policies to help the poor.

The first-hand data were collected through a household survey in June 2020, just after COVID-19 was contained in China, making fi eldwork possible. The survey was conducted in five national and provincial counties designated as poverty stricken, including one in Hubei province, the epicenter of the pandemic. As far as we know, this is the fi rst paper to use fi rst-hand household survey data from a large sample of households in a poor rural area to examine the impact of COVID-19 on employment and income. Moreover, our sample includes rural residents, rural-urban migrants, and residents in small towns. We also managed to include many ethnic minority households in the survey. Besides the household survey, we conducted face-to-face interviews with local stakeholders at the county, township, and village/community levels in all the surveyed counties.

Second, this paper examines the changes not only in wage income but also in total household income and net income. By separating the impacts of the COVID-19 pandemic on personal wage income and household income, the results provide more insights into the impact of COVID-19 on poor rural areas. We use different measures and models of wage change and household income change. The results from different estimations are robust.

The remainder of the paper is organized as follows. Section II presents the data, the analytical framework, and the estimation methods. Section III examines the difference in wage losses among different groups during the COVID-19 pandemic. Section IV explores the difference in household income losses among different groups during the pandemic. Section V provides a conclusion and discusses limitations of the research and possible policy implications.

II. Data and estimation method Second, to analyze the impact of the pandemic on rural and urban areas better, three rural villages and three urban communities in each county were selected as survey sites. Together the survey covered 15 rural villages and 15 urban communities.

Officially poor villages and nonpoor villages were both included, as were urban areas and peri-urban areas. Among the 15 villages included in the survey, seven were offi cially designated as poor. In urban areas, ten communities were located in central urban areas, while fi ve were in peri-urban settings. In each village and urban community, at least 34 households were surveyed. The expected total number of households to be surveyed was 1,020, and the actual number of households surveyed was 1,183, consisting of 5,044 individuals.

Third, due to time and funding limitations along with practical challenges, the survey did not adopt probability sampling during the selection of sample households. In practice, the sample households selected in this assessment included different household types, with basic selection principles. (i) Rural households had to include those that mainly rely on farming or remittances from migrant workers as an income source, or are self-employed. In each village, half of the surveyed households (17 households) had to be registered poor households, and another half had to be nonpoor households.

(ii) Urban households had to include those who mainly relied on wage incomes, and those individuals who were self-employed. In each urban community, surveyed households should cover at least ten migrants or families who rented their homes and at least ten urban households that were self-employed. (iii) Surveyed households ideally had to include both children and the elderly, and had to include families from ethnic minorities and those whose members included individuals with disabilities.

The analytical framework of this study is presented in Figure 1 . The COVID-19

pandemic affected different types of income, namely business income, wage income, transfer income, and farming income. This study focused on wage income. Thus, it first explored the difference in wage changes among different groups during the COVID-19 pandemic. As the pandemic may have affected different sources of income disproportionally, households with different main sources of income could also have been affected disproportionally. Thus, in our final specification, household income change was determined by household characteristics, regional characteristics, and the main source of household income. Based on the analytical framework in Figure 1 , we can examine the differences in household income changes among different groups. As wage income was at the individual level and household income was at the household level, we used different estimation models for the changes in wage and household income. The Wage_Change was measured in two ways. First, we measured the direction of Wage_Change. To estimate the impact of the COVID-19 pandemic on wage income, we asked "During January-May 2020, how did wage income change?" in the questionnaire.

