To test the influence of COVID-19 on the speed of corporate capital structure adjustment, following [12], [8], and [18], we use the classical partial adjustment model to estimate how quickly companies converge towards their target leverage ratio. The model is expressed as follows:

where L e v i , t ∗ is the target leverage ratio of firm i at year t, while L e v i , t refers to the book leverage ratio of firm i at year t. v denotes the leverage adjustment speed, measuring the percentage of the deviation from the target level that firms offset within a year.f_i and year_t represent firm and time fixed effects, respectively. ε i , t is a residual term.

We rearrange Equation (1) as follows:

We construct a multivariable regression model to estimate the target leverage, expressed as:

Because the target leverage is unobservable, the fitted value of Equation (3) is used as an indicator to measure the target, which is expressed as Equation (4).

whereX_i,t is a matrix of control variables affecting financial leverage choices. Among the control variables, we use the level of supply chain finance (SCF). Following [17], we measure it by the ratio of the sum of short-term loans, notes payable and accounts receivable to total assets, because as a novel financing method, SCF may affect the speed of leverage adjustment by alleviating financial constraints. As an essential factor in leverage decisions, firm size (Size) is also included. We consider enterprises’ profitability (ROA) because firms’ leverage has a negative connection with their profitability when their debt exceeds the optimal level ([25]). Asset tangibility (Tang) denotes the ratio of net fixed assets to total assets; enterprises with more fixed assets possess higher credit and better access to financing in the financial market ([10]). Finally, growth potential (Growth), which is the growth rate of total operating revenue over total assets, the non-debt tax shield ratio (Dep), and the median of industry leverage (Median) are all controlled. Specific definitions of these main variables are presented in Table 1.

With Equation (4) submitted into Equation (2), the final partial-adjustment model is constructed as follows:

Our focus is the coefficient of L e v i , t − 1 , because subtracting it from 1 yields the speed of leverage adjustment (SOA).

Table 1. Definitions of variables.

All sample companies are A-share firms listed on the Shanghai and Shenzhen Stock Exchanges. The sample period covers 2016 to 2022, with 2016-2019 defined as the pre-COVID period and 2020-2022 as the during-COVID period. All data used in this paper are obtained from the CSMAR database.

The data are processed as follows: (1) firms in the financial industry and in the information transmission, software, and information technology service industry are removed, because these industries primarily provide financing services and information technology for SCF; (2) firms with debt ratios greater than 1 or less than 0 are excluded; (3) ST and *ST firms are deleted; (4) firms with fewer than seven observations for key variables are dropped; (5) all continuous variables are winsorized at the 1% level in both tails to avoid the influence of extreme values on the scientific validity of the empirical results.

Table 2 presents the descriptive statistics of main variables. As shown, only the leverage ratio (Lev) and firm size (Size) increased significantly during the COVID-19 period compared with the pre-COVID period. In contrast, other variables, such as SCF, ROA, Growth, and Tang, all declined considerably. This suggests that production and business operations were severely damaged, resulting in lower revenue. The change in SCF is particularly noteworthy. During COVID-19, many policies were introduced in China to encourage the development of SCF. In practice, however, the level of SCF decreased. During the pandemic, supply chains were disrupted, undermining the real transaction background and closed-loop risk control on which SCF depends. At the same time, banks actively contracted their business because of declining profits and rising costs. These factors may explain the decline in SCF during the epidemic.

Table 2. Descriptive statistics.

***, **, and * represent rejection of the null hypothesis at the 1%, 5%, and 10% significance levels, respectively. The same applies to the tables below.

We first conduct regressions based on Equation (5) for the full, pre-COVID and during-COVID samples separately. The regression results are reported in Table 3. As shown, the coefficients of Lev_i,t−1 are both positive and significant at the 1% level. The SOA during COVID-19 (71.190%) is lower than that in the pre-pandemic period (83.900%). To examine whether the difference in the Lev_i,t−1 coefficient between columns (1) and (2) is significant, we manually compute the Z-statistic for the coefficients of Lev_i,t-1across the two regressions, and the result is presented at the end of the table. The Z-statistic is 3.026, indicating that the difference in the Lev_i,t-1coefficient between the two columns is statistically significant. These findings suggest that the COVID-19 pandemic slowed down the speed of leverage adjustment, which is consistent with [18].

