Riding out the COVID-19 storm: How government policies affect SMEs in China

Abstract

Based on a nationally representative survey on SMEs in China, we study the impact of government policy interventions on SMEs during the COVID-19 pandemic. Our findings are three-fold. First, relief policies in the form of payment deferrals and exemptions significantly improve SMEs' cash flows and further stimulate their operational recovery. This effect is more pronounced for firms with larger shares of high-skilled employees. Second, financial support policies do not appear to be effective in alleviating SMEs' cash constraints or encouraging the reopening of small businesses, potentially due to difficulties in accessing policy-oriented loans and misallocation of credit. Last, regional and local lock-down policies decrease SMEs' incidence of reopening and delay their expected reopening in the near future, likely by reducing consumer demand. Our findings shed new light on the policy debates on supporting SMEs during the COVID-19 pandemic.

1. Introduction

Small and medium enterprises (SMEs) form an essential part of the economy. They are also the hardest hit by the COVID-19 crisis. While numerous policies have been introduced by governments around the globe to combat the virus and to stimulate the economy, their impact, especially on SMEs, remain understudied. In this paper, we investigate how stabilization and lock-down policies affect SMEs' operating conditions, decisions and expectations.

To study those questions, we utilize the ESIEC survey of over 2000 SMEs from 62 Chinese cities.1 The first wave of the survey was conducted between February 10 and 13 of 2020, after all provincial governments in China declared first-degree state of emergency, and a number of local stabilization and lock-down policies were introduced. The second wave was conducted between May 18 and 25 of 2020, when the spread of COVID-19 had largely been contained, and economic recovery was well under way with support from nationwide stabilization policies.

We explore three types of policies in detail: payment relief, financial support (both of which are stabilization policies) and lock-downs; and focus on SMEs' decisions to reopen, expectations of reopening, self-assessment of future cash balance, and percentage of employees who have returned to work. To examine short-run policy effects, we combine the first wave of the survey with hand-collected information on local policies in early February, which were enacted semi-independently by provincial and city governments and exhibited substantial geographical variations. To examine the medium-run effects of stabilization policies, we apply a propensity-score matching (PSM) method on the May wave of the survey to study how national policies affect firms.

We begin by documenting that the effects of stabilization policies vary by their types. While financial policies do not seem to improve SMEs' operating conditions, SMEs that received payment relief policies are less likely to face short-term liquidity constraints in February: the probability is reduced by 5.8% under social security deferrals and 13% under rent reductions. Furthermore, SMEs under social security deferrals are 5.7% more likely to have reopened by the time of the February wave, and 9.5% more likely to plan on reopening within one month if they have not opened yet. On the other hand, city-level and provincial-level lock-downs in February are associated with a significantly smaller likelihood to reopen (6.8% and 13% respectively), as well as an 11% smaller likelihood of expecting to reopen within one month.

In the medium-run, similar effects are observed for national stabilization policies: SMEs that received social security deferrals and exemptions and employment stabilization subsidies are 3.7% more likely to reopen and 6.4% more likely to have over half of their employees return to work, and SMEs that received rent exemptions are 12% less likely to face short-term cash constraints. While local tax deferrals are ineffective in supporting SME recovery in the short-run, national level tax deferrals and exemptions help relieve SMEs' cash constraints and speed up their reopening in the medium-run. Again, we find no evidence that SMEs that received credit and loan supports see improvements in short-term liquidity or likelihood of reopening.

We perform heterogeneity analyses to disentangle potential channels through which local policies affect SMEs. We first explore the effects of stabilization policies on SMEs with different characteristics. Under local social security deferrals, SMEs with a larger share of highly-skilled workers, which we argue reflects a larger share of formal employment,2 higher social security expenses and more flexible work arrangements, are significantly less likely to face short-term cash constraints and more likely to reopen. We find no evidence that financial support policies especially benefit SMEs with positive account receivables, a proxy for their ex-ante liquidity constraints. This ineffectiveness likely arises from the long-standing difficulties for Chinese SMEs to obtain external financing.3

We then examine the impact of lock-downs on SMEs that face different types of demand. Under provincial highway lock-downs, which hinder the flow of goods across cities and provinces, SMEs relying on a non-local customer base are much less likely to reopen. Under city-wide social-distancing policies, which restrict face-to-face transactions, SMEs that are offline sellers are less likely to reopen. No substantial differences in expected delays in reopening are observed, which is likely because lock-downs are of a temporary nature. These results are consistent with Alexander and Karger (2022), who find that stay-at-home orders lead to a decrease in consumer spending, and with Balla-Elliot, Cullen, Glaeser, Luca, and Stanton (2022), who find that businesses' reopening decisions depend on expected demand.

Our paper is among the early studies on how policy instruments can be used to mitigate the impact of COVID-19 on SMEs, and contributes to a fast growing literature that studies the economic consequences of the pandemic. Several papers use survey evidence to study firms' primary challenges in and responses to the COVID-19 crisis (Balla-Elliot et al., 2022; Barrero, Bloom, & Davis, 2020; Buchheim, Dovern, Krolage, & Link, 2022; Cong, Yang, & Zhang, 2021; Dai et al., 2021; Dai, Mookherjee, Quan, & Zhang, 2021), and some use stock prices to directly evaluate the impact of the crisis on the performance of listed firms with different characteristics (Ding, Levine, Lin, & Xie, 2021; Hassan, Hollander, van Lent, Schwedeler, & Tahoun, 2021). A number of studies examine the effects of policy announcements on the expectations of individuals and SMEs (Baker, Bloom, & Davis, 2016; Baker, Bloom, Davis, & Terry, 2020; Coibion, Gorodnichenko, & Weber, 2020), and others investigate the effects of stay-at-home orders and economic stabilization policies on labor and consumption (Chetty, Friedman, Hendren, & Stepner, 2020; Granja, Makridis, Yannelis, & Zwick, 2020; Mongey, Pilossoph, & Weinberg, 2021). Kawaguchi, Kodama, and Tanaka (2021) find that lump-sum transfers improves firms' expected survival, but not their performance. Our study connects and enriches these two strands of literature by analyzing the effectiveness of government policies from the perspective of SMEs. We find that payment relief can ease SME's liquidity constraints, which proves to be one of firms' predominant concerns (Bartik et al., 2020; Li, Strahan, & Zhang, 2020), and assist SMEs to reopen; that financial policy may not always achieve the desired effect of directing resources towards to the financially vulnerable; and that lock-downs, which are found to depress consumer demand (Alexander & Karger, 2022), further translate to a dampening effect on SMEs' decisions and expectations to reopen. These insights are particularly valuable for policymakers. Our paper also contributes to studies that examine the impact of COVID-19 on the Chinese economy (Chen, Qian, & Wen, 2021; Fang, Wang, & Yang, 2020; He, Pan, & Tanaka, 2020).

This paper is organized as follows. Section 2 outlines the main policies adopted by Chinese cities in early February, which are examined in detail in our empirical analysis. Section 3 describes the ESIEC data and the empirical strategy. Sections 4 analyzes the short-run and medium-run impacts of policies. Section 5 discusses the findings. Section 6 concludes.

2. Overview of policies

Beginning with the lock-down of Wuhan on January 23, 2020, national and regional governments in China prescribed various policies to mitigate the impact of COVID-19. We hand-collect information on local policies for the 62 cities in which the surveyed SMEs are located, from official announcements and news articles. We first focus on the local policies that were announced on or before February 10, 2020, which is the starting date of the February survey. We categorize the policies into two groups: stabilization policies, which aim to provide economic support to firms, and lock-down policies, which aim to contain the spread of COVID-19. There exists substantial variations in the timing, type, scale and intensity of policies across cities in our sample. In Appendix A, we discuss our policy coding procedure, and present a summary of policies that were enacted by each city.

The stabilization and lock-down policies were first proposed and enacted at the local level. Starting from February 20, 2020, the Chinese central government announced a series of nationwide policies to be implemented by regional governments, and many of them were a continuation of and an upgrade over local policies. This led to a convergence of policies on the national level, covering all policies examined in this paper.4 The fortunate timing of the February survey enables us to exploit the narrow window in which variations in local policies existed. The May survey further allows us to study the effect of nationwide stabilization policies in the medium-run.

It is worth noting that most of the policy interventions that were implemented in China can also be found in other countries. Social distancing policies such as stay-at-home orders were widely adopted around the world; deferrals and exemptions of tax and social security payments were adopted in most developed countries such as the United States, Japan, Korea, and major European countries; financial supports such as credit extensions were also provided in many of those countries, with some directly targeting SMEs, such as the Small Business Grant Fund (SBGF) in the United Kingdom and the KfW Fast Loans in Germany.5 Therefore, our study also sheds light on the effectiveness of those common policy practices.

2.1. Stabilization policies

Beginning in early February, provincial and city governments enacted a plethora of policies to economically support firms, including but not restricted to, direct subsidies, improved access to financing, deferral and exemption of payment of expenses, and employment protection. We focus on payment relief and financial support, as they were more commonly adopted across cities. We examine four policies in detail: rent reduction and social security deferral, which provide payment relief; credit guarantee and loan support, which provide financial assistance. We also discuss taxation deferral policies briefly in 4, 5.

