COVID-19 Pandemic and Corporate Liquidity: The Role of SOEs’ Trade Credit Response

Abstract

Although state-owned enterprises are associated with less efficiency and lead to resource misallocation, they may have stabilizing effect in face of a crisis. Exploiting the COVID-19 pandemic as a natural experiment, we study the role of firm ownership in trade credit provision and find robust evidence that SOEs increase their trade credit to downstream firms more than non-SOEs after the outbreak of the pandemic. Moreover, we explore the underlying mechanism and find that better financing ability and multitask of the SOEs contribute to greater trade credit during the pandemic, and the latter plays a more active role. Further analyses show that SOEs’ advantage in trade credit extension is more pronounced in industries with higher external financial dependence and provinces with a higher level of government involvement, suggesting that SOEs might have greater comparative advantage in screening due to its involvements in local economy during crisis periods. Our paper provides new insights into the real effects of SOEs on the economy.

1. Introduction

One of the most notable features in Chinese economy is the prevalence of the state-owned enterprises (SOEs). Although the state ownership has declined dramatically during the past two decades, the share of SOEs in industrial sector is still high (Figure A1 and Figure A2 in the online appendix). The coexistence of SOEs and non-SOEs is one of the most prominent features in Chinese economy even after the privatization process since 1990s. How to explain the existence of SOEs lies in the center to understand China’s reform and development strategy. While an important literature examines the efficiency loss of SOEs in production, investment and asset allocation (Song et al., 2011, Huang et al., 2017), the 2008 financial crisis arouse academic and policy attention on the role of SOEs in stabilizing the economy (Reddy et al., 2016; Liu, 2019, Goldman, 2020). However, despite SOEs’ distinctive characteristics as a supplier and their possible macroeconomic consequences, little is known about whether and how SOEs affect the resilience of domestic private-owned enterprises (POEs) to severe economic shocks.

Figure 1.

Response of trade credit provision to subsidy. Note: The figure shows the estimated response of trade credit provision to subsidy for SOEs and non-SOEs, and the differential response between SOEs and non-SOEs with 95% confidence internal.

In this paper, we examine an important mechanism that may affect domestic POEs’ resilience to negative shocks: the trade credit channel. Specifically, our study focuses on the role of trade credit in SOEs’ alleviation of negative liquidity shocks to downstream firms. We are motivated by three stylized facts documented in the prior literature. First, evidence shows that almost two-thirds of international trade is supported by trade credit (Bank for International Settlements, 2014). In the integrated world weaving of payables and receivables, between-firms borrowing creates a flow of trade credit that runs parallel to the flow of goods and service along supply chains. As an important source of financing for downstream firms, trade credit plays a substantial role in firms’ external financing especially in tight domestic credit periods or in financially less developed economies (Fisman and Love, 2003, Love et al., 2007, Lin and Ye, 2018).

Second, it is well documented in the SOEs’ literature that two related important sources of SOEs’ financing advantage come from their better access to domestic financial markets and easy access to government preferential policies (Song et al., 2011, Hsieh and Song, 2015, Harrison et al., 2019), thus SOEs are less financially constrained than private firms in developing countries. Besides, SOEs will put more emphasis on economic stability in the times of crisis as revealed by the multitask theory of SOEs (Bai et al., 2006). These features of SOEs are taken into consideration in the theoretical model on Chinese economy (Chang et al., 2019, Liu et al., 2021).

Third, the difference in the motivation between SOEs and non-SOEs could shape the differential response of trade credit provision during crisis periods. On the one hand, SOEs will pursue non-economic goals according to the multitask theory of SOEs, which leads to some positive externalities (Putniņš, 2015). For example, SOEs provide trade credit to downstream firms during the crisis. This will help downstream firms to avoid going bankrupt and laying off workers, which enables the economy to recover from the crisis more rapidly. Macroeconomic stability could help to reduce systematic risk, uncertainty, and thus facilitate planning within non-SOEs while maintaining high employment could deter the crime, social upheaval and improve social wellbeing, and thus could be considered as a public good. Due to the implicit relationship between SOEs and the government, SOEs will be used to internalize the externality. On the other hand, the trade credit provision involves risk to the creditor, which will discourage firms to provide trade credit when facing with serious shocks. However, SOEs are more likely to tolerate the risk than non-SOEs. The state as an investor is highly diversified and thus generate a portfolio effect which allows SOEs to accept greater amounts of risk (Grøgaard et al., 2019). Moreover, the bailout of the government and the soft budget constraint mitigate the threat of bankruptcy and thus allow SOEs to tolerate greater amounts of risk (Musacchio et al., 2015, Benito et al., 2016, Rygh, 2018). Therefore, SOEs are more willing to extend trade credit relative to non-SOEs especially when risks are potentially high. In summary, SOEs not only have easy access to credit, but also are more motivated to provide trade credit.

Motivated by these stylized facts, we conjecture that financially less constrained SOEs with multitask are capable of extending more trade credit than non-SOEs during crisis periods. To test this hypothesis, we employ a difference-in-differences (DID) strategy in which we compare trade credit supplied by firms with different ownership positions before and after the start of the COVID-19. Using COVID-19 crisis as a natural experiment, we first provide evidence that SOEs extend more trade credit than domestic private firms after the onset of the COVID-19. Furthermore, we find that better financing ability and multitask of SOE are two mechanisms through which SOEs could extend more trade credit during the pandemic, and multitask mechanism is more important to motivate SOEs to provide trade credit.

Our results contribute to the related literature in the following aspects. First, we identify a trade credit channel through which SOEs could provide liquidity during the crisis. To the best of our knowledge, this channel is still unexplored in the literature on the performance and existence of SOEs. The predominant view is that SOEs tend to generate resource misallocations and harm domestic economies (Kruger, 1990). Yet, our findings show that there may be an economic rationale behind the certain aspects of Chinese SOEs, and these SOEs might have mitigated the negative liquidity shocks of POEs and alleviated negative shocks from crisis to the real economy. As existing literature focus on either the efficiency of SOEs, or easy access to credit (Song et al., 2011, Liu, 2019), we move one step forward by exploring the role of trade credit extended of SOEs in stabilizing the economy during times of crisis. Chen et al. (2021) identify the need for such research, but there is currently limited direct evidence. We therefore add to the production networks literature by directly providing the evidence that SOEs extend more trade credits during the crisis. While our study by no means suggests that the existence of SOEs is optimal, our results do show that the financing advantage and multitask of SOEs, especially in the crisis periods, could affect financial conditions of downstream non-SOEs through production network, indicating a stabilizing effect of SOEs on domestic economy. SOEs could play a positive role during the crisis. However, the low efficiency of SOEs could aggravate the resource misallocation and damage the long-run economic performance. In this paper, we only verify that SOEs are willing to provide more trade credit during the crisis.

Second, our findings complement the studies on the role of state shareholders in the corporate governance. Specifically, we provide some new insights on the role of SOEs and the real effect of SOEs on the economy during the crisis. Besides the easier access to external finance, we show that SOEs might have informational advantage in providing trade credit. Information advantage is the principal element for the liquidity provision of financial intermediary, especially bank (Diamond, 1984, Fama, 1985). Local government has information advantage over central government (Qian et al., 2006, Huang et al., 2017) and SOEs have access to government information and data (Capobianco and Christiansen, 2011). The results confirm that local government-owned SOEs extend more credit provisions relative to central government-owned SOEs after controlling the time-varying factors at regional level. Furthermore, the trade credit providing effects of SOEs are more pronounced in industries with more dependence on external finance and in provinces with higher government involvement. The cost of external financing in these industries and areas tends to be higher, indicating that SOEs are more capable in monitoring in terms of extending trade credit probably due to information advantage in production networks.

We also present evidence on the effect of state-owned shares on trade credit provision during the crisis. The monitoring of principal could affect the agency risk (Jia et al., 2019). Following the literature on blockholder, the larger blockholder with further interests at stake will exert more efforts in monitoring the agency (Dharwadkar et al., 2008). Hence, we check the relationship between state-owned shares and trade credit provision and find that a firm with higher state-owned shares is likely to provide more trade credit after the outbreak of COVID-19.

Third, our study complements to the trade credit literature by investigating the heterogeneity in trade credit provision between SOEs, foreign-owned enterprises (FOEs) and domestic POEs and exploring the impact of COVID-19 shock on firms’ trade credit extension. Cull et al. (2009) show that both SOEs and profitable POEs are likely to redistribute the credit during China’s transition. The trade credit is a substitute for bank loans for firms that are shut out of the formal finance. However, they do not explore the role played by SOEs during the crisis. Lin and Ye (2018) document that FOEs provide more trade credit than local POEs in China during tight domestic credit period and that the global liquidity shock due to the 2007-2008 financial crisis attenuates the willingness of FOEs to provide trade credit. Our study departs from theirs in that we identify SOEs as another type of trade credit provider during a crisis period and the magnitude of trade credit extended by SOEs roughly amounts to those extended by FOEs. Moreover, Wang et al. (2019) investigate the inform financing from the demand side of trade credit. Non-SOEs reply more on trade credit financing and when the external financing retrenches during the financial crisis, they will switch financing with trade credit. Ge and Qiu (2007) present similar findings from the demand side of trade credit. Our work also differs because we focus on the supply side of trade credit based on the multitask theory of SOEs.

Finally, we extend our main analysis to the global financial crisis and our paper is related to a large and growing literature concerning the real effects of crisis, especially the COVID-19 crisis, on firms’ behaviors (Ferrando and Ganoulis, 2020, Guerrieri et al., 2020, Jordà et al., 2020). In particular, our work is closely related to the studies by Garcia-Appendini and Montoriol-Garriga, 2013, Carbó-Valverde et al., 2016, Bureau et al., 2021, Chen et al., 2021 and Al-Hadi and Al-Abri (2022). Garcia-Appendini and Montoriol-Garriga (2013) analyze the effect of the 2007-2008 financial crisis on between-firm liquidity provision. Firms with high pre-crisis cash holdings level are likely to provide trade credit to other firms as compared with low pre-crisis cash holdings. Carbó-Valverde et al. (2016) focus on the effect of crisis on small and midsize enterprises (SMEs), and find that credit constrained SMEs rely on trade credit to finance capital expenditure during the period of credit crunch. Bureau et al. (2021) find a trade credit channel which amplify the demand shock faced by firms during the COVID-19 crisis by examining the daily data of payment defaults on suppliers. Al-Hadi and Al-Abri (2022) present cross-country evidence on firms’ trade credit responses to COVID-19-induced monetary policy and fiscal policy and find that there is a difference between trade credit led by monetary intervention and that induced by fiscal intervention. Chen et al. (2021) also provide international evidence on the relationship between SOEs and trade credit. However, the trade credit provision of SOEs will be different in the case of crisis relative to the case of normal period. We add to this literature by identifying another channel through which the firms’ ownership can significantly impact firms’ trade credit provision. The heterogeneity in trade credit extension between SOEs and POEs could at least partially offset the negative shock brought by COVID-19 on the real economy.

The reminder of this paper is organized as follows. Section 2 introduces the data and describes the empirical identification strategy. Section 3 presents the baseline estimation results exploring the consequences of COVID-19 in shaping SOEs’ behavior on trade credit provision. Section 4 tests the robustness of the baseline results and provides more evidences that support the causal explanation of the effect of COVID-19 shock on SOEs’ trade credit extended to their clients. Section 5 explores the mechanisms. Section 6 extends the analysis of information advantage of SOEs during the pandemic and section 7 concludes.

