Does cash dividend smoothing affect the wealth management products purchased by listed companies?

ABSTRACT Using the regulatory policy on cash dividends implemented in 2013 as the research context, this study develops a difference-in-difference (DID) model using the data of Chinese A-share non-financial listed companies from 2011–2019, and found that cash dividend smoothing reduces the wealth management products (WMPs) purchased by companies. Channel tests demonstrated that cash dividend smoothing reduces the company’s WMPs by increasing the financial leverage and attracting independent-type institutional investors to exert debt governance effect and institutional investor governance effect. Additional analysis found that cash dividend smoothing has a greater effect on WMPs when there are serious managerial agency problems within the firm, the nature of ownership is private, and in regions with dense branches of financial institutions. More importantly, cash dividend smoothing was found to alleviate the crowding-out effect of WMPs on industrial investment and help return the firms’ capital to industrial operations.


Introduction
In recent years, the downward pressure on China's economy has continued to increase and the industrial returns have continued to decline, and it is not uncommon for real companies to frequently purchase wealth management products (WMPs). According to the WIND database, the annual cumulative amount of WMPs purchased by non-financial listed companies in the Shanghai and Shenzhen A-shares increased rapidly from 2.36 billion yuan in 2011 to 1.65 trillion yuan in 2018. In 2019, despite policy regulations and a decline in the yields of WMPs, investments into WMPs still reached 1.45 trillion yuan. Purchasing WMPs is a method of capital allocation for listed companies, whereby the company's cash resources are used to invest in financial products rather than traditional production and operation activities. Studies shows that large-scale participation of listed companies in financial management activities will crowd out the industrial investment of enterprises (Hu et al., 2019;Zheng et al., 2020), and have a negative impact on the fundamentals and sustainable growth of enterprises. In addition, there is also the risk of capital loss hidden behind the appearance of a large number of listed companies purchasing WMPs. The introduction of the 'New Regulations on Asset Management' and 'New Regulations on Wealth Management' in 2018 1 have restricted the ability of bank WMPs, which occupies an absolute proportion, to be a source of 'stable profits without loss'. Once the entrusted funds used by the listed company to purchase WMPs have a substantial default, the company's capital chain will be strained, which will lead to adverse consequences, such as operational difficulties for the company. Therefore, regulating the purchase of WMPs by listed companies has received extensive attention from the government and academia.
The existing literature has recorded two motivations for companies to purchase WMPs: liquidity management and managerial agency (Hao, 2020;Zheng et al., 2020); these two motivations are closely related to the external financing environment and internal governance environment of the company. At the same time, maintaining cash dividend smoothing for listed companies have been shown to have a positive impact on the company's external financing environment and internal governance (Brav et al., 2005;John et al., 2015). Cash dividends have historically been regarded as the communication link between the company and external investors, and is a widely recognised corporate governance tool, which has an important impact on the company's capital allocation behaviour (Y. Chen et al., 2015;Liu et al., 2015;Wang et al., 2014). Dividend smoothing 2 is an important feature of listed companies' dividend policy, whereby the company pays earnings to shareholders in the form of stable cash dividends to ensure that the annual cash dividend fluctuation is within a certain small range relative to the surplus fluctuation. Implementing a stable dividend policy can send a positive signal to the market about the company's fundamentals and help improve the company's equity financing environment (Brav et al., 2005); this can aid in alleviating the financing constraints faced by the company. In addition, dividend smoothing removes excess cash resources from the managers' discretion and forces managers to accept external capital market scrutiny , while attracting institutional investors with special preferences for dividend smoothing to participate in corporate governance (M. Chen et al., 2017;Larkin et al., 2017), and thus, reduce the problem of corporate inefficiency caused by managerial agency conflicts. Therefore, the improvement of the financing environment and internal governance of companies through cash dividend smoothing is expected to have a significant impact on the WMPs purchased by companies.