Rent Reduction. This policy normally granted one to two months of rent exemption to commercial tenants renting state-owned properties. In contrast, owners of privately-owned properties were encouraged, rather than required, to negotiate terms of rent relief with their tenants. We define a city to be enacting rent reduction, if exemptions are granted to SMEs unconditionally.

Social Security/Tax Deferral. Chinese firms are required to pay social security contributions for their employees, and face a comprehensive rate of around 55% of employees' base salary.6 During the COVID-19 crisis, a majority of cities granted social security payment deferrals to firms for a period of up to three months. While some cities granted deferrals automatically, others required firms to apply for and obtain government approval prior to granting deferrals. We define a city to be implementing social security payment deferral if deferrals are granted to SMEs automatically. Tax deferral policies were also implemented in some cities in early February, enabling SMEs to postpone their corporate income tax payments for a period of up to three months.7 A major difference between social security deferrals and tax deferrals is that social security deferrals often applied automatically to SMEs, whereas tax deferrals required SMEs to submit an application that needed approval from tax authorities.

It is worth noting that the nationwide policies concerning social security and tax payments in late February represent a substantial upgrade over the local ones. In addition to deferrals, these policies also provide payment exemptions and employment stabilization subsidies. Our analysis in Subsection 4.2 reveals that these differences may strengthen the effectiveness of social security or taxation policies over the medium run.

Credit Guarantee Support. Credit guarantee schemes are designed to help SMEs gain access to bank loans, and are adopted by more than half of the countries worldwide.8 They provide third-party guarantee on loans borrowed by SMEs, and are responsible for repaying these loans, in part or in full, to the issuing banks in case of default.

During the COVID-19 pandemic, some cities took measures to reduce the threshold that firms must meet in order to obtain guaranteed credit. These measures include, but are not restricted to, demanding state-owned credit guarantee agencies to drop counter guarantee requirements for borrowers,9 to offer discounted fees for their services, or to ask for smaller security deposits. We define a city to be enacting credit guarantee support if it instructs credit guarantee agencies to drop counter-guarantee requirements, or to adopt two or more other measures to improve firms' access to guaranteed credit.10 11

Loan Support. Loan support policies involve one or more of the following: direct provision of credit, interest subsidies, risk compensation, and loan repayment deferrals. The first refers to an increase in the amount of private business loans, or the issuance of emergency relief loans for SMEs. The second refers to subsidies on interest payments on new business loans borrowed in 2020.12 The third refers to an increase in the rate of compensation to banks by city governments for losses from loan defaults.13 The last refers to instructing banks and financial institutions to defer loan repayment or to provide rollover loans for firms with operational difficulties.

2.2. Lock-down policies

Since January 23, 2020, lock-down policies were quickly implemented across the entire country, both at the provincial and the city level. At the provincial level, inter-province and inter-city highways were partially shut down in order to limit traffic. At the city level, social-distancing policies were enacted to reduce human contact.

Provincial Highway Closures. By early February, a number of provinces had taken measures to reduce traffic on inter-province and inter-city highways. These measures include the closing down of highway entrances, exits and toll stations, and mandatory inspection of freight and cargo, thereby substantially restricting the flow of goods and commodities across regions. We define a province to be enacting highway closure if they either shut down toll stations or highway entries and exits.

City-Wide Social Distancing. Among the variety of social-distancing policies implemented by city governments, two measures stood out to be the most widely adopted: 1) the close-down of residential communities, and 2) the shutdown of local public transport. Community close-downs were enforced to varying degrees. Almost all cities required compulsory temperature checks and personal ID inspection upon entry and exit into residential complexes, and prohibited entry by visitors and non-residents. Some cities were much more stringent: they only permitted one member from each household to go out and purchase groceries and other essential items every few days; and prohibited residents from exiting residential complexes unless they needed to go to the hospital, were involved in pandemic prevention and control, or worked in industries closely related to civilian livelihood. We define a city to be enacting community close-down if it adopted both of the more stringent measures. In addition, a number of cities either suspended or reduced the frequency of their public transport services. We define a city to be enacting transport shutdown if it suspended local bus services. For the purpose of our analysis, we define a city to be enforcing social-distancing if it enacts both community close-downs and transport shutdowns.14

3. Data and empirical strategy

3.1. Survey of SMEs

The main dataset we use is a survey of small and medium enterprises in China, named the “Enterprise Survey for Innovation and Entrepreneurship in China” (ESIEC) and conducted by the Center for Enterprise Research at Peking University. Two national surveys were conducted in the field in 2018 and 2019. Two waves of a COVID-19 special survey of sample businesses in the previous two years were conducted by telephone in February and May of 2020. Below, we describe the national and the COVID-19 surveys respectively.

National Survey. The ESIEC national survey takes a random stratified sample of firms from the Firm Registration Database of the State Administration for Industry and Commerce of China, which contains the universe of all newly registered firms in China between 2010 and 2017. The sampling procedure is as follows. First, six nationally representative provinces and centrally-administered municipalities are selected. Second, counties that are provincially representative are selected based on population and total GDP. Five hundred firms are then randomly sampled from each county to form the final ESIEC sample of 6628 firms.15

The ESIEC firms are representative at both the national and provincial level, and are spread across 62 cities in 6 provinces and centrally-administered municipalities: Shanghai, Liaoning, Zhejiang, Henan, Guangdong and Gansu.16 In addition, the ESIEC sample is representative of the industry distribution of all firms from the Firm Registration Database except for the wholesale and retail sector, which is deliberately under-sampled due to the high level of homogeneity among retail firms.17

The 2018 wave covered 6199 firms. The 2019 wave, which was a follow-up on firms that missed the 2018 wave, covered the remaining 429 firms. The baseline survey contains information about firms' annual sales, year of registration, employment, physical addresses, industry, and supply chain characteristics. They also contain information about firm owners' personal characteristics and entrepreneurship history.

COVID-19 Survey. In the first half of 2020, all 6628 firms from the national survey were contacted by phone for a COVID-19 special questionnaire. The first wave of the survey was conducted between February 10 and 13, two weeks after all provinces and centrally-administered municipalities in China declared first-degree state of emergency in response to the COVID-19 outbreak. Interviewers were able to complete and retrieve a total of 2044 responses. Questions were asked about firms' operational conditions, production activities, and the main challenges they faced.

The second wave was conducted between May 18 and 25, after most provinces downgraded their state of emergency to third-degree. Again, all 6628 firms were contacted, and 1961 responses were completed and retrieved. 63% of firms that responded to the February survey also responded to the May survey; and the two waves covered a total of 2838 unique firms. Questions were asked about firms' operational recovery, the impact of COVID-19 on firms' suppliers and customers, the measures they took to adapt, and the types of government relief policies they did receive.18

It should be noted that around two thirds of the firms that appeared in the national survey did not respond to the COVID-19 survey. First, this is partially because the two surveys were conducted in different ways—telephone surveys naturally register a much lower response rate compared to field surveys. Moreover, some SME owners could feel reluctant to complete the questionnaire when facing operational challenges induced by COVID-19, which might lead to sampling bias in our analysis. In Appendix B, we perform a balance check on the characteristics of SMEs that did and did not respond to the February survey, and discuss how the sampling bias can potentially affect our results.19

Table 1 reports summary statistics for key variables. Panel A displays the basic characteristics for SMEs in the February wave, and relevant variables are from the national survey. Panel B shows SMEs' exposure to different types of local policies in February. Panel C displays SMEs' self-reported policy coverage in May. Panel D displays the main outcome variables used in the analysis.

Table 1.

Summary Statistics for Surveyed Firms.

Variables	N	Mean	Std. Dev.
Panel A: Firm Characteristics (February Wave)
Firm Age	2044	5.05	2.25
Number of Employees	1857	17.27	85.05
Total Revenue (10,000 RMB)	1245	729.74	4996.52
Whether Firm Received External Financing in 2018	1355	0.20	0.40
Whether Firm Has Account Receivables	1599	0.39	0.49
High-Skilled Worker (Percent)	1774	0.29	0.38
Whether Firm Rents State-Owned Property	2035	0.15	0.36
Whether Firm Made Online Sales	570	0.68	0.47
Whether Largest Customer is Local	749	0.61	0.49
Trade Volume with Largest Customer (Percent)	1557	15.52	24.84
Panel B: Local Policy Coverage (By February 10)
Social Security Payment Deferral	2044	0.51	0.48
Tax Exemptions or Extensions	2044	0.56	0.50
Rent Reduction for State-Owned Property	2044	0.10	0.38
Credit Guarantee Support	2044	0.24	0.43
Loan Support	2044	0.43	0.49
Highway Closure	2044	0.61	0.49
Social Distancing	2044	0.13	0.34
Panel C: Self-Reported Policy Coverage (May Wave)
Social Security Exemption or Employment Stabilization Subsidies	1711	0.42	0.49
Tax Exemptions or Extensions	1711	0.46	0.50
Rent or Utilities Reductions	1711	0.26	0.44
Credit and Loan Support	1711	0.16	0.36
Panel D: Outcome Variables
Cash Flow Is <1 Month (February)	1466	0.19	0.40
Cash Flow Is <1 Month (May)	1711	0.17	0.37
Open on Survey Date (February)	1861	0.19	0.39
Open on Survey Date (May)	1953	0.79	0.41
Expect to Re-Open within 1 Month (February)	1504	0.39	0.49
Whether Firm Has >50% Employees Return to Work	1953	0.64	0.48

Notes: This table displays summary statistics of key variables.