There were three options for an answer: increased, unchanged, and decreased. Based on these three options, we created a dummy variable to indicate whether the wage decreased. Before asking this question, we also asked about working months and wages during 2019. Although this question is subjective, it relies on a comparison between wages in 2019 and January-May 2020. Second, we measured the magnitude of wage loss. In the questionnaire, we asked "how much did wage income decrease during January-May 2020?" for those reporting wage loss. With these two measures of Wage_ Change, we can explore the direction and the magnitude of wage change. Given the way we asked the questions, these two measures can be treated as the impact of the COVID-19 pandemic on wage income. The basic model of Wage_Change is as follows:

In Equation (1), Household_Characteristics indicates whether the person is from a poor household, an ethnic minority household, or a household in an urban area. Personal_

Characteristics include age, gender, and education level. Job_Characteristics include job location, sector, ownership, and size of the work unit. Regional_Characteristics are county dummies that control regional fi xed effects, α 0 is the constant, and μ represents the random error term. A s the direction of Wage_Change is a dummy variable, we can use logit or probit models to explore the determinants of the direction of Wage_Change. For robustness checks, we use both logit and probit models together with l inear probability models (LPM, a special case of OLS) in practice. The magnitude of wage loss, it equals zero for those without wage loss. So, we use Tobit models to investigate the determinants of the magnitude of wage loss.

Household_Income_Change was measured in two ways. The first is a dummy indicating whether the total household income was reduced in 2020. To estimate the impact of COVID-19 on household income, we asked "in 2020, how did total household income change?" There are three options for the answer: increased, unchanged, and decreased. Based on these three options, we could create a dummy variable to indicate whether the total household income decreased. The second measure is a dummy of whether the household net income was reduced in 2020. In the questionnaire, we also asked "compared with 2019, how did household net income change in 2020?" There are three options: increased, unchanged, and decreased. Based on these three options, we could create a dummy variable to indicate whether the household net income decreased. Before we asked these two questions, we asked for details about the status of whether the household was involved in wage employment, business, or farming activities. We also asked for details about the impact of the COVID-19 pandemic on wage employment, business, and farming activities. Our measures therefore reflect the direction of household income change accurately. The basic model of Household_

Income_Change is as follows:

In Equation (2), Household_Characteristics are variables indicating whether the household is a poor household, an ethnic minority household, or living in an urban area. The Main_Source_of_Household_Income is a categorical variable, including wage employment, self-employment (business), farming, and transfer payments.

Regional_ Characteristics are county dummies that control regional fi xed effects, β 0 is the constant, and e represents the random error term. As the two variables measuring the direction of Household_Income_Change are dummy variables, we use LPM, logit and probit models, to explore the determinants of the direction of total household income reduction and household net income reduction.

One concern about the estimation method is that no variable directly linked to the COVID-19 pandemic was included. The reason is twofold. First, the data only covered fi ve poor counties. Four of them did not have any cases of COVID-19. These counties also implemented a strict lockdown policy during the COVID-19 pandemic.

In this model, county dummies can capture part of the variation in policies related to the COVID-19 pandemic. Other good variables to measure the variation of policies related to the COVID-19 pandemic could not be found. Second, we wanted to explore the difference in the impacts of the COVID-19 pandemic among different groups rather than the determinants of the impacts of the COVID-19 pandemic.

Another concern related to our estimation method is the causality between the COVID-19 pandemic and Wage_Change or Household_Income_Change. We admit that it is diffi cult to quantify the income loss caused by the COVID-19 pandemic. In the questionnaire, we try to produce an accurate estimate of income loss caused by COVID-19 Household_Income_Change. In our sample, business income loss was most likely to be caused by the lockdown and the decrease in the number of customers. Both the lockdown and the reduction in customers are directly linked to COVID-19. Farming income is not affected much by the COVID-19 pandemic. We therefore believe that the household income loss defined in the paper is most likely to have been caused by COVID-19.

(iii) As we conducted the survey face-to-face in June 2020, when COVID-19 had just been contained, the information from questionnaires reflects a comprehensive picture of the impacts of COVID-19 on wage and household income. We also conducted face-to-face interviews with local stakeholders at the county, township, and village/community level in all of the surveyed counties. The information from these interviews confi rms our results from questionnaires, indicating that our results are credible. Table 1 shows that households in rural villages account for 49 percent of the total, with urban communities accounting for 51 percent. The urban and rural classifi cation here was determined by the place of residence, not by hukou status. Rural hukou status was held by 18.6 percent of households who reside in urban communities. Thus, the sample covers not only local rural and local urban residents but also rural-urban migrants, which makes this analysis more comprehensive in terms of sample coverage. 

