Corporate Innovation and Dividend Payout Policy Evidence From China’s Listed Companies

Using the sample of China’s listed companies, this study investigates the effect of corporate innovation on dividend payout policy. We expect that managers from more-innovative ﬁrms would have more conﬁdence in future corporate performance, and hence, be more motivated to signal future proﬁtability in stock market through distributing cash dividends. Employing the number of effective patents to measure corporate innovation, we ﬁnd that the propensity to pay cash dividends is higher among more-innovative ﬁrms. Moreover, more-innovative ﬁrms usually pay more cash dividends to their shareholders than that of less-innovative ﬁrms. The empirical results are robust to different model speciﬁcations and to various subsamples based on ﬁrm size. This study emphasizes the importance of stimulating corporate innovation in both domestic economic growth and shareholder protection. This study sheds light on the role of corporate innovation in major corporate decisions in China, which is the largest emerging economy in the world. Moreover, the empirical ﬁndings of this study also have a policy implication to Chinese regulatory bodies who have been urging ﬁrms listed in domestic stock exchanges to pay cash dividends. Regulatory bodies may spend more effort on encouraging innovation, as corporate innovation motivates the managers of listed ﬁrms to distribute proﬁts to shareholders


I. INTRODUCTION
Innovation and property right protection have been a great challenge to sustainable economic growth in most emerging economies.China has overtaken Japan as the second-largest economy in the world since 2010.However, China's economic growth has been criticized as lacking technology innovation.To stimulate sustainable economic growth and national competitiveness, Chinese government has spent a great effort on promoting the research and development (R&D) activities during the past decade.It is noteworthy that China's patent office received the highest number of applications in the world in 2019 as reported by the World Intellectual Property Organization (WIPO).A report by the National Science Board (NSB), which serves as an advisor to both the US President and the Congress, showed that The associate editor coordinating the review of this manuscript and approving it for publication was Davide Aloini.
the center of high-tech gravity was shifting to Asia, and to China particularly, and that China's share of the world's high-technology manufacturing increased from 8% in 2003 to 24% in 2012, in comparison with the number one US share of 27%.China's high technology manufacturing remains expansive.The total R&D input of high technology manufacturing was 380.4 billion RMB, and the revenue from high technology manufacturing enterprises reached 15.78 trillion RMB in 2019.According to the report by World Intellectual Property Organization, China tops the ranking for both the source (filings by China) and the destination (filed in China) for the four types of intellectual property, namely patents, utility models, trademarks and industrial designs) in 2013.PwC reports that, from 2007 to 2015, in-country R&D spending increased 120 percent for China but only 34 percent for U.S.Moreover, many leading Chinese enterprises have been ranked as the top 500 companies based on their R&D expenses.For example, the R&D expenditure of Huawei, a Chinese technology company, has been ranked as the top 20 highest during the past decade according to the report by European Commission.The rapid growth of patents provides a fertile testing ground for the research upon corporate innovations in China.However, the empirical evidence on the role of corporate innovation in China and other emerging economies is quite limited.
Corporate innovation has become the focus of financial research in the recent years.As a crucial contributor to future corporate profits, innovation is pursued by many firms across a wide spectrum of industries [3], [25], and helps firms create new market spaces [12].Compared with non-innovators, innovative firms usually grow significantly faster [24], [25], have higher market shares [5], and have higher market valuation [4], [7], [23].Given the importance of corporate innovation, it may consequently exert influence on major corporate behaviors.We argue that innovative firms are more likely to distribute cash dividends to their shareholders than that of less-innovative firms.Apart from the agency theory, another view of cash dividend payout is the signaling-effect (Fig. 1).Literature suggests that managers pay cash dividends as a credible signal to stock market for the prediction of future earnings or increase in the future cash flows [18], [19], [20].Since corporate innovation helps firms create new market spaces [12], gain higher market shares [5], [8], and increase market value [23], managers from innovative firms may be more motivated to signal good future performance through cash dividends.
As such, this study investigates the relation between corporate innovation and cash dividends in China during the period from 2007 to 2017.The percentage of Chinese listed firms that pay cash dividends to their shareholders has been increasing since early 2000s.Most studies argue that cash dividends are beneficial to investors, especially minority investors in China.Eun and Huang [9] suggest that cash dividends signal shareholder-friendly corporate governance, and thus individual investors are willing to pay a premium for cash dividend paying stocks in China.Their findings are in line with the signaling-effect of cash dividends.Using the number of effective patents to measure listed firms' innovation, we find that firms with more patents are more likely to pay cash dividends to their shareholders.The dividend payout amounts of more-innovative firms are also higher than that of less-innovative firms.The empirical results are robust to different model specifications and to various subsamples based on firm size.
Existing studies argue that corporate innovation is susceptible to financial constraints, as innovation activities has a long-term investment cycle and requires an assurance to funding sources [4], [25].Accordingly, our empirical findings raise concern about the future innovation activities of listed firms, as paying cash dividends reduce their level of cash holdings.It is noteworthy that the cash dividend payout rates of Chinese listed companies are among the lowest in the world from 1990 to 2004 [27].The dividends/earnings ratios of Chinese listed firms, which were in the range of 25% to 35%, are much lower than that of European firms, which were in the range of 40% to 60% [28], and US firms, which were in the range of 60% to 80% during 1990-2001.We also find that firms with more patents (i.e., with above-median patents) distribute on average 33% of net profit to their shareholders.In general, listed firms retained the majority of their earnings.Therefore, paying cash dividends are less likely to impede corporate innovation in China.
This study contributes to the literature on the role of corporate innovation in major corporate decisions, which has rarely been examined.Moreover, the empirical findings of this study also have a policy implication to Chinese regulatory bodies who have been urging firms listed in domestic stock exchanges to pay cash dividends, as corporate innovation helps motivate listed firms to distribute their profits to shareholders.

II. LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT
As mentioned earlier, corporate innovation promotes growth and boosts market valuation [4], [5], [7], [23], [24].Many existing studies also investigate the determinants and consequences of corporate innovation [1], [3], [25].These studies find that innovation plays an important role in major corporate decisions, such as mergers and acquisitions (M&A).More specifically, acquiring innovation is an important motive for firms undertaking M&A.Furthermore, technological overlaps between acquiring and target firms have a positive effect on the incidence of M&A deals [3].Innovation-driven acquisitions also have better long-term real outcomes, i.e., higher sales growth and better long-term stock market performance [3].
Firms' innovation activities can be measured by using either innovation input (i.e., in-house R&D investment) or output (i.e., the number of effective patents).R&D investment VOLUME 9, 2021 is a critical element of innovation.However, it is noteworthy that R&D investment is associated with a high risk, as some R&D activities may fail.Additionally, R&D data may be severely distorted when various subsidy schemes are provided, especially in China [17].Therefore, most studies employ patents as the proxy for firms' innovation [3], [23].As the output of innovation activities, patents are one of the most-known isolating instruments that prevent competitors from easily imitating valuable firm-specific assets for competitive advantages [22].Therefore, this study employs the number of effective patents to measure listed firms' innovation.
Since corporate innovation is a crucial contributor to future corporate profits and market valuation, and is also a key trigger of M&As, it may affect other corporate behaviors, such as cash dividend payouts.Dividend policy has been of interest in the finance literature since the evolvement of the cash dividend puzzle of Miller and Modigliani [18].The most cited explanation on the cash dividend puzzle is based on the agency theory.Cash dividend payouts may reduce agency costs as cash dividend payouts increase the propensity of using external financing to monitor the managers at relatively lower cost [29].Moreover, managers may act in their own interest at the expense of the shareholders by spending cash for their own benefit, and cash dividend payouts can decrease the source of cash under the manager's control [30].Furthermore, managers usually underinvest because of risk aversion, and smooth dividend payouts using the cash retained from underinvestment [16].
Another view of dividend payout is the signaling-effect, which is more related to this study.The signaling theory argues that managers pay cash dividends as a credible signal to the market for the prediction of future earnings and increase in the future cash flows [18], [19], [31].Signaling-effect has been supported by many studies.For instance, Kalay and Loewenstein [32] and Nissim and Ziv [20] report a positive relationship between dividend changes and listed firms' future earnings in the US.A more recent study by Bozos et al. [33] using data from the London Stock Exchange also supports the dividend signaling theory.Hoang and Hoxha [13], [14] further report that payout policy is influenced by investment policies.
The signaling-effect is a major explanation for the cash dividend payouts of the listed firms in China.Cheng et al. [34] and Chi et al. [35] report that to a certain extent, paying cash dividend could provide a signaling-effect to the stock markets in China.The Chinese listed firms with higher earnings per share (EPS) are claimed to be more likely to pay cash dividends.Furthermore, Eun and Huang [9] argue that cash dividends signal shareholder-friendly corporate governance, and thus the individual investors are willing to pay a premium for cash dividend paying stocks in China.
In conjunction with the signaling theory that listed firms usually pay cash dividends as an indication of possessing positive future prospects, we expect that more innovated listed firms (with more effective patents) would have a higher propensity of distributing profits, and also higher cash dividend payout rates.Consequently, our research hypotheses are: H1: All else being equal, listed firms with more effective patents are more likely to pay cash dividends; H2: All else being equal, listed firms with more effective patents pay higher cash dividends.