From Panel A, we can see that our sample firms are small in size and relatively young, which is unsurprising as they are newly registered between 2010 and 2017 by construction. Total revenue is considerably right-skewed with a median of 55, which is much smaller than the mean of 729.74. A relatively small proportion of firms had access to external financing, or are tenants at state-owned properties.20 Moreover, Panel B and Panel C show that a moderate share of firms are exposed to local economic and lock-down policies. The share for the rent reduction policy is quite small, since only 15% of firms in our sample are inferred to be state-property renters.

3.2. Empirical strategy

Our baseline specification estimates the effects of local policies on SMEs' outcomes:

$$Y_{\mathit{ij}}=\alpha +{\beta }^s{P}_{\mathit{ij}}^s+\gamma X_{\mathit{ij\tau }}+{\mathrm{ϵ}}_{\mathit{ij}}.$$ (1)

Here, Y _ij denotes outcome variables for firm i located in city j that are constructed from the COVID-19 survey. P _ij ^s denotes local policy interventions. X _ijτ is a vector of firm-level control variables taken from the ESIEC national survey of wave τ and capture firm i's basic characteristics: total employment, annual sales, firm age, and whether firm i belongs to the service sector.21 Because the ex ante firm characteristics data were collected across two waves, 2018 and 2019, we interact each control variable with a 18/19 ESIEC survey year indicator to account for the potential impact of data inconsistency. To maintain a decent sample size, we impute missing values using indicator variables.

Short-Run Policy Effects. We utilize the February wave of the COVID-19 survey to study the short-run effects of stabilization and lock-down policies. For lock-down policies, we mainly focus on two outcome variables: whether firm i has reopened by the February survey date; and whether firm i expects to resume operation within one month from the survey date, if it has not reopened yet. For stabilization policies, in addition to the two variables above, we also examine whether firm i has enough cash to sustain its operations for one month, which is a proxy for whether firm i is facing stringent short-term liquidity constraints.

P_ij^s, which denotes local policy interventions, is equal to 1 if a policy s is introduced in city j (where firm i is located) by February 10. Our coefficient of interest is β ^s, the estimated effect of policy s on firm i's outcome. It should be noted that, for stabilization policies, our baseline analysis would provide estimates of the intention-to-treat (ITT) effects, as we only observe the availability of policy support in each city in February rather than actual policy assignments. In the analysis of the impact of lock-down policies, we further include city-level infection rates in our regressions as they may directly affect local lock-down decisions as well as firms' operational decisions.

Medium-Run Policy Effects. We utilize the May wave of the COVID-19 survey to study the medium-run effects of stabilization policies. We mainly focus on three outcome variables: whether firm i has sufficient cash balance for one month of operations; second, whether firm i has reopened by the May survey date; and third, whether >50% of employees have returned to work, conditional on firm i having reopened.

Recall that major stabilization policies introduced by local governments were starting to be replaced by unified national policies since late February, and as a result, geographical policy variations no longer exist by May. Since the May wave asks firm owners whether they received specific policy support, we estimate medium-run policy effects by directly looking at the impact of actual policy coverage on firms' outcomes in May. That is, P _ij ^s is equal to 1 if firm i in city j self-identifies as a recipient of policy s.

Since the assignment of policy support may be highly correlated with firms' characteristics, we adopt a propensity-score-matching (PSM) method to address potential selection bias. We use one-to-one nearest neighbor matching based on firms' basic characteristics (employment, sales, age, service/non-service sector, and the 18/19 ESIEC survey year) as well as their geographical distance to Wuhan and industry proximity to Hubei Province. In contrast to the ITT effects of short-run policies, the PSM method here provides estimates of the average treatment on the treated (ATT) effects of stabilization policies in the medium-run.

4. Policies and SMEs' responses

In this section, we first investigate how stabilization and lock-down policies introduced by city governments in February 2020 relate to SMEs' operating conditions and their owners' reopening expectations, as reported in the February wave of the COVID-19 survey. We then examine the medium-run impact of stabilization policies using the May wave of the survey.

4.1. Short-run effects

4.1.1. Stabilization policies

Recall that stabilization policies can be further divided into two types: payment relief policies that directly improve SMEs' short-term cash flows, including deferrals of social security payments and rent reductions; and financial policies that provide support to SMEs through the banking system, such as the lowering of credit guarantee thresholds and subsidies on interest payments. We find that those two types of policies produce very different effects: while the former alleviates firms' short-term cash constraints and accelerates firms' operational recovery, the latter has little or no impact on those outcomes.

Payment Relief. Our regression analysis suggests that, deferrals of social security payments and exemptions of rent payments both reduce SMEs' short-term cash constraints. Moreover, firms that benefit from social security deferrals are more likely to reopen in early February or to plan on reopening within one month. Fig. 1(a) shows that the social security deferral policy decreases the probability of cash shortage by about 5.8%,22 and that the rent reduction policy decreases this probability by 13% for tenants at state-owned properties. Fig. 1(b) and Fig. 1(c) further demonstrate that social security deferral raises the probability of reopening by 5.7%, and the probability of expecting to reopen in one month by 9.5%. Rent reductions, however, are not associated with any statistically significant differences in the reopening decisions and expectations of firms renting state-owner properties. These findings imply that, direct deferrals or exemptions of scheduled payments can improve firms' short-term cash flow, and ones that are related to labor costs may further stimulate firms' operational recovery.23

Fig. 1.

Local Policy Interventions and SMEs' Responses.

Note: The figures display the estimated effects of local policy interventions on SMEs' survey responses. They examine two sets of policy interventions: lock-down policies, including social distancing and highway closure; and stabilization policies, including social security deferral, rent reduction, credit guarantee and loan support. Figure (a) shows the estimated effects on whether the firm holds less than one month of cash; Figure (b) shows the estimated effects on whether the firm had reopened on February 10; Figure (c) shows the estimated effects on whether the firm expects to reopen within one month, if it has not yet reopened. Bars depict 95% confidence intervals. See Table A4, Table A6 for underlying regression output.

To better understand the effect of social security deferral policies on the operations of SMEs, we further divide our sample into two groups: skill-intensive and non-skill-intensive firms.24 Conceptually, firms with a larger share of high-skilled workers may benefit more from deferrals of social security payments for the following reasons. First, they face higher social security expenses per worker, and are subject to more stringent payment obligations as required by formal employment contracts, which are more likely to apply to well-educated workers (Liang, Appleton, & Song, 2016). Second, their operational decisions may be more sensitive to labor cost shifts, because high-skilled workers are more flexible in their work arrangements (Mongey et al., 2021).

These predictions are tested with a sub-sample analysis, and the results are depicted in Fig. 2(a). Skill-intensive firms are significantly less likely to face short-term cash constraints and are more likely to reopen immediately as a result of social security deferral policies, while the policies do not significantly improve the their reopening expectations. On the contrary, non-skill-intensive firms are insignificantly less likely to be cash constrained and insignificantly more likely to reopen under the social security deferrals; nonetheless, they report significantly improved reopening prospects following the policies. Despite no statistically significant differences between the impact on skill-intensive and non-skill-intensive firms, the results suggest labor-cost-related support policies may be particularly effective in relieving the cash constraints and accelerating the reopening of skill-intensive SMEs or SMEs with higher shares of formal employment.

Fig. 2.

Heterogeneous Effects of Local Policy Interventions.

Note: The figures display the heterogeneous effects of local policy interventions on SMEs' survey responses. Figure (a) shows the effects of social security deferral policies by whether the firm has an above-average percentage of high-skilled workers; Figure (b) shows the effects of credit guarantee policies by whether the firm has positive account receivables on its balance sheet; Figure (c) shows the effects of social distancing policies by whether the firm reports making online sales; Figure (d) shows the effects of highway closure policies by whether the firm's biggest customer is non-local. Bars depict 95% confidence intervals. See Table A4, Table A6 for underlying regression output.

In early February, tax deferral policies were also introduced in a number of cities in our sample. Findings in Table A5 show that, in contrast to the social security deferral policies, the effects of local tax deferral coverage on SME's cash constraints, reopening decisions, and reopening expectation are all statistically insignificant and quantitatively smaller. The ineffectiveness of tax deferrals in the short-run is likely due to the following reasons. First, unlike social security deferrals, which were automatically and immediately available to SMEs, tax deferrals were application-based, and SMEs needed to wait before they could enjoy the benefits. Consequently, the impact of tax deferrals, if any, may take some time to manifest itself.25 Secondly, because the amount of taxes payable to the government is determined by corporate income, SMEs may not be able to fully benefit from the policy simply because their earnings were already reduced during the pandemic in February.