To explore determinants of wage losses during January-May 2020, we control Personal_ Characteristics, Job_Characteristics, and Regional_Characteristics in the models.

Personal_Characteristics include age, gender, and education. Job_Characteristics include job locations, industries, ownership, and sizes of work units. Regional_Characteristics are county dummies and whether the household is located in an urban area. We use LPM, logit, and probit estimations and present marginal effects for each model in Table 4 . The magnitude, direction, and signifi cance level of the marginal effects are very close among the LPM, logit, and probit estimations.

First, there is no significant difference between workers in poor households and workers from nonpoor households after controlling for personal characteristics, jobrelated characteristics, and regional characteristics in the models. Workers from ethnic minority households, however, are still more likely to lose wages than workers from Han households. Second, compared to those working in villages, those working in other counties within the province were more likely to lose wages during the COVID-19 pandemic.

There is no significant difference between those working in the village and those working in local towns and counties or other provinces. One possible reason is that most workers were working within their home provinces. During the COVID-19 pandemic, intercounty transportation was strictly controlled. Those working in other counties within the provinces were more likely to be out of work during the COVID-19

pandemic.

Third, in terms of impacts on particular industries, Table 4 shows no significant differences in the expected changes in wages between workers in manufacturing and workers in construction. Compared to workers in manufacturing, however, workers in all other industries except construction were less likely to lose wages during the COVID-19 pandemic. One reason the manufacturing industry was affected more than most other industries is inadequate demand -a problem that especially affects export fi rms (Dai et al., 2021) .

Fourth, workers in the public sector -which includes not only local governments and the civil service, but also state-owned and collective enterprises -are signifi cantly less likely to lose wages than those in the private sector. Public sector workers are less affected by the COVID-19 pandemic than workers in the private sector. One explanation is that public sector workers are more protected by contracts, social insurance, and

working-time arrangements. The public sector is also more fi nancially stable than the private sector.

Fifth, compared with workers in small or microwork units, workers in middle or large work units are less affected by the COVID-19 pandemic. The larger the work unit, the less the wage loss. One possible explanation is that big work units were more likely to be able to pay wages for workers during the pandemic; thus, their employees are better off compared with other groups in terms of wage income. Workers in big work units are also more likely to be protected by contracts, social insurance, and workingtime arrangements.

Sixth, among the different regions, people in Zhouqu county have the lowest probability of wage reduction. This could be due to the relatively high local employment rate of approximately 51 percent, as wages of local employees are relatively less affected by the pandemic. People in Zhangwan county in Hubei province have the highest probability of wage reduction, because Zhangwan had the most cases of COVID-19

and was thus the most affected by the COVID-19 pandemic among the five counties surveyed. As Zhanwan is located in Hubei province, there were also more preventive measures within Zhangwan than in other counties.

In the questionnaire, we asked "how much did wages decrease during January-May 2020?" for those reporting wage loss. The average wage income loss among those who reported a decrease was RMB9,105, while the median was RMB6,000. As a benchmark, Chinese residents' average annual disposable income per capita in 2019 was RMB30,733, while the median was RMB26,523. It should be noted that the actual amount of Wage_Change depends on pre-COVID levels of income. Thus, certain groups (e.g. women) faced smaller wage losses, possibly because their wages were lower in the fi rst place, making their losses comparably smaller.

We used the Tobit model to explore factors affecting the magnitude of wage loss during January-May 2020. The outcome variable was the log of wage reduction. The values of those with unchanged or increased wages were set to 0. Table 5 shows the same pattern as the results in Table 4 . In absolute terms, workers from poor households experienced the same wage reduction as those from nonpoor households. Wages decreased less for workers from Han households than for workers from ethnic minority households. Notes: ***, **, and * represent significance at the 1, 5, and 10 percent levels, respectively. The outcome variable is the amount of wage loss (in RMB) during January-May 2020. Cluster standard errors at the village or community level are in the second column. Source: Authors' survey. Table 6 shows that nearly two-thirds of all households expected their total income in 2020 to decrease, whereas one-third thought that their incomes would remain unchanged or increase. Table 6 also shows that 62.7 percent of poor households reported a decreased net income, whereas the number for nonpoor households was 68 percent.

