Corporate innovation intention and capital market response : evidence from textual analysis

Abstract

Innovation capability is a centralized reflection of a country's economic strength and one of the fundamental factors that determine the economic development of a country or region. As listed companies are the country's most active innovative enterprises, they are the driving force of innovation activities and Research and Development (R&D) investment in China. The issue of how to correctly evaluate and analyze enterprises’ innovation capability is a matter of extremely important policy and academic significance. In this paper, we develop a new measure of innovation using the text of the annual report, firms’ innovation intention from an ex ante perspective, and explore the impact of innovation intention on firms' future stock price performance and innovation outcomes. As the management board is the key decision maker in formulating and implementing strategy, it shapes the firm's decision preferences. Those decision preferences then influence the firm's strategic choices and implementation, and they can largely determine the firm's innovation capability and innovation outcomes. Our text-based measure gives a useful description of innovation by firms with and without patenting and R&D (research and development). We select Chinese A-share listed companies from 2008 to 2020 and use machine learning to calculate the innovation vectors in their annual reports based on an analysis of the company's core competitiveness, reviews of the company’s operations during the reporting period, and the company's outlook on the future. These are the three parts of a listed company’s annual report best suited to present its innovation intention. By calculating the similarity between the innovation vectors of the annual report and the basic "Standard Innovation 300 Vector" constructed from seven well-known books on innovation, we determine the innovation intention of the company's management and thus construct the innovation intention indicator: InnoIntention. We find that future returns are higher for firms with high innovation intentions. More specifically, we use the method proposed by (Daniel, Grinblatt, Titman and Wermers,1997, hereinafter “DGTW”) to calculate the cumulative abnormal returns (CAR) during a window after the release of annual reports to reflect the impact of annual report information on the stock prices of listed companies. The following seven windows are selected: one week, two weeks, one month, one quarter, two quarters, three quarters, and one year after the release of the annual report. The regression coefficients of innovation intention are statistically significant for time windows greater than or equal to two weeks, controlled by the year and month of the release of the annual report, the industry in which the firm is located, and the firm’s Standardized Unexpected Earnings (SUE). One standard deviation increase in innovation intention results in 0.91%CAR one quarter after the release of the annual report. Further analysis shows that innovation intention remains significantly positive even after it is controlled by indicators of capabilities such as company R&D investment (R&D) and patent quality. This indicates that the inclusion of innovative activities in the company's annual report has independent predictive power for stock returns. We also construct portfolios to test the predictive power of innovation intention for stock returns. We construct long-short portfolios that involve buying stocks with high innovation intention while selling stocks with low innovation intention. After obtaining the monthly returns for each portfolio, we risk-adjust the monthly excess returns of each portfolio using the China 4 factor proposed by (Liu et al., 2019). With equally weighted average returns, the alpha of the stock portfolio with high innovation intention is 0.369% and the alpha of the stock portfolio with low innovation intention is -0.059%. Thus, the risk-adjusted alpha for the entire long-short portfolio is 0.428% per month, corresponding to an annualized alpha of 5.136%. This indicates that the indicators of innovation intention are helpful in constructing investment strategies with excess returns. Our text-based measure performs even better in identifying innovative firms without patenting or firms with missing R&D. We investigate the interaction between innovation intention and hard information (i.e., R&D and the number of patents). We find that for firms with missing R&D when innovation intention increases by one standard deviation, there is a 1.24% increase in CAR one quarter after the annual report and a 2.07% increase in CAR one year after the annual report. For companies with nonmissing R&D, when innovation intention increases by one standard deviation, cross-sectionality will result in a 0.41% increase in CAR one quarter after the annual report and a 0.5% increase in CAR one year after the annual report. In addition, the regression coefficients of the innovation intention indicator (InnoIntention) are non-significant for most time windows, suggesting that for firms with nonmissing R&D, the market's assessment of their innovation intention is based primarily on their hard information (R&D) rather than their innovation intention. We further find that for companies with low R&D outcomes when innovation intention increases by one standard deviation (8.27), there is a 1.24% increase in CAR one quarter after the annual report and a 2.23% increase in CAR one year after the annual report. For companies with higher R&D outcomes, when innovation intention increases by one standard deviation (8.27), there is a 0.66% increase in CAR one quarter after the annual report and a 0.83% increase in CAR one year after the annual report in the cross-section. Next, we discuss the interaction between patent granting and innovation intention. The coefficients of the interaction between the InnoIntention and NoPatent dummy are significantly positive. This indicates that InnoIntention has stronger predictive power for companies with firms with no patent grants. In a one-quarter time window, for companies with no patent, the predictive power of InnoIntention is nearly two times that for firms with patents. In a one-year time window, for companies with no patent, the predictive power of InnoIntention is nearly seven times that for firms with patents. Economically, for firms with no patent, when InnoIntention increases by one standard deviation (8.27), the cross-section results in an increase in CAR by one quarter after the annual report by 1.08% and an increase in CAR one year after the annual report by 1.74%. For firms with patents, when InnoIntention increases by one standard deviation, the cross-section results in an increase in CAR one quarter after the annual report by 0.58% and an increase in CAR one year after the annual report by 0.25%. We verify that innovation intention has significant predictive power for a company's future innovation investment, innovation performance, and financial performance. The increase in the number of patent applications and the increase in the number of patents acquired that correspond to a one standard deviation increase in innovation intention are quite significant economically. For example, after one, two, and three years, the number of applications acquired by public companies improves by 20%, 225%, and 210%, respectively, and the number of patents improves by 17%, 190%, and 380%, respectively, of the median. Finally, we conduct robustness tests. We find that the regression coefficients of InnoIntention for the main board subsample are higher than for the full sample. Due to the high level of R&D in the Growth Enterprise Market (GEM) and Science and Technology Innovation Market (STEM) listed companies, the predictive power of innovation intention for stock returns is weaker for non-GEM and non-STEM listed companies. In the traditional technology industry subsample, the regression coefficient of InnoIntention is significantly positive for all windows. In addition, the regression coefficient of InnoIntention is higher than the corresponding regression coefficient under the full sample. Overall, this research is valuable for testing the pricing efficiency and improving the effectiveness of China's capital market. In addition, it has important reference value for analyzing companies’ innovation.