Financial Support. As shown in the regression analysis, the various forms of financial support do not seem to have achieved their policy goals. Fig. 1(a) shows that, firms located in cities that adopted reductions in credit guarantee requirements or loan support programs do not exhibit improved cash flow conditions. Similarly, those policies also exhibit little correlations with firms' reopening decisions and plans, as presented in Fig. 1(b) and 1(c). All coefficient estimates are statistically insignificant and close to zero in magnitude.

Financial support policies could provide much-needed relief to SMEs with severe cash constraints, while having little impact on others.26 To examine this possibility, we further divide our sample into two groups, based on whether firms are likely facing stringent cash constraints. Conceptually, firms with positive account receivables on their balance sheets are more exposed to cash flow constraints and hence have higher default risks during economic downturns. The effects of financial support policies, if any, should be more pronounced for those firms.

We investigate whether firms with positive account receivables prior to the pandemic respond differently to the credit guarantee policies.27 As shown in Fig. 2(b), the effects of reducing credit guarantee requirements on firms' cash flows and reopening decisions remain statistically insignificant and small in magnitude regardless of whether the firm reports positive account receivables. Results are similar for loan support policies,28 suggesting that financial policies in general fail to target SMEs with more urgent liquidity demands and thus prove insufficient to support the recovery of SME activities.

4.1.2. Lock-down policies

We demonstrate that both sets of lock-down policies, namely provincial highway closures and city-wide social-distancing, negatively affect firms' reopening decisions and their expectations of reopening within one month, controlling for firms' basic characteristics, the severity of COVID-19 at the city-level, and the geographic and industry proximity to Hubei. The estimated effects of each lock-down policy are displayed in Fig. 1(b) and 1(c).29 We find that firms located in cities with strict social distancing rules are on average 6.8% less likely to reopen, and 11% less likely to plan on reopening in the next month if they have not yet reopened. Similarly, the reopening rates of firms facing provincial highway closures are on average 13% lower than their counterparts, and the probability of reopening within one month is also about 11% lower. All estimated coefficients are statistically significant at at least the 10% level.

These findings indicate that lock-down policies not only impede SMEs' concurrent operational recovery, but also undermine their recovery expectations for the near future, which may lead to prolonged economic loss. Our results echo with Dai, Mookherjee, et al. (2021), which documents the negative impacts of COVID-19 restrictions on Chinese SMEs30 ; and directly connect with Alexander and Karger (2022), which demonstrate that stay-at-home orders lead to declines in consumer spending and shop visits. We take the analysis one step down the chain, and show that the effects of local and regional lock-downs can further propagate to the production side and disrupt SMEs' operations.

One potential channel through which the lock-down policies affect firms' reopening decisions by restricting firms' market access: the close-down of residential communities and shut-down of local public transport reduce the incidence of face-to-face transactions, and highway closures increase the transportation costs of delivering to distant customers. If the impact of lock-down policies operates through the negative demand shocks, then the effects of city-level social distancing policies should be smaller for online sellers because those sellers rely less on face-to-face transactions, and the effect of provincial highway closures should be greater for firms with larger non-local sales as inter-city transportation becomes more costly.

As before, we examine the above hypotheses by estimating Eq. 1 separately for each subsample of firms. Fig. 2(c) shows that both online and offline sellers31 are significantly negatively affected by social distancing policies in their reopening decisions and expectations. Nonetheless, the coefficient magnitudes are qualitatively smaller for the online sellers, which is consistent with findings in Cong et al. (2021). We then divide our sample to firms whose biggest customer is non-local, and firms that serve a more local or diversified customer base.32 Fig. 2(d) shows that, even though both groups of firms exhibit significantly lower reopening rates and weaker willingness to reopen soon when facing highway closures, the negative effect on reopening decisions is statistically significantly larger for firms whose biggest customer is non-local. The effect on reopening expectations is also qualitatively larger for those firms, although the difference of the effects is statistically insignificant. While unable to rule out other explanations for the lock-down policy effect, the results suggest that negative demand shocks induced by lock-downs may play a crucial role in SMEs' reopening decisions and expectations, echoing with Balla-Elliot et al. (2022).

4.1.3. Robustness checks

We demonstrate that our main results are robust to a larger set of control variables and that they are not driven by possible unobserved confounding factors.33 In the baseline specification, our controls comprise firm-level characteristics, namely employment, annual sales, firm age, and whether the firm belongs to the service sector.

We first introduce two additional control variables to account for possible spillover effects from Wuhan and the Hubei Province, which are the original epicenter of the pandemic. We compute the travel time between Wuhan and city i, as firms may be affected by emergency measures undertaken in Hubei (Fang et al., 2020); and a proxy that measures firm i's dependence on upstream industries in Hubei, as firms that rely on intermediate inputs produced in Hubei may face supply-chain interruptions. Table A9, Table A10 in the Appendix suggest that the effects stabilization and social-distancing policies remain largely unchanged.

For stabilization policies, we control for city-level economic variables—GDP per capita and the ratio of fiscal expenditure to fiscal revenue—since the implementation of stabilization policies depend on local economic conditions as well as the governments' fiscal budget. To address the effects of alternative policies, we also include a measure of policy intensity, which is the total number of alternative stabilization policies that are enacted in the same city. As shown in Table A9, our baseline results are robust to the addition of those variables. For social-distancing policies, we further control for the opening rate of provincial highways to account for the effects of inter-city traffic conditions. While the baseline result for the probability of reopening disappears, the result for the probability of reopening within the next month is unaffected (see Table A10).

4.2. Medium-run effects of stabilization policies

As discussed previously, the introduction of nationwide stabilization policies began to take place in late February (the medium run). Here, we examine how those policies affect firms' operational conditions in May. The PSM method provides estimates of the average treatment effect on the treated (ATT) for each stabilization category.34

As shown in Table 2 , payment relief policies continue to improve SMEs' operating conditions, whereas financial policies stay ineffective. In particular, the effect of social security deferrals, exemptions or employment stabilization subsidies on SMEs' short-term cash constraints remains negative but is no longer statistically significant.35 Meanwhile, they still improve the reopening probability of treated firms' by about 3.7%, and the probability of having a majority of employees return to work by about 6.4%. It is also worth noting that the tax exemptions and deferrals in May also significantly relieve the treated SMEs' cash constraints by 5.7% and improve their reopening probability by 4%, even though local tax deferral policies were ineffective in February. This finding implies that benefits from applications-based policy programs take time to realize.

Table 2.

Matching Results for Medium-Run Policy Effects.

Panel A: Social Security or Employment Stabilization Subsidies
	Cash <1 Month	Reopen	Labor Recovery >50%
Treatment group	0.158	0.936	0.855
Control group	0.174	0.878	0.793
ATT	−0.028	0.037*	0.064**
	(0.027)	(0.020)	(0.027)
Number of matched pairs	716	716	670
Panel B: Tax Exemptions or Extensions
	Cash <1 Month	Reopen	Labor Recovery >50%
Treatment group	0.135	0.932	0.830
Control group	0.195	0.877	0.811
ATT	−0.057**	0.040**	0.008
	(0.024)	(0.020)	(0.025)
Number of matched pairs	795	795	741
Panel C: Rent or Utilities Reductions
	Cash <1 Month	Reopen	Labor Recovery >50%
Treatment group	0.110	0.945	0.835
Control group	0.206	0.931	0.816
ATT	−0.119**	0.018	0.068
	(0.060)	(0.033)	(0.065)
Number of matched pairs	109	109	103
Panel D: Credit or Loan Supports
	Cash <1 Month	Reopen	Labor Recovery >50%
Treatment group	0.127	0.922	0.830
Control group	0.175	0.899	0.818
ATT	−0.018	0.019	−0.027
	(0.033)	(0.026)	(0.038)
Number of matched pairs	268	268	247

Note: This table reports the estimated average treatment-on-the-treated (ATT) effects of national stabilization policies on SMEs' outcomes, based on the propensity score matching (PSM) method. The matching covariates include SMEs' basic characteristics (sales, employment, age, service sector indicator), geographical distance to Wuhan, and industry dependence on Hubei Province. Panel A shows the effects of social security or employment stabilization subsidies; Panel B shows the effects of rent or utility reductions; Panel C shows the effects of credit or loan supports. Robust standard errors are reported in parentheses; Panel D shows the effects of tax reductions or deferrals. *** p < 0.01, ** p < 0.05, * p < 0.1.

Rent and utility reductions significantly reduce treated firms' probability of facing short-term cash constraints by 11.9%, but do not significantly improve their reopening and labor recovery rates. In contrast, the effects of credit and loan support on firms' outcomes are again statistically insignificant and small in magnitude. Our findings suggest that the medium-run effects of stabilization policies are generally in line with their short-run effects.