In addition, we note that the coefficient of SCF is significantly positive at the 1% level, suggesting that SCF is one of the determinants of capital structure decisions.

Table 3. The speed of leverage adjustment before and during COVID-19.

Note: The parentheses indicate robust standard errors, besides the value at the end table is the Z-statistics. The same applies to the tables below.

The literature shows that leverage adjustment speeds are asymmetric between overleveraged and underleveraged firms, with overleveraged firms tending to adjust faster than underleveraged firms ([7]; [13]). [9] and [18] find that the uncertainty caused by sudden shocks, such as the 2008 financial crisis and COVID-19, significantly reduces the capital structure adjustment speed of leveraged firms.

In this section, the entire sample is divided into overleveraged and underleveraged firms. We define the difference between target leverage and actual leverage as the deviation of actual leverage from the target level. Firms with a deviation greater than 0 are classified as overleveraged, while those with a deviation less than 0 are classified as underleveraged. The split-sample results are reported in Table 4. It can be seen that all coefficients of L e v i , t − 1 are significantly positive at the 1% level. The leverage adjustment speed of overleveraged firms is higher than that of underleveraged firms for pre- and during-COVID periods, respectively, which is consistent with [7] and [13].

Importantly, for underleveraged firms, the leverage adjustment speed decreases from 78.547% in the pre-COVID period to 59.931% during the pandemic, a decline of 18.616%. As the method in Section 3.2, we also manually compute the Z-statistics to test whether the difference in the coefficient of L e v i , t − 1 between columns (1) and (2) is significant. The results at the bottom of the table indicate that the coefficient difference is significant. Similarly, for overleveraged firms, the adjustment speed decreases significantly by 13.039%. These results suggest that underleveraged firms seeking to add leverage are more severely affected by the COVID epidemic than overleveraged firms in terms of capital structure adjustment speed.

Table 4. Asymmetric leverage adjustment speed (underleveraged versus overleveraged firms).

Note:The fixed effect requires at least two observations per firm; otherwise, the firm is omitted. Since some firms have fewer than two observations after grouping, the total number of observations before and during COVID in this table is smaller than the 2691 reported in Table 3. The same applies to the tables below.

In this section, we attempt to explore the asymmetric effect identified in Section 3.3 from the perspective of SCF. For underleveraged companies, the objective is to increase leverage, and SCF may directly raise the debt ratio through new loans, bill financing, or other debt arrangements. For overleveraged companies, the goal is to deleverage, and SCF may indirectly reduce short-term debt repayment pressure through accounts receivable realization, cash flow improvement, and debt replacement. Therefore, we conjecture that SCF promotes corporate capital structure adjustment and alleviates the negative impact of COVID-19 on leverage adjustment speed for both under- and over-leveraged firms.

We divide the underleveraged and overleveraged firms into two groups, respectively, by the median of SCF. Table 5 reports the regression results for the full sample and the split samples according to Equation (5). For the full sample, the speed of leverage adjustment in the low-SCF group significantly decreases by 15.639%, while that in the high-SCF group significantly decreases by 11.323%. For the underleveraged group, the low-SCF firms substantially decrease the speed of leverage adjustment by 31.740%, while the high-SCF firms only decrease by 11.243%. For overleveraged firms, the low- and high-SCF firms decrease the leverage adjustment speed notably by 19.147% and 10.994%, respectively. These results show that the adjustment speed of firms with low SCF declined more than that of firms with high SCF; Notably, the low-SCF underleveraged firms’ adjustment speed drops sharply by 31.740%, more severely than that of other group firms. These findings also suggest that SCF mitigates the decline in leverage adjustment speed for both underleveraged and overleveraged firms during the pandemic, and the mitigating effect is greater for underleveraged firms.

Table 5. The leverage adjustment speed in pre- and during-COVID periods, group by SCF and leverage ratio.