This paper selects A-share listed Internet and related digital service enterprises as the research sample. According to the Guidelines for the Industrial Classification of Listed Companies issued by the JR/T 0020-2024 ([4]), the industrial classification of listed companies includes “I Information Transmission, Software and Information Technology Services”, under which there are subcategories such as “64 Internet and Related Services” and “65 Software and Information Technology Services”. Based on this classification standard and the main business characteristics of listed companies, this paper mainly selects firms whose businesses are closely related to Internet platforms, online games, digital content, cybersecurity, cloud computing, data services, Internet media, software, and information technology services.

Considering the research topic, data availability, and sample continuity, this paper finally selects 25 listed companies as research objects, including East Money, Hithink RoyalFlush Information Network, 360 Security Technology, Kunlun Tech, 37 Interactive Entertainment, Kingnet Network, Perfect World, G-bits Network Technology, Ourpalm, Mango Excellent Media, Giant Network, Zhejiang Century Huatong Group, Beijing Shenzhou Taiyue Software, Youzu Interactive, Wuxi Boton Technology, Shenzhen ZQGame, Hangzhou Shunwang Technology, ChineseAll Digital Publishing Group, Zhejiang Daily Digital Culture Group, Hubei Century Network Technology, Shanghai Ganglian E-Commerce Holdings, Beijing Zhidemai Technology, Daily Interactive, UCloud Technology, and People.cn. These enterprises are involved in Internet financial information services, cybersecurity, online games, digital content, Internet media, cloud computing, and data services, and are therefore closely related to the Internet industry and the digital economy.

The underlying data used in this paper cover the period from 2021 to 2025, while the final regression sample covers fiscal years 2022 to 2025. This is because corporate innovation resilience, denoted as IR, is measured by the growth rate of R&D expenses, and Growth is measured by the growth rate of operating revenue. Both indicators require prior-year values as the calculation base. Therefore, the 2021 data on R&D expenses and operating revenue are used only to calculate the 2022 values of IR and Growth, and 2021 itself is not included as a regression observation. In the regression analysis, Lev, Size, ROA, IR, and Growth are all matched to the fiscal years from 2022 to 2025. As a result, the final regression sample consists of 25 listed firms over four fiscal years from 2022 to 2025, generating 100 firm-year observations.

The sample screening procedure is as follows. First, A-share listed companies in the CSRC category of “I Information Transmission, Software and Information Technology Services” are selected, mainly including firms classified under “64 Internet and Related Services” and “65 Software and Information Technology Services”. Second, firms whose main businesses are not closely related to Internet platforms, online games, digital content, cybersecurity, cloud computing, data services, Internet media, software, or information technology services are excluded. Third, ST and *ST firms are excluded to avoid the influence of abnormal financial conditions. Fourth, firms or observations with missing key variables, including R&D expenses, operating revenue, total assets, total liabilities, and net profit, are excluded. Fifth, observations with obvious abnormalities are excluded. In this paper, obvious abnormalities refer to cases in which key accounting items are missing or inconsistent across data sources, total assets, previous-year R&D expenses, or previous-year operating revenue are non-positive so that ratio or growth-rate indicators cannot be calculated properly, or the reported data contain apparent unit errors or data-entry errors after cross-checking with annual reports and database records. After applying the above screening rules, 25 listed companies are finally retained as the research sample.

The data required in this paper mainly include R&D expenses, operating revenue, total assets, total liabilities, and net profit. These data are mainly obtained from listed companies’ annual reports, Eastmoney Choice data, and periodic reports disclosed on CNINFO. To ensure the reliability and consistency of the empirical analysis, the data are checked and matched across firms and years before the regression analysis is conducted.

1) Dependent variable: corporate innovation resilience. This paper takes corporate innovation resilience as the dependent variable, denoted as IR. Corporate innovation resilience mainly reflects an enterprise’s ability to continuously maintain R&D investment under changes in the external environment or operational pressure. Considering that the innovation activities of Internet enterprises are usually based on R&D investment, this paper measures corporate innovation resilience by the growth rate of R&D expenses. A higher value indicates faster growth in R&D expenses and stronger capability for continuous innovation. The formula is as follows:

IR = (Current-year R&D expenses − Previous-year R&D expenses)/Previous-year R&D expenses

In the empirical analysis, IR is calculated for fiscal years 2022-2025, while the 2021 R&D expense is used only as the base value for calculating the 2022 IR.

2) Explanatory variable: debt leverage. The core explanatory variable in this paper is debt leverage, denoted as Lev. Debt leverage measures the proportion of a firm’s assets financed by debt and reflects its capital structure and financial risk level. This paper uses the asset-liability ratio to measure debt leverage. The formula is as follows:

Lev = Total liabilities/Total assets

3) Control variables. To reduce the influence of other firm characteristics on innovation resilience, this paper selects firm size, profitability, and growth ability as control variables. Firm size is denoted as Size and is measured by the natural logarithm of total assets. Profitability is denoted as ROA and is measured by the ratio of net profit to total assets. Growth ability is denoted as Growth and is measured by the growth rate of operating revenue. The formulas are as follows:

Size = ln(Total assets)

ROA = Net profit/Total assets

Growth = (Current-year operating revenue − Previous-year operating revenue)/Previous-year operating revenue

Similarly, Growth is calculated for fiscal years 2022-2025, and the 2021 operating revenue is used only as the base value for calculating the 2022 revenue growth rate (Table 1).