A continuous and stable cash dividend policy helps investors to form stable expectations, which is of great significance for cultivating investors' long-term value investment 1 In April 2018, the People's Bank of China, the China Banking and Insurance Regulatory Commission, the China Securities Regulatory Commission, and the State Administration of Foreign Exchange jointly issued the 'Guidance on Regulating the Asset Management Business of Financial Institutions' (the new regulation on asset management), which clearly proposes to 'break the rigid payment' system. Article 2 stipulates that the asset management business should be the off-balance sheet business of financial institutions, and that financial institutions shall not undertake to protect capital and income when carrying out asset management business. In case of difficulties in payment, financial institutions shall not advance funds in any form to make payments. In September 2018, the China Banking Regulatory Commission issued the 'Supervision and Administration of Commercial Bank Wealth Management Business' (the new wealth management regulations), which again emphasised that banks should not advertise or promise guaranteed returns when selling wealth management products. Thus, the introduction of the 'new regulations on asset management' and 'new regulations on wealth management' have caused bank wealth management to not be an absolutely safe investment in the future. , which further required listed companies to clarify the cash dividend policy in their articles of association, improve the cash dividend system, and maintain the consistency, reasonableness and stability of cash dividends. Under the background that Chinese regulators pay more and more attention to the stability of the cash dividend policy of listed companies, the cash dividends smoothing of listed companies has been significantly enhanced (Li et al., 2014). This paper examines how the improvement of cash dividends smoothing under the guidance of the cash dividend regulatory policy affects the behaviour of companies to purchase WMPs, and explains the mechanism behind this effect. Additionally, we aim to demonstrate how dividend smoothing moderates the relationship between corporate purchases of WMPs and industrial investment. Using data on WMP purchases by Chinese A-share non-financial listed companies from 2011-2019, dividend smoothing was found to significantly reduce the WMPs purchased by companies. Channel tests demonstrated that dividend smoothing reduces firms' WMPs by increasing the financial leverage and attracting independent institutional investors to play a governance effect. Additional analysis showed that the effect of dividend smoothing on WMPs was more pronounced in firms with serious managerial agency problems, private firms, and in firms located in regions with a high density of financial institution branches. Finally, dividend smoothing was discovered to alleviate the crowding-out effect of corporate purchases of WMPs on industrial investment and that it facilitates the return of corporate financial funds to industrial operations.
The study mainly makes the following three contributions: first, in recent years, the large number of WMPs purchased by listed companies in China has caused many concerns. The literature has discussed the series of negative impacts of listed firms' purchases of WMPs on industrial investment, such as crowding out productive investment (Hu et al., 2019;Zheng et al., 2020) and reducing innovation quality (Hao, 2020); however, coping strategies to mitigate these negative impacts had not been proposed previously. This paper demonstrated that dividend smoothing can reduce companies purchases of WMPs and further alleviate the crowding-out effect of WMPs on industrial investment, which is an important reference value for guiding the return of listed firms' financial capital to industrial operations.
Second, the current research on the economic consequences of cash dividend of listed companies focuses on the static characteristics of cash dividend (e.g. dividend propensity, dividend level); however, the dynamic characteristics of cross-period cash dividend policies (e.g. dividend smoothness) have not received sufficient attention. Only a few studies have examined the effects of dividend smoothing on investor response (M. Chen et al., 2017), capital structure , and agency efficiency (Han & Wu, 2020). This paper extends the knowledge on the economic consequences of dividend smoothing in increasing the efficiency of corporate capital allocation by exploring the impact of dividend smoothing on corporate purchases of WMPs.
Third, the issue of cash dividend smoothing of listed companies in China has received continuous attention from regulatory authorities, and the 'Decision' implemented in 2013 has had a positive impact on the cash dividend smoothing of listed companies (Xie, 2019). This study confirms the effectiveness of the regulatory policy on cash dividends implemented in 2013 and affirms the importance of dividend smoothing in guiding financial capital back to the industrial economy by providing positive feedback for the relevant regulations in the 'Decision' and a realistic reference basis for regulators to further guide listed companies to pay dividends smoothly.