The difference between poor and nonpoor households is statistically signifi cant at the 0.1 level. As for ethnic minority households, 72.7 percent of them reported a decreased income, while for Han households the figure is 65 percent. The difference between ethnic minority and Han households is statistically signifi cant at the 0.05 level.

Although wage workers from poor households and ethnic minority households faced greater wage loss than other households during the COVID-19 pandemic, their total household income was not affected as much as their wage income. Table 7 shows that there was no signifi cant difference in total household income change in 2020 between poor and nonpoor households. This suggests that, compared with nonpoor households, poor households were not more affected by the COVID-19 pandemic in terms of total household income. The probability of a reduction in total household income for ethnic minority households, however, was lower than for Han households. This indicates that, compared with Han households, ethnic minority households were less affected by the COVID-19 pandemic in terms of total household income. From the interviews with local government officials, we learned that it was support from the government that kept poor households from suffering more than other households during the pandemic.

All fi ve survey counties were poverty-stricken counties. For these counties, the central government had a strict policy of eliminating all extreme poverty by the end of 2020.

The local government provided poor households with assistance in cash, employment, education, and medical care. These forms of assistance could help poor households to cope with the pandemic. We admit that these forms of government support are not linked directly to the pandemic. The motivation to support poor and vulnerable households most likely stems from political pressure to eliminate poverty rather than pressure to cope with the pandemic. Nevertheless, the support has helped the poor and vulnerable households in terms of income, employment, and medical care during the pandemic. Notes: ***, **, and * represent signifi cance at the 1, 5, and 10 percent levels, respectively. LPM represents the linear probability model. The outcome variable equals 1 if total household income is expected to decrease in 2020; the outcome variable equals 0 otherwise. Cluster standard errors at the village or community level are in parentheses.

The impact of the COVID-19 pandemic on household incomes is closely related to the main source of household income. The probability of households expecting decreased total income is highest among self-employed households, followed by households with incomes mainly from wage employment. The ratio is relatively low among households relying on agricultural and transfer incomes. For example, compared with households with incomes mainly from wage employment, the probability of expecting decreased total income is 0.18 higher for self-employed households; by contrast, the probability for farming-oriented households is 0.21 lower. In short, the COVID-19 pandemic has strong impacts on non-farm business and wage employment, while its impacts on farming are limited. A similar pattern is also found in Nigeria (Amare et al., 2021) . The large effect of the COVID-19 pandemic on nonfarm businesses is due to the strict lockdown policy for business activities. Moreover, since the Spring Festival is the most important period of the year for business, the lockdown measures hit businesses severely. In contrast, there was relatively little agricultural activity during the lockdown. Accordingly, in our data, the effect of the COVID-19 pandemic on farming is limited.

To compare changes in household net income, we also merge the groups with unchanged income and increased income into a single group. Table 6 shows no significant difference in loss of household net income between poor and nonpoor households or between Han and ethnic minority households.

The results from Table 8 are much like the results in Table 7 . First, the impact of the COVID-19 pandemic on household net income was nearly the same between poor and nonpoor households. Compared with Han households, ethnic minority households were less affected by the pandemic in terms of household net income. As for sources of household income, self-employed households were the most affected, while households with incomes mainly from transfer payments are the least affected. Compared with households with incomes mainly from wage employment, farming-oriented households were less affected in terms of household net income. Notes: ***, **, and * represent signifi cance at the 1, 5, and 10 percent levels, respectively. LPM represents the linear probability model. The outcome variable equals 1 if household net income is expected to decrease in 2020; the outcome variable equals 0 otherwise; cluster standard errors at the village or community level in parentheses.

