Firm ownership structure, asset tangibility and the SA index have been widely used as the measurements of financing constraints ( Almeida et al., 2004 ; Berkowitz et al., 2015 ; Liu et al., 2022 ; Lyu, 2024 ). To gain robust empircal conclusions, we choose all three mearsures in the paper.

According to existing literature, state-owned enterprises (SOEs) have more political connection with credit market and have better access to raise bank loans, while private companies face many difficulties and obstacles in external debts ( Cull et al., 2015 ; Ge et al., 2017 ; Megginson et al., 2014 ). Asset tangibility refers to the proportion of fixed assets to total assets. It’s known that the quantity of fixed assets, which can be used as collateral, is an important basis for lenders to make loan decisions and firms with low tangibility have more barriers to increase debts ( Liu et al., 2022 ). Hence, private companies and low tangibility companies are considered relatively more financially constrained. As the proxy of financing constraints, the SA index is relatively stable and can accurately measure the financing constraints level ( Lyu, 2024 ). Generally, the value of the SA index is smaller than zero, and the smaller the value, the higher the financial constraint degree. The specific calculation formula is as follows：

$SA=-0.737\times \text{size}+0.043\times {\text{size}}^2-0.040\times \text{age}$ (1)

To examine the heterogeneious effects of COVID-19 pandemic on the leverage based on the different degrees of financial constraint, we construct a dynamic panel regression model as follows：

$L{R}_{it}={\text{β}}_0+{\text{β}}_{1}F{C}_{it}\times {\text{During}}_t+{\text{β}}_{2}F{C}_{it}+{\text{β}}_3{\text{During}}_t+\text{γ}{X}_{it}+{\text{μ}}_i+{\text{σ}}_t+{\text{ε}}_{it}$ (2)

where subscript i and t represent company i and year t, respectively. LR_it denotes one of three leverage ratios: Lev (total leverage), shortlev (short-term leverage) and longlev (long-term leverage) ² . FC_it represents one of three measurments of financial constraint discussed in section 2.1. During_t denotes an annual dummy variable which equals to one if the sample year is bigger than 2019, otherwise zero. X_it is a series of control variables including firm size (ln(asset)), firm sales growth (growth), return on assets (ROA), Tobin’s q (Q) and firm age (ln(age)), which have links to capital structure choices ( Petersen & Rajan, 1994 ; Myers & Majluf, 1984 ; Frank & Goyal, 2009 ). μ_i and σ_t represent firm-fixed effect and time-fixed effect, respectively, while εit denotes the random error term.

Our focus is β₁, the coefficient of the interaction term FC_it × During_t. Since a significant β1 indicates the influence of COVID-19 on companies with different degree of financing constraints is differential. To confirm our hypothesis, we certainly expect that it is positive to private × during and negative to tangibility × During and SA × During, which imply that contrary to usual, more financially constrained companies increase more leverage relative to less constrained companies during COVID-19.

All the data used in this paper is from the CSMAR database in Shanghai and Shenzhen Stock Exchange. In China, the COVID-19 broke out at the end of December 2019. In order to stop the spread of the epidemic, the Chinese government immediately implemented many stringent blocking measures ³ , which were not completely released until December 7^th, 2022. So the sample period we choose is from 2017 to 2022. In addition, we process the data as follows: Firstly, the companies with less than 5 observations are deleted; Secondly, all financial and construction companies are deleted; Finally, all key variables are winsorized at the top and bottom 1% level to avoid the influence of outliers.

Table 1 describes the statistical results of main variables before and during

Note: ln(asset) (Ln(age)) is the natural logarithm of total book assets (firm age plus one), growth is revenue growth rate. Total debts are the values in the table multiplied by one billion.

COVID-19 pandemic period. As shown, lev, shortlev and longlev all experience a significant growth ⁴ , while other indicators such as growth, ROA and Q present a certain degree of decline during COVID-19 period, which show that enterprises’ operation is badly affected and companies have to choose to increase financial leverage to tide over the economic crisis caused by the pandemic. In particular, we also noticed that the mean value of lev is 0.346 and the mean value of shortlev is 0.291, while the mean value of longlev is only 0.055. These statistics indicate that the loans of Chinese listed firms are mainly derived from short-term debts.

In this section, we focus on testing whether there is a significant and differential causal relationship between COVID-19 and leverage of companies with different degree of financial constraint by using model (2). Tables 2-4 show the empirical results using ownership structure, asset tangibility and the SA index as the measures of financing constraints, respectively.

As shown in Table 2 , the coefficients of interaction term private*during in column (1) and (2) are both highly significant in 0.016, while it’s not in column (3). These results imply that during COVID-19 period, private firms experience a significant increase in lev and shortlev by 1.6% points relative to non-private firms, while nor in longlev. From Table 3 , we can see the value of the coefficients of tangibility*during in columns (1), (2) and (3) are all significant in −3.9%, −5.1% and 1.2%. These results demonstrate that the lev and shortlev decrease with the increase of asset tangibility during COVID-19 period, while it’s opposite for longlev. Similarly, the coefficients of SA*during in Table 4 are all significant in -0.031, −0.023 and −0.008, which signifies that all three leverages decrease with the SA index increases during COVID-19, but it’s worth noting that the absolute value is much smaller for longlev compared to shortlev.

According to the analysis and discussion above, we can conclude that the lev, mainly shortlev, of companies with high level of financing constraints has experienced a greater growth than that of companies with low level of financing constraints during COVID period. This is probably due to governments’ credit support policies in the pandemic period which make financially constrained firms increase more short-term external financing than usual.

Certainly, there may exist another possibility: more financially constrained companies increase fewer assets compared to less constrained companies in that tough economic period ⁵ . In section 4, we try to make a further test to ascertain whether financially constrained firms exhibit a higher increase in debts during COVID-19.

Note: Private is a dummy variable, private = 1 if a firm is 100% privately owned, and private = 0 otherwise. Tangibility in Table 3 and ln(size) in Table 4 are both continuous variables. The values in parentheses are robust standard errors, and *, **, *** represent significance levels of 10%, 5% and 1%, respectively, the same below.

Continued