2. Data and Econometric Strategy

2.1. Data

This section discusses the data and empirical strategy used to identify the effect of COVID-19 shock on SOEs’ trade credit provision. Our main variables and firm level characteristics data are obtained from Chinese Stock Market Accounting Research (CSMAR) database. We select A-share listed manufacturing companies in both Shanghai and Shenzhen securities exchanges from the 1^st quarter of 2017 to the 4^th quarter of 2020. We exclude observations with missing variables used in this study. The whole sample consists of 21581 firm-year-quarter observations. We winsorize the variables at the 1^st and 99^th percentiles to lessen the influence of outliers.

2.2. Variables

2.2.1. Trade credit provision

To measure our key dependent variable, we follow Levine et al. (2018) and begin with accounts receivable which is a stock entry on balance sheet. We calculate the change in accounts receivable from the beginning of a quarter to the end of this quarter. The positive change suggests that firms provide more goods and services than customers’ pay while the negative change suggests that firms not only receive payments for goods and services that are provided, but also take back between-firm lending from customers.

We calculate the change of accounts receivable to sales ratio (Credit_pro) for each firm. To test our conjecture on the role of SOEs in mitigating the COVID-19 shock, we focus on firms’ willingness to provide trade credit. Besides, we examine the trade credit that firms obtained and calculate the change of accounts payable scaled by cost of goods sold (Credit_rec). We also study net trade credit defined as the change of difference between accounts receivable and payable scaled by sales (Netcredit_pro). Net trade credit reflects the relative willingness to extend trade credit, net of the credit that firms receive themselves.

2.2.2. Firms’ ownership

The key independent variable in our empirical analysis is firms’ ownership. We classify firms based on the nature of equity of the actual controller, the data of which comes from China Listed Firm’s Equity of Nature Research Database, a sub-database of CSMAR. From 2003, the listed firms are required to disclose information on actual controller besides controlling shareholder. Specifically, a firm is classified as a SOE if its actual controller is a SOE, central authority, such as State-owned Assets Supervision and Administration Commission of the State Council or local authority, such as State-owned Assets Supervision and Administration Commission of People’s Government of Beijing Municipality. We construct a dummy variable (Soe) indicating 1 if a firm is a SOE and 0 otherwise.

2.2.3. Other controls

Besides firm’s ownership, there are other firm-specific characteristics that may affect the provision of trade credit, including firms’ age, size, profit margin, liquidity, long-term debt, Tobin’s Q and market concentration which have been emphasized in the literature (Lin and Ye, 2018, Levine et al., 2018). Firm’s age (Age) is the number of years since the establishment and firms’ size (Size) is measured by the total assets adjusted by 2010’s price. We take the logarithm of age and size in the empirical model respectively. The profit margin (Profit) measured by the profit to sales ratio is used to capture firms’ difference in profitability and the liquidity (Liquidity) measured by the share of liquid assets in total assets is used to control for the state of firms’ financial health. Additionally, we include long-term debt to total assets ratio (Debtratio) and the market concentration (Topsale) computed as the total market share of the top five firms at industry-year level to control for the effect of product market structure on trade credit provision. Tobin’s Q (Tobinq) is computed as the share of market value of total assets to book value of total assets.

Table 1 provides summary statistics for these variables from the 1^st quarter of 2017 to the 4^th quarter of 2020. The average age of firms in the sample is about 19 years and the average of size reaches 80.26 million RMB that is 12.30 million USD in terms of exchange rate at the end of 2020. The share of SOEs in the sample is 27.5% and the share of foreign owned enterprises (FOEs) is 3.12% which is quite lower than that of SOEs. Therefore, we mainly focus on the difference between SOEs and non-SOEs in the following empirical analysis. We also compare these variables of SOEs with that of non-SOEs. Table 1 also shows that SOEs are more mature with larger size, more cash holding and smaller Tobin’s Q.

Table 1.

Summary statistics. This table shows the summary statistics of Chinese manufacturing listed firms from the 1^st quarter of 2017 to the 4^th quarter of 2020.The last three columns compare the characteristics of SOEs and non-SOEs. Columns 5-6 present the time-series average of corresponding variables for SOEs sample and non-SOEs sample. The last column presents the difference between the time-series average of corresponding variable for SOEs sample and non-SOEs sample. *** significant at 1%, ** significant at 5%, * significant at 10%.

	Full Sample					Subsample
Variable	Mean	Std. Dev.	p50	Obs		SOE	non-SOEs	Diff.
Credit_pro	0.0195	0.2522	0.0132	21581		-0.0153	0.0328	-0.0481
Credit_rec	0.0052	0.1664	0	21581		0.0019	0.0064	-0.0045
Netcredit_pro	0.0250	0.2655	0.0060	21581		-0.0116	0.0389	-0.0505
Shortdebt	0.0772	0.5651	0.0467	21581		-0.0337	0.1193	-0.1530**
Longdebt	0.0180	0.3170	0	21581		0.0157	0.0180	-0.0023
Equity_1	0.0281	0.1653	0	21581		0.0278	0.0282	-0.0004
Equity_2	0.0012	0.0347	0	21581		0.0012	0.0012	-3.46e-5
Cost_1	0.7243	0.1761	0.7532	21581		0.7784	0.7038	0.0746***
Cost_2	0.9781	0.2710	0.9476	21581		0.9800	0.9774	0.0026
Internal_fin	0.0126	0.7057	0.0423	21581		0.0134	0.0124	0.0010
Age(year)	19.4649	5.5399	19	21581		21.6245	18.6458	2.9787***
Asset(10,000RMB)	8026.2	18069.3	3178.3	21581		14678.0	5503.5	9174.5***
Tobinq	1.8483	1.0719	1.5310	21581		1.7194	1.8972	-0.1778***
Profit	0.0255	2.9126	0.0583	21581		0.0241	0.0260	-0.0019
Liquidity	0.5452	0.1596	0.5450	21581		0.5549	0.5416	0.0133
Debtratio	0.0727	0.0786	0.0491	21581		0.0721	0.0730	-0.0009
Cash	0.1388	0.0891	0.1193	21581		0.1459	0.1361	0.0098**
Topsale	0.2352	0.1128	0.2279	21581		0.2466	0.2308	0.0158**
Soe	0.2750	0.4465	0	21581		-	-	-
Fdi	0.0312	0.1738	0	21581		-	-	-

2.3. Econometric strategy

To analyze the stabilizing effect of SOEs during crisis periods, we exploit the COVID-19 shock which offers a particularly useful setting for this study. In particular, we focus on the difference in trade credit provision behavior between SOEs and non-SOEs as COVID-19 shock unfolds. We employ a method which is similar with Lin and Ye (2018) and Goldman (2020). Specifically, we compare the trade credit provision of firms before and after the outbreak of COVID-19 as a function of government ownership using a difference-in-differences framework. The COVID-19 shock breaks out unexpectedly, and is therefore plausibly exogenous to firms’ behavior.

We are mostly interested in examining the role of firms’ government ownership in mitigating the COVID-19 shock. However, the inferences may be confounded if variation in the ownership as COVDI-19 unfolds is endogenous to unobserved variation in trade credit provision. Our baseline specification, as well as the rest of our analysis described below, is designed to address this issue. Since changes in ownership as the COVID-19 shock unfolds may be related to unobserved changes in trade credit provision, we focus on firms with ownership unchanged during our sample range.

Our baseline specification regresses firm-level quarterly trade credit provision over the 1^st quarter of 2017 to the 4^th quarter of 2020 on the interaction between firm’s ownership dummy and the pandemic indicator, controlling for relevant firm-specific characteristics, firm fixed effects and time fixed effects. Our specification focuses on the supply side of trade credit. We should control for the demand factors from downstream firms, and thus construct a variable indicating the demand of trade credit following Lin and Ye (2018). To this end, we first obtain each downstream industry’s degree of trade credit dependence from Fisman and Love (2003). The construction of industry’s dependence on trade credit follows Rajan and Zingale (1998) which measure the industry’s dependence on external finance, and use the U.S. data on the ratio of accounts payable to total assets. The industry’s dependence on trade credit reflects the technological aspect of dependence on trade credit innate to each industry. Next, we identify the downstream industry for industry i from China’s input-output table and obtain the intermediate input usage in the downstream industry. We calculate the intermediate input usage share of each downstream industry as a share of intermediate input usage produced by industry i to total production of industry i. Finally, we employ the intermediate input usage share as a weight and compute industry i’s downstream trade credit dependence as the weighted average of its downstream industries’ trade credit dependence which is denoted as Tcreditdep. We then include its interaction term with COVID-19 shock dummy to test whether our results are driven by the supply side or the demand side. In sum, to test our hypothesis, we estimate the following specification.

$$Y_{\mathit{it}}={\mu }_i+\alpha \left({{\mathit{Soe}}_i\times Shock}_t\right)+{\beta \left({{\mathit{Tcrditdep}}_i\times Shock}_t\right)+\gamma X_{i,t-1}+\tau }_t+{\epsilon }_{\mathit{it}}$$ (1)

Where Y indicates the dependent variable, such as the amount of trade credit extended by firm i of year-quarter t scaled by its sales. Soe is an ownership dummy that takes the value of 1 for SOEs and 0 for others. Shock is an indicator variable that takes 1 if the quarter is after the outbreak of COVID-19 that is the 1^st quarter of 2020 and afterwards. X is a vector presenting a set of firm-specific control variables. We also include the interaction term of downstream industries’ trade credit dependence (Tcreditdep) with COVID-19 shock dummy to capture the demand of trade credit in the downstream. μ_i and τ_t are firm fixed effects and year-quarter fixed effects respectively. The inclusion of firm fixed effects absorbs the level effect of government ownership, and control for all sources of time-invariant variation in trade credit provision across firms and year-quarter time fixed effects capture the unobserved factors potentially affecting trade credit provision that are common to all firms but vary across time, such as fluctuations of economy as well as the changes in macroeconomic policy. ε_it is an error term. Following Bertrant et al. (2004), standard errors are heteroskedasticity-consistent and clustered at firm level.

We are particularly interested in the coefficient of the interaction term between the SOEs’ dummy and the COVID-19 shock dummy. We expect that SOEs will provide more trade credit after the outbreak of COVID-19. Therefore, a positive coefficient of the interaction term would be consistent with our hypothesis. We also take the net trade credit provision into consideration and examine the differential responses in net trade credit extended between SOEs and non-SOEs to COVID-19 shock. Next, we examine the dynamic effects. On the one hand, if the unobservable factors affecting both ownership and trade credit provision lead to our results, then we could detect this effect before the outbreak of COVID-19. On the other hand, the parallel trend could be checked in this way.

The identification of the main coefficient estimates in equation (1) relies on variation across firms and over time. Furthermore, we identify the coefficients of interest from alternative specifications, for example, by examining the dynamic effects of the pandemic, controlling for potential omitted variables, employing propensity score matching (PSM)-DID approach, using 2008 financial crisis as another exogenous shock and exploring alternative firm ownership measures. We also explore the mechanisms through which firm ownership affects the extension of corporate trade credit.

3. Results

3.1. Baseline results

Table 2 shows the results from estimating equation (1) for trade credit provision based on quarterly firm-level data from the 1^st quarter of 2017 to the 4^th quarter of 2020. Firm fixed effects and year fixed effects are included in column (1). Firm fixed effects and year-quarter fixed effects are included in the rest specifications to capture any time-invariant characteristics of firms that may drive trade credit policy and any macroeconomic fluctuations that may correlate with trade credit provision decisions. In column (3), we add control variables which are lagged to account for the effect of firm specific characteristics on changes in trade credit extension. In column (4), we further include the interaction of downstream trade credit dependence with COVID-19 shock dummy to control for the downstream demand on trade credit.