Liquidity management motivation for companies to purchase WMPs
Due to the imperfections of the financial markets, companies facing financing constraints have a strong incentive to set aside certain liquid assets on account of precautionary motives to cope with the uncertainties of future developments. Purchasing WMPs is one of the ways for companies to manage their liquidity. Initially, listed companies in China mainly reserve cash assets in the form of bank deposits; however, managing liquidity funds in the form of bank deposits have become less efficient in recent years as bank deposit interest rates have dropped significantly as part of the expansionary monetary policy environment. WMPs are a new capital management tool arising from the continuous innovation of financial markets and financial products, which have a flexible investment period and better wealth management yields than bank deposits. Therefore, companies are increasingly willing to manage their liquidity by purchasing WMPs (Hao, 2020). Additionally, different from other financial assets, company's purchase of bank WMPs also helps to strengthen their relationship with banks (Chu & Hu, 2020) and facilitates the company's access to bank loans in the future, such as relationship loans or lower cost loans (Friedmann et al., 2018), and thus, indirectly reduces the company's liquidity risk. Therefore, companies, especially those facing greater financing constraints, are motivated to purchase WMPs for liquidity management purposes.

Managers' self-interest motivation for companies to purchase WMPs
In a modern corporate system where ownership and operation are separated, contractual incompleteness, information asymmetry, and incentive incompatibility between owners and managers breed managerial agency problems (Jensen & Meckling, 1976). Managers, who hold the actual management power, have greater discretion over corporate financial asset allocation and have an incentive to obtain private gains through financial asset allocation (Du et al., 2017). First, the effort and risk taken by managers when investing in high-risk projects cannot be compensated by the benefits gained, therefore, managers are incentivised to adopt lazy or risk-averse investment behaviours to increase leisure while avoiding the loss of personal benefits from project failure (John et al., 2008). WMPs are easy to invest in and consumes less time and energy, therefore, managers who lack supervision tend to abandon difficult and uncertain industrial investment projects to instead invest in the purchase of WMPs (Zheng et al., 2020). Second, self-interested managers tend to want to hold more liquid assets (Vo, 2018) as liquid assets are more flexible when managers pursue their private goals and can reduce the market oversight brought by external financing and ensure the safety of managers' positions (Bates et al., 2009;Jensen & Meckling, 1976). Thus, self-interested managers are motivated to purchase WMPs to expand their control over flexible capital and secure their positions as WMPs have good liquidity and high security (Hao, 2020;Zheng et al., 2020).

Dividend smoothing affects the WMPs purchased by companies: financing effect
According to the dividend signalling theory, there are dissipation costs in signalling, and only high-quality firms can afford to use dividends to mitigate the information asymmetry between the firm and the external market (John & Williams, 1985), and maintain a smooth dividend policy over time, while low-quality firms are unable to imitate this behaviour. Accordingly, investors judge companies based on the smoothness of their cash dividends, and consider companies with smooth dividends as highquality companies and pay a higher premium for them (Brav et al., 2005;Brockman et al., 2022;Guttman et al., 2010). Dividend smoothing sends financial signals to external investors about the company's solid performance, earnings sustainability, and good cash flow (Yang, 2019), thereby enhancing investor confidence and eliciting a positive response in the company's stock price (Brav et al., 2005). A higher stock price will consequently reduce the company's refinancing costs and ease the financing constraints faced by the company.
According to the liquidity management motivation for purchasing WMPs, companies facing financing constraints purchase WMPs mainly due to a precautionary incentive. When a firm faces smaller financing constraints, it can access funds more quickly and cheaply from the outside, which would weaken its precautionary incentive to reserve liquidity assets (Almeida et al., 2004;Lian et al., 2010), and accordingly, the need to purchase WMPs to manage liquidity assets will also reduce accordingly.

Dividend smoothing affects the WMPs purchased by companies: governance effect
Dividend agency theory states that dividend smoothing is an alternative to traditional corporate governance mechanisms that can effectively alleviate managerial agency problems (John et al., 2015;Yang, 2019), and then reduce the self-interest motivation of managers to purchase WMPs. First, dividend smoothing increases the firm's debt financing needs, and thus, significantly increase its leverage . The increase in debt plays a contingent governance role and constrains managers' self-interested behaviour (He & Zhang, 2015). Specifically, debt puts higher demands on the future cash flow and earnings of the company, and the pressure of debt service will motivate managers to focus more on the long-term value of the company and invest actively (Dou & Liu, 2011). This is expected to reduce lazy and conservative investments, and thus, discourage managers from making short-term financial investments based on their self-interest, especially through allocation of cash to financial assets, such as WMPs (Shen & An, 2021). Secondly, dividend smoothing will attract strong governance shareholders. Studies have shown that institutional investors, who mainly pursue safety and stable cash flow, prefer stocks with smooth dividends compared to individual investors, and thus, companies with dividend smoothing can attract and retain institutional investors (M. Chen et al., 2017;Larkin et al., 2017). Institutional investors are investors with information advantages and expertise in the market, and their participation in corporate governance can effectively restrain the self-interested behaviour of managers (Allen et al., 2000;Yin et al., 2018), including inhibiting managers from purchasing WMPs due to laziness, risk aversion, or motivation to maintain position security.
This study proposes the following core hypothesis: H1: Dividend smoothing reduces the WMPs purchased by companies The logical framework of the theoretical analysis is shown in Figure 1.

Data source and sample selection
This (2) companies in the financial sector; (3) No dividend for three consecutive years, and the continuous dividend data and financial data are less than three years; (4) operating profit was negative; and (5) the main variable was missing. Consequently, 18,554 firm-year observations were obtained. The detailed data of the WMPs purchased by company was drawn from the cross-check of the CSMAR database and the WIND database, and the company's financial data were all from the CSMAR database. All continuous variables were winsorised at the 1st and 99th to alleviate the possible effects of the extreme values.