Table 1. Variable definitions.

To examine the impact of debt leverage on corporate innovation resilience, this paper constructs the following regression model:

IR = alpha + beta Lev + gamma Size + delta ROA + theta Growth + epsilon

In this model, IR represents corporate innovation resilience, Lev represents debt leverage, Size represents firm size, ROA represents profitability, Growth represents growth ability, and epsilon is the random error term. If beta is negative, debt leverage may weaken corporate innovation resilience; if beta is significantly negative, the inhibitory effect is statistically supported.

Table 2. Descriptive statistics.

The descriptive statistics show that the number of observations for all main variables is 100, indicating that the sample data are relatively complete. In terms of corporate innovation resilience, the mean value of IR is −0.0024, which is close to zero. This indicates that the growth of R&D expenses among the sample Internet enterprises is generally stable during the research period, but the overall growth momentum is not strong. Its standard deviation is 0.1792, with a minimum of −0.39 and a maximum of 0.59, indicating large differences in R&D investment changes across enterprises and years and suggesting a certain degree of differentiation in innovation resilience. In terms of debt leverage, the mean value of Lev is 0.2897, indicating that the average asset-liability ratio of the sample firms is about 28.97%, which represents a moderate overall debt level. However, its minimum value is 0.08 and its maximum value is 0.87, indicating substantial differences in capital structure among firms and relatively high debt levels for some individual enterprises. In terms of firm size, the mean value of Size is 22.8533 and the standard deviation is 1.2502, suggesting differences in asset scale and resource base among the sample firms. In terms of profitability, the mean value of ROA is 0.0345, indicating an average return on assets of about 3.45% and an overall moderate level of profitability. The minimum value of −0.42 indicates that some enterprises experienced losses. In terms of growth ability, the mean value of Growth is 0.0376, indicating that the average growth rate of operating revenue is about 3.76%, showing slight overall growth. However, the minimum value is −0.41 and the maximum value is 0.73, indicating large differences in growth ability across enterprises. Overall, the sample firms differ in innovation resilience, debt level, profitability, and growth ability, making them suitable for subsequent correlation and regression analyses (Table 2).

Table 3. Correlation analysis.

The correlation analysis shows that the correlation coefficient between debt leverage Lev and corporate innovation resilience IR is −0.0384, indicating a negative correlation. This suggests that enterprises with higher debt levels may have relatively lower growth rates of R&D expenses. However, this result does not pass the significance test, indicating that the simple linear correlation between the two variables is not obvious. The correlation coefficient between firm size Size and IR is 0.2579 and is significant at the 1% level, indicating that larger enterprises usually have stronger resource bases and R&D investment capabilities and therefore higher innovation resilience. The correlation coefficient between profitability ROA and IR is 0.2119 and is significant at the 5% level, indicating that more profitable enterprises are more likely to maintain growth in R&D investment. The correlation coefficient between growth ability Growth and IR is 0.4164 and is significant at the 1% level, indicating that enterprises with faster operating revenue growth also have stronger innovation resilience. In addition, Lev is significantly positively correlated with Size, indicating that larger enterprises tend to have relatively higher debt levels. Lev is significantly negatively correlated with ROA, indicating that firms with higher debt levels may have relatively weaker profitability. Overall, the correlation analysis preliminarily indicates a negative but insignificant relationship between debt leverage and corporate innovation resilience, while firm size, profitability, and growth ability show more obvious positive relationships with innovation resilience (Table 3).

Table 4. Analysis of regression results.

The regression results show that the sample size is 100, the F-statistic is 7.20, and Prob > F is 0.0000, indicating that the model as a whole passes the significance test and has certain explanatory power. The R-squared is 0.2327 and the adjusted R-squared is 0.2004, indicating that the model can explain about 23.27% of the variation in corporate innovation resilience. Regarding the core explanatory variable, the regression coefficient of debt leverage Lev is −0.1045, indicating a negative relationship between debt leverage and corporate innovation resilience. In other words, with other conditions unchanged, the higher the debt level of an enterprise, the lower its growth rate of R&D expenses may be, and the weaker its innovation resilience may become. However, the P-value is 0.312, which does not pass the significance test, indicating that this negative effect is not statistically significant in the sample. Therefore, the research hypothesis is supported only in terms of direction. Regarding the control variables, the coefficient of firm size Size is 0.0357 and is significant at the 5% level, indicating that larger firms have stronger resource bases and R&D investment capabilities, which helps improve innovation resilience. The coefficient of growth ability Growth is 0.2919 and is significant at the 1% level, indicating that the faster the growth of operating revenue, the more obvious the growth in R&D investment and the stronger the innovation resilience. The coefficient of profitability ROA is positive but not statistically significant, suggesting that profitability does not have an obvious direct effect on innovation resilience. Overall, debt leverage has a negative but insignificant impact on corporate innovation resilience, while firm size and growth ability are more important factors affecting the innovation resilience of Internet enterprises (Table 4).