III. RESEARCH DESIGN
The data on corporate patents, dividend payouts and other control variables are obtained from the China Stock Markets and Accounting Research (CSMAR) database.The sample period is from 2007 to 2017, as the data on corporate patents become relatively complete in the recent years.We exclude financial firms to make our sample comparable with other studies.The initial sample includes 12566 observations of corporate patents.After matching with the data on dividend payouts and other firm characteristics, the sample size is reduced to 9093.We further winsorize variables defined as ratios, namely CPS, C/A, PATENT, STATE, GRWOTH, E/A, V/A, EXE_SHARE, CASH, and LEV, at the upper and lower 1% levels.The final sample contains 8207 observations.A definition for these abbreviations is provided in the Appendix.
We employ logistic regressions to examine the effect of corporate patents on the likelihood of paying cash dividends.Dependent variable, namely CD, takes the value of 1 if a firm pays cash dividends in year t, and takes the value of 0 otherwise.PATENT, which is the key independent variable, is defined as the number of effective patents scaled by the book value of total assets (total assets are in millions of Chinese RMB).Following Fama and French [10], GROWTH, V/A, and E/A are included in the model as the basic determinants of the dividend payments.GROWTH is the growth rate of total assets.V/A is the ratio of the aggregate market value to the aggregate book value of total assets.E/A is the ratio of earnings before interest to total assets.To control for the effect of corporate governance on dividend payout policy, we include STATE and EXE_SHARE in our regression analyses.STATE is measured by the percentage of shares held by government (including its agencies) or parent state-owned enterprises [35].EXE_SHARE, which is the percentage of shares held by top executives, controls for the effect of managerial ownership on alleviating principal-agent conflicts [15], [35].
Other control variables are CASH, LEV, and SIZE [13], [26], [35].CASH is net cash flow from operating activities scaled by the book value of assets.LEV is the proxy for financial leverage and is defined as the ratio of total liability to total assets.SIZE is the natural logarithm of total assets.Moreover, since the number of listed firm's patents may differ across industries, we include industry fixed effect in our regression analyses.Year dummies are also included to control for changes in macroeconomic environment common to all listed firms over the sample period.We run the logistic regression below: It is noteworthy that some listed firms may continuously pay dividends to their shareholders.A firm's past dividend pattern may affect its future payout policy.Accordingly, the variable CD is auto-correlated.Therefore, we include the lagged value  of CD (CD it−1 ) in the regression to control for the effect of listed firms' past dividend patterns in another model.Furthermore, we use patent change (i.e., the difference the number of effective patents between current and previous year) as the independent variable in another model.This also helps us find out whether a firm is more likely to pay cash dividends if acquiring more patents in the current year than in the previous year.The model is shown below.
Following Firth et al. [11], we use two variables to measure listed firms' dividend payout ratios: C/A and CPS.C/A is cash dividend per share over the book value of assets per share in a year.CPS is cash dividend per share.The OLS regression models are shown below.
Moreover, we employ another logistic model to confirm the robustness of empirical results derived from models 3 and 4. The dependent variable is CPSDM, which takes the value of 1 if a firm's CPS ratio increases from year t-1 to t, and takes the value of 0 otherwise.The main explanatory variable is PATENT, which is the change in the number of patents from year t-1 to t scaled by the book value of assets.The logistic regression is shown below.