Table 2.

Ownership and trade credit provision. This table reports the estimation results from the baseline regression. The dependent variable is Credit_pro in Panel A and is sales (in log) in Panel B. The main independent variable is Soe interacted with shock which is an indicator variable that equals 1 for SOEs after COVID-19 shock, and 0 otherwise. Tcreditdep denotes trade credit dependence of the downstream industry which captures the demand factors of the downstream firms. Other Controls is a vector of firm-level control variables which are lagged, including firm age (in log), total assets (in log), firm Liquidity (liquid assets to total assets ratio), Profitability (operating profit to sales ratio), Debtratio (long term debt to assets ratio), Tobinq and Topsale (market concentration). The main sample covers the periods from the 1^st quarter of 2017 to the 4^th quarter of 2020. Standard errors (in parentheses) are robust to heteroscedasticity and clustered at firm level. *** significant at 1%, ** significant at 5%, * significant at 10%.

Panel A	Dependent Variable: Credit_pro
	(1)	(2)	(3)	(4)
Soe×Shock	0.1278**	0.1268***	0.1327**	0.1289**
	(0.0513)	(0.0478)	(0.0532)	(0.0517)
Tcreditdep× Shock	No	No	No	Yes
Other Controls	No	No	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	No	No	No
Year-quarter FE	No	Yes	Yes	Yes
Observations	21581	21581	18718	18718
R2	0.0697	0.0701	0.0787	0.0787
Panel B	Dependent Variable: Sales(in log)
	(1)	(2)	(3)	(4)
Soe×Shock	0.0385*	0.0311	0.0609***	0.0619***
	(0.0209)	(0.0208)	(0.0173)	(0.0174)
Tcreditdep× Shock	No	No	No	Yes
Other Controls	No	No	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	No	No	No
Year-quarter FE	No	Yes	Yes	Yes
Observations	21581	21581	18718	18718
R2	0.9187	0.9297	0.9443	0.9444

In Panel A of Table 2, we find a positive and significant coefficient on the Soe×Shock term, suggesting that SOEs increase their trade credit to downstream firms more than non-SOEs after the outbreak of the pandemic. In column 4, the difference-in-differences coefficient on Soe $\times$ Shock is 0.1289 and significant at the 5% level, indicating a differential increase in trade credit (scaled by sales) for SOE firms of 12.89 percent, relative to that for non-SOE firms. This size effect is economically large since it represents 16.6% of the sample average of accounts receivable.

There are two potential explanations for the credit provision of SOEs after COVID-19 shock. SOEs could make a large amount of credit sales or allow customer to pay more slowly. We could verify the increased sales on credit directly. We take the logarithm of sales and use the DID specification to investigate whether SOEs increases sales or not when comparing with non-SOEs after COVID-19 shock. Panel B of Table 2 presents the results. From columns (1) through (4), we find a positive and significant coefficient on the Soe×Shock, suggesting that SOEs indeed extends more trade credit to customers through selling more stuff on credit.

3.2. Accounts payable and net trade credit

To further corroborate our conjecture, we examine the differential responses in accounts payable and net trade credit extended between SOEs and non-SOEs to COVID-19 shock. Table 3 presents the results. The dependent variable in columns (1)-(3) is Credit_rec. The estimated coefficient on Soe×Shock is positive and insignificant in all specifications, which suggests that there is no differential response in accounts payable received between SOEs and non-SOEs to COVID-19 shock. There is no evidence that SOEs make less use of trade credit relative to others after the outbreak of COVID-19. This finding is similar to that in Love et al. (2007). They find that the effect of crisis on trade credit usage and provision is different. Financially less vulnerable firms are willing to provide trade credit relative to financially more vulnerable firms. However, there is no significant difference in the usage of trade credit. The dependent variable in columns (4)-(6) is Netcredit_pro. We find a positive and significant estimated coefficient on Soe×Shock in all specifications, which suggests that SOEs provide more trade credit relative to non-SOEs after the outbreak of COVID-19.

Table 3.

Ownership and net trade credit provision. This table reports the results of regressions in which the dependent variable is trade credit acquisition (Credit_rec) in columns (1)-(3) and net trade credit provision (Netcredit_pro) in columns (4)-(6) respectively. The main independent variable is Soe interacted with shock which is an indicator variable that equals 1 for SOEs after COVID-19 shock, and 0 otherwise. Tcreditdep denotes trade credit dependence of the downstream industry which captures the demand factors of the downstream firms. Other Controls is a vector of firm-level control variables which are lagged, including firm age (in log), total assets (in log), firm Liquidity (liquid assets to total assets ratio), Profitability (operating profit to sales ratio), Debtratio (long term debt to assets ratio), Tobinq and Topsale (market concentration). The main sample covers the periods from the 1^st quarter of 2017 to the 4^th quarter of 2020. Standard errors (in parentheses) are robust to heteroscedasticity and clustered at firm level. *** significant at 1%, ** significant at 5%, * significant at 10%.

Dependent Variable:	Credit_rec				Netcredit_pro
	(1)	(2)	(3)		(4)	(5)	(6)
Soe×Shock	0.0160	0.0084	0.0079		0.1164**	0.1166**	0.1131**
	(0.0199)	(0.0161)	(0.0162)		(0.0497)	(0.0537)	(0.0522)
Tcreditdep× Shock	No	No	Yes		No	No	Yes
Other Controls	No	Yes	Yes		No	Yes	Yes
Firm FE	Yes	Yes	Yes		Yes	Yes	Yes
Year-quarter FE	Yes	Yes	Yes		Yes	Yes	Yes
Observations	21581	18718	18718		21581	18718	18718
R2	0.1521	0.1486	0.1486		0.0714	0.0813	0.0813

3.3. Dynamic effects

The previous results compare firms’ behavior on trade credit before and after the COVID-19 shock. We next studies the dynamics of differential response of SOEs and non-SOEs to COVID-19 shock. We do this by constructing a series of dummy variables in the standard regression to trace out the quarter-by-quarter effects of firms’ ownership on trade credit extension. We estimate the following specification.

$$Y_{\mathit{it}}={\mu }_i+{\sum }_{-3}^3{\alpha }_n({\mathit{Soe}}_i\times {\mathit{Shock}}^n)+\beta X_{i,t-1}{+\tau }_t+{\epsilon }_{\mathit{it}}$$ (2)

Where the COVID-19 shock dummy variables, Shock^-j and Shock^+j equal 0 except as follows: Shock^-j equals 1 for the j^th quarter before the 1^th quarter of 2020 while Shock^+j equals 1 for the j^th quarter after the 1^th quarter of 2020. We interact Soe with these dummies respectively and these interaction terms capture the dynamic effects. We consider the first three quarters after the outbreak of COVID-19, the quarter of the outbreak of COVID-19 and the previous three quarters, and exclude the rest quarters in our sample, and thus estimate the dynamic effect of firms’ ownership on the trade credit provision relative to the quarters at the beginning of our sample. μ_i and τ_t are firm fixed effects and year-quarter fixed effects respectively.

Table 4 shows the results from estimating equation (1) by adding a series of additional interaction terms, Soe×Shock^-j and Soe×Shock^+j. The dependence variable in column (1) is Credit_pro, while column (2) reports the results for Netcredit_pro. As shown, the coefficients on the Soe×Shock^-j in both specifications are insignificantly different from zero, with no trend in the differential response in trade credit between SOEs and non-SOEs prior to COVID-19 shock. Whereas, the coefficients on Soe×Shock⁰, Soe×Shock¹ and Soe×Shock² are positive and statistically significant at the 10% level. It is worth noting that trade credit provided by SOEs increases immediately after the outbreak of COVID-19 relative to non-SOEs and such effect fades away after the 3^rd quarter of the year 2020. This suggest that SOEs in China react promptly to COVID-19 shock. The check on dynamic effect is in line with the previous findings. In consistence with the parallel trend assumption, the result also suggests that our findings are not driven by pre-event trend heterogeneity in trade credit provision between SOEs and non-SOEs.

Table 4.

Ownership and trade credit provision: Dynamic effects. This table reports the dynamic effects of the pandemic on trade credit provision. The dependent variable is trade credit provision (Credit_pro) in column (1) and net trade credit provision (Netcredit_pro) in column (2) respectively. The main independent variable is the interaction of ownership dummy (Soe) with a series of shock dummies (shock ^j). We consider the first three quarters after the outbreak of COVID-19, the quarter of the outbreak of COVID-19 and the previous three quarters, and exclude the rest quarters in our sample, and thus estimate the dynamic effect of firms’ ownership on the trade credit provision relative to the quarters at the beginning of our sample. Tcreditdep denotes trade credit dependence of the downstream industry which captures the demand factors of the downstream firms. Other Controls is a vector of firm-level control variables which are lagged, including firm age (in log), total assets (in log), firm Liquidity (liquid assets to total assets ratio), Profitability (operating profit to sales ratio), Debtratio (long term debt to assets ratio), Tobinq and Topsale (market concentration). The main sample covers the periods from the 1^st quarter of 2017 to the 4^th quarter of 2020. Standard errors (in parentheses) are robust to heteroscedasticity and clustered at firm level. *** significant at 1%, ** significant at 5%, * significant at 10%.

Dependence Variable:	Credit_pro		Netcredit_pro
	(1)		(2)
Soe×Shock3	0.0458		0.0936
	(0.0494)		(0.0751)
Soe×Shock2	0.1335*		0.1362*
	(0.0699)		(0.0749)
Soe×Shock1	0.1218*		0.1550**
	(0.0692)		(0.0759)
Soe×Shock0	0.1699**		0.1313*
	(0.0713)		(0.0789)
Soe×Shock-1	0.0658		0.0552
	(0.1182)		(0.0743)
Soe×Shock-2	0.0809		0.1080
	(0.0796)		(0.0849)
Soe×Shock-3	0.1035		0.1192
	(0.0712)		(0.0834)
Tcreditdep× Shock	Yes		Yes
Other Controls	Yes		Yes
Firm FE	Yes		Yes
Year-quarter FE	Yes		Yes
Observations	18718		18718
R2	0.0701		0.0714

4. Robustness

4.1. Global financial crisis

We extend our main analysis to the global financial crisis (2008) and present evidence supporting our hypothesis. Garcia-Appendini and Montoriol-Garriga (2013) point out that the rich cash firms provide more trade credit during the financial crisis. However, the role of ownership could also affect firm’s motivation on trade credit provision. Chen et al. (2021) present cross-country evidence that SOEs provide more trade credit relative to non-SOEs, and institutional factors could shape firm’s trade credit behavior.

We repeat the main empirical analysis in the context of global financial crisis. The data covers the period from the 1st quarter of 2005 to the 4th quarter of 2010. We construct a shock dummy (Crisis), which equals 1 after 2007, and 0 otherwise. We rerun the estimation following equation (1). The dependent variable is Credit_pro in columns (1)-(2), and is Netcredit_pro in columns (3)-(4), respectively. Table A1 of the online appendix presents the results. In panel A, we only check the difference in the trade credit provision before and after the global financial crisis for SOEs and non-SOEs separately. Columns (1) and (3) focus on non-SOEs and columns (2) and (4) focus on SOEs. We control for a vector of firm-level characteristics, GDP growth rate, market concentration at industry-level, firm fixed effects and quarter fixed effects. The coefficients on Crisis in columns (2) and (4) are positive and significant. However, the coefficients on Crisis in columns (1) and (3) are insignificant. These findings reveal that SOEs provides more trade credit after the outbreak of global financial crisis while non-SOEs do not response in trade credit provision to the crisis. In Panel B, we are interested in the estimated coefficients of Soe×Shock which are positive and significant, suggesting that SOEs provide more trade credit after the global financial crisis relative to non-SOEs. The evidence of global financial crisis further supports our hypothesis.