Model design and variable definition
In November 2013, the SFC emphasised in 'Decision' that 'listed companies should firmly establish the awareness of rewarding shareholders, improve the cash dividend system in strict accordance with the provisions of the Company Law, the Securities Law, and the Articles of Association, and maintain the consistency, reasonableness, and stability of the cash dividend policy'. Studies have shown that the cash dividend regulatory policy implemented in 2013 is an exogenous catered object of the dividend smoothing policy of listed companies, and that it has a positive impact on the company's dividend smoothing (Xie, 2019). Therefore, this paper uses the cash dividend regulatory policy in 2013 as an exogenous shock event affecting the dividend smoothing of listed companies and designs a difference in difference (DID) model to test the main hypothesis. The model is as follows: In the model, the dependent variable is the size of the WMPs purchased by company, which includes two indicators: the total amount of funds occupied by WMPs purchased by company (Finance) and the return from the WMPs purchased by company (Return). Finance is the sum of the amount of each WMP purchased by company i in year t multiplied by the entrusted period, and divided by total assets. Return is the actual total returns received from the WMPs purchased by company i in year t divided by operating profit. It should be noted that the existing literature uses annual cumulative amounts to measure the WMPs purchased by companies; however, due to the large differences in the periodicity of WMPs among companies, 4 the cumulative amounts may not adequately reflect the appropriation of funds by companies' purchases of WMPs. Therefore, we weighted the amount of each WMPs by entrusted period and then summed to improve the comparability of WMPs among companies. Specifically, the amount of WMPs for company i in year t = P n j¼1 ((amount of WMPs amount) itj ×(days of WMPs) itj /365). Treat is the treatment dummy variable. Following B. Liu and Jiang (2021), we divided the treatment group and the control group according to the intensity of the policy (the difference between the mean value of the variable directly affected by the policy in the treatment window minus its mean value in the control window). The dividend smoothing of pre-policy companies were used as a benchmark by assigning a value of 1 to companies whose dividend smoothing improves above the sample median after the policy, and 0 otherwise.
The existing literature documents two methods of cash dividend smoothing. One is the partial adjustment speed calculation (PASC) method, which is modified by Leary and Michaely (2011) based on the partial adjustment model of Lintner (1956). The firm's target dividend level was first estimated, and then the adjustment speed of the actual dividend level approaching the target level was calculated. The other is the free model nonparametric calculation (FMNC) method. This method involves measuring the dividend smoothing based on the volatility of dividends relative to earnings (Leary & Michaely, 2011), which is equal to the standard deviation of dividends per share over the past three years divided by the standard deviation of earnings per share over the past three years. The indicators obtained by the two methods are negative indicators of dividend smoothing, where a larger value indicates a less smooth dividend. The dividend smoothing obtained by the PASC method is more suitable for examining different characteristics of cross-section individuals (Fliers, 2019; Leary & Michaely, 2011), while the dependent variable in this study, the WMPs purchased by companies, has a large variation in the firm's longitudinal time dimension. Dividend smoothing measured by the PASC method may not adequately represent the effect of changes in dividend smoothing on firm purchases of WMPs in the firm's time dimension. Therefore, this study adopts the dividend smoothing measured by the FMNC method, which has been adopted by several scholars including L. Chen et al. (2012), Rangvid et al. (2014), X. Liu and Chen (2016), and Xie (2019) was chosen. Since dividend smoothing is a negative indicator, we assign a negative value to it during calculation for convenience of understanding.
Post is a dummy variable for the cash dividend regulation policy, which takes the value of 1 in 2013 and later, and 0 otherwise. The interaction term Treat×Post reflects the change in the WMPs purchased by companies whose dividend smoothing have increased significantly after the dividend policy.
The definitions for the variable and the specific calculation are shown in Table 1. Table 2 shows the statistics of the characteristics of the WMPs purchased by listed companies in China from 2011-2019. Panel A demonstrates that 5,784 of the 18,554 firmyear observations included in this paper were involved in purchasing WMPs, reflecting the high proportion of non-financial listed companies in China involved in purchasing WMPs. The distribution of the characteristics of the sample of purchased WMPs shows that the average entrustment period of WMPs was concentrated in the 1-6 months range, reflecting the liquidity advantages of WMPs. The average return of WMPs was concentrated in the range of greater than 5%, while the rate of return of the company's total assets was 5.1%. Therefore, compared with industrial investment, companies have the motivation to purchase WMPs to obtain financial returns. Panel B shows that, the total amount of WMPs purchased by listed companies in China from 2011-2019 was 2,660 billion yuan, and the return obtained from the WMPs purchased by companies was 86 billion yuan, reflecting the large amounts of transactions involved in the WMPs purchased by companies. Dividing the types of WMP trustees, the highest percentage of bank WMPs were purchased by listed companies, accounting for about 77.82% (72.21%) of the total amount (return) of WMPs. Dividing the types of WMP returns, listed companies purchased the highest percentage of capital-guaranteed or return-guaranteed WMPs, which  Taking 1 for those firms whose dividend smoothing improves above the sample median after the policy, and 0 otherwise Regap WMPs return rate -industrial investment return rate. Where the rate of return from the WMPs purchased by company is equal to the return of WMPs purchased by company divided by the amount of WMPs in the year. The rate of return on industrial investment is based on Zhang and Zhang (2016), which is equal to the ratio of operating income to operating assets, i.e. (operating income-operating costsoperating taxes and surcharges -period expenses -asset impairment losses)/ (operating capital + fixed assets + net intangible assets) Oprisk