Poor households are especially vulnerable during times of crisis, such as during the COVID-19 pandemic. By the end of 2020, China had lifted all of its poor rural citizens out of extreme poverty. Shocks like the COVID-19 pandemic, however, may put more households at risk of falling back into poverty. In this paper, we examined differences in the impacts of the COVID-19 pandemic on wage and household income among different groups in poor rural China.

Using a first-hand dataset collected in June 2020 from five poverty-stricken counties, we found that COVID-19 had a more signifi cant negative impact on certain groups, characterized by a higher ratio of people reporting wage and household income losses. In terms of wage income, we showed that migrant workers and workers in the private sector, small work units, and manufacturing were affected more than other workers. These groups were already disadvantaged before the COVID-19 pandemic, and are more vulnerable to wage income loss now. As for household income, we found that households relying on farming and transfer payments were less affected, whereas households relying on wage income and business were more affected.

Although the poor and ethnic minority households lost large amounts of wage income due to the COVID-19 pandemic, they did not lose more household income than other types of households. This is probably because poor and ethnic minority households are usually covered by government assistance and rely on transfer payments, making them less vulnerable to income losses during the pandemic. It was the support from the government that kept vulnerable households from suffering more than other households during the COVID-19 pandemic.

Our findings imply that the government can play a strong role in alleviating the negative impacts of the COVID-19 pandemic. We suggest that the government continue to support vulnerable and poor groups, including with cash transfers and assistance with employment, medical care, and children's education.

We should be cautious in interpreting our results. We rely on the survey data from five poverty-stricken counties in poor rural areas. The survey data is not statistically representative and therefore cannot be generalized to the overall population. Nevertheless, as a portrait of a specifi c group of households and the wide-ranging economic effects the COVID-19 pandemic has had on these households and individuals in poor rural areas, analysis of this data provides useful information on the nature and magnitude of the economic impact of the COVID-19 pandemic.

COVID-19 and food security: Panel data evidence from Nigeria

Poverty and COVID-19 in Africa and Latin America

Unequal pain: A sketch of the impact of the Covid-19 pandemic on migrants' employment in China

The impact of COVID-19 on small and medium-sized enterprises (SMEs): Evidence from two-wave phone surveys in China

Lives and livelihoods: Estimates of the global mortality and poverty effects of the Covid-19 pandemic

The short-and long-term impacts of the COVID-19 pandemic on family farms in China: Evidence from a survey of 2324 farms

Falling living standards during the COVID-19 crisis: Quantitative evidence from nine developing countries

Learning loss due to school closures during the COVID-19 pandemic

How to prevent a global food and nutrition security crisis under COVID-19

Food system resilience and COVID-19 -Lessons from the Asian experience

Social protection and jobs responses to COVID-19: A real-time review of country measures

Economic impacts of the COVID-19 lockdown in a remittance-dependent region

COVID-19, poverty and inclusive development

Impact of lockdown measures on the informal economy

Toward inclusive food systems: Pandemics, vulnerable groups, and the role of social protection

Impacts of COVID-19 on global poverty, food security, and diets: Insights from global model scenario analysis

Impacts of the COVID-19 pandemic on rural poverty and policy responses in China

Income-related health inequality among Chinese adults during the COVID-19 pandemic: Evidence based on an online survey

Who loses income during the COVID-19 outbreak? Evidence from China

Lockdowns are protecting China's rural families from COVID-19, but the economic burden is heavy

The COVID-19 pandemic and its impact on the global economy: What does it take to turn crisis into opportunity

Redefi ning vulnerability in the era of COVID-19

The State Council Information Office of the People's Republic of China

How China managed the COVID-19 pandemic

The sustainable development goals report 2021

Health, economic, and social implications of COVID-19 for China's rural population

Tracking the effects of COVID-19 in rural China over time

Rural China under the COVID-19 pandemic: Differentiated impacts, rural-urban inequality and agro-industrialization

Impact of COVID-19 on China's macroeconomy and agri-food system: An economy-wide multiplier model analysis