4.2. Role of cash holdings

We first examine whether the findings are robust when we take the role of cash holdings into consideration. When liquidity in financial market is scarce, the lower opportunity cost of funds encourages cash-rich firms to extend liquidity through an increase in trade credit (Garcia-Appendini and Montoriol-Garriga, 2013). As shown in Table 1, cash to assets ratio of SOEs is statistically greater than that of non-SOEs at the 5% level, indicating that SOEs could perform better in face of COVID-19 shock. To control for the potential effects of cash holdings, we employ pre-pandemic cash to asset ratio, Cash_pre, which denotes the cash to asset ratio one quarter previous to the outbreak of COVID-19.

Table A2 reports the results. The coefficients on Cash_pre×Shock are positive but statistically insignificant in columns (1) and (3). The coefficients on Soe×Shock are all significantly positive, suggesting that even controlling for the difference in cash holdings between SOEs and non-SOEs, SOEs still provide more trade credit relative to non-SOEs, and this finding is not driven by the liquidity insurance of cash-rich firms.

We then consider the role of excess cash holdings. Following Opler et al., 1999, Dittmar and Mahrt-Smith, 2007, the excess cash (Excesscash) is defined as the difference between actual and predicted cash (see more detailed calculation of excess cash in Online Appendix B).

We further control for Excesscashpre×Shock in columns (2) and (4) of Table A2. Excesscashpre denotes the Excesscash one quarter previous to the outbreak of COVID-19. Estimation results show that the coefficients on ExcessCashpre×Shock are not significantly different from zero. The coefficients on Soe×Shock are still significantly positive, suggesting that the willingness of SOEs to provide trade credit is not driven by their cash holdings or excess cash holdings.

4.3. Role of foreign-owned enterprises

As discussed in Lin and Ye (2018), foreign-owned enterprises have financing advantage over private-owned enterprises. During a recession, foreign-owned enterprises will use their advantage in international financial market and provide liquidity to other firms through trade credit channel. If this is the case, our results will be underestimated. To further verify our hypothesis, we construct a new dummy variable Fdi which take value of 1 if the actual controller is foreign agent and 0 otherwise. Actually, FOEs are scarce in our sample, and the share of which is only 3.12%. Panel A in Table A3 presents the results. The dependent variable is Credit_pro in columns (1)-(3) and Netcredit_pro in columns (4)-(6), respectively. The coefficients on Fdi×Shock in all specifications are positive and significant, suggesting that FOEs provide more trade credit relative to POEs indeed. This is consistent with the findings of Lin and Ye (2018). Besides, the coefficients on Soe×Shock are still significantly positive. The magnitude of the coefficients on Soe×Shock do not change much when including Fdi×Shock. Panel B in Table A3 presents the results from estimating equation (1) for trade credit provision based on the sample excluding FOEs. Again, the coefficients on Soe×Shock in all specifications are also significantly positive, indicating the robustness of our baseline results

4.4. Excluding lockdown period

To prevent the escalation of COVID-19 shock, China locked down parts of its cities after the outbreak of COVID-19, which strictly curtailed personal mobility as well as economic activities. On 23^th January of the year 2020, Wuhan prohibited travel in and out of the city to control the spread of COVID-19 pandemic, and lifted outbound travel restrictions on April 8^th. To circumvent the effect of lockdown on supply side of trade credit, we exclude the sample of the 1st quarter in year of 2020 which is the strict lockdown period and examine the robustness of the effect of SOEs on trade credit provision after COVID-19 shock. Estimation results in Table A4 shows that the coefficients on Soe×Shock are positive and statistically significant at the 5% level in all specifications, indicating that SOEs experiencing COVID-19 shock extend more trade credit relative to non-SOEs even excluding the strict lockdown period.

4.5. Analysis based on matched sample

One potential problem relates to the endogeneity of the ownership. We have dropped the firms which have changed the ownership during the analysis period. However, the ownership itself could be correlated with unobservable factors affecting trade credit provision. If the ownership is the result of economic fact leading to changes in trade credit provision, then we should be able to detect the differential response between SOEs and non-SOEs before the COVID19 shock. Pre-trend test in Table 4 indicates that there is no differential response among these two groups of firms before COVID-19 shock.

In the case that the vector of controls is large, it is unlikely to find a similar untreated firm for every treated firm with regard to all characteristics. We therefore adopt a propensity score matching scheme to address this concern. In order to calculate the propensity scores, we use a Logit model to estimate the propensity score for a firm transferred to SOEs based on a broad vector of observable variables including firms’ age, size, liquidity, profit, long-term debt ratio, Tobin’s Q. All control variables are lagged by one period.

After the propensity scores have been estimated, we construct a control group for SOEs by matching algorithms of radius with caliper. The results of balancing test are presented in Table A5, indicating that the matching algorithm successfully meets the key test for balancing. We then check whether the results are robust to the matched sample. Table A6 repents the results. The sample in panel A is constructed using radius with one thousandth caliper and the caliper is set to one ten-thousandth in panel B. The coefficients on Soe×Shock in all specifications are significantly positive, which is in line with our previous findings.

5. Mechanism

SOEs provide more trade credit relative to non-SOEs during the crisis. There are several potential channels through which SOEs changes their behavior on trade credit. The first is the financing advantage possessed by SOEs. The second is the multitask of SOEs which puts more emphasis on economic stability. Besides, information advantage based on financing advantage and better access to government resources might also allow SOEs to extend trade credit during the crisis.

5.1. Better financing ability of SOE

5.1.1. Better access to external finance

To explore the external financing advantage of the SOEs, we examine two main financing sources, debt financing and equity financing. According to political pecking order hypothesis, SOEs are less financial constrained compared with POEs. SOEs function as an intermediary agent which could transfer the financing resource from banks to POEs through trade credit. In this section, we extend our analysis on the differential response between SOEs and non-SOEs to other firm level variables including short-term debt position (Shortdebt) measured by the short-term debt to sales ratio, long-term debt position (Longdebt) measure by the long-term debt to sales ratio, equity issue (Equity_1) measured by a dummy variable which equals 1 if a firm issues shares to its existing shareholders, private placement (Equity_2) measured by a dummy variable which equals 1 if a firm issues shares to a select group of investors. Short-term debt position and long-term debt position capture the debt financing channel while equity issue and private placement capture the equity financing channel.

Panel A in Table 5 presents the results for debt financing channel. The dependent variable in Columns (1)-(3) is short-term debt position. The coefficients of the interaction term Soe×Shock is positive and significant, suggesting that SOEs receive more short-term financing relative to non-SOEs in the front of COVID-19 shock. The dependent variable in columns (4)-(6) is long-term debt position. The coefficients of the interaction term Soe×Shock is insignificantly different from zero. Panel B in Table 5 presents the results for the differential response in equity financing to COVID-19 shock. Columns (1)-(3) investigate whether SOEs obtain more opportunity on right issue after the outbreak of COVID-19 while columns (4)-(6) examine whether SOEs receive more financing from private placement as a response to COVID-19 shock. The coefficients of the interaction term Soe×Shock in all specifications is insignificantly different from zero. These results show that SOEs obtain more short-term financing after the outbreak of COVID-19.

Table 5.

Financial ability of SOEs: External finance mechanism. This table examines the external finance mechanism. In Panel A, the dependent variable is the ratio of short term debt to assets (Shortdebt) in columns (1)-(3) and the ratio of long term debt to assets (Longdebt) in columns (4)-(6), respectively. In Panel B, the dependent variable is equity issue dummy (Equity_1) in columns (1)-(3) and private placement dummy (Equity_2) in columns (4)-(6), respectively. Equity_1 equals 1 if a firm issues shares to its existing shareholders, and Equity_2 equals 1 if a firm issues shares to a select group of investors. The main independent variable is Soe interacted with shock which is an indicator variable that equals 1 for SOEs after COVID-19 shock, and 0 otherwise. Tcreditdep denotes trade credit dependence of the downstream industry which captures the demand factors of the downstream firms. Other Controls is a vector of firm-level control variables which are lagged, including firm age (in log), total assets (in log), firm Liquidity (liquid assets to total assets ratio), Profitability (operating profit to sales ratio), Debtratio (long term debt to assets ratio), Tobinq and Topsale (market concentration). Standard errors (in parentheses) are robust to heteroscedasticity and clustered at firm level. *** significant at 1%, ** significant at 5%, * significant at 10%.

Dependent Variable:	Shortdebt				Longdebt
Panel A	(1)	(2)	(3)		(4)	(5)	(6)
Soe×Shock	0.2676***	0.2218**	0.2251**		0.0136	-0.0080	-0.0081
	(0.0981)	(0.0939)	(0.0959)		(0.0240)	(0.0263)	(0.0266)
Tcreditdep× Shock	No	No	Yes		No	No	Yes
Other Controls	No	Yes	Yes		No	Yes	Yes
Firm FE	Yes	Yes	Yes		Yes	Yes	Yes
Year-quarter FE	Yes	Yes	Yes		Yes	Yes	Yes
Observations	21581	18718	18718		20784	18718	18718
R-squared	0.0781	0.0731	0.0731		0.1077	0.1426	0.1426
Dependent Variable:	Equity_1				Equity_2
Panel B	(1)	(2)	(3)		(4)	(5)	(6)
Soe×Shock	-0.0048	-0.0067	-0.0068		0.0020	0.0023	0.0023
	(0.0061)	(0.0061)	(0.0061)		(0.0013)	(0.0014)	(0.0015)
Tcreditdep× Shock	No	No	Yes		No	No	Yes
Other Controls	No	Yes	Yes		No	Yes	Yes
Firm FE	Yes	Yes	Yes		Yes	Yes	Yes
Year-quarter FE	Yes	Yes	Yes		Yes	Yes	Yes
Observations	21581	18718	18718		21581	18718	18718
R-squared	0.1055	0.1109	0.1110		0.0817	0.0899	0.0900

5.1.2. Internal source of financing

Another aspect of financing advantage of SOEs is that they generate more internal sources of financing. Political connection is an important factor determining the access to government contracts and it has been found to facilitate firms’ access to more financial resources (Shleifer and Vishny, 1994, Faccio et al., 2006). Due to inborn political connection, SOEs obtain preferential treatment in government policies (Harrison et al., 2019). In addition, multitask of SOEs implies that government relies on SOEs to pursue some social objectives (Shleifer, 1998, Bai et al., 2006). Therefore, government could also provide better public services or favorable procurement of production input to connected firms (Ran et al., 2020). SOEs will benefit from cost-saving measures and then generate more internal sources of financing, which in turn, motivates SOEs to provide trade credit through the improvement of liquidity within firms.

To test this hypothesis, we follow Ran et al. (2020) and examine firms’ response of costs to COVID-19 shock. To this end, we construct a measure of costs defined as the ratio between cost of goods sold and total sales (Cost_1). In addition, we consider some other costs correlated with producing and selling goods, such as distribution costs, advertising costs and financing costs. Therefore, we construct another measure of costs defined as the ratio between cost of goods sold plus some indirect costs and total sales (Cost_2).