5. Discussion

Our findings highlight a sharp contrast between the effects of payment relief and financial policies. Payment relief policies, namely social security and tax exemptions or deferrals and rent reductions, can alleviate firms' cash flow shortage or encourage the recovery of SMEs' business activities. On the other hand, financial support has little impact on firms' cash balance and operational decisions in both the short-run and the medium-run.

Differences in the effectiveness of the policies can be attributed to two factors: accessibility and misallocation. First, relief policies are in the form of payment deferrals or exemptions, which automatically apply to all qualified SMEs and become effective almost immediately.36 In contrast, the complexity in the application process for bank loans was highlighted by a number of survey respondents as a practical obstacle to obtaining outside credit. Differences in policy accessibility are also reflected in SMEs' self-reported policy coverage: about 42% firms surveyed in May acknowledged receiving social security exemptions or employment stabilization subsidies, and 46% reported receiving tax exemptions or deferrals, but the percentage is only 16% for credit and loan support.37 Second, several respondents reported that SME loans were mainly granted to firms with connections to the banks or the local government. Hence, the marginal benefits of loans are low for those recipients because they were likely to have enjoyed other policy support both before and during the pandemic. The misallocation channel, which is known to generate inefficiencies (Hsieh & Klenow, 2009; Midrigan & Xu, 2014), can explain the insignificant treatment-on-the-treated effects of financial support policies.

As anecdotal evidence of the misallocation channel, we examine the correlation between firm characteristics and the medium-run coverage of stabilization policies. As shown in Table A11, larger SMEs (in annual sales or employment) are more likely to receive social security and tax exemptions, rent reductions, and financial assistance. In addition, self-employed SMEs are less likely to receive financial support and payment deferrals compared to incorporated SMEs. Firms that have previously borrowed from banks are also more likely to receive credit or loans during the pandemic. Lastly, political connections, measured by if the owner of a SME is a member of the Communist Party of China,38 do not affect policy coverage after controlling for other factors. The results suggest stabilization policies tend to favor larger and more productive SMEs. Financial support, in particular, also prioritizes SMEs with ex-ante banking relationships. An uneven distribution of policy benefits as such could result in misallocation of resources.

We provide suggestive evidence that accessibility or misallocation of credit support may have reduced the effectiveness of credit policies. A question in the May wave of the survey enables us to identify SME owners' demand for credit policies.39 Simple summary statistics suggest a considerable degree of misallocation: only 20% of the SMEs that are in need of credit policy assistance40 actually received it, while 11% of the SMEs that do not view credit policies to be important also received it. We then perform the same PSM analysis on credit policies for the subsample of SMEs that are in need of credit policy. We find that credit policies can indeed alleviate the SMEs' cash constraints in the medium-run, even though they do not produce a statistically significant effect on SMEs' reopening and labor recovery rates (see Table A12). These results suggest that while credit policies may have improved the cash flow of the SMEs with high demands for financial assistance, their efficiency is undermined by the limited financial resources assigned to those firms, and by the allocation of resources to SMEs that did not need them.

Another puzzle that emerges from our findings is that, while both social security deferrals and rent reductions improve SMEs' cash flows, only the former stimulates the recovery of SME activities. We propose two explanations for this result. First, rent reductions decrease SMEs' fixed costs, while social security deferrals decrease variable costs of production. Theoretically, reductions in fixed costs will not affect SMEs' shutdown decisions in the short-term while profit margins remain unchanged. Meanwhile, reductions in variable costs and the subsequent increase in profit margins will partially offset the negative demand shocks and stimulate resumption of production.41 Second, rent reduction policies only apply to renters of state-owned properties, most of which are industrial parks and large-scaled complexes or buildings. Management at these properties would have a strong incentive to align their reopening arrangements with the objectives of the central government, which can override their tenants' intention to reopen. Unobserved factors as such can distort SMEs' incentives and decisions.

6. Conclusion

This paper studies the effects of stabilization and lock-down policies on the recovery of SMEs' activities in China during the COVID-19 pandemic. We combine hand-collected policy schedules with the ESIEC survey data on SMEs to assess the immediate impact of local policy interventions, and apply a propensity score matching method to examine the medium-run effects of national stabilization policies on SMEs' operations. We find that stabilization policies that provide payment relief, including social security payment, tax deferrals or exemptions, and rent reductions, significantly increase SMEs' probability of re-opening, accelerate their resumption of operations, and improve their cash flow conditions. In contrast, financial support policies that provide external financing opportunities, such as lowering credit guarantee thresholds and providing loan subsidies, do not appear to be effective in alleviating SMEs' economic distress. In addition, lock-down policies such as social distancing and highway closures suppress the recovery of SMEs' activities through limiting their access to the market.

Our findings provide preliminary but important insights on policy-making in response to COVID-19. First, direct payment deferrals and exemptions can be more effective than financial policies in supporting small businesses, in the context of the Chinese economy. Differences in policy effects may arise from the accessibility of policy benefits and inefficiency in resource allocation. Second, lock-down policies are a double-edged sword: while effective at reducing health risks, they inevitably hinder the recovery of small businesses and incur economic losses by damaging SMEs' market access. Recognition of those fundamental mechanisms can help improve policy responses to COVID-19.

Appendix A. Additional details on policies

This section presents supplementary details on policy coding and implementation across cities.

A.1. Policy coding rule

The Chinese government consists of a multi-level hierarchy, three levels of which are engaged in policy-making: the central, the provincial level and the city-level (or prefecture level). Provincial and city governments are responsible for a wide range of regional matters, and are highly autonomous in the management of local economies (Xu, 2011). Since China is a large country with substantial regional heterogeneities, the central government usually announces policy guidelines and recommendations for the regional governments, which then issue their own policy implementation procedures that account for local needs and constraints.

We hand-collect information on provincial and city-level lock-down and stabilization policies from official announcements and news reports. While some policy items dictate mandatory action, others merely recommend potential courses of action. We define the former to be policy directives, and the latter to be policy advice, and assume that only policy directives will be dutifully enacted.42 That is, in our analysis, we restrict our attention to policy directives only, and use the terms “policy” and “policy directive” interchangeably.

In our definition, a policy directive contains imperative keyword(s),43 as well as a specific numerical goal, such as the number of months for which an exemption is to be applied or the rate at which loans are to be subsidized. Below, we provide one example of a directive and an advice to illustrate their differences.

• 1.Directive: SMEs that rent real estate property from state-owned enterprises for production and business activities will be exempted from paying two months of rent for February and March.44
• 2.Advice: every state-owned credit guarantee agency should offer fee exemptions to businesses that are severely impacted by the pandemic.45

We then use both provincial and city announcements to determine which policies will be enacted at the city-level. In the simplest case, a policy has the same articulation in both announcements, and we assume it will be enacted in this form. If a policy is articulated differently in each announcement, we assume that it will be enacted according to the city-level articulation. Lastly, if a policy appears in the provincial, but not in the city announcement, we assume that it will be enacted according to the provincial articulation.46

A.2. Policy implementation

Table A1 below displays the cities that have enacted highway shutdown, social distancing, rent reduction, social security deferral and credit guarantee threshold policies on or before February 10, 2020.

Appendix B. Sampling bias

This section examines sampling bias in the February wave of the COVID-19 survey, and discusses how it can affect the baseline results.

To do so, we regress SMEs' response behavior in the February COVID-19 survey on a set of SME characteristics taken from the ESIEC national survey. We make sure to include the four variables used in our heterogeneity analysis. Results are displayed in Table A3. While most variables are uncorrelated with SMEs' survey responses, coefficients on external financing, E-commerce, high-skilled workers and account receivables turn out to be statistically significant. Below, we discuss each of the four sampling issues in greater detail.

First, SMEs that obtained external financing in 2018 are more likely to respond to the survey. Since those SMEs have greater reliance on external credit, they should benefit more from credit support programs in February 2020. In other words, our baseline result that credit policies are ineffective is already upward-biased, and this sampling issue should not affect the interpretation of our results.

Second, SMEs that engage in E-commerce—that is, they have positive online sales—are less likely to respond to the survey. According to Fig. 1, those firms are less affected by lock-down policies. Thus, our baseline result that lock-down policies negatively affect reopening behavior and expectations is again upward-biased, and this sampling issue should not affect our results qualitatively.

Third, SMEs that have an above-average share of high-skilled workers are more likely to respond to the survey. According to Fig. 1, those firms benefit more from social security deferrals in comparison to SMEs with a below-average share of high-skilled workers—they see bigger improvements in short-term liquidity and report a higher chance of reopening. This sampling issue can potentially compromise our baseline results on the effects of social security deferrals, which are favorably biased by the over-representation of firms with a larger share of high-skilled workers. Results from the heterogeneity analysis, on the other hand, still hold.

Last, SMEs with positive account receivables are more likely to respond to the survey. According to Fig. 1, credit policies do not affect firms with positive and zero account receivables differently. Hence, this sampling issue does not change the interpretation of our baseline results.