IV. EMPIRICAL RESULTS
Table 1 shows the descriptive statistics of all variables used in this study.Among 8207 firm-year observations, 73.6% (or 6040) have cash dividend payments.The average C/A (cash dividend per share over total assets share) ratio is 0.016, and the average CPS (cash dividend per share) is 0.127.The minimum value of PATENT is 0, which suggests that some listed firms do not have any patent during the sample period.
Table 2 reports the results of correlations among variables used in regression analyses.The correlations between all pairs of independent variables are below 0.5.As such, the correlations are not high enough to cause multicollinearity.We further perform a VIF test.All VIF factors are below 2, which confirms our argument that the correlations are not high enough to cause multicollinearity.Due to size limit, the result is not reported, but available upon request.Table 3 reports the empirical results of logistic regressions testing the propensity to pay.Dependent variable (CD) is a dummy variable that takes the value of 1 if a firm pays cash dividends in year t, and takes the value of 0 otherwise.In Model 1, the coefficient on PATENT is significantly positive at the 1% level.This indicates listed firms are more likely to pay cash dividends when holding more patents hand.In Model 2, we add one more control variable, which is the lagged value of CD (CD it−1 ), in logistic regression.This is because some listed firms may consecutively pay cash dividends to their shareholders.Hence, their past dividend patterns may affect future payout policy.
As shown in Table 3, although the coefficient on PATENT decreases from 5.149 to 3.682, it is still significantly positive at the 1% level.In Model 3, we use patent change ( PATENT) as the key independent variable.The coefficient on PATENT is also significantly positive at the 1% level.This suggests that listed firms are more likely to pay cash dividends if the number of their patents increases from year t-1 to t.The results of control variables are basically consistent with the literature.More profitable firms are more likely to pay cash dividends, but highly leveraged firms are less likely to pay cash dividends [15], [19].The coefficients on EXESHARE are significantly positive across all regressions, which indicate that managerial holdings are an effective mechanism to mitigate agency problem [2], [21].
Table 4 reports the results of Equations 3, 4, and 5. Panel A shows the regression results of Equations 3 and 4. When CPS is used as the dependent variable (Model 1), the coefficient on PATENT is significantly positive at the 1% level.In Model 2, PATENT is still significantly associated with C/A at the 10% level.Therefore, listed firms with more patents tend to distribute more cash dividends to their shareholders.Panel B reports the result of logistic regression (Equation 5).Dependent variable is a dummy variable that takes the value of 1 if a firm's dividend per share increases from year t-1 to t, and takes the value of 0 otherwise.The coefficient on PATENT (patent change) is significantly positive at the 5% level.This indicates that listed firms are likely to increase their CPS ratios when their patents increase from year t-1 to t.

V. ROBUSTNESS CHECK A. ENDOGENEITY
It is noteworthy that listed firms with high dividend payouts are usually more profitable, and hence, have more resources to innovate.As such, there may be a possibility of reverse causality between corporate innovation and dividend payouts.To control for reverse causality, we employ system  generalized method moments (GMM) techniques developed by Arellano and Bover [36] and Blundell and Bond [37] in dynamic panel data.The system GMM in dynamic panel data can alleviate the possible positive bias in difference GMM due to the persistence of lagged dependent variables as instruments [37].The results of system GMM estimators in dynamic panel data models are reported in  overidentifying restrictions at conventional levels for either specification estimated.Furthermore, there is evidence of first order serial correlation in the residuals, and the AR(2) test statistics (p-value > 0.1) reveals absence of second-order serial correlation in the first-differenced errors.Therefore, the instruments used in the system GMM model are valid.The coefficients on PATENT are still significantly positive when either C/A or CPS is used as dependent variable.Accordingly, our empirical results are confirmed by the system GMM approach.
Propensity-score matching (PSM) is another commonly used estimation procedure that addresses endogeneity and selection bias by matching sample firms with control firms having similar characteristics according to a function of covariates.Following Sun et al. [38], we employ the nearest neighbor (NN), Radius and Kernel techniques to perform the PSM tests.The basic approach to PSM is to first model the determinants of variation in PATENT among firms.We divide our sample into two groups based on the yearly median level of PATENT.The group with above-median PATENT is used as the treated group, and the group with below-median PATENT is used as the control group.We include the holdings of institutional investors and the log value of firm age that may explain the variation in PATENT.
As reported in Panel A of Table 6, the p-values for the t-test indicate that the matching algorithm was successful in achieving balance for all covariates.More specifically, 8 of the 10 t-tests are statistically insignificant between the treated and the control sub-groups in all three PSM techniques.Panel B reports the PSM regression results for the NN (columns 1 to 3), Radius (columns 4 to 6), and Kernel (columns 7 to 9) methods.The coefficients on PATENT are significantly positive across all regressions.Accordingly, the PSM analysis confirms the positive association between corporate innovation and cash dividend payouts.