Panel A of Table 6 presents the results The dependence variable is Cost_1 in columns (1)-(3) and Cost_2 in columns (4)-(6), respectively. We are interested in the coefficient of the interaction term, Soe×Shock which captures the differential response on operating costs of SOEs to the outbreak of COVID-19 relative to non-SOEs. The coefficients of interaction term in columns (1)-(6) are all negative and statistically significant at the 5% level or better. The costs of SOEs response more intensively to COVID-19 shock relative to that of non-SOEs and the difference of costs between SOEs and non-SOEs falls significantly following the onset of COVID-19. These findings suggest that there could be some cost-saving measures specific to SOEs, such as better public services or favorable procurement of production input provided by the government.

Table 6.

Financial ability of SOEs: Internal financing mechanism. This table examines the internal financing ability of SOEs. In Panel A, the dependent variable is the ratio of cost of goods sold to sales (Cost_1) in columns (1) -(3) and the ratio of cost of goods sold plus indirect costs related to good sold to sales (Cost_2) in columns (4)-(6), respectively. In Panel B, the dependent variable is the ratio of the sum of cash stocks, inventories and accounts receivables to sales (Internal_fin) in columns (1)-(3). The main independent variable is Soe interacted with shock which is an indicator variable that equals 1 for SOEs after COVID-19 shock, and 0 otherwise. Tcreditdep denotes trade credit dependence of the downstream industry which captures the demand factors of the downstream firms. Other Controls is a vector of firm-level control variables which are lagged, including firm age (in log), total assets (in log), firm Liquidity (liquid assets to total assets ratio), Profitability (operating profit to sales ratio), Debtratio (long term debt to assets ratio), Tobinq and Topsale (market concentration). Standard errors (in parentheses) are robust to heteroscedasticity and clustered at firm level. *** significant at 1%, ** significant at 5%, * significant at 10%.

Dependent Variable:	Cost_1		Cost_2
Panel A	(1)	(2)	(3)	(4)	(5)	(6)
Soe×Shock	-0.0098***	-0.0078**	-0.0078**	-0.0387***	-0.0344***	-0.0347**
	(0.0038)	(0.0038)	(0.0038)	(0.0079)	(0.0083)	(0.0084)
Tcreditdep× Shock	No	No	Yes	No	No	Yes
Other Controls	No	Yes	Yes	No	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes
Year-quarter FE	Yes	Yes	Yes	Yes	Yes	Yes
Observations	21581	18718	18718	21581	18718	18718
R-squared	0.8115	0.8207	0.8207	0.4361	0.4542	0.4543
Dependent Variable:
Panel B	(1)	(3)
Soe×Shock	0.1472**	0.1674***
	(0.0630)	(0.0645)
Tcreditdep× Shock	No	Yes
Other Controls	No	Yes
Firm FE	Yes	Yes
Year-quarter FE	Yes	Yes
Observations	21546	18690
R-squared	0.0836	0.0944

In addition, we also examine the response of internal source of financing to COVID-19 shock. As discussed, cost-saving measures could improve the internal liquidity of SOEs and thus promote SOEs’ extension of trade credit to their clients relative to non-SOEs. To measure the key variable internal financing (Internal_fin), we follow the existing studies, such as Fazzari and Petersen (1993), and calculate the ratio of changes in a set of proxies (cash stocks, inventories and accounts receivables) to sales for internal financing sources for each firm. Panel B of Table 6 presents the results. The coefficients of interaction term in all columns are all significantly positive, suggesting that SOEs possess better availability of internal finance after COVID-19 shock relative to non-SOEs.

There is still one concern that non-SOEs’ decisions on credit extension may not depend on their financial capacity 1. We address this concern in two aspects. First, we examine the response of SOEs and that of non-SOEs to COVID-19 shock by regressing Credit_pro and Netcredit_pro on shock for SOEs and non-SOEs respectively. Table A7 in the online appendix reports the results. the dependent variable is Credit_pro and Netcredit_pro in columns (1)-(2) and in columns (3)-(4), respectively. The coefficients on Shock for SOE firms are positive and significant. However, the coefficients on Shock for non-SOE firms are insignificant. These findings suggest that SOEs provides more trade credit after the pandemic outbreak while non-SOEs’ trade credit provision does not response much to the shock.

Second, we examine whether domestic non-SOEs with abundant cash will extend more trade credit relative to domestic non-SOEs with less cash by constructing three cash abundant indicators (Cash_d) 2. The first is assigned according to the median of domestic non-SOEs’ cash to assets ratio. Cash_d equals 1 if the cash to assets ratio is above the median, and 0 otherwise. The second and third indicators are assigned according to the 70% quantile and 90% quantile of domestic non-SOEs’ cash to assets ratio respectively. We are interested in the estimated coefficients of the interaction term Cash_d×Shock. Table A9 presents the results. The estimated coefficients of Cash_d×Shock are only significantly positive in columns (3) and (6), suggesting that non-SOEs at top quantile of the cash abundance provide more trade credit. This finding is a little different from Garcia-Appendini and Montoriol-Garriga (2013). Only domestic non-SOEs with cash holdings ranking above 90% quantile show a differential response of trade credit provision to COVID-19 shock relative to other non-SOEs. Given that the average cash to assets ratio of SOEs is 0.15 and the 90% quantile of non-SOEs’ cash abundance is 0.25, this suggests that even if non-SOEs have the same cash abundance with a typical SOE, they will also hesitate to provide trade credit.

We posit that there might be several reasons. First, Chen et al. (2021) document that institutional factors could affect the differential trade credit provision between SOEs and non-SOEs. The less developed financial market and the poor creditor right protection discourage trade credit provision of non-SOEs as trade credit is a kind of loan. The difference of institutional environment between China and the country studied by Garcia-Appendini and Montoriol-Garriga (2013) may shape the differential response of non-SOEs to COVID-19 shock.

Second, financing advantage theory is used to explain our findings from the opportunity side. SOEs could handle more financial resources. However, SOEs with non-economic goals could tolerate more amounts of risk and state with highly diversified investment will generate a portfolio effect (Musacchio et al., 2015, Benito et al., 2016, Rygh, 2018, Grøgaard et al., 2019). Overall, SOEs have advantages in both the opportunity side and the motivation side over non-SOEs.

Third, some studies also document evidences that some financing advantages, such as subsidies are specific to SOEs and also leads to soft budget constraints (Krueger, 1990, Claro, 2006, Eckaus, 2006). Therefore, SOEs are more motivated to provide trade credit during the crisis. We examine the differential response of subsidy between SOEs and non-SOEs during the period of COVID-19 shock following equation (1). After the outbreak of COVID-19, Chinese government reacts rapidly and brings “sizeable and targeted” economic aid to business. Political connection is an important factor determining the flow direction of economic aid and it has been found to facilitate firms’ access to more financial resources (Shleifer and Vishny, 1994, Faccio et al., 2006). Due to inborn political connection, SOEs obtain preferential treatment in government subsidies. In addition, multitask of SOEs implies that government relies on SOEs to pursue some social objectives (Shleifer, 1998, Bai et al., 2006). As a result, government subsidies are primarily provided for SOEs. In summary, compared with non-SOEs, SOEs are favored not only by credit policies, but also by government subsidies (Harrison et al., 2019).

To examine preferential government subsidies, we now compare what happens to government subsidies received by SOEs versus that received by non-SOEs before and after the outbreak of COVID-19. Specifically, government subsidies in our context include subsidies supporting firm’s operation, job stability, loan interest subsidies, tax reduction and COVID-19 subsidies. We examine both the extensive margin and intensive margin effects. First, we construct a dummy (Subsidy) which equal 1 if a firm receives government subsidies and otherwise 0. Second, we calculate government subsidies to sales ratio (Subsidy_r). There are 57.36% firm-year observations of whole sample that receives government subsidies. We address the sample selection issue by estimating an inverse Mill’s ratio which shows the probability that a firm receives government subsidies. We adopt a Probit model first and regress SOEs dummy (Soe) on a set of controls. Next, we calculate the inversed Mill’s ratio bases on the results from the first step regression. Finally, we estimate equation (1) for government subsidies and control for the inversed Mill’s ratio to correct for sample selection bias. Panel A of Table 7 presents the results for the channel of preferential government subsidies. In columns (1)-(4), the dependent variable is Subsidy, a dummy indicating whether a firm receive the government subsidies or not whereas the dependence variable in column (5) is Subsidy_r. In column (5), the inversed Mill’s ratio is controlled to correct the sample selection bias. In columns (1)-(2), we adopt a linear probability model whereas in columns (3)-(4), we adopt a random-effects Probit regression model. The coefficients of interaction term in columns (1)-(4) are all significantly positive and thus reveal an extensive margin effect that SOEs are more likely to receive government subsidies during the crisis. In column (5), the coefficient of interaction term is positive and statistically significant at the 1% level among firms that have received the government subsidies, which suggests an intensive margin effect. The SOEs receive more government subsidies when comparing with non-SOEs during the crisis.

Table 7.

Preferential government subsidies . This table examines advantage of SOEs on obtaining government subsidies. The dependent variable is subsidy dummy (Subsidy) in columns (1)-(4) and subsidy to sales ratio (Subsidy_r) in column (5), respectively. Subsidy dummy (Subsidy) equals 1 if the firm has been granted government subsidies and 0 otherwise. In Panel B, we exclude the COVID-19 subsidies from the government subsidies for a robustness check. The main independent variable is Soe interacted with shock which is an indicator variable that equals 1 for SOEs after COVID-19 shock, and 0 otherwise. Tcreditdep denotes trade credit dependence of the downstream industry which captures the demand factors of the downstream firms. Other Controls is a vector of firm-level control variables which are lagged, including firm age (in log), total assets (in log), firm Liquidity (liquid assets to total assets ratio), Profitability (operating profit to sales ratio), Debtratio (long term debt to assets ratio), Tobinq and Topsale (market concentration). Standard errors (in parentheses) are robust to heteroscedasticity and clustered at firm level. *** significant at 1%, ** significant at 5%, * significant at 10%.

Dependent Variable:	Subsidy						Subsidy_r
Panel A	OLS			Probit
	(1)	(2)		(3)	(4)		(5)
Soe×Shock	0.0553***	0.0712***		0.2955***	0.3635***		0.0869***
	(0.0167)	(0.0200)		(0.0702)	(0.0856)		(0.0327)
Other Controls	No	Yes		No	Yes		Yes
Inverse Mills Ratio	No	No		No	No		Yes
Firm FE	Yes	Yes		No	No		Yes
Year FE	Yes	Yes		Yes	Yes		Yes
Observations	14057	8539		14320	9291		4676
R-squared	0.6166	0.6347					0.5033
Wald Test				361.06	234.39
Dependent Variable:	Subsidy						Subsidy_r
Panel B	OLS			Probit
	(1)	(2)		(3)	(4)		(5)
Soe×Shock	0.0478***	0.0687***		0.2646***	0.3549***		0.0871***
	(0.0166)	(0.0198)		(0.0701)	(0.0855)		(0.0348)
Other Controls	No	Yes		No	Yes		Yes
Inverse Mills’ Ratio	No	No		No	No		Yes
Firm FE	Yes	Yes		No	No		Yes
Year FE	Yes	Yes		Yes	Yes		Yes
Observations	14057	8539		14320	9291		4649
R-squared	0.6188	0.6378					0.5024
Wald Test				390.16	255.64

For robustness check, we exclude the COVID-19 subsidies from the government subsidies as COVID-19 subsidies only appear after the outbreak of COVID-19. Panel B of Table 7 presents the results which are similar with that in Panel A. The coefficients of interaction term, Soe×Shock, in all specifications are significantly positive, suggesting that there are both extensive margin effect and intensive margin effect. This is consistent with our conjecture. SOEs not only have advantages in credit allocation, but also gain preferential government subsidies. In summary, the subsidy presents a similar pattern of the trade credit provision between SOEs and non-SOEs during the period of COVID-19 shock. As financing advantage includes the support of government through the channel of fiscal funding, this finding further supports that SOEs have financing advantage over non-SOEs. SOEs play a similar role as government banks during the crisis. In time of the crisis, banks always hesitate to lend and preserve liquidity because of the increase of uncertainty. However, government banks provide more liquidity during the crisis (Coleman and Feler, 2015, Chen et al., 2016). The government funding specific to government banks leads to the differential lending behavior between government banks and other banks (Ivashina and Scharfstein, 2010, Coleman and Feler, 2015). As presented in Table 7, SOEs have advantage over non-SOEs in terms of the government subsidies that is also one aspect of financial advantage of SOEs (Chen et al., 2021). The state-owned entities including both banks and enterprises will leverage their financial advantage to extend the liquidity during the crisis.