Appendix C. Additional variable definitions

This section includes supplementary details on variable construction.

City-level Economic Variables. Total GDP, population, fiscal revenue and expenditure are extracted from the 2018 city statistical yearbooks, which are officially published by China's National Bureau of Statistics and cover all cities and counties in China. We compute GDP per capita using the ratio between total GDP and population, and use the ratio of fiscal expenditure to revenue to measure the budget of city governments.

Highway Opening Rate. Data on highway opening rates is provided by G7, a fleet management service provider in China that offers IoT/AI services to over 800,000 vehicles. Opening rate is defined as the ratio between the intensity of road freight transport in February 2020 and the intensity of road freight transport in December 2019, where the intensity of road freight transport is measured using satellite images.

Infection Rate. Daily data on the number of confirmed COVID-19 cases for each city since late January are extracted from official announcements. We aggregate the number of cases up to February 10, and divide it by city population to obtain a measure of local COVID-19 infection rate.

Geographical Distance to Wuhan. We use Baidu Map to infer the shortest travel time by car from each city in our sample to the city of Wuhan.

Industry Dependence on Hubei Industries. We use the 2012 Input-Output Table (IO Table) and the 2013 Annual Survey of Industrial Enterprises (ASIE) provided by the National Bureau of Statistics to compute each industry's dependence on the upstream industries in Hubei. We first compute the percentage of outputs from Hubei Province in the total output of each industry based on ASIE data. We then compute the product of the vector of input shares of each industry (from the IO Table), and the vector of Hubei's percentage of output in each upstream industry (from the ASIE data). The measure provides us an estimate of the percentage of inputs from Hubei province in each industry.

Property Ownership. We infer property ownership using firms' addresses in the ESIEC dataset.47 We first extract names of plazas and business and industrial complexes from these addresses. If a name contains keywords such as technology parks and development zones, which are usually developed and run by the local government, then we assume it is state-owned property. A full list of such keywords is as follows: parks, technology parks, software parks, entrepreneur parks, industrial parks, science cities, incubators, enterprise bays, science and innovation centers, clusters, development zones, and experimental zones. If a name contains brand names of private real estate developers such as Wanda, Wanke and Hengda, then we know it is private property. For remaining addresses, we manually locate them using Baidu Map, which is the Chinese counterpart of Google Map. If they belong to a plaza, marketplace or business/industrial complex, we search for the ownership information of companies that developed and/or are operating these places using Tianyancha.cn (a data search platform for information on Chinese enterprises). If such information does not exist, then we assume the property is privately-owned. As an illustrative example, a company in our sample is located at “Hongmei Street, Minhang District, Shanghai”, and its street number corresponds to the Hongqiao Fund Town, a business district that belongs to the Shanghai government.

Appendix D. Additional figures and tables

Fig. A1.

Industry and Geographical Distribution of Sample Firms.

Note: Figure (a) displays the industry distribution of firms in the ESIEC dataset and of all firms in the Firm Registration Database. Figure (b) shows the geographical distribution of firms in the analysis.

Table A1.

Policy Implementation Across Cities.

City	Highway	Social Distancing	Rent	Social Security	Credit	Loan
Shanghai			✔	✔		✔
Hangzhou	✔		✔	✔
Ningbo	✔	✔	✔	✔		✔
Wenzhou	✔	✔	✔	✔
Jiaxing	✔		✔			✔
Shaoxing	✔		✔	✔	✔	✔
Jinhua	✔		✔		✔	✔
Quzhou	✔		✔
Taizhou	✔	✔	✔			✔
Guangzhou			✔	✔	✔	✔
Shaoguan				✔	✔
Shenzhen			✔	✔	✔	✔
Zhuhai			✔	✔		✔
Shantou				✔		✔
Foshan			✔	✔
Jiangmen				✔	✔
Zhanjiang				✔	✔
Maoming			✔	✔
Zhaoqing				✔	✔
Huizhou				✔	✔
Meizhou				✔	✔
Shanwei			✔	✔		✔
Heyuan				✔	✔
Yangjiang				✔	✔
Qingyuan				✔	✔
Dongguan			✔	✔	✔	✔
Zhongshan			✔	✔		✔
Chaozhou				✔	✔
Jieyang				✔
Yunfu				✔
Zhengzhou	✔
Kaifeng	✔		✔
Luoyang	✔		✔		✔	✔
Pingdingshan	✔
Anyang	✔
Xuchang	✔		✔		✔	✔
Luohe	✔		✔		✔
Nanyang	✔	✔
Shangqiu	✔		✔	✔		✔
Xinyang	✔	✔
Zhoukou	✔
Zhumadian	✔
Jiyuan	✔
Shenyang	✔
Dalian	✔		✔	✔
Anshan	✔		✔		✔	✔
Dandong	✔		✔			✔
Yingkou	✔		✔			✔
Fuxin	✔		✔			✔
Liaoyang	✔		✔			✔
Huludao	✔		✔
Lanzhou	✔
Baiyin	✔	✔
Tianshui	✔
Wuwei	✔	✔
Zhangye	✔
Pingliang	✔
Jiuquan	✔
Qingyang	✔	✔
Dingxi	✔			✔
Longnan	✔					✔
Gannan	✔

Table A2.

Examples of Policies in Other Countries and Regions.

	Social Distancing Policies	Economic Policies
U.S.	Stay-at-home order	Paycheck Protection Program (Under CARES Act) Deferral of social security payroll taxes
U.K.	Gathering limits Entertainment venues closed Stay-at-home order	The coronavirus job retention scheme (80% of wages) Deferral of VAT payments due to COVID-19 The Small Business Grant Fund (SBGF): cash grant of £10,000
France	Stay-at-home order Close-down of all non-essential locations	110 billion emergency plan Deferrals of social and/or tax payments Direct tax rebates Deferral of rental payments Rescheduling of bank credits
Germany	Non-essential public services closed Public gatherings banned Stay-at-home order (only for a short time)	Short-time working allowance (over 60% of the missing net wage, full reimbursement of social security contributions) The KfW fast loans for SMEs Grants for micro-enterprises and self-employed persons
Japan	Stay-at-home order (Suggested) close-down of entertainment venues	Business subsidy programs Financial supports (loans and guarantees) Employment adjustment subsidies Deferrals of national tax payments
Korea	Entertainment venues closed Public gatherings banned	Emergency Fund to encourage firms to retain their employees Government guarantees and insurance on loans. Tax credits for rental business owners who made rent cuts for commercial buildings Income and corporate tax reductions for SMEs in special disaster areas VAT reductions
Hong Kong	(Some) entertainment venues closed Public gatherings banned	Reduction of tax payable Deferring tax payments Employment Support Scheme Job creation and job advancement Government rental concessions, fee waivers, provision of loans and loan repayment deferrals to reduce financial burdens
Singapore	Gathering limits Entertainment venues closed Non-essential workplaces closed Stay-at-home order	Jobs Support Scheme Rent and loan deferral Enterprise Singapore's SME Working Capital Loan scheme and Temporary Bridging Loan Programme under the Unity Budget

Notes: This table provides a summary of policies implemented by some other countries and regions, including the United States, the United Kingdom, France, Germany, Japan, Korea, Hong Kong, and Singapore. Sources of information include newspapers, government reports, and professional summaries.

Table A3.

SMEs' Characteristics and their Responses to the COVID-19 Survey.

	Whether SME Responded to February Survey
Sales	0.003
	(0.004)
Employment	−0.006
	(0.005)
Age	−0.002
	(0.003)
External Financing in 2018	0.059***
	(0.022)
Number of Big Suppliers	0.001
	(0.008)
Number of Big Customers	0.010
	(0.006)
Engages in E-Commerce	−0.050*
	(0.028)
High-Skilled Workers Above Average	0.034**
	(0.016)
Biggest Customer is Local	−0.015
	(0.020)
Has Account Receivables	0.026*
	(0.015)
Observations	6653
R-Squared	0.023

Note: This table reports correlations between SMEs' characteristics and whether they responded to the first wave of the COVID-19 survey in February. Regression controls for industry and province fixed-effects. Robust standard errors are clustered at city level. *** p < 0.01, ** p < 0.05, * p < 0.1.

Table A4.

Effects of Stabilization Policies.