Statistics on the characteristics of WMPs purchased by listed companies
Standard deviation of industrial investment return for the last three years, industrial investment return is calculated as above Fixed Fixed assets divided by total assets Cfo Net cash flow from operating activities divided by total assets Size The natural logarithm of total assets Roa Total profit divided by total assets Grow Operating income growth rate Boardsize The total number of board members Top1_5 Total shareholding of the second to fifth largest shareholder divided by the shareholding of the first largest shareholder Mshare Total shareholding ratio of managers Msalary Natural logarithm of the average compensation of the top three executives Indep The percentage of independent directors on a board Dual Taking 1  accounted for 48.87% (49.42%) of the total amount (return) of WMPs, while the second highest percentage of non-capital-guaranteed or non-return-guaranteed WMPs, accounted for 26.80% (24.65%) of the total amount (return) of WMPs. The above statistical results as a whole indicate that the WMPs purchased by listed companies in China have bank-led and low-risk characteristics. Simultaneously, there is a certain proportion of non-capitalguaranteed or non-return-guaranteed WMP purchased by listed companies, which reflects that there may be certain risks hidden in the WMP purchased by listed companies in China. Table 3 presents the variable 'descriptive statistics'. The mean value of Finance was 0.033 and the maximum value was 0.545, which indicated that the amount of WMPs purchased by listed companies is about 3.3% of the total assets on average, and can be up to 54.5% of the total assets. The mean value of Return is 0.027 and the maximum value is 0.700, indicating that the return from the WMPs purchased by companies was about 2.7% of operating profit, and can be up to 61.6% of operating profit. The above results indicate that some listed companies purchase WMPs with great strength and that there may be a tendency for WMPs to crowd out investment in their main business. The statistics of other control variables are basically consistent with previous studies.  (1) is negative and significant at the 1% level, indicating that companies with enhanced dividend smoothing reduce the funds occupied by purchasing WMPs after the implementation of the dividend policy. The coefficient of Treat×Post in column (2) is negative and significant at the 1% level, indicating that the return from the WMPs purchased by companies as a percentage of operating profit decreases for the companies with enhanced dividend smoothing after the implementation of dividend policy. In addition, the coefficient of Treat×Post remained  (3) and (4). The above results validate the core hypothesis that dividend smoothing reduces the WMPs purchased by companies.

Parallel trend test
Since the key premise of applying the DID model is that the treatment and control group samples should satisfy the parallel trend assumption, we performed a parallel trend test. We refer to the method of Wu et al. (2018) and included the interaction term of the year dummy variable with Treat in the regression. For example, the variable Year2011 is 1 when 2011, 0 otherwise, and so on for the rest of the year. To avoid the trap of setting dummy variables, we excluded the dummy variable for 2019. The coefficient of the interaction term was expected to be insignificant prior to the implementation of the policy, i.e. the dependent variables in the treatment and control groups were not expected to differ significantly between the pre-policy years. Table 5 reports the regression results of the parallel trend test. Consistent with our expectation, the coefficients of Treat×Year2011 and Treat×Year2012 were insignificant prior to the implementation of the cash dividend policy in 2013 (2011-2012); this indicates that the DID model in this paper satisfies the parallel trend hypothesis. In addition, after the implementation of the policy (2013-2018), the coefficients of Treat×Year2013 were insignificant, the coefficient of Treat×Year2014 in column (1) was significantly negative, and the coefficients of Treat×Year2015, Treat×Year2016, Treat×Year2017, and Treat×Year2018 in column (1) and column (2) were significantly negative, indicating that dividend smoothing leads to a decrease in WMPs purchased by companies after the policy implementation. The insignificance coefficient of Treat×Year2013 may be due to the delay in the policy effect of the 'Decision' in November 2013, resulting in the insufficient release of the policy's effect on the dividend smoothing of listed companies in that year.