5.2. Multitask of SOEs

The above analysis shows that SOEs provide more trade credit after the COVID-19 shock. However, there is a difference between granting more trade credit due to better access to financing and granting more trade credit due to promoting economic stability. Better access to financial credit is one driving force in motivating the trade credit provision (Garcia-Appendini and Montoriol-Garriga, 2013). However, the multitask theory implies that SOEs may take measures including more trade credit provision during the crisis to maintain economic stability (Chen et al., 2021).

In this section, we conduct several tests to investigate the multitask mechanism. First, we further control for more indicators of financial constraints in the regression to check that given the same financing advantage, whether SOEs provide more trade credit during the crisis. Specifically, we use three indicators of financial constraints including KZ index (Kaplan and Zingales, 1997), SA index (Hadlock and Pierce, 2010) and WW index (Whited and Wu, 2006), the data of which are collected from CSMAR. In Panel A of Table 8 , we find that SOEs provide more trade credit during the crisis relative to non-SOEs after controlling for the indicators of financial constraints (Fincons).

Table 8.

Ownership and trade credit provision: Controlling for financing advantage . This table reports the results of regressions in which the dependent variable is trade credit provision (Credit_pro) in columns (1)-(3) and net trade credit provision (Netcredit_pro) in columns (4)-(6), respectively. The main independent variable is Soe interacted with shock which is an indicator variable that equals 1 for SOEs after COVID-19 shock, and 0 otherwise. Fincons is an indicator for financial constraint and Fincons_b is an indicator for financial constraint in 2019. Columns (1) and (4) refer to KZ index. Columns (2) and (5) refer to SA index while columns (3) and (6) refer to WW index. Tcreditdep denotes trade credit dependence of the downstream industry which captures the demand factors of the downstream firms. Other Controls is a vector of firm-level control variables which are lagged, including firm age (in log), total assets (in log), firm Liquidity (liquid assets to total assets ratio), Profitability (operating profit to sales ratio), Debtratio (long term debt to assets ratio), Tobinq and Topsale (market concentration). Standard errors (in parentheses) are robust to heteroscedasticity and clustered at firm level. *** significant at 1%, ** significant at 5%, * significant at 10%.

Dependent Variable:	Credit_pro				Netcredit_pro
Panel A	KZ	SA	WW		KZ	SA	WW
	(1)	(2)	(3)		(4)	(5)	(6)
Soe×Shock	0.0972*	0.1318**	0.0886***		0.1111**	0.1151**	0.0723**
	(0.0573)	(0.0538)	(0.0301)		(0.0472)	(0.0537)	(0.0302)
Fincons	-0.0096	0.2163	-0.2604		-0.0029	0.2533	-0.0300
	(0.0092)	(0.6826)	(0.2836)		(0.0094)	(0.6685)	(0.2415)
Tcreditdep× Shock	Yes	Yes	Yes		Yes	Yes	Yes
Other Controls	Yes	Yes	Yes		Yes	Yes	Yes
Firm FE	Yes	Yes	Yes		Yes	Yes	Yes
Year-quarter FE	Yes	Yes	Yes		Yes	Yes	Yes
Observations	17818	18711	18125		17818	18711	18125
R-squared	0.0777	0.0786	0.0926		0.0777	0.0812	0.0955
Dependent Variable:	Credit_pro				Netcredit_pro
Panel B	KZ	SA	WW		KZ	SA	WW
	(1)	(2)	(3)		(1)	(2)	(3)
Soe×Shock	0.1041**	0.1330**	0.1028***		0.0905*	0.1147**	0.0857**
	(0.0526)	(0.0536)	(0.0378)		(0.0532)	(0.0540)	(0.0379)
Fincons_b×Shock	0.0090	-0.0515	-0.9310		0.0178	-0.0914	-0.9383
	(0.0287)	(0.0731)	(0.5982)		(0.0283)	(0.0736)	(0.5987)
Tcreditdep× Shock	Yes	Yes	Yes		Yes	Yes	Yes
Other Controls	Yes	Yes	Yes		Yes	Yes	Yes
Firm FE	Yes	Yes	Yes		Yes	Yes	Yes
Year-quarter FE	Yes	Yes	Yes		Yes	Yes	Yes
Observations	17928	18718	17582		17928	18718	17582
R-squared	0.0774	0.0787	0.0786		0.0774	0.0813	0.0809

Second, we interact the three indicators on financial constraints in 2019 (Fincons_b) with the shock dummy. This could check whether SOEs provide more trade credit after considering the differential responses among firms with different degree of financial constraints to the shock. In panel B of Table 8, we also find that SOEs provide more trade credit during the crisis relative to non-SOEs after considering the differential responses to the crisis.

Third, we extend the analysis to SOEs’ sample. Because SOEs are always financial unconstrained, we expect that multitask mechanism rather than financing advantage mechanism motivates the trade credit provision of SOEs during the crisis if we focus on SOEs’ sample. We construct three cash abundant indicators (Cash_ds). The first is assigned according to the median of SOEs’ cash to assets ratio. Cash_ds equals 1 if the cash to assets ratio is above the median, and 0 otherwise. The second and the third indicators are assigned according to the 70% quantile and 90% quantile of SOEs’ cash to assets ratio respectively. This is similar with the analysis on non-SOEs. The Panel A of Table 9 reports the results. The coefficients of Cash_ds×Shock are all insignificant, suggesting that the financing advantage mechanism alone could not explain the trade credit provision of SOEs.

Table 9.

Ownership and trade credit provision: Considering less cash-abundant SOEs. This table reports the results of regressions in which the dependent variable is trade credit provision (Credit_pro) in columns (1)-(3) and net trade credit provision (Netcredit_pro) in columns (4)-(6), respectively. The sample is restricted to SOEs in Panel A. The main independent variable is Cash_ds interacted with shock which is an indicator variable that equals 1 for cash-abundant SOEs after COVID-19 shock, and 0 otherwise. p50, p70 and p90 indicate the median, 70% quantile, 90% quantile threshold for the assignment of Cash_ds respectively. The sample includes both non-SOEs and less cash-abundant SOEs in Panel B. The main independent variable is Soe interacted with shock which is an indicator variable that equals 1 for SOEs after COVID-19 shock, and 0 otherwise. p50, p30 and p10 indicate the median, 30% quantile, 10% quantile threshold for the classification of less cash-abundant SOEs, respectively. Tcreditdep denotes trade credit dependence of the downstream industry which captures the demand factors of the downstream firms. Other Controls is a vector of firm-level control variables which are lagged, including firm age (in log), total assets (in log), firm Liquidity (liquid assets to total assets ratio), Profitability (operating profit to sales ratio), Debtratio (long term debt to assets ratio), Tobinq and Topsale (market concentration). Standard errors (in parentheses) are robust to heteroscedasticity and clustered at firm level. *** significant at 1%, ** significant at 5%, * significant at 10%.

Dependent Variable:	Credit_pro				Netcredit_pro
Panel A	p50	p70	p90		p50	p70	p90
	(1)	(2)	(3)		(4)	(5)	(6)
Cash_ds×Shock	0.0438	0.0485	-0.0069		0.0176	0.0228	-0.0214
	(0.0332)	(0.0301)	(0.0236)		(0.0249)	(0.0293)	(0.0254)
Tcreditdep× Shock	Yes	Yes	Yes		Yes	Yes	Yes
Other Controls	Yes	Yes	Yes		Yes	Yes	Yes
Firm FE	Yes	Yes	Yes		Yes	Yes	Yes
Year-quarter FE	Yes	Yes	Yes		Yes	Yes	Yes
Observations	5265	5265	5265		5265	5265	5265
R-squared	0.1166	0.1166	0.1164		0.1051	0.1051	0.1051
Dependent Variable:	Credit_pro				Netcredit_pro
Panel B	p50	p30	p10		p50	p30	p10
	(1)	(2)	(3)		(4)	(5)	(6)
Soe×Shock	0.1135**	0.1118**	0.1252***		0.1148**	0.1147**	0.1435**
	(0.0517)	(0.0508)	(0.0474)		(0.0547)	(0.0545)	(0.0609)
Tcreditdep× Shock	Yes	Yes	Yes		Yes	Yes	Yes
Other Controls	Yes	Yes	Yes		Yes	Yes	Yes
Firm FE	Yes	Yes	Yes		Yes	Yes	Yes
Year-quarter FE	Yes	Yes	Yes		Yes	Yes	Yes
Observations	15862	15446	14405		15862	15446	14405
R-squared	0.0787	0.0787	0.0788		0.0813	0.0814	0.0814

Forth, we construct a sample including both non-SOEs and less cash-abundant SOEs, and perform the regression following equation (1). We construct three less cash-abundant SOEs’ sample. A SOE is classified as a less cash-abundant SOE if the cash to assets ratio is below the median. Columns (1) and (4) in Panel B of Table 9 present the results. Similarly, a SOE is classified as a less cash-abundant SOE if the cash to assets ratio is below the 30% quantile or 10% quantile. Columns (2) and (5) present the results for the sample classified according to the threshold of 30% quantile while columns (3) and (6) present the results for the sample classified according to the threshold of 10% quantile. In the Panel B of Table 9, we find that less cash-abundant SOEs are also more willing to extend trade credit relative to non-SOEs during the crisis. These results prove that many less cash-abundant SOEs show a differential response in trade credit provision compared to their non-SOE counterparts, and furthermore support the multitasks mechanism. The underlying driving force is that SOEs’ objective function concerns social optimality more.

Finally, we check the trade credit response to government subsidy for SOEs and non-SOEs separately. As shown in Table 7, SOEs receive more subsidies relative to non-SOEs. We take the logarithm of accounts receivable and government subsidy in the empirical model respectively, and thus examine the subsidy elasticity of trade credit. The logarithm of accounts receivable (Logcredit_pro) is the dependent variable while the logarithm of subsidy (Logsubsidy) is the independent variable. We also control for the firm level and industry level characteristics as shown in the baseline analysis. Specifically, we separate the sample into SOEs and non-SOEs, and perform the regression for each year from 2017 to 2020 in each subsample, respectively.

The top panel of Figure 1 shows the estimated coefficients of Logsubsidy for SOEs and non-SOEs, respectively. The estimated coefficients of Logsubsidy for SOEs is large than that for non-SOEs, indicating that the response of trade credit to government subsidy is more pronounced for SOEs. This difference in the response between SOEs and non-SOEs become greater during the crisis. Furthermore, we conduct analysis based on both SOEs and non-SOEs. The key variable is the interaction term of Logsubsidy and Soe. We also perform the regression for each year from 2017 to 2020. The estimated coefficients of the interaction term reveal the differential response of trade credit to government subsidy between SOEs and non-SOEs. The bottom panel of Figure 1 shows the estimated coefficients of the interaction term. As shown, SOEs are more responsive to government subsidies than non-SOEs, especially during the crisis. All these evidences support the importance of multitask mechanism.