B. FIRM SIZE
Literature suggests that large firms usually have more stable profits and market shares, but lack growth opportunities.Those firms tend to distribute more cash dividends to their shareholders [10].On the contrary, small-sized listed firms normally have more growth potential, and hence, prefer to retain their earnings.As such, we further split the entire sample into three subsamples based on the book value total assets: large firms, medium size firms, and small size firms.Table 7 presents the results of regression analyses by firm size.Panel A reports the results of logistic regressions testing the propensity to pay cash dividends.The coefficients on PATENT are significantly positive at the 1% level for large and small size firms.Panel B reports the results of OLS regressions testing the effect of patents on payout rates.The coefficients on PATENT are significantly positive at either the 5% or 10% level across all regressions.As a further robustness test, we use firm fixed effects to replace industry fixed effects, and include random effect instead of fixed effect in our regression analyses.The results remain unchanged.Due to the space limit, those results are not reported, but are available upon requested.

VI. CONCLUSION
Existing studies suggest that corporate innovation helps listed firms gain market shares and boost market value.Such effect may enhance the confidence of managers in future corporate performance.In conjunction with the signaling effect of cash dividends, we argue that managers from innovative firms would have more incentive to pay cash dividends.Using the number of effective patents to measure innovation, we find that corporate innovation increases the propensity of paying cash dividends.Moreover, the number of effective patents is also positively associated with the level of cash dividend payment.The empirical findings are robust to different model specifications and to different subsamples based on firm size.Our findings are also in line with the dividend signaling theory.
China is a civil law country with relatively weak investor protection.Since the establishment of domestic stock exchanges, controlling shareholders' embezzlement has been argued as a major problem of Chinese stock markets [39], [40].Many listed firms prefer to keep profits as retained earnings and do not pay cash dividends.On average, the cash dividend payout rates of Chinese listed companies are among the lowest in the world from 1990 to 2004 [27].As such, China Securities Regulatory Commission (CSRC) has been spending a great effort on encouraging listed firms to pay cash dividends.In 2006, CSRC required listed firms applying for equity refinancing to pay cash dividends of no less than 20% of the distributable profit in the previous three years.In 2013, the CSRC even announced that it would officially interview and pay special attention to non-dividend-paying listed firms.Therefore, the policy implication of this study is that Chinese regulatory bodies may keep stimulate the innovative behaviors by listed firms, as corporate innovation could not only boost domestic economic growth but also cause an increase in cash dividend payouts of listed firms.
This study further sheds light on the role of corporate innovation in dividend payout policy, which is a major corporate behavior.Such effect has rarely been examined by existing studies, especially in emerging economies.This study also adds new evidence to the dividend signaling theory.

FIGURE 1 .
FIGURE 1.This figure shows that managers from more-innovative firms may be more willing to distribute cash dividend to signal further profitability.

Future
studies can contribute to the literature on the effect of corporate innovation on stock dividend payment and share repurchase, which are also quite popular in China and other emerging markets.LIPING LIAO received the Ph.D. degree in management science from the Guangdong University of Technology, China.She is currently a Professor of management with Guangdong Polytechnic Normal University, China.She has published research articles in academic journals.Her research interests include systems engineering, innovation technologies, cyberspace security technology, and so on.KWOK FAI GEOFFREY TSO received the Ph.D. degree in biostatistics from the University of Toronto.His research has appeared in Computational Statistics & Data Analysis, Energy, Social Indicators Research, the Journal of the Association for Information Systems, the Journal of Applied Statistics, Communications in Statistics -Simulation and Computation, and others.His research interests include market research, economic indexes, energy research, business intelligence, and deep networks.JINGJING YANG received the Ph.D. degree in finance from Massey University, New Zealand.He is currently a Professor of finance with the Guangdong University of Foreign Studies, China.He has published research articles in academic journals.His research interests include corporate governance and corporate finance, and so on.

TABLE 3 .
Effects of corporate patents on the likelihood of paying dividends.

TABLE 4 .
Effects of corporate patent on cash dividend payout ratios.

TABLE 5 .
Results of system GMM estimators in dynamic panel data models.

Table 5 .
The Sargan tests (p-value > 0.1) do not suggest rejection of the

TABLE 7 .
Regressions by firm size.

TABLE 8 .
Definition of variables.