PSM+DID test
The PSM+DID test was used to verify the basic test to exclude possible differences in other characteristics between the treatment and control groups. First, all control variables in the baseline model were used as matching variables and a 1:1 matching was performed using the nearest neighbour with put-back matching method. Second, the core hypothesis was tested again using the matched samples. The  Table 6 Panel A shows that before matching, some variables were significantly different between the treatment and control groups. However, after matching, there was no significant difference between the treatment and control groups for all variables except for Msalary, indicating that the matching variables pass the balance test. The results of the matched regressions are presented in Table 6 Panel B. The coefficients of Treat×Post are all significantly negative, which is consistent with the results of the basic test. This result suggests that, after excluding differences in other characteristics of the treatment and control groups, cash dividend smoothing still significantly reduces the WMPs purchased by companies.

Alternative measurements of the main variables
(1) Since the independent variable (dividend smoothing) is an indicator with a three-year period, the mean value of the WMPs amount (WMPs return) for the last three years was similarly selected as the dependent variable.
(2) Since current assets or quick assets may also be a factor affecting the amount of WMPs, the amount of WMPs was standardised by current assets and quick assets of the current and previous periods, respectively. (3) Treat was redefined according to the dividend smoothing of companies before the policy implementation, and take the value of 1 for companies with lower degree of dividend smoothing before the policy (such firms are expected to be more affected by the policy), otherwise 0. The results of the above tests were all consistent with the basic results. 5

Alternative models and samples
(1) Model replacement: Since the dependent variable (WMPs) has a truncated statistical distribution characteristic, the Tobit regression model was replaced for regression again.

Channel analyses
The previous basic test suggests that dividend smoothing reduces the WMPs purchased by companies. We mention two channels in the hypothesis section, the financing effect and the governance effect, each of the two channels were tested to determine the specific channel through which dividend smoothing reduces WMPs.

Dividend smoothing reduces WMPs: financing effect
First, according to the financing hypothesis, dividend smoothing sends a favourable signal to the market, reduces the cost of equity financing, alleviates the firm's financing constraints, and thereby, weakens the firm's incentive to purchase WMPs for liquidity management. We construct the mediation test model set (2) to examine this effect.
In the model, Ecost is the cost of equity financing, which was measured as the inverse of the firm's annual price-to-earnings ratio similar to the study by Lin et al. (2015). Other control variables and model settings remain consistent with the model (1). The regression results are presented in Table 7. Since the basic test has shown that dividend smoothing reduces WMPs purchased by companies, this result is not repeated here. The coefficient of Treat×Post in column (1) was negative but insignificant, indicating that the effect of dividend smoothing in reducing the cost of equity financing is insignificant. This result is not consistent with expectations and fails the test of the mediating effect. Further, the coefficients of Ecost in columns (2) and (3) are both significantly negative, which is contrary to expectations, indicating that companies with lower equity financing costs are more likely to purchase WMPs. The result suggests that firms are not motivated by liquidity management to purchase WMPs. The findings were consistent with Hao (2020), who observed that the WMPs purchased by companies in China is more a reflection of the management's principal-agent problem rather than the firm's need for financial liquidity management.