To further support the multitask mechanism, we use the variations of the degree in shock damaging among regions and industries. If trade credit provision by SOEs is used to achieve economic stability, we should expect a greater amount of trade credit provision of SOEs in provinces with lower GDP growth rate which could lead to higher unemployment. We calculate the gap of GDP growth rate between 2019 and 2020 at first and then construct a new variable Gap_gdp_d which equals 1 if a firm located in the region below the median value of GDP growth gap, and 0 otherwise. We interact Gap_gdp_d with Soe×Shock and also other controls. We regress the trade credit provision on the triple interaction term following equation (1). Columns (1) and (3) in Table A10 report the results. Mech represents Gap_gdp_d in columns (1) and (3). The estimated coefficients of Soe×Shock×Gap_gdp are positive and significant, suggesting that SOEs are more willing to provide trade credit during the COVID-19 shock relative to non-SOEs in the regions with lower GDP growth rate. The results are consistent with the views that SOEs provide more trade credit with aiming to maintain economic stability during a crisis.

Besides, the effect of COVID-19 is more damaging for some industries than for others as revealed by Bae et al. (2021). If SOEs grant more trade credit as a stabilizing influence after COVID-19 shock, we should expect a greater amount of trade credit of SOEs provided to the firms in industries that are more adversely affected by COVID-19 shock. We use the cumulative abnormal return (CAR) to reflect the seriousness during the COVID-19 shock. The calculation of the CAR in this paper is based on the daily rate of return on stock and the daily rate of return on the market, the data of which comes from CSMAR. Specifically, we define the start of pandemic crisis as 18^th February, 2020 following Bae et al. (2021) and estimate the market model using 150 days of returns. The estimation window is from 180 days before the start of pandemic crisis to 30 days before the start. We consider (-24, 24) event windows and finally obtain the CAR of each firm. The industry level CAR (CAR_ind) is the average of firms’ CAR in this industry. The industry that suffers serious damage will have poor performance in stock market. Therefore, we construct a new variable CAR_d which equals 1 if a firm belongs to an industry whose CAR is below the median, and otherwise 0. We interact CAR_d with Soe×Shock and also other controls. Columns (2) and (4) in Table A10 report the results. Mech represents CAR_d in columns (2) and (4). The variable of interest is the triple interaction term, Soe×Shock ×CAR_d, which captures the extent to which SOEs in an industry suffering more loss facilitates trade credit extension during a crisis. The estimated coefficients of triple interaction term in columns (2) and (4) are significantly positive and the interaction term Soe×Shock in columns (2) and (4) are positive but insignificant. This suggests that the trade credit provision of SOEs is mainly driven by SOEs in industries which suffer more serious damage after COVID-19 shock.

6. Extensions

Information advantage is the principal element for the liquidity provision of financial intermediary, especially bank (Diamond, 1984, Fama, 1985). Uncertainty often rises significantly over crisis periods. Based on better financing ability and easier access to public sectors, SOEs may have better monitoring capabilities that facilitate trade credit provision like financial intermediaries due to the involvement in economy and supply chain linkages with the downstream firms, which in turn, helps stabilize the economy during crisis periods.

6.1. Heterogeneity among industries with different external financial dependence

The cost of external financing would be higher in industries that are dependent more on external finance since these industries are more financially constrained. If SOEs helps stabilize the economy during the pandemic, they are expected to extend more trade credit to downstream firms in industries with higher external financial dependence. To examine this conjecture, we first exploit the variation in external finance dependence based on Manova et al. (2015) which is constructed following the approach of Rajan and Zingales (1998). Rajan and Zingale (1998) construct an external financial dependence index using data on the U.S. publicly traded firms in each manufacturing industry at the two-digit level. The proportion of capital expenditures not financed by internal operating cash flows then serves as an index measure of external financial dependence. The basic logic for the rationality of external financial dependence index constructed is that some industries require external finance for technological reasons. As publicly traded firms could exclude the influence of supply-side constraints, this index well reflects the characteristics of industries and is widely applied (Fisman and Love, 2007; Lin and Ye, 2018).

We construct a new dummy variable Exf_d which equals 1 if a firm belongs to an industry whose external financial dependence is above the median, and otherwise 0. We now focus on the triple interaction between firm ownership and the COVID-19 shock and the external financial dependence indicator, Soe×Shock×Exf_d, which captures the extent to which SOEs in an industry relying more on external finance facilitates trade credit extension during a crisis. Table 10 presents the results. The dependence variable is Credit_pro and Netcredit_pro in columns (1) and (2), respectively. Estimation results show that the coefficient on the triple interaction term is positive and statistically significant at the 10% level, indicating that SOEs extend more credit to downstream firms, especially to the downstream firms in industries which depend more on external finance.

Table 10.

Ownership and trade credit provision: External financial dependence. This table reports the results of regressions in which the dependent variable is trade credit provision (Credit_pro) in column (1) and net trade credit provision (Netcredit_pro) in column (2), respectively. The main independent variables are the triple interaction between SOE dummy and the pandemic indicator and Exf_d. Exf_d equals 1 if a firm belongs to industry whose external financial dependence is above the median, and 0 otherwise. Tcreditdep denotes trade credit dependence of the downstream industry which captures the demand factors of the downstream firms. Other Controls is a vector of firm-level control variables which are lagged, including firm age (in log), total assets (in log), firm Liquidity (liquid assets to total assets ratio), Profitability (operating profit to sales ratio), Debtratio (long term debt to assets ratio), Tobinq, Topsale (market concentration) and Tcreditdep×shock. Standard errors (in parentheses) are robust to heteroscedasticity and clustered at firm level. *** significant at 1%, ** significant at 5%, * significant at 10%.

Dependent Variable:	Credit_pro		Netcredit_pro
	(1)		(2)
Soe×Shock×Exf_d	0.1333*		0.1516*
	(0.0769)		(0.0792)
Soe×Shock	0.0473**		0.0176
	(0.0217)		(0.0266)
Shock×Exf_d	0.0471		0.0767
	(0.0569)		(0.0599)
Other Controls	Yes		Yes
Other Controls×Exf_d	Yes		Yes
Firm FE	Yes		Yes
Year-quarter FE	Yes		Yes
Observations	18718		18718
R-squared	0.0788		0.0815

6.2. Heterogeneity among regions with different government involvement

External financial would be more expensive in regions where government involvement is stronger. Government usually relies on SOEs to stabilize the economy during crisis periods. As SOEs have easier access to government information and data (Capobianco and Christiansen, 2011), SOEs will be more informative on the market and production network. Based on the information advantage, SOEs might have better screening or monitoring capabilities that facilitates credit supply. We expect that the effect of COVID-19 shock on SOEs’ trade credit provision will be more pronounced in regions with stronger government involvement.

To this end, we employ the government involvement indicator (Neri_gov) from the dataset of the marketization index for China’s provinces 3. The marketization index is constructed from the National Economic Research Institute (NERI) in China, a widely used measure of the progress of marketization in China (Jia et al., 2019). The original data for the marketization index is from the National Bureau of Statistics combined with the Chinese Enterprises Survey conducted by the Development Research Center of the State Council. Higher Neri_gov means a lower degree of government involvement. We construct a new dummy variable Neri_gov_d which equals 1 if a firm belongs to a region whose Neri_gov is below the median, and 0 otherwise.

Table 11 presents the results. We now focus on the triple interaction between Neri_gov_d and Soe×shock. In column (1), the dependent variable is Credit_pro, whereas the dependent in column (2) is the Netcredit_pro. The variable of interest is the triple interaction term, Soe×Shock×Neri_gov_d which captures the extent to which SOEs in the region with higher government involvement facilitates trade credit extension during a crisis. The results are consistent with the views that SOEs located in the region with higher government involvement provide more trade credit during a crisis. Specifically, column (1) shows that the coefficient on the triple interaction term is significantly positive and this positive association holds when replacing dependent variable Credit_pro with Netcredit_pro, suggesting that SOEs provide more trade credit during the pandemic in the region with higher government involvement.

Table 11.

Ownership and trade credit provision: Heterogeneity by government involvement. This table examines the heterogeneity among provinces with different degrees of government involvement. The dependent variable is trade credit provision (Credit_pro) in column (1) and net trade credit provision (Netcredit_pro) in column (2), respectively. The main independent variable is the triple interaction between firm ownership dummy (Soe) with the pandemic indicator (Shock) and the government involvement indicator (Neri_gov_d). Neri_gov_d equals 1 if a firm is located in a region where the government involvement is above the median, and 0 otherwise. Tcreditdep denotes trade credit dependence of the downstream industry which captures the demand factors of the downstream firms. Other Controls is a vector of firm-level control variables which are lagged, including firm age (in log), total assets (in log), firm Liquidity (liquid assets to total assets ratio), Profitability (operating profit to sales ratio), Debtratio (long term debt to assets ratio), Tobinq, Topsale (market concentration) and Tcreditdep×shock. Standard errors (in parentheses) are robust to heteroscedasticity and clustered at firm level. *** significant at 1%, ** significant at 5%, * significant at 10%.

Dependent Variable:	Credit_pro	Netcredit_pro
	(1)	(2)
Soe×Shock×Neri_gov_d	0.1356*	0.1122*
	(0.0759)	(0.0642)
Soe×Shock	0.0283	0.0329
	(0.0297)	(0.0246)
Shock×Neri_gov_d	0.1038	0.1621*
	(0.0969)	(0.0968)
Other Controls	Yes	Yes
Other Controls×Neri_gov_d	Yes	Yes
Firm FE	Yes	Yes
Year-quarter FE	Yes	Yes
Observations	18718	18718
R-squared	0.0787	0.0813

6.3. Local SOEs vs. Central SOEs

The above analyses reveal that the trade credit provision of SOEs during the crisis responses more pronouncedly in industries with higher dependence on external finance and in provinces with higher government involvement. Meanwhile, the cost of external financing is higher in these industries and provinces. Therefore, SOEs might have greater comparative advantage in monitoring and screening due to the involvement in local economy and the supply chain linkage with the downstream firms. We further illustrate this point by examining the differential response between local government-owned SOEs and central government-owned SOEs to rationalize this hypothesis.

Specifically, we use the information on equity ownership for each firm, and differentiate the local government-owned SOEs from central government-owned SOEs. Local information is key to understanding the efficiency of alternative economic systems and whether production should be centralized or decentralized (Hayek, 1945). Local government has inborn advantage on acquiring local information over central government and local information is more important for the operation of local SOEs (Qian et al., 2006, Huang et al., 2017). Additionally, local government has more information and data, such as official data survey on firms’ operation, data on firms’ tax payment and electricity consumption and so forth. Local SOEs have easier access to government information and data which are unavailable to non-SOEs to a great extent (Capobianco and Christiansen, 2011). Therefore, the information advantage on local economy could be employed by local SOEs during the crisis periods.

We use the information on the hierarchy of SOEs to verify our conjectures. We separate the variable Soe into two variables Soe_Central and Soe_Local according to the hierarchy of SOEs. Soe_Central equals 1 if a firm is central government-owned enterprise, and 0 otherwise. Soe_Local equals 1 if a firm is local government-owned enterprise, and 0 otherwise. Panel A of Table 12 presents the results. The coefficients on Soe_Central×Shock and Soe_Central×Shock are all significantly positive, suggesting that both central government-owned enterprises and local government-owned enterprises are more willing to extend trade credit relative to other firms. Furthermore, the coefficient on Soe_Local×Shock is larger than that on Soe_Central×Shock in all specifications, implying a greater effect of COVID-19 shock on trade credit extension of local government-owned enterprises.