Panel A. Effects of social security policies
	Cash <1 Month			Reopen			Reopen <1 Month
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Social Security Deferrals	−0.058*	−0.100***	−0.038	0.057*	0.136**	0.024	0.095***	0.057	0.106**
	(0.029)	(0.035)	(0.037)	(0.032)	(0.052)	(0.031)	(0.031)	(0.056)	(0.042)
Sample	All	High Skill	Low Skill	All	High Skill	Low Skill	All	High Skill	Low Skill
Observations	1466	487	806	1861	596	1037	1504	462	853
R-Squared	0.018	0.044	0.013	0.038	0.094	0.016	0.074	0.092	0.055
Panel B. Effects of credit guarantee policies
	Cash <1 Month			Reopen			Reopen <1 Month
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Credit Guarantee	0.015	−0.021	0.020	0.018	0.038	0.003	0.036	0.093	0.024
	(0.024)	(0.043)	(0.032)	(0.040)	(0.054)	(0.046)	(0.042)	(0.066)	(0.052)
Sample	All	AR > 0	AR = 0	All	AR > 0	AR = 0	All	AR > 0	AR = 0
Observations	1466	484	668	1861	588	878	1504	487	706
R-Squared	0.014	0.019	0.017	0.033	0.047	0.018	0.067	0.055	0.065
Panel C. Effects of rent reduction policies
	Cash <1 Month			Reopen			Reopen <1 Month
	(1)			(2)			(3)
Rent Reductions	−0.135**			0.017			−0.024
	(0.052)			(0.076)			(0.064)
Sample	State-property renters			State-property renters			State-property renters
Observations	255			305			244
R-Squared	0.109			0.074			0.149
Panel D. Effects of loan supports
	Cash <1 Month			Reopen			Reopen <1 Month
	(1)			(2)			(3)
Loan Supports	−0.039			−0.022			0.017
	(0.029)			(0.037)			(0.036)
Sample	All			All			All
Observations	1466			1861			1504
R-Squared	0.016			0.034			0.066

Note: This table reports the estimated effects of stabilization policies on whether SMEs hold less than one month of cash balance, their reopening status by the survey dates, and whether they expect to reopen in one month, if not reopened yet. Columns 1 and 4 report estimates for all sample firms; columns 2, 3, 5, and 6 report estimates for subsamples of firms. All regressions control for SMEs' basic characteristics (sales, employment, age) interacted with year fixed effects and service-sector fixed effects. Robust standard errors are clustered at city level. *** p < 0.01, ** p < 0.05, * p < 0.1.

Table A5.

Short-Run Effects of Tax Deferral Policies.

	Cash <1 Month	Reopen	Reopen <1 Month
	(1)	(2)	(3)
Tax Deferrals	−0.020	0.018	0.054
	(0.0315)	(0.0320)	(0.0340)
Sample	All	All	All
Observations	1466	1861	1504
R-Squared	0.014	0.033	0.068

Notes: This table reports the estimated effects of tax deferral policies on firms' short-term cash flow, reopening decision and expectations to reopen within one month. Cities that introduced tax deferral policies in early February include: Anshan, Dandong, Shanghai, Hangzhou, Wenzhou, Jiaxing, Shaoxing, Jinhua, Taizhou, Zhengzhou, Kaifeng, Luoyang, Luohe, Shangqiu, Guangzhou, Shaoguan, Shenzhen, Shantou, Foshan, Jiangmen, Zhanjiang, Maoming, Zhaoqing, Huizhou, Meizhou, Yangjiang, Qingyuan, Dongguan, Zhongshan, Chaozhou, Jieyang, Yunfu, and Longnan. All regressions control for SMEs' basic characteristics (sales, employment, age) interacted with year fixed effects, service-sector fixed effect, and city-level infection rates of COVID-19. Robust standard errors are clustered at city level. *** p < 0.01, ** p < 0.05, * p < 0.1.

Table A6.

Effects of Lock-Down Policies.

Panel A. Effects of social distancing policies
	Reopen			Reopen <1 Month
	(1)	(2)	(3)	(4)	(5)	(6)
Social Distancing	−0.068*	−0.107**	−0.245**	−0.114***	−0.171**	−0.345**
	(0.036)	(0.051)	(0.096)	(0.037)	(0.078)	(0.166)
Sample	All	E-Comm >0	E-Comm = 0	All	E-Comm >0	E-Comm = 0
Observations	1806	350	176	1460	277	131
R-Squared	0.042	0.040	0.199	0.074	0.121	0.147
Panel B. Effects of highway closure policies
	Reopen			Reopen <1 Month
	(1)	(2)	(3)	(4)	(5)	(6)
Highway Closure	−0.128***	−0.098***	−0.295***	−0.112***	−0.104***	−0.183**
	(0.027)	(0.026)	(0.056)	(0.036)	(0.038)	(0.077)
Sample	All	Local/Div Customer	Non-local Customer	All	Local/Div Customer	Non-local Customer
Observations	1806	1534	272	1460	1250	210
R-Squared	0.058	0.043	0.190	0.075	0.069	0.123

Notes: This table reports the estimated effects of lock-down policies on SMEs' reopening status by the survey dates, and whether they expect to reopen in one month, if not reopen yet. Columns 1 and 4 report estimates for all sample firms; columns 2, 3, 5, and 6 report estimates for subsamples of firms. All regressions control for SMEs' basic characteristics (sales, employment, age) interacted with year fixed effects, service-sector fixed effect, and city-level infection rates of COVID-19. Robust standard errors are clustered at city level. *** p < 0.01, ** p < 0.05, * p < 0.1.

Table A7.

Heterogeneous Effects of Stabilization and Lock-Down Policies Across Sectors.

Panel A. Stabilization Policies, Group 1
	Social Security Deferral			Rent Reduction
	Cash <1 Month	Reopen	Reopen <1 month	Cash <1 Month	Reopen	Reopen <1 month
	(1)	(2)	(3)	(4)	(5)	(6)
Policy × Agriculture	−0.049	−0.137	0.118	–	0.024	−0.248
	(0.153)	(0.095)	(0.088)	–	(0.405)	(0.333)
Policy × Manufacturing	−0.051	0.081*	0.146**	−0.127	−0.071	0.047
	(0.048)	(0.046)	(0.063)	(0.098)	(0.081)	(0.089)
Policy × Service	−0.056*	0.083**	0.083**	−0.162**	0.087	−0.066
	(0.032)	(0.032)	(0.031)	(0.069)	(0.087)	(0.084)
Observations	1466	1861	1504	255	305	244
R-Squared	0.019	0.053	0.076	0.117	0.085	0.152
Panel B. Stabilization Policies, Group 2
	Credit Guarantee			Loan Supports
	Cash <1 Month	Reopen	Reopen <1 month	Cash <1 Month	Reopen	Reopen <1 month
	(1)	(2)	(3)	(4)	(5)	(6)
Policy × Agriculture	−0.144*	−0.088	0.112	0.103	−0.037	0.021
	(0.076)	(0.157)	(0.127)	(0.135)	(0.102)	(0.139)
Policy × Manufacturing	0.007	0.024	0.020	−0.035	−0.065	0.111
	(0.046)	(0.050)	(0.063)	(0.048)	(0.046)	(0.067)
Policy × Service	0.031	0.032	0.039	−0.046	0.006	−0.011
	(0.028)	(0.045)	(0.042)	(0.030)	(0.040)	(0.034)
Observations	1466	1861	1504	1466	1861	1504
R-Squared	0.017	0.044	0.067	0.018	0.044	0.069
Panel C. Lockdown Policies
	Social Distancing			Highway Closure
	Cash <1 Month	Reopen	Reopen <1 month	Cash <1 Month	Reopen	Reopen <1 month
		(1)	(2)		(3)	(4)
Policy × Agriculture		−0.020	−0.182*		0.119	−0.138
		(0.067)	(0.093)		(0.151)	(0.092)
Policy × Manufacturing		−0.072*	−0.126**		−0.146***	−0.211***
		(0.040)	(0.055)		(0.043)	(0.070)
Policy × Service		−0.086**	−0.099**		−0.147***	−0.070*
		(0.041)	(0.040)		(0.027)	(0.037)
Observations		1806	1460		1806	1460
R-Squared		0.051	0.075		0.072	0.079

Notes: This table reports the heterogeneous effects of policy interventions across different sectors on SMEs' reopening status by the survey dates, and whether they expect to reopen in one month. Panel A displays results for social security deferral and rent reduction, Panel B displays results for credit guarantee and loan supports, and Panel C displays results for lockdown policies. All regressions control for SMEs' basic characteristics (sales, employment, age) interacted with year fixed effects, sector fixed effects, and city-level infection rates of COVID-19. Robust standard errors are clustered at city level. *** p < 0.01, ** p < 0.05, * p < 0.1.

Table A8.

Covariate Balance Summary, for PSM Analysis of Policy Effects.