Dividend smoothing reduces WMPs: governance effects
According to the governance hypothesis, dividend smoothing will force firms to raise funds from the outside to increase financial leverage, and attract and retain institutional investors, thereby playing the role of debt governance and institutional investor governance, and ultimately weakening managers' self-interest incentives to purchase WMPs. Therefore, the mediation test model set (3) was constructed to examine debt governance and institutional investor governance separately.
In the model, Lev is financial leverage, which is equal to total liabilities divided by total assets. Inst is the percentage of institutional investors' shareholding. Other control variables and model settings remained consistent with model (1). Due to the heterogeneity within institutional investors, different types of institutional investors differ in their effectiveness in improving corporate governance. Studies have shown that independent institutional investors are more motivated to participate in corporate governance and are more likely to significantly improve corporate governance (Li & Yan, 2017). Conversely, non-independent institutional investors are expected to be more submissive to management decisions in order to maintain business relationships with the companies, and thus, fails to play a good monitoring and disciplinary role (Aggarwal et al., 2011). Therefore, we subdivide institutional investors (Inst) into two categories: non-independent institutional investors (Depinst) and independent institutional investors (Indepinst). Specifically, drawing on the division method of M. Chen et al. (2017), securities investment funds and qualified foreign investors (QFIIs) were defined as independent institutional investors and other types of institutional investors as non-independent institutional investors.
The regression results for debt governance are presented in Table 8 Panel A. In column (1), the coefficient of Treat×Post is significantly positive, which indicates that dividend smoothing significantly increases the company's financial leverage. In columns (2) and (3), the coefficients of Lev are significantly negative, indicating that debt constraint reduces the firms' WMPs, and the coefficient values of Treat×Post decrease to different degrees compared to Table 4. The results pass the mediating effect test. The above results suggest that dividend smoothing reduces firms' WMPs by playing the role of debt governance. This finding is consistent with the study of Shen and An (2021), whose study showed that debt constraints can significantly inhibit the size of cash-based financial assets held by listed firms.
The regression results for institutional investor governance are presented in Table 8 Panel B. The coefficient of Treat×Post in column (1) is significantly positive, indicating that dividend smoothing significantly attracts institutional investors. After distinguishing between independent and non-independent institutional investors, the coefficient of Treat×Post in column (2) was observed to be negative and insignificant, while the coefficient of Treat×Post in column (3) was significantly positive. This result indicated that dividend smoothing mainly attracts independent institutional investors, while having no significant effect on non-independent institutional investors. Consequently, the mediating role of independent institutional investors were further tested. The results show that the coefficient of Indepinst in column (4) is significantly negative, indicating that independent institutional investors significantly reduce the firms' WMPs amount, and the results pass the mediation effect test. The coefficient of Indepinst in column (5) is insignificant and does not pass the test of the mediating effect. The results indicate that dividend smoothing will attract independent institutional investors with an oversight governance role to exert their governance effect and thus, reduce the WMPs purchased by companies.

Managerial agency problems, dividend smoothing and WMPs purchased by companies
Previous tests showed that the WMPs purchased by listed companies in China reflects managerial agency problems rather than the need of corporate liquidity management. Dividend smoothing suppressed the self-interest motivation of managers to purchase WMPs by playing the role of debt governance and investor governance. However, managerial agency problems vary within different firms, which may lead to differences in the effect that dividend smoothing has on reducing firms' WMPs. In firms with weak managerial agency problems, the firm's capital allocation decisions are inherently more efficient, and thus, there is less room for dividend smoothness to be effective. However, in firms with serious managerial agency problems, self-interested managers have strong incentives to manipulate allocation resources for personal gain (Cornett et al., 2009), and are more opportunistic in purchasing WMP. Thus, the effect of dividend smoothing in reducing firms' WMPs can be more effective in these circumstances. Consequently, the effect of dividend smoothing in reducing firms' WMPs is expected to be more effective in firms with serious managerial agency problems. The managerial agency costs were measured by the ratio of administrative expenses to operating income and dividing the sample into two groups based on the annual industry median of this indicator. The results in Table 9 show that the coefficients of Treat×Post were significantly negative in the group with weak managerial agency problems (columns 1 and 2), and that the coefficients of Treat×Post were also significantly negative in the group with serious managerial agency problems (columns 3 and 4). Comparing the coefficients of Treat×Post of the two groups reveal that there was a significant difference between the coefficients of the groups in columns (1) and (3) (column 3 > column 1), while there was no significant difference between the coefficients of the groups in columns (2) and (4). The results show that the effect of dividend smoothing reducing the WMPs purchased by companies is more pronounced in firms with serious management agency problems than in firms with weak management agency problems.

Nature of ownership, dividend smoothing, and WMPs purchased by companies
The nature of property rights, i.e. whether state-owned or private, is an important characteristic of companies in China, and the different nature of property rights has led to significant differences between state-owned and private enterprises in many aspects, such as resource allocation and corporate governance. State-owned enterprises have been subject to more government intervention, leaving less room for external governance mechanisms, such as capital markets to exert their influence (Bo & Wu, 2009), while private firms are subject to less administrative intervention, and thus, external governance mechanisms are able to more effectively exert their influence. Therefore, we expect the effect of dividend smoothing in reducing firms' WMPs to be better in private firms. We divide the sample into two categories, state-owned and private, for testing based on the nature of the actual controller. The data on the nature of the actual controllers are obtained from the CCER database. The regression results in Table 10 show that the coefficients of Treat×Post were significantly negative in private firms (columns 1 and 2), and that the coefficients of Treat×Post were also significantly negative in state-owned firms (columns 3 and 4). The test for the difference in coefficients between groups found significant difference between the coefficients of Treat×Post columns (1) and (3) (column 1 > column 3), while there was no significant difference between the coefficients of columns (2) and (4). The above results indicate that dividend smoothing is more effective in reducing the WMPs purchased by companies in private firms than in state-owned firms.