Table 12.

Ownership and trade credit provision: Heterogeneity by ownership hierarchy. This table examines the heterogeneous response between SOEs owned by central government and SOEs owned by local government. The dependent variable is trade credit provision (Credit_pro) in columns (1)-(3) of Panel A and in columns (1)-(2) of Panel B, and is net trade credit provision (Netcredit_pro) in columns (4)-(6) of Panel A and in columns (3)-(4) of Panel B, respectively. The main independent variables are the interaction of the pandemic indicator (Shock) with central SOE dummy (Soe_Central) and with local SOE dummy (Soe_Local). Soe_dep is a measure of regional-level SOE dependence calculated as the ratio of sales of state-holding industrial enterprises to total sales of industrial enterprises in 2017. Tcreditdep denotes trade credit dependence of the downstream industry which captures the demand factors of the downstream firms. Other Controls is a vector of firm-level control variables which are lagged, including firm age (in log), total assets (in log), firm Liquidity (liquid assets to total assets ratio), Profitability (operating profit to sales ratio), Debtratio (long term debt to assets ratio), Tobinq and Topsale (market concentration). Standard errors (in parentheses) are robust to heteroscedasticity and clustered at firm level. *** significant at 1%, ** significant at 5%, * significant at 10%.

Dependent Variable:	Credit_pro		Netcredit_pro
Panel A	(1)	(2)	(3)		(4)	(5)	(6)
Soe_Central×Shock	0.0913***	0.0868**	0.0872**		0.0873**	0.0696*	0.0701*
	(0.0350)	(0.0383)	(0.0382)		(0.0401)	(0.0386)	(0.0386)
Soe_Local×Shock	0.1844**	0.1156**	0.1101**		0.1271***	0.1163**	0.1111**
	(0.0930)	(0.0458)	(0.0438)		(0.0456)	(0.0488)	(0.0471)
Tcreditdep× Shock	No	No	Yes		No	No	Yes
Other Controls	No	Yes	Yes		No	Yes	Yes
Firm FE	Yes	Yes	Yes		Yes	Yes	Yes
Year-quarter FE	Yes	Yes	Yes		Yes	Yes	Yes
Observations	21581	18718	18718		21581	18718	18718
R-squared	0.0701	0.0787	0.0787		0.0701	0.0787	0.0787
Dependent Variable:	Credit_pro		Netcredit_pro
Panel B	(1)	(2)		(3)	(4)
Soe_Central×Shock	0.0871**	0.0949		0.0677*	0.0764
	(0.0345)	(0.0573)		(0.0349)	(0.0563)
Soe_Local×Shock	0.1085**	0.0942*		0.1133**	0.1034*
	(0.0430)	(0.0511)		(0.0456)	(0.0534)
Soe_dep×Shock	0.2404			0.2519
	(0.2277)			(0.2241)
Tcreditdep× Shock	Yes	Yes		Yes	Yes
Other Controls	Yes	Yes		Yes	Yes
Firm FE	Yes	Yes		Yes	Yes
Year-quarter FE	Yes	No		Yes	No
Region×Year-quarter FE	No	Yes		No	Yes
Observations	17238	17238		17238	17238
R-squared	0.0786	0.0995		0.0812	0.1019

It is possible, however, local government-owned SOEs provide more trade credit could be caused by the fact that some local areas suffer more serious shock. As local government cares more on local economy, local government-owned SOEs will be used to extend trade credit to downstream firms. We further separate the information advantage effect by interacting SOEs dependence indicator at regional level with COVID-19 shock dummy. SOEs’ activities at regional level could affect local government’s motivation to use SOEs to stabilize local economy. Because the latest data on industrial enterprises sales at regional level is 2017, we calculate the state-holding industrial enterprises sales to total industrial enterprises sales ratio in 2017 (Soe_dep). The data comes from National Bureau of Statistics of China.

We also employ a more stringent specification by interacting region fixed effects with year-quarter fixed effects to control the time-varying factors at regional level. By controlling the time-varying factors at regional level, we could check that given the degree of the damage by COVID-19 shock, whether the local government-owned SOEs provide more trade credit relative to central government-owned SOEs. If the local government-owned SOEs still provide more trade credit, this suggests that the information advantage facilitates local SOEs to extend more trade credit.

Panel B of Table 12 presents the results. Columns (1) and (3) add the interaction term Soe_dep×Shock and columns (2) and (4) add the interaction terms of region fixed effects and year-quarter fixed effects. The coefficients on Soe_local×Shock are all significantly positive and the coefficients on Soe_central×Shock are positive and insignificant in columns (2) and (4), suggesting that only local government-owned enterprises are more willing to extend trade credit relative to downstream firms in more stringent specification. Furthermore, the coefficient on Soe_Local×Shock is significantly larger than that on Soe_Central×Shock in columns (1) and (3), implying a greater effect of COVID-19 shock on trade credit extension of local government-owned enterprises.

In addition, we also present evidence on the effect of state-owned shares on trade credit provision during the crisis. The monitoring of principal could affect the agency risk (Jia et al., 2020). Following the literature on blockholder, the larger blockholder with further interests at stake will exert more efforts in monitoring the agency (Dharwadkar et al., 2008). Hence, we check the effects of state-owned shares on trade credit provision. We use the ownership information on the top ten shareholders and calculate the share of the state-own shareholders (Soe_share) in 2019. We replace the Soe×Shock with Soe_share×Shock in equation (1). The mean of Soe_share of SOEs is 0.4438 while that of non-SOEs is 0.0340. Panel A of Table 13 presents the results. The coefficients on Soe_share×Shock in all specifications are positive and significant, suggesting that a firm with higher state-owned share is likely to provide more trade credit after the outbreak of COVID-19. We further construct a new dummy (Soe_n) that equals 1 if a firm is a SOE and the state-owned share is larger than 50%, and 0 otherwise. Panel B of Table 13 presents the results for Soe_n×Shock. The coefficients on Soe_n×Shock in columns (2)-(6) are positive and significant. All these results indicate that the information advantage could be a potential factor determining SOEs’ trade credit provision during the crisis.

Table 13.

Government share and trade credit provision. This table use government share as an alternative measure for firm ownership. The dependent variable is trade credit provision (Credit_pro) in columns (1)-(3) and net trade credit provision (Netcredit_pro) in columns (4)-(6), respectively. The main independent variables are the interaction of the pandemic indicator (Shock) with proxies for government share (Soe_share and Soe_n). Soe_share is measure as the share of the state-own shareholders in top ten shareholders. Soe_n equals 1 if a firm is a SOE and the state-owned share is larger than 50%, and 0 otherwise. Tcreditdep denotes trade credit dependence of the downstream industry which captures the demand factors of the downstream firms. Other Controls is a vector of firm-level control variables which are lagged, including firm age (in log), total assets (in log), firm Liquidity (liquid assets to total assets ratio), Profitability (operating profit to sales ratio), Debtratio (long term debt to assets ratio), Tobinq and Topsale (market concentration). Standard errors (in parentheses) are robust to heteroscedasticity and clustered at firm level. *** significant at 1%, ** significant at 5%, * significant at 10%.

Dependent Variable:	Credit_pro				Netcredit_pro
Panel A	(1)	(2)	(3)		(4)	(5)	(6)
Soe_share×Shock	0.2117**	0.2683**	0.2636**		0.2208*	0.2361*	0.2319*
	(0.1017)	(0.1242)	(0.1219)		(0.1155)	(0.1265)	(0.1242)
Tcreditdep× Shock	No	No	Yes		No	No	Yes
Other Controls	No	Yes	Yes		No	Yes	Yes
Firm FE	Yes	Yes	Yes		Yes	Yes	Yes
Year-quarter FE	Yes	Yes	Yes		Yes	Yes	Yes
Observations	21581	18718	18718		21581	18718	18718
R-squared	0.0701	0.0787	0.0787		0.0714	0.0813	0.0813
Dependent Variable:	Credit_pro				Netcredit_pro
Panel B	(1)	(2)	(3)		(4)	(5)	(6)
Soe_n×Shock	0.0576	0.0719**	0.0717**		0.0730*	0.0677*	0.0675*
	(0.0364)	(0.0369)	(0.0366)		(0.0422)	(0.0403)	(0.0401)
Tcreditdep× Shock	No	No	Yes		No	No	Yes
Other Controls	No	Yes	Yes		No	Yes	Yes
Firm FE	Yes	Yes	Yes		Yes	Yes	Yes
Year-quarter FE	Yes	Yes	Yes		Yes	Yes	Yes
Observations	21581	18718	18718		21581	18718	18718
R-squared	0.0700	0.0786	0.0787		0.0713	0.0812	0.0813

7. Concluding Remarks

This paper provides evidence that SOEs could function as a stabilizer by extending trade credit to downstream firms during a crisis period. The ongoing COVID-19 pandemic represents an unexpected negative shock to the supply chain and then to the real economy, which makes it an ideal scenario to analyze the role of SOEs during the crisis. We propose a trade credit channel through which COVID-19 shock could affect SOEs’ provision of trade credit to downstream firms in China’s context. SOEs have better access to financial resources and information advantage derived from the local government and thus could play a positive role for the economy during the crisis. Therefore, the effects of COVID-19 shock on the economy might be alleviated through SOEs’ provision of trade credit to downstream firms.

Employing a quarterly sample of Chinese listed manufacturing firms from the 1^st quarter of 2017 to the 4^th quarter of 2020, we find significant evidence in favor of our hypothesis. First, since SOEs are less financially constrained in general, we find that they provide more trade credit relative to domestic private firms during the COVID-19 pandemic period. Second, we find that better financing ability and multitask of SOE are two mechanisms through which SOEs could extend more trade credit during the pandemic, and the latter is more important. Moreover, we find evidence that SOEs’ advantage in trade credit extension is more pronounced in industries with higher external financial dependence and provinces with a higher level of government involvement, suggesting that SOEs might have greater comparative advantage in screening due to its involvement in local economy. However, it is important to note that although we observe an enhancing effect in trade credit provision of SOEs during the pandemic, the findings do not imply that a high level of SOEs is necessarily net beneficial for real economy. We do not have the data on who receives the trade credit and we cannot conduct the analysis on the whole social welfare.

Our result complements the literature on trade credit, corporate governance, and also the impact of the ongoing COVID-19 pandemic. First, we examine the role of firms’ ownership as a determinant of extending trade credit in crisis periods. We argue that this is the key to understand the resilience of Chinese economy after the onset of COVID-19. Previous studies present evidence that state-owned banks could stabilize the economy through countercyclical lending strategy (Coleman and Feler, 2015, Chen et al., 2016). We show that not only state-owned banks but also SOEs have stabilizing effect against negative shock. Second, our findings complement the studies on the role of state shareholders in the corporate governance. Except for the better financial ability, we show that SOEs have better screening capability that facilitates credit supply. Third, our results complement the latest literature concerning the real effects of crisis by identifying another channel through which the firm ownership plays an important role during the global financial crisis and the ongoing pandemic.

Uncited references

Wang et al., 2019, Whited and Wu, 2016.

CRediT authorship contribution statement

Xun Wang: Conceptualization, Methodology, Formal analysis, Writing – review & editing, Supervision. Jingwen Yu: Conceptualization, Data curation, Software, Formal analysis, Writing – original draft.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

Data will be made available on request.