Panel A. Social Security Exemption or Employment Stabilization Subsidies
	Standardized differences		Variance ratio
	Raw	Matched	Raw	Matched
Sales	0.433	0.052	1.770	0.899
Employment	0.562	−0.028	0.996	0.918
Age	0.054	0.010	1.268	1.062
Service Sector Indicator	−0.104	0.051	0.902	0.948
Wave 2018 Indicator	−0.267	−0.013	2.545	1.035
Distance to Wuhan	−0.010	−0.026	0.538	0.658
Ind. Dependence on Hubei	−0.295	−0.048	1.093	1.597
Panel B: Tax Exemptions or Extensions
	Standardized differences		Variance ratio
	Raw	Matched	Raw	Matched
Sales	0.153	0.074	1.203	1.078
Employment	0.282	−0.008	0.920	0.992
Age	−0.016	0.060	0.949	0.943
Service Sector Indicator	−0.013	−0.022	1.013	1.022
Wave 2018 Indicator	−0.043	0.037	1.158	0.892
Distance to Wuhan	−0.029	0.037	0.798	0.961
Ind. Dependence on Hubei	−0.127	0.016	1.222	1.156
Panel C. Rent or Utilities Reductions
	Standardized differences		Variance ratio
	Raw	Matched	Raw	Matched
Sales	0.108	−0.073	1.093	0.821
Employment	0.048	−0.088	1.117	1.120
Age	0.179	−0.088	1.074	0.735
Service Sector Indicator	0.079	0.057	0.958	0.965
Wave 2018 Indicator	−0.267	−0.013	2.545	1.035
Distance to Wuhan	0.051	−0.034	0.595	0.762
Ind. Dependence on Hubei	−0.121	0.037	0.958	0.965
Panel D. Credit or Loan Supports
	Standardized differences		Variance ratio
	Raw	Matched	Raw	Matched
Sales	0.272	−0.011	1.535	0.960
Employment	0.367	−0.010	1.204	0.972
Age	0.100	0.040	0.907	0.931
Service Sector Indicator	−0.122	−0.081	1.116	1.069
Wave 2018 Indicator	−0.104	−0.013	2.545	1.035
Distance to Wuhan	0.025	0.063	0.856	0.866
Ind. Dependence on Hubei	−0.039	0.006	1.096	1.093

Note: This table reports the balance test of covariates in the propensity score matching analysis of policy effects on reopening status. The covariates include firms' basic characteristics (sales, employment, age, and service sector indicator), geographic distance to Wuhan, and industry dependence on Hubei province. The treatment group comprises of firms that self-identify as recipients of corresponding policy supports. Each panel compares the mean and variance of covariates of the treatment and control groups, in raw and balanced data.

Table A9.

Robustness Checks for the Effects of Stabilization Policies.

Panel A. Effects of social security policies
	Cash <1 Month			Reopen			Reopen <1 Month
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Social Security Deferral	−0.051*	−0.050*	−0.053*	0.057*	0.087***	0.075**	0.092***	0.078*	0.102***
	(0.028)	(0.027)	(0.030)	(0.030)	(0.032)	(0.034)	(0.030)	(0.041)	(0.036)
Observations	1466	1433	1466	1861	1806	1861	1504	1460	1504
R-Squared	0.032	0.022	0.019	0.044	0.052	0.040	0.077	0.074	0.074
Additional Controls	Wuhan + Hubei	Economic	Policy Intensity	Wuhan + Hubei	Economic	Policy Intensity	Wuhan + Hubei	Economic	Policy Intensity
Panel B. Effects of credit guarantee policies
	Cash <1 Month			Reopen			Reopen <1 Month
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Credit Guarantee	0.024	0.012	0.030	0.026	0.004	0.018	0.038	0.020	0.023
	(0.023)	(0.024)	(0.020)	(0.040)	(0.040)	(0.043)	(0.041)	(0.045)	(0.043)
Observations	1466	1433	1466	1861	1806	1861	1504	1460	1504
R-Squared	0.029	0.020	0.020	0.040	0.020	0.033	0.070	0.070	0.068
Additional Controls	Wuhan + Hubei	Economic	Policy Intensity	Wuhan + Hubei	Economic	Policy Intensity	Wuhan + Hubei	Economic	Policy Intensity
Panel C. Effects of rent reduction policies
	Cash <1 Month			Reopen			Reopen <1 Month
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Rent Reduction	−0.135**	−0.135**	−0.164***	0.018	−0.020	−0.045	−0.007	−0.047	−0.104
	(0.054)	(0.058)	(0.061)	(0.074)	(0.075)	(0.069)	(0.060)	(0.097)	(0.073)
Observations	255	251	255	305	301	305	244	240	244
R-Squared	0.118	0.109	0.112	0.074	0.114	0.088	0.173	0.157	0.163
Additional Controls	Wuhan + Hubei	Economic	Policy Intensity	Wuhan + Hubei	Economic	Policy Intensity	Wuhan + Hubei	Economic	Policy Intensity
Panel D. Effects of loan support policies
	Cash <1 Month			Reopen			Reopen <1 Month
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Loan Support	−0.034	−0.019	−0.018	−0.021	−0.004	−0.049	0.014	−0.020	−0.029
	(0.029)	(0.036)	(0.030)	(0.037)	(0.037)	(0.044)	(0.035)	(0.034)	(0.039)
Observations	1433	1466	1466	1806	1861	1861	1504	1460	1504
R-Squared	0.030	0.020	0.017	0.040	0.044	0.036	0.069	0.070	0.070
Additional Controls	Wuhan + Hubei	Economic	Policy Intensity	Wuhan + Hubei	Economic	Policy Intensity	Wuhan + Hubei	Economic	Policy Intensity

Notes: This table reports robustness checks of the effects of stabilization policies on firms' short-term cash flow, reopening decision and expectations to reopen within one month. Columns 1, 4 and 7 include geographical distance to Wuhan and industry dependence on Hubei Province. Columns 2, 5 and 8 include city level GDP per capita and ratio of fiscal expenditure to fiscal revenue. Columns 3, 6 and 9 include the number of of other stabilization policies enacted at the city level. All regressions control for SMEs' basic characteristics (sales, employment, age) interacted with year fixed effects, service-sector fixed effect, and city-level infection rates of COVID-19. Robust standard errors are clustered at city level. *** p < 0.01, ** p < 0.05, * p < 0.1.

Table A10.

Robustness Checks for the Effects of Lock-Down Policies.

	Reopen	Reopen <1 Month	Reopen	Reopen <1 Month	Reopen	Reopen <1 Month
	(1)	(2)	(3)	(4)	(5)	(6)
Social Distancing	−0.063*	−0.111***			−0.005	−0.098**
	(0.037)	(0.034)			(0.036)	(0.041)
Highway Closure			−0.129***	−0.100***
			(0.026)	(0.035)
Highway Opening Rate					1.245***	0.332
					(0.285)	(0.303)
Additional Controls	Wuhan + Hubei	Wuhan + Hubei	Wuhan + Hubei	Wuhan + Hubei	Logistics	Logistics
Observations	1806	1460	1806	1460	1806	1460
R-Squared	0.046	0.077	0.063	0.078	0.062	0.075

Notes: This table reports robustness checks of the effects of lock-down policies on SMEs' reopening status by the survey dates, and whether they expect to reopen in one month, if not reopened yet. Columns 1 to 4 include geographical distance to Wuhan and industry dependence on Hubei Province. Columns 5 and 6 include highway opening rate. All regressions control for SMEs' basic characteristics (sales, employment, age) interacted with year fixed effects, service-sector fixed effect, and city-level infection rates of COVID-19. Robust standard errors are clustered at city level. *** p < 0.01, ** p < 0.05, * p < 0.1.

Table A11.

Correlations between Firm Characteristics and Medium-Run Policy Coverage.

	Social Security/Employment	Tax Exemptions/Extensions	Rent Reduction	Credit/Loan support
	(1)	(2)	(3)	(4)
Party Member	−0.035	0.044	−0.002	0.018
	(0.031)	(0.033)	(0.029)	(0.024)
Annual Sales	0.070***	0.005	0.009	0.029**
	(0.017)	(0.017)	(0.016)	(0.013)
Total Employment	0.069***	0.028**	0.044***	0.025***
	(0.011)	(0.011)	(0.010)	(0.008)
Firm Age	0.008	−0.002	0.011**	0.002
	(0.005)	(0.005)	(0.005)	(0.004)
Ex-ante Banking Relationship	−0.008	0.015	−0.077**	0.059*
	(0.040)	(0.042)	(0.037)	(0.031)
Self-employed	−0.237***	−0.240***	0.021	−0.122***
	(0.058)	(0.061)	(0.055)	(0.045)
Observations	1682	1682	1682	1682
R-Squared	0.128	0.059	0.025	0.034

Notes: This table displays the correlation between SME owner's party membership, annual sales, staff size, fim age, registration type, ex-ante banking relationship and coverage of stabilization policies in the medium run. ***, **, * denote statistical significance at 1, 5, and 10% levels.

Table A12.

Matching Results for Medium-Run Policy Effects with High Credit Demand.

	Cash <1 Month	Reopen	Labor Recovery >50%
Treatment group	0.137	0.906	0.824
Control group	0.247	0.904	0.792
ATT	−0.211***	0.006	−0.013
	(0.061)	(0.035)	(0.048)
Number of matched pairs	175	175	159

Note: This table reports the estimated average treatment-on-the-treated (ATT) effects of credit of loan support policies on SMEs' outcomes on subsample with high credit demand, based on the propensity score matching (PSM) method. The matching covariates include SMEs' basic characteristics (sales, employment, age, service sector indicator), geographical distance to Wuhan, and industry dependence on Hubei Province. Robust standard errors are reported in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.