Financial institution branch density, dividend smoothing, and WMPs purchased by companies
China has a vast territory, and there are large regional differences in financial development. Regions with a higher level of financial development have a higher number of spatial clusters of banking and financial institutions. Under the constraint of limited market size, bank financial institutions are prone to attract investors through price competition and lower investment threshold, which enhances the enthusiasm of companies for purchasing WMPs (Xu & Zeng, 2015). Therefore, the effect of dividend smoothing reducing WMPs of companies is expected to be more effective when the companies are located in regions with a high density of financial institution branches.
We measure the regional financial institution branch density by the number of financial institution branches per million population in the province where the company is located, and divide the sample into two groups for discussion based on the annual median of this indicator. Data on the banking financial institutions branches in each region were obtained from the Wind database, and data on the population in each region were obtained from the China Statistical Yearbook. The results in Table 11 show that the coefficient of Treat×Post was significantly negative in column (1) and insignificant in  (2) in regions with a lower density of financial institution branches (columns 1 and 2). In regions with a higher branch density of financial institution branches (columns 3 and 4), the coefficients of Treat×Post were significantly negative. In addition, there was a significant difference between the coefficients of Treat×Post of the groups in columns (1) and (3) (column 3 > column 1), while there was no significant difference between the coefficients of the groups in columns (2) and (4). The above results overall suggest that dividend smoothing reduces the WMPs purchased by companies better in areas with a higher density of financial institution branches than in areas with a lower density of financial institution branches.

Dividend smoothing, WMPs purchased by companies, and corporate industrial investments
Some studies have pointed out that the WMPs purchased by companies with managerial agency problems will squeeze out the firm's productive investments (Hu et al., 2019;Zheng et al., 2020). Accordingly, model (4) was constructed to further examine how the effect of 'dividend smoothing reduces the WMPs purchased by companies' ultimately affects firms' industrial investment.

(4)
Invest in the model is the level of industrial investment, which is equal to the cash paid by the company for the acquisition of fixed assets, intangible assets, and other long-term assets divided by total assets. We also controlled for the level of investment in the previous period in the model, and all other variables in the model were consistent with the previous section. The regression results in Table 12 show that the coefficient of Treat×Post×Finance in column (1) is significantly positive, and that the coefficient of Treat×Post×Return in column (2) is also significantly positive. This indicated that dividend smoothing weakened the crowding out effect of WMPs purchased by companies on the level of industrial investment. Therefore, enhancing the firm's dividend smoothing is expected to facilitate the return of the firm's financial capital to industrial operations.

Conclusion
This study examined whether cash dividend smoothing has a positive effect on constraining firms' purchases of WMPs, and the results confirmed this hypothesis. Channel tests demonstrated that dividend smoothing mainly reduces firms' WMPs by increasing firms' financial leverage and attracting independent institutional investors to play debt governance and institutional investor governance role. Additional analysis demonstrated that the effect of dividend smoothing on WMPs was more significant when there was a serious management agency problem within the firm, when the nature of ownership was private, and when the firm was located in regions with a higher density of financial institutions branches. Finally, the study showed that dividend smoothing mitigates the crowding-out effect of WMPs on industrial real investment, and that it facilitates the return of firms' financial management funds to industrial operations. This paper has the following policy implications: (1) There are two perspectives on the motivation of listed companies to participate in financial investment activities: the 'reservoir' view and the 'agency' view. The results of this study show that the WMPs purchased by listed companies in China reflects the managerial agency problem, which further supports the 'agency' view of purchasing WMPs by companies. Therefore, under the modern enterprise management system of separation of ownership and control, company owners should pay more attention to managers' opportunistic behaviour of purchasing WMPs based on self-interest-based motives.
(2) This study found that the 2013 cash dividend policy, which required companies to increase the smoothness of cash dividend, discouraged companies from purchasing WMPs and mitigated the crowding-out effect of WMP purchases on the firms' industrial investments. Accordingly, regulators should continue to pay attention to the stability of the cash dividend policies of listed companies and further guide listed companies, especially those with serious management agency problems, private companies, and companies in regions with a higher availability of WMPs, to pay smooth dividends in order to regulate managers' decisions to purchase WMPs, and thus, guide irrational financial investment funds back to the industrial economy.